X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fmesa%2Fdrivers%2Fdri%2Fr300%2Fr300_state.c;h=aa9eb5e28bdc25ce49d0c79160030dd13bdcbd9f;hb=34dd4843d7b86eee8e0d18de275dd963ac2e3ffd;hp=53f90c63221d63a88b5f990ccb2e6cb10c66bc0c;hpb=e09c843c137d91a20e5c705c6c4b8a75579d8232;p=mesa.git diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c index 53f90c63221..aa9eb5e28bd 100644 --- a/src/mesa/drivers/dri/r300/r300_state.c +++ b/src/mesa/drivers/dri/r300/r300_state.c @@ -48,6 +48,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "swrast_setup/swrast_setup.h" #include "array_cache/acache.h" #include "tnl/tnl.h" +#include "texformat.h" #include "radeon_ioctl.h" #include "radeon_state.h" @@ -56,6 +57,11 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_state.h" #include "r300_reg.h" #include "r300_program.h" +#include "r300_emit.h" +#include "r300_fixed_pipelines.h" +#include "r300_tex.h" +#include "r300_maos.h" +#include "r300_texprog.h" static void r300AlphaFunc(GLcontext * ctx, GLenum func, GLfloat ref) { @@ -64,7 +70,7 @@ static void r300AlphaFunc(GLcontext * ctx, GLenum func, GLfloat ref) GLubyte refByte; CLAMPED_FLOAT_TO_UBYTE(refByte, ref); - + R300_STATECHANGE(rmesa, at); pp_misc &= ~(R300_ALPHA_TEST_OP_MASK | R300_REF_ALPHA_MASK); @@ -94,6 +100,7 @@ static void r300AlphaFunc(GLcontext * ctx, GLenum func, GLfloat ref) break; case GL_ALWAYS: pp_misc |= R300_ALPHA_TEST_PASS; + //pp_misc &= ~R300_ALPHA_TEST_ENABLE; break; } @@ -104,8 +111,18 @@ static void r300BlendColor(GLcontext * ctx, const GLfloat cf[4]) { GLubyte color[4]; r300ContextPtr rmesa = R300_CONTEXT(ctx); - fprintf(stderr, "%s:%s is not implemented yet. Fixme !\n", __FILE__, __FUNCTION__); - #if 0 + + R300_STATECHANGE(rmesa, unk4E10); + + /* Ordering might be wrong */ + CLAMPED_FLOAT_TO_UBYTE(color[0], cf[0]); + CLAMPED_FLOAT_TO_UBYTE(color[1], cf[1]); + CLAMPED_FLOAT_TO_UBYTE(color[2], cf[2]); + CLAMPED_FLOAT_TO_UBYTE(color[3], cf[3]); + + rmesa->hw.unk4E10.cmd[1]=r300PackColor(4, color[0], color[1], color[2], color[3]); + //fprintf(stderr, "%s:%s is not implemented yet. Fixme !\n", __FILE__, __FUNCTION__); +#if 0 R200_STATECHANGE(rmesa, ctx); CLAMPED_FLOAT_TO_UBYTE(color[0], cf[0]); CLAMPED_FLOAT_TO_UBYTE(color[1], cf[1]); @@ -114,7 +131,7 @@ static void r300BlendColor(GLcontext * ctx, const GLfloat cf[4]) if (rmesa->radeon.radeonScreen->drmSupportsBlendColor) rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCOLOR] = radeonPackColor(4, color[0], color[1], color[2], color[3]); - #endif +#endif } /** @@ -199,14 +216,38 @@ static int blend_factor(GLenum factor, GLboolean is_src) * and GL_FUNC_REVERSE_SUBTRACT will cause wrong results otherwise for * unknown reasons. */ + +/* helper function */ +static void r300_set_blend_cntl(r300ContextPtr rmesa, int func, int eqn, int cbits, int funcA, int eqnA) +{ + GLuint new_ablend, new_cblend; + +#if 0 + fprintf(stderr, "eqnA=%08x funcA=%08x eqn=%08x func=%08x cbits=%08x\n", eqnA, funcA, eqn, func, cbits); +#endif + new_ablend = eqnA | funcA; + new_cblend = eqn | func; + if(funcA == func){ + new_cblend |= R300_BLEND_NO_SEPARATE; + } + new_cblend |= cbits; + + if((new_ablend != rmesa->hw.bld.cmd[R300_BLD_ABLEND]) + || (new_cblend != rmesa->hw.bld.cmd[R300_BLD_CBLEND])){ + R300_STATECHANGE(rmesa, bld); + rmesa->hw.bld.cmd[R300_BLD_ABLEND]=new_ablend; + rmesa->hw.bld.cmd[R300_BLD_CBLEND]=new_cblend; + } +} + static void r300_set_blend_state(GLcontext * ctx) { r300ContextPtr rmesa = R300_CONTEXT(ctx); - #if 0 +#if 0 GLuint cntl = rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] & ~(R300_ROP_ENABLE | R300_ALPHA_BLEND_ENABLE | R300_SEPARATE_ALPHA_ENABLE); - #endif +#endif int func = (R200_BLEND_GL_ONE << R200_SRC_BLEND_SHIFT) | (R200_BLEND_GL_ZERO << R200_DST_BLEND_SHIFT); @@ -215,49 +256,53 @@ static void r300_set_blend_state(GLcontext * ctx) (R200_BLEND_GL_ZERO << R200_DST_BLEND_SHIFT); int eqnA = R200_COMB_FCN_ADD_CLAMP; - R300_STATECHANGE(rmesa, bld); if (rmesa->radeon.radeonScreen->drmSupportsBlendColor) { if (ctx->Color._LogicOpEnabled) { - #if 0 +#if 0 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = cntl | R300_ROP_ENABLE; - #endif - rmesa->hw.bld.cmd[R300_BLD_ABLEND] = eqn | func; - rmesa->hw.bld.cmd[R300_BLD_CBLEND] = eqn | func; +#endif + r300_set_blend_cntl(rmesa, + func, eqn, 0, + func, eqn); return; } else if (ctx->Color.BlendEnabled) { - #if 0 +#if 0 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = cntl | R300_ALPHA_BLEND_ENABLE | R300_SEPARATE_ALPHA_ENABLE; - #endif +#endif } else { - #if 0 +#if 0 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = cntl; - #endif - rmesa->hw.bld.cmd[R300_BLD_ABLEND] = eqn | func; - rmesa->hw.bld.cmd[R300_BLD_CBLEND] = eqn | func; +#endif + r300_set_blend_cntl(rmesa, + func, eqn, 0, + func, eqn); return; } } else { if (ctx->Color._LogicOpEnabled) { - #if 0 +#if 0 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = cntl | R300_ROP_ENABLE; rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCNTL] = eqn | func; - #endif +#endif return; } else if (ctx->Color.BlendEnabled) { - #if 0 +#if 0 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = cntl | R300_ALPHA_BLEND_ENABLE; - #endif +#endif } else { - #if 0 +#if 0 rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = cntl; rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCNTL] = eqn | func; - #endif +#endif + r300_set_blend_cntl(rmesa, + func, eqn, 0, + func, eqn); return; } } @@ -301,9 +346,9 @@ static void r300_set_blend_state(GLcontext * ctx) } if (!rmesa->radeon.radeonScreen->drmSupportsBlendColor) { - #if 0 +#if 0 rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCNTL] = eqn | func; - #endif +#endif return; } @@ -344,14 +389,12 @@ static void r300_set_blend_state(GLcontext * ctx) return; } - rmesa->hw.bld.cmd[R300_BLD_ABLEND] = eqnA | funcA; - rmesa->hw.bld.cmd[R300_BLD_CBLEND] = eqn | func ; - if(rmesa->hw.bld.cmd[R300_BLD_ABLEND] == rmesa->hw.bld.cmd[R300_BLD_CBLEND]){ - rmesa->hw.bld.cmd[R300_BLD_CBLEND] |= R300_BLEND_UNKNOWN | R300_BLEND_ENABLE | R300_BLEND_NO_SEPARATE; - } else { - rmesa->hw.bld.cmd[R300_BLD_CBLEND] |= R300_BLEND_UNKNOWN | R300_BLEND_ENABLE; - } - + r300_set_blend_cntl(rmesa, + func, eqn, R300_BLEND_UNKNOWN | R300_BLEND_ENABLE, + funcA, eqnA); + r300_set_blend_cntl(rmesa, + func, eqn, R300_BLEND_UNKNOWN | R300_BLEND_ENABLE, + funcA, eqnA); } static void r300BlendEquationSeparate(GLcontext * ctx, @@ -389,7 +432,6 @@ static void r300UpdateCulling(GLcontext* ctx) else val |= R300_FRONT_FACE_CCW; } - r300->hw.cul.cmd[R300_CUL_CULL] = val; } @@ -410,8 +452,31 @@ static void r300Enable(GLcontext* ctx, GLenum cap, GLboolean state) state ? "GL_TRUE" : "GL_FALSE"); switch (cap) { + /* Fast track this one... + */ + case GL_TEXTURE_1D: + case GL_TEXTURE_2D: + case GL_TEXTURE_3D: + break; + + case GL_ALPHA_TEST: + R300_STATECHANGE(r300, at); + if (state) { + r300->hw.at.cmd[R300_AT_ALPHA_TEST] |= + R300_ALPHA_TEST_ENABLE; + } else { + r300->hw.at.cmd[R300_AT_ALPHA_TEST] &= + ~R300_ALPHA_TEST_ENABLE; + } + break; + + case GL_BLEND: + case GL_COLOR_LOGIC_OP: + r300_set_blend_state(ctx); + break; + case GL_DEPTH_TEST: - R300_STATECHANGE(r300, zc); + R300_STATECHANGE(r300, zs); if (state) { if (ctx->Depth.Mask) @@ -419,15 +484,58 @@ static void r300Enable(GLcontext* ctx, GLenum cap, GLboolean state) else newval = R300_RB3D_Z_TEST; } else - newval = 0; + newval = R300_RB3D_Z_DISABLED_1; - r300->hw.zc.cmd[R300_ZC_CNTL_0] = newval; + r300->hw.zs.cmd[R300_ZS_CNTL_0] &= R300_RB3D_STENCIL_ENABLE; + r300->hw.zs.cmd[R300_ZS_CNTL_0] |= newval; + break; + + case GL_STENCIL_TEST: + WARN_ONCE("TODO - double side stencil !\n"); + if (r300->state.stencil.hw_stencil) { + R300_STATECHANGE(r300, zs); + if (state) { + r300->hw.zs.cmd[R300_ZS_CNTL_0] |= + R300_RB3D_STENCIL_ENABLE; + } else { + r300->hw.zs.cmd[R300_ZS_CNTL_0] &= + ~R300_RB3D_STENCIL_ENABLE; + } + } else { + FALLBACK(&r300->radeon, RADEON_FALLBACK_STENCIL, state); + } break; case GL_CULL_FACE: r300UpdateCulling(ctx); break; + case GL_POLYGON_OFFSET_POINT: + case GL_POLYGON_OFFSET_LINE: + WARN_ONCE("Don't know how to enable polygon offset point/line. Help me !\n"); + + /* Something is apparently blocking these from working */ + R300_STATECHANGE(r300, unk42B4); + if(state){ + r300->hw.unk42B4.cmd[1] |= ~(3<<0); + } else { + r300->hw.unk42B4.cmd[1] &= (3<<0); + } + break; + + case GL_POLYGON_OFFSET_FILL: + R300_STATECHANGE(r300, unk42B4); + if(state){ + r300->hw.unk42B4.cmd[1] |= (3<<0); + } else { + r300->hw.unk42B4.cmd[1] &= ~(3<<0); + } + break; + + case GL_VERTEX_PROGRAM_ARB: + //TCL_FALLBACK(rmesa->glCtx, R200_TCL_FALLBACK_TCL_DISABLE, state); + break; + default: radeonEnable(ctx, cap, state); return; @@ -470,32 +578,34 @@ static void r300DepthFunc(GLcontext* ctx, GLenum func) { r300ContextPtr r300 = R300_CONTEXT(ctx); - R300_STATECHANGE(r300, zc); + R300_STATECHANGE(r300, zs); + + r300->hw.zs.cmd[R300_ZS_CNTL_1] &= ~(R300_ZS_MASK << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT); switch(func) { case GL_NEVER: - r300->hw.zc.cmd[R300_ZC_CNTL_1] = R300_Z_TEST_NEVER; + r300->hw.zs.cmd[R300_ZS_CNTL_1] |= R300_ZS_NEVER << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT; break; case GL_LESS: - r300->hw.zc.cmd[R300_ZC_CNTL_1] = R300_Z_TEST_LESS; + r300->hw.zs.cmd[R300_ZS_CNTL_1] |= R300_ZS_LESS << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT; break; case GL_EQUAL: - r300->hw.zc.cmd[R300_ZC_CNTL_1] = R300_Z_TEST_EQUAL; + r300->hw.zs.cmd[R300_ZS_CNTL_1] |= R300_ZS_EQUAL << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT; break; case GL_LEQUAL: - r300->hw.zc.cmd[R300_ZC_CNTL_1] = R300_Z_TEST_LEQUAL; + r300->hw.zs.cmd[R300_ZS_CNTL_1] |= R300_ZS_LEQUAL << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT; break; case GL_GREATER: - r300->hw.zc.cmd[R300_ZC_CNTL_1] = R300_Z_TEST_GREATER; + r300->hw.zs.cmd[R300_ZS_CNTL_1] |= R300_ZS_GREATER << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT; break; case GL_NOTEQUAL: - r300->hw.zc.cmd[R300_ZC_CNTL_1] = R300_Z_TEST_NEQUAL; + r300->hw.zs.cmd[R300_ZS_CNTL_1] |= R300_ZS_NOTEQUAL << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT; break; case GL_GEQUAL: - r300->hw.zc.cmd[R300_ZC_CNTL_1] = R300_Z_TEST_GEQUAL; + r300->hw.zs.cmd[R300_ZS_CNTL_1] |= R300_ZS_GEQUAL << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT; break; case GL_ALWAYS: - r300->hw.zc.cmd[R300_ZC_CNTL_1] = R300_Z_TEST_ALWAYS; + r300->hw.zs.cmd[R300_ZS_CNTL_1] |= R300_ZS_ALWAYS << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT; break; } } @@ -513,9 +623,10 @@ static void r300DepthMask(GLcontext* ctx, GLboolean mask) if (!ctx->Depth.Test) return; - R300_STATECHANGE(r300, zc); - r300->hw.zc.cmd[R300_ZC_CNTL_0] = mask - ? R300_RB3D_Z_TEST_AND_WRITE : R300_RB3D_Z_TEST; + R300_STATECHANGE(r300, zs); + r300->hw.zs.cmd[R300_ZS_CNTL_0] &= R300_RB3D_STENCIL_ENABLE; + r300->hw.zs.cmd[R300_ZS_CNTL_0] |= mask + ? R300_RB3D_Z_TEST_AND_WRITE : R300_RB3D_Z_TEST; } @@ -540,10 +651,279 @@ static void r300ColorMask(GLcontext* ctx, static void r300PointSize(GLcontext * ctx, GLfloat size) { r300ContextPtr r300 = R300_CONTEXT(ctx); - + + size = ctx->Point._Size; + + R300_STATECHANGE(r300, ps); + r300->hw.ps.cmd[R300_PS_POINTSIZE] = + ((int)(size * 6) << R300_POINTSIZE_X_SHIFT) | + ((int)(size * 6) << R300_POINTSIZE_Y_SHIFT); + +#if 0 /* r200 reg? */ /* This might need fixing later */ R300_STATECHANGE(r300, vps); r300->hw.vps.cmd[R300_VPS_POINTSIZE] = r300PackFloat32(1.0); +#endif +} + +/* ============================================================= + * Line state + */ +static void r300LineWidth(GLcontext *ctx, GLfloat widthf) +{ + r300ContextPtr r300 = R300_CONTEXT(ctx); + + widthf = ctx->Line._Width; + + R300_STATECHANGE(r300, lcntl); + r300->hw.lcntl.cmd[1] = (int)(widthf * 6.0); + /* Doesnt look very good without this... */ + r300->hw.lcntl.cmd[1] |= R300_LINE_CNT_UNK1; +} + +/* + +glPolygonMode(GL_FRONT_AND_BACK, GL_LINE); : 00000091 ( 1001 0001) +glPolygonMode(GL_FRONT_AND_BACK, GL_POINT); : 00000001 ( 1) + +glPolygonMode(GL_FRONT, GL_LINE); : 00000111 (1 0001 0001) +glPolygonMode(GL_FRONT, GL_POINT); : 00000101 (1 0000 0001) + +glPolygonMode(GL_BACK, GL_LINE); : 000000a1 ( 1010 0001) +glPolygonMode(GL_BACK, GL_POINT); : 00000021 ( 10 0001) + +*/ + +/* exclusive */ +#define PM_NOT_BACK (1<<8) +#define PM_NOT_FRONT (1<<5) + +#define PM_FRONT_LINE (1<<4) +#define PM_BACK_LINE (1<<7) + +static void r300PolygonMode(GLcontext *ctx, GLenum face, GLenum mode) +{ + r300ContextPtr r300 = R300_CONTEXT(ctx); + unsigned long hw_mode=0; + + //hw_mode=r300->hw.unk4288.cmd[1]; + hw_mode |= 1; /* enables point mode by default */ + + switch (ctx->Polygon.FrontMode) { + case GL_LINE: + hw_mode &= ~PM_NOT_FRONT; + hw_mode |= PM_FRONT_LINE; + break; + case GL_POINT: + hw_mode &= ~PM_NOT_FRONT; + hw_mode &= ~PM_FRONT_LINE; + break; + /* I dont think fgl properly handles these... In any case, test program is needed */ + case GL_FILL: + break; + } + + switch (ctx->Polygon.BackMode) { + case GL_LINE: + hw_mode &= ~PM_NOT_BACK; + hw_mode |= PM_BACK_LINE; + break; + case GL_POINT: + hw_mode &= ~PM_NOT_BACK; + hw_mode &= ~PM_BACK_LINE; + break; + case GL_FILL: + break; + } + + if(hw_mode == 1) + hw_mode = 0; + +#if 0 + switch (face) { + case GL_FRONT: + //fprintf(stderr, "front\n"); + hw_mode &= ~PM_NOT_FRONT; + switch (mode) { + case GL_LINE: + hw_mode |= PM_FRONT_LINE; + break; + case GL_POINT: + hw_mode &= ~PM_FRONT_LINE; + break; + case GL_FILL: + break; + } + break; + + case GL_BACK: + //fprintf(stderr, "back\n"); + hw_mode &= ~PM_NOT_BACK; + switch (mode) { + case GL_LINE: + hw_mode |= PM_BACK_LINE; + break; + case GL_POINT: + hw_mode &= ~PM_BACK_LINE; + break; + case GL_FILL: + break; + } + break; + + case GL_FRONT_AND_BACK: + //fprintf(stderr, "front and back\n"); + hw_mode &= ~PM_NOT_FRONT; + hw_mode &= ~PM_NOT_BACK; + switch (mode) { + case GL_LINE: + hw_mode |= PM_FRONT_LINE; + hw_mode |= PM_BACK_LINE; + break; + case GL_POINT: + hw_mode &= ~PM_FRONT_LINE; + hw_mode &= ~PM_BACK_LINE; + break; + case GL_FILL: + hw_mode = 0; + break; + } + break; + } +#endif + + //if( front and back fill) hw_mode=0; + + if(r300->hw.unk4288.cmd[1] != hw_mode){ + R300_STATECHANGE(r300, unk4288); + r300->hw.unk4288.cmd[1] = hw_mode; + } +} + +/* ============================================================= + * Stencil + */ + + static int translate_stencil_func(int func) + { + switch (func) { + case GL_NEVER: + return R300_ZS_NEVER; + break; + case GL_LESS: + return R300_ZS_LESS; + break; + case GL_EQUAL: + return R300_ZS_EQUAL; + break; + case GL_LEQUAL: + return R300_ZS_LEQUAL; + break; + case GL_GREATER: + return R300_ZS_GREATER; + break; + case GL_NOTEQUAL: + return R300_ZS_NOTEQUAL; + break; + case GL_GEQUAL: + return R300_ZS_GEQUAL; + break; + case GL_ALWAYS: + return R300_ZS_ALWAYS; + break; + } + return 0; + } + + static int translate_stencil_op(int op) +{ + switch (op) { + case GL_KEEP: + return R300_ZS_KEEP; + case GL_ZERO: + return R300_ZS_ZERO; + case GL_REPLACE: + return R300_ZS_REPLACE; + case GL_INCR: + return R300_ZS_INCR; + case GL_DECR: + return R300_ZS_DECR; + case GL_INCR_WRAP_EXT: + return R300_ZS_INCR_WRAP; + case GL_DECR_WRAP_EXT: + return R300_ZS_DECR_WRAP; + case GL_INVERT: + return R300_ZS_INVERT; + default: + WARN_ONCE("Do not know how to translate stencil op"); + return R300_ZS_KEEP; + } +} + +static void r300StencilFunc(GLcontext * ctx, GLenum func, + GLint ref, GLuint mask) +{ + r300ContextPtr rmesa = R300_CONTEXT(ctx); + GLuint refmask = ((ctx->Stencil.Ref[0] << R300_RB3D_ZS2_STENCIL_REF_SHIFT) | + (ctx->Stencil.ValueMask[0] << R300_RB3D_ZS2_STENCIL_MASK_SHIFT)); + + GLuint flag; + + R300_STATECHANGE(rmesa, zs); + + rmesa->hw.zs.cmd[R300_ZS_CNTL_1] &= ~( + (R300_ZS_MASK << R300_RB3D_ZS1_FRONT_FUNC_SHIFT) + | (R300_ZS_MASK << R300_RB3D_ZS1_BACK_FUNC_SHIFT)); + + rmesa->hw.zs.cmd[R300_ZS_CNTL_2] &= ~((R300_RB3D_ZS2_STENCIL_MASK << R300_RB3D_ZS2_STENCIL_REF_SHIFT) | + (R300_RB3D_ZS2_STENCIL_MASK << R300_RB3D_ZS2_STENCIL_MASK_SHIFT)); + + flag = translate_stencil_func(ctx->Stencil.Function[0]); + + rmesa->hw.zs.cmd[R300_ZS_CNTL_1] |= (flag << R300_RB3D_ZS1_FRONT_FUNC_SHIFT) + | (flag << R300_RB3D_ZS1_BACK_FUNC_SHIFT); + rmesa->hw.zs.cmd[R300_ZS_CNTL_2] |= refmask; +} + +static void r300StencilMask(GLcontext * ctx, GLuint mask) +{ + r300ContextPtr rmesa = R300_CONTEXT(ctx); + + R300_STATECHANGE(rmesa, zs); + rmesa->hw.zs.cmd[R300_ZS_CNTL_2] &= ~(R300_RB3D_ZS2_STENCIL_MASK << R300_RB3D_ZS2_STENCIL_WRITE_MASK_SHIFT); + rmesa->hw.zs.cmd[R300_ZS_CNTL_2] |= ctx->Stencil.WriteMask[0] << R300_RB3D_ZS2_STENCIL_WRITE_MASK_SHIFT; +} + + +static void r300StencilOp(GLcontext * ctx, GLenum fail, + GLenum zfail, GLenum zpass) +{ + r300ContextPtr rmesa = R300_CONTEXT(ctx); + + R300_STATECHANGE(rmesa, zs); + /* It is easier to mask what's left.. */ + rmesa->hw.zs.cmd[R300_ZS_CNTL_1] &= + (R300_ZS_MASK << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT) | + (R300_ZS_MASK << R300_RB3D_ZS1_FRONT_FUNC_SHIFT) | + (R300_ZS_MASK << R300_RB3D_ZS1_BACK_FUNC_SHIFT); + + rmesa->hw.zs.cmd[R300_ZS_CNTL_1] |= + (translate_stencil_op(ctx->Stencil.FailFunc[0]) << R300_RB3D_ZS1_FRONT_FAIL_OP_SHIFT) + |(translate_stencil_op(ctx->Stencil.ZFailFunc[0]) << R300_RB3D_ZS1_FRONT_ZFAIL_OP_SHIFT) + |(translate_stencil_op(ctx->Stencil.ZPassFunc[0]) << R300_RB3D_ZS1_FRONT_ZPASS_OP_SHIFT) + |(translate_stencil_op(ctx->Stencil.FailFunc[0]) << R300_RB3D_ZS1_BACK_FAIL_OP_SHIFT) + |(translate_stencil_op(ctx->Stencil.ZFailFunc[0]) << R300_RB3D_ZS1_BACK_ZFAIL_OP_SHIFT) + |(translate_stencil_op(ctx->Stencil.ZPassFunc[0]) << R300_RB3D_ZS1_BACK_ZPASS_OP_SHIFT); +} + +static void r300ClearStencil(GLcontext * ctx, GLint s) +{ + r300ContextPtr rmesa = R300_CONTEXT(ctx); + + rmesa->state.stencil.clear = + ((GLuint) ctx->Stencil.Clear | + (R300_RB3D_ZS2_STENCIL_MASK << R300_RB3D_ZS2_STENCIL_MASK_SHIFT) | + (ctx->Stencil.WriteMask[0] << R300_RB3D_ZS2_STENCIL_WRITE_MASK_SHIFT)); } /* ============================================================= @@ -589,7 +969,7 @@ static void r300Viewport(GLcontext * ctx, GLint x, GLint y, * setting below. Could apply deltas to rescue pipelined viewport * values, or keep the originals hanging around. */ - R200_FIREVERTICES(R200_CONTEXT(ctx)); + R300_FIREVERTICES(R300_CONTEXT(ctx)); r300UpdateWindow(ctx); } @@ -598,6 +978,25 @@ static void r300DepthRange(GLcontext * ctx, GLclampd nearval, GLclampd farval) r300UpdateWindow(ctx); } +/* ============================================================= + * Polygon state + */ +static void r300PolygonOffset(GLcontext * ctx, GLfloat factor, GLfloat units) +{ + r300ContextPtr rmesa = R300_CONTEXT(ctx); + GLfloat constant = units * /*rmesa->state.depth.scale*/4; + + factor *= 12; + +/* fprintf(stderr, "%s f:%f u:%f\n", __FUNCTION__, factor, constant); */ + + R300_STATECHANGE(rmesa, zbs); + rmesa->hw.zbs.cmd[R300_ZBS_T_FACTOR] = r300PackFloat32(factor); + rmesa->hw.zbs.cmd[R300_ZBS_T_CONSTANT] = r300PackFloat32(constant); + rmesa->hw.zbs.cmd[R300_ZBS_W_FACTOR] = r300PackFloat32(factor); + rmesa->hw.zbs.cmd[R300_ZBS_W_CONSTANT] = r300PackFloat32(constant); +} + /* Routing and texture-related */ void r300_setup_routing(GLcontext *ctx, GLboolean immediate) @@ -607,73 +1006,133 @@ void r300_setup_routing(GLcontext *ctx, GLboolean immediate) TNLcontext *tnl = TNL_CONTEXT(ctx); struct vertex_buffer *VB = &tnl->vb; r300ContextPtr r300 = R300_CONTEXT(ctx); - - + + /* Stage 1 - input to VAP */ - + /* Assign register number automatically, retaining it in rmesa->state.reg */ - + /* Note: immediate vertex data includes all coordinates. To save bandwidth use either VBUF or state-based vertex generation */ - - #define CONFIGURE_AOS(v, o, r, f) \ + +#define CONFIGURE_AOS(v, o, r, f) \ {\ - if(immediate){ \ - r300->state.aos[count].element_size=4; \ - r300->state.aos[count].stride=4; \ - r300->state.aos[count].ncomponents=4; \ - } else { \ - r300->state.aos[count].element_size=v->size; \ - r300->state.aos[count].stride=v->size; \ - r300->state.aos[count].ncomponents=v->size; \ - } \ - r300->state.aos[count].offset=o; \ - r300->state.aos[count].reg=reg; \ - r300->state.aos[count].format=(f); \ + if (RADEON_DEBUG & DEBUG_STATE)fprintf(stderr, "Enabling "#r "\n"); \ + r300->state.aos[count].aos_size=4; \ + r300->state.aos[count].aos_stride=4; \ + r300->state.aos[count].aos_offset=o; \ + r300->state.aos[count].aos_reg=reg; \ + r300->state.aos[count].aos_format=(f); \ r300->state.vap_reg.r=reg; \ count++; \ reg++; \ } - - /* All offsets are 0 - for use by immediate mode. + + /* All offsets are 0 - for use by immediate mode. Should change later to handle vertex buffers */ - CONFIGURE_AOS(VB->ObjPtr, 0, i_coords, AOS_FORMAT_FLOAT); - CONFIGURE_AOS(VB->ColorPtr[0], 0, i_color[0], AOS_FORMAT_FLOAT_COLOR); + if(r300->current_vp!=NULL){ + + /* VERT_ATTRIB_WEIGHT, VERT_ATTRIB_SIX, VERT_ATTRIB_SEVEN, VERT_ATTRIB_GENERIC0, + VERT_ATTRIB_GENERIC1, VERT_ATTRIB_GENERIC2, VERT_ATTRIB_GENERIC3 */ + r300->state.render_inputs = 0; + + if(r300->current_vp->inputs[VERT_ATTRIB_POS] != -1){ + reg=r300->current_vp->inputs[VERT_ATTRIB_POS]; + CONFIGURE_AOS(VB->ObjPtr, 0, i_coords, AOS_FORMAT_FLOAT); + r300->state.render_inputs |= _TNL_BIT_POS; + } + if(r300->current_vp->inputs[VERT_ATTRIB_NORMAL] != -1){ + reg=r300->current_vp->inputs[VERT_ATTRIB_NORMAL]; + CONFIGURE_AOS(VB->NormalPtr, 0, i_normal, AOS_FORMAT_FLOAT); + r300->state.render_inputs |= _TNL_BIT_NORMAL; + } + if(r300->current_vp->inputs[VERT_ATTRIB_COLOR0] != -1){ + reg=r300->current_vp->inputs[VERT_ATTRIB_COLOR0]; + CONFIGURE_AOS(VB->ColorPtr[0], 0, i_color[0], AOS_FORMAT_FLOAT_COLOR); + r300->state.render_inputs |= _TNL_BIT_COLOR0; + } + if(r300->current_vp->inputs[VERT_ATTRIB_COLOR1] != -1){ + reg=r300->current_vp->inputs[VERT_ATTRIB_COLOR1]; + CONFIGURE_AOS(VB->SecondaryColorPtr[0], 0, i_color[1], AOS_FORMAT_FLOAT_COLOR); + r300->state.render_inputs |= _TNL_BIT_COLOR1; + } + if(r300->current_vp->inputs[VERT_ATTRIB_FOG] != -1){ + reg=r300->current_vp->inputs[VERT_ATTRIB_FOG]; + CONFIGURE_AOS(VB->FogCoordPtr, 0, i_fog, AOS_FORMAT_FLOAT); + r300->state.render_inputs |= _TNL_BIT_FOG; + } + for(i=0;i < ctx->Const.MaxTextureUnits;i++) // tex 7 is last + if(r300->current_vp->inputs[VERT_ATTRIB_TEX0+i] != -1){ + reg=r300->current_vp->inputs[VERT_ATTRIB_TEX0+i]; + CONFIGURE_AOS(VB->TexCoordPtr[i], 0, i_tex[i], AOS_FORMAT_FLOAT); + r300->state.render_inputs |= _TNL_BIT_TEX0<render_inputs & _TNL_BIT_INDEX)) + CONFIGURE_AOS(VB->IndexPtr[0], 0, i_index, AOS_FORMAT_FLOAT); + + if((tnl->render_inputs & _TNL_BIT_POINTSIZE)) + CONFIGURE_AOS(VB->PointSizePtr, 0, i_pointsize, AOS_FORMAT_FLOAT); +#endif + }else{ + + r300->state.render_inputs = tnl->render_inputs; + + if(tnl->render_inputs & _TNL_BIT_POS) + CONFIGURE_AOS(VB->ObjPtr, 0, i_coords, AOS_FORMAT_FLOAT); + if(tnl->render_inputs & _TNL_BIT_NORMAL) + CONFIGURE_AOS(VB->NormalPtr, 0, i_normal, AOS_FORMAT_FLOAT); + + if(tnl->render_inputs & _TNL_BIT_COLOR0) + CONFIGURE_AOS(VB->ColorPtr[0], 0, i_color[0], AOS_FORMAT_FLOAT_COLOR); + if(tnl->render_inputs & _TNL_BIT_COLOR1) + CONFIGURE_AOS(VB->SecondaryColorPtr[0], 0, i_color[1], AOS_FORMAT_FLOAT_COLOR); + + /*if(tnl->render_inputs & _TNL_BIT_FOG) // Causes lock ups when immediate mode is on + CONFIGURE_AOS(VB->FogCoordPtr, 0, i_fog, AOS_FORMAT_FLOAT);*/ + for(i=0;i < ctx->Const.MaxTextureUnits;i++) - if(ctx->Texture.Unit[i].Enabled) + if(tnl->render_inputs & (_TNL_BIT_TEX0<TexCoordPtr[i], 0, i_tex[i], AOS_FORMAT_FLOAT); - + + if(tnl->render_inputs & _TNL_BIT_INDEX) + CONFIGURE_AOS(VB->IndexPtr[0], 0, i_index, AOS_FORMAT_FLOAT); + if(tnl->render_inputs & _TNL_BIT_POINTSIZE) + CONFIGURE_AOS(VB->PointSizePtr, 0, i_pointsize, AOS_FORMAT_FLOAT); + } + r300->state.aos_count=count; - + if (RADEON_DEBUG & DEBUG_STATE) - fprintf(stderr, "aos_count=%d\n", count); - + fprintf(stderr, "aos_count=%d render_inputs=%08x\n", count, r300->state.render_inputs); + + if(count>R300_MAX_AOS_ARRAYS){ fprintf(stderr, "Aieee ! AOS array count exceeded !\n"); exit(-1); } - + /* Implement AOS */ - + /* setup INPUT_ROUTE */ R300_STATECHANGE(r300, vir[0]); for(i=0;i+1state.aos[i].ncomponents-1) - | ((r300->state.aos[i].reg)<<8) - | (r300->state.aos[i].format<<14) - | (((r300->state.aos[i+1].ncomponents-1) - | ((r300->state.aos[i+1].reg)<<8) - | (r300->state.aos[i+1].format<<14))<<16); - + dw=(r300->state.aos[i].aos_size-1) + | ((r300->state.aos[i].aos_reg)<<8) + | (r300->state.aos[i].aos_format<<14) + | (((r300->state.aos[i+1].aos_size-1) + | ((r300->state.aos[i+1].aos_reg)<<8) + | (r300->state.aos[i+1].aos_format<<14))<<16); + if(i+2==count){ dw|=(1<<(13+16)); } r300->hw.vir[0].cmd[R300_VIR_CNTL_0+(i>>1)]=dw; } if(count & 1){ - dw=(r300->state.aos[count-1].ncomponents-1) - | (r300->state.aos[count-1].format<<14) - | ((r300->state.aos[count-1].reg)<<8) + dw=(r300->state.aos[count-1].aos_size-1) + | (r300->state.aos[count-1].aos_format<<14) + | ((r300->state.aos[count-1].aos_reg)<<8) | (1<<13); r300->hw.vir[0].cmd[R300_VIR_CNTL_0+(count>>1)]=dw; //fprintf(stderr, "vir0 dw=%08x\n", dw); @@ -681,40 +1140,40 @@ void r300_setup_routing(GLcontext *ctx, GLboolean immediate) /* Set the rest of INPUT_ROUTE_0 to 0 */ //for(i=((count+1)>>1); i<8; i++)r300->hw.vir[0].cmd[R300_VIR_CNTL_0+i]=(0x0); ((drm_r300_cmd_header_t*)r300->hw.vir[0].cmd)->unchecked_state.count = (count+1)>>1; - - + + /* Mesa assumes that all missing components are from (0, 0, 0, 1) */ - #define ALL_COMPONENTS ((R300_INPUT_ROUTE_SELECT_X<state.aos[i].ncomponents*3))-1; + mask=(1<<(r300->state.aos[i].aos_size*3))-1; dw=(ALL_COMPONENTS & mask) | (ALL_DEFAULT & ~mask) | R300_INPUT_ROUTE_ENABLE; - + /* i+1 */ - mask=(1<<(r300->state.aos[i+1].ncomponents*3))-1; - dw|=( + mask=(1<<(r300->state.aos[i+1].aos_size*3))-1; + dw|=( (ALL_COMPONENTS & mask) | (ALL_DEFAULT & ~mask) | R300_INPUT_ROUTE_ENABLE )<<16; - + r300->hw.vir[1].cmd[R300_VIR_CNTL_0+(i>>1)]=dw; } if(count & 1){ - mask=(1<<(r300->state.aos[count-1].ncomponents*3))-1; + mask=(1<<(r300->state.aos[count-1].aos_size*3))-1; dw=(ALL_COMPONENTS & mask) | (ALL_DEFAULT & ~mask) | R300_INPUT_ROUTE_ENABLE; @@ -724,35 +1183,43 @@ void r300_setup_routing(GLcontext *ctx, GLboolean immediate) /* Set the rest of INPUT_ROUTE_1 to 0 */ //for(i=((count+1)>>1); i<8; i++)r300->hw.vir[1].cmd[R300_VIR_CNTL_0+i]=0x0; ((drm_r300_cmd_header_t*)r300->hw.vir[1].cmd)->unchecked_state.count = (count+1)>>1; - + /* Set up input_cntl */ - + R300_STATECHANGE(r300, vic); r300->hw.vic.cmd[R300_VIC_CNTL_0]=0x5555; /* Hard coded value, no idea what it means */ - - r300->hw.vic.cmd[R300_VIC_CNTL_1]=R300_INPUT_CNTL_POS - | R300_INPUT_CNTL_COLOR; - + + r300->hw.vic.cmd[R300_VIC_CNTL_1]=0; + + if(r300->state.render_inputs & _TNL_BIT_POS) + r300->hw.vic.cmd[R300_VIC_CNTL_1]|=R300_INPUT_CNTL_POS; + + if(r300->state.render_inputs & _TNL_BIT_NORMAL) + r300->hw.vic.cmd[R300_VIC_CNTL_1]|=R300_INPUT_CNTL_NORMAL; + + if(r300->state.render_inputs & _TNL_BIT_COLOR0) + r300->hw.vic.cmd[R300_VIC_CNTL_1]|=R300_INPUT_CNTL_COLOR; + for(i=0;i < ctx->Const.MaxTextureUnits;i++) - if(ctx->Texture.Unit[i].Enabled) + if(r300->state.render_inputs & (_TNL_BIT_TEX0<hw.vic.cmd[R300_VIC_CNTL_1]|=(R300_INPUT_CNTL_TC0<hw.vof.cmd[R300_VOF_CNTL_0]=R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT | R300_VAP_OUTPUT_VTX_FMT_0__COLOR_PRESENT; - + r300->hw.vof.cmd[R300_VOF_CNTL_1]=0; for(i=0;i < ctx->Const.MaxTextureUnits;i++) - if(ctx->Texture.Unit[i].Enabled) + if(r300->state.render_inputs & (_TNL_BIT_TEX0<hw.vof.cmd[R300_VOF_CNTL_1]|=(4<<(3*i)); - + } static r300TexObj default_tex_obj={ filter:R300_TX_MAG_FILTER_LINEAR | R300_TX_MIN_FILTER_LINEAR, pitch: 0x8000, - size: (0xff << R300_TX_WIDTHMASK_SHIFT) + size: (0xff << R300_TX_WIDTHMASK_SHIFT) | (0xff << R300_TX_HEIGHTMASK_SHIFT) | (0x8 << R300_TX_SIZE_SHIFT), format: 0x88a0c, @@ -761,7 +1228,7 @@ static r300TexObj default_tex_obj={ unknown5: 0x0 }; - /* there is probably a system to these value, but, for now, + /* there is probably a system to these value, but, for now, we just try by hand */ static int inline translate_src(int src) @@ -789,71 +1256,71 @@ static int inline translate_src(int src) return 0; } } - -/* I think 357 and 457 are prime numbers.. wiggle them if you get coincidences */ -#define FORMAT_HASH(opRGB, srcRGB, modeRGB, opA, srcA, modeA, format) ( \ - (\ - ((opRGB)<<30) | ((opA)<<28) | \ - ((srcRGB)<< 25) | ((srcA)<<22) | \ - ((modeRGB)) \ - ) \ - ^ ((modeA)*357) \ - ^ (((format)) *457) \ - ) - -static GLuint translate_texture_format(GLcontext *ctx, GLint tex_unit, GLuint format) + +/* r300 doesnt handle GL_CLAMP and GL_MIRROR_CLAMP_EXT correctly when filter is NEAREST. + * Since texwrap produces same results for GL_CLAMP and GL_CLAMP_TO_EDGE we use them instead. + * We need to recalculate wrap modes whenever filter mode is changed because someone might do: + * glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST); + * glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP); + * glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR); + * Since r300 completely ignores R300_TX_CLAMP when either min or mag is nearest it cant handle + * combinations where only one of them is nearest. + */ +static unsigned long gen_fixed_filter(unsigned long f) { - const struct gl_texture_unit *texUnit= &ctx->Texture.Unit[tex_unit]; - int i=0; /* number of alpha args .. */ - switch(FORMAT_HASH( - texUnit->_CurrentCombine->OperandRGB[i] -GL_SRC_COLOR, - translate_src(texUnit->_CurrentCombine->SourceRGB[i]), - texUnit->_CurrentCombine->ModeRGB, - texUnit->_CurrentCombine->OperandA[i] -GL_SRC_ALPHA, - translate_src(texUnit->_CurrentCombine->SourceA[i]), - texUnit->_CurrentCombine->ModeA, - format - )){ - case FORMAT_HASH(0, 1, 0x2100, 0, 1, 0x2100, 0x00088047): - /* tested with: - kfiresaver.kss - */ - return 0x1a0c; - case FORMAT_HASH(0, 1, 0x2100, 0, 1, 0x2100, 0x00077047): - /* tested with: - kfiresaver.kss - */ - return 0x4ba0c; - case FORMAT_HASH(0, 1, 0x2100, 0, 1, 0x2100, 0x00055047): - /* tested with: - kfiresaver.kss - kfountain.kss - */ - return 0x51a0c; - case FORMAT_HASH(0, 1, 0x2100, 0, 4, 0x1e01, 0x00088047): - /* tested with - lesson 06 - lesson 07 - */ - return 0x53a0c; - case FORMAT_HASH(0, 1, 0x2100, 0, 4, 0x1e01, 0x00055047): - /* Can't remember what I tested this with.. - try putting return 0 of you see broken textures which - are not being complained about */ - return 0x53a0c; - default: - { - static int warn_once=1; - if(warn_once){ - fprintf(stderr, "%s:%s Do not know how to translate texture format - help me !\n", - __FILE__, __FUNCTION__); - warn_once=0; - } - } - return 0; + unsigned long mag, min, needs_fixing=0; + //return f; + + /* We ignore MIRROR bit so we dont have to do everything twice */ + if((f & ((7-1) << R300_TX_WRAP_S_SHIFT)) == (R300_TX_CLAMP << R300_TX_WRAP_S_SHIFT)){ + needs_fixing |= 1; + } + if((f & ((7-1) << R300_TX_WRAP_T_SHIFT)) == (R300_TX_CLAMP << R300_TX_WRAP_T_SHIFT)){ + needs_fixing |= 2; + } + if((f & ((7-1) << R300_TX_WRAP_Q_SHIFT)) == (R300_TX_CLAMP << R300_TX_WRAP_Q_SHIFT)){ + needs_fixing |= 4; + } + + if(!needs_fixing) + return f; + + mag=f & R300_TX_MAG_FILTER_MASK; + min=f & R300_TX_MIN_FILTER_MASK; + + /* TODO: Check for anisto filters too */ + if((mag != R300_TX_MAG_FILTER_NEAREST) && (min != R300_TX_MIN_FILTER_NEAREST)) + return f; + + /* r300 cant handle these modes hence we force nearest to linear */ + if((mag == R300_TX_MAG_FILTER_NEAREST) && (min != R300_TX_MIN_FILTER_NEAREST)){ + f &= ~R300_TX_MAG_FILTER_NEAREST; + f |= R300_TX_MAG_FILTER_LINEAR; + return f; + } + + if((min == R300_TX_MIN_FILTER_NEAREST) && (mag != R300_TX_MAG_FILTER_NEAREST)){ + f &= ~R300_TX_MIN_FILTER_NEAREST; + f |= R300_TX_MIN_FILTER_LINEAR; + return f; + } + + /* Both are nearest */ + if(needs_fixing & 1){ + f &= ~((7-1) << R300_TX_WRAP_S_SHIFT); + f |= R300_TX_CLAMP_TO_EDGE << R300_TX_WRAP_S_SHIFT; + } + if(needs_fixing & 2){ + f &= ~((7-1) << R300_TX_WRAP_T_SHIFT); + f |= R300_TX_CLAMP_TO_EDGE << R300_TX_WRAP_T_SHIFT; } + if(needs_fixing & 4){ + f &= ~((7-1) << R300_TX_WRAP_Q_SHIFT); + f |= R300_TX_CLAMP_TO_EDGE << R300_TX_WRAP_Q_SHIFT; + } + return f; } - + void r300_setup_textures(GLcontext *ctx) { int i, mtu; @@ -869,93 +1336,73 @@ void r300_setup_textures(GLcontext *ctx) R300_STATECHANGE(r300, tex.format); R300_STATECHANGE(r300, tex.offset); R300_STATECHANGE(r300, tex.unknown4); - R300_STATECHANGE(r300, tex.unknown5); + R300_STATECHANGE(r300, tex.border_color); r300->state.texture.tc_count=0; - + r300->hw.txe.cmd[R300_TXE_ENABLE]=0x0; - + mtu = r300->radeon.glCtx->Const.MaxTextureUnits; if (RADEON_DEBUG & DEBUG_STATE) fprintf(stderr, "mtu=%d\n", mtu); - - if(mtu>R300_MAX_TEXTURE_UNITS){ - fprintf(stderr, "Aiiee ! mtu=%d is greater than R300_MAX_TEXTURE_UNITS=%d\n", + + if(mtu > R300_MAX_TEXTURE_UNITS) { + fprintf(stderr, "Aiiee ! mtu=%d is greater than R300_MAX_TEXTURE_UNITS=%d\n", mtu, R300_MAX_TEXTURE_UNITS); exit(-1); + } + + for(i=0; i < mtu; i++) { + /*if(ctx->Texture.Unit[i].Enabled == 0) + continue;*/ + if( ((r300->state.render_inputs & (_TNL_BIT_TEX0<Texture.Unit[i].Enabled)!=0) ) { + WARN_ONCE("Mismatch between render_inputs and ctx->Texture.Unit[i].Enabled value.\n"); } - for(i=0;iTexture.Unit[i].Enabled){ + + if(r300->state.render_inputs & (_TNL_BIT_TEX0<state.texture.unit[i].texobj; + //fprintf(stderr, "format=%08x\n", r300->state.texture.unit[i].format); r300->state.texture.tc_count++; - if(t==NULL){ + + if(t == NULL){ fprintf(stderr, "Texture unit %d enabled, but corresponding texobj is NULL, using default object.\n", i); //exit(-1); t=&default_tex_obj; - } + } + + //fprintf(stderr, "t->format=%08x\n", t->format); + if((t->format & 0xffffff00)==0xffffff00) { + WARN_ONCE("unknown texture format (entry %x) encountered. Help me !\n", t->format & 0xff); + //fprintf(stderr, "t->format=%08x\n", t->format); + } + if (RADEON_DEBUG & DEBUG_STATE) fprintf(stderr, "Activating texture unit %d\n", i); max_texture_unit=i; r300->hw.txe.cmd[R300_TXE_ENABLE]|=(1<hw.tex.filter.cmd[R300_TEX_VALUE_0+i]=t->filter; - r300->hw.tex.unknown1.cmd[R300_TEX_VALUE_0+i]=t->pitch; + r300->hw.tex.filter.cmd[R300_TEX_VALUE_0+i]=gen_fixed_filter(t->filter) | (i << 28); + r300->hw.tex.unknown1.cmd[R300_TEX_VALUE_0+i]=0x0; + + /* No idea why linear filtered textures shake when puting random data */ + /*r300->hw.tex.unknown1.cmd[R300_TEX_VALUE_0+i]=(rand()%0xffffffff) & (~0x1fff);*/ r300->hw.tex.size.cmd[R300_TEX_VALUE_0+i]=t->size; r300->hw.tex.format.cmd[R300_TEX_VALUE_0+i]=t->format; + //fprintf(stderr, "t->format=%08x\n", t->format); r300->hw.tex.offset.cmd[R300_TEX_VALUE_0+i]=r300->radeon.radeonScreen->fbLocation+t->offset; r300->hw.tex.unknown4.cmd[R300_TEX_VALUE_0+i]=0x0; - r300->hw.tex.unknown5.cmd[R300_TEX_VALUE_0+i]=0x0; - - - /* We don't know how to set this yet */ - //value from r300_lib.c for RGB24 - //r300->hw.tex.format.cmd[R300_TEX_VALUE_0+i]=0x88a0c; - r300->hw.tex.format.cmd[R300_TEX_VALUE_0+i]=translate_texture_format(ctx, i, t->format); - /* Use the code below to quickly find matching texture - formats. Requires an app that displays the same texture - repeatedly */ - #if 1 - if(r300->hw.tex.format.cmd[R300_TEX_VALUE_0+i]==0){ - static int fmt=1; - static int k=0; - k++; - if(k>200){ - k=0; - fmt+=2; - texUnit = &ctx->Texture.Unit[i]; - fprintf(stderr, "Want to set FORMAT_HASH(%d, %d, 0x%04x, %d, %d, 0x%04x, 0x%08x)\n", - texUnit->_CurrentCombine->OperandRGB[0] -GL_SRC_COLOR, - translate_src(texUnit->_CurrentCombine->SourceRGB[0]), - texUnit->_CurrentCombine->ModeRGB, - texUnit->_CurrentCombine->OperandA[0] -GL_SRC_ALPHA, - translate_src(texUnit->_CurrentCombine->SourceA[0]), - texUnit->_CurrentCombine->ModeA, - t->format - ); - if(fmt>0xff){ - //exit(-1); - fmt=0; - } - //sleep(1); - fprintf(stderr, "Now trying format %08x\n", - 0x00a0c | (fmt<<12)); - fprintf(stderr, "size=%08x\n", t->size); - } - r300->hw.tex.format.cmd[R300_TEX_VALUE_0+i]=0x00a0c | (fmt<<12); - } - #endif - - } - + r300->hw.tex.border_color.cmd[R300_TEX_VALUE_0+i]=t->pp_border_color; } + } + ((drm_r300_cmd_header_t*)r300->hw.tex.filter.cmd)->unchecked_state.count = max_texture_unit+1; ((drm_r300_cmd_header_t*)r300->hw.tex.unknown1.cmd)->unchecked_state.count = max_texture_unit+1; ((drm_r300_cmd_header_t*)r300->hw.tex.size.cmd)->unchecked_state.count = max_texture_unit+1; ((drm_r300_cmd_header_t*)r300->hw.tex.format.cmd)->unchecked_state.count = max_texture_unit+1; ((drm_r300_cmd_header_t*)r300->hw.tex.offset.cmd)->unchecked_state.count = max_texture_unit+1; ((drm_r300_cmd_header_t*)r300->hw.tex.unknown4.cmd)->unchecked_state.count = max_texture_unit+1; - ((drm_r300_cmd_header_t*)r300->hw.tex.unknown5.cmd)->unchecked_state.count = max_texture_unit+1; - + ((drm_r300_cmd_header_t*)r300->hw.tex.border_color.cmd)->unchecked_state.count = max_texture_unit+1; + if (RADEON_DEBUG & DEBUG_STATE) fprintf(stderr, "TX_ENABLE: %08x max_texture_unit=%d\n", r300->hw.txe.cmd[R300_TXE_ENABLE], max_texture_unit); } @@ -963,47 +1410,539 @@ void r300_setup_textures(GLcontext *ctx) void r300_setup_rs_unit(GLcontext *ctx) { r300ContextPtr r300 = R300_CONTEXT(ctx); - int i; + int i, cur_reg; + GLuint interp_magic[8] = { + 0x00, + 0x40, + 0x80, + 0xC0, + 0x00, + 0x00, + 0x00, + 0x00 + }; /* This needs to be rewritten - it is a hack at best */ - + R300_STATECHANGE(r300, ri); R300_STATECHANGE(r300, rc); R300_STATECHANGE(r300, rr); +#if 1 + cur_reg = 0; + for (i=0;iConst.MaxTextureUnits;i++) { + r300->hw.ri.cmd[R300_RI_INTERP_0+i] = 0 + | R300_RS_INTERP_USED + | (cur_reg << R300_RS_INTERP_SRC_SHIFT) + | interp_magic[i]; +// fprintf(stderr, "RS_INTERP[%d] = 0x%x\n", i, r300->hw.ri.cmd[R300_RI_INTERP_0+i]); + + if (r300->state.render_inputs & (_TNL_BIT_TEX0<hw.rr.cmd[R300_RR_ROUTE_0 + cur_reg] = 0 + | R300_RS_ROUTE_ENABLE + | i /* source INTERP */ + | (cur_reg << R300_RS_ROUTE_DEST_SHIFT); +// fprintf(stderr, "RS_ROUTE[%d] = 0x%x\n", cur_reg, r300->hw.rr.cmd[R300_RR_ROUTE_0 + cur_reg]); + cur_reg++; + } + } + if (r300->state.render_inputs & _TNL_BIT_COLOR0) + r300->hw.rr.cmd[R300_RR_ROUTE_0] |= 0 + | R300_RS_ROUTE_0_COLOR + | (cur_reg << R300_RS_ROUTE_0_COLOR_DEST_SHIFT); + r300->hw.rr.cmd[R300_RR_CMD_0] = cmducs(R300_RS_ROUTE_0, cur_reg); +// fprintf(stderr, "ADJ_RR0 = 0x%x\n", r300->hw.rr.cmd[R300_RR_ROUTE_0]); + +// fprintf(stderr, "rendering with %d texture co-ordinate sets\n", cur_reg); + + r300->hw.rc.cmd[1] = 0 + | (cur_reg /* count */ << R300_RS_CNTL_TC_CNT_SHIFT) + | R300_RS_CNTL_0_UNKNOWN_7 + | R300_RS_CNTL_0_UNKNOWN_18; + r300->hw.rc.cmd[2] = (0xC0 | (cur_reg-1) /* index of highest */ ); +#else for(i = 1; i <= 8; ++i) r300->hw.ri.cmd[i] = 0x00d10000; r300->hw.ri.cmd[R300_RI_INTERP_1] |= R300_RS_INTERP_1_UNKNOWN; r300->hw.ri.cmd[R300_RI_INTERP_2] |= R300_RS_INTERP_2_UNKNOWN; r300->hw.ri.cmd[R300_RI_INTERP_3] |= R300_RS_INTERP_3_UNKNOWN; - + +#if 1 + for(i = 2; i <= 8; ++i) + r300->hw.ri.cmd[i] |= 4; +#endif + for(i = 1; i <= 8; ++i) r300->hw.rr.cmd[i] = 0; /* textures enabled ? */ if(r300->state.texture.tc_count>0){ - + /* This code only really works with one set of texture coordinates */ - + /* The second constant is needed to get glxgears display anything .. */ - r300->hw.rc.cmd[1] = R300_RS_CNTL_0_UNKNOWN_7 - | R300_RS_CNTL_0_UNKNOWN_18 + r300->hw.rc.cmd[1] = R300_RS_CNTL_0_UNKNOWN_7 + | R300_RS_CNTL_0_UNKNOWN_18 | (r300->state.texture.tc_count<hw.rc.cmd[2] = 0xc0; - - + + ((drm_r300_cmd_header_t*)r300->hw.rr.cmd)->unchecked_state.count = 1; r300->hw.rr.cmd[R300_RR_ROUTE_0] = 0x24008; - + } else { - + /* The second constant is needed to get glxgears display anything .. */ r300->hw.rc.cmd[1] = R300_RS_CNTL_0_UNKNOWN_7 | R300_RS_CNTL_0_UNKNOWN_18; r300->hw.rc.cmd[2] = 0; - + ((drm_r300_cmd_header_t*)r300->hw.rr.cmd)->unchecked_state.count = 1; r300->hw.rr.cmd[R300_RR_ROUTE_0] = 0x4000; + + } +#endif +} + +#define vpucount(ptr) (((drm_r300_cmd_header_t*)(ptr))->vpu.count) + +#define bump_vpu_count(ptr, new_count) do{\ + drm_r300_cmd_header_t* _p=((drm_r300_cmd_header_t*)(ptr));\ + int _nc=(new_count)/4; \ + if(_nc>_p->vpu.count)_p->vpu.count=_nc;\ + }while(0) + +void static inline setup_vertex_shader_fragment(r300ContextPtr r300, int dest, struct r300_vertex_shader_fragment *vsf) +{ + int i; + + if(vsf->length==0)return; + + if(vsf->length & 0x3){ + fprintf(stderr,"VERTEX_SHADER_FRAGMENT must have length divisible by 4\n"); + exit(-1); + } + + switch((dest>>8) & 0xf){ + case 0: + R300_STATECHANGE(r300, vpi); + for(i=0;ilength;i++) + r300->hw.vpi.cmd[R300_VPI_INSTR_0+i+4*(dest & 0xff)]=(vsf->body.d[i]); + bump_vpu_count(r300->hw.vpi.cmd, vsf->length+4*(dest & 0xff)); + break; + + case 2: + R300_STATECHANGE(r300, vpp); + for(i=0;ilength;i++) + r300->hw.vpp.cmd[R300_VPP_PARAM_0+i+4*(dest & 0xff)]=(vsf->body.d[i]); + bump_vpu_count(r300->hw.vpp.cmd, vsf->length+4*(dest & 0xff)); + break; + case 4: + R300_STATECHANGE(r300, vps); + for(i=0;ilength;i++) + r300->hw.vps.cmd[1+i+4*(dest & 0xff)]=(vsf->body.d[i]); + bump_vpu_count(r300->hw.vps.cmd, vsf->length+4*(dest & 0xff)); + break; + default: + fprintf(stderr, "%s:%s don't know how to handle dest %04x\n", __FILE__, __FUNCTION__, dest); + exit(-1); + } +} + +void r300SetupVertexProgram(r300ContextPtr rmesa); + +/* just a skeleton for now.. */ + +/* Generate a vertex shader that simply transforms vertex and texture coordinates, + while leaving colors intact. Nothing fancy (like lights) + + If implementing lights make a copy first, so it is easy to switch between the two versions */ +void r300GenerateSimpleVertexShader(r300ContextPtr r300) +{ + int i; + + /* Allocate parameters */ + r300->state.vap_param.transform_offset=0x0; /* transform matrix */ + r300->state.vertex_shader.param_offset=0x0; + r300->state.vertex_shader.param_count=0x4; /* 4 vector values - 4x4 matrix */ + + r300->state.vertex_shader.program_start=0x0; + r300->state.vertex_shader.unknown_ptr1=0x4; /* magic value ? */ + r300->state.vertex_shader.program_end=0x0; + + r300->state.vertex_shader.unknown_ptr2=0x0; /* magic value */ + r300->state.vertex_shader.unknown_ptr3=0x4; /* magic value */ + + /* Initialize matrix and vector parameters.. these should really be restructured */ + /* TODO: fix vertex_shader structure */ + r300->state.vertex_shader.matrix[0].length=16; + r300->state.vertex_shader.matrix[1].length=0; + r300->state.vertex_shader.matrix[2].length=0; + r300->state.vertex_shader.vector[0].length=0; + r300->state.vertex_shader.vector[1].length=0; + r300->state.vertex_shader.unknown1.length=0; + r300->state.vertex_shader.unknown2.length=0; + +#define WRITE_OP(oper,source1,source2,source3) {\ + r300->state.vertex_shader.program.body.i[r300->state.vertex_shader.program_end].op=(oper); \ + r300->state.vertex_shader.program.body.i[r300->state.vertex_shader.program_end].src1=(source1); \ + r300->state.vertex_shader.program.body.i[r300->state.vertex_shader.program_end].src2=(source2); \ + r300->state.vertex_shader.program.body.i[r300->state.vertex_shader.program_end].src3=(source3); \ + r300->state.vertex_shader.program_end++; \ + } + + /* Multiply vertex coordinates with transform matrix */ + + WRITE_OP( + EASY_VSF_OP(MUL, 0, ALL, TMP), + VSF_PARAM(3), + VSF_ATTR_W(0), + EASY_VSF_SOURCE(0, W, W, W, W, NONE, NONE) + ) + + WRITE_OP( + EASY_VSF_OP(MUL, 1, ALL, RESULT), + VSF_REG(1), + VSF_ATTR_UNITY(1), + VSF_UNITY(1) + ) + + WRITE_OP( + EASY_VSF_OP(MAD, 0, ALL, TMP), + VSF_PARAM(2), + VSF_ATTR_Z(0), + VSF_TMP(0) + ) + + WRITE_OP( + EASY_VSF_OP(MAD, 0, ALL, TMP), + VSF_PARAM(1), + VSF_ATTR_Y(0), + VSF_TMP(0) + ) + + WRITE_OP( + EASY_VSF_OP(MAD, 0, ALL, RESULT), + VSF_PARAM(0), + VSF_ATTR_X(0), + VSF_TMP(0) + ) + /* Pass through texture coordinates, if any */ + for(i=0;i < r300->radeon.glCtx->Const.MaxTextureUnits;i++) + if(r300->state.render_inputs & (_TNL_BIT_TEX0<state.vap_reg.i_tex[i]); + WRITE_OP( + EASY_VSF_OP(MUL, 2+i, ALL, RESULT), + VSF_REG(r300->state.vap_reg.i_tex[i]), + VSF_ATTR_UNITY(r300->state.vap_reg.i_tex[i]), + VSF_UNITY(r300->state.vap_reg.i_tex[i]) + ) + } + + r300->state.vertex_shader.program_end--; /* r300 wants program length to be one more - no idea why */ + r300->state.vertex_shader.program.length=(r300->state.vertex_shader.program_end+1)*4; + + r300->state.vertex_shader.unknown_ptr1=r300->state.vertex_shader.program_end; /* magic value ? */ + r300->state.vertex_shader.unknown_ptr2=r300->state.vertex_shader.program_end; /* magic value ? */ + r300->state.vertex_shader.unknown_ptr3=r300->state.vertex_shader.program_end; /* magic value ? */ + +} + + +void r300SetupVertexShader(r300ContextPtr rmesa) +{ + GLcontext* ctx = rmesa->radeon.glCtx; + struct r300_vertex_shader_fragment unk4={ + length: 4, + body: { f: { + /*0.0*/(rand()%100)/10.0, + /*0.0*/(rand()%100)/10.0, + /*1.0*/(rand()%100)/10.0, + /*0.0*/(rand()%100)/10.0 + } } + }; + LOCAL_VARS + + /* Reset state, in case we don't use something */ + ((drm_r300_cmd_header_t*)rmesa->hw.vpp.cmd)->vpu.count = 0; + ((drm_r300_cmd_header_t*)rmesa->hw.vpi.cmd)->vpu.count = 0; + ((drm_r300_cmd_header_t*)rmesa->hw.vps.cmd)->vpu.count = 0; + + /* Not sure why this doesnt work... + 0x400 area might have something to do with pixel shaders as it appears right after pfs programming. + 0x406 is set to { 0.0, 0.0, 1.0, 0.0 } most of the time but should change with smooth points and in other rare cases. */ + //setup_vertex_shader_fragment(rmesa, 0x406, &unk4); + if(rmesa->current_vp != NULL){ + r300SetupVertexProgram(rmesa); + return ; + } + +/* This needs to be replaced by vertex shader generation code */ + + +#if 0 + /* textures enabled ? */ + if(rmesa->state.texture.tc_count>0){ + rmesa->state.vertex_shader=SINGLE_TEXTURE_VERTEX_SHADER; + } else { + rmesa->state.vertex_shader=FLAT_COLOR_VERTEX_SHADER; + } +#endif + + r300GenerateSimpleVertexShader(rmesa); + + rmesa->state.vertex_shader.matrix[0].length=16; + memcpy(rmesa->state.vertex_shader.matrix[0].body.f, ctx->_ModelProjectMatrix.m, 16*4); + + setup_vertex_shader_fragment(rmesa, VSF_DEST_PROGRAM, &(rmesa->state.vertex_shader.program)); + + setup_vertex_shader_fragment(rmesa, VSF_DEST_MATRIX0, &(rmesa->state.vertex_shader.matrix[0])); +#if 0 + setup_vertex_shader_fragment(rmesa, VSF_DEST_MATRIX1, &(rmesa->state.vertex_shader.matrix[0])); + setup_vertex_shader_fragment(rmesa, VSF_DEST_MATRIX2, &(rmesa->state.vertex_shader.matrix[0])); + + setup_vertex_shader_fragment(rmesa, VSF_DEST_VECTOR0, &(rmesa->state.vertex_shader.vector[0])); + setup_vertex_shader_fragment(rmesa, VSF_DEST_VECTOR1, &(rmesa->state.vertex_shader.vector[1])); +#endif + +#if 0 + setup_vertex_shader_fragment(rmesa, VSF_DEST_UNKNOWN1, &(rmesa->state.vertex_shader.unknown1)); + setup_vertex_shader_fragment(rmesa, VSF_DEST_UNKNOWN2, &(rmesa->state.vertex_shader.unknown2)); +#endif + + R300_STATECHANGE(rmesa, pvs); + rmesa->hw.pvs.cmd[R300_PVS_CNTL_1]=(rmesa->state.vertex_shader.program_start << R300_PVS_CNTL_1_PROGRAM_START_SHIFT) + | (rmesa->state.vertex_shader.unknown_ptr1 << R300_PVS_CNTL_1_UNKNOWN_SHIFT) + | (rmesa->state.vertex_shader.program_end << R300_PVS_CNTL_1_PROGRAM_END_SHIFT); + rmesa->hw.pvs.cmd[R300_PVS_CNTL_2]=(rmesa->state.vertex_shader.param_offset << R300_PVS_CNTL_2_PARAM_OFFSET_SHIFT) + | (rmesa->state.vertex_shader.param_count << R300_PVS_CNTL_2_PARAM_COUNT_SHIFT); + rmesa->hw.pvs.cmd[R300_PVS_CNTL_3]=(rmesa->state.vertex_shader.unknown_ptr2 << R300_PVS_CNTL_3_PROGRAM_UNKNOWN_SHIFT) + | (rmesa->state.vertex_shader.unknown_ptr3 << 0); + + /* This is done for vertex shader fragments, but also needs to be done for vap_pvs, + so I leave it as a reminder */ +#if 0 + reg_start(R300_VAP_PVS_WAITIDLE,0); + e32(0x00000000); +#endif +} + +void r300SetupVertexProgram(r300ContextPtr rmesa) +{ + GLcontext* ctx = rmesa->radeon.glCtx; + int inst_count; + int param_count; + LOCAL_VARS + + + /* Reset state, in case we don't use something */ + ((drm_r300_cmd_header_t*)rmesa->hw.vpp.cmd)->vpu.count = 0; + ((drm_r300_cmd_header_t*)rmesa->hw.vpi.cmd)->vpu.count = 0; + ((drm_r300_cmd_header_t*)rmesa->hw.vps.cmd)->vpu.count = 0; + + r300VertexProgUpdateParams(ctx, rmesa->current_vp); + + setup_vertex_shader_fragment(rmesa, VSF_DEST_PROGRAM, &(rmesa->current_vp->program)); + + setup_vertex_shader_fragment(rmesa, VSF_DEST_MATRIX0, &(rmesa->current_vp->params)); + +#if 0 + setup_vertex_shader_fragment(rmesa, VSF_DEST_UNKNOWN1, &(rmesa->state.vertex_shader.unknown1)); + setup_vertex_shader_fragment(rmesa, VSF_DEST_UNKNOWN2, &(rmesa->state.vertex_shader.unknown2)); +#endif + + inst_count=rmesa->current_vp->program.length/4 - 1; + param_count=rmesa->current_vp->params.length/4; + + R300_STATECHANGE(rmesa, pvs); + rmesa->hw.pvs.cmd[R300_PVS_CNTL_1]=(0 << R300_PVS_CNTL_1_PROGRAM_START_SHIFT) + | (inst_count/*0*/ << R300_PVS_CNTL_1_UNKNOWN_SHIFT) + | (inst_count << R300_PVS_CNTL_1_PROGRAM_END_SHIFT); + rmesa->hw.pvs.cmd[R300_PVS_CNTL_2]=(0 << R300_PVS_CNTL_2_PARAM_OFFSET_SHIFT) + | (param_count << R300_PVS_CNTL_2_PARAM_COUNT_SHIFT); + rmesa->hw.pvs.cmd[R300_PVS_CNTL_3]=(0/*rmesa->state.vertex_shader.unknown_ptr2*/ << R300_PVS_CNTL_3_PROGRAM_UNKNOWN_SHIFT) + | ((inst_count-rmesa->current_vp->t2rs) /*rmesa->state.vertex_shader.unknown_ptr3*/ << 0); + + /* This is done for vertex shader fragments, but also needs to be done for vap_pvs, + so I leave it as a reminder */ +#if 0 + reg_start(R300_VAP_PVS_WAITIDLE,0); + e32(0x00000000); +#endif +} + + +/* just a skeleton for now.. */ +void r300GenerateTexturePixelShader(r300ContextPtr r300) +{ + int i, mtu; + mtu = r300->radeon.glCtx->Const.MaxTextureUnits; + GLenum envMode; + + int tex_inst=0, alu_inst=0; + + for(i=0;istate.render_inputs & (_TNL_BIT_TEX0<radeon.glCtx->Texture.Unit[i].EnvMode; + //fprintf(stderr, "envMode=%s\n", _mesa_lookup_enum_by_nr(envMode)); + + /* Fetch textured pixel */ + + r300->state.pixel_shader.program.tex.inst[tex_inst]=0x00018000; + tex_inst++; + + switch(r300->radeon.glCtx->Texture.Unit[i]._CurrentCombine->ModeRGB){ + case GL_REPLACE: + WARN_ONCE("ModeA==GL_REPLACE is possibly broken.\n"); + r300->state.pixel_shader.program.alu.inst[alu_inst].inst0= + EASY_PFS_INSTR0(MAD, SRC0C_XYZ, ONE, ZERO); + + r300->state.pixel_shader.program.alu.inst[alu_inst].inst1= + EASY_PFS_INSTR1(0, 0, 0 | PFS_FLAG_CONST, 0 | PFS_FLAG_CONST, NONE, ALL); + break; + case GL_MODULATE: + WARN_ONCE("ModeRGB==GL_MODULATE is possibly broken.\n"); + r300->state.pixel_shader.program.alu.inst[alu_inst].inst0= + EASY_PFS_INSTR0(MAD, SRC0C_XYZ, SRC1C_XYZ, ZERO); + + r300->state.pixel_shader.program.alu.inst[alu_inst].inst1= + EASY_PFS_INSTR1(0, 0, 1, 0 | PFS_FLAG_CONST, NONE, ALL); + + break; + default: + WARN_ONCE("ModeRGB=%s is not implemented yet !\n", + _mesa_lookup_enum_by_nr(r300->radeon.glCtx->Texture.Unit[i]._CurrentCombine->ModeRGB)); + /* PFS_NOP */ + r300->state.pixel_shader.program.alu.inst[alu_inst].inst0= + EASY_PFS_INSTR0(MAD, SRC0C_XYZ, ONE, ZERO); + + r300->state.pixel_shader.program.alu.inst[alu_inst].inst1= + EASY_PFS_INSTR1(0, 0, 0 | PFS_FLAG_CONST, 0 | PFS_FLAG_CONST, NONE, ALL); + } + switch(r300->radeon.glCtx->Texture.Unit[i]._CurrentCombine->ModeA){ + case GL_REPLACE: + WARN_ONCE("ModeA==GL_REPLACE is possibly broken.\n"); + r300->state.pixel_shader.program.alu.inst[alu_inst].inst2= + EASY_PFS_INSTR2(MAD, SRC0A, ONE, ZERO); + + r300->state.pixel_shader.program.alu.inst[alu_inst].inst3= + EASY_PFS_INSTR3(0, 0, 0| PFS_FLAG_CONST, 0 | PFS_FLAG_CONST, OUTPUT); + +#if 0 + fprintf(stderr, "numArgsA=%d sourceA[0]=%s op=%d\n", + r300->radeon.glCtx->Texture.Unit[i]._CurrentCombine->_NumArgsA, + _mesa_lookup_enum_by_nr(r300->radeon.glCtx->Texture.Unit[i]._CurrentCombine->SourceA[0]), + r300->radeon.glCtx->Texture.Unit[i]._CurrentCombine->OperandA[0]-GL_SRC_ALPHA); +#endif + break; + case GL_MODULATE: + WARN_ONCE("ModeA==GL_MODULATE is possibly broken.\n"); + + r300->state.pixel_shader.program.alu.inst[alu_inst].inst2= + EASY_PFS_INSTR2(MAD, SRC0A, SRC1A, ZERO); + + r300->state.pixel_shader.program.alu.inst[alu_inst].inst3= + EASY_PFS_INSTR3(0, 0, 1, 0 | PFS_FLAG_CONST, OUTPUT); + + break; + default: + WARN_ONCE("ModeA=%s is not implemented yet !\n", + _mesa_lookup_enum_by_nr(r300->radeon.glCtx->Texture.Unit[i]._CurrentCombine->ModeA)); + /* PFS_NOP */ + r300->state.pixel_shader.program.alu.inst[alu_inst].inst2= + EASY_PFS_INSTR2(MAD, SRC0A, ONE, ZERO); + + r300->state.pixel_shader.program.alu.inst[alu_inst].inst3= + EASY_PFS_INSTR3(0, 0, 0 | PFS_FLAG_CONST, 0 | PFS_FLAG_CONST, OUTPUT); + + } + + alu_inst++; + } + + r300->state.pixel_shader.program.tex.length=tex_inst; + r300->state.pixel_shader.program.tex_offset=0; + r300->state.pixel_shader.program.tex_end=tex_inst-1; + +#if 0 + /* saturate last instruction, like i915 driver does */ + r300->state.pixel_shader.program.alu.inst[alu_inst-1].inst0|=R300_FPI0_OUTC_SAT; + r300->state.pixel_shader.program.alu.inst[alu_inst-1].inst2|=R300_FPI2_OUTA_SAT; +#endif + + r300->state.pixel_shader.program.alu.length=alu_inst; + r300->state.pixel_shader.program.alu_offset=0; + r300->state.pixel_shader.program.alu_end=alu_inst-1; +} + +void r300SetupPixelShader(r300ContextPtr rmesa) +{ +int i,k; + + /* This needs to be replaced by pixel shader generation code */ + + /* textures enabled ? */ + if(rmesa->state.texture.tc_count>0){ +#if 1 + r300GenerateTextureFragmentShader(rmesa); +#else + rmesa->state.pixel_shader=SINGLE_TEXTURE_PIXEL_SHADER; + r300GenerateTexturePixelShader(rmesa); +#endif + } else { + rmesa->state.pixel_shader=FLAT_COLOR_PIXEL_SHADER; + } + + R300_STATECHANGE(rmesa, fpt); + for(i=0;istate.pixel_shader.program.tex.length;i++) + rmesa->hw.fpt.cmd[R300_FPT_INSTR_0+i]=rmesa->state.pixel_shader.program.tex.inst[i]; + rmesa->hw.fpt.cmd[R300_FPT_CMD_0]=cmducs(R300_PFS_TEXI_0, rmesa->state.pixel_shader.program.tex.length); + +#define OUTPUT_FIELD(st, reg, field) \ + R300_STATECHANGE(rmesa, st); \ + for(i=0;istate.pixel_shader.program.alu.length;i++) \ + rmesa->hw.st.cmd[R300_FPI_INSTR_0+i]=rmesa->state.pixel_shader.program.alu.inst[i].field;\ + rmesa->hw.st.cmd[R300_FPI_CMD_0]=cmducs(reg, rmesa->state.pixel_shader.program.alu.length); + + OUTPUT_FIELD(fpi[0], R300_PFS_INSTR0_0, inst0); + OUTPUT_FIELD(fpi[1], R300_PFS_INSTR1_0, inst1); + OUTPUT_FIELD(fpi[2], R300_PFS_INSTR2_0, inst2); + OUTPUT_FIELD(fpi[3], R300_PFS_INSTR3_0, inst3); +#undef OUTPUT_FIELD + + R300_STATECHANGE(rmesa, fp); + for(i=0;i<4;i++){ + rmesa->hw.fp.cmd[R300_FP_NODE0+i]= + (rmesa->state.pixel_shader.program.node[i].alu_offset << R300_PFS_NODE_ALU_OFFSET_SHIFT) + | (rmesa->state.pixel_shader.program.node[i].alu_end << R300_PFS_NODE_ALU_END_SHIFT) + | (rmesa->state.pixel_shader.program.node[i].tex_offset << R300_PFS_NODE_TEX_OFFSET_SHIFT) + | (rmesa->state.pixel_shader.program.node[i].tex_end << R300_PFS_NODE_TEX_END_SHIFT) + | ( (i==3) ? R300_PFS_NODE_LAST_NODE : 0); } + + /* PFS_CNTL_0 */ + rmesa->hw.fp.cmd[R300_FP_CNTL0]= + (rmesa->state.pixel_shader.program.active_nodes-1) + | (rmesa->state.pixel_shader.program.first_node_has_tex<<3); + /* PFS_CNTL_1 */ + rmesa->hw.fp.cmd[R300_FP_CNTL1]=rmesa->state.pixel_shader.program.temp_register_count; + /* PFS_CNTL_2 */ + rmesa->hw.fp.cmd[R300_FP_CNTL2]= + (rmesa->state.pixel_shader.program.alu_offset << R300_PFS_CNTL_ALU_OFFSET_SHIFT) + | (rmesa->state.pixel_shader.program.alu_end << R300_PFS_CNTL_ALU_END_SHIFT) + | (rmesa->state.pixel_shader.program.tex_offset << R300_PFS_CNTL_TEX_OFFSET_SHIFT) + | (rmesa->state.pixel_shader.program.tex_end << R300_PFS_CNTL_TEX_END_SHIFT); + + R300_STATECHANGE(rmesa, fpp); + for(i=0;istate.pixel_shader.param_length;i++){ + rmesa->hw.fpp.cmd[R300_FPP_PARAM_0+4*i+0]=r300PackFloat32(rmesa->state.pixel_shader.param[i].x); + rmesa->hw.fpp.cmd[R300_FPP_PARAM_0+4*i+1]=r300PackFloat32(rmesa->state.pixel_shader.param[i].y); + rmesa->hw.fpp.cmd[R300_FPP_PARAM_0+4*i+2]=r300PackFloat32(rmesa->state.pixel_shader.param[i].z); + rmesa->hw.fpp.cmd[R300_FPP_PARAM_0+4*i+3]=r300PackFloat32(rmesa->state.pixel_shader.param[i].w); + } + rmesa->hw.fpp.cmd[R300_FPP_CMD_0]=cmducs(R300_PFS_PARAM_0_X, rmesa->state.pixel_shader.param_length); + } /** @@ -1023,7 +1962,53 @@ static void r300InvalidateState(GLcontext * ctx, GLuint new_state) r300ResetHwState(r300); } - +/* Checks that r300ResetHwState actually modifies all states. + Should probably be burried in somewhere else as this file is getting longish. */ +void verify_r300ResetHwState(r300ContextPtr r300, int stage) +{ + struct r300_state_atom* atom; + int i; + drm_r300_cmd_header_t cmd; + + if(stage){ /* mess around with states */ + unsigned long fp1, cb1; + + fp1=r300->hw.fp.cmd[R300_FP_CMD_1]; /* some special cases... */ + cb1=r300->hw.cb.cmd[R300_CB_CMD_1]; + + fprintf(stderr, "verify begin:\n"); + + foreach(atom, &r300->hw.atomlist) { + for(i=1; i < (*atom->check)(r300, atom); i++) + atom->cmd[i]=0xdeadbeef; + } + r300->hw.fp.cmd[R300_FP_CMD_1]=fp1; + r300->hw.cb.cmd[R300_CB_CMD_1]=cb1; + + foreach(atom, &r300->hw.atomlist) { + cmd.u=atom->cmd[0]; + switch(cmd.header.cmd_type){ + case R300_CMD_UNCHECKED_STATE: + case R300_CMD_VPU: + case R300_CMD_PACKET3: + case R300_CMD_END3D: + case R300_CMD_CP_DELAY: + case R300_CMD_DMA_DISCARD: + break; + default: fprintf(stderr, "unknown cmd_type %d in atom %s\n", + cmd.header.cmd_type, atom->name); + } + + } + } else { /* check that they were set */ + foreach(atom, &r300->hw.atomlist) { + for(i=1; i < (*atom->check)(r300, atom); i++) + if(atom->cmd[i]==0xdeadbeef) + fprintf(stderr, "atom %s is untouched\n", atom->name); + } + } +} + /** * Completely recalculates hardware state based on the Mesa state. */ @@ -1035,8 +2020,30 @@ void r300ResetHwState(r300ContextPtr r300) if (RADEON_DEBUG & DEBUG_STATE) fprintf(stderr, "%s\n", __FUNCTION__); + //verify_r300ResetHwState(r300, 1); + + /* This is a place to initialize registers which + have bitfields accessed by different functions + and not all bits are used */ +#if 0 + /* initialize similiar to r200 */ + r300->hw.zs.cmd[R300_ZS_CNTL_0] = 0; + r300->hw.zs.cmd[R300_ZS_CNTL_1] = + (R300_ZS_ALWAYS << R300_RB3D_ZS1_FRONT_FUNC_SHIFT) | + (R300_ZS_KEEP << R300_RB3D_ZS1_FRONT_FAIL_OP_SHIFT) | + (R300_ZS_KEEP << R300_RB3D_ZS1_FRONT_ZPASS_OP_SHIFT) | + (R300_ZS_KEEP << R300_RB3D_ZS1_FRONT_ZFAIL_OP_SHIFT) | + (R300_ZS_ALWAYS << R300_RB3D_ZS1_BACK_FUNC_SHIFT) | + (R300_ZS_KEEP << R300_RB3D_ZS1_BACK_FAIL_OP_SHIFT) | + (R300_ZS_KEEP << R300_RB3D_ZS1_BACK_ZPASS_OP_SHIFT) | + (R300_ZS_KEEP << R300_RB3D_ZS1_BACK_ZFAIL_OP_SHIFT); + r300->hw.zs.cmd[R300_ZS_CNTL_2] = 0x00ffff00; +#endif + + /* go and compute register values from GL state */ + r300UpdateWindow(ctx); - + r300ColorMask(ctx, ctx->Color.ColorMask[RCOMP], ctx->Color.ColorMask[GCOMP], @@ -1046,25 +2053,35 @@ void r300ResetHwState(r300ContextPtr r300) r300Enable(ctx, GL_DEPTH_TEST, ctx->Depth.Test); r300DepthMask(ctx, ctx->Depth.Mask); r300DepthFunc(ctx, ctx->Depth.Func); + + /* stencil */ + r300Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled); + r300StencilMask(ctx, ctx->Stencil.WriteMask[0]); + r300StencilFunc(ctx, ctx->Stencil.Function[0], ctx->Stencil.Ref[0], ctx->Stencil.ValueMask[0]); + r300StencilOp(ctx, ctx->Stencil.FailFunc[0], ctx->Stencil.ZFailFunc[0], ctx->Stencil.ZPassFunc[0]); r300UpdateCulling(ctx); - - r300_setup_routing(ctx, GL_TRUE); - + r300UpdateTextureState(ctx); + +// r300_setup_routing(ctx, GL_TRUE); + r300EmitArrays(ctx, GL_TRUE); /* Just do the routing */ r300_setup_textures(ctx); r300_setup_rs_unit(ctx); - - + + r300SetupVertexShader(r300); + r300SetupPixelShader(r300); + r300_set_blend_state(ctx); - r300AlphaFunc(ctx, ctx->Color.AlphaFunc, ctx->Color.AlphaRef); -//BEGIN: TODO + r300AlphaFunc(ctx, ctx->Color.AlphaFunc, ctx->Color.AlphaRef); + r300Enable(ctx, GL_ALPHA_TEST, ctx->Color.AlphaEnabled); + + /* Initialize magic registers + TODO : learn what they really do, or get rid of + those we don't have to touch */ 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.vte.cmd[1] = R300_VPORT_X_SCALE_ENA | R300_VPORT_X_OFFSET_ENA | R300_VPORT_Y_SCALE_ENA @@ -1076,10 +2093,13 @@ void r300ResetHwState(r300ContextPtr r300) r300->hw.unk2134.cmd[1] = 0x00FFFFFF; r300->hw.unk2134.cmd[2] = 0x00000000; - +#ifdef MESA_BIG_ENDIAN + r300->hw.unk2140.cmd[1] = 0x00000002; +#else r300->hw.unk2140.cmd[1] = 0x00000000; +#endif - #if 0 /* Done in setup routing */ +#if 0 /* Done in setup routing */ ((drm_r300_cmd_header_t*)r300->hw.vir[0].cmd)->unchecked_state.count = 1; r300->hw.vir[0].cmd[1] = 0x21030003; @@ -1088,8 +2108,8 @@ void r300ResetHwState(r300ContextPtr r300) r300->hw.vic.cmd[R300_VIR_CNTL_0] = 0x00000001; r300->hw.vic.cmd[R300_VIR_CNTL_1] = 0x00000405; - #endif - +#endif + r300->hw.unk21DC.cmd[1] = 0xAAAAAAAA; r300->hw.unk221C.cmd[1] = R300_221C_NORMAL; @@ -1104,15 +2124,16 @@ void r300ResetHwState(r300ContextPtr r300) else r300->hw.unk2288.cmd[1] = R300_2288_RV350; - #if 0 +#if 0 r300->hw.vof.cmd[R300_VOF_CNTL_0] = R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT | R300_VAP_OUTPUT_VTX_FMT_0__COLOR_PRESENT; r300->hw.vof.cmd[R300_VOF_CNTL_1] = 0; /* no textures */ - #endif - + + 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; +#endif r300->hw.gb_enable.cmd[1] = R300_GB_POINT_STUFF_ENABLE | R300_GB_LINE_STUFF_ENABLE @@ -1124,6 +2145,10 @@ void r300ResetHwState(r300ContextPtr r300) r300->hw.gb_misc.cmd[R300_GB_MISC_TILE_CONFIG] = R300_GB_TILE_ENABLE | R300_GB_TILE_PIPE_COUNT_R300 | R300_GB_TILE_SIZE_16; + else if (GET_CHIP(r300->radeon.radeonScreen) == RADEON_CHIP_R420) + r300->hw.gb_misc.cmd[R300_GB_MISC_TILE_CONFIG] = R300_GB_TILE_ENABLE + | R300_GB_TILE_PIPE_COUNT_R420 + | R300_GB_TILE_SIZE_16; else r300->hw.gb_misc.cmd[R300_GB_MISC_TILE_CONFIG] = R300_GB_TILE_ENABLE | R300_GB_TILE_PIPE_COUNT_RV300 @@ -1140,13 +2165,22 @@ void r300ResetHwState(r300ContextPtr r300) r300->hw.unk4214.cmd[1] = 0x00050005; + r300PointSize(ctx, 0.0); +#if 0 r300->hw.ps.cmd[R300_PS_POINTSIZE] = (6 << R300_POINTSIZE_X_SHIFT) | (6 << R300_POINTSIZE_Y_SHIFT); +#endif r300->hw.unk4230.cmd[1] = 0x01800000; r300->hw.unk4230.cmd[2] = 0x00020006; r300->hw.unk4230.cmd[3] = r300PackFloat32(1.0 / 192.0); + r300LineWidth(ctx, 0.0); + +#ifdef EXP_C + static int foobar=0; + r300->hw.lsf.cmd[1] = foobar++; //0x3a088889; +#endif r300->hw.unk4260.cmd[1] = 0; r300->hw.unk4260.cmd[2] = r300PackFloat32(0.0); r300->hw.unk4260.cmd[3] = r300PackFloat32(1.0); @@ -1156,7 +2190,8 @@ void r300ResetHwState(r300ContextPtr r300) r300->hw.unk4274.cmd[3] = 0x00000000; r300->hw.unk4274.cmd[4] = 0x00000000; - r300->hw.unk4288.cmd[1] = 0x00000000; + r300PolygonMode(ctx, GL_FRONT, ctx->Polygon.FrontMode); + r300PolygonMode(ctx, GL_BACK, ctx->Polygon.BackMode); r300->hw.unk4288.cmd[2] = 0x00000001; r300->hw.unk4288.cmd[3] = 0x00000000; r300->hw.unk4288.cmd[4] = 0x00000000; @@ -1164,8 +2199,11 @@ void r300ResetHwState(r300ContextPtr r300) r300->hw.unk42A0.cmd[1] = 0x00000000; - r300->hw.unk42B4.cmd[1] = 0x00000000; - + r300PolygonOffset(ctx, ctx->Polygon.OffsetFactor, ctx->Polygon.OffsetUnits); + r300Enable(ctx, GL_POLYGON_OFFSET_POINT, ctx->Polygon.OffsetPoint); + r300Enable(ctx, GL_POLYGON_OFFSET_LINE, ctx->Polygon.OffsetLine); + r300Enable(ctx, GL_POLYGON_OFFSET_FILL, ctx->Polygon.OffsetFill); + r300->hw.unk42C0.cmd[1] = 0x4B7FFFFF; r300->hw.unk42C0.cmd[2] = 0x00000000; @@ -1175,6 +2213,7 @@ void r300ResetHwState(r300ContextPtr r300) r300->hw.unk43E8.cmd[1] = 0x00FFFFFF; +#if 0 r300->hw.fp.cmd[R300_FP_CNTL0] = 0; r300->hw.fp.cmd[R300_FP_CNTL1] = 0; r300->hw.fp.cmd[R300_FP_CNTL2] = 0; @@ -1182,6 +2221,7 @@ void r300ResetHwState(r300ContextPtr r300) r300->hw.fp.cmd[R300_FP_NODE1] = 0; r300->hw.fp.cmd[R300_FP_NODE2] = 0; r300->hw.fp.cmd[R300_FP_NODE3] = 0; +#endif r300->hw.unk46A4.cmd[1] = 0x00001B01; r300->hw.unk46A4.cmd[2] = 0x00001B0F; @@ -1189,6 +2229,7 @@ void r300ResetHwState(r300ContextPtr r300) r300->hw.unk46A4.cmd[4] = 0x00001B0F; r300->hw.unk46A4.cmd[5] = 0x00000001; +#if 0 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)); @@ -1196,6 +2237,7 @@ void r300ResetHwState(r300ContextPtr r300) 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); } +#endif r300->hw.unk4BC0.cmd[1] = 0; @@ -1203,29 +2245,37 @@ void r300ResetHwState(r300ContextPtr r300) r300->hw.unk4BC8.cmd[2] = 0; r300->hw.unk4BC8.cmd[3] = 0; - #if 0 - r300->hw.at.cmd[R300_AT_ALPHA_TEST] = 0; - #endif + r300->hw.at.cmd[R300_AT_UNKNOWN] = 0; r300->hw.unk4BD8.cmd[1] = 0; r300->hw.unk4E00.cmd[1] = 0; - #if 0 +#if 0 r300->hw.bld.cmd[R300_BLD_CBLEND] = 0; r300->hw.bld.cmd[R300_BLD_ABLEND] = 0; - #endif +#endif - r300->hw.unk4E10.cmd[1] = 0; + r300BlendColor(ctx, ctx->Color.BlendColor); 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; - + + /* Again, r300ClearBuffer uses this */ + if(ctx->Visual.doubleBufferMode){ + 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; + } else { + r300->hw.cb.cmd[R300_CB_OFFSET] = + r300->radeon.radeonScreen->frontOffset + + r300->radeon.radeonScreen->fbLocation; + r300->hw.cb.cmd[R300_CB_PITCH] = r300->radeon.radeonScreen->frontPitch + | R300_COLOR_UNKNOWN_22_23; + + } + r300->hw.unk4E50.cmd[1] = 0; r300->hw.unk4E50.cmd[2] = 0; r300->hw.unk4E50.cmd[3] = 0; @@ -1241,13 +2291,14 @@ void r300ResetHwState(r300ContextPtr r300) r300->hw.unk4EA0.cmd[1] = 0x00000000; r300->hw.unk4EA0.cmd[2] = 0xffffffff; - r300->hw.unk4F08.cmd[1] = 0x00FFFF00; - r300->hw.unk4F10.cmd[1] = 0x00000002; // depthbuffer format? r300->hw.unk4F10.cmd[2] = 0x00000000; r300->hw.unk4F10.cmd[3] = 0x00000003; r300->hw.unk4F10.cmd[4] = 0x00000000; + /* experiment a bit */ + r300->hw.unk4F10.cmd[2] = 0x00000001; // depthbuffer format? + r300->hw.zb.cmd[R300_ZB_OFFSET] = r300->radeon.radeonScreen->depthOffset + r300->radeon.radeonScreen->fbLocation; @@ -1262,6 +2313,7 @@ void r300ResetHwState(r300ContextPtr r300) r300->hw.unk4F54.cmd[1] = 0; +#if 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 */ @@ -1274,14 +2326,15 @@ void r300ResetHwState(r300ContextPtr r300) ((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; +#endif r300->hw.vps.cmd[R300_VPS_ZERO_0] = 0; r300->hw.vps.cmd[R300_VPS_ZERO_1] = 0; r300->hw.vps.cmd[R300_VPS_POINTSIZE] = r300PackFloat32(1.0); r300->hw.vps.cmd[R300_VPS_ZERO_3] = 0; - + //END: TODO - + //verify_r300ResetHwState(r300, 0); r300->hw.all_dirty = GL_TRUE; } @@ -1298,31 +2351,34 @@ void r300InitState(r300ContextPtr r300) GLuint depth_fmt; radeonInitState(&r300->radeon); - + switch (ctx->Visual.depthBits) { case 16: r300->state.depth.scale = 1.0 / (GLfloat) 0xffff; depth_fmt = R200_DEPTH_FORMAT_16BIT_INT_Z; - //r300->state.stencil.clear = 0x00000000; + r300->state.stencil.clear = 0x00000000; break; case 24: r300->state.depth.scale = 1.0 / (GLfloat) 0xffffff; depth_fmt = R200_DEPTH_FORMAT_24BIT_INT_Z; - //r300->state.stencil.clear = 0xff000000; + r300->state.stencil.clear = 0x00ff0000; break; default: fprintf(stderr, "Error: Unsupported depth %d... exiting\n", ctx->Visual.depthBits); exit(-1); } - - memset(&(r300->state.texture), 0, sizeof(r300->state.texture)); + /* Only have hw stencil when depth buffer is 24 bits deep */ + r300->state.stencil.hw_stencil = (ctx->Visual.stencilBits > 0 && + ctx->Visual.depthBits == 24); + + memset(&(r300->state.texture), 0, sizeof(r300->state.texture)); + r300ResetHwState(r300); } - /** * Initialize driver's state callback functions */ @@ -1331,7 +2387,7 @@ void r300InitStateFuncs(struct dd_function_table* functions) radeonInitStateFuncs(functions); functions->UpdateState = r300InvalidateState; - //functions->AlphaFunc = r300AlphaFunc; + functions->AlphaFunc = r300AlphaFunc; functions->BlendColor = r300BlendColor; functions->BlendEquationSeparate = r300BlendEquationSeparate; functions->BlendFuncSeparate = r300BlendFuncSeparate; @@ -1342,9 +2398,18 @@ void r300InitStateFuncs(struct dd_function_table* functions) functions->CullFace = r300CullFace; functions->FrontFace = r300FrontFace; + /* Stencil related */ + functions->ClearStencil = r300ClearStencil; + functions->StencilFunc = r300StencilFunc; + functions->StencilMask = r300StencilMask; + functions->StencilOp = r300StencilOp; + /* Viewport related */ functions->Viewport = r300Viewport; functions->DepthRange = r300DepthRange; functions->PointSize = r300PointSize; -} + functions->LineWidth = r300LineWidth; + functions->PolygonOffset = r300PolygonOffset; + functions->PolygonMode = r300PolygonMode; +}