wsim: Send a periodic depth stamp down each queue

In order to measure external load (from other clients), we need a
periodic marker along each queue.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

1 files changed, 68 insertions, 0 deletions

diff --git a/benchmarks/gem_wsim.c b/benchmarks/gem_wsim.c
index b193378e..3da8499b 100644
--- a/benchmarks/gem_wsim.c
+++ b/benchmarks/gem_wsim.c
@@ -925,6 +925,73 @@ static void w_sync_to(struct workload *wrk, struct w_step *w, int target)
 	gem_sync(fd, wrk->steps[target].obj[0].handle);
 }
 
+static void init_status_page(struct workload *wrk)
+{
+	struct drm_i915_gem_exec_object2 obj[2] = {};
+	struct drm_i915_gem_relocation_entry reloc[3] = {};
+	struct drm_i915_gem_execbuffer2 eb = {
+		.buffer_count = 2, .buffers_ptr = to_user_pointer(obj)
+	};
+
+	/* Want to make sure that the balancer has a reasonable view of
+	 * the background busyness of each engine. To do that we occasionally
+	 * send a dummy batch down the pipeline.
+	 */
+
+	if (!wrk->status_page_handle)
+		return;
+
+	obj[0].handle = wrk->status_page_handle;
+	obj[1].relocs_ptr = to_user_pointer(reloc);
+	obj[1].relocation_count = ARRAY_SIZE(reloc);
+
+	reloc[0].offset = sizeof(uint32_t);
+	reloc[1].offset = 5 * sizeof(uint32_t);
+	reloc[2].offset = 10 * sizeof(uint32_t);
+
+	for (int engine = VCS1; engine <= VCS2; engine++) {
+		uint32_t *cs, *base;
+		uint64_t addr;
+
+		reloc[0].delta = VCS_SEQNO_OFFSET(engine);
+		reloc[1].delta = VCS_SEQNO_OFFSET(engine) + sizeof(uint32_t);
+		reloc[2].delta = VCS_SEQNO_OFFSET(engine) + 2 * sizeof(uint32_t);
+
+		obj[1].handle = gem_create(fd, 4096);
+
+		base = gem_mmap__cpu(fd, obj[1].handle, 0, 4096, PROT_WRITE);
+		cs = base;
+
+		addr = reloc[0].presumed_offset + reloc[0].delta;
+		*cs++ = MI_STORE_DWORD_IMM;
+		*cs++ = addr;
+		*cs++ = addr >> 32;
+		*cs++ = ++wrk->seqno[engine];
+
+		addr = reloc[1].presumed_offset + reloc[1].delta;
+		*cs++ = MI_STORE_DWORD_IMM;
+		*cs++ = addr;
+		*cs++ = addr >> 32;
+		*cs++ = *REG(RCS_TIMESTAMP);
+
+		addr = reloc[2].presumed_offset + reloc[2].delta;
+		*cs++ = 0x24 << 23 | 2; /* MI_STORE_REG_MEM */
+		*cs++ = RCS_TIMESTAMP;
+		*cs++ = addr;
+		*cs++ = addr >> 32;
+
+		*cs++ = MI_BATCH_BUFFER_END;
+		munmap(base, 4096);
+
+		eb.flags = eb_engine_map[engine];
+		eb.flags |= I915_EXEC_HANDLE_LUT;
+		eb.flags |= I915_EXEC_NO_RELOC;
+
+		gem_execbuf(fd, &eb);
+		gem_close(fd, obj[1].handle);
+	}
+}
+
 static void
 run_workload(unsigned int id, struct workload *wrk,
 	     bool background, int pipe_fd,
@@ -946,6 +1013,7 @@ run_workload(unsigned int id, struct workload *wrk,
 
 	for (j = 0; run && (background || j < repeat); j++) {
 		clock_gettime(CLOCK_MONOTONIC, &wrk->repeat_start);
+		init_status_page(wrk);
 
 		for (i = 0, w = wrk->steps; run && (i < wrk->nr_steps);
 		     i++, w++) {