i965: convert MI_REPORT_PERF_COUNT to genxml

Also make it available from gen7 only to gen7+.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This commit is contained in:
Lionel Landwerlin 2017-06-15 14:47:12 +01:00
parent a26f8d99a6
commit e5743ee014
3 changed files with 34 additions and 28 deletions

View File

@ -646,6 +646,17 @@ struct brw_context
uint32_t width, uint32_t height,
uint32_t tile_x, uint32_t tile_y);
/**
* Emit an MI_REPORT_PERF_COUNT command packet.
*
* This asks the GPU to write a report of the current OA counter values
* into @bo at the given offset and containing the given @report_id
* which we can cross-reference when parsing the report (gen7+ only).
*/
void (*emit_mi_report_perf_count)(struct brw_context *brw,
struct brw_bo *bo,
uint32_t offset_in_bytes,
uint32_t report_id);
} vtbl;
struct brw_bufmgr *bufmgr;

View File

@ -467,29 +467,6 @@ snapshot_statistics_registers(struct brw_context *brw,
}
}
/**
* Emit an MI_REPORT_PERF_COUNT command packet.
*
* This asks the GPU to write a report of the current OA counter
* values into @bo at the given offset and containing the given
* @report_id which we can cross-reference when parsing the report.
*/
static void
emit_mi_report_perf_count(struct brw_context *brw,
struct brw_bo *bo,
uint32_t offset_in_bytes,
uint32_t report_id)
{
assert(offset_in_bytes % 64 == 0);
BEGIN_BATCH(3);
OUT_BATCH(GEN6_MI_REPORT_PERF_COUNT);
OUT_RELOC(bo, I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION,
offset_in_bytes);
OUT_BATCH(report_id);
ADVANCE_BATCH();
}
/**
* Add a query to the global list of "unaccumulated queries."
*
@ -1001,8 +978,8 @@ brw_begin_perf_query(struct gl_context *ctx,
brw->perfquery.next_query_start_report_id += 2;
/* Take a starting OA counter snapshot. */
emit_mi_report_perf_count(brw, obj->oa.bo, 0,
obj->oa.begin_report_id);
brw->vtbl.emit_mi_report_perf_count(brw, obj->oa.bo, 0,
obj->oa.begin_report_id);
++brw->perfquery.n_active_oa_queries;
/* No already-buffered samples can possibly be associated with this query
@ -1081,9 +1058,9 @@ brw_end_perf_query(struct gl_context *ctx,
*/
if (!obj->oa.results_accumulated) {
/* Take an ending OA counter snapshot. */
emit_mi_report_perf_count(brw, obj->oa.bo,
MI_RPC_BO_END_OFFSET_BYTES,
obj->oa.begin_report_id + 1);
brw->vtbl.emit_mi_report_perf_count(brw, obj->oa.bo,
MI_RPC_BO_END_OFFSET_BYTES,
obj->oa.begin_report_id + 1);
}
--brw->perfquery.n_active_oa_queries;

View File

@ -4201,6 +4201,22 @@ static const struct brw_tracked_state genX(vf_topology) = {
/* ---------------------------------------------------------------------- */
#if GEN_GEN >= 7
static void
genX(emit_mi_report_perf_count)(struct brw_context *brw,
struct brw_bo *bo,
uint32_t offset_in_bytes,
uint32_t report_id)
{
brw_batch_emit(brw, GENX(MI_REPORT_PERF_COUNT), mi_rpc) {
mi_rpc.MemoryAddress = instruction_bo(bo, offset_in_bytes);
mi_rpc.ReportID = report_id;
}
}
#endif
/* ---------------------------------------------------------------------- */
void
genX(init_atoms)(struct brw_context *brw)
{
@ -4536,5 +4552,7 @@ genX(init_atoms)(struct brw_context *brw)
STATIC_ASSERT(ARRAY_SIZE(compute_atoms) <= ARRAY_SIZE(brw->compute_atoms));
brw_copy_pipeline_atoms(brw, BRW_COMPUTE_PIPELINE,
compute_atoms, ARRAY_SIZE(compute_atoms));
brw->vtbl.emit_mi_report_perf_count = genX(emit_mi_report_perf_count);
#endif
}