Skip to contentSkip to conversation

Tests: Add test cases based on multi drm_fd to test sync

PATCH Applied Zhao Yakui 2 participants 4 messages Apr 24, 2014

lore
Share & tools
lore

Zhao Yakui

author
#0 lore raw

v4 cover Tests: Add test cases based on multi drm_fd to test sync

This follows Daniel's advice to add the two test cases based on multi drm_fd to test the ring sync and CPU<->GPU sync.
The Broadwell GT3 machine has two independent BSD rings that can be used to process the video commands. This is implemented in kernel driver and transparent to the user-space. But we still need to check the ring sync and CPU<->GPU sync for the second BSD ring. Two tests are created based on the multi drm_fds to test the sync. Multi drm_fd can assure that the second BSD ring has the opportunity to dispatch the GPU command.

V1->V2: Follow Daniel's comment to add one subtext instead of one individual test case, which is used to test the CPU<->GPU sync under multi BSD rings/

V2->V3: Follow Imre's comment to remove the unnecessary initialization and use igt_assert_f instead of igt_assert

V3->V4: Add gem_multi_bsd_sync_loop.c into the tests/.gitignore

Zhao Yakui (2):
tests: Add one ring sync case based on multi drm_fd to test ring
semaphore sync under multi BSD rings
tests/gem_dummy_reloc_loop: Add one subtest based on multi drm_fd to

    test CPU<->GPU sync under multi BSD rings
 tests/.gitignore                |    1 +
 tests/Makefile.sources          |    1 +
 tests/gem_dummy_reloc_loop.c    |  107 +++++++++++++++++++++++-
 tests/gem_multi_bsd_sync_loop.c |  175 +++++++++++++++++++++++++++++++++++++++
 4 files changed, 283 insertions(+), 1 deletion(-)
 create mode 100644 tests/gem_multi_bsd_sync_loop.c
Signature · 1 line
1.7.10.1
author
#1 lore raw

PATCH v4 1/2 tests: Add one ring sync case based on multi drm_fd to test ring semaphore sync under multi BSD rings

The Broadwell GT3 machine has two independent BSD rings in kernel driver while it is transparent to the user-space driver. In such case it needs to check the ring sync between the two BSD rings. At the same time it also needs to check the sync among the second BSD ring and the other rings.

V2->V3: Follow Imre's comment to remove the unnecessary initialization and use igt_assert_f instead of igt_assert.

V3->V4: Add gem_multi_bsd_sync_loop.c into the tests/.gitignore

---
 tests/.gitignore                |    1 +
 tests/Makefile.sources          |    1 +
 tests/gem_multi_bsd_sync_loop.c |  175 +++++++++++++++++++++++++++++++++++++++
 3 files changed, 177 insertions(+)
 create mode 100644 tests/gem_multi_bsd_sync_loop.c
tests/.gitignore +1 -0
index 146bab0..42690dd 100644
@@ -54,6 +54,7 @@ gem_media_fill
 gem_mmap
 gem_mmap_gtt
 gem_mmap_offset_exhaustion
+gem_multi_bsd_sync_loop
 gem_non_secure_batch
 gem_partial_pwrite_pread
 gem_persistent_relocs
tests/Makefile.sources +1 -0
index c957ace..7cd9ca8 100644
@@ -105,6 +105,7 @@ TESTS_progs = \
 	gem_render_tiled_blits \
 	gem_ring_sync_copy \
 	gem_ring_sync_loop \
+	gem_multi_bsd_sync_loop \
 	gem_seqno_wrap \
 	gem_set_tiling_vs_gtt \
 	gem_set_tiling_vs_pwrite \
tests/gem_multi_bsd_sync_loop.c new file +175 -0

175 lines not on this page · raw message

Signature · 6 lines
1.7.10.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
author
#2 lore raw

PATCH v4 2/2 tests/gem_dummy_reloc_loop: Add one subtest based on multi drm_fd to test CPU<->GPU sync under multi BSD rings

In reply to Zhao Yakui

The Broadwell GT3 machine has two independent BSD rings in kernel driver while it is transparent to the user-space driver. In such case it needs to check the CPU<->GPU sync for the second BSD ring.

V1->V2: Follow Daniel's comment to add one subtext instead of one individual test case, which is used to test the CPU<->GPU sync under multi BSD rings.

V2->V3: Follow Imre's comment to remove the unnecessary initialization and use igt_assert_f instead of igt_assert

---
 tests/gem_dummy_reloc_loop.c |  107 +++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 106 insertions(+), 1 deletion(-)
tests/gem_dummy_reloc_loop.c +106 -1
index a61b59b..4e4dd49 100644
@@ -48,6 +48,13 @@ static drm_intel_bufmgr *bufmgr;
 struct intel_batchbuffer *batch;
 static drm_intel_bo *target_buffer;
 
+#define NUM_FD	50
+
+static int mfd[NUM_FD];
+static drm_intel_bufmgr *mbufmgr[NUM_FD];
+static struct intel_batchbuffer *mbatch[NUM_FD];
+static drm_intel_bo *mbuffer[NUM_FD];
+
 /*
  * Testcase: Basic check of ring<->cpu sync using a dummy reloc
  *
@@ -124,6 +131,50 @@ dummy_reloc_loop_random_ring(int num_rings)
 	}
 }
 
+static void
+dummy_reloc_loop_random_ring_multi_fd(int num_rings)
+{
+	int i;
+	struct intel_batchbuffer *saved_batch;
+
+	saved_batch = batch;
+
+	srandom(0xdeadbeef);
+
+	for (i = 0; i < 0x100000; i++) {
+		int mindex;
+		int ring = random() % num_rings + 1;
+
+		mindex = random() % NUM_FD;
+		batch = mbatch[mindex];
+
+		if (ring == I915_EXEC_RENDER) {
+			BEGIN_BATCH(4);
+			OUT_BATCH(MI_COND_BATCH_BUFFER_END | MI_DO_COMPARE);
+			OUT_BATCH(0xffffffff); /* compare dword */
+			OUT_RELOC(mbuffer[mindex], I915_GEM_DOMAIN_RENDER,
+					I915_GEM_DOMAIN_RENDER, 0);
+			OUT_BATCH(MI_NOOP);
+			ADVANCE_BATCH();
+		} else {
+			BEGIN_BATCH(4);
+			OUT_BATCH(MI_FLUSH_DW | 1);
+			OUT_BATCH(0); /* reserved */
+			OUT_RELOC(mbuffer[mindex], I915_GEM_DOMAIN_RENDER,
+					I915_GEM_DOMAIN_RENDER, 0);
+			OUT_BATCH(MI_NOOP | (1<<22) | (0xf));
+			ADVANCE_BATCH();
+		}
+		intel_batchbuffer_flush_on_ring(batch, ring);
+
+		drm_intel_bo_map(target_buffer, 0);
+		// map to force waiting on rendering
+		drm_intel_bo_unmap(target_buffer);
+	}
+
+	batch = saved_batch;
+}
+
 int fd;
 int devid;
 int num_rings;
@@ -133,6 +184,7 @@ igt_main
 	igt_skip_on_simulation();
 
 	igt_fixture {
+		int i;
 		fd = drm_open_any();
 		devid = intel_get_drm_devid(fd);
 		num_rings = gem_get_num_rings(fd);
@@ -148,6 +200,40 @@ igt_main
 
 		target_buffer = drm_intel_bo_alloc(bufmgr, "target bo", 4096, 4096);
 		igt_assert(target_buffer);
+
+		/* Create multi drm_fd and map one gem object to multi gem_contexts */
+		{
+			unsigned int target_flink;
+			char buffer_name[32];
+			if (dri_bo_flink(target_buffer, &target_flink)) {
+				printf("fail to get flink for target buffer\n");
+				igt_assert_f(0, "fail to create global "
+					     "gem_handle for target buffer\n");
+			}
+			for (i = 0; i < NUM_FD; i++) {
+				sprintf(buffer_name, "Target buffer %d\n", i);
+				mfd[i] = drm_open_any();
+				mbufmgr[i] = drm_intel_bufmgr_gem_init(mfd[i], 4096);
+				igt_assert_f(mbufmgr[i],
+					     "fail to initialize buf manager "
+					     "for drm_fd %d\n",
+					     mfd[i]);
+				drm_intel_bufmgr_gem_enable_reuse(mbufmgr[i]);
+				mbatch[i] = intel_batchbuffer_alloc(mbufmgr[i], devid);
+				igt_assert_f(mbatch[i],
+					     "fail to create batchbuffer "
+					     "for drm_fd %d\n",
+					     mfd[i]);
+				mbuffer[i] = intel_bo_gem_create_from_name(
+								mbufmgr[i],
+								buffer_name,
+								target_flink);
+				igt_assert_f(mbuffer[i],
+					     "fail to create gem bo from global "
+					     "gem_handle %d for drm_fd %d\n",
+					     target_flink, mfd[i]);
+			}
+		}
 	}
 
 	igt_subtest("render") {
@@ -190,8 +276,27 @@ igt_main
 			printf("dummy loop run on random rings completed\n");
 		}
 	}
-
+	igt_subtest("mixed_multi_fd") {
+		if (num_rings > 1) {
+			sleep(2);
+			printf("running dummy loop on random rings based on "
+					"multi drm_fd\n");
+			dummy_reloc_loop_random_ring_multi_fd(num_rings);
+			printf("dummy loop run on random rings based on "
+					"multi drm_fd completed\n");
+		}
+	}
 	igt_fixture {
+		int i;
+		/* Free the buffer/batchbuffer/buffer mgr for multi-fd */
+		{
+			for (i = 0; i < NUM_FD; i++) {
+				dri_bo_unreference(mbuffer[i]);
+				intel_batchbuffer_free(mbatch[i]);
+				drm_intel_bufmgr_destroy(mbufmgr[i]);
+				close(mfd[i]);
+			}
+		}
 		drm_intel_bo_unreference(target_buffer);
 		intel_batchbuffer_free(batch);
 		drm_intel_bufmgr_destroy(bufmgr);
Signature · 1 line
1.7.10.1
#3 lore raw
In reply to Zhao Yakui
Quoted 15 lines · Zhao Yakui

This follows Daniel's advice to add the two test cases based on multi drm_fd to test the ring sync and CPU<->GPU sync.
The Broadwell GT3 machine has two independent BSD rings that can be used to process the video commands. This is implemented in kernel driver and transparent to the user-space. But we still need to check the ring sync and CPU<->GPU sync for the second BSD ring. Two tests are created based on the multi drm_fds to test the sync. Multi drm_fd can assure that the second BSD ring has the opportunity to dispatch the GPU command.

V1->V2: Follow Daniel's comment to add one subtext instead of one individual test case, which is used to test the CPU<->GPU sync under multi BSD rings/

V2->V3: Follow Imre's comment to remove the unnecessary initialization and use igt_assert_f instead of igt_assert

Quoted 7 lines

V3->V4: Add gem_multi_bsd_sync_loop.c into the tests/.gitignore

Zhao Yakui (2):
tests: Add one ring sync case based on multi drm_fd to test ring
semaphore sync under multi BSD rings
tests/gem_dummy_reloc_loop: Add one subtest based on multi drm_fd to

    test CPU<->GPU sync under multi BSD rings

Merged, thanks for the patches.
-Daniel

Quoted 15 lines
 tests/.gitignore                |    1 +
 tests/Makefile.sources          |    1 +
 tests/gem_dummy_reloc_loop.c    |  107 +++++++++++++++++++++++-
 tests/gem_multi_bsd_sync_loop.c |  175 +++++++++++++++++++++++++++++++++++++++
 4 files changed, 283 insertions(+), 1 deletion(-)
 create mode 100644 tests/gem_multi_bsd_sync_loop.c
Signature · 6 lines
1.7.10.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch