#include "tnl/t_context.h"
#include "r300_context.h"
-#include "radeon_ioctl.h"
#include "r300_state.h"
#include "r300_emit.h"
#include "r300_ioctl.h"
#define DEBUG_ALL DEBUG_VERTS
-#if defined(USE_X86_ASM)
-#define COPY_DWORDS( dst, src, nr ) \
-do { \
- int __tmp; \
- __asm__ __volatile__( "rep ; movsl" \
- : "=%c" (__tmp), "=D" (dst), "=S" (__tmp) \
- : "0" (nr), \
- "D" ((long)dst), \
- "S" ((long)src) ); \
-} while (0)
-#else
-#define COPY_DWORDS( dst, src, nr ) \
-do { \
- int j; \
- for ( j = 0 ; j < nr ; j++ ) \
- dst[j] = ((int *)src)[j]; \
- dst += nr; \
-} while (0)
-#endif
-
-static void r300EmitVec4(uint32_t *out, GLvoid * data, int stride, int count)
-{
- int i;
-
- if (RADEON_DEBUG & DEBUG_VERTS)
- fprintf(stderr, "%s count %d stride %d out %p data %p\n",
- __FUNCTION__, count, stride, (void *)out, (void *)data);
-
- if (stride == 4)
- COPY_DWORDS(out, data, count);
- else
- for (i = 0; i < count; i++) {
- out[0] = *(int *)data;
- out++;
- data += stride;
- }
-}
-
-static void r300EmitVec8(uint32_t *out, GLvoid * data, int stride, int count)
-{
- int i;
-
- if (RADEON_DEBUG & DEBUG_VERTS)
- fprintf(stderr, "%s count %d stride %d out %p data %p\n",
- __FUNCTION__, count, stride, (void *)out, (void *)data);
-
- if (stride == 8)
- COPY_DWORDS(out, data, count * 2);
- else
- for (i = 0; i < count; i++) {
- out[0] = *(int *)data;
- out[1] = *(int *)(data + 4);
- out += 2;
- data += stride;
- }
-}
-
-static void r300EmitVec12(uint32_t *out, GLvoid * data, int stride, int count)
-{
- int i;
-
- if (RADEON_DEBUG & DEBUG_VERTS)
- fprintf(stderr, "%s count %d stride %d out %p data %p\n",
- __FUNCTION__, count, stride, (void *)out, (void *)data);
-
- if (stride == 12) {
- COPY_DWORDS(out, data, count * 3);
- }
- else
- for (i = 0; i < count; i++) {
- out[0] = *(int *)data;
- out[1] = *(int *)(data + 4);
- out[2] = *(int *)(data + 8);
- out += 3;
- data += stride;
- }
-}
-
-static void r300EmitVec16(uint32_t *out, GLvoid * data, int stride, int count)
-{
- int i;
-
- if (RADEON_DEBUG & DEBUG_VERTS)
- fprintf(stderr, "%s count %d stride %d out %p data %p\n",
- __FUNCTION__, count, stride, (void *)out, (void *)data);
-
- if (stride == 16)
- COPY_DWORDS(out, data, count * 4);
- else
- for (i = 0; i < count; i++) {
- out[0] = *(int *)data;
- out[1] = *(int *)(data + 4);
- out[2] = *(int *)(data + 8);
- out[3] = *(int *)(data + 12);
- out += 4;
- data += stride;
- }
-}
-
-static void r300EmitVec(GLcontext * ctx, struct r300_aos *aos,
- GLvoid * data, int size, int stride, int count)
-{
- r300ContextPtr rmesa = R300_CONTEXT(ctx);
- uint32_t *out;
- uint32_t bo_size;
-
- memset(aos, 0, sizeof(struct r300_aos));
- if (stride == 0) {
- bo_size = size * 4;
- count = 1;
- aos->stride = 0;
- } else {
- bo_size = size * count * 4;
- aos->stride = size;
- }
- aos->bo = radeon_bo_open(rmesa->radeon.radeonScreen->bom,
- 0, bo_size, 32, RADEON_GEM_DOMAIN_GTT, 0);
- aos->offset = 0;
- aos->components = size;
- aos->count = count;
-
- radeon_bo_map(aos->bo, 1);
- out = (uint32_t*)((char*)aos->bo->ptr + aos->offset);
- switch (size) {
- case 1: r300EmitVec4(out, data, stride, count); break;
- case 2: r300EmitVec8(out, data, stride, count); break;
- case 3: r300EmitVec12(out, data, stride, count); break;
- case 4: r300EmitVec16(out, data, stride, count); break;
- default:
- assert(0);
- break;
- }
- radeon_bo_unmap(aos->bo);
-}
-
#define DW_SIZE(x) ((inputs[tab[(x)]] << R300_DST_VEC_LOC_SHIFT) | \
(attribptr[tab[(x)]]->size - 1) << R300_DATA_TYPE_0_SHIFT)
GLuint r300VAPInputCntl1(GLcontext * ctx, GLuint InputsRead)
{
- r300ContextPtr rmesa = R300_CONTEXT(ctx);
GLuint i, vic_1 = 0;
if (InputsRead & (1 << VERT_ATTRIB_POS))
if (InputsRead & (1 << VERT_ATTRIB_COLOR0))
vic_1 |= R300_INPUT_CNTL_COLOR;
- rmesa->state.texture.tc_count = 0;
for (i = 0; i < ctx->Const.MaxTextureUnits; i++)
if (InputsRead & (1 << (VERT_ATTRIB_TEX0 + i))) {
- rmesa->state.texture.tc_count++;
vic_1 |= R300_INPUT_CNTL_TC0 << i;
}
GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint OutputsWritten)
{
- GLuint i, ret = 0;
+ GLuint i, ret = 0, first_free_texcoord = 0;
for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
if (OutputsWritten & (1 << (VERT_RESULT_TEX0 + i))) {
ret |= (4 << (3 * i));
+ ++first_free_texcoord;
}
}
+ if (OutputsWritten & (1 << VERT_RESULT_FOGC)) {
+ if (first_free_texcoord > 8) {
+ fprintf(stderr, "\tout of free texcoords to write fog coord\n");
+ _mesa_exit(-1);
+ }
+ ret |= 1 << (3 * first_free_texcoord);
+ }
+
return ret;
}
struct r300_vertex_program *prog =
(struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx);
- if (hw_tcl_on) {
+ if (rmesa->options.hw_tcl_enabled) {
inputs = prog->inputs;
InputsRead = prog->key.InputsRead;
OutputsWritten = prog->key.OutputsWritten;
} else {
- inputs = rmesa->state.sw_tcl_inputs;
+ inputs = rmesa->swtcl.sw_tcl_inputs;
DECLARE_RENDERINPUTS(render_inputs_bitset);
RENDERINPUTS_COPY(render_inputs_bitset, tnl->render_inputs_bitset);
if (InputsRead & (1 << i))
inputs[i] = 6 + (i - VERT_ATTRIB_TEX0);
- RENDERINPUTS_COPY(rmesa->state.render_inputs_bitset, render_inputs_bitset);
+ RENDERINPUTS_COPY(rmesa->render_inputs_bitset, render_inputs_bitset);
}
assert(InputsRead);
for (ci = 0; ci < vb->AttribPtr[tab[i]]->size; ci++) {
swizzle[i][ci] = ci;
}
- r300EmitVec(ctx, &rmesa->state.aos[i],
- vb->AttribPtr[tab[i]]->data,
- vb->AttribPtr[tab[i]]->size,
- vb->AttribPtr[tab[i]]->stride, count);
+ rcommon_emit_vector(ctx, &rmesa->radeon.tcl.aos[i],
+ vb->AttribPtr[tab[i]]->data,
+ vb->AttribPtr[tab[i]]->size,
+ vb->AttribPtr[tab[i]]->stride, count);
}
/* Setup INPUT_ROUTE. */
- if (rmesa->radeon.radeonScreen->kernel_mm) {
- R300_STATECHANGE(rmesa, vir[0]);
- rmesa->hw.vir[0].cmd[0] &= 0xC000FFFF;
- rmesa->hw.vir[1].cmd[0] &= 0xC000FFFF;
- rmesa->hw.vir[0].cmd[0] |=
- (r300VAPInputRoute0(&rmesa->hw.vir[0].cmd[R300_VIR_CNTL_0],
- vb->AttribPtr, inputs, tab, nr) & 0x3FFF) << 16;
- R300_STATECHANGE(rmesa, vir[1]);
- rmesa->hw.vir[1].cmd[0] |=
- (r300VAPInputRoute1(&rmesa->hw.vir[1].cmd[R300_VIR_CNTL_0], swizzle,
- nr) & 0x3FFF) << 16;
- } else {
- R300_STATECHANGE(rmesa, vir[0]);
- ((drm_r300_cmd_header_t *) rmesa->hw.vir[0].cmd)->packet0.count =
- r300VAPInputRoute0(&rmesa->hw.vir[0].cmd[R300_VIR_CNTL_0],
- vb->AttribPtr, inputs, tab, nr);
- R300_STATECHANGE(rmesa, vir[1]);
- ((drm_r300_cmd_header_t *) rmesa->hw.vir[1].cmd)->packet0.count =
- r300VAPInputRoute1(&rmesa->hw.vir[1].cmd[R300_VIR_CNTL_0], swizzle,
- nr);
- }
+ if (rmesa->radeon.radeonScreen->kernel_mm) {
+ R300_STATECHANGE(rmesa, vir[0]);
+ rmesa->hw.vir[0].cmd[0] &= 0xC000FFFF;
+ rmesa->hw.vir[1].cmd[0] &= 0xC000FFFF;
+ rmesa->hw.vir[0].cmd[0] |=
+ (r300VAPInputRoute0(&rmesa->hw.vir[0].cmd[R300_VIR_CNTL_0],
+ vb->AttribPtr, inputs, tab, nr) & 0x3FFF) << 16;
+ R300_STATECHANGE(rmesa, vir[1]);
+ rmesa->hw.vir[1].cmd[0] |=
+ (r300VAPInputRoute1(&rmesa->hw.vir[1].cmd[R300_VIR_CNTL_0], swizzle,
+ nr) & 0x3FFF) << 16;
+ } else {
+ R300_STATECHANGE(rmesa, vir[0]);
+ ((drm_r300_cmd_header_t *) rmesa->hw.vir[0].cmd)->packet0.count =
+ r300VAPInputRoute0(&rmesa->hw.vir[0].cmd[R300_VIR_CNTL_0],
+ vb->AttribPtr, inputs, tab, nr);
+ R300_STATECHANGE(rmesa, vir[1]);
+ ((drm_r300_cmd_header_t *) rmesa->hw.vir[1].cmd)->packet0.count =
+ r300VAPInputRoute1(&rmesa->hw.vir[1].cmd[R300_VIR_CNTL_0], swizzle,
+ nr);
+ }
/* Setup INPUT_CNTL. */
R300_STATECHANGE(rmesa, vic);
rmesa->hw.vof.cmd[R300_VOF_CNTL_1] =
r300VAPOutputCntl1(ctx, OutputsWritten);
- rmesa->state.aos_count = nr;
+ rmesa->radeon.tcl.aos_count = nr;
return R300_FALLBACK_NONE;
}
-void r300ReleaseArrays(GLcontext * ctx)
-{
- r300ContextPtr rmesa = R300_CONTEXT(ctx);
- int i;
-
- if (rmesa->state.elt_dma_bo) {
- radeon_bo_unref(rmesa->state.elt_dma_bo);
- rmesa->state.elt_dma_bo = 0;
- }
- for (i = 0; i < rmesa->state.aos_count; i++) {
- if (rmesa->state.aos[i].bo) {
- rmesa->state.aos[i].bo = radeon_bo_unref(rmesa->state.aos[i].bo);
- }
- }
-}
-
void r300EmitCacheFlush(r300ContextPtr rmesa)
{
BATCH_LOCALS(&rmesa->radeon);
- BEGIN_BATCH(4);
+ BEGIN_BATCH_NO_AUTOSTATE(4);
OUT_BATCH_REGVAL(R300_RB3D_DSTCACHE_CTLSTAT,
R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_FREE_3D_TAGS |
R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D);
projects / mesa.git / blobdiff