summaryrefslogtreecommitdiff
path: root/src/mesa/drivers/dri/i965/brw_performance_query_mdapi.c
diff options
context:
space:
mode:
Diffstat (limited to 'src/mesa/drivers/dri/i965/brw_performance_query_mdapi.c')
-rw-r--r--src/mesa/drivers/dri/i965/brw_performance_query_mdapi.c114
1 files changed, 50 insertions, 64 deletions
diff --git a/src/mesa/drivers/dri/i965/brw_performance_query_mdapi.c b/src/mesa/drivers/dri/i965/brw_performance_query_mdapi.c
index 70f69debe98..0676e868b81 100644
--- a/src/mesa/drivers/dri/i965/brw_performance_query_mdapi.c
+++ b/src/mesa/drivers/dri/i965/brw_performance_query_mdapi.c
@@ -24,6 +24,8 @@
#include "brw_defines.h"
#include "brw_performance_query.h"
+#include "perf/gen_perf.h"
+
/**
* Data format expected by MDAPI.
*/
@@ -203,16 +205,19 @@ brw_perf_query_get_mdapi_oa_data(struct brw_context *brw,
}
static void
-fill_mdapi_perf_query_counter(struct brw_perf_query_info *query,
+fill_mdapi_perf_query_counter(struct gen_perf_query_info *query,
const char *name,
uint32_t data_offset,
uint32_t data_size,
- GLenum data_type)
+ enum gen_perf_counter_data_type data_type)
{
- struct brw_perf_query_counter *counter = &query->counters[query->n_counters];
+ struct gen_perf_query_counter *counter = &query->counters[query->n_counters];
+
+ assert(query->n_counters <= query->max_counters);
counter->name = name;
counter->desc = "Raw counter value";
+ counter->type = GEN_PERF_COUNTER_TYPE_RAW;
counter->data_type = data_type;
counter->offset = data_offset;
counter->size = data_size;
@@ -226,19 +231,21 @@ fill_mdapi_perf_query_counter(struct brw_perf_query_info *query,
(uint8_t *) &struct_name.field_name - \
(uint8_t *) &struct_name, \
sizeof(struct_name.field_name), \
- GL_PERFQUERY_COUNTER_DATA_##type_name##_INTEL)
+ GEN_PERF_COUNTER_DATA_TYPE_##type_name)
#define MDAPI_QUERY_ADD_ARRAY_COUNTER(ctx, query, struct_name, field_name, idx, type_name) \
fill_mdapi_perf_query_counter(query, \
ralloc_asprintf(ctx, "%s%i", #field_name, idx), \
(uint8_t *) &struct_name.field_name[idx] - \
(uint8_t *) &struct_name, \
sizeof(struct_name.field_name[0]), \
- GL_PERFQUERY_COUNTER_DATA_##type_name##_INTEL)
+ GEN_PERF_COUNTER_DATA_TYPE_##type_name)
void
brw_perf_query_register_mdapi_oa_query(struct brw_context *brw)
{
const struct gen_device_info *devinfo = &brw->screen->devinfo;
+ struct gen_perf *perf = brw->perfquery.perf;
+ struct gen_perf_query_info *query = NULL;
/* MDAPI requires different structures for pretty much every generation
* (right now we have definitions for gen 7 to 11).
@@ -246,36 +253,22 @@ brw_perf_query_register_mdapi_oa_query(struct brw_context *brw)
if (!(devinfo->gen >= 7 && devinfo->gen <= 11))
return;
- struct brw_perf_query_info *query = brw_perf_query_append_query_info(brw);
-
- query->kind = OA_COUNTERS_RAW;
- query->name = "Intel_Raw_Hardware_Counters_Set_0_Query";
- /* Guid has to matches with MDAPI's. */
- query->guid = "2f01b241-7014-42a7-9eb6-a925cad3daba";
- query->n_counters = 0;
- query->oa_metrics_set_id = 0; /* Set by MDAPI */
-
- int n_counters;
switch (devinfo->gen) {
case 7: {
+ query = gen_perf_query_append_query_info(perf, 1 + 45 + 16 + 7);
query->oa_format = I915_OA_FORMAT_A45_B8_C8;
struct mdapi_gen7_metrics metric_data;
query->data_size = sizeof(metric_data);
- n_counters = 1 + 45 + 16 + 7;
- query->counters =
- rzalloc_array_size(brw->perfquery.queries,
- sizeof(*query->counters), n_counters);
-
MDAPI_QUERY_ADD_COUNTER(query, metric_data, TotalTime, UINT64);
for (int i = 0; i < ARRAY_SIZE(metric_data.ACounters); i++) {
- MDAPI_QUERY_ADD_ARRAY_COUNTER(brw->perfquery.queries,
- query, metric_data, ACounters, i, UINT64);
+ MDAPI_QUERY_ADD_ARRAY_COUNTER(perf->queries, query,
+ metric_data, ACounters, i, UINT64);
}
for (int i = 0; i < ARRAY_SIZE(metric_data.NOACounters); i++) {
- MDAPI_QUERY_ADD_ARRAY_COUNTER(brw->perfquery.queries,
- query, metric_data, NOACounters, i, UINT64);
+ MDAPI_QUERY_ADD_ARRAY_COUNTER(perf->queries, query,
+ metric_data, NOACounters, i, UINT64);
}
MDAPI_QUERY_ADD_COUNTER(query, metric_data, PerfCounter1, UINT64);
MDAPI_QUERY_ADD_COUNTER(query, metric_data, PerfCounter2, UINT64);
@@ -287,25 +280,21 @@ brw_perf_query_register_mdapi_oa_query(struct brw_context *brw)
break;
}
case 8: {
+ query = gen_perf_query_append_query_info(perf, 2 + 36 + 16 + 16);
query->oa_format = I915_OA_FORMAT_A32u40_A4u32_B8_C8;
struct mdapi_gen8_metrics metric_data;
query->data_size = sizeof(metric_data);
- n_counters = 2 + 36 + 16 + 16;
- query->counters =
- rzalloc_array_size(brw->perfquery.queries,
- sizeof(*query->counters), n_counters);
-
MDAPI_QUERY_ADD_COUNTER(query, metric_data, TotalTime, UINT64);
MDAPI_QUERY_ADD_COUNTER(query, metric_data, GPUTicks, UINT64);
for (int i = 0; i < ARRAY_SIZE(metric_data.OaCntr); i++) {
- MDAPI_QUERY_ADD_ARRAY_COUNTER(brw->perfquery.queries,
- query, metric_data, OaCntr, i, UINT64);
+ MDAPI_QUERY_ADD_ARRAY_COUNTER(perf->queries, query,
+ metric_data, OaCntr, i, UINT64);
}
for (int i = 0; i < ARRAY_SIZE(metric_data.NoaCntr); i++) {
- MDAPI_QUERY_ADD_ARRAY_COUNTER(brw->perfquery.queries,
- query, metric_data, NoaCntr, i, UINT64);
+ MDAPI_QUERY_ADD_ARRAY_COUNTER(perf->queries, query,
+ metric_data, NoaCntr, i, UINT64);
}
MDAPI_QUERY_ADD_COUNTER(query, metric_data, BeginTimestamp, UINT64);
MDAPI_QUERY_ADD_COUNTER(query, metric_data, Reserved1, UINT64);
@@ -328,25 +317,21 @@ brw_perf_query_register_mdapi_oa_query(struct brw_context *brw)
case 9:
case 10:
case 11: {
+ query = gen_perf_query_append_query_info(perf, 2 + 36 + 16 + 16 + 16 + 2);
query->oa_format = I915_OA_FORMAT_A32u40_A4u32_B8_C8;
struct mdapi_gen9_metrics metric_data;
query->data_size = sizeof(metric_data);
- n_counters = 2 + 36 + 16 + 16 + 16 + 2;
- query->counters =
- rzalloc_array_size(brw->perfquery.queries,
- sizeof(*query->counters), n_counters);
-
MDAPI_QUERY_ADD_COUNTER(query, metric_data, TotalTime, UINT64);
MDAPI_QUERY_ADD_COUNTER(query, metric_data, GPUTicks, UINT64);
for (int i = 0; i < ARRAY_SIZE(metric_data.OaCntr); i++) {
- MDAPI_QUERY_ADD_ARRAY_COUNTER(brw->perfquery.queries,
- query, metric_data, OaCntr, i, UINT64);
+ MDAPI_QUERY_ADD_ARRAY_COUNTER(perf->queries, query,
+ metric_data, OaCntr, i, UINT64);
}
for (int i = 0; i < ARRAY_SIZE(metric_data.NoaCntr); i++) {
- MDAPI_QUERY_ADD_ARRAY_COUNTER(brw->perfquery.queries,
- query, metric_data, NoaCntr, i, UINT64);
+ MDAPI_QUERY_ADD_ARRAY_COUNTER(perf->queries, query,
+ metric_data, NoaCntr, i, UINT64);
}
MDAPI_QUERY_ADD_COUNTER(query, metric_data, BeginTimestamp, UINT64);
MDAPI_QUERY_ADD_COUNTER(query, metric_data, Reserved1, UINT64);
@@ -365,8 +350,8 @@ brw_perf_query_register_mdapi_oa_query(struct brw_context *brw)
MDAPI_QUERY_ADD_COUNTER(query, metric_data, ReportId, UINT32);
MDAPI_QUERY_ADD_COUNTER(query, metric_data, ReportsCount, UINT32);
for (int i = 0; i < ARRAY_SIZE(metric_data.UserCntr); i++) {
- MDAPI_QUERY_ADD_ARRAY_COUNTER(brw->perfquery.queries,
- query, metric_data, UserCntr, i, UINT64);
+ MDAPI_QUERY_ADD_ARRAY_COUNTER(perf->queries, query,
+ metric_data, UserCntr, i, UINT64);
}
MDAPI_QUERY_ADD_COUNTER(query, metric_data, UserCntrCfgId, UINT32);
MDAPI_QUERY_ADD_COUNTER(query, metric_data, Reserved4, UINT32);
@@ -377,12 +362,15 @@ brw_perf_query_register_mdapi_oa_query(struct brw_context *brw)
break;
}
- assert(query->n_counters <= n_counters);
+ query->kind = GEN_PERF_QUERY_TYPE_RAW;
+ query->name = "Intel_Raw_Hardware_Counters_Set_0_Query";
+ /* Guid has to matches with MDAPI's. */
+ query->guid = "2f01b241-7014-42a7-9eb6-a925cad3daba";
{
/* Accumulation buffer offsets copied from an actual query... */
- const struct brw_perf_query_info *copy_query =
- &brw->perfquery.queries[0];
+ const struct gen_perf_query_info *copy_query =
+ &brw->perfquery.perf->queries[0];
query->gpu_time_offset = copy_query->gpu_time_offset;
query->gpu_clock_offset = copy_query->gpu_clock_offset;
@@ -400,43 +388,41 @@ brw_perf_query_register_mdapi_statistic_query(struct brw_context *brw)
if (!(devinfo->gen >= 7 && devinfo->gen <= 9))
return;
- struct brw_perf_query_info *query = brw_perf_query_append_query_info(brw);
+ struct gen_perf_query_info *query =
+ gen_perf_query_append_query_info(brw->perfquery.perf, MAX_STAT_COUNTERS);
- query->kind = PIPELINE_STATS;
+ query->kind = GEN_PERF_QUERY_TYPE_PIPELINE;
query->name = "Intel_Raw_Pipeline_Statistics_Query";
- query->n_counters = 0;
- query->counters =
- rzalloc_array(brw, struct brw_perf_query_counter, MAX_STAT_COUNTERS);
/* The order has to match mdapi_pipeline_metrics. */
- brw_perf_query_info_add_basic_stat_reg(query, IA_VERTICES_COUNT,
+ gen_perf_query_info_add_basic_stat_reg(query, IA_VERTICES_COUNT,
"N vertices submitted");
- brw_perf_query_info_add_basic_stat_reg(query, IA_PRIMITIVES_COUNT,
+ gen_perf_query_info_add_basic_stat_reg(query, IA_PRIMITIVES_COUNT,
"N primitives submitted");
- brw_perf_query_info_add_basic_stat_reg(query, VS_INVOCATION_COUNT,
+ gen_perf_query_info_add_basic_stat_reg(query, VS_INVOCATION_COUNT,
"N vertex shader invocations");
- brw_perf_query_info_add_basic_stat_reg(query, GS_INVOCATION_COUNT,
+ gen_perf_query_info_add_basic_stat_reg(query, GS_INVOCATION_COUNT,
"N geometry shader invocations");
- brw_perf_query_info_add_basic_stat_reg(query, GS_PRIMITIVES_COUNT,
+ gen_perf_query_info_add_basic_stat_reg(query, GS_PRIMITIVES_COUNT,
"N geometry shader primitives emitted");
- brw_perf_query_info_add_basic_stat_reg(query, CL_INVOCATION_COUNT,
+ gen_perf_query_info_add_basic_stat_reg(query, CL_INVOCATION_COUNT,
"N primitives entering clipping");
- brw_perf_query_info_add_basic_stat_reg(query, CL_PRIMITIVES_COUNT,
+ gen_perf_query_info_add_basic_stat_reg(query, CL_PRIMITIVES_COUNT,
"N primitives leaving clipping");
if (devinfo->is_haswell || devinfo->gen == 8) {
- brw_perf_query_info_add_stat_reg(query, PS_INVOCATION_COUNT, 1, 4,
+ gen_perf_query_info_add_stat_reg(query, PS_INVOCATION_COUNT, 1, 4,
"N fragment shader invocations",
"N fragment shader invocations");
} else {
- brw_perf_query_info_add_basic_stat_reg(query, PS_INVOCATION_COUNT,
+ gen_perf_query_info_add_basic_stat_reg(query, PS_INVOCATION_COUNT,
"N fragment shader invocations");
}
- brw_perf_query_info_add_basic_stat_reg(query, HS_INVOCATION_COUNT,
+ gen_perf_query_info_add_basic_stat_reg(query, HS_INVOCATION_COUNT,
"N TCS shader invocations");
- brw_perf_query_info_add_basic_stat_reg(query, DS_INVOCATION_COUNT,
+ gen_perf_query_info_add_basic_stat_reg(query, DS_INVOCATION_COUNT,
"N TES shader invocations");
if (devinfo->gen >= 7) {
- brw_perf_query_info_add_basic_stat_reg(query, CS_INVOCATION_COUNT,
+ gen_perf_query_info_add_basic_stat_reg(query, CS_INVOCATION_COUNT,
"N compute shader invocations");
}
generated by cgit v1.2.3 (git 2.39.1) at 2024-06-13 16:57:29 +0000