i965: perf: enable GPA query statistics

The combinaison of GPA/MDAPI components expects a particular name &
layout for their pipeline statistics query.

v2: Limit the query GPA/MDAPI statistics to gen7->9 (Lionel)

v3: Add curly braces (Ken)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

diff --git a/src/mesa/drivers/dri/i965/brw_performance_query.c b/src/mesa/drivers/dri/i965/brw_performance_query.c
index ece2ff0ab69effda7c661a689bc07540223a9b7a..77d23133ad466f6413a9fba4b24aa0d28c42a26c 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_performance_query.c
+++ b/src/mesa/drivers/dri/i965/brw_performance_query.c
@@ -2165,6 +2165,7 @@ brw_init_perf_query_info(struct gl_context *ctx)
       return brw->perfquery.n_queries;
 
    init_pipeline_statistic_query_registers(brw);
+   brw_perf_query_register_mdapi_statistic_query(brw);
 
    oa_register = get_register_queries_function(devinfo);
 

diff --git a/src/mesa/drivers/dri/i965/brw_performance_query.h b/src/mesa/drivers/dri/i965/brw_performance_query.h
index 20fdbc0473f51786a0365a82e470e4ef119540d1..66b32c0490b5fb275aa0e4e6db0bf58d27e8d7cc 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_performance_query.h
+++ b/src/mesa/drivers/dri/i965/brw_performance_query.h
@@ -222,6 +222,6 @@ int brw_perf_query_get_mdapi_oa_data(struct brw_context *brw,
                                      size_t data_size,
                                      uint8_t *data);
 void brw_perf_query_register_mdapi_oa_query(struct brw_context *brw);
-
+void brw_perf_query_register_mdapi_statistic_query(struct brw_context *brw);
 
 #endif /* BRW_PERFORMANCE_QUERY_H */

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 f98918ba76cbd7fd682d9ca2dea30c1edba562ef..70f69debe987a2f5cfeadb50243ff62bbe1fbc61 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_performance_query_mdapi.c
+++ b/src/mesa/drivers/dri/i965/brw_performance_query_mdapi.c
@@ -21,6 +21,7 @@
  * IN THE SOFTWARE.
  */
 
+#include "brw_defines.h"
 #include "brw_performance_query.h"
 
 /**
@@ -99,6 +100,20 @@ struct mdapi_gen9_metrics {
    uint32_t Reserved4;
 };
 
+struct mdapi_pipeline_metrics {
+   uint64_t IAVertices;
+   uint64_t IAPrimitives;
+   uint64_t VSInvocations;
+   uint64_t GSInvocations;
+   uint64_t GSPrimitives;
+   uint64_t CInvocations;
+   uint64_t CPrimitives;
+   uint64_t PSInvocations;
+   uint64_t HSInvocations;
+   uint64_t DSInvocations;
+   uint64_t CSInvocations;
+};
+
 int
 brw_perf_query_get_mdapi_oa_data(struct brw_context *brw,
                                  struct brw_perf_query_object *obj,
@@ -376,3 +391,54 @@ brw_perf_query_register_mdapi_oa_query(struct brw_context *brw)
       query->c_offset = copy_query->c_offset;
    }
 }
+
+void
+brw_perf_query_register_mdapi_statistic_query(struct brw_context *brw)
+{
+   const struct gen_device_info *devinfo = &brw->screen->devinfo;
+
+   if (!(devinfo->gen >= 7 && devinfo->gen <= 9))
+      return;
+
+   struct brw_perf_query_info *query = brw_perf_query_append_query_info(brw);
+
+   query->kind = PIPELINE_STATS;
+   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,
+                                          "N vertices submitted");
+   brw_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,
+                                          "N vertex shader invocations");
+   brw_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,
+                                          "N geometry shader primitives emitted");
+   brw_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,
+                                          "N primitives leaving clipping");
+   if (devinfo->is_haswell || devinfo->gen == 8) {
+      brw_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,
+                                             "N fragment shader invocations");
+   }
+   brw_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,
+                                          "N TES shader invocations");
+   if (devinfo->gen >= 7) {
+      brw_perf_query_info_add_basic_stat_reg(query, CS_INVOCATION_COUNT,
+                                             "N compute shader invocations");
+   }
+
+   query->data_size = sizeof(uint64_t) * query->n_counters;
+}