AI/rocm-systems
2
0
포크 0
코드 이슈 풀 리퀘스트 액션 패키지 프로젝트 릴리즈 위키 활동
4,397 커밋 3 브랜치 0 태그
b8ab281765bbfa2e3ef03b7de6cfc1e5af8d9331
커밋 그래프

4397 커밋

이 브랜치
이 브랜치
모든 브랜치
작성자 SHA1 메시지 날짜
German Andryeyev b8ab281765 SWDEV-317061 - Fix 32-bit HSAIL complib load 
Use __stdcall decl to match the library defines

Change-Id: Id4178443d5a9bb4e5401d80b3b5d15c7bbea0330


[ROCm/clr commit: e0a4e0df0e]
 2022-01-04 16:48:18 -05:00
Chauncey Hui da523f184f SWDEV-2 - Change OpenCL version number from 3409 to 3410 
[ROCm/clr commit: c04d650b67]
 2021-12-29 03:00:09 -05:00
haoyuan2 74c05e9bc1 SWDEV-316150 - align with OCL to use wavefront64 on Navi21/22/23/24 
Change-Id: Ic47bc7cc6300c217c02c2bbbda9940a3b5c9597d


[ROCm/clr commit: 3d5d9e2691]
 2021-12-28 12:39:18 -08:00
Chauncey Hui 729e99fc65 SWDEV-2 - Change OpenCL version number from 3408 to 3409 
[ROCm/clr commit: a6948a31f2]
 2021-12-19 03:00:03 -05:00
Saleel Kudchadker 500f6a6513 SWDEV-313306 - Fix Co-operative groups dtests 
Add a state indicator to retain ExternalSignals when needed.
Co-operative group launch uses external signals to indicate a dependency
to the next command.

Change-Id: I6d0daa006e2377c3bbf4aeca0fd5b63c7ac8fbbb


[ROCm/clr commit: 1fbd75b825]
 2021-12-17 12:41:37 -08:00
Chauncey Hui 27a162656e SWDEV-2 - Change OpenCL version number from 3407 to 3408 
[ROCm/clr commit: e5122e3bc7]
 2021-12-15 03:00:05 -05:00
Saleel Kudchadker 42625f0527 SWDEV-313306 - Clear external signals 
Crash was due to the fact that external signal structure was stale even
after destroyign the command. That is because we skipped wait due to a
missing check.
Detect external signals and dispatch a barrier in ReleaseGpuMemoryFence.
Also clear external_signals_ at ProfilingBegin.

Change-Id: I991387edcfe928b511bf5e780988ee131321ed5a


[ROCm/clr commit: 3239222516]
 2021-12-13 23:03:33 -08:00
Chauncey Hui 09b628c4e0 SWDEV-2 - Change OpenCL version number from 3406 to 3407 
[ROCm/clr commit: 1b5de6e562]
 2021-12-11 03:00:08 -05:00
German Andryeyev c759986e28 SWDEV-300515 - Apply CU granularity on the CU count 
Report granularity for possible app query

Change-Id: I98857c6f4cc7ae590927ea35ce57d181abe7860b


[ROCm/clr commit: f613831471]
 2021-12-10 10:47:28 -05:00
Chauncey Hui aac7266191 SWDEV-2 - Change OpenCL version number from 3405 to 3406 
[ROCm/clr commit: 7abd2827b5]
 2021-12-10 03:00:08 -05:00
haoyuan2 248a738674 SWDEV-290298 - add a flag to indicate the primary context active status 
Change-Id: Ia31790706d3f855bc1eedf5ef874e471


[ROCm/clr commit: 439af94dd9]
 2021-12-09 23:28:54 -05:00
Sarbojit Sarkar 4630f3ade0 SWDEV-314254 - Fix for hipMemcpy3D test crash 
Change-Id: Iac70bfe0d351cfb5b56fefc9a6487d3f26f2b4ef


[ROCm/clr commit: aedbad0109]
 2021-12-09 11:46:52 -05:00
Chauncey Hui 9b2abd8370 SWDEV-2 - Change OpenCL version number from 3404 to 3405 
[ROCm/clr commit: 84af5eed4e]
 2021-12-09 03:00:06 -05:00
German Andryeyev 5ad02b78c4 SWDEV-305016 - Improve MGPU scaling in Tensorflow 
Add a threshold for ROCR/SDMA P2P transfers. ROCR copy path
requires extra barriers in compute for synchronization. That costs
extra performance with tiny transfers.
Reduce active wait time to 10us. Tensorflow uses extra thread
per GPU with constant hipEventQuery() calls. Longer active waits
in ROCr affect CPU performance.

Change-Id: I9020358438615fa2d4617f862f00a562f0a588e7


[ROCm/clr commit: 008133cf41]
 2021-12-08 11:59:37 -05:00
Chauncey Hui 421d659e35 SWDEV-2 - Change OpenCL version number from 3403 to 3404 
[ROCm/clr commit: 587cda0c91]
 2021-12-08 03:00:05 -05:00
Julia Jiang 41eef3076e SWDEV-308644 - reorganize extra blit kernel in PAL stack 
Change-Id: I9d853e8d417ef75b522184d83646ec4b9fa8669b


[ROCm/clr commit: 376ea1e293]
 2021-12-07 14:55:20 -05:00
Christophe Paquot 1b87d3a497 SWDEV-307109 - Don't force Persistent for HIP 
With SAM on, don't force Persistent for allocations
in HIP. This makes ROCCLR go down paths we don't want
for HIP.

Change-Id: If54cc16fa891d4cfdc761c6ab21ad707627e822a


[ROCm/clr commit: 5243552768]
 2021-12-06 12:16:08 -05:00
Chauncey Hui 5dab7d3843 SWDEV-2 - Change OpenCL version number from 3402 to 3403 
[ROCm/clr commit: 1a677effa8]
 2021-12-03 03:00:05 -05:00
Satyanvesh Dittakavi b219bf8292 SWDEV-309286 - save the max system mem size in device info 
Change-Id: I9955625aca6ceda059aef6354d909de636b610ba


[ROCm/clr commit: 9dabdcdc3e]
 2021-12-02 10:59:07 -05:00
Chauncey Hui fb217aa6d0 SWDEV-2 - Change OpenCL version number from 3401 to 3402 
[ROCm/clr commit: 1df0ddefe4]
 2021-12-02 03:00:03 -05:00
Saleel Kudchadker 97456a157b SWDEV-308843 - Increase MaxPinnedXferSize to 128 
This allows experimenting with env var GPU_PINNED_XFER_SIZE which is
still at a default of 32MB

Change-Id: I85ade700ed58d498eba29d1737601dc74d4c26a4


[ROCm/clr commit: 3f82b99f5d]
 2021-12-01 20:37:56 -05:00
anusha GodavarthySurya 02e46ec231 SWDEV-284895 - Adding kind metadata and launch init/fini marked kernels 
Change-Id: If2b21c4b98567632c426943e0b69aca8d6f1ec2a


[ROCm/clr commit: 102aa9d6d9]
 2021-12-01 08:17:44 -08:00
Chauncey Hui 0ef7ea3ad9 SWDEV-2 - Change OpenCL version number from 3400 to 3401 
[ROCm/clr commit: 3d56881a14]
 2021-11-30 03:00:03 -05:00
Sarbojit Sarkar c8a6920b91 SWDEV-310181 - Fix for AtoH Memcpy tests failure 
Change-Id: Ibf8c8c01257f0516088d50d5c9f82040ed8fa067


[ROCm/clr commit: 02dc6f9f9a]
 2021-11-29 22:55:23 -05:00
kjayapra-amd f75cfb049a SWDEV-312822 - Fix the globalWorkSize to number of sizeof(var) instead of bytes. 
Change-Id: Ic6b2bbb2e8d4cb6aa8d906d4b93cd06a176160d8


[ROCm/clr commit: d4ad981c0c]
 2021-11-29 17:36:11 -05:00
German Andryeyev 861b9fb84c SWDEV-294669 - Avoid stall when the new signal was created 
Stall in the host thread could occur earlier than the app expects.
Make sure rutnime can grow the signals to the queue size without
any stall. Also adding a new signal to the end of the pool could
break the dependency chain on signal reuse. The new logic will
insert the new signal after current to keep the chain intact.

Change-Id: I9c90b98515907db8b677528263c3e88cd9581a14


[ROCm/clr commit: 102c19adf3]
 2021-11-29 10:08:06 -05:00
Chauncey Hui ed0b9b60ed SWDEV-2 - Change OpenCL version number from 3399 to 3400 
[ROCm/clr commit: bfcb47ca72]
 2021-11-29 03:00:05 -05:00
kjayapra-amd f74515778c SWDEV-312822 - Revert "SWDEV-310187 - Change flag to keep track of aligned sizes instead of expanded patterns." 
This reverts commit 7220267211.

Change-Id: I022c2a8375f9929e9723cec66e1e0b960263fc39


[ROCm/clr commit: 2e9bc8f793]
 2021-11-28 23:39:40 -05:00
Chauncey Hui af0788d2e9 SWDEV-2 - Change OpenCL version number from 3398 to 3399 
[ROCm/clr commit: 2c23a1091f]
 2021-11-27 03:00:08 -05:00
German Andryeyev b0b0c3049f SWDEV-313126 - Use data() method for the base array address 
Reference for the first element can trigger an assert with
_GLIBCXX_ASSERTIONS build

Change-Id: I59c63c052831307edfe5dcc6384798a43e9596dd


[ROCm/clr commit: 6f2e7c3199]
 2021-11-26 09:51:57 -05:00
Chauncey Hui 3577900318 SWDEV-2 - Change OpenCL version number from 3397 to 3398 
[ROCm/clr commit: e4be77cbef]
 2021-11-26 03:00:05 -05:00
Julia Jiang ea2741f631 SWDEV-308644 - merge roc blit kernels 
Change-Id: I378e511959fe17c03fa45066022e9670a4d181f0


[ROCm/clr commit: f5c9ad5b1d]
 2021-11-25 10:07:51 -05:00
Chauncey Hui 4d011fd9ae SWDEV-2 - Change OpenCL version number from 3396 to 3397 
[ROCm/clr commit: 964f5bc502]
 2021-11-25 03:00:04 -05:00
Satyanvesh Dittakavi 1caf1e4936 SWDEV-292714 - Add unique id for each allocation to support HIP_POINTER_ATTRIBUTE_BUFFER_ID 
Change-Id: Ibb3fcb2d0bbbef03525fc884d5b3e9b5f6c11423


[ROCm/clr commit: c56317b2e0]
 2021-11-24 06:10:50 -05:00
Chauncey Hui b3edecda2c SWDEV-2 - Change OpenCL version number from 3395 to 3396 
[ROCm/clr commit: 0ab4ad1310]
 2021-11-24 03:00:04 -05:00
Saleel Kudchadker dc7a87fad8 SWDEV-299893 - Fix PAL/lnx build 
Change-Id: If4408a81d2dc8686c8e805a7697a7d9b2589c864


[ROCm/clr commit: 1e6cc5a2fa]
 2021-11-23 08:09:33 -08:00
Chauncey Hui 3af9ff75f6 SWDEV-2 - Change OpenCL version number from 3394 to 3395 
[ROCm/clr commit: f526e9ea89]
 2021-11-23 03:00:03 -05:00
Sarbojit Sarkar 04745adf7a SWDEV-310181 - Fix for mGPU dtest failure 
Change-Id: Id0898bd45e23f2d637bef25a3e69f26d9dc40785


[ROCm/clr commit: 2afeacc858]
 2021-11-22 01:01:47 -05:00
Chauncey Hui b9d61c4cd8 SWDEV-2 - Change OpenCL version number from 3393 to 3394 
[ROCm/clr commit: 7fdeb37239]
 2021-11-20 03:00:06 -05:00
kjayapra-amd 7220267211 SWDEV-310187 - Change flag to keep track of aligned sizes instead of expanded patterns. 
Change-Id: I763feda8688bb1b7b11033a2a8cba0f69f07167d


[ROCm/clr commit: 8307886644]
 2021-11-19 10:32:40 -05:00
Saleel Kudchadker 04a391004a SWDEV-299893 - Set preferred node affinity 
Set affinity to the closest node of the current GPU. This reduces
the latency to fetch kernel args since device would query the CPU cache
of core which did the dispatch. This behavior is controlled with
AMD_CPU_AFFINITY env var(disabled by default)

Change-Id: I65afba62cb818ea25a311b88d1c0dd5c51330292


[ROCm/clr commit: b192beea52]
 2021-11-19 04:42:42 -05:00
Chauncey Hui 8ff18d5e6f SWDEV-2 - Change OpenCL version number from 3392 to 3393 
[ROCm/clr commit: 6cf5089f37]
 2021-11-19 03:00:05 -05:00
anusha GodavarthySurya 2c1173ea32 SWDEV-240806 - hipGraph performance create new graph commands for every launch 
Change-Id: Ifd4a373d6a76118ae0946238b29accfacbe32937


[ROCm/clr commit: ef1ec6ffde]
 2021-11-19 00:09:47 -05:00
German Andryeyev c116411e00 SWDEV-294669 - Avoid queue drain 
Use slot wait logic for direct dispatch

Change-Id: I431ba1418eb4aa066b9881934f4055b3d338ce3a


[ROCm/clr commit: 8e4101b4fd]
 2021-11-18 13:06:12 -05:00
Chauncey Hui a3e56d08b7 SWDEV-2 - Change OpenCL version number from 3391 to 3392 
[ROCm/clr commit: 874513375f]
 2021-11-18 03:00:10 -05:00
pghafari e7df603b2a SWDEV-297142 - HIP-Vulkan - linux interop buffer 
Change-Id: I0278e56bba632024c214beb9e1758587ccba0927


[ROCm/clr commit: e38a200bf7]
 2021-11-17 06:06:58 -05:00
Chauncey Hui 0e39d6f41a SWDEV-2 - Change OpenCL version number from 3390 to 3391 
[ROCm/clr commit: 13ef90af61]
 2021-11-17 03:00:15 -05:00
German Andryeyev 95c01935b8 SWDEV-305016 - Correct timeout logic 
Timeout logic should be applied always even if the wait is active.

Change-Id: I2e5db7ac8a0f9a0355ad7b40e4227d76fb002aa0


[ROCm/clr commit: 9877fc9dbf]
 2021-11-16 23:04:34 -05:00
Julia Jiang e3f6db3d64 SWDEV-308644 - update blit kernel setup in rocm 
Change-Id: Iaa9ff97b3ed7d379189c359696be932a83cf203c


[ROCm/clr commit: ef3d6f7b28]
 2021-11-15 13:28:07 -05:00
Chauncey Hui a7edd1e407 SWDEV-2 - Change OpenCL version number from 3389 to 3390 
[ROCm/clr commit: 02f36b565b]
 2021-11-13 03:00:04 -05:00