Igt-dev Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Rob Clark <robdclark@gmail.com>
To: igt-dev@lists.freedesktop.org
Cc: Rob Clark <robdclark@chromium.org>
Subject: [igt-dev] [PATCH v2] tests/msm: Add submitoverhead benchmark
Date: Mon,  7 Aug 2023 09:36:57 -0700	[thread overview]
Message-ID: <20230807163659.205382-1-robdclark@gmail.com> (raw)

From: Rob Clark <robdclark@chromium.org>

Something for profiling CPU overhead of submit ioctl.  Not really a
functional test, but it is convenient to re-use the i-g-t infrastructure
for this.

v2: SPDX license, fix meson.build whitespace

Signed-off-by: Rob Clark <robdclark@chromium.org>
---
 tests/meson.build              |  3 +-
 tests/msm/msm_submitoverhead.c | 86 ++++++++++++++++++++++++++++++++++
 2 files changed, 88 insertions(+), 1 deletion(-)
 create mode 100644 tests/msm/msm_submitoverhead.c

diff --git a/tests/meson.build b/tests/meson.build
index 944a0941f9fd..db427a1a4b6e 100644
--- a/tests/meson.build
+++ b/tests/meson.build
@@ -302,7 +302,8 @@ msm_progs = [
 	'msm_mapping',
 	'msm_recovery',
 	'msm_shrink',
-	'msm_submit'
+	'msm_submit',
+	'msm_submitoverhead',
 ]
 
 chamelium_progs = [
diff --git a/tests/msm/msm_submitoverhead.c b/tests/msm/msm_submitoverhead.c
new file mode 100644
index 000000000000..7d92dda7c226
--- /dev/null
+++ b/tests/msm/msm_submitoverhead.c
@@ -0,0 +1,86 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright © 2023 Google, Inc.
+ */
+
+#include "igt.h"
+#include "igt_msm.h"
+
+/*
+ * Not as much a test, as a kernel submit overhead benchmark.  Generates lots
+ * of submit ioctls with various size #s of buffers attached for measuring
+ * and profiling kernel submit CPU overhead.
+ */
+
+#define MAX_BOS 1000
+
+igt_main
+{
+	struct msm_device *dev = NULL;
+	struct msm_pipe *pipe = NULL;
+	struct msm_bo *bos[MAX_BOS];
+	struct drm_msm_gem_submit_bo bos_table[MAX_BOS];
+	static const int sizes[] = {
+		10, 100, 250, 500, 1000,
+	};
+
+	igt_fixture {
+		struct drm_msm_gem_submit req;
+
+		dev = igt_msm_dev_open();
+		pipe = igt_msm_pipe_open(dev, 0);
+		for (int i = 0; i < MAX_BOS; i++) {
+			bos[i] = igt_msm_bo_new(dev, 0x1000, MSM_BO_WC);
+			bos_table[i] = (struct drm_msm_gem_submit_bo) {
+				.handle = bos[i]->handle,
+				/*
+				 * We don't bother testing BO_READ since
+				 * mesa doesn't use that anymore
+				 */
+				.flags  = MSM_SUBMIT_BO_WRITE,
+			};
+		}
+
+		/*
+		 * Prime the pump, so first submit doesn't take the overhead
+		 * of allocating backing pages:
+		 */
+		req = (struct drm_msm_gem_submit) {
+			.flags   = pipe->pipe | MSM_SUBMIT_FENCE_FD_OUT,
+			.queueid = pipe->submitqueue_id,
+			.nr_bos  = ARRAY_SIZE(bos_table),
+			.bos     = VOID2U64(bos_table),
+		};
+		do_ioctl(dev->fd, DRM_IOCTL_MSM_GEM_SUBMIT, &req);
+		igt_wait_and_close(req.fence_fd);
+	}
+
+	for (int i = 0; i < ARRAY_SIZE(sizes); i++) {
+		for (int mode = 0; mode < 2; mode++) {
+			const char *modestr = mode ? "-no-implicit-sync" : "";
+			const uint32_t modeflags = mode ? MSM_SUBMIT_NO_IMPLICIT : 0;
+			igt_subtest_f("submitbench-%u-bos%s", sizes[i], modestr) {
+				struct drm_msm_gem_submit req = {
+					.flags   = pipe->pipe | modeflags,
+					.queueid = pipe->submitqueue_id,
+					.nr_bos  = sizes[i],
+					.bos     = VOID2U64(bos_table),
+				};
+				unsigned iterations = 0;
+				igt_for_milliseconds(2000) {
+					do_ioctl(dev->fd, DRM_IOCTL_MSM_GEM_SUBMIT, &req);
+					iterations++;
+				}
+				igt_info("%u-bos: %u iterations\n", sizes[i], iterations);
+			}
+		}
+	}
+
+	igt_fixture {
+		for (int i = 0; i < MAX_BOS; i++) {
+			igt_msm_bo_free(bos[i]);
+		}
+		igt_msm_pipe_close(pipe);
+		igt_msm_dev_close(dev);
+	}
+}
-- 
2.41.0

             reply	other threads:[~2023-08-07 16:37 UTC|newest]

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-08-07 16:36 Rob Clark [this message]
2023-08-07 18:04 ` [igt-dev] ○ CI.xeBAT: info for tests/msm: Add submitoverhead benchmark (rev2) Patchwork
2023-08-07 18:13 ` [igt-dev] ✓ Fi.CI.BAT: success " Patchwork
2023-08-07 22:59 ` [igt-dev] ✗ Fi.CI.IGT: failure " Patchwork
2023-08-10 19:31   ` Kamil Konieczny
2023-08-11  9:39     ` Yedireswarapu, SaiX Nandan
2023-08-10 19:33 ` [igt-dev] [PATCH v2] tests/msm: Add submitoverhead benchmark Kamil Konieczny
2023-08-11  6:50 ` [igt-dev] ✗ Fi.CI.IGT: failure for tests/msm: Add submitoverhead benchmark (rev2) Patchwork
2023-08-11  7:24 ` Patchwork
2023-08-11  9:31 ` [igt-dev] ✓ Fi.CI.IGT: success " Patchwork
2023-08-11 15:44 ` [igt-dev] [PATCH v2] tests/msm: Add submitoverhead benchmark Kamil Konieczny

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=20230807163659.205382-1-robdclark@gmail.com \
    --to=robdclark@gmail.com \
    --cc=igt-dev@lists.freedesktop.org \
    --cc=robdclark@chromium.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: 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