fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2025-12-31 16:20:13 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

intel/perf: fix OA format selection on MTL

Browse source 
Anything Gfx12.5+ has a different format.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: 90c86fe63e ("intel: add MTL performance metrics")
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22257>
(cherry picked from commit a88aedbfa5)
This commit is contained in:
Lionel Landwerlin 2023-04-03 09:25:44 +03:00 • committed by Dylan Baker
parent 1b6fa2f407
commit 33e4fe697f
3 changed files with 29 additions and 50 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
.pick_status.json
View file

@ -868,7 +868,7 @@
"description": "intel/perf: fix OA format selection on MTL",
"nominated": true,
"nomination_type": 1,
"resolution": 0,
"resolution": 1,
"main_sha": null,
"because_sha": "90c86fe63e94df7719081f86ebee4851ab3fd341"
},

7
src/intel/perf/gen_perf.py
View file

@ -962,12 +962,7 @@ def main():
c("{\n")
c_indent(3)
if gen.chipset == "hsw":
c("struct intel_perf_query_info *query = hsw_query_alloc(perf, %u);\n" % len(counters))
elif gen.chipset.startswith("acm"):
c("struct intel_perf_query_info *query = xehp_query_alloc(perf, %u);\n" % len(counters))
else:
c("struct intel_perf_query_info *query = bdw_query_alloc(perf, %u);\n" % len(counters))
c("struct intel_perf_query_info *query = intel_query_alloc(perf, %u);\n" % len(counters))
c("\n")
c("query->name = \"" + set.name + "\";\n")
c("query->symbol_name = \"" + set.symbol_name + "\";\n")

70
src/intel/perf/intel_perf_setup.h
View file

@ -38,52 +38,36 @@ intel_query_alloc(struct intel_perf_config *perf, int ncounters)
query->n_counters = 0;
query->oa_metrics_set_id = 0; /* determined at runtime, via sysfs */
query->counters = rzalloc_array(query, struct intel_perf_query_counter, ncounters);
return query;
}
static struct intel_perf_query_info *
hsw_query_alloc(struct intel_perf_config *perf, int ncounters)
{
struct intel_perf_query_info *query = intel_query_alloc(perf, ncounters);
query->oa_format = I915_OA_FORMAT_A45_B8_C8;
/* Accumulation buffer offsets... */
query->gpu_time_offset = 0;
query->a_offset = query->gpu_time_offset + 1;
query->b_offset = query->a_offset + 45;
query->c_offset = query->b_offset + 8;
query->perfcnt_offset = query->c_offset + 8;
query->rpstat_offset = query->perfcnt_offset + 2;
return query;
}
if (perf->devinfo.verx10 <= 75) {
query->oa_format = I915_OA_FORMAT_A45_B8_C8;
query->gpu_time_offset = 0;
query->a_offset = query->gpu_time_offset + 1;
query->b_offset = query->a_offset + 45;
query->c_offset = query->b_offset + 8;
query->perfcnt_offset = query->c_offset + 8;
query->rpstat_offset = query->perfcnt_offset + 2;
} else if (perf->devinfo.verx10 <= 120) {
query->oa_format = I915_OA_FORMAT_A32u40_A4u32_B8_C8;
query->gpu_time_offset = 0;
query->gpu_clock_offset = query->gpu_time_offset + 1;
query->a_offset = query->gpu_clock_offset + 1;
query->b_offset = query->a_offset + 36;
query->c_offset = query->b_offset + 8;
query->perfcnt_offset = query->c_offset + 8;
query->rpstat_offset = query->perfcnt_offset + 2;
} else {
query->oa_format = I915_OA_FORMAT_A24u40_A14u32_B8_C8;
query->gpu_time_offset = 0;
query->gpu_clock_offset = query->gpu_time_offset + 1;
query->a_offset = query->gpu_clock_offset + 1;
query->b_offset = query->a_offset + 38;
query->c_offset = query->b_offset + 8;
query->perfcnt_offset = query->c_offset + 8;
query->rpstat_offset = query->perfcnt_offset + 2;
}
static struct intel_perf_query_info *
bdw_query_alloc(struct intel_perf_config *perf, int ncounters)
{
struct intel_perf_query_info *query = intel_query_alloc(perf, ncounters);
query->oa_format = I915_OA_FORMAT_A32u40_A4u32_B8_C8;
/* Accumulation buffer offsets... */
query->gpu_time_offset = 0;
query->gpu_clock_offset = query->gpu_time_offset + 1;
query->a_offset = query->gpu_clock_offset + 1;
query->b_offset = query->a_offset + 36;
query->c_offset = query->b_offset + 8;
query->perfcnt_offset = query->c_offset + 8;
query->rpstat_offset = query->perfcnt_offset + 2;
return query;
}
static struct intel_perf_query_info *
xehp_query_alloc(struct intel_perf_config *perf, int ncounters)
{
struct intel_perf_query_info *query = intel_query_alloc(perf, ncounters);
query->oa_format = I915_OA_FORMAT_A24u40_A14u32_B8_C8;
query->gpu_time_offset = 0;
query->gpu_clock_offset = query->gpu_time_offset + 1;
query->a_offset = query->gpu_clock_offset + 1;
query->b_offset = query->a_offset + 38;
query->c_offset = query->b_offset + 8;
query->perfcnt_offset = query->c_offset + 8;
query->rpstat_offset = query->perfcnt_offset + 2;
return query;
}