From: Chris Wilson <chris@chris-wilson.co.uk> To: intel-gfx@lists.freedesktop.org Cc: igt-dev@lists.freedesktop.org Subject: [PATCH i-g-t 08/25] i915/gem_sync: Make switch-default asymmetric Date: Thu, 14 Mar 2019 14:19:22 +0000 [thread overview] Message-ID: <20190314141939.26246-8-chris@chris-wilson.co.uk> (raw) In-Reply-To: <20190314141939.26246-1-chris@chris-wilson.co.uk> To make the demonstration of the cheeky preemption more impactful, make the second context a nop to contrast the first being 1024 MI_STORE_DWORD_IMM. Then if we execute and wait on the second context before executing the first, the client latency is even more drastically reduced. To more clearly show any effect on wait reordering, measure the alternative path and present both. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> --- tests/i915/gem_sync.c | 40 +++++++++++++++++++++++++++++----------- 1 file changed, 29 insertions(+), 11 deletions(-) diff --git a/tests/i915/gem_sync.c b/tests/i915/gem_sync.c index fb209977d..3e4feff32 100644 --- a/tests/i915/gem_sync.c +++ b/tests/i915/gem_sync.c @@ -651,7 +651,7 @@ switch_ring(int fd, unsigned ring, int num_children, int timeout) struct drm_i915_gem_relocation_entry reloc[1024]; struct drm_i915_gem_execbuffer2 execbuf; } contexts[2]; - double start, elapsed; + double elapsed, baseline; unsigned long cycles; for (int i = 0; i < ARRAY_SIZE(contexts); i++) { @@ -679,7 +679,7 @@ switch_ring(int fd, unsigned ring, int num_children, int timeout) c->object[1].handle = gem_create(fd, sz); c->object[1].relocs_ptr = to_user_pointer(c->reloc); - c->object[1].relocation_count = 1024; + c->object[1].relocation_count = 1024 * i; batch = gem_mmap__cpu(fd, c->object[1].handle, 0, sz, PROT_WRITE | PROT_READ); @@ -688,7 +688,7 @@ switch_ring(int fd, unsigned ring, int num_children, int timeout) memset(c->reloc, 0, sizeof(c->reloc)); b = batch; - for (int r = 0; r < 1024; r++) { + for (int r = 0; r < c->object[1].relocation_count; r++) { uint64_t offset; c->reloc[r].presumed_offset = c->object[0].offset; @@ -722,26 +722,44 @@ switch_ring(int fd, unsigned ring, int num_children, int timeout) } cycles = 0; - elapsed = 0; - start = gettime(); - do { + baseline = 0; + igt_until_timeout(timeout) { do { double this; - gem_execbuf(fd, &contexts[0].execbuf); gem_execbuf(fd, &contexts[1].execbuf); + gem_execbuf(fd, &contexts[0].execbuf); this = gettime(); gem_sync(fd, contexts[1].object[1].handle); - elapsed += gettime() - this; + gem_sync(fd, contexts[0].object[1].handle); + baseline += gettime() - this; + } while (++cycles & 1023); + } + baseline /= cycles; + + cycles = 0; + elapsed = 0; + igt_until_timeout(timeout) { + do { + double this; + gem_execbuf(fd, &contexts[1].execbuf); + gem_execbuf(fd, &contexts[0].execbuf); + + this = gettime(); gem_sync(fd, contexts[0].object[1].handle); + elapsed += gettime() - this; + + gem_sync(fd, contexts[1].object[1].handle); } while (++cycles & 1023); - } while ((gettime() - start) < timeout); - igt_info("%s%sompleted %ld cycles: %.3f us\n", + } + elapsed /= cycles; + + igt_info("%s%sompleted %ld cycles: %.3f us, baseline %.3f us\n", names[child % num_engines] ?: "", names[child % num_engines] ? " c" : "C", - cycles, elapsed*1e6/cycles); + cycles, elapsed*1e6, baseline*1e6); for (int i = 0; i < ARRAY_SIZE(contexts); i++) { gem_close(fd, contexts[i].object[1].handle); -- 2.20.1 _______________________________________________ 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: intel-gfx@lists.freedesktop.org Cc: igt-dev@lists.freedesktop.org Subject: [igt-dev] [PATCH i-g-t 08/25] i915/gem_sync: Make switch-default asymmetric Date: Thu, 14 Mar 2019 14:19:22 +0000 [thread overview] Message-ID: <20190314141939.26246-8-chris@chris-wilson.co.uk> (raw) In-Reply-To: <20190314141939.26246-1-chris@chris-wilson.co.uk> To make the demonstration of the cheeky preemption more impactful, make the second context a nop to contrast the first being 1024 MI_STORE_DWORD_IMM. Then if we execute and wait on the second context before executing the first, the client latency is even more drastically reduced. To more clearly show any effect on wait reordering, measure the alternative path and present both. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> --- tests/i915/gem_sync.c | 40 +++++++++++++++++++++++++++++----------- 1 file changed, 29 insertions(+), 11 deletions(-) diff --git a/tests/i915/gem_sync.c b/tests/i915/gem_sync.c index fb209977d..3e4feff32 100644 --- a/tests/i915/gem_sync.c +++ b/tests/i915/gem_sync.c @@ -651,7 +651,7 @@ switch_ring(int fd, unsigned ring, int num_children, int timeout) struct drm_i915_gem_relocation_entry reloc[1024]; struct drm_i915_gem_execbuffer2 execbuf; } contexts[2]; - double start, elapsed; + double elapsed, baseline; unsigned long cycles; for (int i = 0; i < ARRAY_SIZE(contexts); i++) { @@ -679,7 +679,7 @@ switch_ring(int fd, unsigned ring, int num_children, int timeout) c->object[1].handle = gem_create(fd, sz); c->object[1].relocs_ptr = to_user_pointer(c->reloc); - c->object[1].relocation_count = 1024; + c->object[1].relocation_count = 1024 * i; batch = gem_mmap__cpu(fd, c->object[1].handle, 0, sz, PROT_WRITE | PROT_READ); @@ -688,7 +688,7 @@ switch_ring(int fd, unsigned ring, int num_children, int timeout) memset(c->reloc, 0, sizeof(c->reloc)); b = batch; - for (int r = 0; r < 1024; r++) { + for (int r = 0; r < c->object[1].relocation_count; r++) { uint64_t offset; c->reloc[r].presumed_offset = c->object[0].offset; @@ -722,26 +722,44 @@ switch_ring(int fd, unsigned ring, int num_children, int timeout) } cycles = 0; - elapsed = 0; - start = gettime(); - do { + baseline = 0; + igt_until_timeout(timeout) { do { double this; - gem_execbuf(fd, &contexts[0].execbuf); gem_execbuf(fd, &contexts[1].execbuf); + gem_execbuf(fd, &contexts[0].execbuf); this = gettime(); gem_sync(fd, contexts[1].object[1].handle); - elapsed += gettime() - this; + gem_sync(fd, contexts[0].object[1].handle); + baseline += gettime() - this; + } while (++cycles & 1023); + } + baseline /= cycles; + + cycles = 0; + elapsed = 0; + igt_until_timeout(timeout) { + do { + double this; + gem_execbuf(fd, &contexts[1].execbuf); + gem_execbuf(fd, &contexts[0].execbuf); + + this = gettime(); gem_sync(fd, contexts[0].object[1].handle); + elapsed += gettime() - this; + + gem_sync(fd, contexts[1].object[1].handle); } while (++cycles & 1023); - } while ((gettime() - start) < timeout); - igt_info("%s%sompleted %ld cycles: %.3f us\n", + } + elapsed /= cycles; + + igt_info("%s%sompleted %ld cycles: %.3f us, baseline %.3f us\n", names[child % num_engines] ?: "", names[child % num_engines] ? " c" : "C", - cycles, elapsed*1e6/cycles); + cycles, elapsed*1e6, baseline*1e6); for (int i = 0; i < ARRAY_SIZE(contexts); i++) { gem_close(fd, contexts[i].object[1].handle); -- 2.20.1 _______________________________________________ igt-dev mailing list igt-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/igt-dev
next prev parent reply other threads:[~2019-03-14 14:19 UTC|newest] Thread overview: 51+ messages / expand[flat|nested] mbox.gz Atom feed top 2019-03-14 14:19 [PATCH i-g-t 01/25] i915/gem_create: Always try to create an object of at least one page Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 02/25] lib/i915: Pretty print HW semaphores Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 03/25] lib: Add GPU power measurement Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 04/25] i915/gem_exec_schedule: Measure semaphore power consumption Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 05/25] i915/gem_exec_whisper: Measure total power consumed Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 06/25] i915/gem_exec_schedule: Verify that using HW semaphores doesn't block Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 07/25] i915/gem_exec_nop: poll-sequential requires ordering between rings Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` Chris Wilson [this message] 2019-03-14 14:19 ` [igt-dev] [PATCH i-g-t 08/25] i915/gem_sync: Make switch-default asymmetric Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 09/25] i915/gem_ctx_param: Remove kneecapping Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 10/25] i915/gem_exec_big: Add a single shot test Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 11/25] kms_fence_pin_leak: Ask for the GPU before use Chris Wilson 2019-03-14 14:19 ` [Intel-gfx] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 12/25] drm-uapi: Import i915_drm.h upto 364df3d04d51 Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 13/25] lib/i915: Improve gem_context error messages Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 14/25] i915/gem_ctx_param: Test set/get (copy) VM Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 15/25] i915/gem_ctx_create: Basic checks for constructor properties Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 16/25] i915: Add gem_ctx_clone Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 17/25] i915: Add gem_vm_create Chris Wilson 2019-03-14 14:19 ` [Intel-gfx] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 18/25] i915: Exercise creating context with shared GTT Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 19/25] i915/gem_ctx_switch: Exercise queues Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 20/25] i915/gem_exec_whisper: Fork all-engine tests one-per-engine Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 21/25] i915/gem_exec_whisper: debugfs/next_seqno is defunct Chris Wilson 2019-03-14 14:19 ` [Intel-gfx] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 22/25] i915: Add gem_ctx_engines Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 23/25] i915: Add gem_exec_balancer Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 24/25] i915/gem_exec_balancer: Exercise bonded pairs Chris Wilson 2019-03-14 14:19 ` [igt-dev] " Chris Wilson 2019-03-14 14:19 ` [PATCH i-g-t 25/25] i915/gem_exec_latency: Measure the latency of context switching Chris Wilson 2019-03-14 14:19 ` [Intel-gfx] " Chris Wilson 2019-03-14 15:01 ` [igt-dev] ✓ Fi.CI.BAT: success for series starting with [i-g-t,01/25] i915/gem_create: Always try to create an object of at least one page 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=20190314141939.26246-8-chris@chris-wilson.co.uk \ --to=chris@chris-wilson.co.uk \ --cc=igt-dev@lists.freedesktop.org \ --cc=intel-gfx@lists.freedesktop.org \ /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.