* \author Keith Whitwell <keith@tungstengraphics.com>
*/
-#include "glheader.h"
-#include "mtypes.h"
-#include "colormac.h"
-#include "imports.h"
-#include "macros.h"
-#include "image.h"
+#include "main/glheader.h"
+#include "main/mtypes.h"
+#include "main/colormac.h"
+#include "main/imports.h"
+#include "main/macros.h"
+#include "main/image.h"
#include "swrast_setup/swrast_setup.h"
#include "math/m_translate.h"
#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"
-#ifdef USER_BUFFERS
-#include "r300_mem.h"
-#endif
#if SWIZZLE_X != R300_INPUT_ROUTE_SELECT_X || \
SWIZZLE_Y != R300_INPUT_ROUTE_SELECT_Y || \
#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(GLcontext * ctx, struct r300_dma_region *rvb,
- GLvoid * data, int stride, int count)
-{
- int i;
- int *out = (int *)(rvb->address + rvb->start);
-
- 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(GLcontext * ctx, struct r300_dma_region *rvb,
- GLvoid * data, int stride, int count)
-{
- int i;
- int *out = (int *)(rvb->address + rvb->start);
-
- 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;
- }
-}
+#define DW_SIZE(x) ((inputs[tab[(x)]] << R300_DST_VEC_LOC_SHIFT) | \
+ (attribptr[tab[(x)]]->size - 1) << R300_DATA_TYPE_0_SHIFT)
-static void r300EmitVec12(GLcontext * ctx, struct r300_dma_region *rvb,
- GLvoid * data, int stride, int count)
-{
- int i;
- int *out = (int *)(rvb->address + rvb->start);
-
- 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(GLcontext * ctx, struct r300_dma_region *rvb,
- GLvoid * data, int stride, int count)
-{
- int i;
- int *out = (int *)(rvb->address + rvb->start);
-
- 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_dma_region *rvb,
- GLvoid * data, int size, int stride, int count)
-{
- r300ContextPtr rmesa = R300_CONTEXT(ctx);
-
- if (stride == 0) {
- r300AllocDmaRegion(rmesa, rvb, size * 4, 4);
- count = 1;
- rvb->aos_offset = GET_START(rvb);
- rvb->aos_stride = 0;
- } else {
- r300AllocDmaRegion(rmesa, rvb, size * count * 4, 4);
- rvb->aos_offset = GET_START(rvb);
- rvb->aos_stride = size;
- }
-
- switch (size) {
- case 1:
- r300EmitVec4(ctx, rvb, data, stride, count);
- break;
- case 2:
- r300EmitVec8(ctx, rvb, data, stride, count);
- break;
- case 3:
- r300EmitVec12(ctx, rvb, data, stride, count);
- break;
- case 4:
- r300EmitVec16(ctx, rvb, data, stride, count);
- break;
- default:
- assert(0);
- break;
- }
-}
-
-static GLuint r300VAPInputRoute0(uint32_t * dst, GLvector4f ** attribptr,
+GLuint r300VAPInputRoute0(uint32_t * dst, GLvector4f ** attribptr,
int *inputs, GLint * tab, GLuint nr)
{
GLuint i, dw;
/* type, inputs, stop bit, size */
- for (i = 0; i + 1 < nr; i += 2) {
- dw = R300_INPUT_ROUTE_FLOAT | (inputs[tab[i]] << 8) | (attribptr[tab[i]]->size - 1);
- dw |= (R300_INPUT_ROUTE_FLOAT | (inputs[tab[i + 1]] << 8) | (attribptr[tab[i + 1]]->size - 1)) << 16;
- if (i + 2 == nr) {
- dw |= (1 << (13 + 16));
+ for (i = 0; i < nr; i += 2) {
+ /* make sure input is valid, would lockup the gpu */
+ assert(inputs[tab[i]] != -1);
+ dw = (R300_SIGNED | DW_SIZE(i));
+ if (i + 1 == nr) {
+ dw |= R300_LAST_VEC << R300_DATA_TYPE_0_SHIFT;
+ } else {
+ assert(inputs[tab[i + 1]] != -1);
+ dw |= (R300_SIGNED |
+ DW_SIZE(i + 1)) << R300_DATA_TYPE_1_SHIFT;
+ if (i + 2 == nr) {
+ dw |= R300_LAST_VEC << R300_DATA_TYPE_1_SHIFT;
+ }
}
dst[i >> 1] = dw;
}
- if (nr & 1) {
- dw = R300_INPUT_ROUTE_FLOAT | (inputs[tab[nr - 1]] << 8) | (attribptr[tab[nr - 1]]->size - 1);
- dw |= 1 << 13;
- dst[nr >> 1] = dw;
- }
-
return (nr + 1) >> 1;
}
static GLuint r300VAPInputRoute1Swizzle(int swizzle[4])
{
- return (swizzle[0] << R300_INPUT_ROUTE_X_SHIFT) |
- (swizzle[1] << R300_INPUT_ROUTE_Y_SHIFT) |
- (swizzle[2] << R300_INPUT_ROUTE_Z_SHIFT) |
- (swizzle[3] << R300_INPUT_ROUTE_W_SHIFT);
+ return (swizzle[0] << R300_SWIZZLE_SELECT_X_SHIFT) |
+ (swizzle[1] << R300_SWIZZLE_SELECT_Y_SHIFT) |
+ (swizzle[2] << R300_SWIZZLE_SELECT_Z_SHIFT) |
+ (swizzle[3] << R300_SWIZZLE_SELECT_W_SHIFT);
}
-static GLuint r300VAPInputRoute1(uint32_t * dst, int swizzle[][4], GLuint nr)
+GLuint r300VAPInputRoute1(uint32_t * dst, int swizzle[][4], GLuint nr)
{
- GLuint i;
-
- for (i = 0; i + 1 < nr; i += 2) {
- dst[i >> 1] = r300VAPInputRoute1Swizzle(swizzle[i]) | R300_INPUT_ROUTE_ENABLE;
- dst[i >> 1] |= (r300VAPInputRoute1Swizzle(swizzle[i + 1]) | R300_INPUT_ROUTE_ENABLE) << 16;
- }
+ GLuint i, dw;
- if (nr & 1) {
- dst[nr >> 1] = r300VAPInputRoute1Swizzle(swizzle[nr - 1]) | R300_INPUT_ROUTE_ENABLE;
+ for (i = 0; i < nr; i += 2) {
+ dw = (r300VAPInputRoute1Swizzle(swizzle[i]) |
+ ((R300_WRITE_ENA_X | R300_WRITE_ENA_Y |
+ R300_WRITE_ENA_Z | R300_WRITE_ENA_W) << R300_WRITE_ENA_SHIFT)) << R300_SWIZZLE0_SHIFT;
+ if (i + 1 < nr) {
+ dw |= (r300VAPInputRoute1Swizzle(swizzle[i + 1]) |
+ ((R300_WRITE_ENA_X | R300_WRITE_ENA_Y |
+ R300_WRITE_ENA_Z | R300_WRITE_ENA_W) << R300_WRITE_ENA_SHIFT)) << R300_SWIZZLE1_SHIFT;
+ }
+ dst[i >> 1] = dw;
}
return (nr + 1) >> 1;
}
-static GLuint r300VAPInputCntl0(GLcontext * ctx, GLuint InputsRead)
+GLuint r300VAPInputCntl0(GLcontext * ctx, GLuint InputsRead)
{
/* No idea what this value means. I have seen other values written to
* this register... */
return 0x5555;
}
-static GLuint r300VAPInputCntl1(GLcontext * ctx, GLuint InputsRead)
+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;
}
return vic_1;
}
-static GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint OutputsWritten)
+GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint OutputsWritten)
{
GLuint ret = 0;
ret |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT;
if (OutputsWritten & (1 << VERT_RESULT_COL0))
- ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_PRESENT;
+ ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT;
if (OutputsWritten & (1 << VERT_RESULT_COL1))
ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT;
-#if 0
- if (OutputsWritten & (1 << VERT_RESULT_BFC0))
- ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT;
-
- if (OutputsWritten & (1 << VERT_RESULT_BFC1))
- ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT;
-
- if (OutputsWritten & (1 << VERT_RESULT_FOGC)) ;
-#endif
+ if (OutputsWritten & (1 << VERT_RESULT_BFC0)
+ || OutputsWritten & (1 << VERT_RESULT_BFC1))
+ ret |=
+ R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT |
+ R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT |
+ R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT;
if (OutputsWritten & (1 << VERT_RESULT_PSIZ))
ret |= R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT;
return ret;
}
-static GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint OutputsWritten)
+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);
+ vb->AttribPtr[VERT_ATTRIB_POS] = vb->ClipPtr;
+
assert(RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_POS));
assert(RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_NORMAL) == 0);
- assert(RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_COLOR0));
if (RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_POS)) {
InputsRead |= 1 << VERT_ATTRIB_POS;
}
}
- if (!(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
- /* Fixed, apply to vir0 only */
- memcpy(vir_inputs, inputs, VERT_ATTRIB_MAX * sizeof(int));
- inputs = vir_inputs;
- if (InputsRead & VERT_ATTRIB_POS)
- inputs[VERT_ATTRIB_POS] = 0;
- if (InputsRead & (1 << VERT_ATTRIB_COLOR0))
- inputs[VERT_ATTRIB_COLOR0] = 2;
- if (InputsRead & (1 << VERT_ATTRIB_COLOR1))
- inputs[VERT_ATTRIB_COLOR1] = 3;
- for (i = VERT_ATTRIB_TEX0; i <= VERT_ATTRIB_TEX7; i++)
- if (InputsRead & (1 << i))
- inputs[i] = 6 + (i - VERT_ATTRIB_TEX0);
- }
-
- RENDERINPUTS_COPY(rmesa->state.render_inputs_bitset, render_inputs_bitset);
+ /* Fixed, apply to vir0 only */
+ memcpy(vir_inputs, inputs, VERT_ATTRIB_MAX * sizeof(int));
+ inputs = vir_inputs;
+ if (InputsRead & VERT_ATTRIB_POS)
+ inputs[VERT_ATTRIB_POS] = 0;
+ if (InputsRead & (1 << VERT_ATTRIB_COLOR0))
+ inputs[VERT_ATTRIB_COLOR0] = 2;
+ if (InputsRead & (1 << VERT_ATTRIB_COLOR1))
+ inputs[VERT_ATTRIB_COLOR1] = 3;
+ for (i = VERT_ATTRIB_TEX0; i <= VERT_ATTRIB_TEX7; i++)
+ if (InputsRead & (1 << i))
+ inputs[i] = 6 + (i - VERT_ATTRIB_TEX0);
+
+ RENDERINPUTS_COPY(rmesa->render_inputs_bitset, render_inputs_bitset);
}
assert(InputsRead);
}
for (i = 0; i < nr; i++) {
- int ci, fix, found = 0;
+ int ci;
swizzle[i][0] = SWIZZLE_ZERO;
swizzle[i][1] = SWIZZLE_ZERO;
for (ci = 0; ci < vb->AttribPtr[tab[i]]->size; ci++) {
swizzle[i][ci] = ci;
}
-
- if (r300IsGartMemory(rmesa, vb->AttribPtr[tab[i]]->data, 4)) {
- if (vb->AttribPtr[tab[i]]->stride % 4) {
- return R300_FALLBACK_TCL;
- }
- rmesa->state.aos[i].address = (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);
- rmesa->state.aos[i].aos_stride = vb->AttribPtr[tab[i]]->stride / 4;
- rmesa->state.aos[i].aos_size = vb->AttribPtr[tab[i]]->size;
- } else {
- r300EmitVec(ctx, &rmesa->state.aos[i],
+ 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);
- }
-
- rmesa->state.aos[i].aos_size = vb->AttribPtr[tab[i]]->size;
-
- for (fix = 0; fix <= 4 - vb->AttribPtr[tab[i]]->size; fix++) {
- if ((rmesa->state.aos[i].aos_offset - _mesa_sizeof_type(GL_FLOAT) * fix) % 4) {
- continue;
- }
- found = 1;
- break;
- }
-
- if (found) {
- if (fix > 0) {
- WARN_ONCE("Feeling lucky?\n");
- }
- rmesa->state.aos[i].aos_offset -= _mesa_sizeof_type(GL_FLOAT) * fix;
- 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);
- return R300_FALLBACK_TCL;
- }
}
/* Setup INPUT_ROUTE. */
- 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;
}
-#ifdef USER_BUFFERS
-void r300UseArrays(GLcontext * ctx)
-{
- r300ContextPtr rmesa = R300_CONTEXT(ctx);
- int i;
-
- if (rmesa->state.elt_dma.buf)
- r300_mem_use(rmesa, rmesa->state.elt_dma.buf->id);
-
- for (i = 0; i < rmesa->state.aos_count; i++) {
- if (rmesa->state.aos[i].buf)
- r300_mem_use(rmesa, rmesa->state.aos[i].buf->id);
- }
-}
-#endif
-
-void r300ReleaseArrays(GLcontext * ctx)
+void r300EmitCacheFlush(r300ContextPtr rmesa)
{
- r300ContextPtr rmesa = R300_CONTEXT(ctx);
- int i;
-
- r300ReleaseDmaRegion(rmesa, &rmesa->state.elt_dma, __FUNCTION__);
- for (i = 0; i < rmesa->state.aos_count; i++) {
- r300ReleaseDmaRegion(rmesa, &rmesa->state.aos[i], __FUNCTION__);
- }
+ BATCH_LOCALS(&rmesa->radeon);
+
+ 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);
+ OUT_BATCH_REGVAL(R300_ZB_ZCACHE_CTLSTAT,
+ R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE |
+ R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE);
+ END_BATCH();
+ COMMIT_BATCH();
}