#ifndef R600_PIPE_H
#define R600_PIPE_H
-#include "util/u_blitter.h"
-#include "util/u_slab.h"
-#include "util/u_suballoc.h"
-#include "r600.h"
+#include "radeon/r600_pipe_common.h"
+#include "radeon/r600_cs.h"
+
#include "r600_llvm.h"
#include "r600_public.h"
-#include "r600_resource.h"
-#define R600_NUM_ATOMS 40
+#include "util/u_suballoc.h"
+#include "util/u_double_list.h"
+#include "util/u_transfer.h"
-#define R600_TRACE_CS 0
+#define R600_NUM_ATOMS 73
/* the number of CS dwords for flushing and drawing */
#define R600_MAX_FLUSH_CS_DWORDS 16
-#define R600_MAX_DRAW_CS_DWORDS 34
+#define R600_MAX_DRAW_CS_DWORDS 40
#define R600_TRACE_CS_DWORDS 7
#define R600_MAX_USER_CONST_BUFFERS 13
/* start driver buffers after user buffers */
#define R600_UCP_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS)
-#define R600_TXQ_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 1)
-#define R600_BUFFER_INFO_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 2)
+#define R600_BUFFER_INFO_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 1)
+#define R600_GS_RING_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 2)
+/* Currently R600_MAX_CONST_BUFFERS just fits on the hw, which has a limit
+ * of 16 const buffers.
+ * UCP/SAMPLE_POSITIONS are never accessed by same shader stage so they can use the same id.
+ *
+ * In order to support d3d 11 mandated minimum of 15 user const buffers
+ * we'd have to squash all use cases into one driver buffer.
+ */
+#define R600_SAMPLE_POSITIONS_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS)
-#define R600_MAX_CONST_BUFFER_SIZE 4096
+#define R600_MAX_CONST_BUFFER_SIZE (4096 * sizeof(float[4]))
#ifdef PIPE_ARCH_BIG_ENDIAN
#define R600_BIG_ENDIAN 1
#define R600_BIG_ENDIAN 0
#endif
-#define R600_MAP_BUFFER_ALIGNMENT 64
-
+struct r600_context;
struct r600_bytecode;
struct r600_shader_key;
-/* This encapsulates a state or an operation which can emitted into the GPU
- * command stream. It's not limited to states only, it can be used for anything
- * that wants to write commands into the CS (e.g. cache flushes). */
-struct r600_atom {
- void (*emit)(struct r600_context *ctx, struct r600_atom *state);
- unsigned id;
- unsigned num_dw;
- bool dirty;
-};
-
/* This is an atom containing GPU commands that never change.
* This is supposed to be copied directly into the CS. */
struct r600_command_buffer {
unsigned pa_cl_vs_out_cntl; /* from vertex shader */
unsigned clip_plane_enable; /* from rasterizer */
unsigned clip_dist_write; /* from vertex shader */
+ boolean clip_disable; /* from vertex shader */
};
struct r600_alphatest_state {
struct r600_cs_shader_state {
struct r600_atom atom;
unsigned kernel_index;
+ unsigned pc;
struct r600_pipe_compute *shader;
};
struct r600_config_state {
struct r600_atom atom;
unsigned sq_gpr_resource_mgmt_1;
+ unsigned sq_gpr_resource_mgmt_2;
};
struct r600_stencil_ref
struct r600_viewport_state {
struct r600_atom atom;
struct pipe_viewport_state state;
+ int idx;
};
-struct compute_memory_pool;
-void compute_memory_pool_delete(struct compute_memory_pool* pool);
-struct compute_memory_pool* compute_memory_pool_new(
- struct r600_screen *rscreen);
-
-struct r600_pipe_fences {
- struct r600_resource *bo;
- unsigned *data;
- unsigned next_index;
- /* linked list of preallocated blocks */
- struct list_head blocks;
- /* linked list of freed fences */
- struct list_head pool;
- pipe_mutex mutex;
+struct r600_shader_stages_state {
+ struct r600_atom atom;
+ unsigned geom_enable;
};
-enum r600_msaa_texture_mode {
- /* If the hw can fetch the first sample only (no decompression available).
- * This means MSAA texturing is not fully implemented. */
- MSAA_TEXTURE_SAMPLE_ZERO,
-
- /* If the hw can fetch decompressed MSAA textures.
- * Supported families: R600, R700, Evergreen.
- * Cayman cannot use this, because it cannot do the decompression. */
- MSAA_TEXTURE_DECOMPRESSED,
-
- /* If the hw can fetch compressed MSAA textures, which means shaders can
- * read resolved FMASK. This yields the best performance.
- * Supported families: Evergreen, Cayman. */
- MSAA_TEXTURE_COMPRESSED
+struct r600_gs_rings_state {
+ struct r600_atom atom;
+ unsigned enable;
+ struct pipe_constant_buffer esgs_ring;
+ struct pipe_constant_buffer gsvs_ring;
};
-typedef boolean (*r600g_dma_blit_t)(struct pipe_context *ctx,
- struct pipe_resource *dst,
- unsigned dst_level,
- unsigned dst_x, unsigned dst_y, unsigned dst_z,
- struct pipe_resource *src,
- unsigned src_level,
- const struct pipe_box *src_box);
-
-/* logging */
-#define DBG_TEX_DEPTH (1 << 0)
-#define DBG_COMPUTE (1 << 1)
-/* shaders */
-#define DBG_FS (1 << 8)
-#define DBG_VS (1 << 9)
-#define DBG_GS (1 << 10)
-#define DBG_PS (1 << 11)
-#define DBG_CS (1 << 12)
+/* This must start from 16. */
/* features */
-#define DBG_NO_HYPERZ (1 << 16)
-#define DBG_NO_LLVM (1 << 17)
+#define DBG_LLVM (1 << 29)
+#define DBG_NO_CP_DMA (1 << 30)
+/* shader backend */
+#define DBG_NO_SB (1 << 21)
+#define DBG_SB_CS (1 << 22)
+#define DBG_SB_DRY_RUN (1 << 23)
+#define DBG_SB_STAT (1 << 24)
+#define DBG_SB_DUMP (1 << 25)
+#define DBG_SB_NO_FALLBACK (1 << 26)
+#define DBG_SB_DISASM (1 << 27)
+#define DBG_SB_SAFEMATH (1 << 28)
struct r600_screen {
- struct pipe_screen screen;
- struct radeon_winsys *ws;
- unsigned debug_flags;
- unsigned family;
- enum chip_class chip_class;
- struct radeon_info info;
- bool has_streamout;
+ struct r600_common_screen b;
bool has_msaa;
- bool has_cp_dma;
- enum r600_msaa_texture_mode msaa_texture_support;
- struct r600_tiling_info tiling_info;
- struct r600_pipe_fences fences;
+ bool has_compressed_msaa_texturing;
/*for compute global memory binding, we allocate stuff here, instead of
* buffers.
* XXX: Not sure if this is the best place for global_pool. Also,
* it's not thread safe, so it won't work with multiple contexts. */
struct compute_memory_pool *global_pool;
-#if R600_TRACE_CS
- struct r600_resource *trace_bo;
- uint32_t *trace_ptr;
- unsigned cs_count;
-#endif
- r600g_dma_blit_t dma_blit;
};
struct r600_pipe_sampler_view {
struct pipe_sampler_view base;
+ struct list_head list;
struct r600_resource *tex_resource;
uint32_t tex_resource_words[8];
bool skip_mip_address_reloc;
unsigned clip_plane_enable;
unsigned pa_sc_line_stipple;
unsigned pa_cl_clip_cntl;
+ unsigned pa_su_sc_mode_cntl;
float offset_units;
float offset_scale;
bool offset_enable;
uint32_t dirty_mask;
uint32_t compressed_depthtex_mask; /* which textures are depth */
uint32_t compressed_colortex_mask;
- boolean dirty_txq_constants;
boolean dirty_buffer_constants;
};
uint32_t *buffer_constants;
};
-struct r600_fence {
- struct pipe_reference reference;
- unsigned index; /* in the shared bo */
- struct r600_resource *sleep_bo;
- struct list_head head;
-};
-
-#define FENCE_BLOCK_SIZE 16
-
-struct r600_fence_block {
- struct r600_fence fences[FENCE_BLOCK_SIZE];
- struct list_head head;
-};
-
-#define R600_CONSTANT_ARRAY_SIZE 256
-#define R600_RESOURCE_ARRAY_SIZE 160
-
struct r600_constbuf_state
{
struct r600_atom atom;
struct r600_atom atom;
struct pipe_scissor_state scissor;
bool enable; /* r6xx only */
+ int idx;
};
struct r600_fetch_shader {
struct r600_shader_state {
struct r600_atom atom;
- struct r600_pipe_shader_selector *shader;
-};
-
-struct r600_streamout {
- struct r600_atom begin_atom;
- bool begin_emitted;
- unsigned num_dw_for_end;
-
- unsigned enabled_mask;
- unsigned num_targets;
- struct r600_so_target *targets[PIPE_MAX_SO_BUFFERS];
-
- unsigned append_bitmask;
- bool suspended;
-};
-
-struct r600_ring {
- struct radeon_winsys_cs *cs;
- bool flushing;
- void (*flush)(void *ctx, unsigned flags);
-};
-
-struct r600_rings {
- struct r600_ring gfx;
- struct r600_ring dma;
+ struct r600_pipe_shader *shader;
};
struct r600_context {
- struct pipe_context context;
+ struct r600_common_context b;
struct r600_screen *screen;
- struct radeon_winsys *ws;
- struct r600_rings rings;
struct blitter_context *blitter;
- struct u_upload_mgr *uploader;
- struct u_suballocator *allocator_so_filled_size;
struct u_suballocator *allocator_fetch_shader;
- struct util_slab_mempool pool_transfers;
/* Hardware info. */
- enum radeon_family family;
- enum chip_class chip_class;
boolean has_vertex_cache;
boolean keep_tiling_flags;
unsigned default_ps_gprs, default_vs_gprs;
unsigned r6xx_num_clause_temp_gprs;
- unsigned backend_mask;
- unsigned max_db; /* for OQ */
-
- /* current unaccounted memory usage */
- uint64_t vram;
- uint64_t gtt;
/* Miscellaneous state objects. */
void *custom_dsa_flush;
void *custom_blend_resolve;
void *custom_blend_decompress;
- void *custom_blend_fmask_decompress;
+ void *custom_blend_fastclear;
/* With rasterizer discard, there doesn't have to be a pixel shader.
* In that case, we bind this one: */
void *dummy_pixel_shader;
struct r600_poly_offset_state poly_offset_state;
struct r600_cso_state rasterizer_state;
struct r600_sample_mask sample_mask;
- struct r600_scissor_state scissor;
+ struct r600_scissor_state scissor[16];
struct r600_seamless_cube_map seamless_cube_map;
struct r600_config_state config_state;
struct r600_stencil_ref_state stencil_ref;
struct r600_vgt_state vgt_state;
- struct r600_viewport_state viewport;
+ struct r600_viewport_state viewport[16];
/* Shaders and shader resources. */
struct r600_cso_state vertex_fetch_shader;
struct r600_shader_state vertex_shader;
struct r600_shader_state pixel_shader;
+ struct r600_shader_state geometry_shader;
+ struct r600_shader_state export_shader;
struct r600_cs_shader_state cs_shader_state;
+ struct r600_shader_stages_state shader_stages;
+ struct r600_gs_rings_state gs_rings;
struct r600_constbuf_state constbuf_state[PIPE_SHADER_TYPES];
struct r600_textures_info samplers[PIPE_SHADER_TYPES];
/** Vertex buffers for fetch shaders */
struct r600_vertexbuf_state vertex_buffer_state;
/** Vertex buffers for compute shaders */
struct r600_vertexbuf_state cs_vertex_buffer_state;
- struct r600_streamout streamout;
/* Additional context states. */
- unsigned flags;
unsigned compute_cb_target_mask;
struct r600_pipe_shader_selector *ps_shader;
struct r600_pipe_shader_selector *vs_shader;
+ struct r600_pipe_shader_selector *gs_shader;
struct r600_rasterizer_state *rasterizer;
bool alpha_to_one;
bool force_blend_disable;
boolean dual_src_blend;
unsigned zwritemask;
+ int ps_iter_samples;
/* Index buffer. */
struct pipe_index_buffer index_buffer;
int last_primitive_type; /* Last primitive type used in draw_vbo. */
int last_start_instance;
- /* Queries. */
- /* The list of active queries. Only one query of each type can be active. */
- int num_occlusion_queries;
- /* Keep track of non-timer queries, because they should be suspended
- * during context flushing.
- * The timer queries (TIME_ELAPSED) shouldn't be suspended. */
- struct list_head active_nontimer_queries;
- unsigned num_cs_dw_nontimer_queries_suspend;
- /* If queries have been suspended. */
- bool nontimer_queries_suspended;
-
- /* Render condition. */
- struct pipe_query *current_render_cond;
- unsigned current_render_cond_mode;
- boolean predicate_drawing;
-
+ void *sb_context;
struct r600_isa *isa;
};
cs->cdw += cb->num_dw;
}
-#if R600_TRACE_CS
void r600_trace_emit(struct r600_context *rctx);
-#endif
static INLINE void r600_emit_atom(struct r600_context *rctx, struct r600_atom *atom)
{
- atom->emit(rctx, atom);
+ atom->emit(&rctx->b, atom);
atom->dirty = false;
-#if R600_TRACE_CS
- if (rctx->screen->trace_bo) {
+ if (rctx->screen->b.trace_bo) {
r600_trace_emit(rctx);
}
-#endif
}
static INLINE void r600_set_cso_state(struct r600_cso_state *state, void *cso)
struct r600_command_buffer *cb)
{
state->cb = cb;
- state->atom.num_dw = cb->num_dw;
+ state->atom.num_dw = cb ? cb->num_dw : 0;
r600_set_cso_state(state, cso);
}
+/* compute_memory_pool.c */
+struct compute_memory_pool;
+void compute_memory_pool_delete(struct compute_memory_pool* pool);
+struct compute_memory_pool* compute_memory_pool_new(
+ struct r600_screen *rscreen);
+
+/* evergreen_compute.c */
+void evergreen_set_cs_sampler_view(struct pipe_context *ctx_,
+ unsigned start_slot, unsigned count,
+ struct pipe_sampler_view **views);
+
/* evergreen_state.c */
struct pipe_sampler_view *
evergreen_create_sampler_view_custom(struct pipe_context *ctx,
struct pipe_resource *texture,
const struct pipe_sampler_view *state,
- unsigned width0, unsigned height0);
+ unsigned width0, unsigned height0,
+ unsigned force_level);
void evergreen_init_common_regs(struct r600_command_buffer *cb,
enum chip_class ctx_chip_class,
enum radeon_family ctx_family,
void evergreen_init_state_functions(struct r600_context *rctx);
void evergreen_init_atom_start_cs(struct r600_context *rctx);
void evergreen_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
+void evergreen_update_es_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
+void evergreen_update_gs_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
void evergreen_update_vs_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
void *evergreen_create_db_flush_dsa(struct r600_context *rctx);
void *evergreen_create_resolve_blend(struct r600_context *rctx);
void *evergreen_create_decompress_blend(struct r600_context *rctx);
-void *evergreen_create_fmask_decompress_blend(struct r600_context *rctx);
+void *evergreen_create_fastclear_blend(struct r600_context *rctx);
boolean evergreen_is_format_supported(struct pipe_screen *screen,
enum pipe_format format,
enum pipe_texture_target target,
void evergreen_update_db_shader_control(struct r600_context * rctx);
/* r600_blit.c */
-void r600_copy_buffer(struct pipe_context *ctx, struct pipe_resource *dst, unsigned dstx,
- struct pipe_resource *src, const struct pipe_box *src_box);
void r600_init_blit_functions(struct r600_context *rctx);
-void r600_blit_decompress_depth(struct pipe_context *ctx,
- struct r600_texture *texture,
- struct r600_texture *staging,
- unsigned first_level, unsigned last_level,
- unsigned first_layer, unsigned last_layer,
- unsigned first_sample, unsigned last_sample);
void r600_decompress_depth_textures(struct r600_context *rctx,
struct r600_samplerview_state *textures);
void r600_decompress_color_textures(struct r600_context *rctx,
struct r600_samplerview_state *textures);
-
-/* r600_buffer.c */
-bool r600_init_resource(struct r600_screen *rscreen,
- struct r600_resource *res,
- unsigned size, unsigned alignment,
- bool use_reusable_pool, unsigned usage);
-struct pipe_resource *r600_buffer_create(struct pipe_screen *screen,
- const struct pipe_resource *templ,
- unsigned alignment);
-
-/* r600_pipe.c */
-boolean r600_rings_is_buffer_referenced(struct r600_context *ctx,
- struct radeon_winsys_cs_handle *buf,
- enum radeon_bo_usage usage);
-void *r600_buffer_mmap_sync_with_rings(struct r600_context *ctx,
- struct r600_resource *resource,
- unsigned usage);
-
-/* r600_query.c */
-void r600_init_query_functions(struct r600_context *rctx);
-void r600_suspend_nontimer_queries(struct r600_context *ctx);
-void r600_resume_nontimer_queries(struct r600_context *ctx);
-
-/* r600_resource.c */
-void r600_init_context_resource_functions(struct r600_context *r600);
+void r600_resource_copy_region(struct pipe_context *ctx,
+ struct pipe_resource *dst,
+ unsigned dst_level,
+ unsigned dstx, unsigned dsty, unsigned dstz,
+ struct pipe_resource *src,
+ unsigned src_level,
+ const struct pipe_box *src_box);
/* r600_shader.c */
int r600_pipe_shader_create(struct pipe_context *ctx,
struct r600_pipe_shader *shader,
struct r600_shader_key key);
-#ifdef HAVE_OPENCL
-int r600_compute_shader_create(struct pipe_context * ctx,
- LLVMModuleRef mod, struct r600_bytecode * bytecode);
-#endif
+
void r600_pipe_shader_destroy(struct pipe_context *ctx, struct r600_pipe_shader *shader);
/* r600_state.c */
void r600_init_state_functions(struct r600_context *rctx);
void r600_init_atom_start_cs(struct r600_context *rctx);
void r600_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
+void r600_update_es_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
+void r600_update_gs_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
void r600_update_vs_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
void *r600_create_db_flush_dsa(struct r600_context *rctx);
void *r600_create_resolve_blend(struct r600_context *rctx);
unsigned usage);
void r600_update_db_shader_control(struct r600_context * rctx);
-/* r600_texture.c */
-void r600_init_screen_texture_functions(struct pipe_screen *screen);
-void r600_init_surface_functions(struct r600_context *r600);
-uint32_t r600_translate_texformat(struct pipe_screen *screen, enum pipe_format format,
- const unsigned char *swizzle_view,
- uint32_t *word4_p, uint32_t *yuv_format_p);
-unsigned r600_texture_get_offset(struct r600_texture *rtex,
- unsigned level, unsigned layer);
-struct pipe_surface *r600_create_surface_custom(struct pipe_context *pipe,
- struct pipe_resource *texture,
- const struct pipe_surface *templ,
- unsigned width, unsigned height);
-
-unsigned r600_get_swizzle_combined(const unsigned char *swizzle_format,
- const unsigned char *swizzle_view,
- boolean vtx);
-
/* r600_hw_context.c */
-void r600_emit_streamout_begin(struct r600_context *ctx, struct r600_atom *atom);
-void r600_emit_streamout_end(struct r600_context *ctx);
+void r600_context_gfx_flush(void *context, unsigned flags,
+ struct pipe_fence_handle **fence);
+void r600_begin_new_cs(struct r600_context *ctx);
+void r600_flush_emit(struct r600_context *ctx);
+void r600_need_cs_space(struct r600_context *ctx, unsigned num_dw, boolean count_draw_in);
+void r600_cp_dma_copy_buffer(struct r600_context *rctx,
+ struct pipe_resource *dst, uint64_t dst_offset,
+ struct pipe_resource *src, uint64_t src_offset,
+ unsigned size);
+void evergreen_cp_dma_clear_buffer(struct r600_context *rctx,
+ struct pipe_resource *dst, uint64_t offset,
+ unsigned size, uint32_t clear_value);
+void r600_dma_copy_buffer(struct r600_context *rctx,
+ struct pipe_resource *dst,
+ struct pipe_resource *src,
+ uint64_t dst_offset,
+ uint64_t src_offset,
+ uint64_t size);
/*
* evergreen_hw_context.c
*/
-void evergreen_flush_vgt_streamout(struct r600_context *ctx);
-void evergreen_set_streamout_enable(struct r600_context *ctx, unsigned buffer_enable_bit);
+void evergreen_dma_copy_buffer(struct r600_context *rctx,
+ struct pipe_resource *dst,
+ struct pipe_resource *src,
+ uint64_t dst_offset,
+ uint64_t src_offset,
+ uint64_t size);
/* r600_state_common.c */
void r600_init_common_state_functions(struct r600_context *rctx);
void r600_sampler_states_dirty(struct r600_context *rctx,
struct r600_sampler_states *state);
void r600_constant_buffers_dirty(struct r600_context *rctx, struct r600_constbuf_state *state);
-void r600_streamout_buffers_dirty(struct r600_context *rctx);
-void r600_draw_rectangle(struct blitter_context *blitter,
- int x1, int y1, int x2, int y2, float depth,
- enum blitter_attrib_type type, const union pipe_color_union *attrib);
+void r600_set_sample_locations_constant_buffer(struct r600_context *rctx);
uint32_t r600_translate_stencil_op(int s_op);
uint32_t r600_translate_fill(uint32_t func);
unsigned r600_tex_wrap(unsigned wrap);
unsigned r600_tex_mipfilter(unsigned filter);
unsigned r600_tex_compare(unsigned compare);
bool sampler_state_needs_border_color(const struct pipe_sampler_state *state);
+struct pipe_surface *r600_create_surface_custom(struct pipe_context *pipe,
+ struct pipe_resource *texture,
+ const struct pipe_surface *templ,
+ unsigned width, unsigned height);
+unsigned r600_get_swizzle_combined(const unsigned char *swizzle_format,
+ const unsigned char *swizzle_view,
+ boolean vtx);
+uint32_t r600_translate_texformat(struct pipe_screen *screen, enum pipe_format format,
+ const unsigned char *swizzle_view,
+ uint32_t *word4_p, uint32_t *yuv_format_p);
+uint32_t r600_translate_colorformat(enum chip_class chip, enum pipe_format format);
+uint32_t r600_colorformat_endian_swap(uint32_t colorformat);
+
+/* r600_uvd.c */
+struct pipe_video_codec *r600_uvd_create_decoder(struct pipe_context *context,
+ const struct pipe_video_codec *decoder);
+
+struct pipe_video_buffer *r600_video_buffer_create(struct pipe_context *pipe,
+ const struct pipe_video_buffer *tmpl);
/*
* Helpers for building command buffers
void r600_init_command_buffer(struct r600_command_buffer *cb, unsigned num_dw);
void r600_release_command_buffer(struct r600_command_buffer *cb);
-/*
- * Helpers for emitting state into a command stream directly.
- */
-static INLINE unsigned r600_context_bo_reloc(struct r600_context *ctx,
- struct r600_ring *ring,
- struct r600_resource *rbo,
- enum radeon_bo_usage usage)
-{
- assert(usage);
- /* make sure that all previous ring use are flushed so everything
- * look serialized from driver pov
- */
- if (!ring->flushing) {
- if (ring == &ctx->rings.gfx) {
- if (ctx->rings.dma.cs) {
- /* flush dma ring */
- ctx->rings.dma.flush(ctx, RADEON_FLUSH_ASYNC);
- }
- } else {
- /* flush gfx ring */
- ctx->rings.gfx.flush(ctx, RADEON_FLUSH_ASYNC);
- }
- }
- return ctx->ws->cs_add_reloc(ring->cs, rbo->cs_buf, usage, rbo->domains) * 4;
-}
-
-static INLINE void r600_write_value(struct radeon_winsys_cs *cs, unsigned value)
-{
- cs->buf[cs->cdw++] = value;
-}
-
-static INLINE void r600_write_array(struct radeon_winsys_cs *cs, unsigned num, unsigned *ptr)
-{
- assert(cs->cdw+num <= RADEON_MAX_CMDBUF_DWORDS);
- memcpy(&cs->buf[cs->cdw], ptr, num * sizeof(ptr[0]));
- cs->cdw += num;
-}
-
-static INLINE void r600_write_config_reg_seq(struct radeon_winsys_cs *cs, unsigned reg, unsigned num)
-{
- assert(reg < R600_CONTEXT_REG_OFFSET);
- assert(cs->cdw+2+num <= RADEON_MAX_CMDBUF_DWORDS);
- cs->buf[cs->cdw++] = PKT3(PKT3_SET_CONFIG_REG, num, 0);
- cs->buf[cs->cdw++] = (reg - R600_CONFIG_REG_OFFSET) >> 2;
-}
-
-static INLINE void r600_write_context_reg_seq(struct radeon_winsys_cs *cs, unsigned reg, unsigned num)
-{
- assert(reg >= R600_CONTEXT_REG_OFFSET && reg < R600_CTL_CONST_OFFSET);
- assert(cs->cdw+2+num <= RADEON_MAX_CMDBUF_DWORDS);
- cs->buf[cs->cdw++] = PKT3(PKT3_SET_CONTEXT_REG, num, 0);
- cs->buf[cs->cdw++] = (reg - R600_CONTEXT_REG_OFFSET) >> 2;
-}
-
static INLINE void r600_write_compute_context_reg_seq(struct radeon_winsys_cs *cs, unsigned reg, unsigned num)
{
r600_write_context_reg_seq(cs, reg, num);
cs->buf[cs->cdw++] = (reg - R600_CTL_CONST_OFFSET) >> 2;
}
-static INLINE void r600_write_config_reg(struct radeon_winsys_cs *cs, unsigned reg, unsigned value)
-{
- r600_write_config_reg_seq(cs, reg, 1);
- r600_write_value(cs, value);
-}
-
-static INLINE void r600_write_context_reg(struct radeon_winsys_cs *cs, unsigned reg, unsigned value)
+static INLINE void r600_write_compute_context_reg(struct radeon_winsys_cs *cs, unsigned reg, unsigned value)
{
- r600_write_context_reg_seq(cs, reg, 1);
- r600_write_value(cs, value);
+ r600_write_compute_context_reg_seq(cs, reg, 1);
+ radeon_emit(cs, value);
}
-static INLINE void r600_write_compute_context_reg(struct radeon_winsys_cs *cs, unsigned reg, unsigned value)
+static INLINE void r600_write_context_reg_flag(struct radeon_winsys_cs *cs, unsigned reg, unsigned value, unsigned flag)
{
- r600_write_compute_context_reg_seq(cs, reg, 1);
- r600_write_value(cs, value);
+ if (flag & RADEON_CP_PACKET3_COMPUTE_MODE) {
+ r600_write_compute_context_reg(cs, reg, value);
+ } else {
+ r600_write_context_reg(cs, reg, value);
+ }
}
static INLINE void r600_write_ctl_const(struct radeon_winsys_cs *cs, unsigned reg, unsigned value)
{
r600_write_ctl_const_seq(cs, reg, 1);
- r600_write_value(cs, value);
+ radeon_emit(cs, value);
}
/*
}
#define ALIGN_DIVUP(x, y) (((x) + (y) - 1) / (y))
-static inline unsigned r600_tex_aniso_filter(unsigned filter)
-{
- if (filter <= 1) return 0;
- if (filter <= 2) return 1;
- if (filter <= 4) return 2;
- if (filter <= 8) return 3;
- /* else */ return 4;
-}
-
/* 12.4 fixed-point */
static INLINE unsigned r600_pack_float_12p4(float x)
{
x >= 4096 ? 0xffff : x * 16;
}
-static INLINE uint64_t r600_resource_va(struct pipe_screen *screen, struct pipe_resource *resource)
-{
- struct r600_screen *rscreen = (struct r600_screen*)screen;
- struct r600_resource *rresource = (struct r600_resource*)resource;
-
- return rscreen->ws->buffer_get_virtual_address(rresource->cs_buf);
-}
-
-static INLINE void r600_context_add_resource_size(struct pipe_context *ctx, struct pipe_resource *r)
-{
- struct r600_context *rctx = (struct r600_context *)ctx;
- struct r600_resource *rr = (struct r600_resource *)r;
-
- if (r == NULL) {
- return;
- }
-
- /*
- * The idea is to compute a gross estimate of memory requirement of
- * each draw call. After each draw call, memory will be precisely
- * accounted. So the uncertainty is only on the current draw call.
- * In practice this gave very good estimate (+/- 10% of the target
- * memory limit).
- */
- if (rr->domains & RADEON_DOMAIN_GTT) {
- rctx->gtt += rr->buf->size;
- }
- if (rr->domains & RADEON_DOMAIN_VRAM) {
- rctx->vram += rr->buf->size;
- }
+/* Return if the depth format can be read without the DB->CB copy on r6xx-r7xx. */
+static INLINE bool r600_can_read_depth(struct r600_texture *rtex)
+{
+ return rtex->resource.b.b.nr_samples <= 1 &&
+ (rtex->resource.b.b.format == PIPE_FORMAT_Z16_UNORM ||
+ rtex->resource.b.b.format == PIPE_FORMAT_Z32_FLOAT);
+}
+
+#define V_028A6C_OUTPRIM_TYPE_POINTLIST 0
+#define V_028A6C_OUTPRIM_TYPE_LINESTRIP 1
+#define V_028A6C_OUTPRIM_TYPE_TRISTRIP 2
+
+static INLINE unsigned r600_conv_prim_to_gs_out(unsigned mode)
+{
+ static const int prim_conv[] = {
+ V_028A6C_OUTPRIM_TYPE_POINTLIST,
+ V_028A6C_OUTPRIM_TYPE_LINESTRIP,
+ V_028A6C_OUTPRIM_TYPE_LINESTRIP,
+ V_028A6C_OUTPRIM_TYPE_LINESTRIP,
+ V_028A6C_OUTPRIM_TYPE_TRISTRIP,
+ V_028A6C_OUTPRIM_TYPE_TRISTRIP,
+ V_028A6C_OUTPRIM_TYPE_TRISTRIP,
+ V_028A6C_OUTPRIM_TYPE_TRISTRIP,
+ V_028A6C_OUTPRIM_TYPE_TRISTRIP,
+ V_028A6C_OUTPRIM_TYPE_TRISTRIP,
+ V_028A6C_OUTPRIM_TYPE_LINESTRIP,
+ V_028A6C_OUTPRIM_TYPE_LINESTRIP,
+ V_028A6C_OUTPRIM_TYPE_TRISTRIP,
+ V_028A6C_OUTPRIM_TYPE_TRISTRIP,
+ V_028A6C_OUTPRIM_TYPE_TRISTRIP
+ };
+ assert(mode < Elements(prim_conv));
+
+ return prim_conv[mode];
}
#endif