From: Nirmoy Das <nirmoy.das@linux.intel.com>
To: Lucas De Marchi <lucas.demarchi@intel.com>,
igt-dev@lists.freedesktop.org
Cc: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Subject: Re: [PATCH i-g-t v3 09/10] tests/intel/xe_drm_fdinfo: Use enum with expected load
Date: Wed, 28 Aug 2024 16:56:45 +0200 [thread overview]
Message-ID: <0c57cfcc-844a-4508-a302-bf6fc99b7670@linux.intel.com> (raw)
In-Reply-To: <20240827165449.1706784-10-lucas.demarchi@intel.com>
On 8/27/2024 6:54 PM, Lucas De Marchi wrote:
> Do not let the execution flag dictate the result check, but rather be
> explicit what's being checked: idle or full load.
>
> Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Nirmoy Das<nirmoy.das@intel.com>
> ---
> tests/intel/xe_drm_fdinfo.c | 45 +++++++++++++++++++++++++++----------
> 1 file changed, 33 insertions(+), 12 deletions(-)
>
> diff --git a/tests/intel/xe_drm_fdinfo.c b/tests/intel/xe_drm_fdinfo.c
> index 23edeea40..d45e45c6b 100644
> --- a/tests/intel/xe_drm_fdinfo.c
> +++ b/tests/intel/xe_drm_fdinfo.c
> @@ -68,6 +68,11 @@ IGT_TEST_DESCRIPTION("Read and verify drm client memory consumption and engine u
> #define TEST_TRAILING_IDLE (1 << 1)
> #define TEST_ISOLATION (1 << 2)
>
> +enum expected_load {
> + EXPECTED_LOAD_IDLE,
> + EXPECTED_LOAD_FULL,
> +};
> +
> struct pceu_cycles {
> uint64_t cycles;
> uint64_t total_cycles;
> @@ -471,7 +476,7 @@ spin_ctx_destroy(int fd, struct spin_ctx *ctx)
>
> static void
> check_results(struct pceu_cycles *s1, struct pceu_cycles *s2,
> - int class, unsigned int flags)
> + int class, enum expected_load expected_load)
> {
> double percent;
> u64 den, num;
> @@ -487,11 +492,18 @@ check_results(struct pceu_cycles *s1, struct pceu_cycles *s2,
>
> igt_debug("%s: percent: %f\n", engine_map[class], percent);
>
> - if (flags & TEST_BUSY) {
> + switch (expected_load) {
> + case EXPECTED_LOAD_IDLE:
> + igt_assert_eq(num, 0);
> + break;
> + case EXPECTED_LOAD_FULL:
> + /*
> + * We are still relying on CPU sleep time and there could be
> + * some imprecision when calculating the load. Use a 5% margin.
> + */
> igt_assert_lt_double(95.0, percent);
> igt_assert_lt_double(percent, 105.0);
> - } else {
> - igt_assert_eq(num, 0);
> + break;
> }
> }
>
> @@ -501,6 +513,7 @@ single(int fd, struct drm_xe_engine_class_instance *hwe, unsigned int flags)
> struct pceu_cycles pceu1[2][DRM_XE_ENGINE_CLASS_COMPUTE + 1];
> struct pceu_cycles pceu2[2][DRM_XE_ENGINE_CLASS_COMPUTE + 1];
> struct spin_ctx *ctx = NULL;
> + enum expected_load expected_load;
> uint32_t vm;
> int new_fd;
>
> @@ -525,10 +538,16 @@ single(int fd, struct drm_xe_engine_class_instance *hwe, unsigned int flags)
> if (flags & TEST_ISOLATION)
> read_engine_cycles(new_fd, pceu2[1]);
>
> - check_results(pceu1[0], pceu2[0], hwe->engine_class, flags);
> + expected_load = flags & TEST_BUSY ?
> + EXPECTED_LOAD_FULL : EXPECTED_LOAD_IDLE;
> + check_results(pceu1[0], pceu2[0], hwe->engine_class, expected_load);
>
> if (flags & TEST_ISOLATION) {
> - check_results(pceu1[1], pceu2[1], hwe->engine_class, 0);
> + /*
> + * Load from one client shouldn't spill on another,
> + * so check for idle
> + */
> + check_results(pceu1[1], pceu2[1], hwe->engine_class, EXPECTED_LOAD_IDLE);
> close(new_fd);
> }
>
> @@ -557,9 +576,10 @@ busy_check_all(int fd, struct drm_xe_engine_class_instance *hwe)
> read_engine_cycles(fd, pceu2);
>
> xe_for_each_engine_class(class) {
> - bool idle = hwe->engine_class != class;
> + enum expected_load expected_load = hwe->engine_class != class ?
> + EXPECTED_LOAD_IDLE : EXPECTED_LOAD_FULL;
>
> - check_results(pceu1, pceu2, class, idle ? 0 : TEST_BUSY);
> + check_results(pceu1, pceu2, class, expected_load);
> }
>
> spin_sync_end(fd, ctx);
> @@ -590,7 +610,7 @@ single_destroy_queue(int fd, struct drm_xe_engine_class_instance *hwe)
>
> xe_vm_destroy(fd, vm);
>
> - check_results(pceu1, pceu2, hwe->engine_class, TEST_BUSY);
> + check_results(pceu1, pceu2, hwe->engine_class, EXPECTED_LOAD_FULL);
> }
>
> static void
> @@ -623,12 +643,13 @@ most_busy_check_all(int fd, struct drm_xe_engine_class_instance *hwe)
> read_engine_cycles(fd, pceu2);
>
> xe_for_each_engine_class(class) {
> - bool idle = hwe->engine_class == class;
> + enum expected_load expected_load = hwe->engine_class == class ?
> + EXPECTED_LOAD_IDLE : EXPECTED_LOAD_FULL;
>
> if (!ctx[class])
> continue;
>
> - check_results(pceu1, pceu2, class, idle ? 0 : TEST_BUSY);
> + check_results(pceu1, pceu2, class, expected_load);
> spin_sync_end(fd, ctx[class]);
> spin_ctx_destroy(fd, ctx[class]);
> }
> @@ -668,7 +689,7 @@ all_busy_check_all(int fd)
> if (!ctx[class])
> continue;
>
> - check_results(pceu1, pceu2, class, TEST_BUSY);
> + check_results(pceu1, pceu2, class, EXPECTED_LOAD_FULL);
> spin_sync_end(fd, ctx[class]);
> spin_ctx_destroy(fd, ctx[class]);
> }
next prev parent reply other threads:[~2024-08-28 14:56 UTC|newest]
Thread overview: 26+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-08-27 16:54 [PATCH i-g-t v3 00/10] tests/intel/xe_drm_fdinfo: Fix noise and improve Lucas De Marchi
2024-08-27 16:54 ` [PATCH i-g-t v3 01/10] tests/intel/xe_drm_fdinfo: Extend mercy to the upper end Lucas De Marchi
2024-08-28 14:44 ` Nirmoy Das
2024-08-28 14:46 ` Nirmoy Das
2024-08-27 16:54 ` [PATCH i-g-t v3 02/10] tests/intel/xe_drm_fdinfo: Print timestamp for debug Lucas De Marchi
2024-08-28 14:45 ` Nirmoy Das
2024-08-27 16:54 ` [PATCH i-g-t v3 03/10] lib/igt_core: Add igt_assert_lt_double() Lucas De Marchi
2024-08-28 14:46 ` Nirmoy Das
2024-08-27 16:54 ` [PATCH i-g-t v3 04/10] tests/intel/xe_drm_fdinfo: Use igt_assert_lt_double() Lucas De Marchi
2024-08-28 14:48 ` Nirmoy Das
2024-08-27 16:54 ` [PATCH i-g-t v3 05/10] tests/intel/xe_drm_fdinfo: Be strict on == 0 comparison Lucas De Marchi
2024-08-28 14:51 ` Nirmoy Das
2024-08-27 16:54 ` [PATCH i-g-t v3 06/10] tests/intel/xe_drm_fdinfo: Use usec for batch duration Lucas De Marchi
2024-08-28 14:51 ` Nirmoy Das
2024-08-27 16:54 ` [PATCH i-g-t v3 07/10] tests/intel/xe_drm_fdinfo: Half the execution time Lucas De Marchi
2024-08-28 14:52 ` Nirmoy Das
2024-08-27 16:54 ` [PATCH i-g-t v3 08/10] tests/intel/xe_drm_fdinfo: Remove unused flags Lucas De Marchi
2024-08-28 15:00 ` Nirmoy Das
2024-08-27 16:54 ` [PATCH i-g-t v3 09/10] tests/intel/xe_drm_fdinfo: Use enum with expected load Lucas De Marchi
2024-08-28 14:56 ` Nirmoy Das [this message]
2024-08-27 16:54 ` [PATCH i-g-t v3 10/10] tests/intel/xe_drm_fdinfo: Rename and reorder tests Lucas De Marchi
2024-08-28 15:08 ` Nirmoy Das
2024-08-28 15:14 ` Nirmoy Das
2024-08-27 18:27 ` ✓ CI.xeBAT: success for tests/intel/xe_drm_fdinfo: Fix noise and improve (rev2) Patchwork
2024-08-27 18:38 ` ✗ Fi.CI.BAT: failure " Patchwork
2024-08-28 4:05 ` ✗ CI.xeFULL: " Patchwork
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=0c57cfcc-844a-4508-a302-bf6fc99b7670@linux.intel.com \
--to=nirmoy.das@linux.intel.com \
--cc=igt-dev@lists.freedesktop.org \
--cc=lucas.demarchi@intel.com \
--cc=umesh.nerlige.ramappa@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox