From: Chris Wilson <chris@chris-wilson.co.uk> To: Tvrtko Ursulin <tursulin@ursulin.net>, igt-dev@lists.freedesktop.org Cc: Intel-gfx@lists.freedesktop.org Subject: Re: [igt-dev] [PATCH i-g-t v2] tests/perf_pmu: Improve accuracy by waiting on spinner to start Date: Thu, 15 Mar 2018 16:01:47 +0000 [thread overview] Message-ID: <152112970795.25315.3159070975692005147@mail.alporthouse.com> (raw) In-Reply-To: <20180315154600.13728-1-tvrtko.ursulin@linux.intel.com> Quoting Tvrtko Ursulin (2018-03-15 15:46:00) > static void > -__submit_spin_batch(int gem_fd, > - struct drm_i915_gem_exec_object2 *obj, > +__submit_spin_batch(int gem_fd, igt_spin_t *spin, > const struct intel_execution_engine2 *e) > { > - struct drm_i915_gem_execbuffer2 eb = { > - .buffer_count = 1, > - .buffers_ptr = to_user_pointer(obj), > - .flags = e2ring(gem_fd, e), > - }; > + struct drm_i915_gem_execbuffer2 eb = spin->execbuf; > + > + eb.flags &= ~(0x3f | I915_EXEC_BSD_MASK); > + eb.flags |= e2ring(gem_fd, e); I'm dubious about keeping spin->execbuf for precisely this reason. Almost all the time I want to specify a different execution, and not use the same context, random fences, etc. However, it does give a convenient way to get buffers_count and buffers_ptr, but that is all that is valid (imo) or at least should be judiciously copied from rather than wholesale. eb = { .buffers_ptr = spin->execbuf.buffers_ptr, .buffer_count = spin->execbuf.buffer_count, .flags = e2ring(gem_fd, e) | I915_EXEC_NORELOC, }; > @@ -1545,24 +1575,30 @@ accuracy(int gem_fd, const struct intel_execution_engine2 *e, > > igt_nsec_elapsed(&test_start); > do { > - unsigned int target_idle_us, t_busy; > + unsigned int target_idle_us; > + struct timespec start = { }; > + unsigned long prep_delay_ns; > > /* Restart the spinbatch. */ > + igt_nsec_elapsed(&start); > __rearm_spin_batch(spin); > - __submit_spin_batch(gem_fd, &obj, e); > + __submit_spin_batch(gem_fd, spin, e); > > - /* > - * Note that the submission may be delayed to a > - * tasklet (ksoftirqd) which cannot run until we > - * sleep as we hog the cpu (we are RT). > - */ > + /* Wait for batch to start executing. */ > + __spin_wait(gem_fd, spin); > + prep_delay_ns = igt_nsec_elapsed(&start); > > - t_busy = measured_usleep(busy_us); > + /* PWM busy sleep. */ > + memset(&start, 0, sizeof(start)); > + igt_nsec_elapsed(&start); Can just keep using start, it's already has the current time from calculating prep_delay_ns. > + measured_usleep(busy_us); > igt_spin_batch_end(spin); > - gem_sync(gem_fd, obj.handle); > + gem_sync(gem_fd, spin->handle); > > - total_busy_ns += t_busy; > + total_busy_ns += igt_nsec_elapsed(&start); > + total_idle_ns += prep_delay_ns; Ok. Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> -Chris _______________________________________________ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/intel-gfx
WARNING: multiple messages have this Message-ID (diff)
From: Chris Wilson <chris@chris-wilson.co.uk> To: Tvrtko Ursulin <tursulin@ursulin.net>, igt-dev@lists.freedesktop.org Cc: Intel-gfx@lists.freedesktop.org, Tvrtko Ursulin <tvrtko.ursulin@intel.com> Subject: Re: [igt-dev] [PATCH i-g-t v2] tests/perf_pmu: Improve accuracy by waiting on spinner to start Date: Thu, 15 Mar 2018 16:01:47 +0000 [thread overview] Message-ID: <152112970795.25315.3159070975692005147@mail.alporthouse.com> (raw) In-Reply-To: <20180315154600.13728-1-tvrtko.ursulin@linux.intel.com> Quoting Tvrtko Ursulin (2018-03-15 15:46:00) > static void > -__submit_spin_batch(int gem_fd, > - struct drm_i915_gem_exec_object2 *obj, > +__submit_spin_batch(int gem_fd, igt_spin_t *spin, > const struct intel_execution_engine2 *e) > { > - struct drm_i915_gem_execbuffer2 eb = { > - .buffer_count = 1, > - .buffers_ptr = to_user_pointer(obj), > - .flags = e2ring(gem_fd, e), > - }; > + struct drm_i915_gem_execbuffer2 eb = spin->execbuf; > + > + eb.flags &= ~(0x3f | I915_EXEC_BSD_MASK); > + eb.flags |= e2ring(gem_fd, e); I'm dubious about keeping spin->execbuf for precisely this reason. Almost all the time I want to specify a different execution, and not use the same context, random fences, etc. However, it does give a convenient way to get buffers_count and buffers_ptr, but that is all that is valid (imo) or at least should be judiciously copied from rather than wholesale. eb = { .buffers_ptr = spin->execbuf.buffers_ptr, .buffer_count = spin->execbuf.buffer_count, .flags = e2ring(gem_fd, e) | I915_EXEC_NORELOC, }; > @@ -1545,24 +1575,30 @@ accuracy(int gem_fd, const struct intel_execution_engine2 *e, > > igt_nsec_elapsed(&test_start); > do { > - unsigned int target_idle_us, t_busy; > + unsigned int target_idle_us; > + struct timespec start = { }; > + unsigned long prep_delay_ns; > > /* Restart the spinbatch. */ > + igt_nsec_elapsed(&start); > __rearm_spin_batch(spin); > - __submit_spin_batch(gem_fd, &obj, e); > + __submit_spin_batch(gem_fd, spin, e); > > - /* > - * Note that the submission may be delayed to a > - * tasklet (ksoftirqd) which cannot run until we > - * sleep as we hog the cpu (we are RT). > - */ > + /* Wait for batch to start executing. */ > + __spin_wait(gem_fd, spin); > + prep_delay_ns = igt_nsec_elapsed(&start); > > - t_busy = measured_usleep(busy_us); > + /* PWM busy sleep. */ > + memset(&start, 0, sizeof(start)); > + igt_nsec_elapsed(&start); Can just keep using start, it's already has the current time from calculating prep_delay_ns. > + measured_usleep(busy_us); > igt_spin_batch_end(spin); > - gem_sync(gem_fd, obj.handle); > + gem_sync(gem_fd, spin->handle); > > - total_busy_ns += t_busy; > + total_busy_ns += igt_nsec_elapsed(&start); > + total_idle_ns += prep_delay_ns; Ok. Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> -Chris _______________________________________________ igt-dev mailing list igt-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/igt-dev
next prev parent reply other threads:[~2018-03-15 16:01 UTC|newest] Thread overview: 53+ messages / expand[flat|nested] mbox.gz Atom feed top 2018-03-15 12:56 [PATCH i-g-t] tests/perf_pmu: Improve accuracy by waiting on spinner to start Tvrtko Ursulin 2018-03-15 12:56 ` [igt-dev] " Tvrtko Ursulin 2018-03-15 13:14 ` Chris Wilson 2018-03-15 13:14 ` Chris Wilson 2018-03-15 13:36 ` Tvrtko Ursulin 2018-03-15 13:36 ` Tvrtko Ursulin 2018-03-15 13:45 ` Chris Wilson 2018-03-15 13:45 ` [Intel-gfx] " Chris Wilson 2018-03-15 14:37 ` Tvrtko Ursulin 2018-03-15 14:37 ` Tvrtko Ursulin 2018-03-15 14:46 ` Chris Wilson 2018-03-15 14:46 ` Chris Wilson 2018-03-15 14:53 ` Tvrtko Ursulin 2018-03-15 14:53 ` [Intel-gfx] " Tvrtko Ursulin 2018-03-15 14:58 ` Chris Wilson 2018-03-15 14:58 ` Chris Wilson 2018-03-15 14:03 ` [igt-dev] ✓ Fi.CI.BAT: success for " Patchwork 2018-03-15 15:46 ` [PATCH i-g-t v2] " Tvrtko Ursulin 2018-03-15 15:46 ` [Intel-gfx] " Tvrtko Ursulin 2018-03-15 16:01 ` Chris Wilson [this message] 2018-03-15 16:01 ` [igt-dev] " Chris Wilson 2018-03-16 7:36 ` [PATCH i-g-t v3] " Tvrtko Ursulin 2018-03-16 7:36 ` [igt-dev] " Tvrtko Ursulin 2018-03-16 10:17 ` [PATCH i-g-t v4] " Tvrtko Ursulin 2018-03-16 10:17 ` [igt-dev] " Tvrtko Ursulin 2018-03-16 12:18 ` [PATCH i-g-t v5] " Tvrtko Ursulin 2018-03-16 12:18 ` [igt-dev] " Tvrtko Ursulin 2018-03-16 13:31 ` [PATCH i-g-t v6] " Tvrtko Ursulin 2018-03-16 13:31 ` [Intel-gfx] " Tvrtko Ursulin 2018-03-19 13:56 ` [PATCH i-g-t v7] " Tvrtko Ursulin 2018-03-19 13:56 ` [Intel-gfx] " Tvrtko Ursulin 2018-03-19 14:02 ` Chris Wilson 2018-03-19 14:02 ` [igt-dev] [Intel-gfx] " Chris Wilson 2018-03-19 15:29 ` Chris Wilson 2018-03-19 15:29 ` [igt-dev] [Intel-gfx] " Chris Wilson 2018-03-19 15:33 ` Chris Wilson 2018-03-19 15:33 ` [igt-dev] [Intel-gfx] " Chris Wilson 2018-03-19 16:59 ` [PATCH i-g-t v8] " Tvrtko Ursulin 2018-03-19 16:59 ` [igt-dev] " Tvrtko Ursulin 2018-03-20 13:51 ` [PATCH i-g-t v9] " Tvrtko Ursulin 2018-03-20 13:51 ` [igt-dev] " Tvrtko Ursulin 2018-03-15 16:35 ` [igt-dev] ✗ Fi.CI.IGT: failure for " Patchwork 2018-03-15 16:49 ` [igt-dev] ✗ Fi.CI.BAT: failure for tests/perf_pmu: Improve accuracy by waiting on spinner to start (rev2) Patchwork 2018-03-16 8:02 ` [igt-dev] ✓ Fi.CI.BAT: success for tests/perf_pmu: Improve accuracy by waiting on spinner to start (rev3) Patchwork 2018-03-16 8:52 ` [igt-dev] ✗ Fi.CI.IGT: failure " Patchwork 2018-03-16 11:20 ` [igt-dev] ✓ Fi.CI.BAT: success for tests/perf_pmu: Improve accuracy by waiting on spinner to start (rev4) Patchwork 2018-03-16 12:27 ` [igt-dev] ✗ Fi.CI.IGT: failure " Patchwork 2018-03-16 14:30 ` [igt-dev] ✓ Fi.CI.BAT: success for tests/perf_pmu: Improve accuracy by waiting on spinner to start (rev6) Patchwork 2018-03-19 10:24 ` [igt-dev] ✗ Fi.CI.IGT: failure " Patchwork 2018-03-19 21:12 ` [igt-dev] ✓ Fi.CI.BAT: success for tests/perf_pmu: Improve accuracy by waiting on spinner to start (rev9) Patchwork 2018-03-20 0:26 ` [igt-dev] ✓ Fi.CI.IGT: " Patchwork 2018-03-20 17:12 ` [igt-dev] ✓ Fi.CI.BAT: success for tests/perf_pmu: Improve accuracy by waiting on spinner to start (rev10) Patchwork 2018-03-20 20:31 ` [igt-dev] ✓ Fi.CI.IGT: " 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=152112970795.25315.3159070975692005147@mail.alporthouse.com \ --to=chris@chris-wilson.co.uk \ --cc=Intel-gfx@lists.freedesktop.org \ --cc=igt-dev@lists.freedesktop.org \ --cc=tursulin@ursulin.net \ /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: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.