[Intel-gfx] [PATCH 1/2] drm/i915/selftests: Move gpu energy measurement into its own little lib

[Intel-gfx] [PATCH 1/2] drm/i915/selftests: Move gpu energy measurement into its own little lib

[Chris Wilson]

Move the handy utility to measure the GPU energy consumption using RAPL
msr into a common lib so that it can be reused easily.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/Makefile            |  3 ++-
 drivers/gpu/drm/i915/gt/selftest_rc6.c   | 25 +++++------------------
 drivers/gpu/drm/i915/selftests/librapl.c | 26 ++++++++++++++++++++++++
 drivers/gpu/drm/i915/selftests/librapl.h | 13 ++++++++++++
 4 files changed, 46 insertions(+), 21 deletions(-)
 create mode 100644 drivers/gpu/drm/i915/selftests/librapl.c
 create mode 100644 drivers/gpu/drm/i915/selftests/librapl.h

diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index fa8d78e40dad..dc80a1d4dc9b 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -258,7 +258,8 @@ i915-$(CONFIG_DRM_I915_SELFTEST) += \
 	selftests/igt_live_test.o \
 	selftests/igt_mmap.o \
 	selftests/igt_reset.o \
-	selftests/igt_spinner.o
+	selftests/igt_spinner.o \
+	selftests/librapl.o
 
 # virtual gpu code
 i915-y += i915_vgpu.o
diff --git a/drivers/gpu/drm/i915/gt/selftest_rc6.c b/drivers/gpu/drm/i915/gt/selftest_rc6.c
index 08c3dbd41b12..5d59589d8a54 100644
--- a/drivers/gpu/drm/i915/gt/selftest_rc6.c
+++ b/drivers/gpu/drm/i915/gt/selftest_rc6.c
@@ -11,22 +11,7 @@
 #include "selftest_rc6.h"
 
 #include "selftests/i915_random.h"
-
-static u64 energy_uJ(struct intel_rc6 *rc6)
-{
-	unsigned long long power;
-	u32 units;
-
-	if (rdmsrl_safe(MSR_RAPL_POWER_UNIT, &power))
-		return 0;
-
-	units = (power & 0x1f00) >> 8;
-
-	if (rdmsrl_safe(MSR_PP1_ENERGY_STATUS, &power))
-		return 0;
-
-	return (1000000 * power) >> units; /* convert to uJ */
-}
+#include "selftests/librapl.h"
 
 static u64 rc6_residency(struct intel_rc6 *rc6)
 {
@@ -74,9 +59,9 @@ int live_rc6_manual(void *arg)
 	res[0] = rc6_residency(rc6);
 
 	dt = ktime_get();
-	rc0_power = energy_uJ(rc6);
+	rc0_power = st_energy_uJ();
 	msleep(250);
-	rc0_power = energy_uJ(rc6) - rc0_power;
+	rc0_power = st_energy_uJ() - rc0_power;
 	dt = ktime_sub(ktime_get(), dt);
 	res[1] = rc6_residency(rc6);
 	if ((res[1] - res[0]) >> 10) {
@@ -99,9 +84,9 @@ int live_rc6_manual(void *arg)
 	res[0] = rc6_residency(rc6);
 	intel_uncore_forcewake_flush(rc6_to_uncore(rc6), FORCEWAKE_ALL);
 	dt = ktime_get();
-	rc6_power = energy_uJ(rc6);
+	rc6_power = st_energy_uJ();
 	msleep(100);
-	rc6_power = energy_uJ(rc6) - rc6_power;
+	rc6_power = st_energy_uJ() - rc6_power;
 	dt = ktime_sub(ktime_get(), dt);
 	res[1] = rc6_residency(rc6);
 	if (res[1] == res[0]) {
diff --git a/drivers/gpu/drm/i915/selftests/librapl.c b/drivers/gpu/drm/i915/selftests/librapl.c
new file mode 100644
index 000000000000..2a0c9aae8591
--- /dev/null
+++ b/drivers/gpu/drm/i915/selftests/librapl.c
@@ -0,0 +1,26 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright © 2020 Intel Corporation
+ */
+
+#include <asm/msr.h>
+
+#include "librapl.h"
+
+u64 st_energy_uJ(void)
+{
+	unsigned long long power;
+	u32 units;
+
+	if (rdmsrl_safe(MSR_RAPL_POWER_UNIT, &power))
+		return 0;
+
+	units = (power & 0x1f00) >> 8;
+
+	if (rdmsrl_safe(MSR_PP1_ENERGY_STATUS, &power))
+		return 0;
+
+	return (1000000 * power) >> units; /* convert to uJ */
+}
+
+
diff --git a/drivers/gpu/drm/i915/selftests/librapl.h b/drivers/gpu/drm/i915/selftests/librapl.h
new file mode 100644
index 000000000000..e0d6334b4367
--- /dev/null
+++ b/drivers/gpu/drm/i915/selftests/librapl.h
@@ -0,0 +1,13 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright © 2020 Intel Corporation
+ */
+
+#ifndef SELFTEST_LIBRAPL_H
+#define SELFTEST_LIBRAPL_H
+
+#include <linux/types.h>
+
+u64 st_energy_uJ(void);
+
+#endif /* SELFTEST_LIBRAPL_H */
-- 
2.20.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

[Intel-gfx] [PATCH 2/2] drm/i915/selftests: Check power consumption at min/max frequencies

[Chris Wilson]

A basic premise of RPS is that at lower frequencies, not only do we run
slower, but we save power compared to higher frequencies. For example,
when idle, we set the minimum frequency just in case there is some
residual current. Since the power curve should be a physical
relationship, if we find no power saving it's likely that we've broken
our frequency handling, so test!

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Andi Shyti <andi.shyti@intel.com>
---
 drivers/gpu/drm/i915/gt/selftest_gt_pm.c |   1 +
 drivers/gpu/drm/i915/gt/selftest_rps.c   | 129 +++++++++++++++++++++++
 drivers/gpu/drm/i915/gt/selftest_rps.h   |   1 +
 3 files changed, 131 insertions(+)

diff --git a/drivers/gpu/drm/i915/gt/selftest_gt_pm.c b/drivers/gpu/drm/i915/gt/selftest_gt_pm.c
index c50bb502fe03..0141c334f2ac 100644
--- a/drivers/gpu/drm/i915/gt/selftest_gt_pm.c
+++ b/drivers/gpu/drm/i915/gt/selftest_gt_pm.c
@@ -54,6 +54,7 @@ int intel_gt_pm_live_selftests(struct drm_i915_private *i915)
 	static const struct i915_subtest tests[] = {
 		SUBTEST(live_rc6_manual),
 		SUBTEST(live_rps_interrupt),
+		SUBTEST(live_rps_power),
 		SUBTEST(live_gt_resume),
 	};
 
diff --git a/drivers/gpu/drm/i915/gt/selftest_rps.c b/drivers/gpu/drm/i915/gt/selftest_rps.c
index 5ad5bee6bdba..3c783f9a04d5 100644
--- a/drivers/gpu/drm/i915/gt/selftest_rps.c
+++ b/drivers/gpu/drm/i915/gt/selftest_rps.c
@@ -3,6 +3,8 @@
  * Copyright © 2020 Intel Corporation
  */
 
+#include <linux/sort.h>
+
 #include "intel_engine_pm.h"
 #include "intel_gt_clk.h"
 #include "intel_gt_pm.h"
@@ -10,6 +12,7 @@
 #include "selftest_rps.h"
 #include "selftests/igt_flush_test.h"
 #include "selftests/igt_spinner.h"
+#include "selftests/librapl.h"
 
 static void dummy_rps_work(struct work_struct *wrk)
 {
@@ -224,3 +227,129 @@ int live_rps_interrupt(void *arg)
 
 	return err;
 }
+
+static u64 __measure_power(int duration_ms)
+{
+	u64 nrg, dt;
+
+	dt = -ktime_get();
+	nrg = -st_energy_uJ();
+	msleep(5);
+	nrg += st_energy_uJ();
+	dt += ktime_get();
+
+	return div64_u64(1000 * 1000 * nrg, dt);
+}
+
+static int cmp_u64(const void *A, const void *B)
+{
+	const u64 *a = A, *b = B;
+
+	if (a < b)
+		return -1;
+	else if (a > b)
+		return 1;
+	else
+		return 0;
+}
+
+static u64 measure_power_at(struct intel_rps *rps, int freq)
+{
+	u64 x[5];
+	int i;
+
+	mutex_lock(&rps->lock);
+	GEM_BUG_ON(!rps->active);
+	intel_rps_set(rps, freq);
+	mutex_unlock(&rps->lock);
+
+	msleep(20); /* more than enough time to stabilise! */
+
+	for (i = 0; i < 5; i++)
+		x[i] = __measure_power(5);
+
+	/* A simple triangle filter for better result stability */
+	sort(x, 5, sizeof(*x), cmp_u64, NULL);
+	return div_u64(x[1] + 2 * x[2] + x[3], 4);
+}
+
+int live_rps_power(void *arg)
+{
+	struct intel_gt *gt = arg;
+	struct intel_rps *rps = &gt->rps;
+	void (*saved_work)(struct work_struct *wrk);
+	struct intel_engine_cs *engine;
+	enum intel_engine_id id;
+	struct igt_spinner spin;
+	int err = 0;
+
+	/*
+	 * Our fundamental assumption is that running at lower frequency
+	 * actually saves power. Let's see if our RAPL measurement support
+	 * that theory.
+	 */
+
+	if (!rps->enabled || rps->max_freq <= rps->min_freq)
+		return 0;
+
+	if (igt_spinner_init(&spin, gt))
+		return -ENOMEM;
+
+	intel_gt_pm_wait_for_idle(gt);
+	saved_work = rps->work.func;
+	rps->work.func = dummy_rps_work;
+
+	for_each_engine(engine, gt, id) {
+		struct i915_request *rq;
+		u64 min, max;
+
+		if (!intel_engine_can_store_dword(engine))
+			continue;
+
+		rq = igt_spinner_create_request(&spin,
+						engine->kernel_context,
+						MI_NOOP);
+		if (IS_ERR(rq)) {
+			err = PTR_ERR(rq);
+			break;
+		}
+
+		i915_request_add(rq);
+
+		if (!igt_wait_for_spinner(&spin, rq)) {
+			pr_err("%s: RPS spinner did not start\n",
+			       engine->name);
+			intel_gt_set_wedged(engine->gt);
+			err = -EIO;
+			break;
+		}
+
+		max = measure_power_at(rps, rps->max_freq);
+		min = measure_power_at(rps, rps->min_freq);
+
+		igt_spinner_end(&spin);
+
+		pr_info("%s: min:%llumW @ %uMHz, max:%llumW @ %uMHz\n",
+			engine->name,
+			min, intel_gpu_freq(rps, rps->min_freq),
+			max, intel_gpu_freq(rps, rps->max_freq));
+		if (min >= max) {
+			pr_err("%s: did not conserve power when setting lower frequency!\n",
+			       engine->name);
+			err = -EINVAL;
+			break;
+		}
+
+		if (igt_flush_test(gt->i915)) {
+			err = -EIO;
+			break;
+		}
+	}
+
+	igt_spinner_fini(&spin);
+
+	intel_gt_pm_wait_for_idle(gt);
+	rps->work.func = saved_work;
+
+	return err;
+}
diff --git a/drivers/gpu/drm/i915/gt/selftest_rps.h b/drivers/gpu/drm/i915/gt/selftest_rps.h
index abba66420996..cad515a7f0e5 100644
--- a/drivers/gpu/drm/i915/gt/selftest_rps.h
+++ b/drivers/gpu/drm/i915/gt/selftest_rps.h
@@ -7,5 +7,6 @@
 #define SELFTEST_RPS_H
 
 int live_rps_interrupt(void *arg);
+int live_rps_power(void *arg);
 
 #endif /* SELFTEST_RPS_H */
-- 
2.20.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

[Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for series starting with [1/2] drm/i915/selftests: Move gpu energy measurement into its own little lib

[Patchwork]

== Series Details ==

Series: series starting with [1/2] drm/i915/selftests: Move gpu energy measurement into its own little lib
URL   : https://patchwork.freedesktop.org/series/76044/
State : warning

== Summary ==

$ dim checkpatch origin/drm-tip
04da6c2d287e drm/i915/selftests: Move gpu energy measurement into its own little lib
-:79: WARNING:FILE_PATH_CHANGES: added, moved or deleted file(s), does MAINTAINERS need updating?
#79: 
new file mode 100644

-:109: CHECK:LINE_SPACING: Please don't use multiple blank lines
#109: FILE: drivers/gpu/drm/i915/selftests/librapl.c:26:
+
+

total: 0 errors, 1 warnings, 1 checks, 93 lines checked
05b16cdfef39 drm/i915/selftests: Check power consumption at min/max frequencies
-:61: WARNING:MSLEEP: msleep < 20ms can sleep for up to 20ms; see Documentation/timers/timers-howto.rst
#61: FILE: drivers/gpu/drm/i915/gt/selftest_rps.c:234:
+	msleep(5);

total: 0 errors, 1 warnings, 0 checks, 157 lines checked

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

[Intel-gfx] ✗ Fi.CI.DOCS: warning for series starting with [1/2] drm/i915/selftests: Move gpu energy measurement into its own little lib

[Patchwork]

== Series Details ==

Series: series starting with [1/2] drm/i915/selftests: Move gpu energy measurement into its own little lib
URL   : https://patchwork.freedesktop.org/series/76044/
State : warning

== Summary ==

$ make htmldocs 2>&1 > /dev/null | grep i915
/home/cidrm/kernel/Documentation/gpu/i915.rst:610: WARNING: duplicate label gpu/i915:layout, other instance in /home/cidrm/kernel/Documentation/gpu/i915.rst

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

[Intel-gfx] ✗ Fi.CI.BAT: failure for series starting with [1/2] drm/i915/selftests: Move gpu energy measurement into its own little lib

[Patchwork]

== Series Details ==

Series: series starting with [1/2] drm/i915/selftests: Move gpu energy measurement into its own little lib
URL   : https://patchwork.freedesktop.org/series/76044/
State : failure

== Summary ==

CI Bug Log - changes from CI_DRM_8311 -> Patchwork_17334
====================================================

Summary
-------

  **FAILURE**

  Serious unknown changes coming with Patchwork_17334 absolutely need to be
  verified manually.
  
  If you think the reported changes have nothing to do with the changes
  introduced in Patchwork_17334, please notify your bug team to allow them
  to document this new failure mode, which will reduce false positives in CI.

  External URL: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17334/index.html

Possible new issues
-------------------

  Here are the unknown changes that may have been introduced in Patchwork_17334:

### IGT changes ###

#### Possible regressions ####

  * igt@i915_selftest@live@gt_pm:
    - fi-bsw-n3050:       [PASS][1] -> [DMESG-FAIL][2]
   [1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8311/fi-bsw-n3050/igt@i915_selftest@live@gt_pm.html
   [2]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17334/fi-bsw-n3050/igt@i915_selftest@live@gt_pm.html
    - fi-bsw-nick:        [PASS][3] -> [DMESG-FAIL][4]
   [3]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8311/fi-bsw-nick/igt@i915_selftest@live@gt_pm.html
   [4]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17334/fi-bsw-nick/igt@i915_selftest@live@gt_pm.html
    - fi-byt-n2820:       [PASS][5] -> [DMESG-FAIL][6]
   [5]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8311/fi-byt-n2820/igt@i915_selftest@live@gt_pm.html
   [6]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17334/fi-byt-n2820/igt@i915_selftest@live@gt_pm.html
    - fi-bsw-kefka:       [PASS][7] -> [DMESG-FAIL][8]
   [7]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8311/fi-bsw-kefka/igt@i915_selftest@live@gt_pm.html
   [8]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17334/fi-bsw-kefka/igt@i915_selftest@live@gt_pm.html
    - fi-kbl-soraka:      [PASS][9] -> [DMESG-FAIL][10]
   [9]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8311/fi-kbl-soraka/igt@i915_selftest@live@gt_pm.html
   [10]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17334/fi-kbl-soraka/igt@i915_selftest@live@gt_pm.html
    - fi-byt-j1900:       [PASS][11] -> [DMESG-FAIL][12]
   [11]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8311/fi-byt-j1900/igt@i915_selftest@live@gt_pm.html
   [12]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17334/fi-byt-j1900/igt@i915_selftest@live@gt_pm.html

  
Known issues
------------

  Here are the changes found in Patchwork_17334 that come from known issues:

### IGT changes ###

#### Issues hit ####

  * igt@i915_selftest@live@late_gt_pm:
    - fi-bsw-kefka:       [PASS][13] -> [INCOMPLETE][14] ([i915#1382])
   [13]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8311/fi-bsw-kefka/igt@i915_selftest@live@late_gt_pm.html
   [14]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17334/fi-bsw-kefka/igt@i915_selftest@live@late_gt_pm.html

  
#### Warnings ####

  * igt@amdgpu/amd_prime@i915-to-amd:
    - fi-bwr-2160:        [FAIL][15] ([i915#489]) -> [SKIP][16] ([fdo#109271])
   [15]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_8311/fi-bwr-2160/igt@amdgpu/amd_prime@i915-to-amd.html
   [16]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17334/fi-bwr-2160/igt@amdgpu/amd_prime@i915-to-amd.html

  
  [fdo#109271]: https://bugs.freedesktop.org/show_bug.cgi?id=109271
  [i915#1382]: https://gitlab.freedesktop.org/drm/intel/issues/1382
  [i915#489]: https://gitlab.freedesktop.org/drm/intel/issues/489


Participating hosts (51 -> 44)
------------------------------

  Missing    (7): fi-bxt-dsi fi-hsw-4200u fi-byt-squawks fi-bsw-cyan fi-ctg-p8600 fi-byt-clapper fi-bdw-samus 


Build changes
-------------

  * CI: CI-20190529 -> None
  * Linux: CI_DRM_8311 -> Patchwork_17334

  CI-20190529: 20190529
  CI_DRM_8311: 19367bb5e65eaf0719597b3ff244fd1c2ea12bda @ git://anongit.freedesktop.org/gfx-ci/linux
  IGT_5593: 1c658f5e46598ae93345177d4981ef54704daec6 @ git://anongit.freedesktop.org/xorg/app/intel-gpu-tools
  Patchwork_17334: 05b16cdfef39c7101e890d23a9af07105c9210ba @ git://anongit.freedesktop.org/gfx-ci/linux


== Linux commits ==

05b16cdfef39 drm/i915/selftests: Check power consumption at min/max frequencies
04da6c2d287e drm/i915/selftests: Move gpu energy measurement into its own little lib

== Logs ==

For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_17334/index.html
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [Intel-gfx] [PATCH 2/2] drm/i915/selftests: Check power consumption at min/max frequencies

[Andi Shyti]

Hi Chris,

> +static u64 __measure_power(int duration_ms)
> +{
> +	u64 nrg, dt;
> +
> +	dt = -ktime_get();
> +	nrg = -st_energy_uJ();
> +	msleep(5);

usleep_range?

Reviewed-by: Andi Shyti <andi.shyti@intel.com>

Thanks,
Andi
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx