* Authors:
* Jerome Glisse
*/
-#include "r600_hw_context_priv.h"
+#include "r600_pipe.h"
#include "r600d.h"
#include "util/u_memory.h"
#include <errno.h>
return;
}
-static void r600_init_block(struct r600_context *ctx,
- struct r600_block *block,
- const struct r600_reg *reg, int index, int nreg,
- unsigned opcode, unsigned offset_base)
-{
- int i = index;
- int j, n = nreg;
-
- /* initialize block */
- block->flags = 0;
- block->status |= R600_BLOCK_STATUS_DIRTY; /* dirty all blocks at start */
- block->start_offset = reg[i].offset;
- block->pm4[block->pm4_ndwords++] = PKT3(opcode, n, 0);
- block->pm4[block->pm4_ndwords++] = (block->start_offset - offset_base) >> 2;
- block->reg = &block->pm4[block->pm4_ndwords];
- block->pm4_ndwords += n;
- block->nreg = n;
- block->nreg_dirty = n;
- LIST_INITHEAD(&block->list);
- LIST_INITHEAD(&block->enable_list);
-
- for (j = 0; j < n; j++) {
- if (reg[i+j].flags & REG_FLAG_DIRTY_ALWAYS) {
- block->flags |= REG_FLAG_DIRTY_ALWAYS;
- }
- if (reg[i+j].flags & REG_FLAG_ENABLE_ALWAYS) {
- if (!(block->status & R600_BLOCK_STATUS_ENABLED)) {
- block->status |= R600_BLOCK_STATUS_ENABLED;
- LIST_ADDTAIL(&block->enable_list, &ctx->enable_list);
- LIST_ADDTAIL(&block->list,&ctx->dirty);
- }
- }
- if (reg[i+j].flags & REG_FLAG_FLUSH_CHANGE) {
- block->flags |= REG_FLAG_FLUSH_CHANGE;
- }
-
- if (reg[i+j].flags & REG_FLAG_NEED_BO) {
- block->nbo++;
- assert(block->nbo < R600_BLOCK_MAX_BO);
- block->pm4_bo_index[j] = block->nbo;
- block->pm4[block->pm4_ndwords++] = PKT3(PKT3_NOP, 0, 0);
- block->pm4[block->pm4_ndwords++] = 0x00000000;
- block->reloc[block->nbo].bo_pm4_index = block->pm4_ndwords - 1;
- }
- }
- /* check that we stay in limit */
- assert(block->pm4_ndwords < R600_BLOCK_MAX_REG);
-}
-
-int r600_context_add_block(struct r600_context *ctx, const struct r600_reg *reg, unsigned nreg,
- unsigned opcode, unsigned offset_base)
-{
- struct r600_block *block;
- struct r600_range *range;
- int offset;
-
- for (unsigned i = 0, n = 0; i < nreg; i += n) {
- /* ignore new block balise */
- if (reg[i].offset == GROUP_FORCE_NEW_BLOCK) {
- n = 1;
- continue;
- }
-
- /* register that need relocation are in their own group */
- /* find number of consecutive registers */
- n = 0;
- offset = reg[i].offset;
- while (reg[i + n].offset == offset) {
- n++;
- offset += 4;
- if ((n + i) >= nreg)
- break;
- if (n >= (R600_BLOCK_MAX_REG - 2))
- break;
- }
-
- /* allocate new block */
- block = calloc(1, sizeof(struct r600_block));
- if (block == NULL) {
- return -ENOMEM;
- }
- ctx->nblocks++;
- for (int j = 0; j < n; j++) {
- range = &ctx->range[CTX_RANGE_ID(reg[i + j].offset)];
- /* create block table if it doesn't exist */
- if (!range->blocks)
- range->blocks = calloc(1 << HASH_SHIFT, sizeof(void *));
- if (!range->blocks) {
- free(block);
- return -1;
- }
-
- range->blocks[CTX_BLOCK_ID(reg[i + j].offset)] = block;
- }
-
- r600_init_block(ctx, block, reg, i, n, opcode, offset_base);
-
- }
- return 0;
-}
-
-static const struct r600_reg r600_context_reg_list[] = {
- {R_028D24_DB_HTILE_SURFACE, 0, 0},
- {R_028614_SPI_VS_OUT_ID_0, 0, 0},
- {R_028618_SPI_VS_OUT_ID_1, 0, 0},
- {R_02861C_SPI_VS_OUT_ID_2, 0, 0},
- {R_028620_SPI_VS_OUT_ID_3, 0, 0},
- {R_028624_SPI_VS_OUT_ID_4, 0, 0},
- {R_028628_SPI_VS_OUT_ID_5, 0, 0},
- {R_02862C_SPI_VS_OUT_ID_6, 0, 0},
- {R_028630_SPI_VS_OUT_ID_7, 0, 0},
- {R_028634_SPI_VS_OUT_ID_8, 0, 0},
- {R_028638_SPI_VS_OUT_ID_9, 0, 0},
- {R_0286C4_SPI_VS_OUT_CONFIG, 0, 0},
- {GROUP_FORCE_NEW_BLOCK, 0, 0},
- {R_028858_SQ_PGM_START_VS, REG_FLAG_NEED_BO, 0},
- {GROUP_FORCE_NEW_BLOCK, 0, 0},
- {R_028868_SQ_PGM_RESOURCES_VS, 0, 0},
- {GROUP_FORCE_NEW_BLOCK, 0, 0},
- {R_0288A4_SQ_PGM_RESOURCES_FS, 0, 0},
- {R_0288DC_SQ_PGM_CF_OFFSET_FS, 0, 0},
- {R_028644_SPI_PS_INPUT_CNTL_0, 0, 0},
- {R_028648_SPI_PS_INPUT_CNTL_1, 0, 0},
- {R_02864C_SPI_PS_INPUT_CNTL_2, 0, 0},
- {R_028650_SPI_PS_INPUT_CNTL_3, 0, 0},
- {R_028654_SPI_PS_INPUT_CNTL_4, 0, 0},
- {R_028658_SPI_PS_INPUT_CNTL_5, 0, 0},
- {R_02865C_SPI_PS_INPUT_CNTL_6, 0, 0},
- {R_028660_SPI_PS_INPUT_CNTL_7, 0, 0},
- {R_028664_SPI_PS_INPUT_CNTL_8, 0, 0},
- {R_028668_SPI_PS_INPUT_CNTL_9, 0, 0},
- {R_02866C_SPI_PS_INPUT_CNTL_10, 0, 0},
- {R_028670_SPI_PS_INPUT_CNTL_11, 0, 0},
- {R_028674_SPI_PS_INPUT_CNTL_12, 0, 0},
- {R_028678_SPI_PS_INPUT_CNTL_13, 0, 0},
- {R_02867C_SPI_PS_INPUT_CNTL_14, 0, 0},
- {R_028680_SPI_PS_INPUT_CNTL_15, 0, 0},
- {R_028684_SPI_PS_INPUT_CNTL_16, 0, 0},
- {R_028688_SPI_PS_INPUT_CNTL_17, 0, 0},
- {R_02868C_SPI_PS_INPUT_CNTL_18, 0, 0},
- {R_028690_SPI_PS_INPUT_CNTL_19, 0, 0},
- {R_028694_SPI_PS_INPUT_CNTL_20, 0, 0},
- {R_028698_SPI_PS_INPUT_CNTL_21, 0, 0},
- {R_02869C_SPI_PS_INPUT_CNTL_22, 0, 0},
- {R_0286A0_SPI_PS_INPUT_CNTL_23, 0, 0},
- {R_0286A4_SPI_PS_INPUT_CNTL_24, 0, 0},
- {R_0286A8_SPI_PS_INPUT_CNTL_25, 0, 0},
- {R_0286AC_SPI_PS_INPUT_CNTL_26, 0, 0},
- {R_0286B0_SPI_PS_INPUT_CNTL_27, 0, 0},
- {R_0286B4_SPI_PS_INPUT_CNTL_28, 0, 0},
- {R_0286B8_SPI_PS_INPUT_CNTL_29, 0, 0},
- {R_0286BC_SPI_PS_INPUT_CNTL_30, 0, 0},
- {R_0286C0_SPI_PS_INPUT_CNTL_31, 0, 0},
- {R_0286CC_SPI_PS_IN_CONTROL_0, 0, 0},
- {R_0286D0_SPI_PS_IN_CONTROL_1, 0, 0},
- {R_0286D8_SPI_INPUT_Z, 0, 0},
- {GROUP_FORCE_NEW_BLOCK, 0, 0},
- {R_028840_SQ_PGM_START_PS, REG_FLAG_NEED_BO, 0},
- {GROUP_FORCE_NEW_BLOCK, 0, 0},
- {R_028850_SQ_PGM_RESOURCES_PS, 0, 0},
- {R_028854_SQ_PGM_EXPORTS_PS, 0, 0},
-};
-
-/* initialize */
-void r600_context_fini(struct r600_context *ctx)
-{
- struct r600_block *block;
- struct r600_range *range;
-
- if (ctx->range) {
- for (int i = 0; i < NUM_RANGES; i++) {
- if (!ctx->range[i].blocks)
- continue;
- for (int j = 0; j < (1 << HASH_SHIFT); j++) {
- block = ctx->range[i].blocks[j];
- if (block) {
- for (int k = 0, offset = block->start_offset; k < block->nreg; k++, offset += 4) {
- range = &ctx->range[CTX_RANGE_ID(offset)];
- range->blocks[CTX_BLOCK_ID(offset)] = NULL;
- }
- for (int k = 1; k <= block->nbo; k++) {
- pipe_resource_reference((struct pipe_resource**)&block->reloc[k].bo, NULL);
- }
- free(block);
- }
- }
- free(ctx->range[i].blocks);
- }
- }
- free(ctx->blocks);
-}
-
-int r600_setup_block_table(struct r600_context *ctx)
-{
- /* setup block table */
- int c = 0;
- ctx->blocks = calloc(ctx->nblocks, sizeof(void*));
- if (!ctx->blocks)
- return -ENOMEM;
- for (int i = 0; i < NUM_RANGES; i++) {
- if (!ctx->range[i].blocks)
- continue;
- for (int j = 0, add; j < (1 << HASH_SHIFT); j++) {
- if (!ctx->range[i].blocks[j])
- continue;
-
- add = 1;
- for (int k = 0; k < c; k++) {
- if (ctx->blocks[k] == ctx->range[i].blocks[j]) {
- add = 0;
- break;
- }
- }
- if (add) {
- assert(c < ctx->nblocks);
- ctx->blocks[c++] = ctx->range[i].blocks[j];
- j += (ctx->range[i].blocks[j]->nreg) - 1;
- }
- }
- }
- return 0;
-}
-
-int r600_context_init(struct r600_context *ctx)
-{
- int r;
-
- /* add blocks */
- r = r600_context_add_block(ctx, r600_context_reg_list,
- Elements(r600_context_reg_list), PKT3_SET_CONTEXT_REG, R600_CONTEXT_REG_OFFSET);
- if (r)
- goto out_err;
-
- r = r600_setup_block_table(ctx);
- if (r)
- goto out_err;
-
- ctx->max_db = 4;
- return 0;
-out_err:
- r600_context_fini(ctx);
- return r;
-}
-
void r600_need_cs_space(struct r600_context *ctx, unsigned num_dw,
boolean count_draw_in)
{
for (i = 0; i < R600_NUM_ATOMS; i++) {
if (ctx->atoms[i] && ctx->atoms[i]->dirty) {
num_dw += ctx->atoms[i]->num_dw;
-#if R600_TRACE_CS
if (ctx->screen->trace_bo) {
num_dw += R600_TRACE_CS_DWORDS;
}
-#endif
}
}
- num_dw += ctx->pm4_dirty_cdwords;
-
/* The upper-bound of how much space a draw command would take. */
num_dw += R600_MAX_FLUSH_CS_DWORDS + R600_MAX_DRAW_CS_DWORDS;
-#if R600_TRACE_CS
if (ctx->screen->trace_bo) {
num_dw += R600_TRACE_CS_DWORDS;
}
-#endif
}
/* Count in queries_suspend. */
}
}
-void r600_context_dirty_block(struct r600_context *ctx,
- struct r600_block *block,
- int dirty, int index)
-{
- if ((index + 1) > block->nreg_dirty)
- block->nreg_dirty = index + 1;
-
- if ((dirty != (block->status & R600_BLOCK_STATUS_DIRTY)) || !(block->status & R600_BLOCK_STATUS_ENABLED)) {
- block->status |= R600_BLOCK_STATUS_DIRTY;
- ctx->pm4_dirty_cdwords += block->pm4_ndwords;
- if (!(block->status & R600_BLOCK_STATUS_ENABLED)) {
- block->status |= R600_BLOCK_STATUS_ENABLED;
- LIST_ADDTAIL(&block->enable_list, &ctx->enable_list);
- }
- LIST_ADDTAIL(&block->list,&ctx->dirty);
-
- if (block->flags & REG_FLAG_FLUSH_CHANGE) {
- ctx->flags |= R600_CONTEXT_WAIT_3D_IDLE;
- }
- }
-}
-
-/**
- * If reg needs a reloc, this function will add it to its block's reloc list.
- * @return true if reg needs a reloc, false otherwise
- */
-static bool r600_reg_set_block_reloc(struct r600_pipe_reg *reg)
-{
- unsigned reloc_id;
-
- if (!reg->block->pm4_bo_index[reg->id]) {
- return false;
- }
- /* find relocation */
- reloc_id = reg->block->pm4_bo_index[reg->id];
- pipe_resource_reference(
- (struct pipe_resource**)®->block->reloc[reloc_id].bo,
- ®->bo->b.b);
- reg->block->reloc[reloc_id].bo_usage = reg->bo_usage;
- return true;
-}
-
-/**
- * This function will emit all the registers in state directly to the command
- * stream allowing you to bypass the r600_context dirty list.
- *
- * This is used for dispatching compute shaders to avoid mixing compute and
- * 3D states in the context's dirty list.
- *
- * @param pkt_flags Should be either 0 or RADEON_CP_PACKET3_COMPUTE_MODE. This
- * value will be passed on to r600_context_block_emit_dirty an or'd against
- * the PKT3 headers.
- */
-void r600_context_pipe_state_emit(struct r600_context *ctx,
- struct r600_pipe_state *state,
- unsigned pkt_flags)
-{
- unsigned i;
-
- /* Mark all blocks as dirty:
- * Since two registers can be in the same block, we need to make sure
- * we mark all the blocks dirty before we emit any of them. If we were
- * to mark blocks dirty and emit them in the same loop, like this:
- *
- * foreach (reg in state->regs) {
- * mark_dirty(reg->block)
- * emit_block(reg->block)
- * }
- *
- * Then if we have two registers in this state that are in the same
- * block, we would end up emitting that block twice.
- */
- for (i = 0; i < state->nregs; i++) {
- struct r600_pipe_reg *reg = &state->regs[i];
- /* Mark all the registers in the block as dirty */
- reg->block->nreg_dirty = reg->block->nreg;
- reg->block->status |= R600_BLOCK_STATUS_DIRTY;
- /* Update the reloc for this register if necessary. */
- r600_reg_set_block_reloc(reg);
- }
-
- /* Emit the registers writes */
- for (i = 0; i < state->nregs; i++) {
- struct r600_pipe_reg *reg = &state->regs[i];
- if (reg->block->status & R600_BLOCK_STATUS_DIRTY) {
- r600_context_block_emit_dirty(ctx, reg->block, pkt_flags);
- }
- }
-}
-
-void r600_context_pipe_state_set(struct r600_context *ctx, struct r600_pipe_state *state)
-{
- struct r600_block *block;
- int dirty;
- for (int i = 0; i < state->nregs; i++) {
- unsigned id;
- struct r600_pipe_reg *reg = &state->regs[i];
-
- block = reg->block;
- id = reg->id;
-
- dirty = block->status & R600_BLOCK_STATUS_DIRTY;
-
- if (reg->value != block->reg[id]) {
- block->reg[id] = reg->value;
- dirty |= R600_BLOCK_STATUS_DIRTY;
- }
- if (block->flags & REG_FLAG_DIRTY_ALWAYS)
- dirty |= R600_BLOCK_STATUS_DIRTY;
- if (r600_reg_set_block_reloc(reg)) {
- /* always force dirty for relocs for now */
- dirty |= R600_BLOCK_STATUS_DIRTY;
- }
-
- if (dirty)
- r600_context_dirty_block(ctx, block, dirty, id);
- }
-}
-
-/**
- * @param pkt_flags should be set to RADEON_CP_PACKET3_COMPUTE_MODE if this
- * block will be used for compute shaders.
- */
-void r600_context_block_emit_dirty(struct r600_context *ctx, struct r600_block *block,
- unsigned pkt_flags)
-{
- struct radeon_winsys_cs *cs = ctx->rings.gfx.cs;
- int optional = block->nbo == 0 && !(block->flags & REG_FLAG_DIRTY_ALWAYS);
- int cp_dwords = block->pm4_ndwords, start_dword = 0;
- int new_dwords = 0;
- int nbo = block->nbo;
-
- if (block->nreg_dirty == 0 && optional) {
- goto out;
- }
-
- if (nbo) {
- for (int j = 0; j < block->nreg; j++) {
- if (block->pm4_bo_index[j]) {
- /* find relocation */
- struct r600_block_reloc *reloc = &block->reloc[block->pm4_bo_index[j]];
- if (reloc->bo) {
- block->pm4[reloc->bo_pm4_index] =
- r600_context_bo_reloc(ctx, &ctx->rings.gfx, reloc->bo, reloc->bo_usage);
- } else {
- block->pm4[reloc->bo_pm4_index] = 0;
- }
- nbo--;
- if (nbo == 0)
- break;
-
- }
- }
- }
-
- optional &= (block->nreg_dirty != block->nreg);
- if (optional) {
- new_dwords = block->nreg_dirty;
- start_dword = cs->cdw;
- cp_dwords = new_dwords + 2;
- }
- memcpy(&cs->buf[cs->cdw], block->pm4, cp_dwords * 4);
-
- /* We are applying the pkt_flags after copying the register block to
- * the the command stream, because it is possible this block will be
- * emitted with a different pkt_flags, and we don't want to store the
- * pkt_flags in the block.
- */
- cs->buf[cs->cdw] |= pkt_flags;
- cs->cdw += cp_dwords;
-
- if (optional) {
- uint32_t newword;
-
- newword = cs->buf[start_dword];
- newword &= PKT_COUNT_C;
- newword |= PKT_COUNT_S(new_dwords);
- cs->buf[start_dword] = newword;
- }
-out:
- block->status ^= R600_BLOCK_STATUS_DIRTY;
- block->nreg_dirty = 0;
- LIST_DELINIT(&block->list);
-}
-
void r600_flush_emit(struct r600_context *rctx)
{
struct radeon_winsys_cs *cs = rctx->rings.gfx.cs;
unsigned cp_coher_cntl = 0;
unsigned wait_until = 0;
- unsigned emit_flush = 0;
if (!rctx->flags) {
return;
if (rctx->flags & R600_CONTEXT_FLUSH_AND_INV) {
cs->buf[cs->cdw++] = PKT3(PKT3_EVENT_WRITE, 0, 0);
cs->buf[cs->cdw++] = EVENT_TYPE(EVENT_TYPE_CACHE_FLUSH_AND_INV_EVENT) | EVENT_INDEX(0);
- if (rctx->chip_class >= EVERGREEN) {
- cp_coher_cntl = S_0085F0_CB0_DEST_BASE_ENA(1) |
- S_0085F0_CB1_DEST_BASE_ENA(1) |
- S_0085F0_CB2_DEST_BASE_ENA(1) |
- S_0085F0_CB3_DEST_BASE_ENA(1) |
- S_0085F0_CB4_DEST_BASE_ENA(1) |
- S_0085F0_CB5_DEST_BASE_ENA(1) |
- S_0085F0_CB6_DEST_BASE_ENA(1) |
- S_0085F0_CB7_DEST_BASE_ENA(1) |
- S_0085F0_CB8_DEST_BASE_ENA(1) |
- S_0085F0_CB9_DEST_BASE_ENA(1) |
- S_0085F0_CB10_DEST_BASE_ENA(1) |
- S_0085F0_CB11_DEST_BASE_ENA(1) |
- S_0085F0_DB_DEST_BASE_ENA(1) |
- S_0085F0_TC_ACTION_ENA(1) |
- S_0085F0_CB_ACTION_ENA(1) |
- S_0085F0_DB_ACTION_ENA(1) |
- S_0085F0_SH_ACTION_ENA(1) |
- S_0085F0_SMX_ACTION_ENA(1) |
- S_0085F0_FULL_CACHE_ENA(1);
- } else {
- cp_coher_cntl = S_0085F0_SMX_ACTION_ENA(1) |
- S_0085F0_SH_ACTION_ENA(1) |
- S_0085F0_VC_ACTION_ENA(1) |
- S_0085F0_TC_ACTION_ENA(1) |
- S_0085F0_FULL_CACHE_ENA(1);
- }
- emit_flush = 1;
}
if (rctx->flags & R600_CONTEXT_INVAL_READ_CACHES) {
cp_coher_cntl |= S_0085F0_VC_ACTION_ENA(1) |
S_0085F0_TC_ACTION_ENA(1) |
+ S_0085F0_SH_ACTION_ENA(1) |
S_0085F0_FULL_CACHE_ENA(1);
- emit_flush = 1;
+ }
+
+ if (rctx->flags & R600_CONTEXT_FLUSH_AND_INV_DB) {
+ cp_coher_cntl |= S_0085F0_DB_ACTION_ENA(1) |
+ S_0085F0_DB_DEST_BASE_ENA(1) |
+ S_0085F0_SMX_ACTION_ENA(1);
+ }
+
+ if (rctx->flags & R600_CONTEXT_FLUSH_AND_INV_CB) {
+ cp_coher_cntl |= S_0085F0_CB_ACTION_ENA(1) |
+ S_0085F0_CB0_DEST_BASE_ENA(1) |
+ S_0085F0_CB1_DEST_BASE_ENA(1) |
+ S_0085F0_CB2_DEST_BASE_ENA(1) |
+ S_0085F0_CB3_DEST_BASE_ENA(1) |
+ S_0085F0_CB4_DEST_BASE_ENA(1) |
+ S_0085F0_CB5_DEST_BASE_ENA(1) |
+ S_0085F0_CB6_DEST_BASE_ENA(1) |
+ S_0085F0_CB7_DEST_BASE_ENA(1) |
+ S_0085F0_SMX_ACTION_ENA(1);
+ if (rctx->chip_class >= EVERGREEN)
+ cp_coher_cntl |= S_0085F0_CB8_DEST_BASE_ENA(1) |
+ S_0085F0_CB9_DEST_BASE_ENA(1) |
+ S_0085F0_CB10_DEST_BASE_ENA(1) |
+ S_0085F0_CB11_DEST_BASE_ENA(1);
}
if (rctx->flags & R600_CONTEXT_STREAMOUT_FLUSH) {
S_0085F0_SO2_DEST_BASE_ENA(1) |
S_0085F0_SO3_DEST_BASE_ENA(1) |
S_0085F0_SMX_ACTION_ENA(1);
- emit_flush = 1;
}
- if (emit_flush) {
+ if (cp_coher_cntl) {
cs->buf[cs->cdw++] = PKT3(PKT3_SURFACE_SYNC, 3, 0);
cs->buf[cs->cdw++] = cp_coher_cntl; /* CP_COHER_CNTL */
cs->buf[cs->cdw++] = 0xffffffff; /* CP_COHER_SIZE */
* this will also flush the framebuffer cache
*/
ctx->flags |= R600_CONTEXT_FLUSH_AND_INV |
+ R600_CONTEXT_FLUSH_AND_INV_CB |
+ R600_CONTEXT_FLUSH_AND_INV_DB |
R600_CONTEXT_FLUSH_AND_INV_CB_META |
R600_CONTEXT_FLUSH_AND_INV_DB_META |
R600_CONTEXT_WAIT_3D_IDLE |
}
/* Flush the CS. */
-#if R600_TRACE_CS
- if (ctx->screen->trace_bo) {
- struct r600_screen *rscreen = ctx->screen;
- unsigned i;
-
- for (i = 0; i < cs->cdw; i++) {
- fprintf(stderr, "[%4d] [%5d] 0x%08x\n", rscreen->cs_count, i, cs->buf[i]);
- }
- rscreen->cs_count++;
- }
-#endif
- ctx->ws->cs_flush(ctx->rings.gfx.cs, flags);
-#if R600_TRACE_CS
- if (ctx->screen->trace_bo) {
- struct r600_screen *rscreen = ctx->screen;
- unsigned i;
-
- for (i = 0; i < 10; i++) {
- usleep(5);
- if (!ctx->ws->buffer_is_busy(rscreen->trace_bo->buf, RADEON_USAGE_READWRITE)) {
- break;
- }
- }
- if (i == 10) {
- fprintf(stderr, "timeout on cs lockup likely happen at cs %d dw %d\n",
- rscreen->trace_ptr[1], rscreen->trace_ptr[0]);
- } else {
- fprintf(stderr, "cs %d executed in %dms\n", rscreen->trace_ptr[1], i * 5);
- }
- }
-#endif
+ ctx->ws->cs_flush(ctx->rings.gfx.cs, flags, ctx->screen->cs_count++);
}
void r600_begin_new_cs(struct r600_context *ctx)
{
- struct r600_block *enable_block = NULL;
unsigned shader;
- ctx->pm4_dirty_cdwords = 0;
ctx->flags = 0;
ctx->gtt = 0;
ctx->vram = 0;
ctx->db_misc_state.atom.dirty = true;
ctx->db_state.atom.dirty = true;
ctx->framebuffer.atom.dirty = true;
+ ctx->pixel_shader.atom.dirty = true;
ctx->poly_offset_state.atom.dirty = true;
ctx->vgt_state.atom.dirty = true;
ctx->sample_mask.atom.dirty = true;
ctx->config_state.atom.dirty = true;
ctx->stencil_ref.atom.dirty = true;
ctx->vertex_fetch_shader.atom.dirty = true;
+ ctx->vertex_shader.atom.dirty = true;
ctx->viewport.atom.dirty = true;
if (ctx->blend_state.cso)
r600_resume_nontimer_queries(ctx);
}
- /* set all valid group as dirty so they get reemited on
- * next draw command
- */
- LIST_FOR_EACH_ENTRY(enable_block, &ctx->enable_list, enable_list) {
- if(!(enable_block->status & R600_BLOCK_STATUS_DIRTY)) {
- LIST_ADDTAIL(&enable_block->list,&ctx->dirty);
- enable_block->status |= R600_BLOCK_STATUS_DIRTY;
- }
- ctx->pm4_dirty_cdwords += enable_block->pm4_ndwords;
- enable_block->nreg_dirty = enable_block->nreg;
- }
-
/* Re-emit the draw state. */
ctx->last_primitive_type = -1;
ctx->last_start_instance = -1;
struct radeon_winsys_cs *cs = rctx->rings.gfx.cs;
assert(size);
- assert(rctx->chip_class != R600);
-
- /* CP DMA doesn't work on R600 (flushing seems to be unreliable). */
- if (rctx->chip_class == R600) {
- return;
- }
+ assert(rctx->screen->has_cp_dma);
dst_offset += r600_resource_va(&rctx->screen->screen, dst);
src_offset += r600_resource_va(&rctx->screen->screen, src);
* to resources which are bound right now. */
rctx->flags |= R600_CONTEXT_INVAL_READ_CACHES |
R600_CONTEXT_FLUSH_AND_INV |
+ R600_CONTEXT_FLUSH_AND_INV_CB |
+ R600_CONTEXT_FLUSH_AND_INV_DB |
R600_CONTEXT_FLUSH_AND_INV_CB_META |
R600_CONTEXT_FLUSH_AND_INV_DB_META |
R600_CONTEXT_STREAMOUT_FLUSH |
/* Invalidate the read caches. */
rctx->flags |= R600_CONTEXT_INVAL_READ_CACHES;
+
+ util_range_add(&r600_resource(dst)->valid_buffer_range, dst_offset,
+ dst_offset + size);
}
void r600_need_dma_space(struct r600_context *ctx, unsigned num_dw)
src_offset += csize << shift;
size -= csize;
}
+
+ util_range_add(&rdst->valid_buffer_range, dst_offset,
+ dst_offset + size);
}
projects / mesa.git / blobdiff