Graf commitů

1651 Commity

Autor SHA1 Zpráva Datum
German Andryeyev eef47ca24a SWDEV-353281 - Initial support of memalloc in graph
Add memory allocation support in graph. Current implementation uses
cache from mempool  to hold the allocations which belong to the graph.
Also the resource tracking is disabled at this moment because mempool
operates with hip::Stream objects, but graph has execution with
amd::HostQueue objects.

Change-Id: I54fe3250126d24f5a26ada975f37d429bb4ef17b
2023-01-13 13:06:59 -05:00
Satyanvesh Dittakavi 873bc03034 SWDEV-369552 - SWDEV-369553 - SWDEV-369555 - Fix hipMemAdvise
- API should return error if the size passed is more than the
bytes allocated
- Passing invalid device should return error except for the
advise hipMemAdviseSetReadMostly

Change-Id: I1c4c15136b825215bf1e1f5ee7a9a08e5895d870
2023-01-13 12:53:31 -05:00
Rakesh Roy a427d3cd10 SWDEV-374369 - Fix hipGraphAdd*Node* APIs invalid behavior
- Return hipErrorInvalidValue if dependencies array contains a node from
  different graph

Change-Id: I0dcfaaaeb50a89848ac546a0b86b185977e2128c
2023-01-13 12:09:42 -05:00
Anusha GodavarthySurya 33eef6d95e SWDEV-364576 - use hipLimitMallocHeapSize to set initial heap buffer size
Change-Id: I2f1d3b229bc28571ce6c55f93f8ef73dda516b50
2023-01-12 23:34:29 -05:00
Jaydeep Patel bb486b04c6 SWDEV-374383 - Check from and to if no of dependencies are more than 0, Check from and to node are belongs to graph pass in.
Change-Id: I0cdafb497147f9ac7f84d40203a21e9eefd595a3
2023-01-11 11:43:22 -05:00
Rakesh Roy b321e2afc6 SWDEV-368881 - Implement hipModuleLaunchCooperative* APIs
- Add implementation for hipModuleLaunchCooperativeKernel, hipModuleLaunchCooperativeKernelMultiDevice APIs

Change-Id: I2a80264e458549211c34b413397a08b57d021147
2023-01-10 13:50:20 -05:00
Ajay c6a7ce66bb SWDEV-369556 - Unit_hipMemGetAddressRange_Negative cases
- return value for invalid device ptr
- offset greater than size

Change-Id: I8bcf878c3c1c0a29836ea664bfb9048e1114b74e
2023-01-10 06:57:54 -05:00
Jiabao Xie 03200953e5 SWDEV-374375 - moving nullptr check into hipGraphNode::isValidNode
Change-Id: Ic9264521d6f0b36ffc948ca9feeaa70eb6e98eb6
2023-01-09 16:27:38 -05:00
Jacob Lambert 7dd54c746e SWDEV-356171 - Adding support for name lowering for bitcodes (-fgpu-rdc) via
Comgr mangled name fetching APIs

Change-Id: I10ce2b1288c459e3e61218d68a270aba2148376e
2023-01-09 14:49:51 -05:00
kjayapra-amd 0c8c08cf51 SWDEV-361374 - Adding support for hipPointerSetAttributes
Change-Id: I56710b808854acefca07b3b5898deebee38063e9
2023-01-09 12:36:57 -05:00
Juan Manuel MARTINEZ CAAMAÑO c391af49dc SWDEV-290637 - Specify a working directory when calling hip_embed_pch.sh
Consider the case where cmake is invoked as follows:

    cd <parent_dir>
    cmake -S<source_dir> -B<build_dir> ...

  If `WORKING_DIRECTORY` parameter is not specified in `execute_command`,
  `hip_pch.o` is created outside `<build_dir>`, in `<parent_dir>/hip_pch.o`.

Change-Id: I751f4390628315ad70c36148e09824c19c93ae9e
2023-01-09 03:17:07 -05:00
German f39655c0c8 SWDEV-372757 - Don't destroy null queue in MT
Skip destruction of the null queue for MT. Windows can destroy worker
thread on exit during the stream destruction, causing a race condition.

Change-Id: I3e4bc3ab2b477d38c0aec1fc59e6f1af0f8f6b0f
2023-01-09 00:03:09 -05:00
Jiabao Xie e5490697cc SWDEV-374375 - Added checks to see if hip graph nodes are valid
Change-Id: I9e6eabf0e4a9100de6f9d26bf5b815957f12d4f2
2023-01-06 16:58:56 -05:00
pghafari 7615fba2d5 SWDEV-375135 - adding error string
Change-Id: I73276462627a47d405c8439d8a843ac266fc5d38
2023-01-06 16:31:35 -05:00
pghafari 335a24d1a0 SWDEV-369042 - updating to GetError to match cuda
Change-Id: I0ec2330443b26cb1c8cedba942f31fb1267cd09d
2023-01-06 05:33:41 -05:00
Juan Manuel MARTINEZ CAAMAÑO 537df06c4f SWDEV-290384 - Missing hiprtcResult case in hiprtcGetErrorString
Change-Id: I11d18ff19f6f6f1b36febf87a586bc9ffb95af10
2023-01-06 03:38:44 -05:00
Satyanvesh Dittakavi 516acd4e60 SWDEV-375020 - Revert "SWDEV-370552 - Improve hiprtc performance by passing the right flags to comgr"
This reverts commit 627e689ad7.

Reason: Adding "-Xclang" and "-disable-llvm-passes" in hiprtc
compilation seem to cause perf issues and memory fault in few apps.

Change-Id: I12ff46293c7b7afb24d570351736d884d2881b77
2023-01-06 00:09:06 -05:00
Ajay 3e4ed8d917 SWDEV-369618 - hipOccupancyMaxPotentialBlockSizeVariableSMemWithFlags invalid flag scenario
Change-Id: I07df2640be71a6b0aed160d54e80ffdeb85ec554
2023-01-05 23:40:29 -05:00
Jaydeep Patel 7c32f66e67 SWDEV-374372 - Return invalid value if pitch is less than width in bytes.
Change-Id: I33806d747cd344250d02e217de8e9b6d5a7f83c1
2023-01-05 10:40:14 -05:00
Jaydeep Patel 67792dd858 SWDEV-374373 - Occupied size should be within range.
Change-Id: I20e24c628c51154c2f8f36dca0a6fa4ead1ee6d0
2023-01-05 08:02:20 -05:00
Anusha GodavarthySurya 067c7316ba SWDEV-373173 - Correct max VGPRs and VGPR Granularity in Occupancy calculation for different asics
Change-Id: Ib213ed8767a793a832776fba7c0811c2f023a8ad
2023-01-04 01:34:40 -05:00
German Andryeyev 38ad6234f9 SWDEV-360031 - Erase destroyed stream from capture
Parallel streams can be destroyed during capture. Make sure runtime
removes the destoryed streams from the list of parallel captured streams

Change-Id: I46b6cbb951e8711f0bf02f2826c3b890b2541ffe
2023-01-03 13:34:07 -05:00
Satyanvesh Dittakavi 409f23d905 SWDEV-374371 - hipGraphAddMemsetNode should return error if width is 0
Change-Id: I1cd4be271f65c9cade8dddb6d00191e1b74c96bc
2023-01-03 10:29:16 -05:00
sdashmiz ce60d47b61 SWDEV-370309 - GetNodes should do levelorder
- GetNode should retrun nodes based on the dependency between them

Signed-off-by: sdashmiz <shadi.dashmiz@amd.com>
Change-Id: Ie8ecdd976f8dc9d9b5dc2adc6ae8f879912c8169
2023-01-03 04:54:55 -05:00
Jaydeep Patel f2b0542442 SWDEV-374370 - Propogate element size to blit kernel.
Change-Id: I8a3355c62947265be34e6a2befd26f86b3dc6cd2
2023-01-01 23:54:09 -05:00
pghafari a643c199de SWDEV-356549 - del Null strm in destroyAllStreams
Change-Id: Ie04172ee39e6333247d07791dffda753a8b9ef41
2022-12-29 15:45:54 -05:00
Rakesh Roy 95b2441cbc SWDEV-374789 - Validate HostQueue before accessing it
- hip::getQueue will return nullptr when comgr fails to load blit kernels.

Change-Id: I3bea1f5c13301d4a35e40ab7c6faa73f9c6aecf4
2022-12-28 01:04:10 -05:00
pghafari 2c906460cb SWDEV-369580 - hipModuleGetTexRef matching cuda
Change-Id: I3f071fe0b289c0ba8316c691f066f47d12b156b5
2022-12-27 13:58:26 -05:00
Ioannis Assiouras 1278cf555e SWDEV-374393, SWDEV-373631 - Added input checks for hipMemcpy2D/3D
Change-Id: I5ae57326c35b599d9ecec05e2a22bc6086027322
2022-12-22 11:21:55 -05:00
Ioannis Assiouras 99d2d016c1 SWDEV-374355, SWDEV-374356, SWDEV-374357, SWDEV-374358, SWDEV-374359 - Added input sanity checks to hipExternalMemoryGetMappedBuffer
Change-Id: I61e43dece3fb9bd961e92db8e015f08170327538
2022-12-22 10:46:10 +00:00
Jaydeep Patel c4003ef62a SWDEV-374398 - Pitch should not be less than width for both src and dst.
Change-Id: Ib8fa8e3bd0c655f3de213b3798d369ba866b2f01
2022-12-21 23:08:55 -05:00
Anusha GodavarthySurya 707ca92a80 SWDEV-373147 - Return error when getDeviceKernel returns nullptr
Change-Id: Ie9fcaa2bff87cb8ef363dc7e8362b74b8725b3b6
2022-12-21 22:14:10 -05:00
Ioannis Assiouras d29a2e545d SWDEV-374392 - Make hipMemCpy2/3DAsync synchronous for any HostToHost and DeviceToPageableHost
Change-Id: Iaa411681a49427a9e9b3a06e90abb0b32f5a6851
2022-12-21 14:59:39 -05:00
Ioannis Assiouras b83271b4ca SWDEV-374402 - Added check for invalid events in hipExtLaunchKernel
Change-Id: I8dfd69aa1ef38971a72b98c723d2099347a66d11
2022-12-21 12:37:12 -05:00
Jaydeep Patel 845bf0adb4 SWDEV-374382 - Change hip error type to context is destroyed instead for invalid stream.
Change-Id: I5ed22463204d7bff3741599eeced3d030da78fc4
2022-12-21 12:54:39 +00:00
Jaydeep Patel 148f67a36d SWDEV-374394 - Correct hip error for invalid stream.
Change-Id: I96949c58e050cb3149ab8434aebe0f1a97a6e857
2022-12-20 21:27:56 -05:00
Ioannis Assiouras f4f377bcaa SWDEV-374402 - Added input stream check for hipExtLaunchKernel
Change-Id: I42915c3b9b10b4a4a5b3ae74b6a2dccc8fddbf7f
2022-12-20 14:29:58 -05:00
Ioannis Assiouras 97c5f772c6 SWDEV-374400, SWDEV-374404 - Change error message in hipFuncGetAttribute(s) to match nvidia.
Change-Id: I95b69a49ced24115f0aebc288ea067a49b770aca
2022-12-20 12:49:17 -05:00
pghafari 3b143cf3a2 SWDEV-369575, SWDEV-369568, SWDEV-369574 - matching
matching cuda return value

hipModuleLaunchKernel and hipExtModuleLaunchKernel matching with cuda in
negative cases

Change-Id: I636070924c81116dccbfa74f693061c60530de45
2022-12-20 12:24:23 -05:00
Jaydeep Patel 027f71ccfb SWDEV-374395 - Correct rhs.
Change-Id: I433d60344a6d23bc48f4db58d5501af89bc1989c
2022-12-20 11:38:11 +00:00
Satyanvesh Dittakavi 627e689ad7 SWDEV-370552 - Improve hiprtc performance by passing the right flags to comgr
Change-Id: I99674b033d1248d9b34ac598b4780bd84a38c279
2022-12-19 06:16:56 -05:00
pghafari c0f779fdf1 SWDEV-369576, SWDEV-369578, SWDEV-369577, SWDEV-369579, SWDEV-369563 - matching cuda return value
SWDEV-369576, SWDEV-369578, SWDEV-369579, SWDEV-369563 - matching cuda return value

Change-Id: I21b598e959571979307238c485492eb7fe2388cc
2022-12-16 01:12:49 -05:00
kjayapra-amd 2f8847bfbd SWDEV-363865 - Use COMGR path to unbundle code object in HIPRTC.
Change-Id: Ifca29ea0c6ff319c3f584397e542a1658d220c4a
2022-12-15 07:16:55 -05:00
Jacob Lambert 2835ff1a63 SWDEV-322000 - Removing scoped locks in code regions with Comgr calls
Comgr now internally locks to avoid multi-threading issues with LLVM

Change-Id: I74b903d373af55c2efade17a9caf31fda3ca1795
2022-12-14 12:40:45 -05:00
Ioannis Assiouras 30e740023b SWDEV-369548 - Disallow opening an ipc mem handle in the originating process.
Change-Id: Icce5e41a220566141bd7d92b7199d477f60bed8e
2022-12-14 04:52:21 -05:00
Ioannis Assiouras 1274f2a54c SWDEV-373235 - Fixed segfault when nullptr function is passed to hipKernelNameRef
Change-Id: If23fbbd19ad92fd0b3fa888dca09e7fe705c0c56
2022-12-13 07:09:03 -05:00
Anusha GodavarthySurya 0fcc5557bd SWDEV-325711 - Node can be enabled/disabled only for kernel, memcpy and memset nodes
Change-Id: Id9c19a14655732d61fccee9c3d1ffa0614123468
2022-12-13 05:08:38 +00:00
Anusha GodavarthySurya 5bcfb07c85 SWDEV-372521 - Fix copy of unused kernel attributes
Change-Id: Ib93f27d68cedac2182ba6dac0c234cc71c2c4229
2022-12-13 03:44:10 +00:00
kjayapra-amd f7c194edf7 SWDEV-371904 - Adding pseudo fine grain flag to hsa memory allocation for device fine grained memory.
Change-Id: I8c605d6bc3ac905bfffc6de3a56911151ef0e663
2022-12-12 11:40:00 -05:00
pghafari 698f52acd8 SWDEV-369566, SWDEV-369565, SWDEV-369564 - match with cuda return
Change-Id: I308f5344a64e5a06aa89685de9348f6722c196da
2022-12-12 09:53:18 -05:00