AI/rocm-systems
2
0
Fork 0
Koodi Ongelmat Pull-pyynnöt Actions Paketit Projektit Julkaisut Wiki Toiminta
3 820 Commitit 3 Branchit 0 Tagit
d943cae31f1b0989ae23a1b0c4f4fc2aa6be3bf2
Commit Graph

16 Commitit

Tekijä SHA1 Viesti Päivämäärä
Jason Tang d943cae31f Add CommandKindString to the log 
Change-Id: Ie23123a85cff82b1732da85f5bffbff6958c02e5
 2020-10-26 09:16:03 -04:00
Laurent Morichetti 080dcfe857 Improve queueLock and lastCmdLock 
Reduce the size of the queueLock and lastCmdLock critical sections
to improve lock contention performance. The smaller the critical
sections are the better.

lasCmdLock is still needed to guarantee that getLastEnqueueCommand_
can retain the command before it is swapped out and released.

Change-Id: Id35d4a77c035b2da0de4c15568b153d49e958bb7
 2020-09-01 18:09:31 -04:00
Laurent Morichetti c95c613edc Fix indentation with clang-format 
Change-Id: I7aeadef3c613d5efc31a98e666bfb819ae34bdf5
 2020-09-01 18:09:19 -04:00
Saleel Kudchadker 1c24072d13 Revert "SWDEV-241977 [ROCm QA] Random Soft hang observed while running TF and Caffe2 benchmarks" 
This reverts commit ce038f3163.

Change-Id: Ib56493c92eca793f1dfb6f1cbefb32f0b4f65e89
 2020-09-01 18:09:10 -04:00
Alex Xie ce038f3163 SWDEV-241977 [ROCm QA] Random Soft hang observed while running TF and Caffe2 benchmarks 
Change-Id: I42016c11db15411b86e7b8130d6ba557bc22dbb7
 2020-07-22 02:03:48 -04:00
Christophe Paquot 3d15a1e291 Make append and setLastQueuedCommand atomic 
Two threads can enqueue to the same HostQueue (HostQueue::enqueue)
and result in last queued command being the first one reachine queue_.enqueue

NOTE: Temporarly make setLastQueuedCommand empty function to pass the build

Change-Id: Id09c3a28d184986f52b2ec86a2f6a18c40df1f0b
 2020-07-14 18:22:45 -04:00
German Andryeyev c18892a590 Remove extra barriers 
Don't flush current batch if the dependent wait is a nop

Change-Id: I8a8722b9011fe042c1a4ce195938290fc75e7c86
 2020-06-22 12:41:02 -04:00
German Andryeyev 44bc0cb35d Revert "Avoid lock for last queued command" 
This reverts commit dc4e09a63a.

Reason for revert: <INSERT REASONING HERE>

Change-Id: Ie10442c9447f010bb90c679b6cffca5b48b8d054
 2020-06-04 18:08:17 -04:00
German Andryeyev dc4e09a63a Avoid lock for last queued command 
Use atomics for last queued command update

Change-Id: I759e9d78ea72f23c0d45dbede6250b231e122276
 2020-05-29 11:06:55 -04:00
Christophe Paquot 0782acabb5 Use a dedicated lock for last queued command set/get 
Change-Id: If3d2144841c7863cf7afe2ca85aea62e0a3a33c7
 2020-05-28 12:49:39 -07:00
Aryan Salmanpour fed94b8604 Add support for setting CU mask on ROCclr for ROCm backend 
Change-Id: I0dbe2eeb33467fc0f24b26929119c10e9b455da7
 2020-05-15 14:23:43 -04:00
Payam 1b6f21ad9a removing AMD emails per palamida scan 
Change-Id: If7307f5b1f81a43f2725ec5abd3b8989cbddbcc5
 2020-03-11 21:26:55 -04:00
German Andryeyev 77ff161324 SWDEV-193956 - [hipclang-vdi-rocm][perf] 
Retain command before a wait.

Change-Id: I0ab8f1352af32c96567dad9e31b3687bdf407bf6
 2020-02-21 16:43:21 -05:00
German Andryeyev a66d09f5a3 SWDEV-193956 - [hipclang-vdi-rocm][perf] 
~45% to 50% of Performance drop on rocBLAS_int8 test

Use the last command in the queue for a wait.
Add extra print information about processed commands.
Add an option to disable file location printing.

Change-Id: I4187883e1a90e571fde3128af98368108fda8785
 2020-02-21 15:21:15 -05:00
Laurent Morichetti b4c6143a2f Update copyright info 
Change-Id: Ia4f9ff0f5f873b4223a8cca154188bb0d2f1abba
 2020-02-04 09:26:14 -08:00
Laurent Morichetti 20c7173849 Merge branch 'origin/pghafari/vdi-prototype' into lmoriche/amd-master 
Change-Id: Id3b833d405596735becb3346f3b08c6da57033fe
 2020-01-30 20:12:13 -08:00