#include "util/u_suballoc.h"
#include "amd/common/sid.h"
-#define SI_MAX_STREAMS 4
-
static const struct si_query_ops query_hw_ops;
struct si_hw_query_params {
struct pipe_fence_handle *fence;
};
-static void si_query_sw_destroy(struct si_screen *sscreen,
+static void si_query_sw_destroy(struct si_context *sctx,
struct si_query *squery)
{
struct si_query_sw *query = (struct si_query_sw *)squery;
- sscreen->b.fence_reference(&sscreen->b, &query->fence, NULL);
+ sctx->b.screen->fence_reference(sctx->b.screen, &query->fence, NULL);
FREE(query);
}
query->begin_result =
p_atomic_read(&sctx->screen->num_shader_cache_hits);
break;
+ case SI_QUERY_PD_NUM_PRIMS_ACCEPTED:
+ query->begin_result = sctx->compute_num_verts_accepted;
+ break;
+ case SI_QUERY_PD_NUM_PRIMS_REJECTED:
+ query->begin_result = sctx->compute_num_verts_rejected;
+ break;
+ case SI_QUERY_PD_NUM_PRIMS_INELIGIBLE:
+ query->begin_result = sctx->compute_num_verts_ineligible;
+ break;
case SI_QUERY_GPIN_ASIC_ID:
case SI_QUERY_GPIN_NUM_SIMD:
case SI_QUERY_GPIN_NUM_RB:
query->end_result =
p_atomic_read(&sctx->screen->num_shader_cache_hits);
break;
+ case SI_QUERY_PD_NUM_PRIMS_ACCEPTED:
+ query->end_result = sctx->compute_num_verts_accepted;
+ break;
+ case SI_QUERY_PD_NUM_PRIMS_REJECTED:
+ query->end_result = sctx->compute_num_verts_rejected;
+ break;
+ case SI_QUERY_PD_NUM_PRIMS_INELIGIBLE:
+ query->end_result = sctx->compute_num_verts_ineligible;
+ break;
case SI_QUERY_GPIN_ASIC_ID:
case SI_QUERY_GPIN_NUM_SIMD:
case SI_QUERY_GPIN_NUM_RB:
result->u64 = (query->end_result - query->begin_result) * 100 /
(query->end_time - query->begin_time);
return true;
+ case SI_QUERY_PD_NUM_PRIMS_ACCEPTED:
+ case SI_QUERY_PD_NUM_PRIMS_REJECTED:
+ case SI_QUERY_PD_NUM_PRIMS_INELIGIBLE:
+ result->u64 = ((unsigned)query->end_result -
+ (unsigned)query->begin_result) / 3;
+ return true;
case SI_QUERY_GPIN_ASIC_ID:
result->u32 = 0;
return true;
}
-void si_query_hw_destroy(struct si_screen *sscreen,
- struct si_query *squery)
+void si_query_hw_destroy(struct si_context *sctx, struct si_query *squery)
{
struct si_query_hw *query = (struct si_query_hw *)squery;
- si_query_buffer_destroy(sscreen, &query->buffer);
+ si_query_buffer_destroy(sctx->screen, &query->buffer);
si_resource_reference(&query->workaround_buf, NULL);
FREE(squery);
}
emit_sample_streamout(cs, va + 32 * stream, stream);
break;
case PIPE_QUERY_TIME_ELAPSED:
- si_cp_release_mem(sctx, V_028A90_BOTTOM_OF_PIPE_TS, 0,
+ si_cp_release_mem(sctx, cs, V_028A90_BOTTOM_OF_PIPE_TS, 0,
EOP_DST_SEL_MEM, EOP_INT_SEL_NONE,
EOP_DATA_SEL_TIMESTAMP, NULL, va,
0, query->b.type);
si_update_occlusion_query_state(sctx, query->b.type, 1);
si_update_prims_generated_query_state(sctx, query->b.type, 1);
+ if (query->b.type == PIPE_QUERY_PIPELINE_STATISTICS)
+ sctx->num_pipeline_stat_queries++;
+
if (query->b.type != SI_QUERY_TIME_ELAPSED_SDMA)
si_need_gfx_cs_space(sctx);
va += 8;
/* fall through */
case PIPE_QUERY_TIMESTAMP:
- si_cp_release_mem(sctx, V_028A90_BOTTOM_OF_PIPE_TS, 0,
+ si_cp_release_mem(sctx, cs, V_028A90_BOTTOM_OF_PIPE_TS, 0,
EOP_DST_SEL_MEM, EOP_INT_SEL_NONE,
EOP_DATA_SEL_TIMESTAMP, NULL, va,
0, query->b.type);
RADEON_PRIO_QUERY);
if (fence_va) {
- si_cp_release_mem(sctx, V_028A90_BOTTOM_OF_PIPE_TS, 0,
+ si_cp_release_mem(sctx, cs, V_028A90_BOTTOM_OF_PIPE_TS, 0,
EOP_DST_SEL_MEM, EOP_INT_SEL_NONE,
EOP_DATA_SEL_VALUE_32BIT,
query->buffer.buf, fence_va, 0x80000000,
si_update_occlusion_query_state(sctx, query->b.type, -1);
si_update_prims_generated_query_state(sctx, query->b.type, -1);
+
+ if (query->b.type == PIPE_QUERY_PIPELINE_STATISTICS)
+ sctx->num_pipeline_stat_queries--;
}
static void emit_set_predicate(struct si_context *ctx,
if (!query)
return;
+ if (ctx->screen->use_ngg_streamout &&
+ (query->b.type == PIPE_QUERY_SO_OVERFLOW_PREDICATE ||
+ query->b.type == PIPE_QUERY_SO_OVERFLOW_ANY_PREDICATE)) {
+ assert(!"not implemented");
+ }
+
invert = ctx->render_cond_invert;
flag_wait = ctx->render_cond_mode == PIPE_RENDER_COND_WAIT ||
ctx->render_cond_mode == PIPE_RENDER_COND_BY_REGION_WAIT;
query_type != SI_QUERY_TIME_ELAPSED_SDMA))
return si_query_sw_create(query_type);
+ if (sscreen->use_ngg_streamout &&
+ (query_type == PIPE_QUERY_PRIMITIVES_EMITTED ||
+ query_type == PIPE_QUERY_PRIMITIVES_GENERATED ||
+ query_type == PIPE_QUERY_SO_STATISTICS ||
+ query_type == PIPE_QUERY_SO_OVERFLOW_PREDICATE ||
+ query_type == PIPE_QUERY_SO_OVERFLOW_ANY_PREDICATE))
+ return gfx10_sh_query_create(sscreen, query_type, index);
+
return si_query_hw_create(sscreen, query_type, index);
}
struct si_context *sctx = (struct si_context *)ctx;
struct si_query *squery = (struct si_query *)query;
- squery->ops->destroy(sctx->screen, squery);
+ squery->ops->destroy(sctx, squery);
}
-static boolean si_begin_query(struct pipe_context *ctx,
- struct pipe_query *query)
+static bool si_begin_query(struct pipe_context *ctx,
+ struct pipe_query *query)
{
struct si_context *sctx = (struct si_context *)ctx;
struct si_query *squery = (struct si_query *)query;
if (!query->buffer.buf)
return false;
- LIST_ADDTAIL(&query->b.active_list, &sctx->active_queries);
+ list_addtail(&query->b.active_list, &sctx->active_queries);
sctx->num_cs_dw_queries_suspend += query->b.num_cs_dw_suspend;
return true;
}
si_query_hw_emit_stop(sctx, query);
if (!(query->flags & SI_QUERY_HW_FLAG_NO_START)) {
- LIST_DELINIT(&query->b.active_list);
+ list_delinit(&query->b.active_list);
sctx->num_cs_dw_queries_suspend -= query->b.num_cs_dw_suspend;
}
.resume = si_query_hw_resume,
};
-static boolean si_get_query_result(struct pipe_context *ctx,
- struct pipe_query *query, boolean wait,
- union pipe_query_result *result)
+static bool si_get_query_result(struct pipe_context *ctx,
+ struct pipe_query *query, bool wait,
+ union pipe_query_result *result)
{
struct si_context *sctx = (struct si_context *)ctx;
struct si_query *squery = (struct si_query *)query;
static void si_get_query_result_resource(struct pipe_context *ctx,
struct pipe_query *query,
- boolean wait,
+ bool wait,
enum pipe_query_value_type result_type,
int index,
struct pipe_resource *resource,
return true;
}
-static void si_restore_qbo_state(struct si_context *sctx,
- struct si_qbo_state *st)
-{
- sctx->b.bind_compute_state(&sctx->b, st->saved_compute);
-
- sctx->b.set_constant_buffer(&sctx->b, PIPE_SHADER_COMPUTE, 0, &st->saved_const0);
- pipe_resource_reference(&st->saved_const0.buffer, NULL);
-
- sctx->b.set_shader_buffers(&sctx->b, PIPE_SHADER_COMPUTE, 0, 3, st->saved_ssbo,
- st->saved_ssbo_writable_mask);
- for (unsigned i = 0; i < 3; ++i)
- pipe_resource_reference(&st->saved_ssbo[i].buffer, NULL);
-}
-
static void si_query_hw_get_result_resource(struct si_context *sctx,
struct si_query *squery,
bool wait,
static void si_render_condition(struct pipe_context *ctx,
struct pipe_query *query,
- boolean condition,
+ bool condition,
enum pipe_render_cond_flag mode)
{
struct si_context *sctx = (struct si_context *)ctx;
X("GPU-surf-sync-busy", GPU_SURF_SYNC_BUSY, UINT64, AVERAGE),
X("GPU-cp-dma-busy", GPU_CP_DMA_BUSY, UINT64, AVERAGE),
X("GPU-scratch-ram-busy", GPU_SCRATCH_RAM_BUSY, UINT64, AVERAGE),
+
+ X("pd-num-prims-accepted", PD_NUM_PRIMS_ACCEPTED, UINT64, AVERAGE),
+ X("pd-num-prims-rejected", PD_NUM_PRIMS_REJECTED, UINT64, AVERAGE),
+ X("pd-num-prims-ineligible", PD_NUM_PRIMS_INELIGIBLE,UINT64, AVERAGE),
};
#undef X
static unsigned si_get_num_queries(struct si_screen *sscreen)
{
/* amdgpu */
- if (sscreen->info.drm_major == 3) {
+ if (sscreen->info.is_amdgpu) {
if (sscreen->info.chip_class >= GFX8)
return ARRAY_SIZE(si_driver_query_list);
else
sctx->b.end_query = si_end_query;
sctx->b.get_query_result = si_get_query_result;
sctx->b.get_query_result_resource = si_get_query_result_resource;
- sctx->atoms.s.render_cond.emit = si_emit_query_predication;
- if (((struct si_screen*)sctx->b.screen)->info.num_render_backends > 0)
- sctx->b.render_condition = si_render_condition;
+ if (sctx->has_graphics) {
+ sctx->atoms.s.render_cond.emit = si_emit_query_predication;
+ sctx->b.render_condition = si_render_condition;
+ }
- LIST_INITHEAD(&sctx->active_queries);
+ list_inithead(&sctx->active_queries);
}
void si_init_screen_query_functions(struct si_screen *sscreen)
projects / mesa.git / blobdiff