batchbuffers failing to execute
I think I've got a concrete test here for something I've been trying to track down for a while now: it appears that when the GPU is too busy, batches get dropped and sometimes the GPU hangs.
This really became clear when I was testing a patch for citybench to "improve" the swap throttling -- a broken version that I submitted caused no throttling to occur at all. The app, which renders a predefined set of frames as fast as possible, would now frequently stutter in the middle. The results were insanely fast (40-60% better!). But it didn't make sense for there to be a stutter if the set of frames rendered is predefined.
So, I wrote this little patch series to try to test my guess that my batches were getting dropped. The plan is: Emit a dword write every (render) batch, and after a while go look and see if they all landed. Corresponding to the stutters is:
Mesa: Initializing x86-64 optimizations
batch 42 didn't report: 0xd0d0d0d0 instead of 0x0000002a batch 43 didn't report: 0xd0d0d0d0 instead of 0x0000002b batch 44 didn't report: 0xd0d0d0d0 instead of 0x0000002c batch 45 didn't report: 0xd0d0d0d0 instead of 0x0000002d batch 46 didn't report: 0xd0d0d0d0 instead of 0x0000002e batch 47 didn't report: 0xd0d0d0d0 instead of 0x0000002f batch 48 didn't report: 0xd0d0d0d0 instead of 0x00000030 batch 49 didn't report: 0xd0d0d0d0 instead of 0x00000031 batch 50 didn't report: 0xd0d0d0d0 instead of 0x00000032 batch 51 didn't report: 0xd0d0d0d0 instead of 0x00000033 batch 52 didn't report: 0xd0d0d0d0 instead of 0x00000034 batch 53 didn't report: 0xd0d0d0d0 instead of 0x00000035 shutting up
The cool thing is apparently I don't need citybench, either. This paste came from vblank_mode=0 glxgears fullscreened on my 1600x900 panel (default size doesn't reproduce the failure).
This is gen7, danvet's kernel de67cba65944f26c0f147035bd62e30c5f456b96 rebased on top of cherry-picks of the 4 ivb workaround patches. A revert of a71d8d94525e8fd855c0466fb586ae1cb008f3a2 doesn't help.
I've also run this on a test libdrm that returns NULL pointers on map failure and asserts that exec and pwrite don't return errors.
PATCH 1/2 Try to ensure that batches emitted actually complete.
--- src/mesa/drivers/dri/i965/brw_state_upload.c | 2 + src/mesa/drivers/dri/intel/intel_batchbuffer.c | 73 ++++++++++++++++++++++++ src/mesa/drivers/dri/intel/intel_batchbuffer.h | 2 + 3 files changed, 77 insertions(+), 0 deletions(-)
src/mesa/drivers/dri/i965/brw_state_upload.c +2 -0
@@ -435,6 +435,8 @@ void brw_upload_state(struct brw_context *brw)int i;static int dirty_count =0 ; +record_batch (intel); + state->mesa |= brw->intel.NewGLState; brw->intel.NewGLState =0 ;
src/mesa/drivers/dri/intel/intel_batchbuffer.c +73 -0
94 lines not on this page · raw message
src/mesa/drivers/dri/intel/intel_batchbuffer.h +2 -0
@@ -132,6 +132,8 @@ intel_batchbuffer_advance(struct intel_context *intel)void intel_batchbuffer_cached_advance (struct intel_context *intel); +void record_batch (struct intel_context *intel); +/* Here are the crusty old macros, to be removed: */ #define BATCH_LOCALS
Signature · 1 line
1.7.9
PATCH 2/2 intel: Disable swap throttling entirely.
In reply to Eric AnholtThis reveals some major breakage in the kernel.
--- src/mesa/drivers/dri/intel/intel_context.c | 22 ---------------------- 1 files changed, 0 insertions(+), 22 deletions(-)
src/mesa/drivers/dri/intel/intel_context.c +0 -22
@@ -411,28 +411,6 @@ intel_prepare_render(struct intel_context *intel)*/ if (intel->is_front_buffer_rendering) intel->front_buffer_dirty = true; - -/* Wait for the swapbuffers before the one we just emitted, so we -* don't get too many swaps outstanding for apps that are GPU-heavy -* but not CPU-heavy. -* -* We're using intelDRI2Flush (called from the loader before -* swapbuffer) and glFlush (for front buffer rendering) as the -* indicator that a frame is done and then throttle when we get -* here as we prepare to render the next frame. At this point for -* round trips for swap/copy and getting new buffers are done and -* we'll spend less time waiting on the GPU. -* -* Unfortunately, we don't have a handle to the batch containing -* the swap, and getting our hands on that doesn't seem worth it, -* so we just us the first batch we emitted after the last swap. -*/ -if (intel->need_throttle && intel->first_post_swapbuffers_batch) { -drm_intel_bo_wait_rendering (intel->first_post_swapbuffers_batch); -drm_intel_bo_unreference (intel->first_post_swapbuffers_batch); - intel->first_post_swapbuffers_batch =NULL ; - intel->need_throttle = false; - } } static void
Signature · 1 line
1.7.9
Quoted 3 lines · Eric Anholt
This is gen7, danvet's kernel
de67cba65944f26c0f147035bd62e30c5f456b96rebased on top of cherry-picks of the 4 ivb workaround patches. A revert ofa71d8d94525e8fd855c0466fb586ae1cb008f3a2doesn't help.
You need the bug fix that was meant to be applied before that patch id:[email protected].
See also https://bugs.freedesktop.org/show_bug.cgi?id=45492
-Chris
Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
Quoted 9 lines · Chris Wilson
Quoted 3 lines · Eric Anholt
This is gen7, danvet's kernel
de67cba65944f26c0f147035bd62e30c5f456b96rebased on top of cherry-picks of the 4 ivb workaround patches. A revert ofa71d8d94525e8fd855c0466fb586ae1cb008f3a2doesn't help.You need the bug fix that was meant to be applied before that patch id:[email protected].
Meh, stupid me somehow got the idea that we'd check the new request->tail before the b0rked autoreport_head and hence merging just the new code for -next would fix this desaster without merging in -fixes. Otherwise I'd been slightly more clear in my -next announcement ...
/me hides in shame
-Daniel
Signature · 3 lines
Daniel Vetter Mail: [email protected] Mobile: +41 (0)79 365 57 48