Commit Graph

122 Commits

Author SHA1 Message Date
kjayapra-amd 9c75797d9f SWDEV-249499 - Add message to Guarantee assert code in hip runtime.
Change-Id: I1850463675bbc3e6ff5e3cef8211b1fd40e55f96
2020-08-24 17:35:39 -04:00
Todd tiantuo Li 99eb486937 SWDEV-240803 - add hipFuncSetSharedMemConfig
Change-Id: I160b04677b3e7b99b3981ae7ecc84a0e3811d5e8
2020-08-20 18:18:24 -04:00
Todd tiantuo Li efcb882e19 move hipFuncSetCacheConfig from hip_device to hip_module
Change-Id: If5a930e1210e76531f5ba9a0e5f5ec98ad473a19
2020-08-17 17:39:01 -04:00
Aryan Salmanpour 4f400bc5e9 SWDEV-248057 - fix the calculation of allGridSize used in multi_grid_group() API
Change-Id: Ib470094e28dcacaa4769dc5c7ab08924f5b7fa41
2020-08-17 12:52:40 -04:00
Jatin Chaudhary 1340b3f07f Adding Anyorder flag to HIP
Change-Id: Ie20931541b3febe23fa9ac36ebc0c90de75a5f0a
2020-08-17 08:46:52 -04:00
Todd tiantuo Li 9dfe15a843 SWDEV-240803 - add hipFuncSetAttribute and hipFuncAttribute
Change-Id: I3f4d67b19d89fd348fa5b884af4a2542ee4aba60
2020-08-14 17:39:29 -04:00
kjayapra-amd 6f8909be34 SWDEV-237658 - Do not attach events to user invisible Marker.
Change-Id: I3a74fd2614d6e2e1d51543471bbd5ae51a713a47
2020-08-14 10:55:34 -04:00
Freddy Paul 6e50c0f755 Enable developers to build hip without CMAKE_PREFIX_PATH
With recent changes in ROCclr to take HSA cmake target it require
HIP to have CMAKE_PREFIX_PATH added for install location for HSA,
amd_comgr,llvm.

User can provide CMAKE_PREFIX_PATH if above
packages are built/installed at custom location as:
cmake -DCMAKE_PREFIX_PATH="<path_hsa> <path_amd_comgr> <path_llvm>"

For default ROCm installation adding a default path will help.

Change-Id: Iefa443dc198da76b57422a641c9821c63cde4922
2020-08-13 23:13:40 -04:00
Jason Tang e2169ce6d0 SWDEV-232197 - Remove refering to gfxipVersion_
Change-Id: I91b6b972e26b18dabf6deda5a142bff491c8bb45
2020-08-12 16:34:18 -04:00
German Andryeyev 505877acea Allow prefetch on the null stream
Change-Id: I87e950460361fab65ab140988fc2288bcb194f08
2020-08-12 13:07:43 -04:00
Rahul Garg d620a25b64 Return error from hipMallocManaged for size =0
Change-Id: Ida0cecf05e1df731176d41e59422d1e606cc9bd9
2020-08-10 18:25:31 +00:00
Vlad Sytchenko 844dab3b22 Don't align 2D image buffer size
Cuda allows creating 2D textures from memory with unaligned size.
Aligning it will cause us to overshoot the size of the underlying
allocation.

Change-Id: Ia20dec2105e36ca4a8448c9c6299d0001b388d25
2020-08-09 15:53:52 -04:00
Jatin Chaudhary 36b420ce68 Fixing the memchannel calculation
Change-Id: Ib5f5931cc0e1ee5b7cb356c2ede0cb951ee8c9de
2020-08-07 05:34:23 -04:00
Saleel Kudchadker 2ef062480b Enable queue profile only if we attach a profiler
Submit explicit profile marker for hipEventRecord to record
timestamps. Enable explicit signal profiling if the API specifies
start and stop events.

Toggle this with env var HIP_FORCE_QUEUE_PROFILING=0

Change-Id: Iae449a63ec3ebf6c2880e65d7b1dd1031a29018f
2020-08-06 11:28:16 -04:00
Payam 39d681b6be adding duration to print log for hip api calls
Change-Id: Icffa8fcacfb8ba24861d980b9553faeee4efec8c
2020-08-06 06:42:21 -04:00
Sarbojit Sarkar d768406ec4 Added free device memory info
Recently there were few OOM(out of Memory) issues poped. Adding trace will help know genuine OOM issue. 
Change-Id: Ie2a56bba95cead24caf1af5c807e73c383c2f5d9
2020-08-06 01:16:36 -04:00
Anusha Godavarthy Surya 9ea54f7797 SWDEV-243961 Added overflow check for globalWorkSize
Change-Id: I153d99398eb4619baad8c0a52466e0d5d5f898ca
2020-08-05 11:18:07 -04:00
Jason Tang 678b3eeed5 SWDEV-243576 - Fix hipMemcpy regression.
If the queue device doesn't match the device on any memory object, use the queue device from the memory object.

Change-Id: I5fdcf00494f8391574f4443332c01788b8da44ef
2020-07-28 03:04:00 -04:00
Rahul Garg dc904abd8a Remove libhip_hcc symlinks to libamdhip64
Change-Id: I7b5b47fc7ca9760316c2424c16b454d7919c665c
2020-07-27 23:58:29 +00:00
jujiang 448008348e SWDEV-242344-hipMemcpyAll failed on MGPU
Change-Id: I01671e322c84213964d3d48bd6b5dc704dba731a
2020-07-27 13:30:30 -04:00
Tao Sang ef9176db03 Remove __gnu_h2f_ieee and __gnu_f2h_ieee
Change-Id: Ide24d245d851e20961020323e52f33322a33fff9
2020-07-23 14:50:24 -04:00
Vlad Sytchenko ed26013ec1 2D/3D copy optimizations
SWDEV-244798

If {src/dst} ptr is marked as hipMemoryTypeHost, check if the memory was
prepinned. If it was, upgrade the copy type to hipMemoryTypeDevice to
avoid extra pinning.

Change-Id: Id287ef5b14ae67dfbcf80c4caa1b08a311191948
2020-07-23 11:24:16 -04:00
kjayapra-amd dab399f22c SWDEV-236178 - Remove use of old routines functions()/modules().
Change-Id: I3a1ed967227c91b0d8cdf39e1360ade685e0bc73
2020-07-22 14:55:02 -04:00
kjayapra-amd 4d42a969fd SWDEV-240800 - Adding initial support for hipDeviceGetP2PAttribute. Some attr support still pending.
Change-Id: I0611aed136270db497dfa374144f6f5e35352a8f
2020-07-22 14:54:34 -04:00
Payam f8e6b228a7 modify cmake to strip debug symbols from rel blds
Change-Id: Ia67261fc3c4a20f3f1bbe439a70384a3f6aefa91
2020-07-22 11:53:54 -04:00
Tao Sang fc5ccbe13d Support numa if libnume-dev is installed
Let rocclr decide numa dependence

Change-Id: I65bdfba7ec0d06b550f86632318bcfd1f765cfa9
2020-07-22 08:55:28 -04:00
Saleel Kudchadker f897f003cf Print devPtr and size for hip*Symbol* api
Change-Id: I07be6949cca005700ee1d1708d86220c793c9728
2020-07-21 13:49:10 -04:00
Rahul Garg 0fdec26a29 Handle size 0 symbols
Change-Id: Ie282b37cf6bbccae4611dfc17c5c7f31977feea7
2020-07-21 10:09:26 -04:00
Aryan Salmanpour 2e72f989bd Add missing stream null check for some hipStreamCreate APIs
Change-Id: I716d71e4ec59b0bd7922869bfa0ed908c22c289e
2020-07-20 18:07:03 -04:00
jujiang 3b7e1b928e SWDEV-242337-Fix hipHostRegister test failure for MGPU
Change-Id: I748dac0c5a47fa00184f065d9ef716573aa70b12
2020-07-20 11:40:49 -04:00
German Andryeyev d491dbd796 Grab the event lock before fetching the last command
Hip applications assume that hipEventRecord called from multiple
threads will contain exactly the last queued command to the stream.

Change-Id: I1da3259f143d7670d0870d9a47c08e32336b2222
2020-07-20 02:01:16 -04:00
Christophe Paquot 0d83b385d1 Remove use of setLastQueuedCommand
Change-Id: I07bf54dc09905d9a17b059a476df9282eb6f1f51
2020-07-17 11:45:25 -04:00
Evgeny 920e6597b2 cleanup: unneeded '\'
Change-Id: I0276312fdb0e4829d678de4281ccfd12eb7f232e
2020-07-16 01:57:07 -05:00
Ashutosh Mishra ab88cbbb8a Adding "BUILD_SHARED_LIBS" flag to cmake files
JIRA : SWDEV-234471
With this change compilation of the ibamdhip64 to be decied by arguments to build_hip_on_rocclr.sh file
build_hip_on_rocclr.sh -r (current) shall conitnue to create libamdhip64.so;
build_hip_on_rocclr.sh -r a (new) shall create libamdhip64.a
other options remain the same

Changed the targets according to requirement of stage2 dependencies.
Changed find_lib with find_package

Adpated the review comments

Change-Id: I719918f73ab33cd8031a32db6fe9b8adaa1b4b87
2020-07-16 00:29:14 -04:00
Sarbojit Sarkar b2b5a821df host pointer check [SWDEV-243104, SWDEV-243206]
Change-Id: Id58dabfe5c35e5f8e42d224e700a431a4a97d003
2020-07-14 07:16:12 -04:00
Tao Sang 66fca8be23 Implement HIP_ENABLE_DEFERRED_LOADING in right way
Change-Id: Ia28dbdea5cde93b3ca69f44c6a3c18ec8a730d42
2020-07-11 18:14:03 -04:00
kjayapra-amd 7233bb75da SWDEV-240165 - Move all amd::MemObjMap_ reference to ROCclr and only allow base ptr to get ipc handle.
Change-Id: I4f11e56ba455a7ce5a066c0e9e49657c5220a211
2020-07-09 21:19:50 -04:00
German Andryeyev b9b982247e Host device context should return a valid queue
Host memory allocations use shared device context, which allows
to have access from any device. That breaks default queue detection.

Change-Id: I27ad1234d49ab0ea8f5edd6f2e3ffa93c47d4b1f
2020-07-09 10:56:33 -04:00
Aryan Salmanpour 9121a74ac1 add an error message before aborting if a wrong GPU target was used for compilation
Change-Id: I7b41e93d420dbc47978238a681f2956a0def01cb
2020-07-07 14:50:44 -04:00
Jatin Chaudhary 24554e6755 Fix error: cannot pass object of non-trivial type string through variadic function when building in debug mode
Change-Id: Idd55ed422addc815fb56c57b05bd4afa9563f2ac
2020-07-06 00:48:40 -04:00
kjayapra-amd 6f15d8ca6b SWDEV-236178 - Fixing func_attributes that was over written on redesign.
Change-Id: I157913ae4fa4a3a149817e901f10475b7750188b
2020-07-01 11:52:45 -04:00
German Andryeyev f2102adc99 Remove hipMemcpyDeviceToDevice check.
There is hipMemcpyDefault that requires validation also.

Change-Id: I5ae0ffa1c023b7774547acbdc049b84a0c2cb4c5
2020-06-30 12:59:54 -04:00
Evgeny a19960e79a HIP-PROF: fixing pointers tracing
Change-Id: Id2c35972e1d134012980c95a3649ddfdf9303494
2020-06-23 14:02:35 -04:00
agodavar 93da1b1f11 SWDEV-239297:hipMemset fix, getMemoryObject with unaligned ptr
Change-Id: I37b8139c5099d37b42f34d728a3d694fdf9cb55d
2020-06-23 08:56:09 -04:00
German Andryeyev 44c1df802e Disable flush on marker
Change-Id: I04d0fd995fb4b29a6bc372e1543902a053017d6a
2020-06-22 13:27:39 -04:00
Tao Sang 0eebbb4f87 Fix static lib crashing issue
Initialize HIP_ENABLE_DEFERRED_LOADING = 1

Change-Id: I432ccf1926b953c4391fdd086cde7da2380962bb
2020-06-22 13:24:40 -04:00
Tao Sang 63051ca2e1 Support numa policy set by user
Add hipHostMallocNumaUser flag to hipHostMalloc() in order to support
numa policy set by user.

Change-Id: I6d70ed539a5f97f27187f2242b68849c0e27e4d6
2020-06-19 21:23:58 -04:00
Vlad Sytchenko e9033edc10 Correct texture element size calculations
hipChannelFormatDesc contains size of each channel in bits. There's 8 bits per byte, not 4, so we should divide by 8.

Change-Id: I38ae2d28ad54740823c98908ca2afecad0ba0183
2020-06-19 11:39:03 -04:00
Vlad Sytchenko 205e4c1c5f Support creating textures from user ptr with offset
Change-Id: Ied0af9a842fc6ef067e068a9d89dde63ff5dfce1
2020-06-17 14:10:14 -04:00
Vlad Sytchenko a6aef2a36a Fix -Wsign-compare warnings
Change-Id: I874dc007ac657c25a72c6752c1a2da74c028a822
2020-06-17 11:20:45 -04:00