diff options
-rw-r--r-- | src/mesa/drivers/dri/r300/r300_context.h | 23 | ||||
-rw-r--r-- | src/mesa/drivers/dri/r300/r300_emit.c | 151 | ||||
-rw-r--r-- | src/mesa/drivers/dri/r300/r300_render.c | 72 |
3 files changed, 56 insertions, 190 deletions
diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h index 13b943f75fc..dbcd5d04d6d 100644 --- a/src/mesa/drivers/dri/r300/r300_context.h +++ b/src/mesa/drivers/dri/r300/r300_context.h @@ -790,28 +790,6 @@ struct r300_fragment_program { #define REG_COLOR0 1 #define REG_TEX0 2 -struct dt { - GLint size; - GLenum type; - GLsizei stride; - void *data; -}; - -struct radeon_vertex_buffer { - int Count; - void *Elts; - int elt_size; - int elt_min, elt_max; /* debug */ - - struct dt AttribPtr[VERT_ATTRIB_MAX]; - - const struct _mesa_prim *Primitive; - GLuint PrimitiveCount; - GLint LockFirst; - GLsizei LockCount; - int lock_uptodate; -}; - struct r300_state { struct r300_depthbuffer_state depth; struct r300_texture_state texture; @@ -820,7 +798,6 @@ struct r300_state { struct r300_pfs_compile_state pfs_compile; struct r300_dma_region aos[R300_MAX_AOS_ARRAYS]; int aos_count; - struct radeon_vertex_buffer VB; GLuint *Elts; struct r300_dma_region elt_dma; diff --git a/src/mesa/drivers/dri/r300/r300_emit.c b/src/mesa/drivers/dri/r300/r300_emit.c index 2c26069f9b3..9fb712f7b8f 100644 --- a/src/mesa/drivers/dri/r300/r300_emit.c +++ b/src/mesa/drivers/dri/r300/r300_emit.c @@ -223,86 +223,48 @@ static void r300EmitVec(GLcontext * ctx, } -static GLuint t_type(struct dt *dt) +#define R300_VIR0_AOS_SIZE_SHIFT 0 +#define R300_VIR0_AOS_INPUT_SHIFT 8 +#define R300_VIR0_AOS_STOP_SHIFT 13 +#define R300_VIR0_AOS_TYPE_SHIFT 14 +#define R300_VIR0_HIGH_SHIFT 16 + +// Pack 4 elemets in a 16 bit (aos_size first 8, input next 5, 1 stop bit(Whild gues), aos_type last 2); +static inline GLuint t_vir_pack(GLvector4f ** dt, int *inputs, int i) { - switch (dt->type) { - case GL_UNSIGNED_BYTE: - return AOS_FORMAT_UBYTE; - case GL_SHORT: - return AOS_FORMAT_USHORT; - case GL_FLOAT: - return AOS_FORMAT_FLOAT; - default: - assert(0); - break; - } - - return AOS_FORMAT_FLOAT; -} - -static GLuint t_vir0_size(struct dt *dt) -{ - switch (dt->type) { - case GL_UNSIGNED_BYTE: - return 4; - case GL_SHORT: - return 7; - case GL_FLOAT: - return dt->size - 1; - default: - assert(0); - break; - } - - return 0; + GLuint dw; + dw = (dt[i]->size - 1) << R300_VIR0_AOS_SIZE_SHIFT; + dw |= inputs[i] << R300_VIR0_AOS_INPUT_SHIFT; + //dw |= t_type(&dt[i]) << R300_VIR0_AOS_TYPE_SHIFT; + return dw; } -static GLuint t_aos_size(struct dt *dt) -{ - switch (dt->type) { - case GL_UNSIGNED_BYTE: - return 1; - case GL_SHORT: - return 2; - case GL_FLOAT: - return dt->size; - default: - assert(0); - break; - } - - return 0; -} - -static GLuint t_vir0(uint32_t * dst, struct dt *dt, int *inputs, +static GLuint t_vir0(uint32_t * dst, GLvector4f ** dt, int *inputs, GLint * tab, GLuint nr) { - GLuint i, dw; + GLuint i, dw, dwInternel; for (i = 0; i + 1 < nr; i += 2) { - dw = t_vir0_size(&dt[tab[i]]) | (inputs[tab[i]] << 8) | - (t_type(&dt[tab[i]]) << 14); - dw |= - (t_vir0_size(&dt[tab[i + 1]]) | - (inputs[tab[i + 1]] << 8) | (t_type(&dt[tab[i + 1]]) - << 14)) << 16; + dw = t_vir_pack(dt, inputs, tab[i]); + dwInternel = t_vir_pack(dt, inputs, tab[i + 1]); + dw |= dwInternel << R300_VIR0_HIGH_SHIFT; if (i + 2 == nr) { - dw |= (1 << (13 + 16)); + dw |= + (1 << + (R300_VIR0_AOS_STOP_SHIFT + R300_VIR0_HIGH_SHIFT)); } - dst[i >> 1] = dw; + dst[i >> 1] = dw; // Is the same as i/2 } if (nr & 1) { - dw = t_vir0_size(&dt[tab[nr - 1]]) | (inputs[tab[nr - 1]] - << 8) | - (t_type(&dt[tab[nr - 1]]) << 14); - dw |= 1 << 13; + dw = t_vir_pack(dt, inputs, tab[nr - 1]); + dw |= 1 << R300_VIR0_AOS_STOP_SHIFT; dst[nr >> 1] = dw; } - return (nr + 1) >> 1; + return (nr + 1) >> 1; // Is the same as (nr+1)/2 } static GLuint t_swizzle(int swizzle[4]) @@ -331,11 +293,6 @@ static GLuint t_vir1(uint32_t * dst, int swizzle[][4], GLuint nr) return (nr + 1) >> 1; } -static GLuint t_emit_size(struct dt *dt) -{ - return dt->size; -} - static GLuint t_vic(GLcontext * ctx, GLuint InputsRead) { r300ContextPtr r300 = R300_CONTEXT(ctx); @@ -369,9 +326,10 @@ int r300EmitArrays(GLcontext * ctx) { r300ContextPtr rmesa = R300_CONTEXT(ctx); r300ContextPtr r300 = rmesa; - struct radeon_vertex_buffer *VB = &rmesa->state.VB; + TNLcontext *tnl = TNL_CONTEXT(ctx); + struct vertex_buffer *vb = &tnl->vb; GLuint nr; - GLuint count = VB->Count; + GLuint count = vb->Count; GLuint i; GLuint InputsRead = 0, OutputsWritten = 0; int *inputs = NULL; @@ -466,57 +424,38 @@ int r300EmitArrays(GLcontext * ctx) swizzle[i][2] = SWIZZLE_ZERO; swizzle[i][3] = SWIZZLE_ONE; - for (ci = 0; ci < VB->AttribPtr[tab[i]].size; ci++) + for (ci = 0; ci < vb->AttribPtr[tab[i]]->size; ci++) swizzle[i][ci] = ci; -#if MESA_BIG_ENDIAN -#define SWAP_INT(a, b) do { \ - int __temp; \ - __temp = a;\ - a = b; \ - b = __temp; \ -} while (0) - - if (VB->AttribPtr[tab[i]].type == GL_UNSIGNED_BYTE) { - SWAP_INT(swizzle[i][0], swizzle[i][3]); - SWAP_INT(swizzle[i][1], swizzle[i][2]); - } -#endif /* MESA_BIG_ENDIAN */ - - if (r300IsGartMemory(rmesa, VB->AttribPtr[tab[i]].data, + if (r300IsGartMemory(rmesa, vb->AttribPtr[tab[i]]->data, /*(count-1)*stride */ 4)) { - if (VB->AttribPtr[tab[i]].stride % 4) + if (vb->AttribPtr[tab[i]]->stride % 4) return R300_FALLBACK_TCL; rmesa->state.aos[i].address = - VB->AttribPtr[tab[i]].data; + (void *)(vb->AttribPtr[tab[i]]->data); rmesa->state.aos[i].start = 0; rmesa->state.aos[i].aos_offset = r300GartOffsetFromVirtual(rmesa, - VB-> - AttribPtr[tab[i]].data); + vb-> + AttribPtr[tab[i]]->data); rmesa->state.aos[i].aos_stride = - VB->AttribPtr[tab[i]].stride / 4; + vb->AttribPtr[tab[i]]->stride / 4; rmesa->state.aos[i].aos_size = - t_emit_size(&VB->AttribPtr[tab[i]]); + vb->AttribPtr[tab[i]]->size; } else { - /* TODO: r300EmitVec can only handle 4 byte vectors */ - if (VB->AttribPtr[tab[i]].type != GL_FLOAT) - return R300_FALLBACK_TCL; - r300EmitVec(ctx, &rmesa->state.aos[i], - VB->AttribPtr[tab[i]].data, - t_emit_size(&VB->AttribPtr[tab[i]]), - VB->AttribPtr[tab[i]].stride, count); + vb->AttribPtr[tab[i]]->data, + vb->AttribPtr[tab[i]]->size, + vb->AttribPtr[tab[i]]->stride, count); } - rmesa->state.aos[i].aos_size = - t_aos_size(&VB->AttribPtr[tab[i]]); + rmesa->state.aos[i].aos_size = vb->AttribPtr[tab[i]]->size; - comp_size = _mesa_sizeof_type(VB->AttribPtr[tab[i]].type); + comp_size = _mesa_sizeof_type(GL_FLOAT); - for (fix = 0; fix <= 4 - VB->AttribPtr[tab[i]].size; fix++) { + for (fix = 0; fix <= 4 - vb->AttribPtr[tab[i]]->size; fix++) { if ((rmesa->state.aos[i].aos_offset - comp_size * fix) % 4) continue; @@ -532,14 +471,14 @@ int r300EmitArrays(GLcontext * ctx) rmesa->state.aos[i].aos_offset -= comp_size * fix; - for (ci = 0; ci < VB->AttribPtr[tab[i]].size; ci++) + for (ci = 0; ci < vb->AttribPtr[tab[i]]->size; ci++) swizzle[i][ci] += fix; } else { WARN_ONCE ("Cannot handle offset %x with stride %d, comp %d\n", rmesa->state.aos[i].aos_offset, rmesa->state.aos[i].aos_stride, - VB->AttribPtr[tab[i]].size); + vb->AttribPtr[tab[i]]->size); return R300_FALLBACK_TCL; } } @@ -547,7 +486,7 @@ int r300EmitArrays(GLcontext * ctx) /* setup INPUT_ROUTE */ R300_STATECHANGE(r300, vir[0]); ((drm_r300_cmd_header_t *) r300->hw.vir[0].cmd)->packet0.count = - t_vir0(&r300->hw.vir[0].cmd[R300_VIR_CNTL_0], VB->AttribPtr, + t_vir0(&r300->hw.vir[0].cmd[R300_VIR_CNTL_0], vb->AttribPtr, inputs, tab, nr); R300_STATECHANGE(r300, vir[1]); diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c index cc13e9a5304..6cec2bb1ea7 100644 --- a/src/mesa/drivers/dri/r300/r300_render.c +++ b/src/mesa/drivers/dri/r300/r300_render.c @@ -301,6 +301,8 @@ static void r300RunRenderPrimitive(r300ContextPtr rmesa, GLcontext * ctx, int start, int end, int prim) { int type, num_verts; + TNLcontext *tnl = TNL_CONTEXT(ctx); + struct vertex_buffer *vb = &tnl->vb; type = r300PrimitiveType(rmesa, ctx, prim); num_verts = r300NumVerts(rmesa, end - start, prim); @@ -308,88 +310,36 @@ static void r300RunRenderPrimitive(r300ContextPtr rmesa, GLcontext * ctx, if (type < 0 || num_verts <= 0) return; - if (rmesa->state.VB.Elts) { + if (vb->Elts) { r300EmitAOS(rmesa, rmesa->state.aos_count, start); if (num_verts > 65535) { /* not implemented yet */ WARN_ONCE("Too many elts\n"); return; } - r300EmitElts(ctx, rmesa->state.VB.Elts, num_verts, - rmesa->state.VB.elt_size); + r300EmitElts(ctx, vb->Elts, num_verts, 4); r300FireEB(rmesa, rmesa->state.elt_dma.aos_offset, - num_verts, type, rmesa->state.VB.elt_size); + num_verts, type, 4); } else { r300EmitAOS(rmesa, rmesa->state.aos_count, start); r300FireAOS(rmesa, num_verts, type); } } -#define CONV_VB(a, b) rvb->AttribPtr[(a)].size = vb->b->size, \ - rvb->AttribPtr[(a)].type = GL_FLOAT, \ - rvb->AttribPtr[(a)].stride = vb->b->stride, \ - rvb->AttribPtr[(a)].data = vb->b->data - -static void radeon_vb_to_rvb(r300ContextPtr rmesa, - struct radeon_vertex_buffer *rvb, - struct vertex_buffer *vb) -{ - int i; - GLcontext *ctx; - ctx = rmesa->radeon.glCtx; - - memset(rvb, 0, sizeof(*rvb)); - - rvb->Elts = vb->Elts; - rvb->elt_size = 4; - rvb->elt_min = 0; - rvb->elt_max = vb->Count; - - rvb->Count = vb->Count; - - if (hw_tcl_on) { - CONV_VB(VERT_ATTRIB_POS, ObjPtr); - } else { - assert(vb->ClipPtr); - CONV_VB(VERT_ATTRIB_POS, ClipPtr); - } - - CONV_VB(VERT_ATTRIB_NORMAL, NormalPtr); - CONV_VB(VERT_ATTRIB_COLOR0, ColorPtr[0]); - CONV_VB(VERT_ATTRIB_COLOR1, SecondaryColorPtr[0]); - CONV_VB(VERT_ATTRIB_FOG, FogCoordPtr); - - for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) - CONV_VB(VERT_ATTRIB_TEX0 + i, TexCoordPtr[i]); - - for (i = 0; i < MAX_VERTEX_PROGRAM_ATTRIBS; i++) - CONV_VB(VERT_ATTRIB_GENERIC0 + i, - AttribPtr[VERT_ATTRIB_GENERIC0 + i]); - - rvb->Primitive = vb->Primitive; - rvb->PrimitiveCount = vb->PrimitiveCount; - rvb->LockFirst = rvb->LockCount = 0; - rvb->lock_uptodate = GL_FALSE; -} - static GLboolean r300RunRender(GLcontext * ctx, struct tnl_pipeline_stage *stage) { r300ContextPtr rmesa = R300_CONTEXT(ctx); - struct radeon_vertex_buffer *VB = &rmesa->state.VB; int i; int cmd_reserved = 0; int cmd_written = 0; drm_radeon_cmd_header_t *cmd = NULL; + TNLcontext *tnl = TNL_CONTEXT(ctx); + struct vertex_buffer *vb = &tnl->vb; if (RADEON_DEBUG & DEBUG_PRIMS) fprintf(stderr, "%s\n", __FUNCTION__); - if (stage) { - TNLcontext *tnl = TNL_CONTEXT(ctx); - radeon_vb_to_rvb(rmesa, VB, &tnl->vb); - } - r300UpdateShaders(rmesa); if (r300EmitArrays(ctx)) return GL_TRUE; @@ -404,10 +354,10 @@ static GLboolean r300RunRender(GLcontext * ctx, r300EmitState(rmesa); - for (i = 0; i < VB->PrimitiveCount; i++) { - GLuint prim = _tnl_translate_prim(&VB->Primitive[i]); - GLuint start = VB->Primitive[i].start; - GLuint end = VB->Primitive[i].start + VB->Primitive[i].count; + for (i = 0; i < vb->PrimitiveCount; i++) { + GLuint prim = _tnl_translate_prim(&vb->Primitive[i]); + GLuint start = vb->Primitive[i].start; + GLuint end = vb->Primitive[i].start + vb->Primitive[i].count; r300RunRenderPrimitive(rmesa, ctx, start, end, prim); } |