[Intel-gfx] [PATCH] igt: Add benchmark for read-read optimisation

[Chris Wilson]

Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
---
 tests/.gitignore            |   1 +
 tests/Makefile.sources      |   1 +
 tests/gem_read_read_speed.c | 193 ++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 195 insertions(+)
 create mode 100644 tests/gem_read_read_speed.c

diff --git a/tests/.gitignore b/tests/.gitignore
index dd2bff6..d2918a5 100644
--- a/tests/.gitignore
+++ b/tests/.gitignore
@@ -77,6 +77,7 @@ gem_pread_after_blit
 gem_pwrite
 gem_pwrite_pread
 gem_readwrite
+gem_read_read_speed
 gem_reg_read
 gem_reloc_overflow
 gem_reloc_vs_gpu
diff --git a/tests/Makefile.sources b/tests/Makefile.sources
index 9353faa..c32c37a 100644
--- a/tests/Makefile.sources
+++ b/tests/Makefile.sources
@@ -54,6 +54,7 @@ TESTS_progs_M = \
 	gem_pwrite \
 	gem_pwrite_pread \
 	gem_readwrite \
+	gem_read_read_speed \
 	gem_reloc_overflow \
 	gem_reloc_vs_gpu \
 	gem_render_copy_redux \
diff --git a/tests/gem_read_read_speed.c b/tests/gem_read_read_speed.c
new file mode 100644
index 0000000..0fbbd03
--- /dev/null
+++ b/tests/gem_read_read_speed.c
@@ -0,0 +1,193 @@
+/*
+ * Copyright © 2015 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+/** @file gem_read_read_speed.c
+ *
+ * This is a test of performance with multiple readers from the same source.
+ */
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <fcntl.h>
+#include <inttypes.h>
+#include <errno.h>
+#include <time.h>
+#include <sys/stat.h>
+#include <sys/time.h>
+#include <sys/wait.h>
+
+#include <drm.h>
+
+#include "ioctl_wrappers.h"
+#include "drmtest.h"
+#include "igt_aux.h"
+#include "igt_core.h"
+#include "igt_gt.h"
+#include "intel_bufmgr.h"
+#include "intel_batchbuffer.h"
+#include "intel_io.h"
+#include "intel_chipset.h"
+
+IGT_TEST_DESCRIPTION("Test of pread/pwrite behavior when writing to active"
+		     " buffers.");
+
+const int width = 512, height = 512;
+igt_render_copyfunc_t rendercopy;
+struct intel_batchbuffer *batch;
+
+static int gem_param(int fd, int name)
+{
+	drm_i915_getparam_t gp;
+	int v = -1; /* No param uses the sign bit, reserve it for errors */
+
+	memset(&gp, 0, sizeof(gp));
+	gp.param = name;
+	gp.value = &v;
+	if (drmIoctl(fd, DRM_IOCTL_I915_GETPARAM, &gp))
+		return -1;
+
+	return v;
+}
+
+static int semaphores_enabled(int fd)
+{
+	FILE *file;
+	int detected = -1;
+	int ret;
+
+	ret = gem_param(fd, 20);
+	if (ret != -1)
+		return ret > 0;
+
+	file = fopen("/sys/module/i915/parameters/semaphores", "r");
+	if (file) {
+		int value;
+		if (fscanf(file, "%d", &value) == 1)
+			detected = value;
+		fclose(file);
+	}
+
+	return detected;
+}
+
+static void rcs_copy_bo(drm_intel_bo *dst, drm_intel_bo *src)
+{
+	struct igt_buf d = {
+		.bo = dst,
+		.size = width * height * 4,
+		.num_tiles = width * height * 4,
+		.stride = width * 4,
+	}, s = {
+		.bo = src,
+		.size = width * height * 4,
+		.num_tiles = width * height * 4,
+		.stride = width * 4,
+	};
+	uint32_t swizzle;
+
+	drm_intel_bo_get_tiling(dst, &d.tiling, &swizzle);
+	drm_intel_bo_get_tiling(src, &s.tiling, &swizzle);
+
+	rendercopy(batch, NULL,
+		   &s, 0, 0,
+		   width, height,
+		   &d, 0, 0);
+}
+
+static void bcs_copy_bo(drm_intel_bo *dst, drm_intel_bo *src)
+{
+	intel_blt_copy(batch,
+		       src, 0, 0, 4*width,
+		       dst, 0, 0, 4*width,
+		       width, height, 32);
+}
+
+static void
+set_bo(drm_intel_bo *bo, uint32_t val)
+{
+	int size = width * height;
+	uint32_t *vaddr;
+
+	do_or_die(drm_intel_gem_bo_map_gtt(bo));
+	vaddr = bo->virtual;
+	while (size--)
+		*vaddr++ = val;
+	drm_intel_bo_unmap(bo);
+}
+
+static double elapsed(const struct timespec *start,
+		      const struct timespec *end,
+		      int loop)
+{
+	return (1e6*(end->tv_sec - start->tv_sec) + (end->tv_nsec - start->tv_nsec)/1000)/loop;
+}
+
+
+igt_main
+{
+	drm_intel_bo *src = NULL, *bcs = NULL, *rcs = NULL;
+	drm_intel_bufmgr *bufmgr;
+	int fd;
+
+	igt_skip_on_simulation();
+
+	igt_fixture {
+		int devid;
+
+		fd = drm_open_any();
+		igt_info("Semaphores: %d\n", semaphores_enabled(fd));
+
+		devid = intel_get_drm_devid(fd);
+
+		rendercopy = igt_get_render_copyfunc(devid);
+		igt_require(rendercopy);
+
+		bufmgr = drm_intel_bufmgr_gem_init(fd, 4096);
+		igt_assert(bufmgr);
+
+		batch =  intel_batchbuffer_alloc(bufmgr, devid);
+
+		src = drm_intel_bo_alloc(bufmgr, "src", 4*width*height, 0);
+		bcs = drm_intel_bo_alloc(bufmgr, "bcs", 4*width*height, 0);
+		rcs = drm_intel_bo_alloc(bufmgr, "rcs", 4*width*height, 0);
+
+		set_bo(src, 0xdeadbeef);
+	}
+
+	igt_subtest("read-read") {
+		struct timespec start, end;
+		int loops = 1000;
+
+		clock_gettime(CLOCK_MONOTONIC, &start);
+		for (int i = 0; i < loops; i++) {
+			rcs_copy_bo(rcs, src);
+			bcs_copy_bo(bcs, src);
+		}
+		drm_intel_gem_bo_start_gtt_access(src, true);
+		clock_gettime(CLOCK_MONOTONIC, &end);
+
+		igt_info("Time to read-read %dk:		%7.3fµs\n",
+			 4*width*height/1024, elapsed(&start, &end, loops));
+	}
+}
-- 
2.1.4