[Mesa-dev] [PATCH 0/4] RadeonSI: Upload constants to VRAM via SDMA

[Dieter Nützel]

Am 12.02.2019 05:10, schrieb Dieter Nützel:
> Am 12.02.2019 03:22, schrieb Dieter Nützel:
>> Am 12.02.2019 00:40, schrieb Dieter Nützel:
>>> Sorry that I step in so late, but the whole family recover slowly 
>>> from
>>> a bad flu...
>>> 
>>> Tried your 'latest" three series altogether with my Polaris 20 
>>> (NIR!).
>>> UH and UV hang after some seconds reliable. VM faults. Have to dig
>>> deeper in (remote) to get some logs.
>> 
>> UH
>> 
>> [47001.185090] amdgpu 0000:01:00.0: GPU fault detected: 147 0x0b384801
>> for process heaven_x64 pid 18565 thread heaven_x64:cs0 pid 18586
>> [47001.185094] amdgpu 0000:01:00.0:
>> VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x0373EF67
>> [47001.185096] amdgpu 0000:01:00.0:
>> VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x06048001
>> [47001.185098] amdgpu 0000:01:00.0: VM fault (0x01, vmid 3, pasid
>> 32786) at page 57929575, read from 'TC4' (0x54433400) (72)
>> [47011.401741] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx
>> timeout, signaled seq=11380701, emitted seq=11380703
>> [47011.401784] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* Process
>> information: process  pid 0 thread  pid 0
>> [47011.401787] amdgpu 0000:01:00.0: GPU reset begin!
>> [47021.631605] [drm:amdgpu_dm_atomic_check [amdgpu]] *ERROR*
>> [CRTC:49:crtc-0] hw_done or flip_done timed out
> 
> These GPU faults are SOLVED after reverting the SDMA (1-4) series.

So I gave this a second change with LLVM 9.0 git.
+ some other patches

e83af67eed7 (HEAD -> master) ac: use new LLVM 8 intrinsic when loading 
16-bit values
7f32d569ffc ac: add ac_build_llvm8_tbuffer_load() helper
037bda54a7d nir: remove simple dead if detection from nir_opt_dead_cf()
51fe88ff1ab radeonsi/nir: set shader_buffers_declared properly
e66a73aa1a6 radeonsi/nir: set colors_read properly
83955dfc81a radeonsi/nir: set input_usage_mask properly
4c355a562db radeonsi: use SDMA for uploading data through const_uploader
6855f871e47 gallium/u_upload_mgr: allow use of FLUSH_EXPLICIT with 
persistent mappings
2116355fc01 gallium/u_threaded: always unmap const_uploader
6e70cce39f3 st/mesa: always unmap the uploader in st_atom_array.c
22a88ca1d92 radeonsi: re-initialize query buffers if they are reused
6775665e5ee (origin/master, origin/HEAD) spirv: Eliminate dead 
input/output variables after translation.

UH
run some sences but (same? - Yes.) GPU fault. - Shit, sadly overwritten 
my dmesg.log. :-(

UV
run some sences but (same? - Yes.) GPU fault.

Unigine Valley Benchmark 1.0 (1.0)Unigine~# world_load valley/valley
Loading "valley/valley.cpp" 126ms
Loading "valley/valley.mat" 72 materials 1160ms
Loading "valley/sound/sound.prop" 142 properties 1ms
Loading "valley/valley.world" 2253ms
valley_x64: ../src/gallium/auxiliary/util/u_inlines.h:81: 
pipe_reference_described: Assertion `count != 1' failed.

[ 1079.415836] amdgpu 0000:01:00.0: GPU fault detected: 147 0x0ca04801 
for process valley_x64 pid 18050 thread valley_x64:cs0 pid 18071
[ 1079.415841] amdgpu 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR  
  0x094A9594
[ 1079.415842] amdgpu 0000:01:00.0:   
VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x08048001
[ 1079.415845] amdgpu 0000:01:00.0: VM fault (0x01, vmid 4, pasid 32769) 
at page 155882900, read from 'TC4' (0x54433400) (72)
[ 1089.543336] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx 
timeout, signaled seq=91489, emitted seq=91491
[ 1089.543379] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* Process 
information: process valley_x64 pid 18050 thread valley_x64:cs0 pid 
18071
[ 1089.543382] amdgpu 0000:01:00.0: GPU reset begin!
[ 1099.773342] [drm:amdgpu_dm_atomic_check [amdgpu]] *ERROR* 
[CRTC:49:crtc-0] hw_done or flip_done timed out

Hope that helps some.

Dieter