[Intel-gfx] [PATCH i-g-t] tools/intel_gpu_top: Fixup imc event parsing

Chris Wilson chris at chris-wilson.co.uk
Thu Nov 26 15:48:56 UTC 2020


After combining rapl_parse and imc_parse into a single pmu_parse, I left
the "energy-" prefixes used by rapl (but not imc) in place. Lift the
prefix to rapl_open() so that pmu_parse() does work for both rapl and
imc!

Reported-by: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
Fixes: d0b71b967ccd ("tools/intel_gpu_top: Consolidate imc to use pmu_counter")
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
---
 tools/intel_gpu_top.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/tools/intel_gpu_top.c b/tools/intel_gpu_top.c
index 5d42a2fad..3ff9236ed 100644
--- a/tools/intel_gpu_top.c
+++ b/tools/intel_gpu_top.c
@@ -151,13 +151,13 @@ static int pmu_parse(struct pmu_counter *pmu, const char *path, const char *str)
 
 	result &= igt_sysfs_scanf(dir, "type", "%"PRIu64, &pmu->type) == 1;
 
-	snprintf(buf, sizeof(buf) - 1, "events/energy-%s", str);
+	snprintf(buf, sizeof(buf) - 1, "events/%s", str);
 	result &= igt_sysfs_scanf(dir, buf, "event=%"PRIx64, &pmu->config) == 1;
 
-	snprintf(buf, sizeof(buf) - 1, "events/energy-%s.scale", str);
+	snprintf(buf, sizeof(buf) - 1, "events/%s.scale", str);
 	result &= igt_sysfs_scanf(dir, buf, "%lf", &pmu->scale) == 1;
 
-	snprintf(buf, sizeof(buf) - 1, "events/energy-%s.unit", str);
+	snprintf(buf, sizeof(buf) - 1, "events/%s.unit", str);
 	result &= igt_sysfs_scanf(dir, buf, "%127s", buf) == 1;
 	pmu->units = strdup(buf);
 
@@ -217,13 +217,13 @@ rapl_open(struct pmu_counter *pmu,
 static void gpu_power_open(struct pmu_counter *pmu,
 			   struct engines *engines)
 {
-	rapl_open(pmu, "gpu", engines);
+	rapl_open(pmu, "energy-gpu", engines);
 }
 
 static void pkg_power_open(struct pmu_counter *pmu,
 			   struct engines *engines)
 {
-	rapl_open(pmu, "pkg", engines);
+	rapl_open(pmu, "energy-pkg", engines);
 }
 
 static uint64_t
-- 
2.29.2



More information about the Intel-gfx mailing list