From: Nicolai Haehnle Date: Tue, 28 Sep 2004 10:59:23 +0000 (+0000) Subject: Initial revision X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=f30d53e0498a558100a1b9d9651ff375a7b3c7b4;p=mesa.git Initial revision --- diff --git a/src/mesa/drivers/dri/r300/Makefile b/src/mesa/drivers/dri/r300/Makefile new file mode 100644 index 00000000000..3cad8c080fe --- /dev/null +++ b/src/mesa/drivers/dri/r300/Makefile @@ -0,0 +1,72 @@ +# src/mesa/drivers/dri/r300/Makefile + +TOP = ../../../../.. +include $(TOP)/configs/current + +LIBNAME = r300_dri.so +DEFINES += -DCOMPILE_R300 + +MINIGLX_SOURCES = server/radeon_dri.c + +COMMON_SOURCES = \ + ../../common/driverfuncs.c \ + ../common/mm.c \ + ../common/utils.c \ + ../common/texmem.c \ + ../common/vblank.c \ + ../common/xmlconfig.c \ + ../common/dri_util.c \ + ../common/glcontextmodes.c + +DRIVER_SOURCES = \ + radeon_screen.c \ + radeon_context.c \ + radeon_ioctl.c \ + radeon_lock.c \ + radeon_span.c \ + \ + r300_context.c \ + r300_ioctl.c \ + r300_cmdbuf.c \ + r300_state.c \ + \ + r200_context.c \ + r200_ioctl.c \ + r200_state.c \ + r200_state_init.c \ + r200_cmdbuf.c \ + r200_pixel.c \ + r200_tex.c \ + r200_texmem.c \ + r200_texstate.c \ + r200_tcl.c \ + r200_swtcl.c \ + r200_maos.c \ + r200_sanity.c \ + r200_vtxfmt.c \ + r200_vtxfmt_c.c \ + r200_vtxfmt_sse.c \ + r200_vtxfmt_x86.c + +C_SOURCES = $(COMMON_SOURCES) $(DRIVER_SOURCES) + +X86_SOURCES = r200_vtxtmp_x86.S + +SYMLINKS = \ + server/radeon_dri.c \ + server/radeon_dri.h \ + server/radeon.h \ + server/radeon_macros.h \ + server/radeon_reg.h + +##### TARGETS ##### + + +include ../Makefile.template + +$(SYMLINKS): + mkdir -p server + for i in $(SYMLINKS) ; do rm -f $$i && test -f ../radeon/$$i && ln -s ../../radeon/$$i $$i ; done + +symlinks: $(SYMLINKS) + diff --git a/src/mesa/drivers/dri/r300/r200_context.h b/src/mesa/drivers/dri/r300/r200_context.h new file mode 100644 index 00000000000..b5ba1f9258e --- /dev/null +++ b/src/mesa/drivers/dri/r300/r200_context.h @@ -0,0 +1,840 @@ +/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_context.h,v 1.2 2002/12/16 16:18:54 dawes Exp $ */ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Keith Whitwell + */ + +#ifndef __R200_CONTEXT_H__ +#define __R200_CONTEXT_H__ + +#ifdef GLX_DIRECT_RENDERING + +#include "tnl/t_vertex.h" +#include "drm.h" +#include "radeon_drm.h" +#include "dri_util.h" +#include "texmem.h" + +#include "macros.h" +#include "mtypes.h" +#include "colormac.h" +#include "r200_reg.h" +#include "radeon_context.h" + +#define ENABLE_HW_3D_TEXTURE 1 /* XXX this is temporary! */ + +struct r200_context; +typedef struct r200_context r200ContextRec; +typedef struct r200_context *r200ContextPtr; + +#include "mm.h" + +/* The blit width for texture uploads + */ +#define BLIT_WIDTH_BYTES 1024 + +/* Use the templated vertex format: + */ +#define COLOR_IS_RGBA +#define TAG(x) r200##x +#include "tnl_dd/t_dd_vertex.h" +#undef TAG + +typedef void (*r200_tri_func) (r200ContextPtr, + r200Vertex *, r200Vertex *, r200Vertex *); + +typedef void (*r200_line_func) (r200ContextPtr, r200Vertex *, r200Vertex *); + +typedef void (*r200_point_func) (r200ContextPtr, r200Vertex *); + +struct r200_depthbuffer_state { + GLfloat scale; +}; + +struct r200_scissor_state { + drm_clip_rect_t rect; + GLboolean enabled; + + GLuint numClipRects; /* Cliprects active */ + GLuint numAllocedClipRects; /* Cliprects available */ + drm_clip_rect_t *pClipRects; +}; + +struct r200_stencilbuffer_state { + GLboolean hwBuffer; + GLuint clear; /* rb3d_stencilrefmask value */ +}; + +struct r200_stipple_state { + GLuint mask[32]; +}; + +#define TEX_0 0x1 +#define TEX_1 0x2 +#define TEX_2 0x4 +#define TEX_3 0x8 +#define TEX_4 0x10 +#define TEX_5 0x20 +#define TEX_ALL 0x3f + +typedef struct r200_tex_obj r200TexObj, *r200TexObjPtr; + +/* Texture object in locally shared texture space. + */ +struct r200_tex_obj { + driTextureObject base; + + GLuint bufAddr; /* Offset to start of locally + shared texture block */ + + GLuint dirty_state; /* Flags (1 per texunit) for + whether or not this texobj + has dirty hardware state + (pp_*) that needs to be + brought into the + texunit. */ + + drm_radeon_tex_image_t image[6][RADEON_MAX_TEXTURE_LEVELS]; + /* Six, for the cube faces */ + + GLuint pp_txfilter; /* hardware register values */ + GLuint pp_txformat; + GLuint pp_txformat_x; + GLuint pp_txoffset; /* Image location in texmem. + All cube faces follow. */ + GLuint pp_txsize; /* npot only */ + GLuint pp_txpitch; /* npot only */ + GLuint pp_border_color; + GLuint pp_cubic_faces; /* cube face 1,2,3,4 log2 sizes */ + + GLboolean border_fallback; +}; + +struct r200_texture_env_state { + r200TexObjPtr texobj; + GLenum format; + GLenum envMode; +}; + +#define R200_MAX_TEXTURE_UNITS 6 + +struct r200_texture_state { + struct r200_texture_env_state unit[R200_MAX_TEXTURE_UNITS]; +}; + +struct r200_state_atom { + struct r200_state_atom *next, *prev; + const char *name; /* for debug */ + int cmd_size; /* size in bytes */ + GLuint idx; + int *cmd; /* one or more cmd's */ + int *lastcmd; /* one or more cmd's */ + int *savedcmd; /* one or more cmd's */ + GLboolean dirty; + GLboolean(*check) (GLcontext *, int); /* is this state active? */ +}; + +/* Trying to keep these relatively short as the variables are becoming + * extravagently long. Drop the driver name prefix off the front of + * everything - I think we know which driver we're in by now, and keep the + * prefix to 3 letters unless absolutely impossible. + */ + +#define CTX_CMD_0 0 +#define CTX_PP_MISC 1 +#define CTX_PP_FOG_COLOR 2 +#define CTX_RE_SOLID_COLOR 3 +#define CTX_RB3D_BLENDCNTL 4 +#define CTX_RB3D_DEPTHOFFSET 5 +#define CTX_RB3D_DEPTHPITCH 6 +#define CTX_RB3D_ZSTENCILCNTL 7 +#define CTX_CMD_1 8 +#define CTX_PP_CNTL 9 +#define CTX_RB3D_CNTL 10 +#define CTX_RB3D_COLOROFFSET 11 +#define CTX_CMD_2 12 /* why */ +#define CTX_RB3D_COLORPITCH 13 /* why */ +#define CTX_STATE_SIZE_OLDDRM 14 +#define CTX_CMD_3 14 +#define CTX_RB3D_BLENDCOLOR 15 +#define CTX_RB3D_ABLENDCNTL 16 +#define CTX_RB3D_CBLENDCNTL 17 +#define CTX_STATE_SIZE_NEWDRM 18 + +#define SET_CMD_0 0 +#define SET_SE_CNTL 1 +#define SET_RE_CNTL 2 /* replace se_coord_fmt */ +#define SET_STATE_SIZE 3 + +#define VTE_CMD_0 0 +#define VTE_SE_VTE_CNTL 1 +#define VTE_STATE_SIZE 2 + +#define LIN_CMD_0 0 +#define LIN_RE_LINE_PATTERN 1 +#define LIN_RE_LINE_STATE 2 +#define LIN_CMD_1 3 +#define LIN_SE_LINE_WIDTH 4 +#define LIN_STATE_SIZE 5 + +#define MSK_CMD_0 0 +#define MSK_RB3D_STENCILREFMASK 1 +#define MSK_RB3D_ROPCNTL 2 +#define MSK_RB3D_PLANEMASK 3 +#define MSK_STATE_SIZE 4 + +#define VPT_CMD_0 0 +#define VPT_SE_VPORT_XSCALE 1 +#define VPT_SE_VPORT_XOFFSET 2 +#define VPT_SE_VPORT_YSCALE 3 +#define VPT_SE_VPORT_YOFFSET 4 +#define VPT_SE_VPORT_ZSCALE 5 +#define VPT_SE_VPORT_ZOFFSET 6 +#define VPT_STATE_SIZE 7 + +#define ZBS_CMD_0 0 +#define ZBS_SE_ZBIAS_FACTOR 1 +#define ZBS_SE_ZBIAS_CONSTANT 2 +#define ZBS_STATE_SIZE 3 + +#define MSC_CMD_0 0 +#define MSC_RE_MISC 1 +#define MSC_STATE_SIZE 2 + +#define TAM_CMD_0 0 +#define TAM_DEBUG3 1 +#define TAM_STATE_SIZE 2 + +#define TEX_CMD_0 0 +#define TEX_PP_TXFILTER 1 /*2c00 */ +#define TEX_PP_TXFORMAT 2 /*2c04 */ +#define TEX_PP_TXFORMAT_X 3 /*2c08 */ +#define TEX_PP_TXSIZE 4 /*2c0c */ +#define TEX_PP_TXPITCH 5 /*2c10 */ +#define TEX_PP_BORDER_COLOR 6 /*2c14 */ +#define TEX_CMD_1 7 +#define TEX_PP_TXOFFSET 8 /*2d00 */ +#define TEX_STATE_SIZE 9 + +#define CUBE_CMD_0 0 /* 1 register follows */ +#define CUBE_PP_CUBIC_FACES 1 /* 0x2c18 */ +#define CUBE_CMD_1 2 /* 5 registers follow */ +#define CUBE_PP_CUBIC_OFFSET_F1 3 /* 0x2d04 */ +#define CUBE_PP_CUBIC_OFFSET_F2 4 /* 0x2d08 */ +#define CUBE_PP_CUBIC_OFFSET_F3 5 /* 0x2d0c */ +#define CUBE_PP_CUBIC_OFFSET_F4 6 /* 0x2d10 */ +#define CUBE_PP_CUBIC_OFFSET_F5 7 /* 0x2d14 */ +#define CUBE_STATE_SIZE 8 + +#define PIX_CMD_0 0 +#define PIX_PP_TXCBLEND 1 +#define PIX_PP_TXCBLEND2 2 +#define PIX_PP_TXABLEND 3 +#define PIX_PP_TXABLEND2 4 +#define PIX_STATE_SIZE 5 + +#define TF_CMD_0 0 +#define TF_TFACTOR_0 1 +#define TF_TFACTOR_1 2 +#define TF_TFACTOR_2 3 +#define TF_TFACTOR_3 4 +#define TF_TFACTOR_4 5 +#define TF_TFACTOR_5 6 +#define TF_STATE_SIZE 7 + +#define TCL_CMD_0 0 +#define TCL_LIGHT_MODEL_CTL_0 1 +#define TCL_LIGHT_MODEL_CTL_1 2 +#define TCL_PER_LIGHT_CTL_0 3 +#define TCL_PER_LIGHT_CTL_1 4 +#define TCL_PER_LIGHT_CTL_2 5 +#define TCL_PER_LIGHT_CTL_3 6 +#define TCL_CMD_1 7 +#define TCL_UCP_VERT_BLEND_CTL 8 +#define TCL_STATE_SIZE 9 + +#define MSL_CMD_0 0 +#define MSL_MATRIX_SELECT_0 1 +#define MSL_MATRIX_SELECT_1 2 +#define MSL_MATRIX_SELECT_2 3 +#define MSL_MATRIX_SELECT_3 4 +#define MSL_MATRIX_SELECT_4 5 +#define MSL_STATE_SIZE 6 + +#define TCG_CMD_0 0 +#define TCG_TEX_PROC_CTL_2 1 +#define TCG_TEX_PROC_CTL_3 2 +#define TCG_TEX_PROC_CTL_0 3 +#define TCG_TEX_PROC_CTL_1 4 +#define TCG_TEX_CYL_WRAP_CTL 5 +#define TCG_STATE_SIZE 6 + +#define MTL_CMD_0 0 +#define MTL_EMMISSIVE_RED 1 +#define MTL_EMMISSIVE_GREEN 2 +#define MTL_EMMISSIVE_BLUE 3 +#define MTL_EMMISSIVE_ALPHA 4 +#define MTL_AMBIENT_RED 5 +#define MTL_AMBIENT_GREEN 6 +#define MTL_AMBIENT_BLUE 7 +#define MTL_AMBIENT_ALPHA 8 +#define MTL_DIFFUSE_RED 9 +#define MTL_DIFFUSE_GREEN 10 +#define MTL_DIFFUSE_BLUE 11 +#define MTL_DIFFUSE_ALPHA 12 +#define MTL_SPECULAR_RED 13 +#define MTL_SPECULAR_GREEN 14 +#define MTL_SPECULAR_BLUE 15 +#define MTL_SPECULAR_ALPHA 16 +#define MTL_CMD_1 17 +#define MTL_SHININESS 18 +#define MTL_STATE_SIZE 19 + +#define VAP_CMD_0 0 +#define VAP_SE_VAP_CNTL 1 +#define VAP_STATE_SIZE 2 + +/* Replaces a lot of packet info from radeon + */ +#define VTX_CMD_0 0 +#define VTX_VTXFMT_0 1 +#define VTX_VTXFMT_1 2 +#define VTX_TCL_OUTPUT_VTXFMT_0 3 +#define VTX_TCL_OUTPUT_VTXFMT_1 4 +#define VTX_CMD_1 5 +#define VTX_TCL_OUTPUT_COMPSEL 6 +#define VTX_CMD_2 7 +#define VTX_STATE_CNTL 8 +#define VTX_STATE_SIZE 9 + +#define VTX_COLOR(v,n) (((v)>>(R200_VTX_COLOR_0_SHIFT+(n)*2))&\ + R200_VTX_COLOR_MASK) + +/** + * Given the \c R200_SE_VTX_FMT_1 for the current vertex state, determine + * how many components are in texture coordinate \c n. + */ +#define VTX_TEXn_COUNT(v,n) (((v) >> (3 * n)) & 0x07) + +#define MAT_CMD_0 0 +#define MAT_ELT_0 1 +#define MAT_STATE_SIZE 17 + +#define GRD_CMD_0 0 +#define GRD_VERT_GUARD_CLIP_ADJ 1 +#define GRD_VERT_GUARD_DISCARD_ADJ 2 +#define GRD_HORZ_GUARD_CLIP_ADJ 3 +#define GRD_HORZ_GUARD_DISCARD_ADJ 4 +#define GRD_STATE_SIZE 5 + +/* position changes frequently when lighting in modelpos - separate + * out to new state item? + */ +#define LIT_CMD_0 0 +#define LIT_AMBIENT_RED 1 +#define LIT_AMBIENT_GREEN 2 +#define LIT_AMBIENT_BLUE 3 +#define LIT_AMBIENT_ALPHA 4 +#define LIT_DIFFUSE_RED 5 +#define LIT_DIFFUSE_GREEN 6 +#define LIT_DIFFUSE_BLUE 7 +#define LIT_DIFFUSE_ALPHA 8 +#define LIT_SPECULAR_RED 9 +#define LIT_SPECULAR_GREEN 10 +#define LIT_SPECULAR_BLUE 11 +#define LIT_SPECULAR_ALPHA 12 +#define LIT_POSITION_X 13 +#define LIT_POSITION_Y 14 +#define LIT_POSITION_Z 15 +#define LIT_POSITION_W 16 +#define LIT_DIRECTION_X 17 +#define LIT_DIRECTION_Y 18 +#define LIT_DIRECTION_Z 19 +#define LIT_DIRECTION_W 20 +#define LIT_ATTEN_QUADRATIC 21 +#define LIT_ATTEN_LINEAR 22 +#define LIT_ATTEN_CONST 23 +#define LIT_ATTEN_XXX 24 +#define LIT_CMD_1 25 +#define LIT_SPOT_DCD 26 +#define LIT_SPOT_DCM 27 +#define LIT_SPOT_EXPONENT 28 +#define LIT_SPOT_CUTOFF 29 +#define LIT_SPECULAR_THRESH 30 +#define LIT_RANGE_CUTOFF 31 /* ? */ +#define LIT_ATTEN_CONST_INV 32 +#define LIT_STATE_SIZE 33 + +/* Fog + */ +#define FOG_CMD_0 0 +#define FOG_R 1 +#define FOG_C 2 +#define FOG_D 3 +#define FOG_PAD 4 +#define FOG_STATE_SIZE 5 + +/* UCP + */ +#define UCP_CMD_0 0 +#define UCP_X 1 +#define UCP_Y 2 +#define UCP_Z 3 +#define UCP_W 4 +#define UCP_STATE_SIZE 5 + +/* GLT - Global ambient + */ +#define GLT_CMD_0 0 +#define GLT_RED 1 +#define GLT_GREEN 2 +#define GLT_BLUE 3 +#define GLT_ALPHA 4 +#define GLT_STATE_SIZE 5 + +/* EYE + */ +#define EYE_CMD_0 0 +#define EYE_X 1 +#define EYE_Y 2 +#define EYE_Z 3 +#define EYE_RESCALE_FACTOR 4 +#define EYE_STATE_SIZE 5 + +/* CST - constant state + */ +#define CST_CMD_0 0 +#define CST_PP_CNTL_X 1 +#define CST_CMD_1 2 +#define CST_RB3D_DEPTHXY_OFFSET 3 +#define CST_CMD_2 4 +#define CST_RE_AUX_SCISSOR_CNTL 5 +#define CST_CMD_3 6 +#define CST_RE_SCISSOR_TL_0 7 +#define CST_RE_SCISSOR_BR_0 8 +#define CST_CMD_4 9 +#define CST_SE_VAP_CNTL_STATUS 10 +#define CST_CMD_5 11 +#define CST_RE_POINTSIZE 12 +#define CST_CMD_6 13 +#define CST_SE_TCL_INPUT_VTX_0 14 +#define CST_SE_TCL_INPUT_VTX_1 15 +#define CST_SE_TCL_INPUT_VTX_2 16 +#define CST_SE_TCL_INPUT_VTX_3 17 +#define CST_STATE_SIZE 18 + +struct r200_hw_state { + /* Head of the linked list of state atoms. */ + struct r200_state_atom atomlist; + + /* Hardware state, stored as cmdbuf commands: + * -- Need to doublebuffer for + * - reviving state after loss of context + * - eliding noop statechange loops? (except line stipple count) + */ + struct r200_state_atom ctx; + struct r200_state_atom set; + struct r200_state_atom vte; + struct r200_state_atom lin; + struct r200_state_atom msk; + struct r200_state_atom vpt; + struct r200_state_atom vap; + struct r200_state_atom vtx; + struct r200_state_atom tcl; + struct r200_state_atom msl; + struct r200_state_atom tcg; + struct r200_state_atom msc; + struct r200_state_atom cst; + struct r200_state_atom tam; + struct r200_state_atom tf; + struct r200_state_atom tex[6]; + struct r200_state_atom cube[6]; + struct r200_state_atom zbs; + struct r200_state_atom mtl[2]; + struct r200_state_atom mat[9]; + struct r200_state_atom lit[8]; /* includes vec, scl commands */ + struct r200_state_atom ucp[6]; + struct r200_state_atom pix[6]; /* pixshader stages */ + struct r200_state_atom eye; /* eye pos */ + struct r200_state_atom grd; /* guard band clipping */ + struct r200_state_atom fog; + struct r200_state_atom glt; + + int max_state_size; /* Number of bytes necessary for a full state emit. */ + GLboolean is_dirty, all_dirty; +}; + +struct r200_colorbuffer_state { + int roundEnable; +}; + +struct r200_state { + /* Derived state for internal purposes: + */ + struct r200_colorbuffer_state color; + struct r200_depthbuffer_state depth; + struct r200_scissor_state scissor; + struct r200_stencilbuffer_state stencil; + struct r200_stipple_state stipple; + struct r200_texture_state texture; +}; + +/* Need refcounting on dma buffers: + */ +struct r200_dma_buffer { + int refcount; /* the number of retained regions in buf */ + drmBufPtr buf; +}; + +#define GET_START(rvb) (rmesa->radeon.radeonScreen->gart_buffer_offset + \ + (rvb)->address - rmesa->dma.buf0_address + \ + (rvb)->start) + +/* A retained region, eg vertices for indexed vertices. + */ +struct r200_dma_region { + struct r200_dma_buffer *buf; + char *address; /* == buf->address */ + int start, end, ptr; /* offsets from start of buf */ + int aos_start; + int aos_stride; + int aos_size; +}; + +struct r200_dma { + /* Active dma region. Allocations for vertices and retained + * regions come from here. Also used for emitting random vertices, + * these may be flushed by calling flush_current(); + */ + struct r200_dma_region current; + + void (*flush) (r200ContextPtr); + + char *buf0_address; /* start of buf[0], for index calcs */ + GLuint nr_released_bufs; /* flush after so many buffers released */ +}; + +#define R200_CMD_BUF_SZ (8*1024) + +struct r200_store { + GLuint statenr; + GLuint primnr; + char cmd_buf[R200_CMD_BUF_SZ]; + int cmd_used; + int elts_start; +}; + +/* r200_tcl.c + */ +struct r200_tcl_info { + GLuint vertex_format; + GLint last_offset; + GLuint hw_primitive; + + struct r200_dma_region *aos_components[8]; + GLuint nr_aos_components; + + GLuint *Elts; + + struct r200_dma_region indexed_verts; + struct r200_dma_region obj; + struct r200_dma_region rgba; + struct r200_dma_region spec; + struct r200_dma_region fog; + struct r200_dma_region tex[R200_MAX_TEXTURE_UNITS]; + struct r200_dma_region norm; +}; + +/* r200_swtcl.c + */ +struct r200_swtcl_info { + GLuint RenderIndex; + + /** + * Size of a hardware vertex. This is calculated when \c ::vertex_attrs is + * installed in the Mesa state vector. + */ + GLuint vertex_size; + + /** + * Attributes instructing the Mesa TCL pipeline where / how to put vertex + * data in the hardware buffer. + */ + struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX]; + + /** + * Number of elements of \c ::vertex_attrs that are actually used. + */ + GLuint vertex_attr_count; + + /** + * Cached pointer to the buffer where Mesa will store vertex data. + */ + GLubyte *verts; + + /* Fallback rasterization functions + */ + r200_point_func draw_point; + r200_line_func draw_line; + r200_tri_func draw_tri; + + GLuint hw_primitive; + GLenum render_primitive; + GLuint numverts; + + /** + * Offset of the 4UB color data within a hardware (swtcl) vertex. + */ + GLuint coloroffset; + + /** + * Offset of the 3UB specular color data within a hardware (swtcl) vertex. + */ + GLuint specoffset; + + /** + * Should Mesa project vertex data or will the hardware do it? + */ + GLboolean needproj; + + struct r200_dma_region indexed_verts; +}; + +struct r200_ioctl { + GLuint vertex_offset; + GLuint vertex_size; +}; + +#define R200_MAX_PRIMS 64 + +/* Want to keep a cache of these around. Each is parameterized by + * only a single value which has only a small range. Only expect a + * few, so just rescan the list each time? + */ +struct dynfn { + struct dynfn *next, *prev; + int key[2]; + char *code; +}; + +struct dfn_lists { + struct dynfn Vertex2f; + struct dynfn Vertex2fv; + struct dynfn Vertex3f; + struct dynfn Vertex3fv; + struct dynfn Color4ub; + struct dynfn Color4ubv; + struct dynfn Color3ub; + struct dynfn Color3ubv; + struct dynfn Color4f; + struct dynfn Color4fv; + struct dynfn Color3f; + struct dynfn Color3fv; + struct dynfn SecondaryColor3ubEXT; + struct dynfn SecondaryColor3ubvEXT; + struct dynfn SecondaryColor3fEXT; + struct dynfn SecondaryColor3fvEXT; + struct dynfn Normal3f; + struct dynfn Normal3fv; + struct dynfn TexCoord3f; + struct dynfn TexCoord3fv; + struct dynfn TexCoord2f; + struct dynfn TexCoord2fv; + struct dynfn TexCoord1f; + struct dynfn TexCoord1fv; + struct dynfn MultiTexCoord3fARB; + struct dynfn MultiTexCoord3fvARB; + struct dynfn MultiTexCoord2fARB; + struct dynfn MultiTexCoord2fvARB; + struct dynfn MultiTexCoord1fARB; + struct dynfn MultiTexCoord1fvARB; +}; + +struct dfn_generators { + struct dynfn *(*Vertex2f) (GLcontext *, const int *); + struct dynfn *(*Vertex2fv) (GLcontext *, const int *); + struct dynfn *(*Vertex3f) (GLcontext *, const int *); + struct dynfn *(*Vertex3fv) (GLcontext *, const int *); + struct dynfn *(*Color4ub) (GLcontext *, const int *); + struct dynfn *(*Color4ubv) (GLcontext *, const int *); + struct dynfn *(*Color3ub) (GLcontext *, const int *); + struct dynfn *(*Color3ubv) (GLcontext *, const int *); + struct dynfn *(*Color4f) (GLcontext *, const int *); + struct dynfn *(*Color4fv) (GLcontext *, const int *); + struct dynfn *(*Color3f) (GLcontext *, const int *); + struct dynfn *(*Color3fv) (GLcontext *, const int *); + struct dynfn *(*SecondaryColor3ubEXT) (GLcontext *, const int *); + struct dynfn *(*SecondaryColor3ubvEXT) (GLcontext *, const int *); + struct dynfn *(*SecondaryColor3fEXT) (GLcontext *, const int *); + struct dynfn *(*SecondaryColor3fvEXT) (GLcontext *, const int *); + struct dynfn *(*Normal3f) (GLcontext *, const int *); + struct dynfn *(*Normal3fv) (GLcontext *, const int *); + struct dynfn *(*TexCoord3f) (GLcontext *, const int *); + struct dynfn *(*TexCoord3fv) (GLcontext *, const int *); + struct dynfn *(*TexCoord2f) (GLcontext *, const int *); + struct dynfn *(*TexCoord2fv) (GLcontext *, const int *); + struct dynfn *(*TexCoord1f) (GLcontext *, const int *); + struct dynfn *(*TexCoord1fv) (GLcontext *, const int *); + struct dynfn *(*MultiTexCoord3fARB) (GLcontext *, const int *); + struct dynfn *(*MultiTexCoord3fvARB) (GLcontext *, const int *); + struct dynfn *(*MultiTexCoord2fARB) (GLcontext *, const int *); + struct dynfn *(*MultiTexCoord2fvARB) (GLcontext *, const int *); + struct dynfn *(*MultiTexCoord1fARB) (GLcontext *, const int *); + struct dynfn *(*MultiTexCoord1fvARB) (GLcontext *, const int *); +}; + +struct r200_prim { + GLuint start; + GLuint end; + GLuint prim; +}; + + /* A maximum total of 29 elements per vertex: 3 floats for position, 3 + * floats for normal, 4 floats for color, 4 bytes for secondary color, + * 3 floats for each texture unit (18 floats total). + * + * we maybe need add. 4 to prevent segfault if someone specifies + * GL_TEXTURE6/GL_TEXTURE7 (esp. for the codegen-path) (FIXME: ) + * + * The position data is never actually stored here, so 3 elements could be + * trimmed out of the buffer. + */ + +#define R200_MAX_VERTEX_SIZE ((3*6)+11) + +struct r200_vbinfo { + GLint counter, initial_counter; + GLint *dmaptr; + void (*notify) (void); + GLint vertex_size; + + union { + float f; + int i; + r200_color_t color; + } vertex[R200_MAX_VERTEX_SIZE]; + + GLfloat *normalptr; + GLfloat *floatcolorptr; + r200_color_t *colorptr; + GLfloat *floatspecptr; + r200_color_t *specptr; + GLfloat *texcoordptr[8]; /* 6 (TMU) + 2 for r200_vtxfmt_c.c when GL_TEXTURE6/7 */ + + GLenum *prim; /* &ctx->Driver.CurrentExecPrimitive */ + GLuint primflags; + GLboolean enabled; /* *_NO_VTXFMT / *_NO_TCL env vars */ + GLboolean installed; + GLboolean fell_back; + GLboolean recheck; + GLint nrverts; + GLuint vtxfmt_0, vtxfmt_1; + + GLuint installed_vertex_format; + GLuint installed_color_3f_sz; + + struct r200_prim primlist[R200_MAX_PRIMS]; + int nrprims; + + struct dfn_lists dfn_cache; + struct dfn_generators codegen; + GLvertexformat vtxfmt; +}; + +/** + * R200 context structure. + */ +struct r200_context { + struct radeon_context radeon; /* parent class, must be first */ + + /* Driver and hardware state management + */ + struct r200_hw_state hw; + struct r200_state state; + + /* Texture object bookkeeping + */ + unsigned nr_heaps; + driTexHeap *texture_heaps[R200_NR_TEX_HEAPS]; + driTextureObject swapped; + int texture_depth; + float initialMaxAnisotropy; + + /* Rasterization and vertex state: + */ + GLuint NewGLState; + + /* Vertex buffers + */ + struct r200_ioctl ioctl; + struct r200_dma dma; + struct r200_store store; + GLboolean save_on_next_unlock; + + /* Clientdata textures; + */ + GLuint prefer_gart_client_texturing; + + /* TCL stuff + */ + GLmatrix TexGenMatrix[R200_MAX_TEXTURE_UNITS]; + GLboolean recheck_texgen[R200_MAX_TEXTURE_UNITS]; + GLboolean TexGenNeedNormals[R200_MAX_TEXTURE_UNITS]; + GLuint TexMatEnabled; + GLuint TexMatCompSel; + GLuint TexGenEnabled; + GLuint TexGenInputs; + GLuint TexGenCompSel; + GLmatrix tmpmat; + + /* r200_tcl.c + */ + struct r200_tcl_info tcl; + + /* r200_swtcl.c + */ + struct r200_swtcl_info swtcl; + + /* r200_vtxfmt.c + */ + struct r200_vbinfo vb; +}; + +#define R200_CONTEXT(ctx) ((r200ContextPtr)(ctx->DriverCtx)) + +extern void r200DestroyContext(__DRIcontextPrivate * driContextPriv); +extern GLboolean r200CreateContext(const __GLcontextModes * glVisual, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate); + +#endif +#endif /* __R200_CONTEXT_H__ */ diff --git a/src/mesa/drivers/dri/r300/r200_ioctl.h b/src/mesa/drivers/dri/r300/r200_ioctl.h new file mode 100644 index 00000000000..db7bd7697f1 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r200_ioctl.h @@ -0,0 +1,204 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Keith Whitwell + */ + +#ifndef __R200_IOCTL_H__ +#define __R200_IOCTL_H__ + +#include "simple_list.h" +#include "radeon_dri.h" +#include "radeon_lock.h" + +#include "xf86drm.h" +#include "drm.h" +#include "r200_context.h" +#include "radeon_drm.h" + +extern void r200EmitState(r200ContextPtr rmesa); +extern void r200EmitVertexAOS(r200ContextPtr rmesa, + GLuint vertex_size, GLuint offset); + +extern void r200EmitVbufPrim(r200ContextPtr rmesa, + GLuint primitive, GLuint vertex_nr); + +extern void r200FlushElts(r200ContextPtr rmesa); + +extern GLushort *r200AllocEltsOpenEnded(r200ContextPtr rmesa, + GLuint primitive, GLuint min_nr); + +extern void r200EmitAOS(r200ContextPtr rmesa, + struct r200_dma_region **regions, + GLuint n, GLuint offset); + +extern void r200EmitBlit(r200ContextPtr rmesa, + GLuint color_fmt, + GLuint src_pitch, + GLuint src_offset, + GLuint dst_pitch, + GLuint dst_offset, + GLint srcx, GLint srcy, + GLint dstx, GLint dsty, GLuint w, GLuint h); + +extern void r200EmitWait(r200ContextPtr rmesa, GLuint flags); + +extern void r200FlushCmdBuf(r200ContextPtr rmesa, const char *); +extern int r200FlushCmdBufLocked(r200ContextPtr rmesa, const char *caller); +extern void r200Flush(GLcontext * ctx); + +extern void r200RefillCurrentDmaRegion(r200ContextPtr rmesa); + +extern void r200AllocDmaRegion(r200ContextPtr rmesa, + struct r200_dma_region *region, + int bytes, int alignment); + +extern void r200AllocDmaRegionVerts(r200ContextPtr rmesa, + struct r200_dma_region *region, + int numverts, int vertsize, int alignment); + +extern void r200ReleaseDmaRegion(r200ContextPtr rmesa, + struct r200_dma_region *region, + const char *caller); + +extern void r200WaitForVBlank(r200ContextPtr rmesa); +extern void r200InitIoctlFuncs(struct dd_function_table *functions); + +extern void *r200AllocateMemoryMESA(__DRInativeDisplay * dpy, int scrn, + GLsizei size, GLfloat readfreq, + GLfloat writefreq, GLfloat priority); +extern void r200FreeMemoryMESA(__DRInativeDisplay * dpy, int scrn, + GLvoid * pointer); +extern GLuint r200GetMemoryOffsetMESA(__DRInativeDisplay * dpy, int scrn, + const GLvoid * pointer); + +extern GLboolean r200IsGartMemory(r200ContextPtr rmesa, const GLvoid * pointer, + GLint size); + +extern GLuint r200GartOffsetFromVirtual(r200ContextPtr rmesa, + const GLvoid * pointer); + +void r200SaveHwState(r200ContextPtr radeon); +void r200SetUpAtomList(r200ContextPtr rmesa); + +/* ================================================================ + * Helper macros: + */ + +/* Close off the last primitive, if it exists. + */ +#define R200_NEWPRIM( rmesa ) \ +do { \ + if ( rmesa->dma.flush ) \ + rmesa->dma.flush( rmesa ); \ +} while (0) + +/* Can accomodate several state changes and primitive changes without + * actually firing the buffer. + */ +#define R200_STATECHANGE( rmesa, ATOM ) \ +do { \ + R200_NEWPRIM( rmesa ); \ + rmesa->hw.ATOM.dirty = GL_TRUE; \ + rmesa->hw.is_dirty = GL_TRUE; \ +} while (0) + +#define R200_DB_STATE( ATOM ) \ + memcpy( rmesa->hw.ATOM.lastcmd, rmesa->hw.ATOM.cmd, \ + rmesa->hw.ATOM.cmd_size * 4) + +static __inline int R200_DB_STATECHANGE(r200ContextPtr rmesa, + struct r200_state_atom *atom) +{ + if (memcmp(atom->cmd, atom->lastcmd, atom->cmd_size * 4)) { + int *tmp; + R200_NEWPRIM(rmesa); + atom->dirty = GL_TRUE; + rmesa->hw.is_dirty = GL_TRUE; + tmp = atom->cmd; + atom->cmd = atom->lastcmd; + atom->lastcmd = tmp; + return 1; + } else + return 0; +} + +/* Fire the buffered vertices no matter what. + */ +#define R200_FIREVERTICES( r200 ) \ +do { \ + if ( (r200)->store.cmd_used || (r200)->dma.flush ) { \ + radeonFlush( (r200)->radeon.glCtx ); \ + } \ +} while (0) + +/* Command lengths. Note that any time you ensure ELTS_BUFSZ or VBUF_BUFSZ + * are available, you will also be adding an rmesa->state.max_state_size because + * r200EmitState is called from within r200EmitVbufPrim and r200FlushElts. + */ +#define AOS_BUFSZ(nr) ((3 + ((nr / 2) * 3) + ((nr & 1) * 2)) * sizeof(int)) +#define VERT_AOS_BUFSZ (5 * sizeof(int)) +#define ELTS_BUFSZ(nr) (12 + nr * 2) +#define VBUF_BUFSZ (3 * sizeof(int)) + +/* Ensure that a minimum amount of space is available in the command buffer. + * This is used to ensure atomicity of state updates with the rendering requests + * that rely on them. + * + * An alternative would be to implement a "soft lock" such that when the buffer + * wraps at an inopportune time, we grab the lock, flush the current buffer, + * and hang on to the lock until the critical section is finished and we flush + * the buffer again and unlock. + */ +static __inline void r200EnsureCmdBufSpace(r200ContextPtr rmesa, int bytes) +{ + if (rmesa->store.cmd_used + bytes > R200_CMD_BUF_SZ) + r200FlushCmdBuf(rmesa, __FUNCTION__); + assert(bytes <= R200_CMD_BUF_SZ); +} + +/* Alloc space in the command buffer + */ +static __inline char *r200AllocCmdBuf(r200ContextPtr rmesa, + int bytes, const char *where) +{ + char *head; + + if (rmesa->store.cmd_used + bytes > R200_CMD_BUF_SZ) + r200FlushCmdBuf(rmesa, where); + + head = rmesa->store.cmd_buf + rmesa->store.cmd_used; + rmesa->store.cmd_used += bytes; + assert(rmesa->store.cmd_used <= R200_CMD_BUF_SZ); + return head; +} + +#endif /* __R200_IOCTL_H__ */ diff --git a/src/mesa/drivers/dri/r300/r200_reg.h b/src/mesa/drivers/dri/r300/r200_reg.h new file mode 100644 index 00000000000..1336e961ac6 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r200_reg.h @@ -0,0 +1,1423 @@ +/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_reg.h,v 1.2 2002/12/16 16:18:54 dawes Exp $ */ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +*/ + +#ifndef _R200_REG_H_ +#define _R200_REG_H_ + +#define R200_PP_MISC 0x1c14 +#define R200_REF_ALPHA_MASK 0x000000ff +#define R200_ALPHA_TEST_FAIL (0 << 8) +#define R200_ALPHA_TEST_LESS (1 << 8) +#define R200_ALPHA_TEST_LEQUAL (2 << 8) +#define R200_ALPHA_TEST_EQUAL (3 << 8) +#define R200_ALPHA_TEST_GEQUAL (4 << 8) +#define R200_ALPHA_TEST_GREATER (5 << 8) +#define R200_ALPHA_TEST_NEQUAL (6 << 8) +#define R200_ALPHA_TEST_PASS (7 << 8) +#define R200_ALPHA_TEST_OP_MASK (7 << 8) +#define R200_CHROMA_FUNC_FAIL (0 << 16) +#define R200_CHROMA_FUNC_PASS (1 << 16) +#define R200_CHROMA_FUNC_NEQUAL (2 << 16) +#define R200_CHROMA_FUNC_EQUAL (3 << 16) +#define R200_CHROMA_KEY_NEAREST (0 << 18) +#define R200_CHROMA_KEY_ZERO (1 << 18) +#define R200_RIGHT_HAND_CUBE_D3D (0 << 24) +#define R200_RIGHT_HAND_CUBE_OGL (1 << 24) +#define R200_PP_FOG_COLOR 0x1c18 +#define R200_FOG_COLOR_MASK 0x00ffffff +#define R200_FOG_VERTEX (0 << 24) +#define R200_FOG_TABLE (1 << 24) +#define R200_FOG_USE_DEPTH (0 << 25) +#define R200_FOG_USE_W (1 << 25) +#define R200_FOG_USE_DIFFUSE_ALPHA (2 << 25) +#define R200_FOG_USE_SPEC_ALPHA (3 << 25) +#define R200_FOG_USE_VTX_FOG (4 << 25) +#define R200_FOG_USE_MASK (7 << 25) +#define R200_RE_SOLID_COLOR 0x1c1c +#define R200_RB3D_BLENDCNTL 0x1c20 +#define R200_COMB_FCN_MASK (7 << 12) +#define R200_COMB_FCN_ADD_CLAMP (0 << 12) +#define R200_COMB_FCN_ADD_NOCLAMP (1 << 12) +#define R200_COMB_FCN_SUB_CLAMP (2 << 12) +#define R200_COMB_FCN_SUB_NOCLAMP (3 << 12) +#define R200_COMB_FCN_MIN (4 << 12) +#define R200_COMB_FCN_MAX (5 << 12) +#define R200_COMB_FCN_RSUB_CLAMP (6 << 12) +#define R200_COMB_FCN_RSUB_NOCLAMP (7 << 12) +#define R200_BLEND_GL_ZERO (32) +#define R200_BLEND_GL_ONE (33) +#define R200_BLEND_GL_SRC_COLOR (34) +#define R200_BLEND_GL_ONE_MINUS_SRC_COLOR (35) +#define R200_BLEND_GL_DST_COLOR (36) +#define R200_BLEND_GL_ONE_MINUS_DST_COLOR (37) +#define R200_BLEND_GL_SRC_ALPHA (38) +#define R200_BLEND_GL_ONE_MINUS_SRC_ALPHA (39) +#define R200_BLEND_GL_DST_ALPHA (40) +#define R200_BLEND_GL_ONE_MINUS_DST_ALPHA (41) +#define R200_BLEND_GL_SRC_ALPHA_SATURATE (42) /* src factor only */ +#define R200_BLEND_GL_CONST_COLOR (43) +#define R200_BLEND_GL_ONE_MINUS_CONST_COLOR (44) +#define R200_BLEND_GL_CONST_ALPHA (45) +#define R200_BLEND_GL_ONE_MINUS_CONST_ALPHA (46) +#define R200_BLEND_MASK (63) +#define R200_SRC_BLEND_SHIFT (16) +#define R200_DST_BLEND_SHIFT (24) +#define R200_RB3D_DEPTHOFFSET 0x1c24 +#define R200_RB3D_DEPTHPITCH 0x1c28 +#define R200_DEPTHPITCH_MASK 0x00001ff8 +#define R200_DEPTH_ENDIAN_NO_SWAP (0 << 18) +#define R200_DEPTH_ENDIAN_WORD_SWAP (1 << 18) +#define R200_DEPTH_ENDIAN_DWORD_SWAP (2 << 18) +#define R200_RB3D_ZSTENCILCNTL 0x1c2c +#define R200_DEPTH_FORMAT_MASK (0xf << 0) +#define R200_DEPTH_FORMAT_16BIT_INT_Z (0 << 0) +#define R200_DEPTH_FORMAT_24BIT_INT_Z (2 << 0) +#define R200_DEPTH_FORMAT_24BIT_FLOAT_Z (3 << 0) +#define R200_DEPTH_FORMAT_32BIT_INT_Z (4 << 0) +#define R200_DEPTH_FORMAT_32BIT_FLOAT_Z (5 << 0) +#define R200_DEPTH_FORMAT_24BIT_FLOAT_W (9 << 0) +#define R200_DEPTH_FORMAT_32BIT_FLOAT_W (11 << 0) +#define R200_Z_TEST_NEVER (0 << 4) +#define R200_Z_TEST_LESS (1 << 4) +#define R200_Z_TEST_LEQUAL (2 << 4) +#define R200_Z_TEST_EQUAL (3 << 4) +#define R200_Z_TEST_GEQUAL (4 << 4) +#define R200_Z_TEST_GREATER (5 << 4) +#define R200_Z_TEST_NEQUAL (6 << 4) +#define R200_Z_TEST_ALWAYS (7 << 4) +#define R200_Z_TEST_MASK (7 << 4) +#define R200_STENCIL_TEST_NEVER (0 << 12) +#define R200_STENCIL_TEST_LESS (1 << 12) +#define R200_STENCIL_TEST_LEQUAL (2 << 12) +#define R200_STENCIL_TEST_EQUAL (3 << 12) +#define R200_STENCIL_TEST_GEQUAL (4 << 12) +#define R200_STENCIL_TEST_GREATER (5 << 12) +#define R200_STENCIL_TEST_NEQUAL (6 << 12) +#define R200_STENCIL_TEST_ALWAYS (7 << 12) +#define R200_STENCIL_TEST_MASK (0x7 << 12) +#define R200_STENCIL_FAIL_KEEP (0 << 16) +#define R200_STENCIL_FAIL_ZERO (1 << 16) +#define R200_STENCIL_FAIL_REPLACE (2 << 16) +#define R200_STENCIL_FAIL_INC (3 << 16) +#define R200_STENCIL_FAIL_DEC (4 << 16) +#define R200_STENCIL_FAIL_INVERT (5 << 16) +#define R200_STENCIL_FAIL_INC_WRAP (6 << 16) +#define R200_STENCIL_FAIL_DEC_WRAP (7 << 16) +#define R200_STENCIL_FAIL_MASK (0x7 << 16) +#define R200_STENCIL_ZPASS_KEEP (0 << 20) +#define R200_STENCIL_ZPASS_ZERO (1 << 20) +#define R200_STENCIL_ZPASS_REPLACE (2 << 20) +#define R200_STENCIL_ZPASS_INC (3 << 20) +#define R200_STENCIL_ZPASS_DEC (4 << 20) +#define R200_STENCIL_ZPASS_INVERT (5 << 20) +#define R200_STENCIL_ZPASS_INC_WRAP (6 << 20) +#define R200_STENCIL_ZPASS_DEC_WRAP (7 << 20) +#define R200_STENCIL_ZPASS_MASK (0x7 << 20) +#define R200_STENCIL_ZFAIL_KEEP (0 << 24) +#define R200_STENCIL_ZFAIL_ZERO (1 << 24) +#define R200_STENCIL_ZFAIL_REPLACE (2 << 24) +#define R200_STENCIL_ZFAIL_INC (3 << 24) +#define R200_STENCIL_ZFAIL_DEC (4 << 24) +#define R200_STENCIL_ZFAIL_INVERT (5 << 24) +#define R200_STENCIL_ZFAIL_INC_WRAP (6 << 24) +#define R200_STENCIL_ZFAIL_DEC_WRAP (7 << 24) +#define R200_STENCIL_ZFAIL_MASK (0x7 << 24) +#define R200_Z_WRITE_ENABLE (1 << 30) +/*gap*/ +#define R200_PP_CNTL 0x1c38 +#define R200_TEX_0_ENABLE 0x00000010 +#define R200_TEX_1_ENABLE 0x00000020 +#define R200_TEX_2_ENABLE 0x00000040 +#define R200_TEX_3_ENABLE 0x00000080 +#define R200_TEX_4_ENABLE 0x00000100 +#define R200_TEX_5_ENABLE 0x00000200 +#define R200_TEX_ENABLE_MASK 0x000003f0 +#define R200_FILTER_ROUND_MODE_MASK 0x00000400 +#define R200_TEX_BLEND_7_ENABLE 0x00000800 +#define R200_TEX_BLEND_0_ENABLE 0x00001000 +#define R200_TEX_BLEND_1_ENABLE 0x00002000 +#define R200_TEX_BLEND_2_ENABLE 0x00004000 +#define R200_TEX_BLEND_3_ENABLE 0x00008000 +#define R200_TEX_BLEND_4_ENABLE 0x00010000 +#define R200_TEX_BLEND_5_ENABLE 0x00020000 +#define R200_TEX_BLEND_6_ENABLE 0x00040000 +#define R200_MULTI_PASS_ENABLE 0x00080000 +#define R200_SPECULAR_ENABLE 0x00200000 +#define R200_FOG_ENABLE 0x00400000 +#define R200_ALPHA_TEST_ENABLE 0x00800000 +#define R200_ANTI_ALIAS_NONE 0x00000000 +#define R200_ANTI_ALIAS_LINE 0x01000000 +#define R200_ANTI_ALIAS_POLY 0x02000000 +#define R200_ANTI_ALIAS_MASK 0x03000000 +#define R200_RB3D_CNTL 0x1c3c +#define R200_ALPHA_BLEND_ENABLE (1 << 0) +#define R200_PLANE_MASK_ENABLE (1 << 1) +#define R200_DITHER_ENABLE (1 << 2) +#define R200_ROUND_ENABLE (1 << 3) +#define R200_SCALE_DITHER_ENABLE (1 << 4) +#define R200_DITHER_INIT (1 << 5) +#define R200_ROP_ENABLE (1 << 6) +#define R200_STENCIL_ENABLE (1 << 7) +#define R200_Z_ENABLE (1 << 8) +#define R200_DEPTH_XZ_OFFEST_ENABLE (1 << 9) +#define R200_COLOR_FORMAT_ARGB1555 (3 << 10) +#define R200_COLOR_FORMAT_RGB565 (4 << 10) +#define R200_COLOR_FORMAT_ARGB8888 (6 << 10) +#define R200_COLOR_FORMAT_RGB332 (7 << 10) +#define R200_COLOR_FORMAT_Y8 (8 << 10) +#define R200_COLOR_FORMAT_RGB8 (9 << 10) +#define R200_COLOR_FORMAT_YUV422_VYUY (11 << 10) +#define R200_COLOR_FORMAT_YUV422_YVYU (12 << 10) +#define R200_COLOR_FORMAT_aYUV444 (14 << 10) +#define R200_COLOR_FORMAT_ARGB4444 (15 << 10) +#define R200_CLRCMP_FLIP_ENABLE (1 << 14) +#define R200_SEPARATE_ALPHA_ENABLE (1 << 16) +#define R200_RB3D_COLOROFFSET 0x1c40 +#define R200_COLOROFFSET_MASK 0xfffffff0 +#define R200_RE_WIDTH_HEIGHT 0x1c44 +#define R200_RE_WIDTH_SHIFT 0 +#define R200_RE_HEIGHT_SHIFT 16 +#define R200_RB3D_COLORPITCH 0x1c48 +#define R200_COLORPITCH_MASK 0x000001ff8 +#define R200_COLOR_ENDIAN_NO_SWAP (0 << 18) +#define R200_COLOR_ENDIAN_WORD_SWAP (1 << 18) +#define R200_COLOR_ENDIAN_DWORD_SWAP (2 << 18) +#define R200_SE_CNTL 0x1c4c +#define R200_FFACE_CULL_CW (0 << 0) +#define R200_FFACE_CULL_CCW (1 << 0) +#define R200_FFACE_CULL_DIR_MASK (1 << 0) +#define R200_BFACE_CULL (0 << 1) +#define R200_BFACE_SOLID (3 << 1) +#define R200_FFACE_CULL (0 << 3) +#define R200_FFACE_SOLID (3 << 3) +#define R200_FFACE_CULL_MASK (3 << 3) +#define R200_FLAT_SHADE_VTX_0 (0 << 6) +#define R200_FLAT_SHADE_VTX_1 (1 << 6) +#define R200_FLAT_SHADE_VTX_2 (2 << 6) +#define R200_FLAT_SHADE_VTX_LAST (3 << 6) +#define R200_DIFFUSE_SHADE_SOLID (0 << 8) +#define R200_DIFFUSE_SHADE_FLAT (1 << 8) +#define R200_DIFFUSE_SHADE_GOURAUD (2 << 8) +#define R200_DIFFUSE_SHADE_MASK (3 << 8) +#define R200_ALPHA_SHADE_SOLID (0 << 10) +#define R200_ALPHA_SHADE_FLAT (1 << 10) +#define R200_ALPHA_SHADE_GOURAUD (2 << 10) +#define R200_ALPHA_SHADE_MASK (3 << 10) +#define R200_SPECULAR_SHADE_SOLID (0 << 12) +#define R200_SPECULAR_SHADE_FLAT (1 << 12) +#define R200_SPECULAR_SHADE_GOURAUD (2 << 12) +#define R200_SPECULAR_SHADE_MASK (3 << 12) +#define R200_FOG_SHADE_SOLID (0 << 14) +#define R200_FOG_SHADE_FLAT (1 << 14) +#define R200_FOG_SHADE_GOURAUD (2 << 14) +#define R200_FOG_SHADE_MASK (3 << 14) +#define R200_ZBIAS_ENABLE_POINT (1 << 16) +#define R200_ZBIAS_ENABLE_LINE (1 << 17) +#define R200_ZBIAS_ENABLE_TRI (1 << 18) +#define R200_WIDELINE_ENABLE (1 << 20) +#define R200_VTX_PIX_CENTER_D3D (0 << 27) +#define R200_VTX_PIX_CENTER_OGL (1 << 27) +#define R200_ROUND_MODE_TRUNC (0 << 28) +#define R200_ROUND_MODE_ROUND (1 << 28) +#define R200_ROUND_MODE_ROUND_EVEN (2 << 28) +#define R200_ROUND_MODE_ROUND_ODD (3 << 28) +#define R200_ROUND_PREC_16TH_PIX (0 << 30) +#define R200_ROUND_PREC_8TH_PIX (1 << 30) +#define R200_ROUND_PREC_4TH_PIX (2 << 30) +#define R200_ROUND_PREC_HALF_PIX (3 << 30) +#define R200_RE_CNTL 0x1c50 +#define R200_STIPPLE_ENABLE 0x1 +#define R200_SCISSOR_ENABLE 0x2 +#define R200_PATTERN_ENABLE 0x4 +#define R200_PERSPECTIVE_ENABLE 0x8 +#define R200_POINT_SMOOTH 0x20 +#define R200_VTX_STQ0_D3D 0x00010000 +#define R200_VTX_STQ1_D3D 0x00040000 +#define R200_VTX_STQ2_D3D 0x00100000 +#define R200_VTX_STQ3_D3D 0x00400000 +#define R200_VTX_STQ4_D3D 0x01000000 +#define R200_VTX_STQ5_D3D 0x04000000 +/* gap */ +#define R200_RE_STIPPLE_ADDR 0x1cc8 +#define R200_RE_STIPPLE_DATA 0x1ccc +#define R200_RE_LINE_PATTERN 0x1cd0 +#define R200_LINE_PATTERN_MASK 0x0000ffff +#define R200_LINE_REPEAT_COUNT_SHIFT 16 +#define R200_LINE_PATTERN_START_SHIFT 24 +#define R200_LINE_PATTERN_LITTLE_BIT_ORDER (0 << 28) +#define R200_LINE_PATTERN_BIG_BIT_ORDER (1 << 28) +#define R200_LINE_PATTERN_AUTO_RESET (1 << 29) +#define R200_RE_LINE_STATE 0x1cd4 +#define R200_LINE_CURRENT_PTR_SHIFT 0 +#define R200_LINE_CURRENT_COUNT_SHIFT 8 +#define R200_RE_SCISSOR_TL_0 0x1cd8 +#define R200_RE_SCISSOR_BR_0 0x1cdc +#define R200_RE_SCISSOR_TL_1 0x1ce0 +#define R200_RE_SCISSOR_BR_1 0x1ce4 +#define R200_RE_SCISSOR_TL_2 0x1ce8 +#define R200_RE_SCISSOR_BR_2 0x1cec +/* gap */ +#define R200_RB3D_DEPTHXY_OFFSET 0x1d60 +#define R200_DEPTHX_SHIFT 0 +#define R200_DEPTHY_SHIFT 16 +/* gap */ +#define R200_RB3D_STENCILREFMASK 0x1d7c +#define R200_STENCIL_REF_SHIFT 0 +#define R200_STENCIL_REF_MASK (0xff << 0) +#define R200_STENCIL_MASK_SHIFT 16 +#define R200_STENCIL_VALUE_MASK (0xff << 16) +#define R200_STENCIL_WRITEMASK_SHIFT 24 +#define R200_STENCIL_WRITE_MASK (0xff << 24) +#define R200_RB3D_ROPCNTL 0x1d80 +#define R200_ROP_MASK (15 << 8) +#define R200_ROP_CLEAR (0 << 8) +#define R200_ROP_NOR (1 << 8) +#define R200_ROP_AND_INVERTED (2 << 8) +#define R200_ROP_COPY_INVERTED (3 << 8) +#define R200_ROP_AND_REVERSE (4 << 8) +#define R200_ROP_INVERT (5 << 8) +#define R200_ROP_XOR (6 << 8) +#define R200_ROP_NAND (7 << 8) +#define R200_ROP_AND (8 << 8) +#define R200_ROP_EQUIV (9 << 8) +#define R200_ROP_NOOP (10 << 8) +#define R200_ROP_OR_INVERTED (11 << 8) +#define R200_ROP_COPY (12 << 8) +#define R200_ROP_OR_REVERSE (13 << 8) +#define R200_ROP_OR (14 << 8) +#define R200_ROP_SET (15 << 8) +#define R200_RB3D_PLANEMASK 0x1d84 +/* gap */ +#define R200_SE_VPORT_XSCALE 0x1d98 +#define R200_SE_VPORT_XOFFSET 0x1d9c +#define R200_SE_VPORT_YSCALE 0x1da0 +#define R200_SE_VPORT_YOFFSET 0x1da4 +#define R200_SE_VPORT_ZSCALE 0x1da8 +#define R200_SE_VPORT_ZOFFSET 0x1dac +#define R200_SE_ZBIAS_FACTOR 0x1db0 +#define R200_SE_ZBIAS_CONSTANT 0x1db4 +#define R200_SE_LINE_WIDTH 0x1db8 +#define R200_LINE_WIDTH_SHIFT 0x00000000 +#define R200_MINPOINTSIZE_SHIFT 0x00000010 +/* gap */ +#define R200_SE_VAP_CNTL 0x2080 +#define R200_VAP_TCL_ENABLE 0x00000001 +#define R200_VAP_SINGLE_BUF_STATE_ENABLE 0x00000010 +#define R200_VAP_FORCE_W_TO_ONE 0x00010000 +#define R200_VAP_D3D_TEX_DEFAULT 0x00020000 +#define R200_VAP_VF_MAX_VTX_NUM__SHIFT 18 +#define R200_VAP_DX_CLIP_SPACE_DEF 0x00400000 +#define R200_SE_VF_CNTL 0x2084 +#define R200_VF_PRIM_NONE 0x00000000 +#define R200_VF_PRIM_POINTS 0x00000001 +#define R200_VF_PRIM_LINES 0x00000002 +#define R200_VF_PRIM_LINE_STRIP 0x00000003 +#define R200_VF_PRIM_TRIANGLES 0x00000004 +#define R200_VF_PRIM_TRIANGLE_FAN 0x00000005 +#define R200_VF_PRIM_TRIANGLE_STRIP 0x00000006 +#define R200_VF_PRIM_RECT_LIST 0x00000008 +#define R200_VF_PRIM_3VRT_POINTS 0x00000009 +#define R200_VF_PRIM_3VRT_LINES 0x0000000a +#define R200_VF_PRIM_POINT_SPRITES 0x0000000b +#define R200_VF_PRIM_LINE_LOOP 0x0000000c +#define R200_VF_PRIM_QUADS 0x0000000d +#define R200_VF_PRIM_QUAD_STRIP 0x0000000e +#define R200_VF_PRIM_POLYGON 0x0000000f +#define R200_VF_PRIM_MASK 0x0000000f +#define R200_VF_PRIM_WALK_IND 0x00000010 +#define R200_VF_PRIM_WALK_LIST 0x00000020 +#define R200_VF_PRIM_WALK_RING 0x00000030 +#define R200_VF_PRIM_WALK_MASK 0x00000030 +#define R200_VF_COLOR_ORDER_RGBA 0x00000040 +#define R200_VF_TCL_OUTPUT_VTX_ENABLE 0x00000200 +#define R200_VF_INDEX_SZ_4 0x00000800 +#define R200_VF_VERTEX_NUMBER_MASK 0xffff0000 +#define R200_VF_VERTEX_NUMBER_SHIFT 16 +#define R200_SE_VTX_FMT_0 0x2088 +#define R200_VTX_XY 0 /* always have xy */ +#define R200_VTX_Z0 (1<<0) +#define R200_VTX_W0 (1<<1) +#define R200_VTX_WEIGHT_COUNT_SHIFT (2) +#define R200_VTX_PV_MATRIX_SEL (1<<5) +#define R200_VTX_N0 (1<<6) +#define R200_VTX_POINT_SIZE (1<<7) +#define R200_VTX_DISCRETE_FOG (1<<8) +#define R200_VTX_SHININESS_0 (1<<9) +#define R200_VTX_SHININESS_1 (1<<10) +#define R200_VTX_COLOR_NOT_PRESENT 0 +#define R200_VTX_PK_RGBA 1 +#define R200_VTX_FP_RGB 2 +#define R200_VTX_FP_RGBA 3 +#define R200_VTX_COLOR_MASK 3 +#define R200_VTX_COLOR_0_SHIFT 11 +#define R200_VTX_COLOR_1_SHIFT 13 +#define R200_VTX_COLOR_2_SHIFT 15 +#define R200_VTX_COLOR_3_SHIFT 17 +#define R200_VTX_COLOR_4_SHIFT 19 +#define R200_VTX_COLOR_5_SHIFT 21 +#define R200_VTX_COLOR_6_SHIFT 23 +#define R200_VTX_COLOR_7_SHIFT 25 +#define R200_VTX_XY1 (1<<28) +#define R200_VTX_Z1 (1<<29) +#define R200_VTX_W1 (1<<30) +#define R200_VTX_N1 (1<<31) +#define R200_SE_VTX_FMT_1 0x208c +#define R200_VTX_TEX0_COMP_CNT_SHIFT 0 +#define R200_VTX_TEX1_COMP_CNT_SHIFT 3 +#define R200_VTX_TEX2_COMP_CNT_SHIFT 6 +#define R200_VTX_TEX3_COMP_CNT_SHIFT 9 +#define R200_VTX_TEX4_COMP_CNT_SHIFT 12 +#define R200_VTX_TEX5_COMP_CNT_SHIFT 15 +#define R200_SE_TCL_OUTPUT_VTX_FMT_0 0x2090 +#define R200_SE_TCL_OUTPUT_VTX_FMT_1 0x2094 +/* gap */ +#define R200_SE_VTE_CNTL 0x20b0 +#define R200_VPORT_X_SCALE_ENA 0x00000001 +#define R200_VPORT_X_OFFSET_ENA 0x00000002 +#define R200_VPORT_Y_SCALE_ENA 0x00000004 +#define R200_VPORT_Y_OFFSET_ENA 0x00000008 +#define R200_VPORT_Z_SCALE_ENA 0x00000010 +#define R200_VPORT_Z_OFFSET_ENA 0x00000020 +#define R200_VTX_XY_FMT 0x00000100 +#define R200_VTX_Z_FMT 0x00000200 +#define R200_VTX_W0_FMT 0x00000400 +#define R200_VTX_W0_NORMALIZE 0x00000800 +#define R200_VTX_ST_DENORMALIZED 0x00001000 +/* gap */ +#define R200_SE_VTX_NUM_ARRAYS 0x20c0 +#define R200_SE_VTX_AOS_ATTR01 0x20c4 +#define R200_SE_VTX_AOS_ADDR0 0x20c8 +#define R200_SE_VTX_AOS_ADDR1 0x20cc +#define R200_SE_VTX_AOS_ATTR23 0x20d0 +#define R200_SE_VTX_AOS_ADDR2 0x20d4 +#define R200_SE_VTX_AOS_ADDR3 0x20d8 +#define R200_SE_VTX_AOS_ATTR45 0x20dc +#define R200_SE_VTX_AOS_ADDR4 0x20e0 +#define R200_SE_VTX_AOS_ADDR5 0x20e4 +#define R200_SE_VTX_AOS_ATTR67 0x20e8 +#define R200_SE_VTX_AOS_ADDR6 0x20ec +#define R200_SE_VTX_AOS_ADDR7 0x20f0 +#define R200_SE_VTX_AOS_ATTR89 0x20f4 +#define R200_SE_VTX_AOS_ADDR8 0x20f8 +#define R200_SE_VTX_AOS_ADDR9 0x20fc +#define R200_SE_VTX_AOS_ATTR1011 0x2100 +#define R200_SE_VTX_AOS_ADDR10 0x2104 +#define R200_SE_VTX_AOS_ADDR11 0x2108 +#define R200_SE_VF_MAX_VTX_INDX 0x210c +#define R200_SE_VF_MIN_VTX_INDX 0x2110 +/* gap */ +#define R200_SE_VAP_CNTL_STATUS 0x2140 +#define R200_VC_NO_SWAP (0 << 0) +#define R200_VC_16BIT_SWAP (1 << 0) +#define R200_VC_32BIT_SWAP (2 << 0) +/* gap */ +#define R200_SE_VTX_STATE_CNTL 0x2180 +#define R200_VSC_COLOR_0_ASSEMBLY_CNTL_SHIFT 0x00000000 +#define R200_VSC_COLOR_1_ASSEMBLY_CNTL_SHIFT 0x00000002 +#define R200_VSC_COLOR_2_ASSEMBLY_CNTL_SHIFT 0x00000004 +#define R200_VSC_COLOR_3_ASSEMBLY_CNTL_SHIFT 0x00000006 +#define R200_VSC_COLOR_4_ASSEMBLY_CNTL_SHIFT 0x00000008 +#define R200_VSC_COLOR_5_ASSEMBLY_CNTL_SHIFT 0x0000000a +#define R200_VSC_COLOR_6_ASSEMBLY_CNTL_SHIFT 0x0000000c +#define R200_VSC_COLOR_7_ASSEMBLY_CNTL_SHIFT 0x0000000e +#define R200_VSC_UPDATE_USER_COLOR_0_ENABLE 0x00010000 +#define R200_VSC_UPDATE_USER_COLOR_1_ENABLE 0x00020000 +/* gap */ +#define R200_SE_TCL_VECTOR_INDX_REG 0x2200 +#define R200_SE_TCL_VECTOR_DATA_REG 0x2204 +#define R200_SE_TCL_SCALAR_INDX_REG 0x2208 +#define R200_SE_TCL_SCALAR_DATA_REG 0x220c +/* gap */ +#define R200_SE_TCL_MATRIX_SEL_0 0x2230 +#define R200_MODELVIEW_0_SHIFT (0) +#define R200_MODELVIEW_1_SHIFT (8) +#define R200_MODELVIEW_2_SHIFT (16) +#define R200_MODELVIEW_3_SHIFT (24) +#define R200_SE_TCL_MATRIX_SEL_1 0x2234 +#define R200_IT_MODELVIEW_0_SHIFT (0) +#define R200_IT_MODELVIEW_1_SHIFT (8) +#define R200_IT_MODELVIEW_2_SHIFT (16) +#define R200_IT_MODELVIEW_3_SHIFT (24) +#define R200_SE_TCL_MATRIX_SEL_2 0x2238 +#define R200_MODELPROJECT_0_SHIFT (0) +#define R200_MODELPROJECT_1_SHIFT (8) +#define R200_MODELPROJECT_2_SHIFT (16) +#define R200_MODELPROJECT_3_SHIFT (24) +#define R200_SE_TCL_MATRIX_SEL_3 0x223c +#define R200_TEXMAT_0_SHIFT 0 +#define R200_TEXMAT_1_SHIFT 8 +#define R200_TEXMAT_2_SHIFT 16 +#define R200_TEXMAT_3_SHIFT 24 +#define R200_SE_TCL_MATRIX_SEL_4 0x2240 +#define R200_TEXMAT_4_SHIFT 0 +#define R200_TEXMAT_5_SHIFT 8 +/* gap */ +#define R200_SE_TCL_OUTPUT_VTX_COMP_SEL 0x2250 +#define R200_OUTPUT_XYZW (1<<0) +#define R200_OUTPUT_COLOR_0 (1<<8) +#define R200_OUTPUT_COLOR_1 (1<<9) +#define R200_OUTPUT_TEX_0 (1<<16) +#define R200_OUTPUT_TEX_1 (1<<17) +#define R200_OUTPUT_TEX_2 (1<<18) +#define R200_OUTPUT_TEX_3 (1<<19) +#define R200_OUTPUT_TEX_4 (1<<20) +#define R200_OUTPUT_TEX_5 (1<<21) +#define R200_OUTPUT_TEX_MASK (0x3f<<16) +#define R200_OUTPUT_DISCRETE_FOG (1<<24) +#define R200_OUTPUT_PT_SIZE (1<<25) +#define R200_FORCE_INORDER_PROC (1<<31) +#define R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_0 0x2254 +#define R200_VERTEX_POSITION_ADDR__SHIFT 0x00000000 +#define R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_1 0x2258 +#define R200_VTX_COLOR_0_ADDR__SHIFT 0x00000000 +#define R200_VTX_COLOR_1_ADDR__SHIFT 0x00000008 +#define R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_2 0x225c +#define R200_VTX_TEX_0_ADDR__SHIFT 0x00000000 +#define R200_VTX_TEX_1_ADDR__SHIFT 0x00000008 +#define R200_VTX_TEX_2_ADDR__SHIFT 0x00000010 +#define R200_VTX_TEX_3_ADDR__SHIFT 0x00000018 +#define R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_3 0x2260 +#define R200_VTX_TEX_4_ADDR__SHIFT 0x00000000 +#define R200_VTX_TEX_5_ADDR__SHIFT 0x00000008 + +/* gap */ +#define R200_SE_TCL_LIGHT_MODEL_CTL_0 0x2268 +#define R200_LIGHTING_ENABLE (1<<0) +#define R200_LIGHT_IN_MODELSPACE (1<<1) +#define R200_LOCAL_VIEWER (1<<2) +#define R200_NORMALIZE_NORMALS (1<<3) +#define R200_RESCALE_NORMALS (1<<4) +#define R200_SPECULAR_LIGHTS (1<<5) +#define R200_DIFFUSE_SPECULAR_COMBINE (1<<6) +#define R200_LIGHT_ALPHA (1<<7) +#define R200_LOCAL_LIGHT_VEC_GL (1<<8) +#define R200_LIGHT_NO_NORMAL_AMBIENT_ONLY (1<<9) +#define R200_LIGHT_TWOSIDE (1<<10) +#define R200_FRONT_SHININESS_SOURCE_SHIFT (0xb) +#define R200_BACK_SHININESS_SOURCE_SHIFT (0xd) +#define R200_LM0_SOURCE_MATERIAL_0 (0) +#define R200_LM0_SOURCE_MATERIAL_1 (1) +#define R200_LM0_SOURCE_VERTEX_SHININESS_0 (2) +#define R200_LM0_SOURCE_VERTEX_SHININESS_1 (3) +#define R200_SE_TCL_LIGHT_MODEL_CTL_1 0x226c +#define R200_LM1_SOURCE_LIGHT_PREMULT (0) +#define R200_LM1_SOURCE_MATERIAL_0 (1) +#define R200_LM1_SOURCE_VERTEX_COLOR_0 (2) +#define R200_LM1_SOURCE_VERTEX_COLOR_1 (3) +#define R200_LM1_SOURCE_VERTEX_COLOR_2 (4) +#define R200_LM1_SOURCE_VERTEX_COLOR_3 (5) +#define R200_LM1_SOURCE_VERTEX_COLOR_4 (6) +#define R200_LM1_SOURCE_VERTEX_COLOR_5 (7) +#define R200_LM1_SOURCE_VERTEX_COLOR_6 (8) +#define R200_LM1_SOURCE_VERTEX_COLOR_7 (9) +#define R200_LM1_SOURCE_MATERIAL_1 (0xf) +#define R200_FRONT_EMISSIVE_SOURCE_SHIFT (0) +#define R200_FRONT_AMBIENT_SOURCE_SHIFT (4) +#define R200_FRONT_DIFFUSE_SOURCE_SHIFT (8) +#define R200_FRONT_SPECULAR_SOURCE_SHIFT (12) +#define R200_BACK_EMISSIVE_SOURCE_SHIFT (16) +#define R200_BACK_AMBIENT_SOURCE_SHIFT (20) +#define R200_BACK_DIFFUSE_SOURCE_SHIFT (24) +#define R200_BACK_SPECULAR_SOURCE_SHIFT (28) +#define R200_SE_TCL_PER_LIGHT_CTL_0 0x2270 +#define R200_LIGHT_0_ENABLE (1<<0) +#define R200_LIGHT_0_ENABLE_AMBIENT (1<<1) +#define R200_LIGHT_0_ENABLE_SPECULAR (1<<2) +#define R200_LIGHT_0_IS_LOCAL (1<<3) +#define R200_LIGHT_0_IS_SPOT (1<<4) +#define R200_LIGHT_0_DUAL_CONE (1<<5) +#define R200_LIGHT_0_ENABLE_RANGE_ATTEN (1<<6) +#define R200_LIGHT_0_CONSTANT_RANGE_ATTEN (1<<7) +#define R200_LIGHT_1_ENABLE (1<<16) +#define R200_LIGHT_1_ENABLE_AMBIENT (1<<17) +#define R200_LIGHT_1_ENABLE_SPECULAR (1<<18) +#define R200_LIGHT_1_IS_LOCAL (1<<19) +#define R200_LIGHT_1_IS_SPOT (1<<20) +#define R200_LIGHT_1_DUAL_CONE (1<<21) +#define R200_LIGHT_1_ENABLE_RANGE_ATTEN (1<<22) +#define R200_LIGHT_1_CONSTANT_RANGE_ATTEN (1<<23) +#define R200_LIGHT_0_SHIFT (0) +#define R200_LIGHT_1_SHIFT (16) +#define R200_SE_TCL_PER_LIGHT_CTL_1 0x2274 +#define R200_LIGHT_2_SHIFT (0) +#define R200_LIGHT_3_SHIFT (16) +#define R200_SE_TCL_PER_LIGHT_CTL_2 0x2278 +#define R200_LIGHT_4_SHIFT (0) +#define R200_LIGHT_5_SHIFT (16) +#define R200_SE_TCL_PER_LIGHT_CTL_3 0x227c +#define R200_LIGHT_6_SHIFT (0) +#define R200_LIGHT_7_SHIFT (16) +/* gap */ +#define R200_SE_TCL_TEX_PROC_CTL_2 0x22a8 +#define R200_TEXGEN_0_COMP_MASK_SHIFT (0) +#define R200_TEXGEN_1_COMP_MASK_SHIFT (4) +#define R200_TEXGEN_2_COMP_MASK_SHIFT (8) +#define R200_TEXGEN_3_COMP_MASK_SHIFT (12) +#define R200_TEXGEN_4_COMP_MASK_SHIFT (16) +#define R200_TEXGEN_5_COMP_MASK_SHIFT (20) +#define R200_SE_TCL_TEX_PROC_CTL_3 0x22ac +#define R200_TEXGEN_0_INPUT_TEX_SHIFT (0) +#define R200_TEXGEN_1_INPUT_TEX_SHIFT (4) +#define R200_TEXGEN_2_INPUT_TEX_SHIFT (8) +#define R200_TEXGEN_3_INPUT_TEX_SHIFT (12) +#define R200_TEXGEN_4_INPUT_TEX_SHIFT (16) +#define R200_TEXGEN_5_INPUT_TEX_SHIFT (20) +#define R200_SE_TCL_TEX_PROC_CTL_0 0x22b0 +#define R200_TEXGEN_TEXMAT_0_ENABLE (1<<0) +#define R200_TEXGEN_TEXMAT_1_ENABLE (1<<1) +#define R200_TEXGEN_TEXMAT_2_ENABLE (1<<2) +#define R200_TEXGEN_TEXMAT_3_ENABLE (1<<3) +#define R200_TEXGEN_TEXMAT_4_ENABLE (1<<4) +#define R200_TEXGEN_TEXMAT_5_ENABLE (1<<5) +#define R200_TEXMAT_0_ENABLE (1<<8) +#define R200_TEXMAT_1_ENABLE (1<<9) +#define R200_TEXMAT_2_ENABLE (1<<10) +#define R200_TEXMAT_3_ENABLE (1<<11) +#define R200_TEXMAT_4_ENABLE (1<<12) +#define R200_TEXMAT_5_ENABLE (1<<13) +#define R200_TEXGEN_FORCE_W_TO_ONE (1<<16) +#define R200_SE_TCL_TEX_PROC_CTL_1 0x22b4 +#define R200_TEXGEN_INPUT_MASK (0xf) +#define R200_TEXGEN_INPUT_TEXCOORD_0 (0) +#define R200_TEXGEN_INPUT_TEXCOORD_1 (1) +#define R200_TEXGEN_INPUT_TEXCOORD_2 (2) +#define R200_TEXGEN_INPUT_TEXCOORD_3 (3) +#define R200_TEXGEN_INPUT_TEXCOORD_4 (4) +#define R200_TEXGEN_INPUT_TEXCOORD_5 (5) +#define R200_TEXGEN_INPUT_OBJ (8) +#define R200_TEXGEN_INPUT_EYE (9) +#define R200_TEXGEN_INPUT_EYE_NORMAL (0xa) +#define R200_TEXGEN_INPUT_EYE_REFLECT (0xb) +#define R200_TEXGEN_INPUT_SPHERE (0xd) +#define R200_TEXGEN_0_INPUT_SHIFT (0) +#define R200_TEXGEN_1_INPUT_SHIFT (4) +#define R200_TEXGEN_2_INPUT_SHIFT (8) +#define R200_TEXGEN_3_INPUT_SHIFT (12) +#define R200_TEXGEN_4_INPUT_SHIFT (16) +#define R200_TEXGEN_5_INPUT_SHIFT (20) +#define R200_SE_TC_TEX_CYL_WRAP_CTL 0x22b8 +/* gap */ +#define R200_SE_TCL_UCP_VERT_BLEND_CTL 0x22c0 +#define R200_UCP_IN_CLIP_SPACE (1<<0) +#define R200_UCP_IN_MODEL_SPACE (1<<1) +#define R200_UCP_ENABLE_0 (1<<2) +#define R200_UCP_ENABLE_1 (1<<3) +#define R200_UCP_ENABLE_2 (1<<4) +#define R200_UCP_ENABLE_3 (1<<5) +#define R200_UCP_ENABLE_4 (1<<6) +#define R200_UCP_ENABLE_5 (1<<7) +#define R200_TCL_FOG_MASK (3<<8) +#define R200_TCL_FOG_DISABLE (0<<8) +#define R200_TCL_FOG_EXP (1<<8) +#define R200_TCL_FOG_EXP2 (2<<8) +#define R200_TCL_FOG_LINEAR (3<<8) +#define R200_RNG_BASED_FOG (1<<10) +#define R200_CLIP_DISABLE (1<<11) +#define R200_CULL_FRONT_IS_CW (0<<28) +#define R200_CULL_FRONT_IS_CCW (1<<28) +#define R200_CULL_FRONT (1<<29) +#define R200_CULL_BACK (1<<30) +#define R200_SE_TCL_POINT_SPRITE_CNTL 0x22c4 +/* gap */ +#define R200_SE_VTX_ST_POS_0_X_4 0x2300 +#define R200_SE_VTX_ST_POS_0_Y_4 0x2304 +#define R200_SE_VTX_ST_POS_0_Z_4 0x2308 +#define R200_SE_VTX_ST_POS_0_W_4 0x230c +#define R200_SE_VTX_ST_NORM_0_X 0x2310 +#define R200_SE_VTX_ST_NORM_0_Y 0x2314 +#define R200_SE_VTX_ST_NORM_0_Z 0x2318 +#define R200_SE_VTX_ST_PVMS 0x231c +#define R200_SE_VTX_ST_CLR_0_R 0x2320 +#define R200_SE_VTX_ST_CLR_0_G 0x2324 +#define R200_SE_VTX_ST_CLR_0_B 0x2328 +#define R200_SE_VTX_ST_CLR_0_A 0x232c +#define R200_SE_VTX_ST_CLR_1_R 0x2330 +#define R200_SE_VTX_ST_CLR_1_G 0x2334 +#define R200_SE_VTX_ST_CLR_1_B 0x2338 +#define R200_SE_VTX_ST_CLR_1_A 0x233c +#define R200_SE_VTX_ST_CLR_2_R 0x2340 +#define R200_SE_VTX_ST_CLR_2_G 0x2344 +#define R200_SE_VTX_ST_CLR_2_B 0x2348 +#define R200_SE_VTX_ST_CLR_2_A 0x234c +#define R200_SE_VTX_ST_CLR_3_R 0x2350 +#define R200_SE_VTX_ST_CLR_3_G 0x2354 +#define R200_SE_VTX_ST_CLR_3_B 0x2358 +#define R200_SE_VTX_ST_CLR_3_A 0x235c +#define R200_SE_VTX_ST_CLR_4_R 0x2360 +#define R200_SE_VTX_ST_CLR_4_G 0x2364 +#define R200_SE_VTX_ST_CLR_4_B 0x2368 +#define R200_SE_VTX_ST_CLR_4_A 0x236c +#define R200_SE_VTX_ST_CLR_5_R 0x2370 +#define R200_SE_VTX_ST_CLR_5_G 0x2374 +#define R200_SE_VTX_ST_CLR_5_B 0x2378 +#define R200_SE_VTX_ST_CLR_5_A 0x237c +#define R200_SE_VTX_ST_CLR_6_R 0x2380 +#define R200_SE_VTX_ST_CLR_6_G 0x2384 +#define R200_SE_VTX_ST_CLR_6_B 0x2388 +#define R200_SE_VTX_ST_CLR_6_A 0x238c +#define R200_SE_VTX_ST_CLR_7_R 0x2390 +#define R200_SE_VTX_ST_CLR_7_G 0x2394 +#define R200_SE_VTX_ST_CLR_7_B 0x2398 +#define R200_SE_VTX_ST_CLR_7_A 0x239c +#define R200_SE_VTX_ST_TEX_0_S 0x23a0 +#define R200_SE_VTX_ST_TEX_0_T 0x23a4 +#define R200_SE_VTX_ST_TEX_0_R 0x23a8 +#define R200_SE_VTX_ST_TEX_0_Q 0x23ac +#define R200_SE_VTX_ST_TEX_1_S 0x23b0 +#define R200_SE_VTX_ST_TEX_1_T 0x23b4 +#define R200_SE_VTX_ST_TEX_1_R 0x23b8 +#define R200_SE_VTX_ST_TEX_1_Q 0x23bc +#define R200_SE_VTX_ST_TEX_2_S 0x23c0 +#define R200_SE_VTX_ST_TEX_2_T 0x23c4 +#define R200_SE_VTX_ST_TEX_2_R 0x23c8 +#define R200_SE_VTX_ST_TEX_2_Q 0x23cc +#define R200_SE_VTX_ST_TEX_3_S 0x23d0 +#define R200_SE_VTX_ST_TEX_3_T 0x23d4 +#define R200_SE_VTX_ST_TEX_3_R 0x23d8 +#define R200_SE_VTX_ST_TEX_3_Q 0x23dc +#define R200_SE_VTX_ST_TEX_4_S 0x23e0 +#define R200_SE_VTX_ST_TEX_4_T 0x23e4 +#define R200_SE_VTX_ST_TEX_4_R 0x23e8 +#define R200_SE_VTX_ST_TEX_4_Q 0x23ec +#define R200_SE_VTX_ST_TEX_5_S 0x23f0 +#define R200_SE_VTX_ST_TEX_5_T 0x23f4 +#define R200_SE_VTX_ST_TEX_5_R 0x23f8 +#define R200_SE_VTX_ST_TEX_5_Q 0x23fc +#define R200_SE_VTX_ST_PNT_SPRT_SZ 0x2400 +#define R200_SE_VTX_ST_DISC_FOG 0x2404 +#define R200_SE_VTX_ST_SHININESS_0 0x2408 +#define R200_SE_VTX_ST_SHININESS_1 0x240c +#define R200_SE_VTX_ST_BLND_WT_0 0x2410 +#define R200_SE_VTX_ST_BLND_WT_1 0x2414 +#define R200_SE_VTX_ST_BLND_WT_2 0x2418 +#define R200_SE_VTX_ST_BLND_WT_3 0x241c +#define R200_SE_VTX_ST_POS_1_X 0x2420 +#define R200_SE_VTX_ST_POS_1_Y 0x2424 +#define R200_SE_VTX_ST_POS_1_Z 0x2428 +#define R200_SE_VTX_ST_POS_1_W 0x242c +#define R200_SE_VTX_ST_NORM_1_X 0x2430 +#define R200_SE_VTX_ST_NORM_1_Y 0x2434 +#define R200_SE_VTX_ST_NORM_1_Z 0x2438 +#define R200_SE_VTX_ST_USR_CLR_0_R 0x2440 +#define R200_SE_VTX_ST_USR_CLR_0_G 0x2444 +#define R200_SE_VTX_ST_USR_CLR_0_B 0x2448 +#define R200_SE_VTX_ST_USR_CLR_0_A 0x244c +#define R200_SE_VTX_ST_USR_CLR_1_R 0x2450 +#define R200_SE_VTX_ST_USR_CLR_1_G 0x2454 +#define R200_SE_VTX_ST_USR_CLR_1_B 0x2458 +#define R200_SE_VTX_ST_USR_CLR_1_A 0x245c +#define R200_SE_VTX_ST_CLR_0_PKD 0x2460 +#define R200_SE_VTX_ST_CLR_1_PKD 0x2464 +#define R200_SE_VTX_ST_CLR_2_PKD 0x2468 +#define R200_SE_VTX_ST_CLR_3_PKD 0x246c +#define R200_SE_VTX_ST_CLR_4_PKD 0x2470 +#define R200_SE_VTX_ST_CLR_5_PKD 0x2474 +#define R200_SE_VTX_ST_CLR_6_PKD 0x2478 +#define R200_SE_VTX_ST_CLR_7_PKD 0x247c +#define R200_SE_VTX_ST_POS_0_X_2 0x2480 +#define R200_SE_VTX_ST_POS_0_Y_2 0x2484 +#define R200_SE_VTX_ST_PAR_CLR_LD 0x2488 +#define R200_SE_VTX_ST_USR_CLR_PKD 0x248c +#define R200_SE_VTX_ST_POS_0_X_3 0x2490 +#define R200_SE_VTX_ST_POS_0_Y_3 0x2494 +#define R200_SE_VTX_ST_POS_0_Z_3 0x2498 +#define R200_SE_VTX_ST_END_OF_PKT 0x249c +/* gap */ +#define R200_RE_POINTSIZE 0x2648 +#define R200_POINTSIZE_SHIFT 0 +#define R200_MAXPOINTSIZE_SHIFT 16 +/* gap */ +#define R200_RE_TOP_LEFT 0x26c0 +#define R200_RE_LEFT_SHIFT 0 +#define R200_RE_TOP_SHIFT 16 +#define R200_RE_MISC 0x26c4 +#define R200_STIPPLE_COORD_MASK 0x1f +#define R200_STIPPLE_X_OFFSET_SHIFT 0 +#define R200_STIPPLE_X_OFFSET_MASK (0x1f << 0) +#define R200_STIPPLE_Y_OFFSET_SHIFT 8 +#define R200_STIPPLE_Y_OFFSET_MASK (0x1f << 8) +#define R200_STIPPLE_LITTLE_BIT_ORDER (0 << 16) +#define R200_STIPPLE_BIG_BIT_ORDER (1 << 16) +/* gap */ +#define R200_RE_AUX_SCISSOR_CNTL 0x26f0 +#define R200_EXCLUSIVE_SCISSOR_0 0x01000000 +#define R200_EXCLUSIVE_SCISSOR_1 0x02000000 +#define R200_EXCLUSIVE_SCISSOR_2 0x04000000 +#define R200_SCISSOR_ENABLE_0 0x10000000 +#define R200_SCISSOR_ENABLE_1 0x20000000 +#define R200_SCISSOR_ENABLE_2 0x40000000 +/* gap */ +#define R200_PP_TXFILTER_0 0x2c00 +#define R200_MAG_FILTER_NEAREST (0 << 0) +#define R200_MAG_FILTER_LINEAR (1 << 0) +#define R200_MAG_FILTER_MASK (1 << 0) +#define R200_MIN_FILTER_NEAREST (0 << 1) +#define R200_MIN_FILTER_LINEAR (1 << 1) +#define R200_MIN_FILTER_NEAREST_MIP_NEAREST (2 << 1) +#define R200_MIN_FILTER_NEAREST_MIP_LINEAR (3 << 1) +#define R200_MIN_FILTER_LINEAR_MIP_NEAREST (6 << 1) +#define R200_MIN_FILTER_LINEAR_MIP_LINEAR (7 << 1) +#define R200_MIN_FILTER_ANISO_NEAREST (8 << 1) +#define R200_MIN_FILTER_ANISO_LINEAR (9 << 1) +#define R200_MIN_FILTER_ANISO_NEAREST_MIP_NEAREST (10 << 1) +#define R200_MIN_FILTER_ANISO_NEAREST_MIP_LINEAR (11 << 1) +#define R200_MIN_FILTER_MASK (15 << 1) +#define R200_MAX_ANISO_1_TO_1 (0 << 5) +#define R200_MAX_ANISO_2_TO_1 (1 << 5) +#define R200_MAX_ANISO_4_TO_1 (2 << 5) +#define R200_MAX_ANISO_8_TO_1 (3 << 5) +#define R200_MAX_ANISO_16_TO_1 (4 << 5) +#define R200_MAX_ANISO_MASK (7 << 5) +#define R200_MAX_MIP_LEVEL_MASK (0x0f << 16) +#define R200_MAX_MIP_LEVEL_SHIFT 16 +#define R200_YUV_TO_RGB (1 << 20) +#define R200_YUV_TEMPERATURE_COOL (0 << 21) +#define R200_YUV_TEMPERATURE_HOT (1 << 21) +#define R200_YUV_TEMPERATURE_MASK (1 << 21) +#define R200_WRAPEN_S (1 << 22) +#define R200_CLAMP_S_WRAP (0 << 23) +#define R200_CLAMP_S_MIRROR (1 << 23) +#define R200_CLAMP_S_CLAMP_LAST (2 << 23) +#define R200_CLAMP_S_MIRROR_CLAMP_LAST (3 << 23) +#define R200_CLAMP_S_CLAMP_BORDER (4 << 23) +#define R200_CLAMP_S_MIRROR_CLAMP_BORDER (5 << 23) +#define R200_CLAMP_S_CLAMP_GL (6 << 23) +#define R200_CLAMP_S_MIRROR_CLAMP_GL (7 << 23) +#define R200_CLAMP_S_MASK (7 << 23) +#define R200_WRAPEN_T (1 << 26) +#define R200_CLAMP_T_WRAP (0 << 27) +#define R200_CLAMP_T_MIRROR (1 << 27) +#define R200_CLAMP_T_CLAMP_LAST (2 << 27) +#define R200_CLAMP_T_MIRROR_CLAMP_LAST (3 << 27) +#define R200_CLAMP_T_CLAMP_BORDER (4 << 27) +#define R200_CLAMP_T_MIRROR_CLAMP_BORDER (5 << 27) +#define R200_CLAMP_T_CLAMP_GL (6 << 27) +#define R200_CLAMP_T_MIRROR_CLAMP_GL (7 << 27) +#define R200_CLAMP_T_MASK (7 << 27) +#define R200_KILL_LT_ZERO (1 << 30) +#define R200_BORDER_MODE_OGL (0 << 31) +#define R200_BORDER_MODE_D3D (1 << 31) +#define R200_PP_TXFORMAT_0 0x2c04 +#define R200_TXFORMAT_I8 (0 << 0) +#define R200_TXFORMAT_AI88 (1 << 0) +#define R200_TXFORMAT_RGB332 (2 << 0) +#define R200_TXFORMAT_ARGB1555 (3 << 0) +#define R200_TXFORMAT_RGB565 (4 << 0) +#define R200_TXFORMAT_ARGB4444 (5 << 0) +#define R200_TXFORMAT_ARGB8888 (6 << 0) +#define R200_TXFORMAT_RGBA8888 (7 << 0) +#define R200_TXFORMAT_Y8 (8 << 0) +#define R200_TXFORMAT_AVYU4444 (9 << 0) +#define R200_TXFORMAT_VYUY422 (10 << 0) +#define R200_TXFORMAT_YVYU422 (11 << 0) +#define R200_TXFORMAT_DXT1 (12 << 0) +#define R200_TXFORMAT_DXT23 (14 << 0) +#define R200_TXFORMAT_DXT45 (15 << 0) +#define R200_TXFORMAT_FORMAT_MASK (31 << 0) +#define R200_TXFORMAT_FORMAT_SHIFT 0 +#define R200_TXFORMAT_ALPHA_IN_MAP (1 << 6) +#define R200_TXFORMAT_NON_POWER2 (1 << 7) +#define R200_TXFORMAT_WIDTH_MASK (15 << 8) +#define R200_TXFORMAT_WIDTH_SHIFT 8 +#define R200_TXFORMAT_HEIGHT_MASK (15 << 12) +#define R200_TXFORMAT_HEIGHT_SHIFT 12 +#define R200_TXFORMAT_F5_WIDTH_MASK (15 << 16) /* cube face 5 */ +#define R200_TXFORMAT_F5_WIDTH_SHIFT 16 +#define R200_TXFORMAT_F5_HEIGHT_MASK (15 << 20) +#define R200_TXFORMAT_F5_HEIGHT_SHIFT 20 +#define R200_TXFORMAT_ST_ROUTE_STQ0 (0 << 24) +#define R200_TXFORMAT_ST_ROUTE_STQ1 (1 << 24) +#define R200_TXFORMAT_ST_ROUTE_STQ2 (2 << 24) +#define R200_TXFORMAT_ST_ROUTE_STQ3 (3 << 24) +#define R200_TXFORMAT_ST_ROUTE_STQ4 (4 << 24) +#define R200_TXFORMAT_ST_ROUTE_STQ5 (5 << 24) +#define R200_TXFORMAT_ST_ROUTE_MASK (7 << 24) +#define R200_TXFORMAT_ST_ROUTE_SHIFT 24 +#define R200_TXFORMAT_ALPHA_MASK_ENABLE (1 << 28) +#define R200_TXFORMAT_CHROMA_KEY_ENABLE (1 << 29) +#define R200_TXFORMAT_CUBIC_MAP_ENABLE (1 << 30) +#define R200_PP_TXFORMAT_X_0 0x2c08 +#define R200_DEPTH_LOG2_MASK (0xf << 0) +#define R200_DEPTH_LOG2_SHIFT 0 +#define R200_VOLUME_FILTER_SHIFT 4 +#define R200_VOLUME_FILTER_MASK (1 << 4) +#define R200_VOLUME_FILTER_NEAREST (0 << 4) +#define R200_VOLUME_FILTER_LINEAR (1 << 4) +#define R200_WRAPEN_Q (1 << 8) +#define R200_CLAMP_Q_WRAP (0 << 9) +#define R200_CLAMP_Q_MIRROR (1 << 9) +#define R200_CLAMP_Q_CLAMP_LAST (2 << 9) +#define R200_CLAMP_Q_MIRROR_CLAMP_LAST (3 << 9) +#define R200_CLAMP_Q_CLAMP_BORDER (4 << 9) +#define R200_CLAMP_Q_MIRROR_CLAMP_BORDER (5 << 9) +#define R200_CLAMP_Q_CLAMP_GL (6 << 9) +#define R200_CLAMP_Q_MIRROR_CLAMP_GL (7 << 9) +#define R200_CLAMP_Q_MASK (7 << 9) +#define R200_MIN_MIP_LEVEL_MASK (0xff << 12) +#define R200_MIN_MIP_LEVEL_SHIFT 12 +#define R200_TEXCOORD_NONPROJ (0 << 16) +#define R200_TEXCOORD_CUBIC_ENV (1 << 16) +#define R200_TEXCOORD_VOLUME (2 << 16) +#define R200_TEXCOORD_PROJ (3 << 16) +#define R200_TEXCOORD_DEPTH (4 << 16) +#define R200_TEXCOORD_1D_PROJ (5 << 16) +#define R200_TEXCOORD_1D (6 << 16) +#define R200_TEXCOORD_ZERO (7 << 16) +#define R200_TEXCOORD_MASK (7 << 16) +#define R200_LOD_BIAS_MASK (0xfff80000) +#define R200_LOD_BIAS_SHIFT 19 +#define R200_PP_TXSIZE_0 0x2c0c /* NPOT only */ +#define R200_PP_TXPITCH_0 0x2c10 /* NPOT only */ +#define R200_PP_BORDER_COLOR_0 0x2c14 +#define R200_PP_CUBIC_FACES_0 0x2c18 +#define R200_FACE_WIDTH_1_SHIFT 0 +#define R200_FACE_HEIGHT_1_SHIFT 4 +#define R200_FACE_WIDTH_1_MASK (0xf << 0) +#define R200_FACE_HEIGHT_1_MASK (0xf << 4) +#define R200_FACE_WIDTH_2_SHIFT 8 +#define R200_FACE_HEIGHT_2_SHIFT 12 +#define R200_FACE_WIDTH_2_MASK (0xf << 8) +#define R200_FACE_HEIGHT_2_MASK (0xf << 12) +#define R200_FACE_WIDTH_3_SHIFT 16 +#define R200_FACE_HEIGHT_3_SHIFT 20 +#define R200_FACE_WIDTH_3_MASK (0xf << 16) +#define R200_FACE_HEIGHT_3_MASK (0xf << 20) +#define R200_FACE_WIDTH_4_SHIFT 24 +#define R200_FACE_HEIGHT_4_SHIFT 28 +#define R200_FACE_WIDTH_4_MASK (0xf << 24) +#define R200_FACE_HEIGHT_4_MASK (0xf << 28) +#define R200_PP_TXFILTER_1 0x2c20 +#define R200_PP_TXFORMAT_1 0x2c24 +#define R200_PP_TXFORMAT_X_1 0x2c28 +#define R200_PP_TXSIZE_1 0x2c2c +#define R200_PP_TXPITCH_1 0x2c30 +#define R200_PP_BORDER_COLOR_1 0x2c34 +#define R200_PP_CUBIC_FACES_1 0x2c38 +#define R200_PP_TXFILTER_2 0x2c40 +#define R200_PP_TXFORMAT_2 0x2c44 +#define R200_PP_TXSIZE_2 0x2c4c +#define R200_PP_TXFORMAT_X_2 0x2c48 +#define R200_PP_TXPITCH_2 0x2c50 +#define R200_PP_BORDER_COLOR_2 0x2c54 +#define R200_PP_CUBIC_FACES_2 0x2c58 +#define R200_PP_TXFILTER_3 0x2c60 +#define R200_PP_TXFORMAT_3 0x2c64 +#define R200_PP_TXSIZE_3 0x2c6c +#define R200_PP_TXFORMAT_X_3 0x2c68 +#define R200_PP_TXPITCH_3 0x2c70 +#define R200_PP_BORDER_COLOR_3 0x2c74 +#define R200_PP_CUBIC_FACES_3 0x2c78 +#define R200_PP_TXFILTER_4 0x2c80 +#define R200_PP_TXFORMAT_4 0x2c84 +#define R200_PP_TXSIZE_4 0x2c8c +#define R200_PP_TXFORMAT_X_4 0x2c88 +#define R200_PP_TXPITCH_4 0x2c90 +#define R200_PP_BORDER_COLOR_4 0x2c94 +#define R200_PP_CUBIC_FACES_4 0x2c98 +#define R200_PP_TXFILTER_5 0x2ca0 +#define R200_PP_TXFORMAT_5 0x2ca4 +#define R200_PP_TXSIZE_5 0x2cac +#define R200_PP_TXFORMAT_X_5 0x2ca8 +#define R200_PP_TXPITCH_5 0x2cb0 +#define R200_PP_BORDER_COLOR_5 0x2cb4 +#define R200_PP_CUBIC_FACES_5 0x2cb8 +/* gap */ +#define R200_PP_CNTL_X 0x2cc4 +/* gap */ +#define R200_PP_TXOFFSET_0 0x2d00 +#define R200_TXO_ENDIAN_NO_SWAP (0 << 0) +#define R200_TXO_ENDIAN_BYTE_SWAP (1 << 0) +#define R200_TXO_ENDIAN_WORD_SWAP (2 << 0) +#define R200_TXO_ENDIAN_HALFDW_SWAP (3 << 0) +#define R200_TXO_OFFSET_MASK 0xffffffe0 +#define R200_TXO_OFFSET_SHIFT 5 +#define R200_PP_CUBIC_OFFSET_F1_0 0x2d04 +#define R200_PP_CUBIC_OFFSET_F2_0 0x2d08 +#define R200_PP_CUBIC_OFFSET_F3_0 0x2d0c +#define R200_PP_CUBIC_OFFSET_F4_0 0x2d10 +#define R200_PP_CUBIC_OFFSET_F5_0 0x2d14 +#define R200_PP_TXOFFSET_1 0x2d18 +#define R200_PP_CUBIC_OFFSET_F1_1 0x2d1c +#define R200_PP_CUBIC_OFFSET_F2_1 0x2d20 +#define R200_PP_CUBIC_OFFSET_F3_1 0x2d24 +#define R200_PP_CUBIC_OFFSET_F4_1 0x2d28 +#define R200_PP_CUBIC_OFFSET_F5_1 0x2d2c +#define R200_PP_TXOFFSET_2 0x2d30 +#define R200_PP_CUBIC_OFFSET_F1_2 0x2d34 +#define R200_PP_CUBIC_OFFSET_F2_2 0x2d38 +#define R200_PP_CUBIC_OFFSET_F3_2 0x2d3c +#define R200_PP_CUBIC_OFFSET_F4_2 0x2d40 +#define R200_PP_CUBIC_OFFSET_F5_2 0x2d44 +#define R200_PP_TXOFFSET_3 0x2d48 +#define R200_PP_CUBIC_OFFSET_F1_3 0x2d4c +#define R200_PP_CUBIC_OFFSET_F2_3 0x2d50 +#define R200_PP_CUBIC_OFFSET_F3_3 0x2d54 +#define R200_PP_CUBIC_OFFSET_F4_3 0x2d58 +#define R200_PP_CUBIC_OFFSET_F5_3 0x2d5c +#define R200_PP_TXOFFSET_4 0x2d60 +#define R200_PP_CUBIC_OFFSET_F1_4 0x2d64 +#define R200_PP_CUBIC_OFFSET_F2_4 0x2d68 +#define R200_PP_CUBIC_OFFSET_F3_4 0x2d6c +#define R200_PP_CUBIC_OFFSET_F4_4 0x2d70 +#define R200_PP_CUBIC_OFFSET_F5_4 0x2d74 +#define R200_PP_TXOFFSET_5 0x2d78 +#define R200_PP_CUBIC_OFFSET_F1_5 0x2d7c +#define R200_PP_CUBIC_OFFSET_F2_5 0x2d80 +#define R200_PP_CUBIC_OFFSET_F3_5 0x2d84 +#define R200_PP_CUBIC_OFFSET_F4_5 0x2d88 +#define R200_PP_CUBIC_OFFSET_F5_5 0x2d8c +/* gap */ +#define R200_PP_TAM_DEBUG3 0x2d9c +/* gap */ +#define R200_PP_TFACTOR_0 0x2ee0 +#define R200_PP_TFACTOR_1 0x2ee4 +#define R200_PP_TFACTOR_2 0x2ee8 +#define R200_PP_TFACTOR_3 0x2eec +#define R200_PP_TFACTOR_4 0x2ef0 +#define R200_PP_TFACTOR_5 0x2ef4 +/* gap */ +#define R200_PP_TXCBLEND_0 0x2f00 +#define R200_TXC_ARG_A_ZERO (0) +#define R200_TXC_ARG_A_CURRENT_COLOR (2) +#define R200_TXC_ARG_A_CURRENT_ALPHA (3) +#define R200_TXC_ARG_A_DIFFUSE_COLOR (4) +#define R200_TXC_ARG_A_DIFFUSE_ALPHA (5) +#define R200_TXC_ARG_A_SPECULAR_COLOR (6) +#define R200_TXC_ARG_A_SPECULAR_ALPHA (7) +#define R200_TXC_ARG_A_TFACTOR_COLOR (8) +#define R200_TXC_ARG_A_TFACTOR_ALPHA (9) +#define R200_TXC_ARG_A_R0_COLOR (10) +#define R200_TXC_ARG_A_R0_ALPHA (11) +#define R200_TXC_ARG_A_R1_COLOR (12) +#define R200_TXC_ARG_A_R1_ALPHA (13) +#define R200_TXC_ARG_A_R2_COLOR (14) +#define R200_TXC_ARG_A_R2_ALPHA (15) +#define R200_TXC_ARG_A_R3_COLOR (16) +#define R200_TXC_ARG_A_R3_ALPHA (17) +#define R200_TXC_ARG_A_R4_COLOR (18) +#define R200_TXC_ARG_A_R4_ALPHA (19) +#define R200_TXC_ARG_A_R5_COLOR (20) +#define R200_TXC_ARG_A_R5_ALPHA (21) +#define R200_TXC_ARG_A_TFACTOR1_COLOR (26) +#define R200_TXC_ARG_A_TFACTOR1_ALPHA (27) +#define R200_TXC_ARG_A_MASK (31 << 0) +#define R200_TXC_ARG_A_SHIFT 0 +#define R200_TXC_ARG_B_ZERO (0<<5) +#define R200_TXC_ARG_B_CURRENT_COLOR (2<<5) +#define R200_TXC_ARG_B_CURRENT_ALPHA (3<<5) +#define R200_TXC_ARG_B_DIFFUSE_COLOR (4<<5) +#define R200_TXC_ARG_B_DIFFUSE_ALPHA (5<<5) +#define R200_TXC_ARG_B_SPECULAR_COLOR (6<<5) +#define R200_TXC_ARG_B_SPECULAR_ALPHA (7<<5) +#define R200_TXC_ARG_B_TFACTOR_COLOR (8<<5) +#define R200_TXC_ARG_B_TFACTOR_ALPHA (9<<5) +#define R200_TXC_ARG_B_R0_COLOR (10<<5) +#define R200_TXC_ARG_B_R0_ALPHA (11<<5) +#define R200_TXC_ARG_B_R1_COLOR (12<<5) +#define R200_TXC_ARG_B_R1_ALPHA (13<<5) +#define R200_TXC_ARG_B_R2_COLOR (14<<5) +#define R200_TXC_ARG_B_R2_ALPHA (15<<5) +#define R200_TXC_ARG_B_R3_COLOR (16<<5) +#define R200_TXC_ARG_B_R3_ALPHA (17<<5) +#define R200_TXC_ARG_B_R4_COLOR (18<<5) +#define R200_TXC_ARG_B_R4_ALPHA (19<<5) +#define R200_TXC_ARG_B_R5_COLOR (20<<5) +#define R200_TXC_ARG_B_R5_ALPHA (21<<5) +#define R200_TXC_ARG_B_TFACTOR1_COLOR (26<<5) +#define R200_TXC_ARG_B_TFACTOR1_ALPHA (27<<5) +#define R200_TXC_ARG_B_MASK (31 << 5) +#define R200_TXC_ARG_B_SHIFT 5 +#define R200_TXC_ARG_C_ZERO (0<<10) +#define R200_TXC_ARG_C_CURRENT_COLOR (2<<10) +#define R200_TXC_ARG_C_CURRENT_ALPHA (3<<10) +#define R200_TXC_ARG_C_DIFFUSE_COLOR (4<<10) +#define R200_TXC_ARG_C_DIFFUSE_ALPHA (5<<10) +#define R200_TXC_ARG_C_SPECULAR_COLOR (6<<10) +#define R200_TXC_ARG_C_SPECULAR_ALPHA (7<<10) +#define R200_TXC_ARG_C_TFACTOR_COLOR (8<<10) +#define R200_TXC_ARG_C_TFACTOR_ALPHA (9<<10) +#define R200_TXC_ARG_C_R0_COLOR (10<<10) +#define R200_TXC_ARG_C_R0_ALPHA (11<<10) +#define R200_TXC_ARG_C_R1_COLOR (12<<10) +#define R200_TXC_ARG_C_R1_ALPHA (13<<10) +#define R200_TXC_ARG_C_R2_COLOR (14<<10) +#define R200_TXC_ARG_C_R2_ALPHA (15<<10) +#define R200_TXC_ARG_C_R3_COLOR (16<<10) +#define R200_TXC_ARG_C_R3_ALPHA (17<<10) +#define R200_TXC_ARG_C_R4_COLOR (18<<10) +#define R200_TXC_ARG_C_R4_ALPHA (19<<10) +#define R200_TXC_ARG_C_R5_COLOR (20<<10) +#define R200_TXC_ARG_C_R5_ALPHA (21<<10) +#define R200_TXC_ARG_C_TFACTOR1_COLOR (26<<10) +#define R200_TXC_ARG_C_TFACTOR1_ALPHA (27<<10) +#define R200_TXC_ARG_C_MASK (31 << 10) +#define R200_TXC_ARG_C_SHIFT 10 +#define R200_TXC_COMP_ARG_A (1 << 16) +#define R200_TXC_COMP_ARG_A_SHIFT (16) +#define R200_TXC_BIAS_ARG_A (1 << 17) +#define R200_TXC_SCALE_ARG_A (1 << 18) +#define R200_TXC_NEG_ARG_A (1 << 19) +#define R200_TXC_COMP_ARG_B (1 << 20) +#define R200_TXC_COMP_ARG_B_SHIFT (20) +#define R200_TXC_BIAS_ARG_B (1 << 21) +#define R200_TXC_SCALE_ARG_B (1 << 22) +#define R200_TXC_NEG_ARG_B (1 << 23) +#define R200_TXC_COMP_ARG_C (1 << 24) +#define R200_TXC_COMP_ARG_C_SHIFT (24) +#define R200_TXC_BIAS_ARG_C (1 << 25) +#define R200_TXC_SCALE_ARG_C (1 << 26) +#define R200_TXC_NEG_ARG_C (1 << 27) +#define R200_TXC_OP_MADD (0 << 28) +#define R200_TXC_OP_CND0 (2 << 28) +#define R200_TXC_OP_LERP (3 << 28) +#define R200_TXC_OP_DOT3 (4 << 28) +#define R200_TXC_OP_DOT4 (5 << 28) +#define R200_TXC_OP_CONDITIONAL (6 << 28) +#define R200_TXC_OP_DOT2_ADD (7 << 28) +#define R200_TXC_OP_MASK (7 << 28) +#define R200_PP_TXCBLEND2_0 0x2f04 +#define R200_TXC_TFACTOR_SEL_SHIFT 0 +#define R200_TXC_TFACTOR_SEL_MASK 0x7 +#define R200_TXC_TFACTOR1_SEL_SHIFT 4 +#define R200_TXC_TFACTOR1_SEL_MASK (0x7 << 4) +#define R200_TXC_SCALE_SHIFT 8 +#define R200_TXC_SCALE_MASK (7 << 8) +#define R200_TXC_SCALE_1X (0 << 8) +#define R200_TXC_SCALE_2X (1 << 8) +#define R200_TXC_SCALE_4X (2 << 8) +#define R200_TXC_SCALE_8X (3 << 8) +#define R200_TXC_SCALE_INV2 (5 << 8) +#define R200_TXC_SCALE_INV4 (6 << 8) +#define R200_TXC_SCALE_INV8 (7 << 8) +#define R200_TXC_CLAMP_SHIFT 12 +#define R200_TXC_CLAMP_MASK (3 << 12) +#define R200_TXC_CLAMP_WRAP (0 << 12) +#define R200_TXC_CLAMP_0_1 (1 << 12) +#define R200_TXC_CLAMP_8_8 (2 << 12) +#define R200_TXC_OUTPUT_REG_MASK (7 << 16) +#define R200_TXC_OUTPUT_REG_NONE (0 << 16) +#define R200_TXC_OUTPUT_REG_R0 (1 << 16) +#define R200_TXC_OUTPUT_REG_R1 (2 << 16) +#define R200_TXC_OUTPUT_REG_R2 (3 << 16) +#define R200_TXC_OUTPUT_REG_R3 (4 << 16) +#define R200_TXC_OUTPUT_REG_R4 (5 << 16) +#define R200_TXC_OUTPUT_REG_R5 (6 << 16) +#define R200_TXC_OUTPUT_MASK_MASK (7 << 20) +#define R200_TXC_OUTPUT_MASK_RGB (0 << 20) +#define R200_TXC_OUTPUT_MASK_RG (1 << 20) +#define R200_TXC_OUTPUT_MASK_RB (2 << 20) +#define R200_TXC_OUTPUT_MASK_R (3 << 20) +#define R200_TXC_OUTPUT_MASK_GB (4 << 20) +#define R200_TXC_OUTPUT_MASK_G (5 << 20) +#define R200_TXC_OUTPUT_MASK_B (6 << 20) +#define R200_TXC_OUTPUT_MASK_NONE (7 << 20) +#define R200_TXC_REPL_NORMAL 0 +#define R200_TXC_REPL_RED 1 +#define R200_TXC_REPL_GREEN 2 +#define R200_TXC_REPL_BLUE 3 +#define R200_TXC_REPL_ARG_A_SHIFT 26 +#define R200_TXC_REPL_ARG_A_MASK (3 << 26) +#define R200_TXC_REPL_ARG_B_SHIFT 28 +#define R200_TXC_REPL_ARG_B_MASK (3 << 28) +#define R200_TXC_REPL_ARG_C_SHIFT 30 +#define R200_TXC_REPL_ARG_C_MASK (3 << 30) +#define R200_PP_TXABLEND_0 0x2f08 +#define R200_TXA_ARG_A_ZERO (0) +#define R200_TXA_ARG_A_CURRENT_ALPHA (2) /* guess */ +#define R200_TXA_ARG_A_CURRENT_BLUE (3) /* guess */ +#define R200_TXA_ARG_A_DIFFUSE_ALPHA (4) +#define R200_TXA_ARG_A_DIFFUSE_BLUE (5) +#define R200_TXA_ARG_A_SPECULAR_ALPHA (6) +#define R200_TXA_ARG_A_SPECULAR_BLUE (7) +#define R200_TXA_ARG_A_TFACTOR_ALPHA (8) +#define R200_TXA_ARG_A_TFACTOR_BLUE (9) +#define R200_TXA_ARG_A_R0_ALPHA (10) +#define R200_TXA_ARG_A_R0_BLUE (11) +#define R200_TXA_ARG_A_R1_ALPHA (12) +#define R200_TXA_ARG_A_R1_BLUE (13) +#define R200_TXA_ARG_A_R2_ALPHA (14) +#define R200_TXA_ARG_A_R2_BLUE (15) +#define R200_TXA_ARG_A_R3_ALPHA (16) +#define R200_TXA_ARG_A_R3_BLUE (17) +#define R200_TXA_ARG_A_R4_ALPHA (18) +#define R200_TXA_ARG_A_R4_BLUE (19) +#define R200_TXA_ARG_A_R5_ALPHA (20) +#define R200_TXA_ARG_A_R5_BLUE (21) +#define R200_TXA_ARG_A_TFACTOR1_ALPHA (26) +#define R200_TXA_ARG_A_TFACTOR1_BLUE (27) +#define R200_TXA_ARG_A_MASK (31 << 0) +#define R200_TXA_ARG_A_SHIFT 0 +#define R200_TXA_ARG_B_ZERO (0<<5) +#define R200_TXA_ARG_B_CURRENT_ALPHA (2<<5) /* guess */ +#define R200_TXA_ARG_B_CURRENT_BLUE (3<<5) /* guess */ +#define R200_TXA_ARG_B_DIFFUSE_ALPHA (4<<5) +#define R200_TXA_ARG_B_DIFFUSE_BLUE (5<<5) +#define R200_TXA_ARG_B_SPECULAR_ALPHA (6<<5) +#define R200_TXA_ARG_B_SPECULAR_BLUE (7<<5) +#define R200_TXA_ARG_B_TFACTOR_ALPHA (8<<5) +#define R200_TXA_ARG_B_TFACTOR_BLUE (9<<5) +#define R200_TXA_ARG_B_R0_ALPHA (10<<5) +#define R200_TXA_ARG_B_R0_BLUE (11<<5) +#define R200_TXA_ARG_B_R1_ALPHA (12<<5) +#define R200_TXA_ARG_B_R1_BLUE (13<<5) +#define R200_TXA_ARG_B_R2_ALPHA (14<<5) +#define R200_TXA_ARG_B_R2_BLUE (15<<5) +#define R200_TXA_ARG_B_R3_ALPHA (16<<5) +#define R200_TXA_ARG_B_R3_BLUE (17<<5) +#define R200_TXA_ARG_B_R4_ALPHA (18<<5) +#define R200_TXA_ARG_B_R4_BLUE (19<<5) +#define R200_TXA_ARG_B_R5_ALPHA (20<<5) +#define R200_TXA_ARG_B_R5_BLUE (21<<5) +#define R200_TXA_ARG_B_TFACTOR1_ALPHA (26<<5) +#define R200_TXA_ARG_B_TFACTOR1_BLUE (27<<5) +#define R200_TXA_ARG_B_MASK (31 << 5) +#define R200_TXA_ARG_B_SHIFT 5 +#define R200_TXA_ARG_C_ZERO (0<<10) +#define R200_TXA_ARG_C_CURRENT_ALPHA (2<<10) /* guess */ +#define R200_TXA_ARG_C_CURRENT_BLUE (3<<10) /* guess */ +#define R200_TXA_ARG_C_DIFFUSE_ALPHA (4<<10) +#define R200_TXA_ARG_C_DIFFUSE_BLUE (5<<10) +#define R200_TXA_ARG_C_SPECULAR_ALPHA (6<<10) +#define R200_TXA_ARG_C_SPECULAR_BLUE (7<<10) +#define R200_TXA_ARG_C_TFACTOR_ALPHA (8<<10) +#define R200_TXA_ARG_C_TFACTOR_BLUE (9<<10) +#define R200_TXA_ARG_C_R0_ALPHA (10<<10) +#define R200_TXA_ARG_C_R0_BLUE (11<<10) +#define R200_TXA_ARG_C_R1_ALPHA (12<<10) +#define R200_TXA_ARG_C_R1_BLUE (13<<10) +#define R200_TXA_ARG_C_R2_ALPHA (14<<10) +#define R200_TXA_ARG_C_R2_BLUE (15<<10) +#define R200_TXA_ARG_C_R3_ALPHA (16<<10) +#define R200_TXA_ARG_C_R3_BLUE (17<<10) +#define R200_TXA_ARG_C_R4_ALPHA (18<<10) +#define R200_TXA_ARG_C_R4_BLUE (19<<10) +#define R200_TXA_ARG_C_R5_ALPHA (20<<10) +#define R200_TXA_ARG_C_R5_BLUE (21<<10) +#define R200_TXA_ARG_C_TFACTOR1_ALPHA (26<<10) +#define R200_TXA_ARG_C_TFACTOR1_BLUE (27<<10) +#define R200_TXA_ARG_C_MASK (31 << 10) +#define R200_TXA_ARG_C_SHIFT 10 +#define R200_TXA_COMP_ARG_A (1 << 16) +#define R200_TXA_COMP_ARG_A_SHIFT (16) +#define R200_TXA_BIAS_ARG_A (1 << 17) +#define R200_TXA_SCALE_ARG_A (1 << 18) +#define R200_TXA_NEG_ARG_A (1 << 19) +#define R200_TXA_COMP_ARG_B (1 << 20) +#define R200_TXA_COMP_ARG_B_SHIFT (20) +#define R200_TXA_BIAS_ARG_B (1 << 21) +#define R200_TXA_SCALE_ARG_B (1 << 22) +#define R200_TXA_NEG_ARG_B (1 << 23) +#define R200_TXA_COMP_ARG_C (1 << 24) +#define R200_TXA_COMP_ARG_C_SHIFT (24) +#define R200_TXA_BIAS_ARG_C (1 << 25) +#define R200_TXA_SCALE_ARG_C (1 << 26) +#define R200_TXA_NEG_ARG_C (1 << 27) +#define R200_TXA_OP_MADD (0 << 28) +#define R200_TXA_OP_CND0 (2 << 28) +#define R200_TXA_OP_LERP (3 << 28) +#define R200_TXA_OP_CONDITIONAL (6 << 28) +#define R200_TXA_OP_MASK (7 << 28) +#define R200_PP_TXABLEND2_0 0x2f0c +#define R200_TXA_TFACTOR_SEL_SHIFT 0 +#define R200_TXA_TFACTOR_SEL_MASK 0x7 +#define R200_TXA_TFACTOR1_SEL_SHIFT 4 +#define R200_TXA_TFACTOR1_SEL_MASK (0x7 << 4) +#define R200_TXA_SCALE_SHIFT 8 +#define R200_TXA_SCALE_MASK (7 << 8) +#define R200_TXA_SCALE_1X (0 << 8) +#define R200_TXA_SCALE_2X (1 << 8) +#define R200_TXA_SCALE_4X (2 << 8) +#define R200_TXA_SCALE_8X (3 << 8) +#define R200_TXA_SCALE_INV2 (5 << 8) +#define R200_TXA_SCALE_INV4 (6 << 8) +#define R200_TXA_SCALE_INV8 (7 << 8) +#define R200_TXA_CLAMP_SHIFT 12 +#define R200_TXA_CLAMP_MASK (3 << 12) +#define R200_TXA_CLAMP_WRAP (0 << 12) +#define R200_TXA_CLAMP_0_1 (1 << 12) +#define R200_TXA_CLAMP_8_8 (2 << 12) +#define R200_TXA_OUTPUT_REG_MASK (7 << 16) +#define R200_TXA_OUTPUT_REG_NONE (0 << 16) +#define R200_TXA_OUTPUT_REG_R0 (1 << 16) +#define R200_TXA_OUTPUT_REG_R1 (2 << 16) +#define R200_TXA_OUTPUT_REG_R2 (3 << 16) +#define R200_TXA_OUTPUT_REG_R3 (4 << 16) +#define R200_TXA_OUTPUT_REG_R4 (5 << 16) +#define R200_TXA_OUTPUT_REG_R5 (6 << 16) +#define R200_TXA_DOT_ALPHA (1 << 20) +#define R200_TXA_REPL_NORMAL 0 +#define R200_TXA_REPL_RED 1 +#define R200_TXA_REPL_GREEN 2 +#define R200_TXA_REPL_ARG_A_SHIFT 26 +#define R200_TXA_REPL_ARG_A_MASK (3 << 26) +#define R200_TXA_REPL_ARG_B_SHIFT 28 +#define R200_TXA_REPL_ARG_B_MASK (3 << 28) +#define R200_TXA_REPL_ARG_C_SHIFT 30 +#define R200_TXA_REPL_ARG_C_MASK (3 << 30) +#define R200_PP_TXCBLEND_1 0x2f10 +#define R200_PP_TXCBLEND2_1 0x2f14 +#define R200_PP_TXABLEND_1 0x2f18 +#define R200_PP_TXABLEND2_1 0x2f1c +#define R200_PP_TXCBLEND_2 0x2f20 +#define R200_PP_TXCBLEND2_2 0x2f24 +#define R200_PP_TXABLEND_2 0x2f28 +#define R200_PP_TXABLEND2_2 0x2f2c +#define R200_PP_TXCBLEND_3 0x2f30 +#define R200_PP_TXCBLEND2_3 0x2f34 +#define R200_PP_TXABLEND_3 0x2f38 +#define R200_PP_TXABLEND2_3 0x2f3c +#define R200_PP_TXCBLEND_4 0x2f40 +#define R200_PP_TXCBLEND2_4 0x2f44 +#define R200_PP_TXABLEND_4 0x2f48 +#define R200_PP_TXABLEND2_4 0x2f4c +#define R200_PP_TXCBLEND_5 0x2f50 +#define R200_PP_TXCBLEND2_5 0x2f54 +#define R200_PP_TXABLEND_5 0x2f58 +#define R200_PP_TXABLEND2_5 0x2f5c +#define R200_PP_TXCBLEND_6 0x2f60 +#define R200_PP_TXCBLEND2_6 0x2f64 +#define R200_PP_TXABLEND_6 0x2f68 +#define R200_PP_TXABLEND2_6 0x2f6c +#define R200_PP_TXCBLEND_7 0x2f70 +#define R200_PP_TXCBLEND2_7 0x2f74 +#define R200_PP_TXABLEND_7 0x2f78 +#define R200_PP_TXABLEND2_7 0x2f7c +/* gap */ +#define R200_RB3D_BLENDCOLOR 0x3218 /* ARGB 8888 */ +#define R200_RB3D_ABLENDCNTL 0x321C /* see BLENDCTL */ +#define R200_RB3D_CBLENDCNTL 0x3220 /* see BLENDCTL */ + +/* + * Offsets in TCL vector state. NOTE: Hardwiring matrix positions. + * Multiple contexts could collaberate to eliminate state bouncing. + */ +#define R200_VS_LIGHT_AMBIENT_ADDR 0x00000028 +#define R200_VS_LIGHT_DIFFUSE_ADDR 0x00000030 +#define R200_VS_LIGHT_SPECULAR_ADDR 0x00000038 +#define R200_VS_LIGHT_DIRPOS_ADDR 0x00000040 +#define R200_VS_LIGHT_HWVSPOT_ADDR 0x00000048 +#define R200_VS_LIGHT_ATTENUATION_ADDR 0x00000050 +#define R200_VS_SPOT_DUAL_CONE 0x00000058 +#define R200_VS_GLOBAL_AMBIENT_ADDR 0x0000005C +#define R200_VS_FOG_PARAM_ADDR 0x0000005D +#define R200_VS_EYE_VECTOR_ADDR 0x0000005E +#define R200_VS_UCP_ADDR 0x00000060 +#define R200_VS_PNT_SPRITE_VPORT_SCALE 0x00000068 +#define R200_VS_MATRIX_0_MV 0x00000080 +#define R200_VS_MATRIX_1_INV_MV 0x00000084 +#define R200_VS_MATRIX_2_MVP 0x00000088 +#define R200_VS_MATRIX_3_TEX0 0x0000008C +#define R200_VS_MATRIX_4_TEX1 0x00000090 +#define R200_VS_MATRIX_5_TEX2 0x00000094 +#define R200_VS_MATRIX_6_TEX3 0x00000098 +#define R200_VS_MATRIX_7_TEX4 0x0000009C +#define R200_VS_MATRIX_8_TEX5 0x000000A0 +#define R200_VS_MAT_0_EMISS 0x000000B0 +#define R200_VS_MAT_0_AMB 0x000000B1 +#define R200_VS_MAT_0_DIF 0x000000B2 +#define R200_VS_MAT_0_SPEC 0x000000B3 +#define R200_VS_MAT_1_EMISS 0x000000B4 +#define R200_VS_MAT_1_AMB 0x000000B5 +#define R200_VS_MAT_1_DIF 0x000000B6 +#define R200_VS_MAT_1_SPEC 0x000000B7 +#define R200_VS_EYE2CLIP_MTX 0x000000B8 +#define R200_VS_PNT_SPRITE_ATT_CONST 0x000000BC +#define R200_VS_PNT_SPRITE_EYE_IN_MODEL 0x000000BD +#define R200_VS_PNT_SPRITE_CLAMP 0x000000BE +#define R200_VS_MAX 0x000001C0 + +/* + * Offsets in TCL scalar state + */ +#define R200_SS_LIGHT_DCD_ADDR 0x00000000 +#define R200_SS_LIGHT_DCM_ADDR 0x00000008 +#define R200_SS_LIGHT_SPOT_EXPONENT_ADDR 0x00000010 +#define R200_SS_LIGHT_SPOT_CUTOFF_ADDR 0x00000018 +#define R200_SS_LIGHT_SPECULAR_THRESH_ADDR 0x00000020 +#define R200_SS_LIGHT_RANGE_CUTOFF_SQRD 0x00000028 +#define R200_SS_LIGHT_RANGE_ATT_CONST 0x00000030 +#define R200_SS_VERT_GUARD_CLIP_ADJ_ADDR 0x00000080 +#define R200_SS_VERT_GUARD_DISCARD_ADJ_ADDR 0x00000081 +#define R200_SS_HORZ_GUARD_CLIP_ADJ_ADDR 0x00000082 +#define R200_SS_HORZ_GUARD_DISCARD_ADJ_ADDR 0x00000083 +#define R200_SS_MAT_0_SHININESS 0x00000100 +#define R200_SS_MAT_1_SHININESS 0x00000101 + +/* + * Matrix indices + */ +#define R200_MTX_MV 0 +#define R200_MTX_IMV 1 +#define R200_MTX_MVP 2 +#define R200_MTX_TEX0 3 +#define R200_MTX_TEX1 4 +#define R200_MTX_TEX2 5 +#define R200_MTX_TEX3 6 +#define R200_MTX_TEX4 7 +#define R200_MTX_TEX5 8 + +/* Color formats for 2d packets + */ +#define R200_CP_COLOR_FORMAT_CI8 2 +#define R200_CP_COLOR_FORMAT_ARGB1555 3 +#define R200_CP_COLOR_FORMAT_RGB565 4 +#define R200_CP_COLOR_FORMAT_ARGB8888 6 +#define R200_CP_COLOR_FORMAT_RGB332 7 +#define R200_CP_COLOR_FORMAT_RGB8 9 +#define R200_CP_COLOR_FORMAT_ARGB4444 15 + +/* + * CP type-3 packets + */ +#define R200_CP_CMD_NOP 0xC0001000 +#define R200_CP_CMD_NEXT_CHAR 0xC0001900 +#define R200_CP_CMD_PLY_NEXTSCAN 0xC0001D00 +#define R200_CP_CMD_SET_SCISSORS 0xC0001E00 +#define R200_CP_CMD_LOAD_MICROCODE 0xC0002400 +#define R200_CP_CMD_WAIT_FOR_IDLE 0xC0002600 +#define R200_CP_CMD_3D_DRAW_VBUF 0xC0002800 +#define R200_CP_CMD_3D_DRAW_IMMD 0xC0002900 +#define R200_CP_CMD_3D_DRAW_INDX 0xC0002A00 +#define R200_CP_CMD_LOAD_PALETTE 0xC0002C00 +#define R200_CP_CMD_3D_LOAD_VBPNTR 0xC0002F00 +#define R200_CP_CMD_INDX_BUFFER 0xC0003300 +#define R200_CP_CMD_3D_DRAW_VBUF_2 0xC0003400 +#define R200_CP_CMD_3D_DRAW_IMMD_2 0xC0003500 +#define R200_CP_CMD_3D_DRAW_INDX_2 0xC0003600 +#define R200_CP_CMD_PAINT 0xC0009100 +#define R200_CP_CMD_BITBLT 0xC0009200 +#define R200_CP_CMD_SMALLTEXT 0xC0009300 +#define R200_CP_CMD_HOSTDATA_BLT 0xC0009400 +#define R200_CP_CMD_POLYLINE 0xC0009500 +#define R200_CP_CMD_POLYSCANLINES 0xC0009800 +#define R200_CP_CMD_PAINT_MULTI 0xC0009A00 +#define R200_CP_CMD_BITBLT_MULTI 0xC0009B00 +#define R200_CP_CMD_TRANS_BITBLT 0xC0009C00 + +#endif diff --git a/src/mesa/drivers/dri/r300/r200_sanity.h b/src/mesa/drivers/dri/r300/r200_sanity.h new file mode 100644 index 00000000000..30852b49092 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r200_sanity.h @@ -0,0 +1,7 @@ +#ifndef R200_SANITY_H +#define R200_SANITY_H + +extern int r200SanityCmdBuffer(r200ContextPtr rmesa, + int nbox, drm_clip_rect_t * boxes); + +#endif diff --git a/src/mesa/drivers/dri/r300/r200_state.h b/src/mesa/drivers/dri/r300/r200_state.h new file mode 100644 index 00000000000..50bb51af6c4 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r200_state.h @@ -0,0 +1,61 @@ +/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_state.h,v 1.2 2002/11/05 17:46:08 tsi Exp $ */ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Keith Whitwell + */ + +#ifndef __R200_STATE_H__ +#define __R200_STATE_H__ + +#ifdef GLX_DIRECT_RENDERING + +#include "r200_context.h" + +extern void r200InitState(r200ContextPtr rmesa); +extern void r200InitStateFuncs(struct dd_function_table *functions); +extern void r200InitTnlFuncs(GLcontext * ctx); + +extern void r200UpdateMaterial(GLcontext * ctx); + +extern void r200SetCliprects(r200ContextPtr rmesa, GLenum mode); +extern void r200RecalcScissorRects(r200ContextPtr rmesa); +extern void r200UpdateViewportOffset(GLcontext * ctx); +extern void r200UpdateWindow(GLcontext * ctx); + +extern void r200ValidateState(GLcontext * ctx); + +extern void r200PrintDirty(r200ContextPtr rmesa, const char *msg); + +extern void r200LightingSpaceChange(GLcontext * ctx); + +#endif +#endif diff --git a/src/mesa/drivers/dri/r300/r200_tcl.h b/src/mesa/drivers/dri/r300/r200_tcl.h new file mode 100644 index 00000000000..4528cf152b2 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r200_tcl.h @@ -0,0 +1,50 @@ +/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_tcl.h,v 1.2 2002/12/16 16:18:55 dawes Exp $ */ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Keith Whitwell + */ + +#ifndef __R200_TCL_H__ +#define __R200_TCL_H__ + +#ifdef GLX_DIRECT_RENDERING + +#include "r200_context.h" + +extern void r200TclPrimitive(GLcontext * ctx, GLenum prim, int hw_prim); +extern void r200EmitEltPrimitive(GLcontext * ctx, GLuint first, GLuint last, + GLuint flags); +extern void r200EmitPrimitive(GLcontext * ctx, GLuint first, GLuint last, + GLuint flags); + +#endif +#endif diff --git a/src/mesa/drivers/dri/r300/r200_tex.h b/src/mesa/drivers/dri/r300/r200_tex.h new file mode 100644 index 00000000000..bc0dbc3fbb1 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r200_tex.h @@ -0,0 +1,51 @@ +/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_tex.h,v 1.1 2002/10/30 12:51:53 alanh Exp $ */ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Keith Whitwell + */ + +#ifndef __R200_TEX_H__ +#define __R200_TEX_H__ + +#ifdef GLX_DIRECT_RENDERING + +extern void r200UpdateTextureState(GLcontext * ctx); + +extern int r200UploadTexImages(r200ContextPtr rmesa, r200TexObjPtr t, + GLuint face); + +extern void r200DestroyTexObj(r200ContextPtr rmesa, r200TexObjPtr t); + +extern void r200InitTextureFuncs(struct dd_function_table *functions); + +#endif +#endif /* __R200_TEX_H__ */ diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c new file mode 100644 index 00000000000..7c51e10a171 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c @@ -0,0 +1,489 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Nicolai Haehnle + */ + +#include "glheader.h" +#include "state.h" +#include "imports.h" +#include "macros.h" +#include "context.h" +#include "swrast/swrast.h" +#include "simple_list.h" + +#include "drm.h" +#include "radeon_drm.h" + +#include "radeon_ioctl.h" +#include "r300_context.h" +#include "r300_ioctl.h" +#include "r300_reg.h" +#include "r300_cmdbuf.h" + + +/** + * Send the current command buffer via ioctl to the hardware. + */ +int r300FlushCmdBuf(r300ContextPtr r300, const char* caller) +{ + int ret; + int i; + drm_radeon_cmd_buffer_t cmd; + int start; + + if (r300->radeon.lost_context) + start = 0; + else + start = r300->cmdbuf.count_reemit; + + if (RADEON_DEBUG & DEBUG_IOCTL) { + fprintf(stderr, "%s from %s\n", __FUNCTION__, caller); + + if (RADEON_DEBUG & DEBUG_VERBOSE) + for (i = start; i < r300->cmdbuf.count_used; ++i) + fprintf(stderr, "%d: %08x\n", i, + r300->cmdbuf.cmd_buf[i]); + } + + LOCK_HARDWARE(&r300->radeon); + + cmd.buf = (char*)(r300->cmdbuf.cmd_buf + start); + cmd.bufsz = (r300->cmdbuf.count_used - start) * 4; + +#if 0 // TODO: scissors + if (rmesa->state.scissor.enabled) { + cmd.nbox = rmesa->state.scissor.numClipRects; + cmd.boxes = (drm_clip_rect_t *) rmesa->state.scissor.pClipRects; + } else { +#endif + cmd.nbox = r300->radeon.numClipRects; + cmd.boxes = (drm_clip_rect_t *) r300->radeon.pClipRects; +#if 0 + } +#endif + + ret = drmCommandWrite(r300->radeon.dri.fd, + DRM_RADEON_CMDBUF, &cmd, sizeof(cmd)); + if (ret) { + UNLOCK_HARDWARE(&r300->radeon); + fprintf(stderr, "drmCommandWrite: %d\n", ret); + exit(-1); + } + + if (RADEON_DEBUG & DEBUG_SYNC) { + fprintf(stderr, "Syncing in %s\n\n", __FUNCTION__); + radeonWaitForIdleLocked(&r300->radeon); + } + + UNLOCK_HARDWARE(&r300->radeon); + + r300->cmdbuf.count_used = 0; + r300->cmdbuf.count_reemit = 0; + + return ret; +} + + +static void print_state_atom(struct r300_state_atom *state, int dwords) +{ + int i; + + fprintf(stderr, " emit %s/%d/%d\n", state->name, dwords, state->cmd_size); + + if (RADEON_DEBUG & DEBUG_VERBOSE) + for (i = 0; i < dwords; i++) + fprintf(stderr, " %s[%d]: %08X\n", state->name, i, + state->cmd[i]); +} + +/** + * Emit all atoms with a dirty field equal to dirty. + * + * The caller must have ensured that there is enough space in the command + * buffer. + */ +static __inline__ void r300DoEmitState(r300ContextPtr r300, GLboolean dirty) +{ + struct r300_state_atom* atom; + uint32_t* dest; + + dest = r300->cmdbuf.cmd_buf + r300->cmdbuf.count_used; + + if (RADEON_DEBUG & DEBUG_STATE) { + foreach(atom, &r300->hw.atomlist) { + if ((atom->dirty || r300->hw.all_dirty) == dirty) { + int dwords = (*atom->check)(r300, atom); + + if (dwords) + print_state_atom(atom, dwords); + else + fprintf(stderr, " skip state %s\n", + atom->name); + } + } + } + + foreach(atom, &r300->hw.atomlist) { + if ((atom->dirty || r300->hw.all_dirty) == dirty) { + int dwords = (*atom->check)(r300, atom); + + if (dwords) { + memcpy(dest, atom->cmd, dwords*4); + dest += dwords; + r300->cmdbuf.count_used += dwords; + atom->dirty = GL_FALSE; + } + } + } +} + + +/** + * Copy dirty hardware state atoms into the command buffer. + * + * We also copy out clean state if we're at the start of a buffer. That makes + * it easy to recover from lost contexts. + */ +void r300EmitState(r300ContextPtr r300) +{ + if (RADEON_DEBUG & (DEBUG_STATE | DEBUG_PRIMS)) + fprintf(stderr, "%s\n", __FUNCTION__); + + if (r300->cmdbuf.count_used && !r300->hw.is_dirty && !r300->hw.all_dirty) + return; + + /* To avoid going across the entire set of states multiple times, just check + * for enough space for the case of emitting all state, and inline the + * r300AllocCmdBuf code here without all the checks. + */ + r300EnsureCmdBufSpace(r300, r300->hw.max_state_size, __FUNCTION__); + + if (!r300->cmdbuf.count_used) { + if (RADEON_DEBUG & DEBUG_STATE) + fprintf(stderr, "Begin reemit state\n"); + + r300DoEmitState(r300, GL_FALSE); + r300->cmdbuf.count_reemit = r300->cmdbuf.count_used; + } + + if (RADEON_DEBUG & DEBUG_STATE) + fprintf(stderr, "Begin dirty state\n"); + + r300DoEmitState(r300, GL_TRUE); + + assert(r300->cmdbuf.count_used < r300->cmdbuf.size); + + r300->hw.is_dirty = GL_FALSE; + r300->hw.all_dirty = GL_FALSE; +} + + +static __inline__ uint32_t cmducs(int reg, int count) +{ + drm_r300_cmd_header_t cmd; + + cmd.unchecked_state.cmd_type = R300_CMD_UNCHECKED_STATE; + cmd.unchecked_state.count = count; + cmd.unchecked_state.reghi = ((unsigned int)reg & 0xFF00) >> 8; + cmd.unchecked_state.reglo = ((unsigned int)reg & 0x00FF); + + return cmd.u; +} + +static __inline__ uint32_t cmdvpu(int addr, int count) +{ + drm_r300_cmd_header_t cmd; + + cmd.vpu.cmd_type = R300_CMD_VPU; + cmd.vpu.count = count; + cmd.vpu.adrhi = ((unsigned int)addr & 0xFF00) >> 8; + cmd.vpu.adrlo = ((unsigned int)addr & 0x00FF); + + return cmd.u; +} + +#define CHECK( NM, COUNT ) \ +static int check_##NM( r300ContextPtr r300, \ + struct r300_state_atom* atom ) \ +{ \ + (void) atom; (void) r300; \ + return (COUNT); \ +} + +#define ucscount(ptr) (((drm_r300_cmd_header_t*)(ptr))->unchecked_state.count) +#define vpucount(ptr) (((drm_r300_cmd_header_t*)(ptr))->vpu.count) + +CHECK( always, atom->cmd_size ) +CHECK( never, 0 ) +CHECK( variable, ucscount(atom->cmd) ? (1 + ucscount(atom->cmd)) : 0 ) +CHECK( vpu, vpucount(atom->cmd) ? (1 + vpucount(atom->cmd)*4) : 0 ) + +#undef ucscount + +#define ALLOC_STATE( ATOM, CHK, SZ, NM, IDX ) \ + do { \ + r300->hw.ATOM.cmd_size = SZ; \ + r300->hw.ATOM.cmd = (uint32_t*)CALLOC(SZ * sizeof(uint32_t)); \ + r300->hw.ATOM.name = NM; \ + r300->hw.ATOM.idx = IDX; \ + r300->hw.ATOM.check = check_##CHK; \ + r300->hw.ATOM.dirty = GL_FALSE; \ + r300->hw.max_state_size += SZ; \ + } while (0) + + +/** + * Allocate memory for the command buffer and initialize the state atom + * list. Note that the initial hardware state is set by r300InitState(). + */ +void r300InitCmdBuf(r300ContextPtr r300) +{ + int size; + + r300->hw.max_state_size = 0; + + /* Initialize state atoms */ + ALLOC_STATE( vpt, always, R300_VPT_CMDSIZE, "vpt", 0 ); + r300->hw.vpt.cmd[R300_VPT_CMD_0] = cmducs(R300_SE_VPORT_XSCALE, 6); + ALLOC_STATE( unk2080, always, 2, "unk2080", 0 ); + r300->hw.unk2080.cmd[0] = cmducs(0x2080, 1); + ALLOC_STATE( ovf, always, R300_OVF_CMDSIZE, "ovf", 0 ); + r300->hw.ovf.cmd[R300_OVF_CMD_0] = cmducs(R300_VAP_OUTPUT_VTX_FMT_0, 2); + ALLOC_STATE( unk20B0, always, 3, "unk20B0", 0 ); + r300->hw.unk20B0.cmd[0] = cmducs(0x20B0, 2); + ALLOC_STATE( unk2134, always, 3, "unk2134", 0 ); + r300->hw.unk2134.cmd[0] = cmducs(0x2134, 2); + ALLOC_STATE( unk2140, always, 2, "unk2140", 0 ); + r300->hw.unk2140.cmd[0] = cmducs(0x2140, 1); + ALLOC_STATE( vir[0], variable, R300_VIR_CMDSIZE, "vir/0", 0 ); + r300->hw.vir[0].cmd[R300_VIR_CMD_0] = cmducs(R300_VAP_INPUT_ROUTE_0_0, 1); + ALLOC_STATE( vir[1], variable, R300_VIR_CMDSIZE, "vir/1", 1 ); + r300->hw.vir[1].cmd[R300_VIR_CMD_0] = cmducs(R300_VAP_INPUT_ROUTE_1_0, 1); + ALLOC_STATE( vic, always, R300_VIC_CMDSIZE, "vic", 0 ); + r300->hw.vic.cmd[R300_VIC_CMD_0] = cmducs(R300_VAP_INPUT_CNTL_0, 2); + ALLOC_STATE( unk21DC, always, 2, "unk21DC", 0 ); + r300->hw.unk21DC.cmd[0] = cmducs(0x21DC, 1); + ALLOC_STATE( unk221C, always, 2, "unk221C", 0 ); + r300->hw.unk221C.cmd[0] = cmducs(0x221C, 1); + ALLOC_STATE( unk2220, always, 5, "unk2220", 0 ); + r300->hw.unk2220.cmd[0] = cmducs(0x2220, 4); + ALLOC_STATE( unk2288, always, 2, "unk2288", 0 ); + r300->hw.unk2288.cmd[0] = cmducs(0x2288, 1); + ALLOC_STATE( pvs, always, R300_PVS_CMDSIZE, "pvs", 0 ); + r300->hw.pvs.cmd[R300_PVS_CMD_0] = cmducs(R300_VAP_PVS_CNTL_1, 3); + ALLOC_STATE( unk4008, always, 2, "unk4008", 0 ); + r300->hw.unk4008.cmd[0] = cmducs(0x4008, 1); + ALLOC_STATE( unk4010, always, 6, "unk4010", 0 ); + r300->hw.unk4010.cmd[0] = cmducs(0x4010, 5); + ALLOC_STATE( txe, always, R300_TXE_CMDSIZE, "txe", 0 ); + r300->hw.txe.cmd[R300_TXE_CMD_0] = cmducs(R300_TX_ENABLE, 1); + ALLOC_STATE( unk4200, always, 5, "unk4200", 0 ); + r300->hw.unk4200.cmd[0] = cmducs(0x4200, 4); + ALLOC_STATE( unk4214, always, 2, "unk4214", 0 ); + r300->hw.unk4214.cmd[0] = cmducs(0x4214, 1); + ALLOC_STATE( unk4230, always, 4, "unk4230", 0 ); + r300->hw.unk4230.cmd[0] = cmducs(0x4230, 3); + ALLOC_STATE( unk4260, always, 4, "unk4260", 0 ); + r300->hw.unk4260.cmd[0] = cmducs(0x4260, 3); + ALLOC_STATE( unk4274, always, 5, "unk4274", 0 ); + r300->hw.unk4274.cmd[0] = cmducs(0x4274, 4); + ALLOC_STATE( unk4288, always, 6, "unk4288", 0 ); + r300->hw.unk4288.cmd[0] = cmducs(0x4288, 5); + ALLOC_STATE( unk42A0, always, 2, "unk42A0", 0 ); + r300->hw.unk42A0.cmd[0] = cmducs(0x42A0, 1); + ALLOC_STATE( unk42B4, always, 3, "unk42B4", 0 ); + r300->hw.unk42B4.cmd[0] = cmducs(0x42B4, 2); + ALLOC_STATE( unk42C0, always, 3, "unk42C0", 0 ); + r300->hw.unk42C0.cmd[0] = cmducs(0x42C0, 2); + ALLOC_STATE( rc, always, R300_RC_CMDSIZE, "rc", 0 ); + r300->hw.rc.cmd[R300_RC_CMD_0] = cmducs(R300_RS_CNTL_0, 2); + ALLOC_STATE( ri, always, R300_RI_CMDSIZE, "ri", 0 ); + r300->hw.ri.cmd[R300_RI_CMD_0] = cmducs(R300_RS_INTERP_0, 8); + ALLOC_STATE( rr, variable, R300_RR_CMDSIZE, "rr", 0 ); + r300->hw.rr.cmd[R300_RR_CMD_0] = cmducs(R300_RS_ROUTE_0, 1); + ALLOC_STATE( unk43A4, always, 3, "unk43A4", 0 ); + r300->hw.unk43A4.cmd[0] = cmducs(0x43A4, 2); + ALLOC_STATE( unk43E0, always, 4, "unk43E0", 0 ); + r300->hw.unk43E0.cmd[0] = cmducs(0x43E0, 3); + ALLOC_STATE( fp, always, R300_FP_CMDSIZE, "fp", 0 ); + r300->hw.fp.cmd[R300_FP_CMD_0] = cmducs(R300_PFS_CNTL_0, 3); + r300->hw.fp.cmd[R300_FP_CMD_1] = cmducs(R300_PFS_NODE_0, 4); + ALLOC_STATE( unk46A4, always, 6, "unk46A4", 0 ); + r300->hw.unk46A4.cmd[0] = cmducs(0x46A4, 5); + ALLOC_STATE( fpi[0], variable, R300_FPI_CMDSIZE, "fpi/0", 0 ); + r300->hw.fpi[0].cmd[R300_FPI_CMD_0] = cmducs(R300_PFS_INSTR0_0, 1); + ALLOC_STATE( fpi[1], variable, R300_FPI_CMDSIZE, "fpi/1", 1 ); + r300->hw.fpi[1].cmd[R300_FPI_CMD_0] = cmducs(R300_PFS_INSTR1_0, 1); + ALLOC_STATE( fpi[2], variable, R300_FPI_CMDSIZE, "fpi/2", 2 ); + r300->hw.fpi[2].cmd[R300_FPI_CMD_0] = cmducs(R300_PFS_INSTR2_0, 1); + ALLOC_STATE( fpi[3], variable, R300_FPI_CMDSIZE, "fpi/3", 3 ); + r300->hw.fpi[3].cmd[R300_FPI_CMD_0] = cmducs(R300_PFS_INSTR3_0, 1); + ALLOC_STATE( unk4BC0, always, 2, "unk4BC0", 0 ); + r300->hw.unk4BC0.cmd[0] = cmducs(0x4BC0, 1); + ALLOC_STATE( unk4BC8, always, 4, "unk4BC8", 0 ); + r300->hw.unk4BC8.cmd[0] = cmducs(0x4BC8, 3); + ALLOC_STATE( at, always, R300_AT_CMDSIZE, "at", 0 ); + r300->hw.at.cmd[R300_AT_CMD_0] = cmducs(R300_PP_ALPHA_TEST, 1); + ALLOC_STATE( unk4BD8, always, 2, "unk4BD8", 0 ); + r300->hw.unk4BD8.cmd[0] = cmducs(0x4BD8, 1); + ALLOC_STATE( unk4E00, always, 2, "unk4E00", 0 ); + r300->hw.unk4E00.cmd[0] = cmducs(0x4E00, 1); + ALLOC_STATE( bld, always, R300_BLD_CMDSIZE, "bld", 0 ); + r300->hw.bld.cmd[R300_BLD_CMD_0] = cmducs(R300_RB3D_CBLEND, 2); + ALLOC_STATE( cmk, always, R300_CMK_CMDSIZE, "cmk", 0 ); + r300->hw.cmk.cmd[R300_CMK_CMD_0] = cmducs(R300_RB3D_COLORMASK, 1); + ALLOC_STATE( unk4E10, always, 4, "unk4E10", 0 ); + r300->hw.unk4E10.cmd[0] = cmducs(0x4E10, 3); + ALLOC_STATE( cb, always, R300_CB_CMDSIZE, "cb", 0 ); + r300->hw.cb.cmd[R300_CB_CMD_0] = cmducs(R300_RB3D_COLOROFFSET0, 1); + r300->hw.cb.cmd[R300_CB_CMD_1] = cmducs(R300_RB3D_COLORPITCH0, 1); + ALLOC_STATE( unk4E50, always, 10, "unk4E50", 0 ); + r300->hw.unk4E50.cmd[0] = cmducs(0x4E50, 9); + ALLOC_STATE( unk4E88, always, 2, "unk4E88", 0 ); + r300->hw.unk4E88.cmd[0] = cmducs(0x4E88, 1); + ALLOC_STATE( zc, always, R300_ZC_CMDSIZE, "zc", 0 ); + r300->hw.zc.cmd[R300_ZC_CMD_0] = cmducs(R300_RB3D_ZCNTL_0, 2); + ALLOC_STATE( unk4F08, always, 6, "unk4F08", 0 ); + r300->hw.unk4F08.cmd[0] = cmducs(0x4F08, 5); + ALLOC_STATE( zb, always, R300_ZB_CMDSIZE, "zb", 0 ); + r300->hw.zb.cmd[R300_ZB_CMD_0] = cmducs(R300_RB3D_DEPTHOFFSET, 2); + ALLOC_STATE( unk4F28, always, 2, "unk4F28", 0 ); + r300->hw.unk4F28.cmd[0] = cmducs(0x4F28, 1); + ALLOC_STATE( unk4F30, always, 3, "unk4F30", 0 ); + r300->hw.unk4F30.cmd[0] = cmducs(0x4F30, 2); + ALLOC_STATE( unk4F44, always, 2, "unk4F44", 0 ); + r300->hw.unk4F44.cmd[0] = cmducs(0x4F44, 1); + ALLOC_STATE( unk4F54, always, 2, "unk4F54", 0 ); + r300->hw.unk4F54.cmd[0] = cmducs(0x4F54, 1); + + ALLOC_STATE( vpi, vpu, R300_VPI_CMDSIZE, "vpi", 0 ); + r300->hw.vpi.cmd[R300_VPI_CMD_0] = cmdvpu(R300_PVS_UPLOAD_PROGRAM, 0); + ALLOC_STATE( vpp, vpu, R300_VPP_CMDSIZE, "vpp", 0 ); + r300->hw.vpp.cmd[R300_VPP_CMD_0] = cmdvpu(R300_PVS_UPLOAD_PARAMETERS, 0); + + /* Setup the atom linked list */ + make_empty_list(&r300->hw.atomlist); + r300->hw.atomlist.name = "atom-list"; + + insert_at_tail(&r300->hw.atomlist, &r300->hw.vpt); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk2080); + insert_at_tail(&r300->hw.atomlist, &r300->hw.ovf); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk20B0); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk2134); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk2140); + insert_at_tail(&r300->hw.atomlist, &r300->hw.vir[0]); + insert_at_tail(&r300->hw.atomlist, &r300->hw.vir[1]); + insert_at_tail(&r300->hw.atomlist, &r300->hw.vic); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk21DC); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk221C); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk2220); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk2288); + insert_at_tail(&r300->hw.atomlist, &r300->hw.pvs); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4008); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4010); + insert_at_tail(&r300->hw.atomlist, &r300->hw.txe); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4200); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4214); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4230); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4260); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4274); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4288); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk42A0); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk42B4); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk42C0); + insert_at_tail(&r300->hw.atomlist, &r300->hw.rc); + insert_at_tail(&r300->hw.atomlist, &r300->hw.ri); + insert_at_tail(&r300->hw.atomlist, &r300->hw.rr); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk43A4); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk43E0); + insert_at_tail(&r300->hw.atomlist, &r300->hw.fp); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk46A4); + insert_at_tail(&r300->hw.atomlist, &r300->hw.fpi[0]); + insert_at_tail(&r300->hw.atomlist, &r300->hw.fpi[1]); + insert_at_tail(&r300->hw.atomlist, &r300->hw.fpi[2]); + insert_at_tail(&r300->hw.atomlist, &r300->hw.fpi[3]); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4BC0); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4BC8); + insert_at_tail(&r300->hw.atomlist, &r300->hw.at); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4BD8); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4E00); + insert_at_tail(&r300->hw.atomlist, &r300->hw.bld); + insert_at_tail(&r300->hw.atomlist, &r300->hw.cmk); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4E10); + insert_at_tail(&r300->hw.atomlist, &r300->hw.cb); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4E50); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4E88); + insert_at_tail(&r300->hw.atomlist, &r300->hw.zc); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4F08); + insert_at_tail(&r300->hw.atomlist, &r300->hw.zb); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4F28); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4F30); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4F44); + insert_at_tail(&r300->hw.atomlist, &r300->hw.unk4F54); + + insert_at_tail(&r300->hw.atomlist, &r300->hw.vpi); + insert_at_tail(&r300->hw.atomlist, &r300->hw.vpp); + + r300->hw.is_dirty = GL_TRUE; + r300->hw.all_dirty = GL_TRUE; + + /* Initialize command buffer */ + size = 256 * driQueryOptioni(&r300->radeon.optionCache, "command_buffer_size"); + if (size < 2*r300->hw.max_state_size) + size = 2*r300->hw.max_state_size; + + if (RADEON_DEBUG & DEBUG_IOCTL) + fprintf(stderr, + "Allocating %d bytes command buffer (max state is %d bytes)\n", + size*4, r300->hw.max_state_size*4); + + r300->cmdbuf.size = size; + r300->cmdbuf.cmd_buf = (uint32_t*)CALLOC(size*4); + r300->cmdbuf.count_used = 0; + r300->cmdbuf.count_reemit = 0; +} + + +/** + * Destroy the command buffer and state atoms. + */ +void r300DestroyCmdBuf(r300ContextPtr r300) +{ + struct r300_state_atom* atom; + + FREE(r300->cmdbuf.cmd_buf); + + foreach(atom, &r300->hw.atomlist) { + FREE(atom->cmd); + } +} + diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.h b/src/mesa/drivers/dri/r300/r300_cmdbuf.h new file mode 100644 index 00000000000..32277257534 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.h @@ -0,0 +1,102 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Nicolai Haehnle + */ + +#ifndef __R300_CMDBUF_H__ +#define __R300_CMDBUF_H__ + +#include "r300_context.h" + + +extern int r300FlushCmdBuf(r300ContextPtr r300, const char* caller); + +extern void r300EmitState(r300ContextPtr r300); + +extern void r300InitCmdBuf(r300ContextPtr r300); +extern void r300DestroyCmdBuf(r300ContextPtr r300); + + +/** + * Make sure that enough space is available in the command buffer + * by flushing if necessary. + */ +static __inline__ void r300EnsureCmdBufSpace(r300ContextPtr r300, + int dwords, const char* caller) +{ + assert(dwords < r300->cmdbuf.size); + + if (r300->cmdbuf.count_used + dwords > r300->cmdbuf.size) + r300FlushCmdBuf(r300, caller); +} + + +/** + * Allocate the given number of dwords in the command buffer and return + * a pointer to the allocated area. + * When necessary, these functions cause a flush. r300AllocCmdBuf() also + * causes state reemission after a flush. This is necessary to ensure + * correct hardware state after an unlock. + */ +static __inline__ uint32_t* r300RawAllocCmdBuf(r300ContextPtr r300, + int dwords, const char* caller) +{ + uint32_t* ptr; + + r300EnsureCmdBufSpace(r300, dwords, caller); + + ptr = &r300->cmdbuf.cmd_buf[r300->cmdbuf.count_used]; + r300->cmdbuf.count_used += dwords; + return ptr; +} + +static __inline__ uint32_t* r300AllocCmdBuf(r300ContextPtr r300, + int dwords, const char* caller) +{ + uint32_t* ptr; + + r300EnsureCmdBufSpace(r300, dwords, caller); + + if (!r300->cmdbuf.count_used) { + if (RADEON_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "Reemit state after flush (from %s)\n", + caller); + r300EmitState(r300); + } + + ptr = &r300->cmdbuf.cmd_buf[r300->cmdbuf.count_used]; + r300->cmdbuf.count_used += dwords; + return ptr; +} + + +#endif /* __R300_CMDBUF_H__ */ diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c new file mode 100644 index 00000000000..3604b1181e8 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_context.c @@ -0,0 +1,308 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Keith Whitwell + * Nicolai Haehnle + */ + +#include "glheader.h" +#include "api_arrayelt.h" +#include "context.h" +#include "simple_list.h" +#include "imports.h" +#include "matrix.h" +#include "extensions.h" +#include "state.h" + +#include "swrast/swrast.h" +#include "swrast_setup/swrast_setup.h" +#include "array_cache/acache.h" + +#include "tnl/tnl.h" +#include "tnl/t_pipeline.h" + +#include "drivers/common/driverfuncs.h" + +#include "radeon_ioctl.h" +#include "radeon_span.h" +#include "r300_context.h" +#include "r300_cmdbuf.h" +#include "r300_state.h" +#include "r300_ioctl.h" + +#include "vblank.h" +#include "utils.h" +#include "xmlpool.h" /* for symbolic values of enum-type options */ + + +/* Extension strings exported by the R300 driver. + */ +static const char *const card_extensions[] = { + "GL_ARB_multisample", + "GL_ARB_multitexture", + "GL_ARB_texture_border_clamp", + "GL_ARB_texture_compression", + "GL_ARB_texture_cube_map", + "GL_ARB_texture_env_add", + "GL_ARB_texture_env_combine", + "GL_ARB_texture_env_dot3", + "GL_ARB_texture_mirrored_repeat", + "GL_ARB_vertex_buffer_object", + "GL_ARB_vertex_program", + "GL_EXT_blend_equation_separate", + "GL_EXT_blend_func_separate", + "GL_EXT_blend_minmax", + "GL_EXT_blend_subtract", + "GL_EXT_secondary_color", + "GL_EXT_stencil_wrap", + "GL_EXT_texture_edge_clamp", + "GL_EXT_texture_env_combine", + "GL_EXT_texture_env_dot3", + "GL_EXT_texture_filter_anisotropic", + "GL_EXT_texture_lod_bias", + "GL_EXT_texture_mirror_clamp", + "GL_EXT_texture_rectangle", + "GL_ATI_texture_env_combine3", + "GL_ATI_texture_mirror_once", + "GL_MESA_pack_invert", + "GL_MESA_ycbcr_texture", + "GL_NV_blend_square", + "GL_NV_vertex_program", + "GL_SGIS_generate_mipmap", + NULL +}; + +//extern const struct tnl_pipeline_stage _r300_render_stage; +//extern const struct tnl_pipeline_stage _r300_tcl_stage; + +static const struct tnl_pipeline_stage *r300_pipeline[] = { + + /* Try and go straight to t&l + */ + //&_r200_tcl_stage, + + /* Catch any t&l fallbacks + */ + &_tnl_vertex_transform_stage, + &_tnl_normal_transform_stage, + &_tnl_lighting_stage, + &_tnl_fog_coordinate_stage, + &_tnl_texgen_stage, + &_tnl_texture_transform_stage, + &_tnl_vertex_program_stage, + + /* Try again to go to tcl? + * - no good for asymmetric-twoside (do with multipass) + * - no good for asymmetric-unfilled (do with multipass) + * - good for material + * - good for texgen + * - need to manipulate a bit of state + * + * - worth it/not worth it? + */ + + /* Else do them here. + */ + //&_r300_render_stage, + &_tnl_render_stage, /* FALLBACK */ + 0, +}; + + +/* Create the device specific rendering context. + */ +GLboolean r300CreateContext(const __GLcontextModes * glVisual, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate) +{ + __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; + radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private); + struct dd_function_table functions; + r300ContextPtr r300; + GLcontext *ctx; + int i; + int tcl_mode; + + assert(glVisual); + assert(driContextPriv); + assert(screen); + + /* Allocate the R300 context */ + r300 = (r300ContextPtr) CALLOC(sizeof(*r300)); + if (!r300) + return GL_FALSE; + + /* Parse configuration files. + * Do this here so that initialMaxAnisotropy is set before we create + * the default textures. + */ + driParseConfigFiles(&r300->radeon.optionCache, &screen->optionCache, + screen->driScreen->myNum, "r300"); + + /* Init default driver functions then plug in our R300-specific functions + * (the texture functions are especially important) + */ + _mesa_init_driver_functions(&functions); + r300InitIoctlFuncs(&functions); + //r200InitStateFuncs(&functions); + //r200InitTextureFuncs(&functions); + + if (!radeonInitContext(&r300->radeon, &functions, + glVisual, driContextPriv, sharedContextPrivate)) { + FREE(r300); + return GL_FALSE; + } + + /* Init r300 context data */ + + /* Set the maximum texture size small enough that we can guarentee that + * all texture units can bind a maximal texture and have them both in + * texturable memory at once. + */ + + ctx = r300->radeon.glCtx; + ctx->Const.MaxTextureImageUnits = driQueryOptioni(&r300->radeon.optionCache, + "texture_image_units"); + ctx->Const.MaxTextureCoordUnits = driQueryOptioni(&r300->radeon.optionCache, + "texture_coord_units"); + ctx->Const.MaxTextureUnits = MIN2(ctx->Const.MaxTextureImageUnits, + ctx->Const.MaxTextureCoordUnits); + ctx->Const.MaxTextureMaxAnisotropy = 16.0; + + /* No wide points. + */ + ctx->Const.MinPointSize = 1.0; + ctx->Const.MinPointSizeAA = 1.0; + ctx->Const.MaxPointSize = 1.0; + ctx->Const.MaxPointSizeAA = 1.0; + + ctx->Const.MinLineWidth = 1.0; + ctx->Const.MinLineWidthAA = 1.0; + ctx->Const.MaxLineWidth = 1.0; + ctx->Const.MaxLineWidthAA = 1.0; + + /* Initialize the software rasterizer and helper modules. + */ + _swrast_CreateContext(ctx); + _ac_CreateContext(ctx); + _tnl_CreateContext(ctx); + _swsetup_CreateContext(ctx); + _ae_create_context(ctx); + +#if 0 + /* Install the customized pipeline: + */ + _tnl_destroy_pipeline(ctx); + _tnl_install_pipeline(ctx, r200_pipeline); + ctx->Driver.FlushVertices = r200FlushVertices; +#endif + + /* Try and keep materials and vertices separate: + */ + _tnl_isolate_materials(ctx, GL_TRUE); + + /* Configure swrast and TNL to match hardware characteristics: + */ + _swrast_allow_pixel_fog(ctx, GL_FALSE); + _swrast_allow_vertex_fog(ctx, GL_TRUE); + _tnl_allow_pixel_fog(ctx, GL_FALSE); + _tnl_allow_vertex_fog(ctx, GL_TRUE); + + driInitExtensions(ctx, card_extensions, GL_TRUE); + + radeonInitSpanFuncs(ctx); + r300InitCmdBuf(r300); + r300InitState(r300); +#if 0 + /* plug in a few more device driver functions */ + /* XXX these should really go right after _mesa_init_driver_functions() */ + r200InitPixelFuncs(ctx); + r200InitTnlFuncs(ctx); + r200InitSwtcl(ctx); +#endif + TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline; + + tcl_mode = driQueryOptioni(&r300->radeon.optionCache, "tcl_mode"); + if (1 || + driQueryOptionb(&r300->radeon.optionCache, "no_rast")) { + fprintf(stderr, "disabling 3D acceleration\n"); + FALLBACK(&r300->radeon, RADEON_FALLBACK_DISABLE, 1); + } + if (tcl_mode == DRI_CONF_TCL_SW || + !(r300->radeon.radeonScreen->chipset & RADEON_CHIPSET_TCL)) { + if (r300->radeon.radeonScreen->chipset & RADEON_CHIPSET_TCL) { + r300->radeon.radeonScreen->chipset &= ~RADEON_CHIPSET_TCL; + fprintf(stderr, "Disabling HW TCL support\n"); + } + TCL_FALLBACK(r300->radeon.glCtx, RADEON_TCL_FALLBACK_TCL_DISABLE, 1); + } + + return GL_TRUE; +} + +/* Destroy the device specific context. + */ +void r300DestroyContext(__DRIcontextPrivate * driContextPriv) +{ + GET_CURRENT_CONTEXT(ctx); + r300ContextPtr r300 = (r300ContextPtr) driContextPriv->driverPrivate; + radeonContextPtr current = ctx ? RADEON_CONTEXT(ctx) : NULL; + + /* check if we're deleting the currently bound context */ + if (&r300->radeon == current) { + radeonFlush(r300->radeon.glCtx); + _mesa_make_current2(NULL, NULL, NULL); + } + + /* Free r300 context resources */ + assert(r300); /* should never be null */ + + if (r300) { + GLboolean release_texture_heaps; + + release_texture_heaps = (r300->radeon.glCtx->Shared->RefCount == 1); + _swsetup_DestroyContext(r300->radeon.glCtx); + _tnl_DestroyContext(r300->radeon.glCtx); + _ac_DestroyContext(r300->radeon.glCtx); + _swrast_DestroyContext(r300->radeon.glCtx); + + r300DestroyCmdBuf(r300); + + /* free the Mesa context */ + r300->radeon.glCtx->DriverCtx = NULL; + _mesa_destroy_context(r300->radeon.glCtx); + + /* free the option cache */ + driDestroyOptionCache(&r300->radeon.optionCache); + + FREE(r300); + } +} diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h new file mode 100644 index 00000000000..b8523b67722 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_context.h @@ -0,0 +1,312 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Keith Whitwell + * Nicolai Haehnle + */ + +#ifndef __R300_CONTEXT_H__ +#define __R300_CONTEXT_H__ + +#include "tnl/t_vertex.h" +#include "drm.h" +#include "radeon_drm.h" +#include "dri_util.h" +#include "texmem.h" + +#include "macros.h" +#include "mtypes.h" +#include "colormac.h" +#include "radeon_context.h" + +struct r300_context; +typedef struct r300_context r300ContextRec; +typedef struct r300_context *r300ContextPtr; + +#include "radeon_lock.h" +#include "mm.h" + + +static __inline__ uint32_t r300PackFloat32(float fl) +{ + union { float fl; uint32_t u; } u; + + u.fl = fl; + return u.u; +} + +/** + * A block of hardware state. + * + * When check returns non-zero, the returned number of dwords must be + * copied verbatim into the command buffer in order to update a state atom + * when it is dirty. + */ +struct r300_state_atom { + struct r300_state_atom *next, *prev; + const char* name; /* for debug */ + int cmd_size; /* maximum size in dwords */ + GLuint idx; /* index in an array (e.g. textures) */ + uint32_t* cmd; + GLboolean dirty; + + int (*check)(r300ContextPtr, struct r300_state_atom* atom); +}; + + +#define R300_VPT_CMD_0 0 +#define R300_VPT_XSCALE 1 +#define R300_VPT_XOFFSET 2 +#define R300_VPT_YSCALE 3 +#define R300_VPT_YOFFSET 4 +#define R300_VPT_ZSCALE 5 +#define R300_VPT_ZOFFSET 6 +#define R300_VPT_CMDSIZE 7 + +#define R300_OVF_CMD_0 0 +#define R300_OVF_FMT_0 1 +#define R300_OVF_FMT_1 2 +#define R300_OVF_CMDSIZE 3 + +#define R300_VIR_CMD_0 0 /* vir is variable size (at least 1) */ +#define R300_VIR_CNTL_0 1 +#define R300_VIR_CNTL_1 2 +#define R300_VIR_CNTL_2 3 +#define R300_VIR_CNTL_3 4 +#define R300_VIR_CNTL_4 5 +#define R300_VIR_CNTL_5 6 +#define R300_VIR_CNTL_6 7 +#define R300_VIR_CNTL_7 8 +#define R300_VIR_CMDSIZE 9 + +#define R300_VIC_CMD_0 0 +#define R300_VIC_CNTL_0 1 +#define R300_VIC_CNTL_1 2 +#define R300_VIC_CMDSIZE 3 + +#define R300_PVS_CMD_0 0 +#define R300_PVS_CNTL_1 1 +#define R300_PVS_CNTL_2 2 +#define R300_PVS_CNTL_3 3 +#define R300_PVS_CMDSIZE 4 + +#define R300_TXE_CMD_0 0 +#define R300_TXE_ENABLE 1 +#define R300_TXE_CMDSIZE 2 + +#define R300_RC_CMD_0 0 +#define R300_RC_CNTL_0 1 +#define R300_RC_CNTL_1 2 +#define R300_RC_CMDSIZE 3 + +#define R300_RI_CMD_0 0 +#define R300_RI_INTERP_0 1 +#define R300_RI_INTERP_1 2 +#define R300_RI_INTERP_2 3 +#define R300_RI_INTERP_3 4 +#define R300_RI_INTERP_4 5 +#define R300_RI_INTERP_5 6 +#define R300_RI_INTERP_6 7 +#define R300_RI_INTERP_7 8 +#define R300_RI_CMDSIZE 9 + +#define R300_RR_CMD_0 0 /* rr is variable size (at least 1) */ +#define R300_RR_ROUTE_0 1 +#define R300_RR_ROUTE_1 2 +#define R300_RR_ROUTE_2 3 +#define R300_RR_ROUTE_3 4 +#define R300_RR_ROUTE_4 5 +#define R300_RR_ROUTE_5 6 +#define R300_RR_ROUTE_6 7 +#define R300_RR_ROUTE_7 8 +#define R300_RR_CMDSIZE 9 + +#define R300_FP_CMD_0 0 +#define R300_FP_CNTL0 1 +#define R300_FP_CNTL1 2 +#define R300_FP_CNTL2 3 +#define R300_FP_CMD_1 4 +#define R300_FP_NODE0 5 +#define R300_FP_NODE1 6 +#define R300_FP_NODE2 7 +#define R300_FP_NODE3 8 +#define R300_FP_CMDSIZE 9 + +#define R300_FPI_CMD_0 0 +#define R300_FPI_INSTR_0 1 +#define R300_FPI_CMDSIZE 65 + +#define R300_AT_CMD_0 0 +#define R300_AT_ALPHA_TEST 1 +#define R300_AT_CMDSIZE 2 + +#define R300_BLD_CMD_0 0 +#define R300_BLD_CBLEND 1 +#define R300_BLD_ABLEND 2 +#define R300_BLD_CMDSIZE 3 + +#define R300_CMK_CMD_0 0 +#define R300_CMK_COLORMASK 1 +#define R300_CMK_CMDSIZE 2 + +#define R300_CB_CMD_0 0 +#define R300_CB_OFFSET 1 +#define R300_CB_CMD_1 2 +#define R300_CB_PITCH 3 +#define R300_CB_CMDSIZE 4 + +#define R300_ZC_CMD_0 0 +#define R300_ZC_CNTL_0 1 +#define R300_ZC_CNTL_1 2 +#define R300_ZC_CMDSIZE 3 + +#define R300_ZB_CMD_0 0 +#define R300_ZB_OFFSET 1 +#define R300_ZB_PITCH 2 +#define R300_ZB_CMDSIZE 3 + +#define R300_VPI_CMD_0 0 +#define R300_VPI_INSTR_0 1 +#define R300_VPI_CMDSIZE 1025 /* 256 16 byte instructions */ + +#define R300_VPP_CMD_0 0 +#define R300_VPP_PARAM_0 1 +#define R300_VPP_CMDSIZE 1025 /* 256 4-component parameters */ + +/** + * Cache for hardware register state. + */ +struct r300_hw_state { + struct r300_state_atom atomlist; + + GLboolean is_dirty; + GLboolean all_dirty; + int max_state_size; /* in dwords */ + + struct r300_state_atom vpt; /* viewport (1D98) */ + struct r300_state_atom unk2080; /* (2080) */ + struct r300_state_atom ovf; /* output vertex format (2090) */ + struct r300_state_atom unk20B0; /* (20B0) */ + struct r300_state_atom unk2134; /* (2134) */ + struct r300_state_atom unk2140; /* (2140) */ + struct r300_state_atom vir[2]; /* vap input route (2150/21E0) */ + struct r300_state_atom vic; /* vap input control (2180) */ + struct r300_state_atom unk21DC; /* (21DC) */ + struct r300_state_atom unk221C; /* (221C) */ + struct r300_state_atom unk2220; /* (2220) */ + struct r300_state_atom unk2288; /* (2288) */ + struct r300_state_atom pvs; /* pvs_cntl (22D0) */ + struct r300_state_atom unk4008; /* (4008) */ + struct r300_state_atom unk4010; /* (4010) */ + struct r300_state_atom txe; /* tex enable (4104) */ + struct r300_state_atom unk4200; /* (4200) */ + struct r300_state_atom unk4214; /* (4214) */ + // what about UNKNOWN_421C? (see r300_reg.h) + struct r300_state_atom unk4230; /* (4230) */ + struct r300_state_atom unk4260; /* (4260) */ + struct r300_state_atom unk4274; /* (4274) */ + struct r300_state_atom unk4288; /* (4288) */ + struct r300_state_atom unk42A0; /* (42A0) */ + struct r300_state_atom unk42B4; /* (42B4) */ + struct r300_state_atom unk42C0; /* (42C0) */ + struct r300_state_atom rc; /* rs control (4300) */ + struct r300_state_atom ri; /* rs interpolators (4310) */ + struct r300_state_atom rr; /* rs route (4330) */ + struct r300_state_atom unk43A4; /* (43A4) */ + struct r300_state_atom unk43E0; /* (43E0) */ + struct r300_state_atom fp; /* fragment program cntl + nodes (4600) */ + struct r300_state_atom unk46A4; /* (46A4) */ + struct r300_state_atom fpi[4]; /* fp instructions (46C0/47C0/48C0/49C0) */ + struct r300_state_atom unk4BC0; /* (4BC0) */ + struct r300_state_atom unk4BC8; /* (4BC8) */ + struct r300_state_atom at; /* alpha test (4BD4) */ + struct r300_state_atom unk4BD8; /* (4BD8) */ + struct r300_state_atom unk4E00; /* (4E00) */ + struct r300_state_atom bld; /* blending (4E04) */ + struct r300_state_atom cmk; /* colormask (4E0C) */ + struct r300_state_atom unk4E10; /* (4E10) */ + struct r300_state_atom cb; /* colorbuffer (4E28) */ + struct r300_state_atom unk4E50; /* (4E50) */ + struct r300_state_atom unk4E88; /* (4E88) */ + struct r300_state_atom zc; /* z control (4F00) */ + struct r300_state_atom unk4F08; /* (4F08) */ + struct r300_state_atom zb; /* z buffer (4F20) */ + struct r300_state_atom unk4F28; /* (4F28) */ + struct r300_state_atom unk4F30; /* (4F30) */ + struct r300_state_atom unk4F44; /* (4F44) */ + struct r300_state_atom unk4F54; /* (4F54) */ + + struct r300_state_atom vpi; /* vp instructions */ + struct r300_state_atom vpp; /* vp parameters */ +}; + + +/** + * This structure holds the command buffer while it is being constructed. + * + * The first batch of commands in the buffer is always the state that needs + * to be re-emitted when the context is lost. This batch can be skipped + * otherwise. + */ +struct r300_cmdbuf { + int size; /* DWORDs allocated for buffer */ + uint32_t* cmd_buf; + int count_used; /* DWORDs filled so far */ + int count_reemit; /* size of re-emission batch */ +}; + + +/** + * State cache + */ +struct r300_state { +}; + + +/** + * R300 context structure. + */ +struct r300_context { + struct radeon_context radeon; /* parent class, must be first */ + + struct r300_hw_state hw; + struct r300_cmdbuf cmdbuf; + struct r300_state state; +}; + +#define R300_CONTEXT(ctx) ((r300ContextPtr)(ctx->DriverCtx)) + +extern void r300DestroyContext(__DRIcontextPrivate * driContextPriv); +extern GLboolean r300CreateContext(const __GLcontextModes * glVisual, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate); + +#endif /* __R300_CONTEXT_H__ */ diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.c b/src/mesa/drivers/dri/r300/r300_ioctl.c new file mode 100644 index 00000000000..a59b1efc9ea --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_ioctl.c @@ -0,0 +1,236 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Keith Whitwell + * Nicolai Haehnle + */ + +#include +#include + +#include "glheader.h" +#include "imports.h" +#include "macros.h" +#include "context.h" +#include "swrast/swrast.h" + +#include "r300_context.h" +#include "radeon_ioctl.h" +#include "r300_ioctl.h" +#include "r300_cmdbuf.h" +#include "r300_state.h" +#include "r300_program.h" +#include "radeon_reg.h" + +#include "vblank.h" + + +static void r300ClearColorBuffer(r300ContextPtr r300, int buffer) +{ + GLcontext* ctx = r300->radeon.glCtx; + __DRIdrawablePrivate *dPriv = r300->radeon.dri.drawable; + int i; + GLuint cboffset, cbpitch; + drm_r300_cmd_header_t* cmd; + + if (RADEON_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s: %s buffer\n", __FUNCTION__, + buffer ? "back" : "front"); + + if (buffer) { + cboffset = r300->radeon.radeonScreen->backOffset; + cbpitch = r300->radeon.radeonScreen->backPitch; + } else { + cboffset = r300->radeon.radeonScreen->frontOffset; + cbpitch = r300->radeon.radeonScreen->frontPitch; + } + cboffset = r300->radeon.radeonScreen->fbLocation; + + R300_STATECHANGE(r300, vpt); + r300->hw.vpt.cmd[R300_VPT_XSCALE] = r300PackFloat32(1.0); + r300->hw.vpt.cmd[R300_VPT_XOFFSET] = r300PackFloat32(dPriv->x); + r300->hw.vpt.cmd[R300_VPT_YSCALE] = r300PackFloat32(1.0); + r300->hw.vpt.cmd[R300_VPT_YOFFSET] = r300PackFloat32(dPriv->y); + r300->hw.vpt.cmd[R300_VPT_ZSCALE] = r300PackFloat32(1.0); + r300->hw.vpt.cmd[R300_VPT_ZOFFSET] = r300PackFloat32(0.0); + + R300_STATECHANGE(r300, cb); + r300->hw.cb.cmd[R300_CB_OFFSET] = cboffset; + r300->hw.cb.cmd[R300_CB_PITCH] = cbpitch | R300_COLOR_UNKNOWN_22_23; + + R300_STATECHANGE(r300, unk221C); + r300->hw.unk221C.cmd[1] = R300_221C_CLEAR; + + R300_STATECHANGE(r300, ri); + for(i = 1; i <= 8; ++i) + r300->hw.ri.cmd[i] = R300_RS_INTERP_USED; + + R300_STATECHANGE(r300, rr); + ((drm_r300_cmd_header_t*)r300->hw.rr.cmd)->unchecked_state.count = 1; + r300->hw.rr.cmd[1] = 0x00004000; + + R300_STATECHANGE(r300, fp); + r300->hw.fp.cmd[R300_FP_CNTL0] = 0; /* 1 pass, no textures */ + r300->hw.fp.cmd[R300_FP_CNTL1] = 0; /* no temporaries */ + r300->hw.fp.cmd[R300_FP_CNTL2] = 0; /* no offset, one ALU instr */ + r300->hw.fp.cmd[R300_FP_NODE0] = 0; + r300->hw.fp.cmd[R300_FP_NODE1] = 0; + r300->hw.fp.cmd[R300_FP_NODE2] = 0; + r300->hw.fp.cmd[R300_FP_NODE3] = R300_PFS_NODE_LAST_NODE; + + R300_STATECHANGE(r300, fpi[0]); + R300_STATECHANGE(r300, fpi[1]); + R300_STATECHANGE(r300, fpi[2]); + R300_STATECHANGE(r300, fpi[3]); + ((drm_r300_cmd_header_t*)r300->hw.fpi[0].cmd)->unchecked_state.count = 1; + ((drm_r300_cmd_header_t*)r300->hw.fpi[1].cmd)->unchecked_state.count = 1; + ((drm_r300_cmd_header_t*)r300->hw.fpi[2].cmd)->unchecked_state.count = 1; + ((drm_r300_cmd_header_t*)r300->hw.fpi[3].cmd)->unchecked_state.count = 1; + + /* MOV o0, t0 */ + r300->hw.fpi[0].cmd[1] = FP_INSTRC(MAD, FP_ARGC(SRC0C_XYZ), FP_ARGC(ONE), FP_ARGC(ZERO)); + r300->hw.fpi[1].cmd[1] = FP_SELC(0,NO,XYZ,FP_TMP(0),0,0); + r300->hw.fpi[2].cmd[1] = FP_INSTRA(MAD, FP_ARGA(SRC0A), FP_ARGA(ONE), FP_ARGA(ZERO)); + r300->hw.fpi[3].cmd[1] = FP_SELA(0,NO,W,FP_TMP(0),0,0); + + R300_STATECHANGE(r300, pvs); + r300->hw.pvs.cmd[R300_PVS_CNTL_1] = + (0 << R300_PVS_CNTL_1_PROGRAM_START_SHIFT) | + (0 << R300_PVS_CNTL_1_UNKNOWN_SHIFT) | + (1 << R300_PVS_CNTL_1_PROGRAM_END_SHIFT); + r300->hw.pvs.cmd[R300_PVS_CNTL_2] = 0; /* no parameters */ + r300->hw.pvs.cmd[R300_PVS_CNTL_3] = + (1 << R300_PVS_CNTL_3_PROGRAM_UNKNOWN_SHIFT); + + R300_STATECHANGE(r300, vpi); + ((drm_r300_cmd_header_t*)r300->hw.vpi.cmd)->unchecked_state.count = 8; + + /* MOV o0, i0; */ + r300->hw.vpi.cmd[1] = VP_OUT(ADD,OUT,0,XYZW); + r300->hw.vpi.cmd[2] = VP_IN(IN,0); + r300->hw.vpi.cmd[3] = VP_ZERO(); + r300->hw.vpi.cmd[4] = 0; + + /* MOV o1, i1; */ + r300->hw.vpi.cmd[5] = VP_OUT(ADD,OUT,1,XYZW); + r300->hw.vpi.cmd[6] = VP_IN(IN,1); + r300->hw.vpi.cmd[7] = VP_ZERO(); + r300->hw.vpi.cmd[8] = 0; + + /* Make sure we have enough space */ + r300EnsureCmdBufSpace(r300, r300->hw.max_state_size + 9, __FUNCTION__); + + r300EmitState(r300); + + cmd = (drm_r300_cmd_header_t*)r300AllocCmdBuf(r300, 9, __FUNCTION__); + cmd[0].packet3.cmd_type = R300_CMD_PACKET3; + cmd[0].packet3.packet = R300_CMD_PACKET3_CLEAR; + cmd[1].u = r300PackFloat32(dPriv->w / 2.0); /* my guess is as good as yours */ + cmd[2].u = r300PackFloat32(dPriv->h / 2.0); + cmd[3].u = r300PackFloat32(0.0); + cmd[4].u = r300PackFloat32(1.0); + cmd[5].u = r300PackFloat32(ctx->Color.ClearColor[0]); + cmd[6].u = r300PackFloat32(ctx->Color.ClearColor[1]); + cmd[7].u = r300PackFloat32(ctx->Color.ClearColor[2]); + cmd[8].u = r300PackFloat32(ctx->Color.ClearColor[3]); +} + + +/** + * Buffer clear + */ +static void r300Clear(GLcontext * ctx, GLbitfield mask, GLboolean all, + GLint cx, GLint cy, GLint cw, GLint ch) +{ + r300ContextPtr r300 = R300_CONTEXT(ctx); + __DRIdrawablePrivate *dPriv = r300->radeon.dri.drawable; + int flags = 0; + int swapped; + + if (RADEON_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s: all=%d cx=%d cy=%d cw=%d ch=%d\n", + __FUNCTION__, all, cx, cy, cw, ch); + + { + LOCK_HARDWARE(&r300->radeon); + UNLOCK_HARDWARE(&r300->radeon); + if (dPriv->numClipRects == 0) + return; + } + + if (mask & DD_FRONT_LEFT_BIT) { + flags |= DD_FRONT_LEFT_BIT; + mask &= ~DD_FRONT_LEFT_BIT; + } + + if (mask & DD_BACK_LEFT_BIT) { + flags |= DD_BACK_LEFT_BIT; + mask &= ~DD_BACK_LEFT_BIT; + } + + if (mask) { + if (RADEON_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "%s: swrast clear, mask: %x\n", + __FUNCTION__, mask); + _swrast_Clear(ctx, mask, all, cx, cy, cw, ch); + } + + swapped = r300->radeon.doPageFlip && (r300->radeon.sarea->pfCurrentPage == 1); + + if (flags & DD_FRONT_LEFT_BIT) + r300ClearColorBuffer(r300, swapped); + + if (flags & DD_BACK_LEFT_BIT) + r300ClearColorBuffer(r300, swapped ^ 1); + + /* Recalculate the hardware set. This could be done more efficiently, + * but do keep it like this for now. + */ + r300ResetHwState(r300); +} + +void r300Flush(GLcontext * ctx) +{ + r300ContextPtr r300 = R300_CONTEXT(ctx); + + if (RADEON_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s\n", __FUNCTION__); + + if (r300->cmdbuf.count_used > r300->cmdbuf.count_reemit) + r300FlushCmdBuf(r300, __FUNCTION__); +} + +void r300InitIoctlFuncs(struct dd_function_table *functions) +{ + functions->Clear = r300Clear; + functions->Finish = radeonFinish; + functions->Flush = r300Flush; +} diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.h b/src/mesa/drivers/dri/r300/r300_ioctl.h new file mode 100644 index 00000000000..26ca07cce29 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_ioctl.h @@ -0,0 +1,42 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Keith Whitwell + * Nicolai Haehnle + */ + +#ifndef __R300_IOCTL_H__ +#define __R300_IOCTL_H__ + +extern void r300Flush(GLcontext * ctx); +extern void r300InitIoctlFuncs(struct dd_function_table *functions); + +#endif /* __R300_IOCTL_H__ */ diff --git a/src/mesa/drivers/dri/r300/r300_program.h b/src/mesa/drivers/dri/r300/r300_program.h new file mode 100644 index 00000000000..b78ac17d223 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_program.h @@ -0,0 +1,148 @@ +/* +Copyright (C) 2004 Nicolai Haehnle. All Rights Reserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Nicolai Haehnle + */ + +#ifndef __R300_PROGRAM_H__ +#define __R300_PROGRAM_H__ + +#include "r300_reg.h" + +/** + * Vertex program helper macros + */ + +/* Produce out dword */ +#define VP_OUTCLASS_TMP R300_VPI_OUT_REG_CLASS_TEMPORARY +#define VP_OUTCLASS_OUT R300_VPI_OUT_REG_CLASS_RESULT + +#define VP_OUTMASK_X R300_VPI_OUT_WRITE_X +#define VP_OUTMASK_Y R300_VPI_OUT_WRITE_Y +#define VP_OUTMASK_Z R300_VPI_OUT_WRITE_Z +#define VP_OUTMASK_W R300_VPI_OUT_WRITE_W +#define VP_OUTMASK_XY (VP_OUTMASK_X|VP_OUTMASK_Y) +#define VP_OUTMASK_XZ (VP_OUTMASK_X|VP_OUTMASK_Z) +#define VP_OUTMASK_XW (VP_OUTMASK_X|VP_OUTMASK_W) +#define VP_OUTMASK_XYZ (VP_OUTMASK_XY|VP_OUTMASK_Z) +#define VP_OUTMASK_XYW (VP_OUTMASK_XY|VP_OUTMASK_W) +#define VP_OUTMASK_XZW (VP_OUTMASK_XZ|VP_OUTMASK_W) +#define VP_OUTMASK_XYZW (VP_OUTMASK_XYZ|VP_OUTMASK_W) +#define VP_OUTMASK_YZ (VP_OUTMASK_Y|VP_OUTMASK_Z) +#define VP_OUTMASK_YW (VP_OUTMASK_Y|VP_OUTMASK_W) +#define VP_OUTMASK_YZW (VP_OUTMASK_YZ|VP_OUTMASK_W) +#define VP_OUTMASK_ZW (VP_OUTMASK_Z|VP_OUTMASK_W) + +#define VP_OUT(instr,outclass,outidx,outmask) \ + (R300_VPI_OUT_OP_##instr | \ + ((outidx) << R300_VPI_OUT_REG_INDEX_SHIFT) | \ + VP_OUTCLASS_##outclass | \ + VP_OUTMASK_##outmask) + +/* Produce in dword */ +#define VP_INCLASS_TMP R300_VPI_IN_REG_CLASS_TEMPORARY +#define VP_INCLASS_IN R300_VPI_IN_REG_CLASS_ATTRIBUTE +#define VP_INCLASS_CONST R300_VPI_IN_REG_CLASS_PARAMETER + +#define VP_IN(class,idx) \ + (((idx) << R300_VPI_IN_REG_INDEX_SHIFT) | \ + VP_INCLASS_##class | \ + (R300_VPI_IN_SELECT_X << R300_VPI_IN_X_SHIFT) | \ + (R300_VPI_IN_SELECT_Y << R300_VPI_IN_Y_SHIFT) | \ + (R300_VPI_IN_SELECT_Z << R300_VPI_IN_Z_SHIFT) | \ + (R300_VPI_IN_SELECT_W << R300_VPI_IN_W_SHIFT)) +#define VP_ZERO() \ + ((R300_VPI_IN_SELECT_ZERO << R300_VPI_IN_X_SHIFT) | \ + (R300_VPI_IN_SELECT_ZERO << R300_VPI_IN_Y_SHIFT) | \ + (R300_VPI_IN_SELECT_ZERO << R300_VPI_IN_Z_SHIFT) | \ + (R300_VPI_IN_SELECT_ZERO << R300_VPI_IN_W_SHIFT)) +#define VP_ONE() \ + ((R300_VPI_IN_SELECT_ONE << R300_VPI_IN_X_SHIFT) | \ + (R300_VPI_IN_SELECT_ONE << R300_VPI_IN_Y_SHIFT) | \ + (R300_VPI_IN_SELECT_ONE << R300_VPI_IN_Z_SHIFT) | \ + (R300_VPI_IN_SELECT_ONE << R300_VPI_IN_W_SHIFT)) + +#define VP_NEG(in,comp) ((in) ^ (R300_VPI_IN_NEG_##comp)) +#define VP_NEGALL(in,comp) VP_NEG(VP_NEG(VP_NEG(VP_NEG((in),X),Y),Z),W) + +/** + * Fragment program helper macros + */ + +/* Produce unshifted source selectors */ +#define FP_TMP(idx) (idx) +#define FP_CONST(idx) ((idx) | (1 << 5)) + +/* Produce source/dest selector dword */ +#define FP_SELC_MASK_NO 0 +#define FP_SELC_MASK_X 1 +#define FP_SELC_MASK_Y 2 +#define FP_SELC_MASK_XY 3 +#define FP_SELC_MASK_Z 4 +#define FP_SELC_MASK_XZ 5 +#define FP_SELC_MASK_YZ 6 +#define FP_SELC_MASK_XYZ 7 + +#define FP_SELC(destidx,regmask,outmask,src0,src1,src2) \ + (((destidx) << R300_FPI1_DSTC_SHIFT) | \ + (FP_SELC_MASK_##regmask << 23) | \ + (FP_SELC_MASK_##outmask << 26) | \ + ((src0) << R300_FPI1_SRC0C_SHIFT) | \ + ((src1) << R300_FPI1_SRC1C_SHIFT) | \ + ((src2) << R300_FPI1_SRC2C_SHIFT)) + +#define FP_SELA_MASK_NO 0 +#define FP_SELA_MASK_W 1 + +#define FP_SELA(destidx,regmask,outmask,src0,src1,src2) \ + (((destidx) << R300_FPI3_DSTA_SHIFT) | \ + (FP_SELA_MASK_##regmask << 23) | \ + (FP_SELA_MASK_##outmask << 24) | \ + ((src0) << R300_FPI3_SRC0A_SHIFT) | \ + ((src1) << R300_FPI3_SRC1A_SHIFT) | \ + ((src2) << R300_FPI3_SRC2A_SHIFT)) + +/* Produce unshifted argument selectors */ +#define FP_ARGC(source) R300_FPI0_ARGC_##source +#define FP_ARGA(source) R300_FPI2_ARGA_##source +#define FP_ABS(arg) ((arg) | (1 << 6)) +#define FP_NEG(arg) ((arg) ^ (1 << 5)) + +/* Produce instruction dword */ +#define FP_INSTRC(opcode,arg0,arg1,arg2) \ + (R300_FPI0_OUTC_##opcode | \ + ((arg0) << R300_FPI0_ARG0C_SHIFT) | \ + ((arg1) << R300_FPI0_ARG1C_SHIFT) | \ + ((arg2) << R300_FPI0_ARG2C_SHIFT)) + +#define FP_INSTRA(opcode,arg0,arg1,arg2) \ + (R300_FPI2_OUTA_##opcode | \ + ((arg0) << R300_FPI2_ARG0A_SHIFT) | \ + ((arg1) << R300_FPI2_ARG1A_SHIFT) | \ + ((arg2) << R300_FPI2_ARG2A_SHIFT)) + +#endif /* __R300_PROGRAM_H__ */ diff --git a/src/mesa/drivers/dri/r300/r300_reg.h b/src/mesa/drivers/dri/r300/r300_reg.h new file mode 100644 index 00000000000..90c5158b8b0 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_reg.h @@ -0,0 +1,847 @@ +#ifndef _R300_REG_H +#define _R300_REG_H + +/* +This file contains registers and constants for the R300. They have been +found mostly by examining command buffers captured using glxtest, as well +as by extrapolating some known registers and constants from the R200. + +I am fairly certain that they are correct unless stated otherwise in comments. +*/ + +#define R300_SE_VPORT_XSCALE 0x1D98 +#define R300_SE_VPORT_XOFFSET 0x1D9C +#define R300_SE_VPORT_YSCALE 0x1DA0 +#define R300_SE_VPORT_YOFFSET 0x1DA4 +#define R300_SE_VPORT_ZSCALE 0x1DA8 +#define R300_SE_VPORT_ZOFFSET 0x1DAC + + +// BEGIN: Wild guesses +#define R300_VAP_OUTPUT_VTX_FMT_0 0x2090 +# define R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT (1<<0) +# define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_PRESENT (1<<1) +# define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT (1<<2) // GUESS +# define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT (1<<3) // GUESS +# define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT (1<<4) // GUESS +# define R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT (1<<16) // GUESS + +#define R300_VAP_OUTPUT_VTX_FMT_1 0x2094 +# define R300_VAP_OUTPUT_VTX_FMT_1__TEX_0_COMP_CNT_SHIFT 0 +# define R300_VAP_OUTPUT_VTX_FMT_1__TEX_1_COMP_CNT_SHIFT 3 +# define R300_VAP_OUTPUT_VTX_FMT_1__TEX_2_COMP_CNT_SHIFT 6 +# define R300_VAP_OUTPUT_VTX_FMT_1__TEX_3_COMP_CNT_SHIFT 9 +# define R300_VAP_OUTPUT_VTX_FMT_1__TEX_4_COMP_CNT_SHIFT 12 +# define R300_VAP_OUTPUT_VTX_FMT_1__TEX_5_COMP_CNT_SHIFT 15 +# define R300_VAP_OUTPUT_VTX_FMT_1__TEX_6_COMP_CNT_SHIFT 18 +# define R300_VAP_OUTPUT_VTX_FMT_1__TEX_7_COMP_CNT_SHIFT 21 +// END + +// BEGIN: Vertex data assembly - lots of uncertainties +/* gap */ +// Where do we get our vertex data? +// +// Vertex data either comes either from immediate mode registers or from +// vertex arrays. +// There appears to be no mixed mode (though we can force the pitch of +// vertex arrays to 0, effectively reusing the same element over and over +// again). +// +// Immediate mode is controlled by the INPUT_CNTL registers. I am not sure +// if these registers influence vertex array processing. +// +// Vertex arrays are controlled via the 3D_LOAD_VBPNTR packet3. +// +// In both cases, vertex attributes are then passed through INPUT_ROUTE. + +// Beginning with INPUT_ROUTE_0_0 is a list of WORDs that route vertex data +// into the vertex processor's input registers. +// The first word routes the first input, the second word the second, etc. +// The corresponding input is routed into the register with the given index. +// The list is ended by a word with INPUT_ROUTE_END set. +// +// Always set COMPONENTS_4 in immediate mode. +#define R300_VAP_INPUT_ROUTE_0_0 0x2150 +# define R300_INPUT_ROUTE_COMPONENTS_1 (0 << 0) +# define R300_INPUT_ROUTE_COMPONENTS_2 (1 << 0) +# define R300_INPUT_ROUTE_COMPONENTS_3 (2 << 0) +# define R300_INPUT_ROUTE_COMPONENTS_4 (3 << 0) +# define R300_INPUT_ROUTE_COMPONENTS_RGBA (4 << 0) // GUESS +# define R300_VAP_INPUT_ROUTE_IDX_SHIFT 8 +# define R300_VAP_INPUT_ROUTE_IDX_MASK (31 << 8) // GUESS +# define R300_VAP_INPUT_ROUTE_END (1 << 13) +# define R300_INPUT_ROUTE_IMMEDIATE_MODE (0 << 14) // GUESS +# define R300_INPUT_ROUTE_FLOAT (1 << 14) // GUESS +# define R300_INPUT_ROUTE_UNSIGNED_BYTE (2 << 14) // GUESS +# define R300_INPUT_ROUTE_FLOAT_COLOR (3 << 14) // GUESS +#define R300_VAP_INPUT_ROUTE_0_1 0x2154 +#define R300_VAP_INPUT_ROUTE_0_2 0x2158 +#define R300_VAP_INPUT_ROUTE_0_3 0x215C + +/* gap */ +// Notes: +// - always set up to produce at least two attributes: +// if vertex program uses only position, fglrx will set normal, too +// - INPUT_CNTL_0_COLOR and INPUT_CNTL_COLOR bits are always equal +#define R300_VAP_INPUT_CNTL_0 0x2180 +# define R300_INPUT_CNTL_0_COLOR 0x00000001 +#define R300_VAP_INPUT_CNTL_1 0x2184 +# define R300_INPUT_CNTL_POS 0x00000001 +# define R300_INPUT_CNTL_NORMAL 0x00000002 +# define R300_INPUT_CNTL_COLOR 0x00000004 +# define R300_INPUT_CNTL_TC0 0x00000400 +# define R300_INPUT_CNTL_TC1 0x00000800 +# define R300_INPUT_CNTL_TC2 0x00001000 // GUESS +# define R300_INPUT_CNTL_TC3 0x00002000 // GUESS +# define R300_INPUT_CNTL_TC4 0x00004000 // GUESS +# define R300_INPUT_CNTL_TC5 0x00008000 // GUESS +# define R300_INPUT_CNTL_TC6 0x00010000 // GUESS +# define R300_INPUT_CNTL_TC7 0x00020000 // GUESS + +/* gap */ +// Words parallel to INPUT_ROUTE_0; All words that are active in INPUT_ROUTE_0 +// are set to a swizzling bit pattern, other words are 0. +// +// In immediate mode, the pattern is always set to xyzw. In vertex array +// mode, the swizzling pattern is e.g. used to set zw components in texture +// coordinates with only tweo components. +#define R300_VAP_INPUT_ROUTE_1_0 0x21E0 +# define R300_INPUT_ROUTE_SELECT_X 0 +# define R300_INPUT_ROUTE_SELECT_Y 1 +# define R300_INPUT_ROUTE_SELECT_Z 2 +# define R300_INPUT_ROUTE_SELECT_W 3 +# define R300_INPUT_ROUTE_SELECT_ZERO 4 +# define R300_INPUT_ROUTE_SELECT_ONE 5 +# define R300_INPUT_ROUTE_SELECT_MASK 7 +# define R300_INPUT_ROUTE_X_SHIFT 0 +# define R300_INPUT_ROUTE_Y_SHIFT 3 +# define R300_INPUT_ROUTE_Z_SHIFT 6 +# define R300_INPUT_ROUTE_W_SHIFT 9 +# define R300_INPUT_ROUTE_ENABLE (15 << 12) +#define R300_VAP_INPUT_ROUTE_1_1 0x21E4 +#define R300_VAP_INPUT_ROUTE_1_2 0x21E8 +#define R300_VAP_INPUT_ROUTE_1_3 0x21EC + +// END + +/* gap */ +// BEGIN: Upload vertex program and data +// The programmable vertex shader unit has a memory bank of unknown size +// that can be written to in 16 byte units by writing the address into +// UPLOAD_ADDRESS, followed by data in UPLOAD_DATA (multiples of 4 DWORDs). +// +// Pointers into the memory bank are always in multiples of 16 bytes. +// +// The memory bank is divided into areas with fixed meaning. +// +// Starting at address UPLOAD_PROGRAM: Vertex program instructions. +// Native limits reported by drivers from ATI suggest size 256 (i.e. 4KB), +// whereas the difference between known addresses suggests size 512. +// +// Starting at address UPLOAD_PARAMETERS: Vertex program parameters. +// Native reported limits and the VPI layout suggest size 256, whereas +// difference between known addresses suggests size 512. +// +// Multiple vertex programs and parameter sets can be loaded at once, +// which could explain the size discrepancy. +#define R300_VAP_PVS_UPLOAD_ADDRESS 0x2200 +# define R300_PVS_UPLOAD_PROGRAM 0x00000000 +# define R300_PVS_UPLOAD_PARAMETERS 0x00000200 +# define R300_PVS_UPLOAD_UNKNOWN 0x00000400 +/* gap */ +#define R300_VAP_PVS_UPLOAD_DATA 0x2208 +// END + +/* gap */ +// I do not know the purpose of this register. However, I do know that +// it is set to 221C_CLEAR for clear operations and to 221C_NORMAL +// for normal rendering. +#define R300_VAP_UNKNOWN_221C 0x221C +# define R300_221C_NORMAL 0x00000000 +# define R300_221C_CLEAR 0x0001C000 + +/* gap */ +// Sometimes, END_OF_PKT and 0x2284=0 are the only commands sent between +// rendering commands and overwriting vertex program parameters. +// Therefore, I suspect writing zero to 0x2284 synchronizes the engine and +// avoids bugs caused by still running shaders reading bad data from memory. +#define R300_VAP_PVS_WAITIDLE 0x2284 // GUESS + +// Absolutely no clue what this register is about. +#define R300_VAP_UNKNOWN_2288 0x2288 +# define R300_2288_R300 0x00750000 // -- nh +# define R300_2288_RV350 0x0000FFFF // -- Vladimir + +/* gap */ +// Addresses are relative to the vertex program instruction area of the +// memory bank. PROGRAM_END points to the last instruction of the active +// program +// +// The meaning of the two UNKNOWN fields is obviously not known. However, +// experiments so far have shown that both *must* point to an instruction +// inside the vertex program, otherwise the GPU locks up. +// fglrx usually sets CNTL_3_UNKNOWN to the end of the program and +// CNTL_1_UNKNOWN somewhere in the middle, but the criteria are not clear. +#define R300_VAP_PVS_CNTL_1 0x22D0 +# define R300_PVS_CNTL_1_PROGRAM_START_SHIFT 0 +# define R300_PVS_CNTL_1_UNKNOWN_SHIFT 10 +# define R300_PVS_CNTL_1_PROGRAM_END_SHIFT 20 +// Addresses are relative the the vertex program parameters area. +#define R300_VAP_PVS_CNTL_2 0x22D4 +# define R300_PVS_CNTL_2_PARAM_OFFSET_SHIFT 0 +# define R300_PVS_CNTL_2_PARAM_COUNT_SHIFT 16 +#define R300_VAP_PVS_CNTL_3 0x22D8 +# define R300_PVS_CNTL_3_PROGRAM_UNKNOWN_SHIFT 10 + +// The entire range from 0x2300 to 0x2AC inclusive seems to be used for +// immediate vertices +#define R300_VAP_VTX_COLOR_R 0x2464 +#define R300_VAP_VTX_COLOR_G 0x2468 +#define R300_VAP_VTX_COLOR_B 0x246C +#define R300_VAP_VTX_POS_0_X_1 0x2490 // used for glVertex2*() +#define R300_VAP_VTX_POS_0_Y_1 0x2494 +#define R300_VAP_VTX_COLOR_PKD 0x249C // RGBA +#define R300_VAP_VTX_POS_0_X_2 0x24A0 // used for glVertex3*() +#define R300_VAP_VTX_POS_0_Y_2 0x24A4 +#define R300_VAP_VTX_POS_0_Z_2 0x24A8 +#define R300_VAP_VTX_END_OF_PKT 0x24AC // write 0 to indicate end of packet? + +/* gap */ +// BEGIN: !unverified! +#define R300_GB_TILE_CONFIG 0x4018 +#define R300_GB_TILE_ENABLE (1 << 0) +#define R300_GB_TILE_PIPE_COUNT_R300 (0 << 1) +#define R300_GB_TILE_PIPE_COUNT_RV300 (3 << 1) +#define R300_GB_TILE_SIZE_8 (0 << 4) +#define R300_GB_TILE_SIZE_16 (1 << 4) +#define R300_GB_TILE_SIZE_32 (2 << 4) +#define R300_GB_SUPER_SIZE_1 (0 << 6) +#define R300_GB_SUPER_SIZE_2 (1 << 6) +#define R300_GB_SUPER_SIZE_4 (2 << 6) +#define R300_GB_SUPER_SIZE_8 (3 << 6) +#define R300_GB_SUPER_SIZE_16 (4 << 6) +#define R300_GB_SUPER_SIZE_32 (5 << 6) +#define R300_GB_SUPER_SIZE_64 (6 << 6) +#define R300_GB_SUPER_SIZE_128 (7 << 6) +#define R300_GB_SUPER_X_SHIFT 9 // 3 bits wide +#define R300_GB_SUPER_Y_SHIFT 12 // 3 bits wide +#define R300_GB_SUPER_TILE_A (0 << 15) +#define R300_GB_SUPER_TILE_B (1 << 15) +#define R300_GB_SUBPIXEL_1_12 (0 << 16) +#define R300_GB_SUBPIXEL_1_16 (1 << 16) +// END + +/* gap */ +// The upper enable bits are guessed, based on fglrx reported limits. +#define R300_TX_ENABLE 0x4104 +# define R300_TX_ENABLE_0 (1 << 0) +# define R300_TX_ENABLE_1 (1 << 1) +# define R300_TX_ENABLE_2 (1 << 2) +# define R300_TX_ENABLE_3 (1 << 3) +# define R300_TX_ENABLE_4 (1 << 4) +# define R300_TX_ENABLE_5 (1 << 5) +# define R300_TX_ENABLE_6 (1 << 6) +# define R300_TX_ENABLE_7 (1 << 7) +# define R300_TX_ENABLE_8 (1 << 8) +# define R300_TX_ENABLE_9 (1 << 9) +# define R300_TX_ENABLE_10 (1 << 10) +# define R300_TX_ENABLE_11 (1 << 11) +# define R300_TX_ENABLE_12 (1 << 12) +# define R300_TX_ENABLE_13 (1 << 13) +# define R300_TX_ENABLE_14 (1 << 14) +# define R300_TX_ENABLE_15 (1 << 15) + +// No idea what the purpose is, but it is set to 421C_CLEAR just before +// issuing the clear command and then reset to 421C_NORMAL afterwards. +#define R300_UNKNOWN_421C 0x421C +# define R300_421C_NORMAL 0x00060006 +# define R300_421C_CLEAR 0x0F000B40 + +// BEGIN: Rasterization / Interpolators - many guesses +// So far, 0_UNKOWN_7 has always been set. +// 0_UNKNOWN_18 has always been set except for clear operations. +// TC_CNT is the number of incoming texture coordinate sets (i.e. it depends +// on the vertex program, *not* the fragment program) +#define R300_RS_CNTL_0 0x4300 +# define R300_RS_CNTL_TC_CNT_SHIFT 2 +# define R300_RS_CNTL_TC_CNT_MASK (7 << 2) +# define R300_RS_CNTL_0_UNKNOWN_7 (1 << 7) +# define R300_RS_CNTL_0_UNKNOWN_18 (1 << 18) +// Guess: RS_CNTL_1 holds the index of the highest used RS_ROUTE_n register. +#define R300_RS_CNTL_1 0x4304 + +/* gap */ +// Only used for texture coordinates (color seems to be always interpolated). +// Use the source field to route texture coordinate input from the vertex program +// to the desired interpolator. Note that the source field is relative to the +// outputs the vertex program *actually* writes. If a vertex program only writes +// texcoord[1], this will be source index 0. +// Set INTERP_USED on all interpolators that produce data used by the +// fragment program. INTERP_USED looks like a swizzling mask, but +// I haven't seen it used that way. +// +// Note: The _UNKNOWN constants are always set in their respective register. +// I don't know if this is necessary. +#define R300_RS_INTERP_0 0x4310 +#define R300_RS_INTERP_1 0x4314 +# define R300_RS_INTERP_1_UNKNOWN 0x40 +#define R300_RS_INTERP_2 0x4318 +# define R300_RS_INTERP_2_UNKNOWN 0x80 +#define R300_RS_INTERP_3 0x431C +# define R300_RS_INTERP_3_UNKNOWN 0xC0 +#define R300_RS_INTERP_4 0x4320 +#define R300_RS_INTERP_5 0x4324 +#define R300_RS_INTERP_6 0x4328 +#define R300_RS_INTERP_7 0x432C +# define R300_RS_INTERP_SRC_SHIFT 2 +# define R300_RS_INTERP_SRC_MASK (7 << 2) +# define R300_RS_INTERP_USED 0x00D10000 + +// These DWORDs control how vertex data is routed into fragment program +// registers, after interpolators. +#define R300_RS_ROUTE_0 0x4330 +#define R300_RS_ROUTE_1 0x4334 +#define R300_RS_ROUTE_2 0x4338 +#define R300_RS_ROUTE_3 0x433C // GUESS +#define R300_RS_ROUTE_4 0x4340 // GUESS +#define R300_RS_ROUTE_5 0x4344 // GUESS +#define R300_RS_ROUTE_6 0x4348 // GUESS +#define R300_RS_ROUTE_7 0x434C // GUESS +# define R300_RS_ROUTE_SOURCE_INTERP_0 0 +# define R300_RS_ROUTE_SOURCE_INTERP_1 1 +# define R300_RS_ROUTE_SOURCE_INTERP_2 2 +# define R300_RS_ROUTE_SOURCE_INTERP_3 3 +# define R300_RS_ROUTE_SOURCE_INTERP_4 4 +# define R300_RS_ROUTE_SOURCE_INTERP_5 5 // GUESS +# define R300_RS_ROUTE_SOURCE_INTERP_6 6 // GUESS +# define R300_RS_ROUTE_SOURCE_INTERP_7 7 // GUESS +# define R300_RS_ROUTE_ENABLE (1 << 3) // GUESS +# define R300_RS_ROUTE_DEST_SHIFT 6 +# define R300_RS_ROUTE_DEST_MASK (31 << 6) // GUESS + +// Special handling for color: When the fragment program uses color, +// the ROUTE_0_COLOR bit is set and ROUTE_0_COLOR_DEST contains the +// color register index. +# define R300_RS_ROUTE_0_COLOR (1 << 14) +# define R300_RS_ROUTE_0_COLOR_DEST_SHIFT (1 << 17) +# define R300_RS_ROUTE_0_COLOR_DEST_MASK (31 << 6) // GUESS +// END + +// BEGIN: Texture specification +// The texture specification dwords are grouped by meaning and not by texture unit. +// This means that e.g. the offset for texture image unit N is found in register +// TX_OFFSET_0 + (4*N) +#define R300_TX_FILTER_0 0x4400 +# define R300_TX_REPEAT 0 +# define R300_TX_CLAMP_TO_EDGE 1 +# define R300_TX_CLAMP 2 +# define R300_TX_CLAMP_TO_BORDER 3 + +# define R300_TX_WRAP_S_SHIFT 1 +# define R300_TX_WRAP_S_MASK (3 << 1) +# define R300_TX_WRAP_T_SHIFT 4 +# define R300_TX_WRAP_T_MASK (3 << 4) +# define R300_TX_MAG_FILTER_NEAREST (1 << 9) +# define R300_TX_MAG_FILTER_LINEAR (2 << 9) +# define R300_TX_MAG_FILTER_MASK (3 << 9) +# define R300_TX_MIN_FILTER_NEAREST (1 << 11) +# define R300_TX_MIN_FILTER_LINEAR (2 << 11) +#define R300_TX_UNK1_0 0x4440 +#define R300_TX_SIZE_0 0x4480 +# define R300_TX_WIDTHMASK_SHIFT 0 +# define R300_TX_WIDTHMASK_MASK (2047 << 0) +# define R300_TX_HEIGHTMASK_SHIFT 11 +# define R300_TX_HEIGHTMASK_MASK (2047 << 11) +# define R300_TX_SIZE_SHIFT 26 // largest of width, height +# define R300_TX_SIZE_MASK (15 << 26) +#define R300_TX_FORMAT_0 0x44C0 +#define R300_TX_OFFSET_0 0x4540 +// BEGIN: Guess from R200 +# define R300_TXO_ENDIAN_NO_SWAP (0 << 0) +# define R300_TXO_ENDIAN_BYTE_SWAP (1 << 0) +# define R300_TXO_ENDIAN_WORD_SWAP (2 << 0) +# define R300_TXO_ENDIAN_HALFDW_SWAP (3 << 0) +# define R300_TXO_OFFSET_MASK 0xffffffe0 +# define R300_TXO_OFFSET_SHIFT 5 +// END +#define R300_TX_UNK4_0 0x4580 +#define R300_TX_UNK5_0 0x45C0 +// END + +// BEGIN: Fragment program instruction set +// Fragment programs are written directly into register space. +// There are separate instruction streams for texture instructions and ALU +// instructions. +// In order to synchronize these streams, the program is divided into up +// to 4 nodes. Each node begins with a number of TEX operations, followed +// by a number of ALU operations. +// The first node can have zero TEX ops, all subsequent nodes must have at least +// one TEX ops. +// All nodes must have at least one ALU op. +// +// The index of the last node is stored in PFS_CNTL_0: A value of 0 means +// 1 node, a value of 3 means 4 nodes. +// The total amount of instructions is defined in PFS_CNTL_2. The offsets are +// offsets into the respective instruction streams, while *_END points to the +// last instruction relative to this offset. +#define R300_PFS_CNTL_0 0x4600 +# define R300_PFS_CNTL_LAST_NODES_SHIFT 0 +# define R300_PFS_CNTL_LAST_NODES_MASK (3 << 0) +# define R300_PFS_CNTL_FIRST_NODE_HAS_TEX (1 << 3) +#define R300_PFS_CNTL_1 0x4604 +// There is an unshifted value here which has so far always been equal to the +// index of the highest used temporary register. +#define R300_PFS_CNTL_2 0x4608 +# define R300_PFS_CNTL_ALU_OFFSET_SHIFT 0 +# define R300_PFS_CNTL_ALU_OFFSET_MASK (63 << 0) +# define R300_PFS_CNTL_ALU_END_SHIFT 6 +# define R300_PFS_CNTL_ALU_END_MASK (63 << 0) +# define R300_PFS_CNTL_TEX_OFFSET_SHIFT 12 +# define R300_PFS_CNTL_TEX_OFFSET_MASK (31 << 12) // GUESS +# define R300_PFS_CNTL_TEX_END_SHIFT 18 +# define R300_PFS_CNTL_TEX_END_MASK (31 << 18) // GUESS + +/* gap */ +// Nodes are stored backwards. The last active node is always stored in +// PFS_NODE_3. +// Example: In a 2-node program, NODE_0 and NODE_1 are set to 0. The +// first node is stored in NODE_2, the second node is stored in NODE_3. +// +// Offsets are relative to the master offset from PFS_CNTL_2. +// LAST_NODE is set for the last node, and only for the last node. +#define R300_PFS_NODE_0 0x4610 +#define R300_PFS_NODE_1 0x4614 +#define R300_PFS_NODE_2 0x4618 +#define R300_PFS_NODE_3 0x461C +# define R300_PFS_NODE_ALU_OFFSET_SHIFT 0 +# define R300_PFS_NODE_ALU_OFFSET_MASK (63 << 0) +# define R300_PFS_NODE_ALU_END_SHIFT 6 +# define R300_PFS_NODE_ALU_END_MASK (63 << 6) +# define R300_PFS_NODE_TEX_OFFSET_SHIFT 12 +# define R300_PFS_NODE_TEX_OFFSET_MASK (31 << 12) +# define R300_PFS_NODE_TEX_END_SHIFT 17 +# define R300_PFS_NODE_TEX_END_MASK (31 << 17) +# define R300_PFS_NODE_LAST_NODE (1 << 22) + +// TEX +// As far as I can tell, texture instructions cannot write into output +// registers directly. A subsequent ALU instruction is always necessary, +// even if it's just MAD o0, r0, 1, 0 +#define R300_PFS_TEXI_0 0x4620 +# define R300_FPITX_SRC_SHIFT 0 +# define R300_FPITX_SRC_MASK (31 << 0) +# define R300_FPITX_SRC_CONST (1 << 5) // GUESS +# define R300_FPITX_DST_SHIFT 6 +# define R300_FPITX_DST_MASK (31 << 6) +# define R300_FPITX_IMAGE_SHIFT 11 +# define R300_FPITX_IMAGE_MASK (15 << 11) // GUESS based on layout and native limits + +// ALU +// The ALU instructions register blocks are enumerated according to the order +// in which fglrx. I assume there is space for 64 instructions, since +// each block has space for a maximum of 64 DWORDs, and this matches reported +// native limits. +// +// The basic functional block seems to be one MAD for each color and alpha, +// and an adder that adds all components after the MUL. +// - ADD, MUL, MAD etc.: use MAD with appropriate neutral operands +// - DP4: Use OUTC_DP4, OUTA_DP4 +// - DP3: Use OUTC_DP3, OUTA_DP4, appropriate alpha operands +// - DPH: Use OUTC_DP4, OUTA_DP4, appropriate alpha operands +// - CMP: If ARG2 < 0, return ARG1, else return ARG0 +// - FLR: use FRC+MAD +// - XPD: use MAD+MAD +// - SGE, SLT: use MAD+CMP +// - RSQ: use ABS modifier for argument +// - Use OUTC_REPL_ALPHA to write results of an alpha-only operation (e.g. RCP) +// into color register +// - apparently, there's no quick DST operation +// - fglrx set FPI2_UNKNOWN_31 on a "MAD fragment.color, tmp0, tmp1, tmp2" +// - fglrx set FPI2_UNKNOWN_31 on a "MAX r2, r1, c0" +// - fglrx once set FPI0_UNKNOWN_31 on a "FRC r1, r1" +// +// Operand selection +// First stage selects three sources from the available registers and +// constant parameters. This is defined in INSTR1 (color) and INSTR3 (alpha). +// fglrx sorts the three source fields: Registers before constants, +// lower indices before higher indices; I do not know whether this is necessary. +// fglrx fills unused sources with "read constant 0" +// According to specs, you cannot select more than two different constants. +// +// Second stage selects the operands from the sources. This is defined in +// INSTR0 (color) and INSTR2 (alpha). You can also select the special constants +// zero and one. +// Swizzling and negation happens in this stage, as well. +// +// Important: Color and alpha seem to be mostly separate, i.e. their sources +// selection appears to be fully independent (the register storage is probably +// physically split into a color and an alpha section). +// However (because of the apparent physical split), there is some interaction +// WRT swizzling. If, for example, you want to load an R component into an +// Alpha operand, this R component is taken from a *color* source, not from +// an alpha source. The corresponding register doesn't even have to appear in +// the alpha sources list. (I hope this alll makes sense to you) +// +// Destination selection +// The destination register index is in FPI1 (color) and FPI3 (alpha) together +// with enable bits. +// There are separate enable bits for writing into temporary registers +// (DSTC_REG_*/DSTA_REG) and and program output registers (DSTC_OUTPUT_*/DSTA_OUTPUT). +// You can write to both at once, or not write at all (the same index +// must be used for both). +// +// Note: There is a special form for LRP +// - Argument order is the same as in ARB_fragment_program. +// - Operation is MAD +// - ARG1 is set to ARGC_SRC1C_LRP/ARGC_SRC1A_LRP +// - Set FPI0/FPI2_SPECIAL_LRP +// Arbitrary LRP (including support for swizzling) requires vanilla MAD+MAD +#define R300_PFS_INSTR1_0 0x46C0 +# define R300_FPI1_SRC0C_SHIFT 0 +# define R300_FPI1_SRC0C_MASK (31 << 0) +# define R300_FPI1_SRC0C_CONST (1 << 5) +# define R300_FPI1_SRC1C_SHIFT 6 +# define R300_FPI1_SRC1C_MASK (31 << 6) +# define R300_FPI1_SRC1C_CONST (1 << 11) +# define R300_FPI1_SRC2C_SHIFT 12 +# define R300_FPI1_SRC2C_MASK (31 << 12) +# define R300_FPI1_SRC2C_CONST (1 << 17) +# define R300_FPI1_DSTC_SHIFT 18 +# define R300_FPI1_DSTC_MASK (31 << 18) +# define R300_FPI1_DSTC_REG_X (1 << 23) +# define R300_FPI1_DSTC_REG_Y (1 << 24) +# define R300_FPI1_DSTC_REG_Z (1 << 25) +# define R300_FPI1_DSTC_OUTPUT_X (1 << 26) +# define R300_FPI1_DSTC_OUTPUT_Y (1 << 27) +# define R300_FPI1_DSTC_OUTPUT_Z (1 << 28) + +#define R300_PFS_INSTR3_0 0x47C0 +# define R300_FPI3_SRC0A_SHIFT 0 +# define R300_FPI3_SRC0A_MASK (31 << 0) +# define R300_FPI3_SRC0A_CONST (1 << 5) +# define R300_FPI3_SRC1A_SHIFT 6 +# define R300_FPI3_SRC1A_MASK (31 << 6) +# define R300_FPI3_SRC1A_CONST (1 << 11) +# define R300_FPI3_SRC2A_SHIFT 12 +# define R300_FPI3_SRC2A_MASK (31 << 12) +# define R300_FPI3_SRC2A_CONST (1 << 17) +# define R300_FPI3_DSTA_SHIFT 18 +# define R300_FPI3_DSTA_MASK (31 << 18) +# define R300_FPI3_DSTA_REG (1 << 23) +# define R300_FPI3_DSTA_OUTPUT (1 << 24) + +#define R300_PFS_INSTR0_0 0x48C0 +# define R300_FPI0_ARGC_SRC0C_XYZ 0 +# define R300_FPI0_ARGC_SRC0C_XXX 1 +# define R300_FPI0_ARGC_SRC0C_YYY 2 +# define R300_FPI0_ARGC_SRC0C_ZZZ 3 +# define R300_FPI0_ARGC_SRC1C_XYZ 4 +# define R300_FPI0_ARGC_SRC1C_XXX 5 +# define R300_FPI0_ARGC_SRC1C_YYY 6 +# define R300_FPI0_ARGC_SRC1C_ZZZ 7 +# define R300_FPI0_ARGC_SRC2C_XYZ 8 +# define R300_FPI0_ARGC_SRC2C_XXX 9 +# define R300_FPI0_ARGC_SRC2C_YYY 10 +# define R300_FPI0_ARGC_SRC2C_ZZZ 11 +# define R300_FPI0_ARGC_SRC0A 12 +# define R300_FPI0_ARGC_SRC1A 13 +# define R300_FPI0_ARGC_SRC2A 14 +# define R300_FPI0_ARGC_SRC1C_LRP 15 +# define R300_FPI0_ARGC_ZERO 20 +# define R300_FPI0_ARGC_ONE 21 +# define R300_FPI0_ARGC_HALF 22 // GUESS +# define R300_FPI0_ARGC_SRC0C_YZX 23 +# define R300_FPI0_ARGC_SRC1C_YZX 24 +# define R300_FPI0_ARGC_SRC2C_YZX 25 +# define R300_FPI0_ARGC_SRC0C_ZXY 26 +# define R300_FPI0_ARGC_SRC1C_ZXY 27 +# define R300_FPI0_ARGC_SRC2C_ZXY 28 +# define R300_FPI0_ARGC_SRC0CA_WZY 29 +# define R300_FPI0_ARGC_SRC1CA_WZY 30 +# define R300_FPI0_ARGC_SRC2CA_WZY 31 + +# define R300_FPI0_ARG0C_SHIFT 0 +# define R300_FPI0_ARG0C_MASK (31 << 0) +# define R300_FPI0_ARG0C_NEG (1 << 5) +# define R300_FPI0_ARG0C_ABS (1 << 6) +# define R300_FPI0_ARG1C_SHIFT 7 +# define R300_FPI0_ARG1C_MASK (31 << 7) +# define R300_FPI0_ARG1C_NEG (1 << 12) +# define R300_FPI0_ARG1C_ABS (1 << 13) +# define R300_FPI0_ARG2C_SHIFT 14 +# define R300_FPI0_ARG2C_MASK (31 << 14) +# define R300_FPI0_ARG2C_NEG (1 << 19) +# define R300_FPI0_ARG2C_ABS (1 << 20) +# define R300_FPI0_SPECIAL_LRP (1 << 21) +# define R300_FPI0_OUTC_MAD (0 << 23) +# define R300_FPI0_OUTC_DP3 (1 << 23) +# define R300_FPI0_OUTC_DP4 (2 << 23) +# define R300_FPI0_OUTC_MIN (4 << 23) +# define R300_FPI0_OUTC_MAX (5 << 23) +# define R300_FPI0_OUTC_CMP (8 << 23) +# define R300_FPI0_OUTC_FRC (9 << 23) +# define R300_FPI0_OUTC_REPL_ALPHA (10 << 23) +# define R300_FPI0_OUTC_SAT (1 << 30) +# define R300_FPI0_UNKNOWN_31 (1 << 31) + +#define R300_PFS_INSTR2_0 0x49C0 +# define R300_FPI2_ARGA_SRC0C_X 0 +# define R300_FPI2_ARGA_SRC0C_Y 1 +# define R300_FPI2_ARGA_SRC0C_Z 2 +# define R300_FPI2_ARGA_SRC1C_X 3 +# define R300_FPI2_ARGA_SRC1C_Y 4 +# define R300_FPI2_ARGA_SRC1C_Z 5 +# define R300_FPI2_ARGA_SRC2C_X 6 +# define R300_FPI2_ARGA_SRC2C_Y 7 +# define R300_FPI2_ARGA_SRC2C_Z 8 +# define R300_FPI2_ARGA_SRC0A 9 +# define R300_FPI2_ARGA_SRC1A 10 +# define R300_FPI2_ARGA_SRC2A 11 +# define R300_FPI2_ARGA_SRC1A_LRP 15 +# define R300_FPI2_ARGA_ZERO 16 +# define R300_FPI2_ARGA_ONE 17 +# define R300_FPI2_ARGA_HALF 18 // GUESS + +# define R300_FPI2_ARG0A_SHIFT 0 +# define R300_FPI2_ARG0A_MASK (31 << 0) +# define R300_FPI2_ARG0A_NEG (1 << 5) +# define R300_FPI2_ARG1A_SHIFT 7 +# define R300_FPI2_ARG1A_MASK (31 << 7) +# define R300_FPI2_ARG1A_NEG (1 << 12) +# define R300_FPI2_ARG2A_SHIFT 14 +# define R300_FPI2_AEG2A_MASK (31 << 14) +# define R300_FPI2_ARG2A_NEG (1 << 19) +# define R300_FPI2_SPECIAL_LRP (1 << 21) +# define R300_FPI2_OUTA_MAD (0 << 23) +# define R300_FPI2_OUTA_DP4 (1 << 23) +# define R300_RPI2_OUTA_MIN (2 << 23) +# define R300_RPI2_OUTA_MAX (3 << 23) +# define R300_FPI2_OUTA_CMP (6 << 23) +# define R300_FPI2_OUTA_FRC (7 << 23) +# define R300_FPI2_OUTA_EX2 (8 << 23) +# define R300_FPI2_OUTA_LG2 (9 << 23) +# define R300_FPI2_OUTA_RCP (10 << 23) +# define R300_FPI2_OUTA_RSQ (11 << 23) +# define R300_FPI2_OUTA_SAT (1 << 30) +# define R300_FPI2_UNKNOWN_31 (1 << 31) +// END + +/* gap */ +#define R300_PP_ALPHA_TEST 0x4BD4 +# define R300_REF_ALPHA_MASK 0x000000ff +# define R300_ALPHA_TEST_FAIL (0 << 8) +# define R300_ALPHA_TEST_LESS (1 << 8) +# define R300_ALPHA_TEST_LEQUAL (2 << 8) +# define R300_ALPHA_TEST_EQUAL (3 << 8) +# define R300_ALPHA_TEST_GEQUAL (4 << 8) +# define R300_ALPHA_TEST_GREATER (5 << 8) +# define R300_ALPHA_TEST_NEQUAL (6 << 8) +# define R300_ALPHA_TEST_PASS (7 << 8) +# define R300_ALPHA_TEST_OP_MASK (7 << 8) +# define R300_ALPHA_TEST_ENABLE (1 << 11) + +/* gap */ +// Fragment program parameters in 7.16 floating point +#define R300_PFS_PARAM_0_X 0x4C00 +#define R300_PFS_PARAM_0_Y 0x4C04 +#define R300_PFS_PARAM_0_Z 0x4C08 +#define R300_PFS_PARAM_0_W 0x4C0C +// GUESS: PARAM_31 is last, based on native limits reported by fglrx +#define R300_PFS_PARAM_31_X 0x4DF0 +#define R300_PFS_PARAM_31_Y 0x4DF4 +#define R300_PFS_PARAM_31_Z 0x4DF8 +#define R300_PFS_PARAM_31_W 0x4DFC + +// Notes: +// - AFAIK fglrx always sets BLEND_UNKNOWN when blending is used in the application +// - AFAIK fglrx always sets BLEND_NO_SEPARATE when CBLEND and ABLEND are set to the same +// function (both registers are always set up completely in any case) +// - Most blend flags are simply copied from R200 and not tested yet +#define R300_RB3D_CBLEND 0x4E04 +#define R300_RB3D_ABLEND 0x4E08 + /* the following only appear in CBLEND */ +# define R300_BLEND_ENABLE (1 << 0) +# define R300_BLEND_UNKNOWN (3 << 1) +# define R300_BLEND_NO_SEPARATE (1 << 3) + /* the following are shared between CBLEND and ABLEND */ +# define R300_FCN_MASK (3 << 12) +# define R300_COMB_FCN_ADD_CLAMP (0 << 12) +# define R300_COMB_FCN_ADD_NOCLAMP (1 << 12) +# define R300_COMB_FCN_SUB_CLAMP (2 << 12) +# define R300_COMB_FCN_SUB_NOCLAMP (3 << 12) +# define R300_SRC_BLEND_GL_ZERO (32 << 16) +# define R300_SRC_BLEND_GL_ONE (33 << 16) +# define R300_SRC_BLEND_GL_SRC_COLOR (34 << 16) +# define R300_SRC_BLEND_GL_ONE_MINUS_SRC_COLOR (35 << 16) +# define R300_SRC_BLEND_GL_DST_COLOR (36 << 16) +# define R300_SRC_BLEND_GL_ONE_MINUS_DST_COLOR (37 << 16) +# define R300_SRC_BLEND_GL_SRC_ALPHA (38 << 16) +# define R300_SRC_BLEND_GL_ONE_MINUS_SRC_ALPHA (39 << 16) +# define R300_SRC_BLEND_GL_DST_ALPHA (40 << 16) +# define R300_SRC_BLEND_GL_ONE_MINUS_DST_ALPHA (41 << 16) +# define R300_SRC_BLEND_GL_SRC_ALPHA_SATURATE (42 << 16) +# define R300_SRC_BLEND_MASK (63 << 16) +# define R300_DST_BLEND_GL_ZERO (32 << 24) +# define R300_DST_BLEND_GL_ONE (33 << 24) +# define R300_DST_BLEND_GL_SRC_COLOR (34 << 24) +# define R300_DST_BLEND_GL_ONE_MINUS_SRC_COLOR (35 << 24) +# define R300_DST_BLEND_GL_DST_COLOR (36 << 24) +# define R300_DST_BLEND_GL_ONE_MINUS_DST_COLOR (37 << 24) +# define R300_DST_BLEND_GL_SRC_ALPHA (38 << 24) +# define R300_DST_BLEND_GL_ONE_MINUS_SRC_ALPHA (39 << 24) +# define R300_DST_BLEND_GL_DST_ALPHA (40 << 24) +# define R300_DST_BLEND_GL_ONE_MINUS_DST_ALPHA (41 << 24) +# define R300_DST_BLEND_MASK (63 << 24) +#define R300_RB3D_COLORMASK 0x4E0C +# define R300_COLORMASK0_B (1<<0) +# define R300_COLORMASK0_G (1<<1) +# define R300_COLORMASK0_R (1<<2) +# define R300_COLORMASK0_A (1<<3) + +/* gap */ +#define R300_RB3D_COLOROFFSET0 0x4E28 +# define R300_COLOROFFSET_MASK 0xFFFFFFF0 // GUESS +#define R300_RB3D_COLOROFFSET1 0x4E2C // GUESS +#define R300_RB3D_COLOROFFSET2 0x4E30 // GUESS +#define R300_RB3D_COLOROFFSET3 0x4E34 // GUESS +/* gap */ +// Bit 16: Larger tiles +// Bit 17: 4x2 tiles +// Bit 18: Extremely weird tile like, but some pixels duplicated? +#define R300_RB3D_COLORPITCH0 0x4E38 +# define R300_COLORPITCH_MASK 0x00001FF8 // GUESS +# define R300_COLOR_TILE_ENABLE (1 << 16) // GUESS +# define R300_COLOR_MICROTILE_ENABLE (1 << 17) // GUESS +# define R300_COLOR_ENDIAN_NO_SWAP (0 << 18) // GUESS +# define R300_COLOR_ENDIAN_WORD_SWAP (1 << 18) // GUESS +# define R300_COLOR_ENDIAN_DWORD_SWAP (2 << 18) // GUESS +# define R300_COLOR_UNKNOWN_22_23 (3 << 22) // GUESS: Format? +#define R300_RB3D_COLORPITCH1 0x4E3C // GUESS +#define R300_RB3D_COLORPITCH2 0x4E40 // GUESS +#define R300_RB3D_COLORPITCH3 0x4E44 // GUESS + +/* gap */ +// Guess by Vladimir. +// Set to 0A before 3D operations, set to 02 afterwards. +#define R300_RB3D_DSTCACHE_CTLSTAT 0x4E4C +# define R300_RB3D_DSTCACHE_02 0x00000002 +# define R300_RB3D_DSTCACHE_0A 0x0000000A + +/* gap */ +// There seems to be no "write only" setting, so use Z-test = ALWAYS for this. +#define R300_RB3D_ZCNTL_0 0x4F00 +# define R300_RB3D_Z_DISABLED_1 0x00000010 // GUESS +# define R300_RB3D_Z_DISABLED_2 0x00000014 // GUESS +# define R300_RB3D_Z_TEST 0x00000012 +# define R300_RB3D_Z_TEST_AND_WRITE 0x00000016 +#define R300_RB3D_ZCNTL_1 0x4F04 +# define R300_RB3D_Z_TEST_NEVER (0 << 0) // GUESS (based on R200) +# define R300_RB3D_Z_TEST_LESS (1 << 0) +# define R300_RB3D_Z_TEST_LEQUAL (2 << 0) +# define R300_RB3D_Z_TEST_EQUAL (3 << 0) // GUESS +# define R300_RB3D_Z_TEST_GEQUAL (4 << 0) // GUESS +# define R300_RB3D_Z_TEST_GREATER (5 << 0) // GUESS +# define R300_RB3D_Z_TEST_NEQUAL (6 << 0) +# define R300_RB3D_Z_TEST_ALWAYS (7 << 0) +# define R300_RB3D_Z_TEST_MASK (7 << 0) +/* gap */ +#define R300_RB3D_DEPTHOFFSET 0x4F20 +#define R300_RB3D_DEPTHPITCH 0x4F24 +# define R300_DEPTHPITCH_MASK 0x00001FF8 // GUESS +# define R300_DEPTH_TILE_ENABLE (1 << 16) // GUESS +# define R300_DEPTH_MICROTILE_ENABLE (1 << 17) // GUESS +# define R300_DEPTH_ENDIAN_NO_SWAP (0 << 18) // GUESS +# define R300_DEPTH_ENDIAN_WORD_SWAP (1 << 18) // GUESS +# define R300_DEPTH_ENDIAN_DWORD_SWAP (2 << 18) // GUESS + +// BEGIN: Vertex program instruction set +// Every instruction is four dwords long: +// DWORD 0: output and opcode +// DWORD 1: first argument +// DWORD 2: second argument +// DWORD 3: third argument +// +// Notes: +// - ABS r, a is implemented as MAX r, a, -a +// - MOV is implemented as ADD to zero +// - XPD is implemented as MUL + MAD +// - FLR is implemented as FRC + ADD +// - apparently, fglrx tries to schedule instructions so that there is at least +// one instruction between the write to a temporary and the first read +// from said temporary; however, violations of this scheduling are allowed +// - register indices seem to be unrelated with OpenGL aliasing to conventional state +// - only one attribute and one parameter can be loaded at a time; however, the +// same attribute/parameter can be used for more than one argument +// - the second software argument for POW is the third hardware argument (no idea why) +// - MAD with only temporaries as input seems to use VPI_OUT_SELECT_MAD_2 +// +// There is some magic surrounding LIT: +// The single argument is replicated across all three inputs, but swizzled: +// First argument: xyzy +// Second argument: xyzx +// Third argument: xyzw +// Whenever the result is used later in the fragment program, fglrx forces x and w +// to be 1.0 in the input selection; I don't know whether this is strictly necessary +#define R300_VPI_OUT_OP_DOT (1 << 0) +#define R300_VPI_OUT_OP_MUL (2 << 0) +#define R300_VPI_OUT_OP_ADD (3 << 0) +#define R300_VPI_OUT_OP_MAD (4 << 0) +#define R300_VPI_OUT_OP_FRC (6 << 0) +#define R300_VPI_OUT_OP_MAX (7 << 0) +#define R300_VPI_OUT_OP_MIN (8 << 0) +#define R300_VPI_OUT_OP_SGE (9 << 0) +#define R300_VPI_OUT_OP_SLT (10 << 0) +#define R300_VPI_OUT_OP_EXP (65 << 0) +#define R300_VPI_OUT_OP_LOG (66 << 0) +#define R300_VPI_OUT_OP_LIT (68 << 0) +#define R300_VPI_OUT_OP_POW (69 << 0) +#define R300_VPI_OUT_OP_RCP (70 << 0) +#define R300_VPI_OUT_OP_RSQ (72 << 0) +#define R300_VPI_OUT_OP_EX2 (75 << 0) +#define R300_VPI_OUT_OP_LG2 (76 << 0) +#define R300_VPI_OUT_OP_MAD_2 (128 << 0) + +#define R300_VPI_OUT_REG_CLASS_TEMPORARY (0 << 8) +#define R300_VPI_OUT_REG_CLASS_RESULT (2 << 8) +#define R300_VPI_OUT_REG_CLASS_MASK (31 << 8) + +#define R300_VPI_OUT_REG_INDEX_SHIFT 13 +#define R300_VPI_OUT_REG_INDEX_MASK (31 << 13) // GUESS based on fglrx native limits + +#define R300_VPI_OUT_WRITE_X (1 << 20) +#define R300_VPI_OUT_WRITE_Y (1 << 21) +#define R300_VPI_OUT_WRITE_Z (1 << 22) +#define R300_VPI_OUT_WRITE_W (1 << 23) + +#define R300_VPI_IN_REG_CLASS_TEMPORARY (0 << 0) +#define R300_VPI_IN_REG_CLASS_ATTRIBUTE (1 << 0) +#define R300_VPI_IN_REG_CLASS_PARAMETER (2 << 0) +#define R300_VPI_IN_REG_CLASS_NONE (9 << 0) +#define R300_VPI_IN_REG_CLASS_MASK (31 << 0) // GUESS + +#define R300_VPI_IN_REG_INDEX_SHIFT 5 +#define R300_VPI_IN_REG_INDEX_MASK (255 << 5) // GUESS based on fglrx native limits + +// The R300 can select components from the input register arbitrarily. +// Use the following constants, shifted by the component shift you +// want to select +#define R300_VPI_IN_SELECT_X 0 +#define R300_VPI_IN_SELECT_Y 1 +#define R300_VPI_IN_SELECT_Z 2 +#define R300_VPI_IN_SELECT_W 3 +#define R300_VPI_IN_SELECT_ZERO 4 +#define R300_VPI_IN_SELECT_ONE 5 +#define R300_VPI_IN_SELECT_MASK 7 + +#define R300_VPI_IN_X_SHIFT 13 +#define R300_VPI_IN_Y_SHIFT 16 +#define R300_VPI_IN_Z_SHIFT 19 +#define R300_VPI_IN_W_SHIFT 22 + +#define R300_VPI_IN_NEG_X (1 << 25) +#define R300_VPI_IN_NEG_Y (1 << 26) +#define R300_VPI_IN_NEG_Z (1 << 27) +#define R300_VPI_IN_NEG_W (1 << 28) +// END + +#endif // _R300_REG_H diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c new file mode 100644 index 00000000000..e1cb31da40d --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_state.c @@ -0,0 +1,329 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. +Copyright (C) 2004 Nicolai Haehnle. +All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Nicolai Haehnle + */ + +#include "glheader.h" +#include "state.h" +#include "imports.h" +#include "macros.h" +#include "context.h" +#include "dd.h" +#include "simple_list.h" + +#include "api_arrayelt.h" +#include "swrast/swrast.h" +#include "swrast_setup/swrast_setup.h" +#include "array_cache/acache.h" +#include "tnl/tnl.h" + +#include "radeon_ioctl.h" +#include "r300_context.h" +#include "r300_ioctl.h" +#include "r300_state.h" +#include "r300_reg.h" +#include "r300_program.h" + + +/** + * Called by Mesa after an internal state update. + */ +static void r300InvalidateState(GLcontext * ctx, GLuint new_state) +{ + r300ContextPtr r300 = R300_CONTEXT(ctx); + + _swrast_InvalidateState(ctx, new_state); + _swsetup_InvalidateState(ctx, new_state); + _ac_InvalidateState(ctx, new_state); + _tnl_InvalidateState(ctx, new_state); + _ae_invalidate_state(ctx, new_state); + + /* Go inefficiency! */ + r300ResetHwState(r300); +} + + +/** + * Completely recalculates hardware state based on the Mesa state. + */ +void r300ResetHwState(r300ContextPtr r300) +{ + int i; + + if (RADEON_DEBUG & DEBUG_STATE) + fprintf(stderr, "%s\n", __FUNCTION__); + + { + __DRIdrawablePrivate *dPriv = r300->radeon.dri.drawable; + GLfloat xoffset = dPriv ? (GLfloat) dPriv->x : 0; + GLfloat yoffset = dPriv ? (GLfloat) dPriv->y + dPriv->h : 0; + const GLfloat *v = r300->radeon.glCtx->Viewport._WindowMap.m; + + r300->hw.vpt.cmd[R300_VPT_XSCALE] = + r300PackFloat32(v[MAT_SX]); + r300->hw.vpt.cmd[R300_VPT_XOFFSET] = + r300PackFloat32(v[MAT_TX] + xoffset); + r300->hw.vpt.cmd[R300_VPT_YSCALE] = + r300PackFloat32(-v[MAT_SY]); + r300->hw.vpt.cmd[R300_VPT_YOFFSET] = + r300PackFloat32(-v[MAT_TY] + yoffset); + r300->hw.vpt.cmd[R300_VPT_YSCALE] = + r300PackFloat32(v[MAT_SZ]); + r300->hw.vpt.cmd[R300_VPT_YOFFSET] = + r300PackFloat32(v[MAT_TZ]); + } + +//BEGIN: TODO + r300->hw.unk2080.cmd[1] = 0x0030045A; + + r300->hw.ovf.cmd[R300_OVF_FMT_0] = 0x00000003; + r300->hw.ovf.cmd[R300_OVF_FMT_1] = 0x00000000; + + r300->hw.unk20B0.cmd[1] = 0x0000040A; + r300->hw.unk20B0.cmd[2] = 0x00000008; + + r300->hw.unk2134.cmd[1] = 0x00FFFFFF; + r300->hw.unk2134.cmd[2] = 0x00000000; + + r300->hw.unk2140.cmd[1] = 0x00000000; + + ((drm_r300_cmd_header_t*)r300->hw.vir[0].cmd)->unchecked_state.count = 1; + r300->hw.vir[0].cmd[1] = 0x21030003; + + ((drm_r300_cmd_header_t*)r300->hw.vir[1].cmd)->unchecked_state.count = 1; + r300->hw.vir[1].cmd[1] = 0xF688F688; + + r300->hw.vic.cmd[R300_VIR_CNTL_0] = 0x00000001; + r300->hw.vic.cmd[R300_VIR_CNTL_1] = 0x00000405; + + r300->hw.unk21DC.cmd[1] = 0xAAAAAAAA; + + r300->hw.unk221C.cmd[1] = R300_221C_NORMAL; + + r300->hw.unk2220.cmd[1] = r300PackFloat32(1.0); + r300->hw.unk2220.cmd[2] = r300PackFloat32(1.0); + r300->hw.unk2220.cmd[3] = r300PackFloat32(1.0); + r300->hw.unk2220.cmd[4] = r300PackFloat32(1.0); + + if (GET_CHIP(r300->radeon.radeonScreen) == RADEON_CHIP_R300) + r300->hw.unk2288.cmd[1] = R300_2288_R300; + else + r300->hw.unk2288.cmd[1] = R300_2288_RV350; + + r300->hw.pvs.cmd[R300_PVS_CNTL_1] = 0; + r300->hw.pvs.cmd[R300_PVS_CNTL_2] = 0; + r300->hw.pvs.cmd[R300_PVS_CNTL_3] = 0; + + r300->hw.unk4008.cmd[1] = 0x00000007; + + r300->hw.unk4010.cmd[1] = 0x66666666; + r300->hw.unk4010.cmd[2] = 0x06666666; + if (GET_CHIP(r300->radeon.radeonScreen) == RADEON_CHIP_R300) + r300->hw.unk4010.cmd[3] = 0x00000017; + else + r300->hw.unk4010.cmd[3] = 0x00000011; + r300->hw.unk4010.cmd[4] = 0x00000000; + r300->hw.unk4010.cmd[5] = 0x00000000; + + r300->hw.txe.cmd[R300_TXE_ENABLE] = 0; + + r300->hw.unk4200.cmd[1] = r300PackFloat32(0.0); + r300->hw.unk4200.cmd[2] = r300PackFloat32(0.0); + r300->hw.unk4200.cmd[3] = r300PackFloat32(1.0); + r300->hw.unk4200.cmd[4] = r300PackFloat32(1.0); + + r300->hw.unk4214.cmd[1] = 0x00050005; + + r300->hw.unk4230.cmd[1] = 0x01800000; + r300->hw.unk4230.cmd[2] = 0x00020006; + r300->hw.unk4230.cmd[3] = r300PackFloat32(1.0 / 192.0); + + r300->hw.unk4260.cmd[1] = 0; + r300->hw.unk4260.cmd[2] = r300PackFloat32(0.0); + r300->hw.unk4260.cmd[3] = r300PackFloat32(1.0); + + r300->hw.unk4274.cmd[1] = 0x00000002; + r300->hw.unk4274.cmd[2] = 0x0003AAAA; + r300->hw.unk4274.cmd[3] = 0x00000000; + r300->hw.unk4274.cmd[4] = 0x00000000; + + r300->hw.unk4288.cmd[1] = 0x00000000; + r300->hw.unk4288.cmd[2] = 0x00000001; + r300->hw.unk4288.cmd[3] = 0x00000000; + r300->hw.unk4288.cmd[4] = 0x00000000; + r300->hw.unk4288.cmd[5] = 0x00000000; + + r300->hw.unk42A0.cmd[1] = 0x00000000; + + r300->hw.unk42B4.cmd[1] = 0x00000000; + r300->hw.unk42B4.cmd[2] = 0x00000000; + + r300->hw.unk42C0.cmd[1] = 0x4B7FFFFF; + r300->hw.unk42C0.cmd[2] = 0x00000000; + + r300->hw.rc.cmd[1] = R300_RS_CNTL_0_UNKNOWN_7; + r300->hw.rc.cmd[2] = 0; + + for(i = 1; i <= 8; ++i) + r300->hw.ri.cmd[i] = 0; + + ((drm_r300_cmd_header_t*)r300->hw.rr.cmd)->unchecked_state.count = 1; + for(i = 1; i <= 8; ++i) + r300->hw.rr.cmd[1] = 0; + + r300->hw.unk43A4.cmd[1] = 0x0000001C; + r300->hw.unk43A4.cmd[2] = 0x2DA49525; + + r300->hw.unk43E0.cmd[1] = 0x00000000; + r300->hw.unk43E0.cmd[2] = 0x03FFFFFF; + r300->hw.unk43E0.cmd[3] = 0x00FFFFFF; + + r300->hw.fp.cmd[R300_FP_CNTL0] = 0; + r300->hw.fp.cmd[R300_FP_CNTL1] = 0; + r300->hw.fp.cmd[R300_FP_CNTL2] = 0; + r300->hw.fp.cmd[R300_FP_NODE0] = 0; + r300->hw.fp.cmd[R300_FP_NODE1] = 0; + r300->hw.fp.cmd[R300_FP_NODE2] = 0; + r300->hw.fp.cmd[R300_FP_NODE3] = 0; + + r300->hw.unk46A4.cmd[1] = 0x00001B01; + r300->hw.unk46A4.cmd[2] = 0x00001B0F; + r300->hw.unk46A4.cmd[3] = 0x00001B0F; + r300->hw.unk46A4.cmd[4] = 0x00001B0F; + r300->hw.unk46A4.cmd[5] = 0x00000001; + + for(i = 1; i <= 64; ++i) { + /* create NOP instructions */ + r300->hw.fpi[0].cmd[i] = FP_INSTRC(MAD, FP_ARGC(SRC0C_XYZ), FP_ARGC(ONE), FP_ARGC(ZERO)); + r300->hw.fpi[1].cmd[i] = FP_SELC(0,XYZ,NO,FP_TMP(0),0,0); + r300->hw.fpi[2].cmd[i] = FP_INSTRA(MAD, FP_ARGA(SRC0A), FP_ARGA(ONE), FP_ARGA(ZERO)); + r300->hw.fpi[3].cmd[i] = FP_SELA(0,W,NO,FP_TMP(0),0,0); + } + + r300->hw.unk4BC0.cmd[1] = 0; + + r300->hw.unk4BC8.cmd[1] = 0; + r300->hw.unk4BC8.cmd[2] = 0; + r300->hw.unk4BC8.cmd[3] = 0; + + r300->hw.at.cmd[R300_AT_ALPHA_TEST] = 0; + + r300->hw.unk4BD8.cmd[1] = 0; + + r300->hw.unk4E00.cmd[1] = 0; + + r300->hw.bld.cmd[R300_BLD_CBLEND] = 0; + r300->hw.bld.cmd[R300_BLD_ABLEND] = 0; + + r300->hw.cmk.cmd[R300_CMK_COLORMASK] = 0xF; + + r300->hw.unk4E10.cmd[1] = 0; + r300->hw.unk4E10.cmd[2] = 0; + r300->hw.unk4E10.cmd[3] = 0; + + r300->hw.cb.cmd[R300_CB_OFFSET] = + r300->radeon.radeonScreen->backOffset + + r300->radeon.radeonScreen->fbLocation; + r300->hw.cb.cmd[R300_CB_PITCH] = r300->radeon.radeonScreen->backPitch + | R300_COLOR_UNKNOWN_22_23; + + r300->hw.unk4E50.cmd[1] = 0; + r300->hw.unk4E50.cmd[2] = 0; + r300->hw.unk4E50.cmd[3] = 0; + r300->hw.unk4E50.cmd[4] = 0; + r300->hw.unk4E50.cmd[5] = 0; + r300->hw.unk4E50.cmd[6] = 0; + r300->hw.unk4E50.cmd[7] = 0; + r300->hw.unk4E50.cmd[8] = 0; + r300->hw.unk4E50.cmd[9] = 0; + + r300->hw.unk4E88.cmd[1] = 0; + + r300->hw.zc.cmd[R300_ZC_CNTL_0] = 0; + r300->hw.zc.cmd[R300_ZC_CNTL_1] = 0; + + r300->hw.unk4F08.cmd[1] = 0x00FFFF00; + r300->hw.unk4F08.cmd[2] = 0x00000002; + r300->hw.unk4F08.cmd[3] = 0x00000000; + r300->hw.unk4F08.cmd[4] = 0x00000003; + r300->hw.unk4F08.cmd[5] = 0x00000000; + + r300->hw.zb.cmd[R300_ZB_OFFSET] = + r300->radeon.radeonScreen->depthOffset + + r300->radeon.radeonScreen->fbLocation; + r300->hw.zb.cmd[R300_ZB_PITCH] = r300->radeon.radeonScreen->depthPitch; + + r300->hw.unk4F28.cmd[1] = 0; + + r300->hw.unk4F30.cmd[1] = 0; + r300->hw.unk4F30.cmd[2] = 0; + + r300->hw.unk4F44.cmd[1] = 0; + + r300->hw.unk4F54.cmd[1] = 0; + + ((drm_r300_cmd_header_t*)r300->hw.vpi.cmd)->vpu.count = 0; + for(i = 1; i < R300_VPI_CMDSIZE; i += 4) { + /* MOV t0, t0 */ + r300->hw.vpi.cmd[i+0] = VP_OUT(ADD,TMP,0,XYZW); + r300->hw.vpi.cmd[i+1] = VP_IN(TMP,0); + r300->hw.vpi.cmd[i+2] = VP_ZERO(); + r300->hw.vpi.cmd[i+3] = VP_ZERO(); + } + + ((drm_r300_cmd_header_t*)r300->hw.vpp.cmd)->vpu.count = 0; + for(i = 1; i < R300_VPP_CMDSIZE; ++i) + r300->hw.vpp.cmd[i] = 0; +//END: TODO + + r300->hw.all_dirty = GL_TRUE; +} + + +/** + * Calculate initial hardware state and register state functions. + * Assumes that the command buffer and state atoms have been + * initialized already. + */ +void r300InitState(r300ContextPtr r300) +{ + struct dd_function_table* functions; + + r300ResetHwState(r300); + + /* Setup state functions */ + functions = &r300->radeon.glCtx->Driver; + functions->UpdateState = r300InvalidateState; +} + diff --git a/src/mesa/drivers/dri/r300/r300_state.h b/src/mesa/drivers/dri/r300/r300_state.h new file mode 100644 index 00000000000..59d76e33df7 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_state.h @@ -0,0 +1,49 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Nicolai Haehnle + */ + +#ifndef __R300_STATE_H__ +#define __R300_STATE_H__ + +#include "r300_context.h" + +#define R300_STATECHANGE(r300, atom) \ + do { \ + r300->hw.atom.dirty = GL_TRUE; \ + r300->hw.is_dirty = GL_TRUE; \ + } while(0) + +extern void r300ResetHwState(r300ContextPtr r300); +extern void r300InitState(r300ContextPtr r300); + +#endif /* __R300_STATE_H__ */ diff --git a/src/mesa/drivers/dri/r300/radeon_context.c b/src/mesa/drivers/dri/r300/radeon_context.c new file mode 100644 index 00000000000..d04045d9717 --- /dev/null +++ b/src/mesa/drivers/dri/r300/radeon_context.c @@ -0,0 +1,302 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/** + * \file radeon_context.c + * Common context initialization. + * + * \author Keith Whitwell + */ + +#include + +#include "glheader.h" +#include "imports.h" +#include "context.h" +#include "state.h" +#include "matrix.h" + +#include "drivers/common/driverfuncs.h" +#include "swrast/swrast.h" + +#include "radeon_screen.h" +#include "r200_context.h" +#include "radeon_ioctl.h" +#include "radeon_macros.h" +#include "radeon_reg.h" +#include "r200_state.h" + +#include "utils.h" +#include "vblank.h" +#include "xmlpool.h" /* for symbolic values of enum-type options */ + +#define DRIVER_DATE "20040924" + + +/* Return various strings for glGetString(). + */ +static const GLubyte *radeonGetString(GLcontext * ctx, GLenum name) +{ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + static char buffer[128]; + + switch (name) { + case GL_VENDOR: + return (GLubyte *) "Tungsten Graphics, Inc."; + + case GL_RENDERER: + { + unsigned offset; + GLuint agp_mode = radeon->radeonScreen->IsPCI ? 0 : + radeon->radeonScreen->AGPMode; + const char* chipname; + + if (IS_FAMILY_R300(radeon)) + chipname = "R300"; + else + chipname = "R200"; + + offset = driGetRendererString(buffer, chipname, DRIVER_DATE, + agp_mode); + + sprintf(&buffer[offset], " %sTCL", + !(radeon->TclFallback & RADEON_TCL_FALLBACK_TCL_DISABLE) + ? "" : "NO-"); + + return (GLubyte *) buffer; + } + + default: + return NULL; + } +} + + +/* Return the width and height of the given buffer. + */ +static void radeonGetBufferSize(GLframebuffer * buffer, + GLuint * width, GLuint * height) +{ + GET_CURRENT_CONTEXT(ctx); + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + + LOCK_HARDWARE(radeon); + *width = radeon->dri.drawable->w; + *height = radeon->dri.drawable->h; + UNLOCK_HARDWARE(radeon); +} + + +/* Initialize the driver's misc functions. + */ +static void radeonInitDriverFuncs(struct dd_function_table *functions) +{ + functions->GetBufferSize = radeonGetBufferSize; + functions->ResizeBuffers = _swrast_alloc_buffers; + functions->GetString = radeonGetString; +} + + +static int get_ust_nop(int64_t * ust) +{ + *ust = 1; + return 0; +} + +/** + * Create and initialize all common fields of the context, + * including the Mesa context itself. + */ +GLboolean radeonInitContext(radeonContextPtr radeon, + struct dd_function_table* functions, + const __GLcontextModes * glVisual, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate) +{ + __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; + radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private); + GLcontext* ctx; + GLcontext* shareCtx; + int fthrottle_mode; + + /* Fill in additional standard functions. */ + radeonInitDriverFuncs(functions); + + /* Allocate and initialize the Mesa context */ + if (sharedContextPrivate) + shareCtx = ((radeonContextPtr)sharedContextPrivate)->glCtx; + else + shareCtx = NULL; + radeon->glCtx = _mesa_create_context(glVisual, shareCtx, + functions, (void *)radeon); + if (!radeon->glCtx) + return GL_FALSE; + + ctx = radeon->glCtx; + driContextPriv->driverPrivate = radeon; + + /* DRI fields */ + radeon->dri.context = driContextPriv; + radeon->dri.screen = sPriv; + radeon->dri.drawable = NULL; /* Set by XMesaMakeCurrent */ + radeon->dri.hwContext = driContextPriv->hHWContext; + radeon->dri.hwLock = &sPriv->pSAREA->lock; + radeon->dri.fd = sPriv->fd; + radeon->dri.drmMinor = sPriv->drmMinor; + + radeon->radeonScreen = screen; + radeon->sarea = (drm_radeon_sarea_t *) ((GLubyte *) sPriv->pSAREA + + screen->sarea_priv_offset); + + /* Setup IRQs */ + fthrottle_mode = driQueryOptioni(&radeon->optionCache, "fthrottle_mode"); + radeon->iw.irq_seq = -1; + radeon->irqsEmitted = 0; + radeon->do_irqs = (radeon->dri.drmMinor >= 6 && + fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS && + radeon->radeonScreen->irq); + + radeon->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS); + + if (!radeon->do_irqs) + fprintf(stderr, + "IRQ's not enabled, falling back to %s: %d %d %d\n", + radeon->do_usleeps ? "usleeps" : "busy waits", + radeon->dri.drmMinor, + fthrottle_mode, radeon->radeonScreen->irq); + + radeon->vblank_flags = (radeon->radeonScreen->irq != 0) + ? driGetDefaultVBlankFlags(&radeon->optionCache) : VBLANK_FLAG_NO_IRQ; + + radeon->get_ust = + (PFNGLXGETUSTPROC) glXGetProcAddress((const GLubyte *) + "__glXGetUST"); + if (radeon->get_ust == NULL) { + radeon->get_ust = get_ust_nop; + } + (*radeon->get_ust) (&radeon->swap_ust); + + return GL_TRUE; +} + + +/** + * Swap front and back buffer. + */ +void radeonSwapBuffers(__DRIdrawablePrivate * dPriv) +{ + if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { + radeonContextPtr radeon; + GLcontext *ctx; + + radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate; + ctx = radeon->glCtx; + + if (ctx->Visual.doubleBufferMode) { + _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ + if (radeon->doPageFlip) { + radeonPageFlip(dPriv); + } else { + radeonCopyBuffer(dPriv); + } + } + } else { + /* XXX this shouldn't be an error but we can't handle it for now */ + _mesa_problem(NULL, "%s: drawable has no context!", + __FUNCTION__); + } +} + + +/* Force the context `c' to be the current context and associate with it + * buffer `b'. + */ +GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv, + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv) +{ + if (driContextPriv) { + radeonContextPtr radeon = + (radeonContextPtr) driContextPriv->driverPrivate; + + if (RADEON_DEBUG & DEBUG_DRI) + fprintf(stderr, "%s ctx %p\n", __FUNCTION__, + radeon->glCtx); + + if (radeon->dri.drawable != driDrawPriv) { + driDrawableInitVBlank(driDrawPriv, + radeon->vblank_flags); + radeon->dri.drawable = driDrawPriv; + + if (IS_FAMILY_R200(radeon)) { + r200UpdateWindow(radeon->glCtx); + r200UpdateViewportOffset(radeon->glCtx); + } + } + + _mesa_make_current2(radeon->glCtx, + (GLframebuffer *) driDrawPriv-> + driverPrivate, + (GLframebuffer *) driReadPriv-> + driverPrivate); + + if (!radeon->glCtx->Viewport.Width) { + _mesa_set_viewport(radeon->glCtx, 0, 0, + driDrawPriv->w, driDrawPriv->h); + } + + _mesa_update_state(radeon->glCtx); + + if (IS_FAMILY_R200(radeon)) + r200ValidateState(radeon->glCtx); + + } else { + if (RADEON_DEBUG & DEBUG_DRI) + fprintf(stderr, "%s ctx is null\n", __FUNCTION__); + _mesa_make_current(0, 0); + } + + if (RADEON_DEBUG & DEBUG_DRI) + fprintf(stderr, "End %s\n", __FUNCTION__); + return GL_TRUE; +} + +/* Force the context `c' to be unbound from its buffer. + */ +GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv) +{ + radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate; + + if (RADEON_DEBUG & DEBUG_DRI) + fprintf(stderr, "%s ctx %p\n", __FUNCTION__, + radeon->glCtx); + + return GL_TRUE; +} + diff --git a/src/mesa/drivers/dri/r300/radeon_context.h b/src/mesa/drivers/dri/r300/radeon_context.h new file mode 100644 index 00000000000..d51cf4c09d1 --- /dev/null +++ b/src/mesa/drivers/dri/r300/radeon_context.h @@ -0,0 +1,222 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Keith Whitwell + * Nicolai Haehnle + */ + +#ifndef __RADEON_CONTEXT_H__ +#define __RADEON_CONTEXT_H__ + +#include "mtypes.h" +#include "radeon_screen.h" +#include "drm.h" +#include "dri_util.h" + +struct radeon_context; +typedef struct radeon_context radeonContextRec; +typedef struct radeon_context* radeonContextPtr; + +static __inline GLuint radeonPackColor(GLuint cpp, + GLubyte r, GLubyte g, GLubyte b, GLubyte a) +{ + switch (cpp) { + case 2: + return PACK_COLOR_565(r, g, b); + case 4: + return PACK_COLOR_8888(a, r, g, b); + default: + return 0; + } +} + + +/* Rasterizing fallbacks */ +/* See correponding strings in r200_swtcl.c */ +#define RADEON_FALLBACK_TEXTURE 0x01 +#define RADEON_FALLBACK_DRAW_BUFFER 0x02 +#define RADEON_FALLBACK_STENCIL 0x04 +#define RADEON_FALLBACK_RENDER_MODE 0x08 +#define RADEON_FALLBACK_BLEND_EQ 0x10 +#define RADEON_FALLBACK_BLEND_FUNC 0x20 +#define RADEON_FALLBACK_DISABLE 0x40 +#define RADEON_FALLBACK_BORDER_MODE 0x80 + +extern void radeonFallback(GLcontext * ctx, GLuint bit, GLboolean mode); + +#define FALLBACK( radeon, bit, mode ) do { \ + if ( 0 ) fprintf( stderr, "FALLBACK in %s: #%d=%d\n", \ + __FUNCTION__, bit, mode ); \ + radeonFallback( (radeon)->glCtx, bit, mode ); \ +} while (0) + + +/* TCL fallbacks */ +extern void radeonTclFallback(GLcontext * ctx, GLuint bit, GLboolean mode); + +#define RADEON_TCL_FALLBACK_RASTER 0x0001 /* rasterization */ +#define RADEON_TCL_FALLBACK_UNFILLED 0x0002 /* unfilled tris */ +#define RADEON_TCL_FALLBACK_LIGHT_TWOSIDE 0x0004 /* twoside tris */ +#define RADEON_TCL_FALLBACK_MATERIAL 0x0008 /* material in vb */ +#define RADEON_TCL_FALLBACK_TEXGEN_0 0x0010 /* texgen, unit 0 */ +#define RADEON_TCL_FALLBACK_TEXGEN_1 0x0020 /* texgen, unit 1 */ +#define RADEON_TCL_FALLBACK_TEXGEN_2 0x0040 /* texgen, unit 2 */ +#define RADEON_TCL_FALLBACK_TEXGEN_3 0x0080 /* texgen, unit 3 */ +#define RADEON_TCL_FALLBACK_TEXGEN_4 0x0100 /* texgen, unit 4 */ +#define RADEON_TCL_FALLBACK_TEXGEN_5 0x0200 /* texgen, unit 5 */ +#define RADEON_TCL_FALLBACK_TCL_DISABLE 0x0400 /* user disable */ +#define RADEON_TCL_FALLBACK_BITMAP 0x0800 /* draw bitmap with points */ +#define RADEON_TCL_FALLBACK_VERTEX_PROGRAM 0x1000 /* vertex program active */ + +#define TCL_FALLBACK( ctx, bit, mode ) radeonTclFallback( ctx, bit, mode ) + + +struct radeon_dri_mirror { + __DRIcontextPrivate *context; /* DRI context */ + __DRIscreenPrivate *screen; /* DRI screen */ + __DRIdrawablePrivate *drawable; /* DRI drawable bound to this ctx */ + + drm_context_t hwContext; + drm_hw_lock_t *hwLock; + int fd; + int drmMinor; +}; + +/** + * Derived state for internal purposes. + */ +struct radeon_colorbuffer_state { + GLuint clear; + GLint drawOffset, drawPitch; +}; + +struct radeon_pixel_state { + GLint readOffset, readPitch; +}; + +struct radeon_state { + struct radeon_colorbuffer_state color; + struct radeon_pixel_state pixel; +}; + +/** + * Common per-context variables shared by R200 and R300. + * R200- and R300-specific code "derive" their own context from this + * structure. + */ +struct radeon_context { + GLcontext *glCtx; /* Mesa context */ + radeonScreenPtr radeonScreen; /* Screen private DRI data */ + + /* Fallback state */ + GLuint Fallback; + GLuint TclFallback; + + /* Page flipping */ + GLuint doPageFlip; + + /* Drawable, cliprect and scissor information */ + GLuint numClipRects; /* Cliprects for the draw buffer */ + drm_clip_rect_t *pClipRects; + unsigned int lastStamp; + GLboolean lost_context; + drm_radeon_sarea_t *sarea; /* Private SAREA data */ + + /* Mirrors of some DRI state */ + struct radeon_dri_mirror dri; + + /* Busy waiting */ + GLuint do_usleeps; + GLuint do_irqs; + GLuint irqsEmitted; + drm_radeon_irq_wait_t iw; + + /* VBI / buffer swap */ + GLuint vbl_seq; + GLuint vblank_flags; + + int64_t swap_ust; + int64_t swap_missed_ust; + + GLuint swap_count; + GLuint swap_missed_count; + + PFNGLXGETUSTPROC get_ust; + + /* Derived state */ + struct radeon_state state; + + /* Configuration cache + */ + driOptionCache optionCache; +}; + +#define RADEON_CONTEXT(glctx) ((radeonContextPtr)(ctx->DriverCtx)) + +extern void radeonSwapBuffers(__DRIdrawablePrivate * dPriv); +extern GLboolean radeonInitContext(radeonContextPtr radeon, + struct dd_function_table* functions, + const __GLcontextModes * glVisual, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate); +extern GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv, + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv); +extern GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv); + +/* ================================================================ + * Debugging: + */ +#define DO_DEBUG 1 + +#if DO_DEBUG +extern int RADEON_DEBUG; +#else +#define RADEON_DEBUG 0 +#endif + +#define DEBUG_TEXTURE 0x0001 +#define DEBUG_STATE 0x0002 +#define DEBUG_IOCTL 0x0004 +#define DEBUG_PRIMS 0x0008 +#define DEBUG_VERTS 0x0010 +#define DEBUG_FALLBACKS 0x0020 +#define DEBUG_VFMT 0x0040 +#define DEBUG_CODEGEN 0x0080 +#define DEBUG_VERBOSE 0x0100 +#define DEBUG_DRI 0x0200 +#define DEBUG_DMA 0x0400 +#define DEBUG_SANITY 0x0800 +#define DEBUG_SYNC 0x1000 +#define DEBUG_PIXEL 0x2000 +#define DEBUG_MEMORY 0x4000 + +#endif /* __RADEON_CONTEXT_H__ */ diff --git a/src/mesa/drivers/dri/r300/radeon_ioctl.c b/src/mesa/drivers/dri/r300/radeon_ioctl.c new file mode 100644 index 00000000000..0caf44569dd --- /dev/null +++ b/src/mesa/drivers/dri/r300/radeon_ioctl.c @@ -0,0 +1,348 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Keith Whitwell + */ + +#include +#include + +#include "glheader.h" +#include "imports.h" +#include "macros.h" +#include "context.h" +#include "swrast/swrast.h" + +#include "r200_context.h" +#include "r200_state.h" +#include "radeon_ioctl.h" +#include "r200_ioctl.h" +#include "r300_ioctl.h" +#include "r200_tcl.h" +#include "r200_sanity.h" +#include "radeon_reg.h" + +#include "vblank.h" + +static void radeonWaitForIdle(radeonContextPtr radeon); + +/* ================================================================ + * SwapBuffers with client-side throttling + */ + +static uint32_t radeonGetLastFrame(radeonContextPtr radeon) +{ + drm_radeon_getparam_t gp; + int ret; + uint32_t frame; + + gp.param = RADEON_PARAM_LAST_FRAME; + gp.value = (int *)&frame; + ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_GETPARAM, + &gp, sizeof(gp)); + if (ret) { + fprintf(stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__, + ret); + exit(1); + } + + return frame; +} + +static void radeonEmitIrqLocked(radeonContextPtr radeon) +{ + drm_radeon_irq_emit_t ie; + int ret; + + ie.irq_seq = &radeon->iw.irq_seq; + ret = drmCommandWriteRead(radeon->dri.fd, DRM_RADEON_IRQ_EMIT, + &ie, sizeof(ie)); + if (ret) { + fprintf(stderr, "%s: drmRadeonIrqEmit: %d\n", __FUNCTION__, + ret); + exit(1); + } +} + +static void radeonWaitIrq(radeonContextPtr radeon) +{ + int ret; + + do { + ret = drmCommandWrite(radeon->dri.fd, DRM_RADEON_IRQ_WAIT, + &radeon->iw, sizeof(radeon->iw)); + } while (ret && (errno == EINTR || errno == EAGAIN)); + + if (ret) { + fprintf(stderr, "%s: drmRadeonIrqWait: %d\n", __FUNCTION__, + ret); + exit(1); + } +} + +static void radeonWaitForFrameCompletion(radeonContextPtr radeon) +{ + drm_radeon_sarea_t *sarea = radeon->sarea; + + if (radeon->do_irqs) { + if (radeonGetLastFrame(radeon) < sarea->last_frame) { + if (!radeon->irqsEmitted) { + while (radeonGetLastFrame(radeon) < + sarea->last_frame) ; + } else { + UNLOCK_HARDWARE(radeon); + radeonWaitIrq(radeon); + LOCK_HARDWARE(radeon); + } + radeon->irqsEmitted = 10; + } + + if (radeon->irqsEmitted) { + radeonEmitIrqLocked(radeon); + radeon->irqsEmitted--; + } + } else { + while (radeonGetLastFrame(radeon) < sarea->last_frame) { + UNLOCK_HARDWARE(radeon); + if (radeon->do_usleeps) + DO_USLEEP(1); + LOCK_HARDWARE(radeon); + } + } +} + +/* Copy the back color buffer to the front color buffer. + */ +void radeonCopyBuffer(const __DRIdrawablePrivate * dPriv) +{ + radeonContextPtr radeon; + GLint nbox, i, ret; + GLboolean missed_target; + int64_t ust; + + assert(dPriv); + assert(dPriv->driContextPriv); + assert(dPriv->driContextPriv->driverPrivate); + + radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate; + + if (RADEON_DEBUG & DEBUG_IOCTL) { + fprintf(stderr, "\n%s( %p )\n\n", __FUNCTION__, + (void *)radeon->glCtx); + } + + if (IS_FAMILY_R200(radeon)) + R200_FIREVERTICES((r200ContextPtr)radeon); + + LOCK_HARDWARE(radeon); + + /* Throttle the frame rate -- only allow one pending swap buffers + * request at a time. + */ + radeonWaitForFrameCompletion(radeon); + UNLOCK_HARDWARE(radeon); + driWaitForVBlank(dPriv, &radeon->vbl_seq, radeon->vblank_flags, + &missed_target); + LOCK_HARDWARE(radeon); + + nbox = dPriv->numClipRects; /* must be in locked region */ + + for (i = 0; i < nbox;) { + GLint nr = MIN2(i + RADEON_NR_SAREA_CLIPRECTS, nbox); + drm_clip_rect_t *box = dPriv->pClipRects; + drm_clip_rect_t *b = radeon->sarea->boxes; + GLint n = 0; + + for (; i < nr; i++) { + *b++ = box[i]; + n++; + } + radeon->sarea->nbox = n; + + ret = drmCommandNone(radeon->dri.fd, DRM_RADEON_SWAP); + + if (ret) { + fprintf(stderr, "DRM_R200_SWAP_BUFFERS: return = %d\n", + ret); + UNLOCK_HARDWARE(radeon); + exit(1); + } + } + + UNLOCK_HARDWARE(radeon); + + if (IS_FAMILY_R200(radeon)) + ((r200ContextPtr)radeon)->hw.all_dirty = GL_TRUE; + + radeon->swap_count++; + (*radeon->get_ust) (&ust); + if (missed_target) { + radeon->swap_missed_count++; + radeon->swap_missed_ust = ust - radeon->swap_ust; + } + + radeon->swap_ust = ust; + + sched_yield(); +} + +void radeonPageFlip(const __DRIdrawablePrivate * dPriv) +{ + radeonContextPtr radeon; + GLint ret; + GLboolean missed_target; + + assert(dPriv); + assert(dPriv->driContextPriv); + assert(dPriv->driContextPriv->driverPrivate); + + radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate; + + if (RADEON_DEBUG & DEBUG_IOCTL) { + fprintf(stderr, "%s: pfCurrentPage: %d\n", __FUNCTION__, + radeon->sarea->pfCurrentPage); + } + + if (IS_FAMILY_R200(radeon)) + R200_FIREVERTICES((r200ContextPtr)radeon); + LOCK_HARDWARE(radeon); + + if (!dPriv->numClipRects) { + UNLOCK_HARDWARE(radeon); + usleep(10000); /* throttle invisible client 10ms */ + return; + } + + /* Need to do this for the perf box placement: + */ + { + drm_clip_rect_t *box = dPriv->pClipRects; + drm_clip_rect_t *b = radeon->sarea->boxes; + b[0] = box[0]; + radeon->sarea->nbox = 1; + } + + /* Throttle the frame rate -- only allow a few pending swap buffers + * request at a time. + */ + radeonWaitForFrameCompletion(radeon); + UNLOCK_HARDWARE(radeon); + driWaitForVBlank(dPriv, &radeon->vbl_seq, radeon->vblank_flags, + &missed_target); + if (missed_target) { + radeon->swap_missed_count++; + (void)(*radeon->get_ust) (&radeon->swap_missed_ust); + } + LOCK_HARDWARE(radeon); + + ret = drmCommandNone(radeon->dri.fd, DRM_RADEON_FLIP); + + UNLOCK_HARDWARE(radeon); + + if (ret) { + fprintf(stderr, "DRM_RADEON_FLIP: return = %d\n", ret); + exit(1); + } + + radeon->swap_count++; + (void)(*radeon->get_ust) (&radeon->swap_ust); + + if (radeon->sarea->pfCurrentPage == 1) { + radeon->state.color.drawOffset = radeon->radeonScreen->frontOffset; + radeon->state.color.drawPitch = radeon->radeonScreen->frontPitch; + } else { + radeon->state.color.drawOffset = radeon->radeonScreen->backOffset; + radeon->state.color.drawPitch = radeon->radeonScreen->backPitch; + } + + if (IS_FAMILY_R200(radeon)) { + r200ContextPtr r200 = (r200ContextPtr)radeon; + + R200_STATECHANGE(r200, ctx); + r200->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] = radeon->state.color.drawOffset + + radeon->radeonScreen->fbLocation; + r200->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = radeon->state.color.drawPitch; + } +} + +void radeonWaitForIdleLocked(radeonContextPtr radeon) +{ + int ret; + int i = 0; + + do { + ret = drmCommandNone(radeon->dri.fd, DRM_RADEON_CP_IDLE); + if (ret) + DO_USLEEP(1); + } while (ret && ++i < 100); + + if (ret < 0) { + UNLOCK_HARDWARE(radeon); + fprintf(stderr, "Error: R200 timed out... exiting\n"); + exit(-1); + } +} + +static void radeonWaitForIdle(radeonContextPtr radeon) +{ + LOCK_HARDWARE(radeon); + radeonWaitForIdleLocked(radeon); + UNLOCK_HARDWARE(radeon); +} + +void radeonFlush(GLcontext * ctx) +{ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + + if (IS_FAMILY_R300(radeon)) + r300Flush(ctx); + else + r200Flush(ctx); +} + + +/* Make sure all commands have been sent to the hardware and have + * completed processing. + */ +void radeonFinish(GLcontext * ctx) +{ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + + radeonFlush(ctx); + + if (radeon->do_irqs) { + LOCK_HARDWARE(radeon); + radeonEmitIrqLocked(radeon); + UNLOCK_HARDWARE(radeon); + radeonWaitIrq(radeon); + } else + radeonWaitForIdle(radeon); +} diff --git a/src/mesa/drivers/dri/r300/radeon_ioctl.h b/src/mesa/drivers/dri/r300/radeon_ioctl.h new file mode 100644 index 00000000000..7c19325dd3a --- /dev/null +++ b/src/mesa/drivers/dri/r300/radeon_ioctl.h @@ -0,0 +1,53 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Keith Whitwell + */ + +#ifndef __RADEON_IOCTL_H__ +#define __RADEON_IOCTL_H__ + +#include "simple_list.h" +#include "radeon_dri.h" +#include "radeon_lock.h" + +#include "xf86drm.h" +#include "drm.h" +#include "r200_context.h" +#include "radeon_drm.h" + +extern void radeonCopyBuffer(const __DRIdrawablePrivate * drawable); +extern void radeonPageFlip(const __DRIdrawablePrivate * drawable); +extern void radeonFlush(GLcontext * ctx); +extern void radeonFinish(GLcontext * ctx); +extern void radeonWaitForIdleLocked(radeonContextPtr radeon); + +#endif /* __RADEON_IOCTL_H__ */ diff --git a/src/mesa/drivers/dri/r300/radeon_lock.c b/src/mesa/drivers/dri/r300/radeon_lock.c new file mode 100644 index 00000000000..b5fb20ddb44 --- /dev/null +++ b/src/mesa/drivers/dri/r300/radeon_lock.c @@ -0,0 +1,159 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Keith Whitwell + */ +#include + +#include "r200_context.h" +#include "radeon_lock.h" +#include "r200_tex.h" +#include "r200_state.h" +#include "radeon_ioctl.h" + +#if DEBUG_LOCKING +char *prevLockFile = NULL; +int prevLockLine = 0; +#endif + +/* Turn on/off page flipping according to the flags in the sarea: + */ +static void radeonUpdatePageFlipping(radeonContextPtr radeon) +{ + int use_back; + + radeon->doPageFlip = radeon->sarea->pfState; + + use_back = (radeon->glCtx->Color._DrawDestMask == DD_BACK_LEFT_BIT); + use_back ^= (radeon->sarea->pfCurrentPage == 1); + + if (use_back) { + radeon->state.color.drawOffset = radeon->radeonScreen->backOffset; + radeon->state.color.drawPitch = radeon->radeonScreen->backPitch; + } else { + radeon->state.color.drawOffset = radeon->radeonScreen->frontOffset; + radeon->state.color.drawPitch = radeon->radeonScreen->frontPitch; + } +} + +/** + * Called by radeonGetLock() after the lock has been obtained. + */ +static void r200RegainedLock(r200ContextPtr r200) +{ + __DRIdrawablePrivate *dPriv = r200->radeon.dri.drawable; + __DRIscreenPrivate *sPriv = r200->radeon.dri.screen; + int i; + + if (r200->radeon.lastStamp != dPriv->lastStamp) { + radeonUpdatePageFlipping(&r200->radeon); + R200_STATECHANGE(r200, ctx); + r200->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] = + r200->radeon.state.color.drawOffset + + r200->radeon.radeonScreen->fbLocation; + r200->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = + r200->radeon.state.color.drawPitch; + + if (r200->radeon.glCtx->Color._DrawDestMask == DD_BACK_LEFT_BIT) + r200SetCliprects(r200, GL_BACK_LEFT); + else + r200SetCliprects(r200, GL_FRONT_LEFT); + r200UpdateViewportOffset(r200->radeon.glCtx); + r200->radeon.lastStamp = dPriv->lastStamp; + } + + for (i = 0; i < r200->nr_heaps; i++) { + DRI_AGE_TEXTURES(r200->texture_heaps[i]); + } +} + +static void r300RegainedLock(radeonContextPtr radeon) +{ + __DRIdrawablePrivate *dPriv = radeon->dri.drawable; + __DRIscreenPrivate *sPriv = radeon->dri.screen; + int i; + + if (radeon->lastStamp != dPriv->lastStamp) { + radeonUpdatePageFlipping(radeon); + +#if 0 + if (radeon->glCtx->Color._DrawDestMask == DD_BACK_LEFT_BIT) + r200SetCliprects(r200, GL_BACK_LEFT); + else + r200SetCliprects(r200, GL_FRONT_LEFT); + r200UpdateViewportOffset(r200->radeon.glCtx); +#endif + radeon->lastStamp = dPriv->lastStamp; + } + +#if 0 + for (i = 0; i < r200->nr_heaps; i++) { + DRI_AGE_TEXTURES(r200->texture_heaps[i]); + } +#endif +} + +/* Update the hardware state. This is called if another context has + * grabbed the hardware lock, which includes the X server. This + * function also updates the driver's window state after the X server + * moves, resizes or restacks a window -- the change will be reflected + * in the drawable position and clip rects. Since the X server grabs + * the hardware lock when it changes the window state, this routine will + * automatically be called after such a change. + */ +void radeonGetLock(radeonContextPtr radeon, GLuint flags) +{ + __DRIdrawablePrivate *dPriv = radeon->dri.drawable; + __DRIscreenPrivate *sPriv = radeon->dri.screen; + drm_radeon_sarea_t *sarea = radeon->sarea; + + drmGetLock(radeon->dri.fd, radeon->dri.hwContext, flags); + + /* The window might have moved, so we might need to get new clip + * rects. + * + * NOTE: This releases and regrabs the hw lock to allow the X server + * to respond to the DRI protocol request for new drawable info. + * Since the hardware state depends on having the latest drawable + * clip rects, all state checking must be done _after_ this call. + */ + DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); + + if (sarea->ctx_owner != radeon->dri.hwContext) + sarea->ctx_owner = radeon->dri.hwContext; + + if (IS_FAMILY_R300(radeon)) + r300RegainedLock(radeon); + else + r200RegainedLock((r200ContextPtr)radeon); + + radeon->lost_context = GL_TRUE; +} diff --git a/src/mesa/drivers/dri/r300/radeon_lock.h b/src/mesa/drivers/dri/r300/radeon_lock.h new file mode 100644 index 00000000000..3359d7822c0 --- /dev/null +++ b/src/mesa/drivers/dri/r300/radeon_lock.h @@ -0,0 +1,113 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Keith Whitwell + */ + +#ifndef __RADEON_LOCK_H__ +#define __RADEON_LOCK_H__ + +#include "r200_ioctl.h" + +extern void radeonGetLock(radeonContextPtr radeon, GLuint flags); + +/* Turn DEBUG_LOCKING on to find locking conflicts. + */ +#define DEBUG_LOCKING 0 + +#if DEBUG_LOCKING +extern char *prevLockFile; +extern int prevLockLine; + +#define DEBUG_LOCK() \ + do { \ + prevLockFile = (__FILE__); \ + prevLockLine = (__LINE__); \ + } while (0) + +#define DEBUG_RESET() \ + do { \ + prevLockFile = 0; \ + prevLockLine = 0; \ + } while (0) + +#define DEBUG_CHECK_LOCK() \ + do { \ + if ( prevLockFile ) { \ + fprintf( stderr, \ + "LOCK SET!\n\tPrevious %s:%d\n\tCurrent: %s:%d\n", \ + prevLockFile, prevLockLine, __FILE__, __LINE__ ); \ + exit( 1 ); \ + } \ + } while (0) + +#else + +#define DEBUG_LOCK() +#define DEBUG_RESET() +#define DEBUG_CHECK_LOCK() + +#endif + +/* + * !!! We may want to separate locks from locks with validation. This + * could be used to improve performance for those things commands that + * do not do any drawing !!! + */ + +/* Lock the hardware and validate our state. + */ +#define LOCK_HARDWARE( radeon ) \ + do { \ + char __ret = 0; \ + DEBUG_CHECK_LOCK(); \ + DRM_CAS( (radeon)->dri.hwLock, (radeon)->dri.hwContext, \ + (DRM_LOCK_HELD | (radeon)->dri.hwContext), __ret ); \ + if ( __ret ) \ + radeonGetLock( (radeon), 0 ); \ + DEBUG_LOCK(); \ + } while (0) + +#define UNLOCK_HARDWARE( radeon ) \ + do { \ + DRM_UNLOCK( (radeon)->dri.fd, \ + (radeon)->dri.hwLock, \ + (radeon)->dri.hwContext ); \ + DEBUG_RESET(); \ + if (IS_FAMILY_R200((radeon))) { \ + r200ContextPtr __r200 = (r200ContextPtr)(radeon); \ + if (__r200->save_on_next_unlock) \ + r200SaveHwState( __r200 ); \ + __r200->save_on_next_unlock = GL_FALSE; \ + } \ + } while (0) + +#endif /* __RADEON_LOCK_H__ */ diff --git a/src/mesa/drivers/dri/r300/radeon_screen.c b/src/mesa/drivers/dri/r300/radeon_screen.c new file mode 100644 index 00000000000..557f3a341f2 --- /dev/null +++ b/src/mesa/drivers/dri/r300/radeon_screen.c @@ -0,0 +1,817 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/** + * \file radeon_screen.c + * Screen initialization functions for the R200 driver. + * + * \author Keith Whitwell + */ + +#include + +#include "glheader.h" +#include "imports.h" +#include "context.h" + +#define STANDALONE_MMIO +#include "radeon_screen.h" +#include "r200_context.h" +#include "r300_context.h" +#include "radeon_ioctl.h" +#include "r200_ioctl.h" +#include "radeon_macros.h" +#include "radeon_reg.h" + +#include "utils.h" +#include "vblank.h" +#include "GL/internal/dri_interface.h" + +/* R200 configuration + */ +#include "xmlpool.h" + +const char __driR200ConfigOptions[] = +DRI_CONF_BEGIN + DRI_CONF_SECTION_PERFORMANCE + DRI_CONF_TCL_MODE(DRI_CONF_TCL_CODEGEN) + DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) + DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) + DRI_CONF_MAX_TEXTURE_UNITS(4, 2, 6) + DRI_CONF_SECTION_END + DRI_CONF_SECTION_QUALITY + DRI_CONF_TEXTURE_DEPTH(DRI_CONF_TEXTURE_DEPTH_FB) + DRI_CONF_DEF_MAX_ANISOTROPY(1.0, "1.0,2.0,4.0,8.0,16.0") + DRI_CONF_NO_NEG_LOD_BIAS(false) + DRI_CONF_COLOR_REDUCTION(DRI_CONF_COLOR_REDUCTION_DITHER) + DRI_CONF_ROUND_MODE(DRI_CONF_ROUND_TRUNC) + DRI_CONF_DITHER_MODE(DRI_CONF_DITHER_XERRORDIFF) + DRI_CONF_SECTION_END + DRI_CONF_SECTION_DEBUG + DRI_CONF_NO_RAST(false) + DRI_CONF_SECTION_END + DRI_CONF_SECTION_SOFTWARE + DRI_CONF_ARB_VERTEX_PROGRAM(true) + DRI_CONF_NV_VERTEX_PROGRAM(false) + DRI_CONF_SECTION_END +DRI_CONF_END; +static const GLuint __driR200NConfigOptions = 13; + +/* TODO: integrate these into xmlpool.h! */ +#define DRI_CONF_MAX_TEXTURE_IMAGE_UNITS(def,min,max) \ +DRI_CONF_OPT_BEGIN_V(texture_image_units,int,def, # min ":" # max ) \ + DRI_CONF_DESC(en,"Number of texture image units") \ + DRI_CONF_DESC(de,"Anzahl der Textureinheiten") \ +DRI_CONF_OPT_END + +#define DRI_CONF_MAX_TEXTURE_COORD_UNITS(def,min,max) \ +DRI_CONF_OPT_BEGIN_V(texture_coord_units,int,def, # min ":" # max ) \ + DRI_CONF_DESC(en,"Number of texture coordinate units") \ + DRI_CONF_DESC(de,"Anzahl der Texturkoordinateneinheiten") \ +DRI_CONF_OPT_END + +#define DRI_CONF_COMMAND_BUFFER_SIZE(def,min,max) \ +DRI_CONF_OPT_BEGIN_V(command_buffer_size,int,def, # min ":" # max ) \ + DRI_CONF_DESC(en,"Size of command buffer (in KB)") \ + DRI_CONF_DESC(de,"Grösse des Befehlspuffers (in KB)") \ +DRI_CONF_OPT_END + + +const char __driR300ConfigOptions[] = +DRI_CONF_BEGIN + DRI_CONF_SECTION_PERFORMANCE + DRI_CONF_TCL_MODE(DRI_CONF_TCL_CODEGEN) + DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) + DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) + DRI_CONF_MAX_TEXTURE_IMAGE_UNITS(16, 2, 16) + DRI_CONF_MAX_TEXTURE_COORD_UNITS(8, 2, 8) + DRI_CONF_COMMAND_BUFFER_SIZE(8, 8, 32) + DRI_CONF_SECTION_END + DRI_CONF_SECTION_QUALITY + DRI_CONF_TEXTURE_DEPTH(DRI_CONF_TEXTURE_DEPTH_FB) + DRI_CONF_DEF_MAX_ANISOTROPY(1.0, "1.0,2.0,4.0,8.0,16.0") + DRI_CONF_NO_NEG_LOD_BIAS(false) + DRI_CONF_COLOR_REDUCTION(DRI_CONF_COLOR_REDUCTION_DITHER) + DRI_CONF_ROUND_MODE(DRI_CONF_ROUND_TRUNC) + DRI_CONF_DITHER_MODE(DRI_CONF_DITHER_XERRORDIFF) + DRI_CONF_SECTION_END + DRI_CONF_SECTION_DEBUG + DRI_CONF_NO_RAST(false) + DRI_CONF_SECTION_END +DRI_CONF_END; +static const GLuint __driR300NConfigOptions = 13; + + +#ifndef RADEON_DEBUG +int RADEON_DEBUG = 0; +#endif + +static const struct dri_debug_control debug_control[] = { + {"fall", DEBUG_FALLBACKS}, + {"tex", DEBUG_TEXTURE}, + {"ioctl", DEBUG_IOCTL}, + {"prim", DEBUG_PRIMS}, + {"vert", DEBUG_VERTS}, + {"state", DEBUG_STATE}, + {"code", DEBUG_CODEGEN}, + {"vfmt", DEBUG_VFMT}, + {"vtxf", DEBUG_VFMT}, + {"verb", DEBUG_VERBOSE}, + {"dri", DEBUG_DRI}, + {"dma", DEBUG_DMA}, + {"san", DEBUG_SANITY}, + {"sync", DEBUG_SYNC}, + {"pix", DEBUG_PIXEL}, + {"mem", DEBUG_MEMORY}, + {NULL, 0} +}; + +#if 1 +/* Including xf86PciInfo.h introduces a bunch of errors... + */ +#define PCI_CHIP_R200_QD 0x5144 /* why do they have r200 names? */ +#define PCI_CHIP_R200_QE 0x5145 /* Those are all standard radeons */ +#define PCI_CHIP_R200_QF 0x5146 +#define PCI_CHIP_R200_QG 0x5147 +#define PCI_CHIP_R200_QY 0x5159 +#define PCI_CHIP_R200_QZ 0x515A +#define PCI_CHIP_R200_LW 0x4C57 +#define PCI_CHIP_R200_LY 0x4C59 +#define PCI_CHIP_R200_LZ 0x4C5A +#define PCI_CHIP_RV200_QW 0x5157 /* Radeon 7500 - not an R200 at all */ +#define PCI_CHIP_RV200_QX 0x5158 +#define PCI_CHIP_RS100_4136 0x4136 /* IGP RS100, RS200, RS250 are not R200 */ +#define PCI_CHIP_RS200_4137 0x4137 +#define PCI_CHIP_RS250_4237 0x4237 +#define PCI_CHIP_RS100_4336 0x4336 +#define PCI_CHIP_RS200_4337 0x4337 +#define PCI_CHIP_RS250_4437 0x4437 +#define PCI_CHIP_RS300_5834 0x5834 /* All RS300's are R200 */ +#define PCI_CHIP_RS300_5835 0x5835 +#define PCI_CHIP_RS300_5836 0x5836 +#define PCI_CHIP_RS300_5837 0x5837 +#define PCI_CHIP_R200_BB 0x4242 /* r200 (non-derived) start */ +#define PCI_CHIP_R200_BC 0x4243 +#define PCI_CHIP_R200_QH 0x5148 +#define PCI_CHIP_R200_QI 0x5149 +#define PCI_CHIP_R200_QJ 0x514A +#define PCI_CHIP_R200_QK 0x514B +#define PCI_CHIP_R200_QL 0x514C +#define PCI_CHIP_R200_QM 0x514D +#define PCI_CHIP_R200_QN 0x514E +#define PCI_CHIP_R200_QO 0x514F /* r200 (non-derived) end */ +/* are the R200 Qh (0x5168) and following needed too? They are not in xf86PciInfo.h + but in the pci database. Maybe just secondary ports or something ? */ + +#define PCI_CHIP_R300_AD 0x4144 +#define PCI_CHIP_R300_AE 0x4145 +#define PCI_CHIP_R300_AF 0x4146 +#define PCI_CHIP_R300_AG 0x4147 +#define PCI_CHIP_RV350_AP 0x4150 +#define PCI_CHIP_RV350_NP 0x4E50 +#define PCI_CHIP_R300_ND 0x4E44 +#define PCI_CHIP_R300_NE 0x4E45 +#define PCI_CHIP_R300_NF 0x4E46 +#define PCI_CHIP_R300_NG 0x4E47 +#define PCI_CHIP_R350_NH 0x4E48 +#endif + +#ifdef USE_NEW_INTERFACE +static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; +#endif /* USE_NEW_INTERFACE */ + +static radeonScreenPtr __radeonScreen; + +static int getSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo); + +#ifdef USE_NEW_INTERFACE +static __GLcontextModes *radeonFillInModes(unsigned pixel_bits, + unsigned depth_bits, + unsigned stencil_bits, + GLboolean have_back_buffer) +{ + __GLcontextModes *modes; + __GLcontextModes *m; + unsigned num_modes; + unsigned depth_buffer_factor; + unsigned back_buffer_factor; + GLenum fb_format; + GLenum fb_type; + + /* Right now GLX_SWAP_COPY_OML isn't supported, but it would be easy + * enough to add support. Basically, if a context is created with an + * fbconfig where the swap method is GLX_SWAP_COPY_OML, pageflipping + * will never be used. + */ + static const GLenum back_buffer_modes[] = { + GLX_NONE, GLX_SWAP_UNDEFINED_OML /*, GLX_SWAP_COPY_OML */ + }; + + uint8_t depth_bits_array[2]; + uint8_t stencil_bits_array[2]; + + depth_bits_array[0] = depth_bits; + depth_bits_array[1] = depth_bits; + + /* Just like with the accumulation buffer, always provide some modes + * with a stencil buffer. It will be a sw fallback, but some apps won't + * care about that. + */ + stencil_bits_array[0] = 0; + stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits; + + depth_buffer_factor = ((depth_bits != 0) + || (stencil_bits != 0)) ? 2 : 1; + back_buffer_factor = (have_back_buffer) ? 2 : 1; + + num_modes = depth_buffer_factor * back_buffer_factor * 4; + + if (pixel_bits == 16) { + fb_format = GL_RGB; + fb_type = GL_UNSIGNED_SHORT_5_6_5; + } else { + fb_format = GL_BGRA; + fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; + } + + modes = (*create_context_modes) (num_modes, sizeof(__GLcontextModes)); + m = modes; + if (!driFillInModes(&m, fb_format, fb_type, + depth_bits_array, stencil_bits_array, + depth_buffer_factor, back_buffer_modes, + back_buffer_factor, GLX_TRUE_COLOR)) { + fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, + __LINE__); + return NULL; + } + + if (!driFillInModes(&m, fb_format, fb_type, + depth_bits_array, stencil_bits_array, + depth_buffer_factor, back_buffer_modes, + back_buffer_factor, GLX_DIRECT_COLOR)) { + fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, + __LINE__); + return NULL; + } + + /* Mark the visual as slow if there are "fake" stencil bits. + */ + for (m = modes; m != NULL; m = m->next) { + if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) { + m->visualRating = GLX_SLOW_CONFIG; + } + } + + return modes; +} +#endif /* USE_NEW_INTERFACE */ + +/* Create the device specific screen private data struct. + */ +static radeonScreenPtr radeonCreateScreen(__DRIscreenPrivate * sPriv) +{ + radeonScreenPtr screen; + RADEONDRIPtr dri_priv = (RADEONDRIPtr) sPriv->pDevPriv; + unsigned char *RADEONMMIO; + + /* Allocate the private area */ + screen = (radeonScreenPtr) CALLOC(sizeof(*screen)); + if (!screen) { + __driUtilMessage + ("%s: Could not allocate memory for screen structure", + __FUNCTION__); + return NULL; + } + +#if DO_DEBUG + RADEON_DEBUG = driParseDebugString(getenv("RADEON_DEBUG"), debug_control); +#endif + + /* Get family and potential quirks from the PCI device ID. + */ + switch (dri_priv->deviceID) { + case PCI_CHIP_R200_QD: + case PCI_CHIP_R200_QE: + case PCI_CHIP_R200_QF: + case PCI_CHIP_R200_QG: + case PCI_CHIP_R200_QY: + case PCI_CHIP_R200_QZ: + case PCI_CHIP_RV200_QW: + case PCI_CHIP_RV200_QX: + case PCI_CHIP_R200_LW: + case PCI_CHIP_R200_LY: + case PCI_CHIP_R200_LZ: + case PCI_CHIP_RS100_4136: + case PCI_CHIP_RS200_4137: + case PCI_CHIP_RS250_4237: + case PCI_CHIP_RS100_4336: + case PCI_CHIP_RS200_4337: + case PCI_CHIP_RS250_4437: + __driUtilMessage("radeonCreateScreen(): Device isn't an r200!\n"); + FREE(screen); + return NULL; + + case PCI_CHIP_RS300_5834: + case PCI_CHIP_RS300_5835: + case PCI_CHIP_RS300_5836: + case PCI_CHIP_RS300_5837: + screen->chipset = RADEON_CHIP_UNREAL_R200; + break; + + case PCI_CHIP_R200_BB: + case PCI_CHIP_R200_BC: + case PCI_CHIP_R200_QH: + case PCI_CHIP_R200_QI: + case PCI_CHIP_R200_QJ: + case PCI_CHIP_R200_QK: + case PCI_CHIP_R200_QL: + case PCI_CHIP_R200_QM: + case PCI_CHIP_R200_QN: + case PCI_CHIP_R200_QO: + screen->chipset = RADEON_CHIP_REAL_R200 | RADEON_CHIPSET_TCL; + break; + + /* TODO: Check all those chips for the exact flags required. + */ + case PCI_CHIP_R300_AD: + case PCI_CHIP_R300_AE: + case PCI_CHIP_R300_AF: + case PCI_CHIP_R300_AG: + case PCI_CHIP_RV350_AP: + case PCI_CHIP_RV350_NP: + screen->chipset = RADEON_CHIP_RV350; + break; + + case PCI_CHIP_R300_ND: /* confirmed -- nh */ + case PCI_CHIP_R300_NE: + case PCI_CHIP_R300_NF: + case PCI_CHIP_R300_NG: + case PCI_CHIP_R350_NH: + screen->chipset = RADEON_CHIP_R300; + break; + + default: + fprintf(stderr, + "Unknown device ID %04X, please report. Assuming plain R300.\n", + dri_priv->deviceID); + screen->chipset = RADEON_CHIP_R300; + } + + /* Parse configuration */ + if (GET_CHIP(screen) >= RADEON_CHIP_R300) { + driParseOptionInfo(&screen->optionCache, + __driR300ConfigOptions, __driR300NConfigOptions); + } else { + driParseOptionInfo(&screen->optionCache, + __driR200ConfigOptions, __driR200NConfigOptions); + } + + /* This is first since which regions we map depends on whether or + * not we are using a PCI card. + */ + screen->IsPCI = dri_priv->IsPCI; + + { + int ret; + drm_radeon_getparam_t gp; + + gp.param = RADEON_PARAM_GART_BUFFER_OFFSET; + gp.value = &screen->gart_buffer_offset; + + ret = drmCommandWriteRead(sPriv->fd, DRM_RADEON_GETPARAM, + &gp, sizeof(gp)); + if (ret) { + FREE(screen); + fprintf(stderr, + "drmRadeonGetParam (RADEON_PARAM_GART_BUFFER_OFFSET): %d\n", + ret); + return NULL; + } + + if (sPriv->drmMinor >= 6) { + gp.param = RADEON_PARAM_GART_BASE; + gp.value = &screen->gart_base; + + ret = + drmCommandWriteRead(sPriv->fd, DRM_RADEON_GETPARAM, + &gp, sizeof(gp)); + if (ret) { + FREE(screen); + fprintf(stderr, + "drmR200GetParam (RADEON_PARAM_GART_BASE): %d\n", + ret); + return NULL; + } + + gp.param = RADEON_PARAM_IRQ_NR; + gp.value = &screen->irq; + + ret = + drmCommandWriteRead(sPriv->fd, DRM_RADEON_GETPARAM, + &gp, sizeof(gp)); + if (ret) { + FREE(screen); + fprintf(stderr, + "drmRadeonGetParam (RADEON_PARAM_IRQ_NR): %d\n", + ret); + return NULL; + } + + /* Check if kernel module is new enough to support cube maps */ + screen->drmSupportsCubeMaps = (sPriv->drmMinor >= 7); + /* Check if kernel module is new enough to support blend color and + separate blend functions/equations */ + screen->drmSupportsBlendColor = (sPriv->drmMinor >= 11); + + } + } + + screen->mmio.handle = dri_priv->registerHandle; + screen->mmio.size = dri_priv->registerSize; + if (drmMap(sPriv->fd, + screen->mmio.handle, screen->mmio.size, &screen->mmio.map)) { + FREE(screen); + __driUtilMessage("%s: drmMap failed\n", __FUNCTION__); + return NULL; + } + + RADEONMMIO = screen->mmio.map; + + screen->status.handle = dri_priv->statusHandle; + screen->status.size = dri_priv->statusSize; + if (drmMap(sPriv->fd, + screen->status.handle, + screen->status.size, &screen->status.map)) { + drmUnmap(screen->mmio.map, screen->mmio.size); + FREE(screen); + __driUtilMessage("%s: drmMap (2) failed\n", __FUNCTION__); + return NULL; + } + screen->scratch = (__volatile__ uint32_t *) + ((GLubyte *) screen->status.map + RADEON_SCRATCH_REG_OFFSET); + + screen->buffers = drmMapBufs(sPriv->fd); + if (!screen->buffers) { + drmUnmap(screen->status.map, screen->status.size); + drmUnmap(screen->mmio.map, screen->mmio.size); + FREE(screen); + __driUtilMessage("%s: drmMapBufs failed\n", __FUNCTION__); + return NULL; + } + + if (dri_priv->gartTexHandle && dri_priv->gartTexMapSize) { + + screen->gartTextures.handle = dri_priv->gartTexHandle; + screen->gartTextures.size = dri_priv->gartTexMapSize; + if (drmMap(sPriv->fd, + screen->gartTextures.handle, + screen->gartTextures.size, + (drmAddressPtr) & screen->gartTextures.map)) { + drmUnmapBufs(screen->buffers); + drmUnmap(screen->status.map, screen->status.size); + drmUnmap(screen->mmio.map, screen->mmio.size); + FREE(screen); + __driUtilMessage + ("%s: drmMAP failed for GART texture area\n", + __FUNCTION__); + return NULL; + } + + screen->gart_texture_offset = + dri_priv->gartTexOffset + + (screen->IsPCI ? INREG(RADEON_AIC_LO_ADDR) + : ((INREG(RADEON_MC_AGP_LOCATION) & 0x0ffffU) << 16)); + } + + screen->cpp = dri_priv->bpp / 8; + screen->AGPMode = dri_priv->AGPMode; + + screen->fbLocation = (INREG(RADEON_MC_FB_LOCATION) & 0xffff) << 16; + + if (sPriv->drmMinor >= 10) { + drm_radeon_setparam_t sp; + + sp.param = RADEON_SETPARAM_FB_LOCATION; + sp.value = screen->fbLocation; + + drmCommandWrite(sPriv->fd, DRM_RADEON_SETPARAM, + &sp, sizeof(sp)); + } + + screen->frontOffset = dri_priv->frontOffset; + screen->frontPitch = dri_priv->frontPitch; + screen->backOffset = dri_priv->backOffset; + screen->backPitch = dri_priv->backPitch; + screen->depthOffset = dri_priv->depthOffset; + screen->depthPitch = dri_priv->depthPitch; + + screen->texOffset[RADEON_LOCAL_TEX_HEAP] = dri_priv->textureOffset + + screen->fbLocation; + screen->texSize[RADEON_LOCAL_TEX_HEAP] = dri_priv->textureSize; + screen->logTexGranularity[RADEON_LOCAL_TEX_HEAP] = + dri_priv->log2TexGran; + + if (!screen->gartTextures.map) { + screen->numTexHeaps = RADEON_NR_TEX_HEAPS - 1; + screen->texOffset[RADEON_GART_TEX_HEAP] = 0; + screen->texSize[RADEON_GART_TEX_HEAP] = 0; + screen->logTexGranularity[RADEON_GART_TEX_HEAP] = 0; + } else { + screen->numTexHeaps = RADEON_NR_TEX_HEAPS; + screen->texOffset[RADEON_GART_TEX_HEAP] = + screen->gart_texture_offset; + screen->texSize[RADEON_GART_TEX_HEAP] = + dri_priv->gartTexMapSize; + screen->logTexGranularity[RADEON_GART_TEX_HEAP] = + dri_priv->log2GARTTexGran; + } + + screen->driScreen = sPriv; + screen->sarea_priv_offset = dri_priv->sarea_priv_offset; + + if (driCompareGLXAPIVersion(20030813) >= 0) { + PFNGLXSCRENABLEEXTENSIONPROC glx_enable_extension = + (PFNGLXSCRENABLEEXTENSIONPROC) + glXGetProcAddress((const GLubyte *) + "__glXScrEnableExtension"); + void *const psc = sPriv->psc->screenConfigs; + + if (glx_enable_extension != NULL) { + if (screen->irq != 0) { + (*glx_enable_extension) (psc, + "GLX_SGI_swap_control"); + (*glx_enable_extension) (psc, + "GLX_SGI_video_sync"); + (*glx_enable_extension) (psc, + "GLX_MESA_swap_control"); + } + + (*glx_enable_extension) (psc, + "GLX_MESA_swap_frame_usage"); + + if (driCompareGLXAPIVersion(20030818) >= 0) { + sPriv->psc->allocateMemory = + (void *)r200AllocateMemoryMESA; + sPriv->psc->freeMemory = + (void *)r200FreeMemoryMESA; + sPriv->psc->memoryOffset = + (void *)r200GetMemoryOffsetMESA; + + (*glx_enable_extension) (psc, + "GLX_MESA_allocate_memory"); + } + + if (driCompareGLXAPIVersion(20030915) >= 0) { + (*glx_enable_extension) (psc, + "GLX_SGIX_fbconfig"); + (*glx_enable_extension) (psc, + "GLX_OML_swap_method"); + } + } + } + return screen; +} + +/* Destroy the device specific screen private data struct. + */ +static void radeonDestroyScreen(__DRIscreenPrivate * sPriv) +{ + radeonScreenPtr screen = (radeonScreenPtr) sPriv->private; + + if (!screen) + return; + + if (screen->gartTextures.map) { + drmUnmap(screen->gartTextures.map, screen->gartTextures.size); + } + drmUnmapBufs(screen->buffers); + drmUnmap(screen->status.map, screen->status.size); + drmUnmap(screen->mmio.map, screen->mmio.size); + + /* free all option information */ + driDestroyOptionInfo(&screen->optionCache); + + FREE(screen); + sPriv->private = NULL; +} + +/* Initialize the driver specific screen private data. + */ +static GLboolean radeonInitDriver(__DRIscreenPrivate * sPriv) +{ + __radeonScreen = radeonCreateScreen(sPriv); + + sPriv->private = (void *)__radeonScreen; + + return sPriv->private ? GL_TRUE : GL_FALSE; +} + +/** + * Create and initialize the Mesa and driver specific pixmap buffer + * data. + * + * \todo This function (and its interface) will need to be updated to support + * pbuffers. + */ +static GLboolean +radeonCreateBuffer(__DRIscreenPrivate * driScrnPriv, + __DRIdrawablePrivate * driDrawPriv, + const __GLcontextModes * mesaVis, GLboolean isPixmap) +{ + if (isPixmap) { + return GL_FALSE; /* not implemented */ + } else { + const GLboolean swDepth = GL_FALSE; + const GLboolean swAlpha = GL_FALSE; + const GLboolean swAccum = mesaVis->accumRedBits > 0; + const GLboolean swStencil = mesaVis->stencilBits > 0 && + mesaVis->depthBits != 24; + driDrawPriv->driverPrivate = (void *) + _mesa_create_framebuffer(mesaVis, + swDepth, + swStencil, swAccum, swAlpha); + return (driDrawPriv->driverPrivate != NULL); + } +} + +static void radeonDestroyBuffer(__DRIdrawablePrivate * driDrawPriv) +{ + _mesa_destroy_framebuffer((GLframebuffer *) (driDrawPriv-> + driverPrivate)); +} + + +/** + * Choose the appropriate CreateContext function based on the chipset. + */ +static GLboolean radeonCreateContext(const __GLcontextModes * glVisual, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPriv) +{ + __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; + radeonScreenPtr screen = (radeonScreenPtr) (sPriv->private); + int chip = GET_CHIP(screen); + + if (chip >= RADEON_CHIP_R300) + return r300CreateContext(glVisual, driContextPriv, sharedContextPriv); + else + return r200CreateContext(glVisual, driContextPriv, sharedContextPriv); +} + + +/** + * Choose the appropriate DestroyContext function based on the chipset. + */ +static void radeonDestroyContext(__DRIcontextPrivate * driContextPriv) +{ + radeonContextPtr radeon = (radeonContextPtr) driContextPriv->driverPrivate; + int chip = GET_CHIP(radeon->radeonScreen); + + if (chip >= RADEON_CHIP_R300) + return r300DestroyContext(driContextPriv); + else + return r200DestroyContext(driContextPriv); +} + + +static const struct __DriverAPIRec radeonAPI = { + .InitDriver = radeonInitDriver, + .DestroyScreen = radeonDestroyScreen, + .CreateContext = radeonCreateContext, + .DestroyContext = radeonDestroyContext, + .CreateBuffer = radeonCreateBuffer, + .DestroyBuffer = radeonDestroyBuffer, + .SwapBuffers = radeonSwapBuffers, + .MakeCurrent = radeonMakeCurrent, + .UnbindContext = radeonUnbindContext, + .GetSwapInfo = getSwapInfo, + .GetMSC = driGetMSC32, + .WaitForMSC = driWaitForMSC32, + .WaitForSBC = NULL, + .SwapBuffersMSC = NULL +}; + +/* + * This is the bootstrap function for the driver. + * The __driCreateScreen name is the symbol that libGL.so fetches. + * Return: pointer to a __DRIscreenPrivate. + * + */ +#if !defined(DRI_NEW_INTERFACE_ONLY) +void *__driCreateScreen(Display * dpy, int scrn, __DRIscreen * psc, + int numConfigs, __GLXvisualConfig * config) +{ + __DRIscreenPrivate *psp; + psp = + __driUtilCreateScreen(dpy, scrn, psc, numConfigs, config, &radeonAPI); + return (void *)psp; +} +#endif /* !defined(DRI_NEW_INTERFACE_ONLY) */ + +/** + * This is the bootstrap function for the driver. libGL supplies all of the + * requisite information about the system, and the driver initializes itself. + * This routine also fills in the linked list pointed to by \c driver_modes + * with the \c __GLcontextModes that the driver can support for windows or + * pbuffers. + * + * \return A pointer to a \c __DRIscreenPrivate on success, or \c NULL on + * failure. + */ +#ifdef USE_NEW_INTERFACE +void *__driCreateNewScreen(__DRInativeDisplay * dpy, int scrn, + __DRIscreen * psc, const __GLcontextModes * modes, + const __DRIversion * ddx_version, + const __DRIversion * dri_version, + const __DRIversion * drm_version, + const __DRIframebuffer * frame_buffer, + drmAddress pSAREA, int fd, int internal_api_version, + __GLcontextModes ** driver_modes) +{ + __DRIscreenPrivate *psp; + static const __DRIversion ddx_expected = { 4, 0, 0 }; + static const __DRIversion dri_expected = { 4, 0, 0 }; + static const __DRIversion drm_expected = { 1, 11, 1 }; + + if (!driCheckDriDdxDrmVersions2("R300", + dri_version, &dri_expected, + ddx_version, &ddx_expected, + drm_version, &drm_expected)) { + return NULL; + } + + psp = __driUtilCreateNewScreen(dpy, scrn, psc, NULL, + ddx_version, dri_version, drm_version, + frame_buffer, pSAREA, fd, + internal_api_version, &radeonAPI); + if (psp != NULL) { + create_context_modes = (PFNGLXCREATECONTEXTMODES) + glXGetProcAddress((const GLubyte *) + "__glXCreateContextModes"); + if (create_context_modes != NULL) { + RADEONDRIPtr dri_priv = (RADEONDRIPtr) psp->pDevPriv; + *driver_modes = radeonFillInModes(dri_priv->bpp, + (dri_priv->bpp == + 16) ? 16 : 24, + (dri_priv->bpp == + 16) ? 0 : 8, + (dri_priv->backOffset != + dri_priv-> + depthOffset)); + } + } + + return (void *)psp; +} +#endif /* USE_NEW_INTERFACE */ + +/** + * Get information about previous buffer swaps. + */ +static int getSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) +{ + radeonContextPtr radeon; + + if ((dPriv == NULL) || (dPriv->driContextPriv == NULL) + || (dPriv->driContextPriv->driverPrivate == NULL) + || (sInfo == NULL)) { + return -1; + } + + radeon = (radeonContextPtr) dPriv->driContextPriv->driverPrivate; + sInfo->swap_count = radeon->swap_count; + sInfo->swap_ust = radeon->swap_ust; + sInfo->swap_missed_count = radeon->swap_missed_count; + + sInfo->swap_missed_usage = (sInfo->swap_missed_count != 0) + ? driCalculateSwapUsage(dPriv, 0, radeon->swap_missed_ust) + : 0.0; + + return 0; +} diff --git a/src/mesa/drivers/dri/r300/radeon_screen.h b/src/mesa/drivers/dri/r300/radeon_screen.h new file mode 100644 index 00000000000..a1a019162ab --- /dev/null +++ b/src/mesa/drivers/dri/r300/radeon_screen.h @@ -0,0 +1,111 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Keith Whitwell + */ + +#ifndef __RADEON_SCREEN_H__ +#define __RADEON_SCREEN_H__ + +#ifdef GLX_DIRECT_RENDERING + +#include "xf86drm.h" +#include "drm.h" +#include "radeon_drm.h" +#include "dri_util.h" +#include "xmlconfig.h" + +typedef struct { + drm_handle_t handle; /* Handle to the DRM region */ + drmSize size; /* Size of the DRM region */ + drmAddress map; /* Mapping of the DRM region */ +} radeonRegionRec, *radeonRegionPtr; + +/* chipset features */ +#define RADEON_CHIP_UNREAL_R200 0 +#define RADEON_CHIP_REAL_R200 1 +#define RADEON_CHIP_R300 2 +#define RADEON_CHIP_RV350 3 +#define RADEON_CHIP_MASK 0x0f + +#define RADEON_CHIPSET_TCL (1 << 8) + +#define GET_CHIP(screen) ((screen)->chipset & RADEON_CHIP_MASK) +#define IS_FAMILY_R200(radeon) (GET_CHIP((radeon)->radeonScreen) < RADEON_CHIP_R300) +#define IS_FAMILY_R300(radeon) (GET_CHIP((radeon)->radeonScreen) >= RADEON_CHIP_R300) + +#define R200_NR_TEX_HEAPS 2 + +typedef struct { + int chipset; + int cpp; + int IsPCI; /* Current card is a PCI card */ + int AGPMode; + unsigned int irq; /* IRQ number (0 means none) */ + + unsigned int fbLocation; + unsigned int frontOffset; + unsigned int frontPitch; + unsigned int backOffset; + unsigned int backPitch; + + unsigned int depthOffset; + unsigned int depthPitch; + + /* Shared texture data */ + int numTexHeaps; + int texOffset[R200_NR_TEX_HEAPS]; + int texSize[R200_NR_TEX_HEAPS]; + int logTexGranularity[R200_NR_TEX_HEAPS]; + + radeonRegionRec mmio; + radeonRegionRec status; + radeonRegionRec gartTextures; + + drmBufMapPtr buffers; + + __volatile__ int32_t *scratch; + + __DRIscreenPrivate *driScreen; + unsigned int sarea_priv_offset; + unsigned int gart_buffer_offset; /* offset in card memory space */ + unsigned int gart_texture_offset; /* offset in card memory space */ + unsigned int gart_base; + + GLboolean drmSupportsCubeMaps; /* need radeon kernel module >=1.7 */ + GLboolean drmSupportsBlendColor; /* need radeon kernel module >= 1.11 */ + + /* Configuration cache with default values for all contexts */ + driOptionCache optionCache; +} radeonScreenRec, *radeonScreenPtr; + +#endif +#endif /* __RADEON_SCREEN_H__ */ diff --git a/src/mesa/drivers/dri/r300/radeon_span.c b/src/mesa/drivers/dri/r300/radeon_span.c new file mode 100644 index 00000000000..9046320cc1b --- /dev/null +++ b/src/mesa/drivers/dri/r300/radeon_span.c @@ -0,0 +1,520 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Keith Whitwell + */ + +#include "glheader.h" +#include "imports.h" +#include "swrast/swrast.h" +#include "colormac.h" + +#include "r200_context.h" +#include "radeon_ioctl.h" +#include "r300_ioctl.h" +#include "radeon_span.h" + +#define DBG 0 + +#define LOCAL_VARS \ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); \ + radeonScreenPtr radeonScreen = radeon->radeonScreen; \ + __DRIscreenPrivate *sPriv = radeon->dri.screen; \ + __DRIdrawablePrivate *dPriv = radeon->dri.drawable; \ + GLuint pitch = radeonScreen->frontPitch * radeonScreen->cpp; \ + GLuint height = dPriv->h; \ + char *buf = (char *)(sPriv->pFB + \ + radeon->state.color.drawOffset + \ + (dPriv->x * radeonScreen->cpp) + \ + (dPriv->y * pitch)); \ + char *read_buf = (char *)(sPriv->pFB + \ + radeon->state.pixel.readOffset + \ + (dPriv->x * radeonScreen->cpp) + \ + (dPriv->y * pitch)); \ + GLuint p; \ + (void) read_buf; (void) buf; (void) p + +#define LOCAL_DEPTH_VARS \ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); \ + radeonScreenPtr radeonScreen = radeon->radeonScreen; \ + __DRIscreenPrivate *sPriv = radeon->dri.screen; \ + __DRIdrawablePrivate *dPriv = radeon->dri.drawable; \ + GLuint height = dPriv->h; \ + GLuint xo = dPriv->x; \ + GLuint yo = dPriv->y; \ + char *buf = (char *)(sPriv->pFB + radeon->radeonScreen->depthOffset); \ + GLuint pitch = radeon->radeonScreen->depthPitch; \ + (void) buf; (void) pitch + +#define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS + +#define CLIPPIXEL( _x, _y ) \ + ((_x >= minx) && (_x < maxx) && (_y >= miny) && (_y < maxy)) + +#define CLIPSPAN( _x, _y, _n, _x1, _n1, _i ) \ + if ( _y < miny || _y >= maxy ) { \ + _n1 = 0, _x1 = x; \ + } else { \ + _n1 = _n; \ + _x1 = _x; \ + if ( _x1 < minx ) _i += (minx-_x1), n1 -= (minx-_x1), _x1 = minx; \ + if ( _x1 + _n1 >= maxx ) n1 -= (_x1 + n1 - maxx); \ + } + +#define Y_FLIP( _y ) (height - _y - 1) + +#define HW_LOCK() + +#define HW_CLIPLOOP() \ + do { \ + __DRIdrawablePrivate *dPriv = radeon->dri.drawable; \ + int _nc = dPriv->numClipRects; \ + \ + while ( _nc-- ) { \ + int minx = dPriv->pClipRects[_nc].x1 - dPriv->x; \ + int miny = dPriv->pClipRects[_nc].y1 - dPriv->y; \ + int maxx = dPriv->pClipRects[_nc].x2 - dPriv->x; \ + int maxy = dPriv->pClipRects[_nc].y2 - dPriv->y; + +#define HW_ENDCLIPLOOP() \ + } \ + } while (0) + +#define HW_UNLOCK() + +/* ================================================================ + * Color buffer + */ + +/* 16 bit, RGB565 color spanline and pixel functions + */ +#define INIT_MONO_PIXEL(p, color) \ + p = PACK_COLOR_565( color[0], color[1], color[2] ) + +#define WRITE_RGBA( _x, _y, r, g, b, a ) \ + *(GLushort *)(buf + _x*2 + _y*pitch) = ((((int)r & 0xf8) << 8) | \ + (((int)g & 0xfc) << 3) | \ + (((int)b & 0xf8) >> 3)) + +#define WRITE_PIXEL( _x, _y, p ) \ + *(GLushort *)(buf + _x*2 + _y*pitch) = p + +#define READ_RGBA( rgba, _x, _y ) \ + do { \ + GLushort p = *(GLushort *)(read_buf + _x*2 + _y*pitch); \ + rgba[0] = ((p >> 8) & 0xf8) * 255 / 0xf8; \ + rgba[1] = ((p >> 3) & 0xfc) * 255 / 0xfc; \ + rgba[2] = ((p << 3) & 0xf8) * 255 / 0xf8; \ + rgba[3] = 0xff; \ + } while (0) + +#define TAG(x) radeon##x##_RGB565 +#include "spantmp.h" + +/* 32 bit, ARGB8888 color spanline and pixel functions + */ +#undef INIT_MONO_PIXEL +#define INIT_MONO_PIXEL(p, color) \ + p = PACK_COLOR_8888( color[3], color[0], color[1], color[2] ) + +#define WRITE_RGBA( _x, _y, r, g, b, a ) \ +do { \ + *(GLuint *)(buf + _x*4 + _y*pitch) = ((b << 0) | \ + (g << 8) | \ + (r << 16) | \ + (a << 24) ); \ +} while (0) + +#define WRITE_PIXEL( _x, _y, p ) \ +do { \ + *(GLuint *)(buf + _x*4 + _y*pitch) = p; \ +} while (0) + +#define READ_RGBA( rgba, _x, _y ) \ +do { \ + volatile GLuint *ptr = (volatile GLuint *)(read_buf + _x*4 + _y*pitch); \ + GLuint p = *ptr; \ + rgba[0] = (p >> 16) & 0xff; \ + rgba[1] = (p >> 8) & 0xff; \ + rgba[2] = (p >> 0) & 0xff; \ + rgba[3] = (p >> 24) & 0xff; \ +} while (0) + +#define TAG(x) radeon##x##_ARGB8888 +#include "spantmp.h" + +/* ================================================================ + * Depth buffer + */ + +/* The Radeon family has depth tiling on all the time, so we have to convert + * the x,y coordinates into the memory bus address (mba) in the same + * manner as the engine. In each case, the linear block address (ba) + * is calculated, and then wired with x and y to produce the final + * memory address. + */ + +#define BIT(x,b) ((x & (1<>b) +static GLuint radeon_mba_z32(radeonContextPtr radeon, GLint x, GLint y) +{ + GLuint pitch = radeon->radeonScreen->depthPitch; + GLuint b = + ((y & 0x3FF) >> 4) * ((pitch & 0xFFF) >> 5) + ((x & 0x3FF) >> 5); + GLuint a = + (BIT(x, 0) << 2) | (BIT(y, 0) << 3) | (BIT(x, 1) << 4) | (BIT(y, 1) + << 5) | + (BIT(x, 3) << 6) | (BIT(x, 4) << 7) | (BIT(x, 2) << 8) | (BIT(y, 2) + << 9) | + (BIT(y, 3) << 10) | + (((pitch & 0x20) ? (b & 0x01) : ((b & 0x01) ^ (BIT(y, 4)))) << 11) | + ((b >> 1) << 12); + return a; +} + +static GLuint radeon_mba_z16(radeonContextPtr radeon, GLint x, GLint y) +{ + GLuint pitch = radeon->radeonScreen->depthPitch; + GLuint b = + ((y & 0x3FF) >> 4) * ((pitch & 0xFFF) >> 6) + ((x & 0x3FF) >> 6); + GLuint a = + (BIT(x, 0) << 1) | (BIT(y, 0) << 2) | (BIT(x, 1) << 3) | (BIT(y, 1) + << 4) | + (BIT(x, 2) << 5) | (BIT(x, 4) << 6) | (BIT(x, 5) << 7) | (BIT(x, 3) + << 8) | + (BIT(y, 2) << 9) | (BIT(y, 3) << 10) | + (((pitch & 0x40) ? (b & 0x01) : ((b & 0x01) ^ (BIT(y, 4)))) << 11) | + ((b >> 1) << 12); + return a; +} + + +/* 16-bit depth buffer functions + */ +#define WRITE_DEPTH( _x, _y, d ) \ + *(GLushort *)(buf + radeon_mba_z16( radeon, _x + xo, _y + yo )) = d; + +#define READ_DEPTH( d, _x, _y ) \ + d = *(GLushort *)(buf + radeon_mba_z16( radeon, _x + xo, _y + yo )); + +#define TAG(x) radeon##x##_16_TILE +#include "depthtmp.h" + +/* 24 bit depth, 8 bit stencil depthbuffer functions + */ +#define WRITE_DEPTH( _x, _y, d ) \ +do { \ + GLuint offset = radeon_mba_z32( radeon, _x + xo, _y + yo ); \ + GLuint tmp = *(GLuint *)(buf + offset); \ + tmp &= 0xff000000; \ + tmp |= ((d) & 0x00ffffff); \ + *(GLuint *)(buf + offset) = tmp; \ +} while (0) + +#define READ_DEPTH( d, _x, _y ) \ + d = *(GLuint *)(buf + radeon_mba_z32( radeon, _x + xo, \ + _y + yo )) & 0x00ffffff; + +#define TAG(x) radeon##x##_24_8_TILE +#include "depthtmp.h" + +/* 16-bit depth buffer functions + */ +#define WRITE_DEPTH( _x, _y, d ) \ + *(GLushort *)(buf + (_x + xo)*2 + (_y + yo)*pitch ) = d; + +#define READ_DEPTH( d, _x, _y ) \ + d = *(GLushort *)(buf + (_x + xo)*2 + (_y + yo)*pitch ); + +#define TAG(x) radeon##x##_16_LINEAR +#include "depthtmp.h" + +/* 24 bit depth, 8 bit stencil depthbuffer functions + */ +#define WRITE_DEPTH( _x, _y, d ) \ +do { \ + GLuint offset = (_x + xo)*4 + (_y + yo)*pitch; \ + GLuint tmp = *(GLuint *)(buf + offset); \ + tmp &= 0xff000000; \ + tmp |= ((d) & 0x00ffffff); \ + *(GLuint *)(buf + offset) = tmp; \ +} while (0) + +#define READ_DEPTH( d, _x, _y ) \ + d = *(GLuint *)(buf + (_x + xo)*4 + (_y + yo)*pitch) & 0x00ffffff; + +#define TAG(x) radeon##x##_24_8_LINEAR +#include "depthtmp.h" + +/* ================================================================ + * Stencil buffer + */ + +/* 24 bit depth, 8 bit stencil depthbuffer functions + */ +#define WRITE_STENCIL( _x, _y, d ) \ +do { \ + GLuint offset = radeon_mba_z32( radeon, _x + xo, _y + yo ); \ + GLuint tmp = *(GLuint *)(buf + offset); \ + tmp &= 0x00ffffff; \ + tmp |= (((d) & 0xff) << 24); \ + *(GLuint *)(buf + offset) = tmp; \ +} while (0) + +#define READ_STENCIL( d, _x, _y ) \ +do { \ + GLuint offset = radeon_mba_z32( radeon, _x + xo, _y + yo ); \ + GLuint tmp = *(GLuint *)(buf + offset); \ + tmp &= 0xff000000; \ + d = tmp >> 24; \ +} while (0) + +#define TAG(x) radeon##x##_24_8_TILE +#include "stenciltmp.h" + +/* 24 bit depth, 8 bit stencil depthbuffer functions + */ +#define WRITE_STENCIL( _x, _y, d ) \ +do { \ + GLuint offset = (_x + xo)*4 + (_y + yo)*pitch; \ + GLuint tmp = *(GLuint *)(buf + offset); \ + tmp &= 0x00ffffff; \ + tmp |= (((d) & 0xff) << 24); \ + *(GLuint *)(buf + offset) = tmp; \ +} while (0) + +#define READ_STENCIL( d, _x, _y ) \ +do { \ + GLuint offset = (_x + xo)*4 + (_y + yo)*pitch; \ + GLuint tmp = *(GLuint *)(buf + offset); \ + tmp &= 0xff000000; \ + d = tmp >> 24; \ +} while (0) + +#define TAG(x) radeon##x##_24_8_LINEAR +#include "stenciltmp.h" + +/* + * This function is called to specify which buffer to read and write + * for software rasterization (swrast) fallbacks. This doesn't necessarily + * correspond to glDrawBuffer() or glReadBuffer() calls. + */ +static void radeonSetBuffer(GLcontext * ctx, + GLframebuffer * colorBuffer, GLuint bufferBit) +{ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + + switch (bufferBit) { + case DD_FRONT_LEFT_BIT: + if (radeon->doPageFlip && radeon->sarea->pfCurrentPage == 1) { + radeon->state.pixel.readOffset = + radeon->radeonScreen->backOffset; + radeon->state.pixel.readPitch = + radeon->radeonScreen->backPitch; + radeon->state.color.drawOffset = + radeon->radeonScreen->backOffset; + radeon->state.color.drawPitch = + radeon->radeonScreen->backPitch; + } else { + radeon->state.pixel.readOffset = + radeon->radeonScreen->frontOffset; + radeon->state.pixel.readPitch = + radeon->radeonScreen->frontPitch; + radeon->state.color.drawOffset = + radeon->radeonScreen->frontOffset; + radeon->state.color.drawPitch = + radeon->radeonScreen->frontPitch; + } + break; + case DD_BACK_LEFT_BIT: + if (radeon->doPageFlip && radeon->sarea->pfCurrentPage == 1) { + radeon->state.pixel.readOffset = + radeon->radeonScreen->frontOffset; + radeon->state.pixel.readPitch = + radeon->radeonScreen->frontPitch; + radeon->state.color.drawOffset = + radeon->radeonScreen->frontOffset; + radeon->state.color.drawPitch = + radeon->radeonScreen->frontPitch; + } else { + radeon->state.pixel.readOffset = + radeon->radeonScreen->backOffset; + radeon->state.pixel.readPitch = + radeon->radeonScreen->backPitch; + radeon->state.color.drawOffset = + radeon->radeonScreen->backOffset; + radeon->state.color.drawPitch = + radeon->radeonScreen->backPitch; + } + break; + default: + _mesa_problem(ctx, "Bad bufferBit in %s", __FUNCTION__); + break; + } +} + +/* Move locking out to get reasonable span performance (10x better + * than doing this in HW_LOCK above). WaitForIdle() is the main + * culprit. + */ + +static void radeonSpanRenderStart(GLcontext * ctx) +{ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + + if (IS_FAMILY_R200(radeon)) + R200_FIREVERTICES((r200ContextPtr)radeon); + else + r300Flush(ctx); + + LOCK_HARDWARE(radeon); + radeonWaitForIdleLocked(radeon); + + /* Read & rewrite the first pixel in the frame buffer. This should + * be a noop, right? In fact without this conform fails as reading + * from the framebuffer sometimes produces old results -- the + * on-card read cache gets mixed up and doesn't notice that the + * framebuffer has been updated. + * + * In the worst case this is buggy too as p might get the wrong + * value first time, so really need a hidden pixel somewhere for this. + */ + { + int p; + volatile int *read_buf = + (volatile int *)(radeon->dri.screen->pFB + + radeon->state.pixel.readOffset); + p = *read_buf; + *read_buf = p; + } +} + +static void radeonSpanRenderFinish(GLcontext * ctx) +{ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + + _swrast_flush(ctx); + UNLOCK_HARDWARE(radeon); +} + +void radeonInitSpanFuncs(GLcontext * ctx) +{ + r200ContextPtr rmesa = R200_CONTEXT(ctx); + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + struct swrast_device_driver *swdd = + _swrast_GetDeviceDriverReference(ctx); + + swdd->SetBuffer = radeonSetBuffer; + + switch (radeon->radeonScreen->cpp) { + case 2: + swdd->WriteRGBASpan = radeonWriteRGBASpan_RGB565; + swdd->WriteRGBSpan = radeonWriteRGBSpan_RGB565; + swdd->WriteMonoRGBASpan = radeonWriteMonoRGBASpan_RGB565; + swdd->WriteRGBAPixels = radeonWriteRGBAPixels_RGB565; + swdd->WriteMonoRGBAPixels = radeonWriteMonoRGBAPixels_RGB565; + swdd->ReadRGBASpan = radeonReadRGBASpan_RGB565; + swdd->ReadRGBAPixels = radeonReadRGBAPixels_RGB565; + break; + + case 4: + swdd->WriteRGBASpan = radeonWriteRGBASpan_ARGB8888; + swdd->WriteRGBSpan = radeonWriteRGBSpan_ARGB8888; + swdd->WriteMonoRGBASpan = radeonWriteMonoRGBASpan_ARGB8888; + swdd->WriteRGBAPixels = radeonWriteRGBAPixels_ARGB8888; + swdd->WriteMonoRGBAPixels = radeonWriteMonoRGBAPixels_ARGB8888; + swdd->ReadRGBASpan = radeonReadRGBASpan_ARGB8888; + swdd->ReadRGBAPixels = radeonReadRGBAPixels_ARGB8888; + break; + + default: + break; + } + + if (IS_FAMILY_R300(radeon)) + { + switch (radeon->glCtx->Visual.depthBits) { + case 16: + swdd->ReadDepthSpan = radeonReadDepthSpan_16_LINEAR; + swdd->WriteDepthSpan = radeonWriteDepthSpan_16_LINEAR; + swdd->WriteMonoDepthSpan = radeonWriteMonoDepthSpan_16_LINEAR; + swdd->ReadDepthPixels = radeonReadDepthPixels_16_LINEAR; + swdd->WriteDepthPixels = radeonWriteDepthPixels_16_LINEAR; + break; + + case 24: + swdd->ReadDepthSpan = radeonReadDepthSpan_24_8_LINEAR; + swdd->WriteDepthSpan = radeonWriteDepthSpan_24_8_LINEAR; + swdd->WriteMonoDepthSpan = radeonWriteMonoDepthSpan_24_8_LINEAR; + swdd->ReadDepthPixels = radeonReadDepthPixels_24_8_LINEAR; + swdd->WriteDepthPixels = radeonWriteDepthPixels_24_8_LINEAR; + + swdd->ReadStencilSpan = radeonReadStencilSpan_24_8_LINEAR; + swdd->WriteStencilSpan = radeonWriteStencilSpan_24_8_LINEAR; + swdd->ReadStencilPixels = radeonReadStencilPixels_24_8_LINEAR; + swdd->WriteStencilPixels = radeonWriteStencilPixels_24_8_LINEAR; + break; + + default: + break; + } + } + else + { + switch (radeon->glCtx->Visual.depthBits) { + case 16: + swdd->ReadDepthSpan = radeonReadDepthSpan_16_TILE; + swdd->WriteDepthSpan = radeonWriteDepthSpan_16_TILE; + swdd->WriteMonoDepthSpan = radeonWriteMonoDepthSpan_16_TILE; + swdd->ReadDepthPixels = radeonReadDepthPixels_16_TILE; + swdd->WriteDepthPixels = radeonWriteDepthPixels_16_TILE; + break; + + case 24: + swdd->ReadDepthSpan = radeonReadDepthSpan_24_8_TILE; + swdd->WriteDepthSpan = radeonWriteDepthSpan_24_8_TILE; + swdd->WriteMonoDepthSpan = radeonWriteMonoDepthSpan_24_8_TILE; + swdd->ReadDepthPixels = radeonReadDepthPixels_24_8_TILE; + swdd->WriteDepthPixels = radeonWriteDepthPixels_24_8_TILE; + + swdd->ReadStencilSpan = radeonReadStencilSpan_24_8_TILE; + swdd->WriteStencilSpan = radeonWriteStencilSpan_24_8_TILE; + swdd->ReadStencilPixels = radeonReadStencilPixels_24_8_TILE; + swdd->WriteStencilPixels = radeonWriteStencilPixels_24_8_TILE; + break; + + default: + break; + } + } + + swdd->SpanRenderStart = radeonSpanRenderStart; + swdd->SpanRenderFinish = radeonSpanRenderFinish; +} diff --git a/src/mesa/drivers/dri/r300/radeon_span.h b/src/mesa/drivers/dri/r300/radeon_span.h new file mode 100644 index 00000000000..e6a0f983f07 --- /dev/null +++ b/src/mesa/drivers/dri/r300/radeon_span.h @@ -0,0 +1,43 @@ +/* +Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. + +The Weather Channel (TM) funded Tungsten Graphics to develop the +initial release of the Radeon 8500 driver under the XFree86 license. +This notice must be preserved. + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial +portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. +IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Keith Whitwell + */ + +#ifndef __RADEON_SPAN_H__ +#define __RADEON_SPAN_H__ + +#ifdef GLX_DIRECT_RENDERING + +extern void radeonInitSpanFuncs(GLcontext * ctx); + +#endif +#endif