[Mesa-dev] [PATCH v2 5/5] i965: perf: enable GPA query statistics

[Lionel Landwerlin]

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)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin at intel.com>
---
 src/mesa/drivers/dri/i965/brw_performance_query.c  |  1 +
 .../drivers/dri/i965/brw_performance_query_mdapi.c | 64 ++++++++++++++++++++++
 .../drivers/dri/i965/brw_performance_query_mdapi.h |  2 +-
 3 files changed, 66 insertions(+), 1 deletion(-)

diff --git a/src/mesa/drivers/dri/i965/brw_performance_query.c b/src/mesa/drivers/dri/i965/brw_performance_query.c
index aed85417845..a7e3f114d8a 100644
--- a/src/mesa/drivers/dri/i965/brw_performance_query.c
+++ b/src/mesa/drivers/dri/i965/brw_performance_query.c
@@ -2144,6 +2144,7 @@ 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_mdapi.c b/src/mesa/drivers/dri/i965/brw_performance_query_mdapi.c
index cc315ceb4f8..7e254e7a058 100644
--- 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"
 #include "brw_performance_query_mdapi.h"
 
@@ -100,6 +101,20 @@ 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,
@@ -377,3 +392,52 @@ 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;
+}
diff --git a/src/mesa/drivers/dri/i965/brw_performance_query_mdapi.h b/src/mesa/drivers/dri/i965/brw_performance_query_mdapi.h
index e6da3a279e6..42cb5654a4d 100644
--- a/src/mesa/drivers/dri/i965/brw_performance_query_mdapi.h
+++ b/src/mesa/drivers/dri/i965/brw_performance_query_mdapi.h
@@ -34,6 +34,6 @@ int brw_perf_query_get_mdapi_oa_data(struct brw_context *brw,
                                      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_MDAPI_H */
-- 
2.16.3