benchmark/gem_wait: poc for benchmarking i915_wait_request overhead
One scenario under recent discussion is that of having a thundering herd
in i915_wait_request - where the overhead of waking up every waiter for
every batchbuffer was significantly impacting customer throughput. This
benchmark tries to replicate something to that effect by having a large
number of consumers generating a busy load (a large copy followed by
lots of small copies to generate lots of interrupts) and tries to wait
upon all the consumers concurrenctly (to reproduce the thundering herd
effect). To measure the overhead, we have a bunch of cpu hogs - less
kernel overhead in waiting should allow more CPU throughput.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
diff --git a/benchmarks/.gitignore b/benchmarks/.gitignore
index 3937d46..67afa18 100644
--- a/benchmarks/.gitignore
+++ b/benchmarks/.gitignore
@@ -7,6 +7,7 @@
gem_mmap
gem_prw
gem_userptr_benchmark
+gem_wait
intel_upload_blit_large
intel_upload_blit_large_gtt
intel_upload_blit_large_map
diff --git a/benchmarks/Makefile.am b/benchmarks/Makefile.am
index 15e55b4..9ba0fda 100644
--- a/benchmarks/Makefile.am
+++ b/benchmarks/Makefile.am
@@ -8,3 +8,6 @@
benchmarks_LTLIBRARIES = gem_exec_tracer.la
gem_exec_tracer_la_LDFLAGS = -module -avoid-version -no-undefined
gem_exec_tracer_la_LIBADD = -ldl
+
+gem_wait_CFLAGS = $(AM_CFLAGS) $(THREAD_CFLAGS)
+gem_wait_LDADD = $(LDADD) -lpthread
diff --git a/benchmarks/Makefile.sources b/benchmarks/Makefile.sources
index 94e0ea3..453f223 100644
--- a/benchmarks/Makefile.sources
+++ b/benchmarks/Makefile.sources
@@ -14,5 +14,6 @@
gem_mmap \
gem_prw \
gem_userptr_benchmark \
+ gem_wait \
kms_vblank \
$(NULL)
diff --git a/benchmarks/gem_wait.c b/benchmarks/gem_wait.c
new file mode 100644
index 0000000..912ffad
--- /dev/null
+++ b/benchmarks/gem_wait.c
@@ -0,0 +1,302 @@
+/*
+ * Copyright © 2011 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:
+ * Chris Wilson <chris@chris-wilson.co.uk>
+ *
+ */
+
+#include "igt.h"
+#include <unistd.h>
+#include <stdlib.h>
+#include <stdint.h>
+#include <stdio.h>
+#include <string.h>
+#include <fcntl.h>
+#include <inttypes.h>
+#include <pthread.h>
+#include <signal.h>
+#include <errno.h>
+#include <sys/stat.h>
+#include <sys/ioctl.h>
+#include <sys/time.h>
+#include <time.h>
+#include <math.h>
+
+#include "drm.h"
+
+#define LOCAL_I915_EXEC_NO_RELOC (1<<11)
+#define LOCAL_I915_EXEC_HANDLE_LUT (1<<12)
+
+#define COPY_BLT_CMD (2<<29|0x53<<22|0x6)
+#define BLT_WRITE_ALPHA (1<<21)
+#define BLT_WRITE_RGB (1<<20)
+#define BLT_SRC_TILED (1<<15)
+#define BLT_DST_TILED (1<<11)
+
+#define OBJECT 1024*1024
+
+static int has_64bit_reloc;
+
+static int gem_linear_blt(int fd,
+ uint32_t *batch,
+ int offset,
+ uint32_t src,
+ uint32_t dst,
+ uint32_t length,
+ struct drm_i915_gem_relocation_entry *reloc)
+{
+ uint32_t *b = batch + offset/4;
+ int height = length / (16 * 1024);
+
+ igt_assert_lte(height, 1 << 16);
+
+ if (height) {
+ int i = 0;
+ b[i++] = COPY_BLT_CMD | BLT_WRITE_ALPHA | BLT_WRITE_RGB;
+ if (has_64bit_reloc)
+ b[i-1]+=2;
+ b[i++] = 0xcc << 16 | 1 << 25 | 1 << 24 | (16*1024);
+ b[i++] = 0;
+ b[i++] = height << 16 | (4*1024);
+ b[i++] = 0;
+ reloc->offset = (b-batch+4) * sizeof(uint32_t);
+ reloc->delta = 0;
+ reloc->target_handle = dst;
+ reloc->read_domains = I915_GEM_DOMAIN_RENDER;
+ reloc->write_domain = I915_GEM_DOMAIN_RENDER;
+ reloc->presumed_offset = 0;
+ reloc++;
+ if (has_64bit_reloc)
+ b[i++] = 0; /* FIXME */
+
+ b[i++] = 0;
+ b[i++] = 16*1024;
+ b[i++] = 0;
+ reloc->offset = (b-batch+7) * sizeof(uint32_t);
+ if (has_64bit_reloc)
+ reloc->offset += sizeof(uint32_t);
+ reloc->delta = 0;
+ reloc->target_handle = src;
+ reloc->read_domains = I915_GEM_DOMAIN_RENDER;
+ reloc->write_domain = 0;
+ reloc->presumed_offset = 0;
+ reloc++;
+ if (has_64bit_reloc)
+ b[i++] = 0; /* FIXME */
+
+ b += i;
+ length -= height * 16*1024;
+ }
+
+ if (length) {
+ int i = 0;
+ b[i++] = COPY_BLT_CMD | BLT_WRITE_ALPHA | BLT_WRITE_RGB;
+ if (has_64bit_reloc)
+ b[i-1]+=2;
+ b[i++] = 0xcc << 16 | 1 << 25 | 1 << 24 | (16*1024);
+ b[i++] = height << 16;
+ b[i++] = (1+height) << 16 | (length / 4);
+ b[i++] = 0;
+ reloc->offset = (b-batch+4) * sizeof(uint32_t);
+ reloc->delta = 0;
+ reloc->target_handle = dst;
+ reloc->read_domains = I915_GEM_DOMAIN_RENDER;
+ reloc->write_domain = I915_GEM_DOMAIN_RENDER;
+ reloc->presumed_offset = 0;
+ reloc++;
+ if (has_64bit_reloc)
+ b[i++] = 0; /* FIXME */
+
+ b[i++] = height << 16;
+ b[i++] = 16*1024;
+ b[i++] = 0;
+ reloc->offset = (b-batch+7) * sizeof(uint32_t);
+ if (has_64bit_reloc)
+ reloc->offset += sizeof(uint32_t);
+ reloc->delta = 0;
+ reloc->target_handle = src;
+ reloc->read_domains = I915_GEM_DOMAIN_RENDER;
+ reloc->write_domain = 0;
+ reloc->presumed_offset = 0;
+ reloc++;
+ if (has_64bit_reloc)
+ b[i++] = 0; /* FIXME */
+
+ b += i;
+ }
+
+ b[0] = MI_BATCH_BUFFER_END;
+ b[1] = 0;
+
+ return (b+2 - batch) * sizeof(uint32_t);
+}
+
+static int __gem_execbuf(int fd, struct drm_i915_gem_execbuffer2 *execbuf)
+{
+ int err = 0;
+ if (drmIoctl(fd, DRM_IOCTL_I915_GEM_EXECBUFFER2, execbuf))
+ err = -errno;
+ return err;
+}
+
+static void waiter(int child)
+{
+ struct drm_i915_gem_execbuffer2 execbuf;
+ struct drm_i915_gem_exec_object2 exec[3];
+ struct drm_i915_gem_relocation_entry reloc[2*4096/32];
+ uint32_t *buf, handle, src, dst;
+ int fd, len, gen, nreloc, last = 0;
+ int ring;
+
+ fd = drm_open_driver(DRIVER_INTEL);
+ handle = gem_create(fd, 4096);
+ buf = gem_mmap__cpu(fd, handle, 0, 4096, PROT_WRITE);
+
+ gen = intel_gen(intel_get_drm_devid(fd));
+ has_64bit_reloc = gen >= 8;
+
+ src = gem_create(fd, OBJECT);
+ dst = gem_create(fd, OBJECT);
+
+ len = gem_linear_blt(fd, buf, 0, 0, 1, OBJECT, reloc);
+
+ memset(exec, 0, sizeof(exec));
+ exec[0].handle = src;
+ exec[1].handle = dst;
+
+ exec[2].handle = handle;
+ if (has_64bit_reloc)
+ exec[2].relocation_count = len > 56 ? 4 : 2;
+ else
+ exec[2].relocation_count = len > 40 ? 4 : 2;
+ exec[2].relocs_ptr = (uintptr_t)reloc;
+
+ ring = 0;
+ if (gen >= 6)
+ ring = I915_EXEC_BLT;
+
+ memset(&execbuf, 0, sizeof(execbuf));
+ execbuf.buffers_ptr = (uintptr_t)exec;
+ execbuf.buffer_count = 3;
+ execbuf.batch_len = len;
+ execbuf.flags = ring;
+ execbuf.flags |= LOCAL_I915_EXEC_HANDLE_LUT;
+ execbuf.flags |= LOCAL_I915_EXEC_NO_RELOC;
+
+ if (__gem_execbuf(fd, &execbuf)) {
+ gem_set_domain(fd, handle, I915_GEM_DOMAIN_CPU, I915_GEM_DOMAIN_CPU);
+ len = gem_linear_blt(fd, buf, 0, src, dst, OBJECT, reloc);
+ igt_assert(len == execbuf.batch_len);
+ execbuf.flags = ring;
+ gem_execbuf(fd, &execbuf);
+ }
+ gem_sync(fd, handle);
+
+ if (execbuf.flags & LOCAL_I915_EXEC_HANDLE_LUT) {
+ src = 0;
+ dst = 1;
+ }
+
+ nreloc = exec[2].relocation_count;
+ while (execbuf.batch_len + len <= 4096) {
+ last = len - 8;
+ len = gem_linear_blt(fd, buf, len - 8,
+ src, dst, OBJECT,
+ reloc + exec[2].relocation_count);
+ exec[2].relocation_count += nreloc;
+ }
+ munmap(buf, 4096);
+
+ execbuf.batch_len = len;
+ gem_execbuf(fd, &execbuf);
+ usleep(0);
+
+ execbuf.batch_len = len - last;
+ execbuf.batch_start_offset = last;
+ for (int loop = 0; loop < 16*1024; loop++)
+ gem_execbuf(fd, &execbuf);
+ usleep(0);
+
+ execbuf.batch_len = len;
+ execbuf.batch_start_offset = 0;
+ gem_execbuf(fd, &execbuf);
+
+ gem_sync(fd, handle);
+
+ close(fd);
+}
+
+static void *thread(void *arg)
+{
+ uint64_t *c = arg;
+
+ pthread_setcanceltype(PTHREAD_CANCEL_ASYNCHRONOUS, NULL);
+
+ while (1)
+ __sync_fetch_and_add(c, 1);
+
+ return NULL;
+}
+
+static int run(int num_waiters)
+{
+ int num_cpus = sysconf(_SC_NPROCESSORS_ONLN);
+ pthread_t *threads = calloc(num_cpus, sizeof(pthread_t));
+ uint64_t *counters = calloc(num_cpus, 8*sizeof(uint64_t));
+ uint64_t count;
+
+ for (int n = 0; n < num_cpus; n++)
+ pthread_create(&threads[n], NULL, thread, &counters[8*n]);
+
+ igt_fork(child, num_waiters)
+ waiter(child);
+ igt_waitchildren();
+
+ count = 0;
+ for (int n = 0; n < num_cpus; n++) {
+ pthread_cancel(threads[n]);
+ __sync_synchronize();
+ count += counters[8*n];
+ }
+ printf("%llu\n", (long long unsigned)count);
+
+ return 0;
+}
+
+int main(int argc, char **argv)
+{
+ int num_waiters = 128;
+ int c;
+
+ while ((c = getopt (argc, argv, "w:")) != -1) {
+ switch (c) {
+ case 'w':
+ num_waiters = atoi(optarg);
+ if (num_waiters < 1)
+ num_waiters = 1;
+ break;
+ }
+ }
+
+ return run(num_waiters);
+}