[Intel-gfx] [RFC] benchmarks/gem_slice_shutdown: microbenchmark for slice shutdown delays
Oscar Mateo
oscar.mateo at intel.com
Tue May 2 15:08:18 UTC 2017
Cc: Dmitry Rogozhkin <dmitry.v.rogozhkin at intel.com>
Cc: Chris Wilson <chris at chris-wilson.co.uk>
Signed-off-by: Oscar Mateo <oscar.mateo at intel.com>
---
benchmarks/Makefile.sources | 1 +
benchmarks/gem_slice_shutdown.c | 295 ++++++++++++++++++++++++++++++++++++++++
2 files changed, 296 insertions(+)
create mode 100644 benchmarks/gem_slice_shutdown.c
diff --git a/benchmarks/Makefile.sources b/benchmarks/Makefile.sources
index 3a94115..591b5ae 100644
--- a/benchmarks/Makefile.sources
+++ b/benchmarks/Makefile.sources
@@ -13,6 +13,7 @@ benchmarks_prog_list = \
gem_mmap \
gem_prw \
gem_set_domain \
+ gem_slice_shutdown \
gem_syslatency \
gem_wsim \
kms_vblank \
diff --git a/benchmarks/gem_slice_shutdown.c b/benchmarks/gem_slice_shutdown.c
new file mode 100644
index 0000000..dcb17c1
--- /dev/null
+++ b/benchmarks/gem_slice_shutdown.c
@@ -0,0 +1,295 @@
+/*
+ * Copyright © 2017 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.
+ *
+ * Authors:
+ * Oscar Mateo <oscar.mateo at intel.com>
+ *
+ */
+
+/*
+ * This tool measures time to change the configuration of number of slices
+ */
+
+#include <unistd.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <string.h>
+#include <fcntl.h>
+#include <inttypes.h>
+#include <errno.h>
+#include <sys/stat.h>
+#include <sys/ioctl.h>
+#include <sys/time.h>
+#include <time.h>
+#include <assert.h>
+
+#include "drm.h"
+#include "ioctl_wrappers.h"
+#include "drmtest.h"
+#include "intel_io.h"
+#include "igt_stats.h"
+#include "intel_chipset.h"
+#include "intel_bufmgr.h"
+
+#define GEN8_R_PWR_CLK_STATE (0x20C8)
+#define GEN8_RPCS_ENABLE (1 << 31)
+#define GEN8_RPCS_S_CNT_ENABLE (1 << 18)
+#define GEN8_RPCS_S_CNT_SHIFT 15
+#define GEN8_RPCS_S_CNT_MASK (0x7 << GEN8_RPCS_S_CNT_SHIFT)
+#define GEN8_RPCS_SS_CNT_ENABLE (1 << 11)
+#define GEN8_RPCS_SS_CNT_SHIFT 8
+#define GEN8_RPCS_SS_CNT_MASK (0x7 << GEN8_RPCS_SS_CNT_SHIFT)
+#define GEN8_RPCS_EU_MAX_SHIFT 4
+#define GEN8_RPCS_EU_MAX_MASK (0xf << GEN8_RPCS_EU_MAX_SHIFT)
+#define GEN8_RPCS_EU_MIN_SHIFT 0
+#define GEN8_RPCS_EU_MIN_MASK (0xf << GEN8_RPCS_EU_MIN_SHIFT)
+
+static const char *yesno(bool x)
+{
+ return x ? "yes" : "no";
+}
+
+static void print_rpcs_config(uint32_t rpcs_config)
+{
+ bool rpcs_enable, s_enable, ss_enable;
+ uint s_count, ss_count, eu_max, eu_min;
+
+ rpcs_enable = rpcs_config & GEN8_RPCS_ENABLE;
+ s_enable = rpcs_config & GEN8_RPCS_S_CNT_ENABLE;
+ ss_enable = rpcs_config & GEN8_RPCS_SS_CNT_ENABLE;
+ s_count = (rpcs_config & GEN8_RPCS_S_CNT_MASK) >> GEN8_RPCS_S_CNT_SHIFT;
+ ss_count = (rpcs_config & GEN8_RPCS_SS_CNT_MASK) >> GEN8_RPCS_SS_CNT_SHIFT;
+ eu_max = (rpcs_config & GEN8_RPCS_EU_MAX_MASK) >> GEN8_RPCS_EU_MAX_SHIFT;
+ eu_min = (rpcs_config & GEN8_RPCS_EU_MIN_MASK) >> GEN8_RPCS_EU_MIN_SHIFT;
+ printf("RPCS enabled: %s\n", yesno(rpcs_enable));
+ printf("Slice count enabled: %s, count: %u\n", yesno(s_enable), s_count);
+ printf("Subslice count enabled: %s, count: %u\n", yesno(ss_enable), ss_count);
+ printf("EU max: %u, min: %u\n", eu_max, eu_min);
+}
+
+static void init_buffer(drm_intel_bufmgr *bufmgr,
+ struct igt_buf *buf,
+ uint32_t size)
+{
+ buf->bo = drm_intel_bo_alloc(bufmgr, "", size, 4096);
+ buf->size = size;
+ buf->tiling = I915_TILING_NONE;
+ buf->stride = 4096;
+}
+
+static double elapsed(const struct timespec *start, const struct timespec *end)
+{
+ return 1e6*(end->tv_sec - start->tv_sec) + 1e-3*(end->tv_nsec - start->tv_nsec);
+}
+
+#define MI_STORE_REGISTER_MEM_64_BIT_ADDR ((0x24 << 23) | 2)
+
+static uint32_t do_read_pwrclk_state(drm_intel_bufmgr *bufmgr,
+ struct intel_batchbuffer *batch,
+ drm_intel_context *context)
+{
+ uint32_t rpcs_config;
+ uint32_t *data;
+ drm_intel_bo *dst_bo;
+
+ dst_bo = drm_intel_bo_alloc(bufmgr, "dst", 4, 4096);
+
+ BEGIN_BATCH(3, 1);
+ OUT_BATCH(MI_STORE_REGISTER_MEM_64_BIT_ADDR);
+ OUT_BATCH(GEN8_R_PWR_CLK_STATE);
+ OUT_RELOC(dst_bo, I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER, 0);
+ ADVANCE_BATCH();
+
+ intel_batchbuffer_flush_with_context(batch, context);
+
+ drm_intel_bo_map(dst_bo, 1);
+
+ data = dst_bo->virtual;
+ rpcs_config = *data;
+
+ drm_intel_bo_unmap(dst_bo);
+
+ drm_intel_bo_unreference(dst_bo);
+
+ return rpcs_config;
+}
+
+#define LOCAL_MI_LOAD_REGISTER_IMM (0x22 << 23)
+
+#define GFX_OP_PIPE_CONTROL(len) ((0x3<<29)|(0x3<<27)|(0x2<<24)|((len)-2))
+#define PIPE_CONTROL_CS_STALL (1<<20)
+#define PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH (1<<12)
+#define PIPE_CONTROL_FLUSH_ENABLE (1<<7)
+#define PIPE_CONTROL_DC_FLUSH_ENABLE (1<<5)
+#define PIPE_CONTROL_DEPTH_CACHE_FLUSH (1<<0)
+
+static void emit_config_slice_count(drm_intel_bufmgr *bufmgr,
+ struct intel_batchbuffer *batch,
+ drm_intel_context *context,
+ uint32_t rpcs_config)
+{
+ drm_intel_bo *dst_bo;
+
+ dst_bo = drm_intel_bo_alloc(bufmgr, "scratch", 4, 4096);
+
+ BEGIN_BATCH(9, 1);
+ OUT_BATCH(LOCAL_MI_LOAD_REGISTER_IMM | 1);
+ OUT_BATCH(GEN8_R_PWR_CLK_STATE);
+ OUT_BATCH(rpcs_config);
+ OUT_BATCH(GFX_OP_PIPE_CONTROL(6));
+ OUT_BATCH(PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH |
+ PIPE_CONTROL_DEPTH_CACHE_FLUSH |
+ PIPE_CONTROL_DC_FLUSH_ENABLE |
+ PIPE_CONTROL_FLUSH_ENABLE |
+ PIPE_CONTROL_CS_STALL);
+ OUT_RELOC(dst_bo, I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER, 0);
+ OUT_BATCH(0);
+ OUT_BATCH(0);
+ OUT_BATCH(0);
+ ADVANCE_BATCH();
+
+ drm_intel_bo_unreference(dst_bo);
+}
+
+static void do_test(igt_render_copyfunc_t rendercopy,
+ drm_intel_bufmgr *bufmgr,
+ drm_intel_context *context,
+ struct intel_batchbuffer *batch,
+ uint32_t rpcs_config,
+ struct igt_buf *src,
+ struct igt_buf *dst,
+ uint c)
+{
+ rpcs_config &= ~GEN8_RPCS_S_CNT_MASK;
+ rpcs_config |= (c << GEN8_RPCS_S_CNT_SHIFT);
+
+ emit_config_slice_count(bufmgr, batch, context, rpcs_config);
+ rendercopy(batch, context, src, 0, 0, 0, 0, dst, 0, 0);
+ drm_intel_bo_wait_rendering(dst->bo);
+}
+
+static void do_measures(igt_render_copyfunc_t rendercopy,
+ drm_intel_bufmgr *bufmgr,
+ drm_intel_context *context,
+ struct intel_batchbuffer *batch,
+ uint32_t rpcs_config,
+ struct igt_buf *src,
+ struct igt_buf *dst,
+ uint c1, uint c2)
+{
+ uint c = c1;
+ struct igt_mean real[2], baseline[2];
+ struct timespec start, end;
+ int i, j;
+
+ for (i = 0; i < 2; i++) {
+ igt_mean_init(&real[i]);
+ igt_mean_init(&baseline[i]);
+ }
+
+ /* Initial config and warm up */
+ do_test(rendercopy, bufmgr, context, batch, rpcs_config, src, dst, c);
+
+#define LOOPS 1000
+ for (j = 0; j < LOOPS; j++) {
+ c = c2;
+ for (i = 0; i < 2; i++) {
+ clock_gettime(CLOCK_MONOTONIC, &start);
+ do_test(rendercopy, bufmgr, context, batch,
+ rpcs_config, src, dst, c);
+ clock_gettime(CLOCK_MONOTONIC, &end);
+
+ igt_mean_add(&real[i], elapsed(&start, &end));
+
+ clock_gettime(CLOCK_MONOTONIC, &start);
+ do_test(rendercopy, bufmgr, context, batch,
+ rpcs_config, src, dst, c);
+ clock_gettime(CLOCK_MONOTONIC, &end);
+
+ igt_mean_add(&baseline[i], elapsed(&start, &end));
+
+ c = c1;
+ }
+ }
+
+ printf("Slice poweron time (%u -> %u): %7.3f us\n", c1, c2,
+ igt_mean_get(&real[0]) - igt_mean_get(&baseline[0]));
+ printf("Slice shutdown time (%u -> %u): %7.3f us\n", c2, c1,
+ igt_mean_get(&real[1]) - igt_mean_get(&baseline[1]));
+
+ printf("\n");
+}
+
+static void slice_shutdown(int fd, int devid)
+{
+ igt_render_copyfunc_t rendercopy = igt_get_render_copyfunc(devid);
+ drm_intel_bufmgr *bufmgr = drm_intel_bufmgr_gem_init(fd, 4096);
+ drm_intel_context *context = drm_intel_gem_context_create(bufmgr);
+ struct intel_batchbuffer *batch = intel_batchbuffer_alloc(bufmgr, devid);
+ struct igt_buf src, dst;
+ uint32_t rpcs_config;
+ bool rpcs_enable, s_enable;
+ int s_count;
+ uint i, j;
+
+ init_buffer(bufmgr, &src, 10*4096*4096);
+ init_buffer(bufmgr, &dst, 10*4096*4096);
+
+ rpcs_config = do_read_pwrclk_state(bufmgr, batch, context);
+ print_rpcs_config(rpcs_config);
+
+ rpcs_enable = rpcs_config & GEN8_RPCS_ENABLE;
+ s_enable = rpcs_config & GEN8_RPCS_S_CNT_ENABLE;
+
+ if (!rpcs_enable || !s_enable) {
+ fprintf(stderr, "We need slice count to be already enabled\n");
+ exit(-1);
+ }
+
+ s_count = (rpcs_config & GEN8_RPCS_S_CNT_MASK) >> GEN8_RPCS_S_CNT_SHIFT;
+
+ for (i = 1; i <= s_count; i++)
+ for (j = i + 1; j <= s_count; j++)
+ do_measures(rendercopy, bufmgr, context, batch,
+ rpcs_config, &src, &dst, i, j);
+
+ intel_batchbuffer_free(batch);
+ drm_intel_gem_context_destroy(context);
+ drm_intel_bufmgr_destroy(bufmgr);
+}
+
+int main(int argc, char **argv)
+{
+ int fd, devid;
+
+ fd = drm_open_driver(DRIVER_INTEL);
+
+ devid = intel_get_drm_devid(fd);
+ if (intel_gen(devid) < 8) {
+ fprintf(stderr, "gen8+ required, yours is gen%u\n",
+ intel_gen(devid));
+ exit(-1);
+ }
+
+ slice_shutdown(fd, devid);
+}
--
1.9.1
More information about the Intel-gfx
mailing list