Граф коммитов

5088 Коммитов

Автор SHA1 Сообщение Дата
Reshabh Sharma bb62f8553f Output file name should not change flags picked for compiler (#1938)
Fixes SWDEV-207362,

The output file name should not contribute to picking up the right flags for the compiler. This fix solves issues when the output has conflicting extensions which confuses hipcc to treat them as the source files and add the required flags for them.

PS: Output file refers to the file followed by -o

Change-Id: I1095966c11143ad73e81fabc35b4e9de5d3afada
Example: hipcc test.o -o test.hip will add the flags for .hip compilation ignoring the fact that it is an output file
2020-03-31 16:13:43 -04:00
Michael Hong Bin Liao 5b96f38c1e Merge "[hipcc] Remove the previous workaround." into amd-master-next 2020-03-31 10:05:56 -04:00
Michael LIAO b72196613a [vdi] Fix hipGetSymbol{Address|Size}
- Use symbol value as the qeury key. Compared to the symbol name, the
  symbol value is more robust as developers may use unqualified or
  qualified identifiers. It also removes the mangling and/or demangling
  requirement for the runtime API.

Change-Id: I9d4259f3842612c7cc98551269fc2092d8b5c19e
2020-03-31 00:26:53 -04:00
Christophe Paquot 94a7ef6ed1 Do not retry to allocate when OOM. Shouldn't be needed since we idle on Free.
SWDEV-229214

Change-Id: I183006f409388e3c7981f2569649d01d6378be46
2020-03-30 12:49:48 -07:00
kjayapra-amd b081912962 SWDEV-216213 - Use different static & dynamic module maps for faster lookup.
Change-Id: Ia605e76a411ad5be04046b9d61f1ac111d49bb4a
2020-03-30 14:28:07 -04:00
Anusha Godavarthy Surya caa083357a Merge "Update Enable/Disable peers to match cuda behaviour" into amd-master-next 2020-03-30 13:46:42 -04:00
agodavar bdb3a4b393 Update Enable/Disable peers to match cuda behaviour
Change-Id: I67194ccf77a0019368579ff7d95b7790fcf228f3
2020-03-30 12:49:16 -04:00
Tao Sang 1697cfc790 Merge "enable HCC printf when using hip-clang" into amd-master-next 2020-03-30 11:37:23 -04:00
Michael LIAO f1566f6dc5 [hipcc] Remove the previous workaround.
Change-Id: Ie1c2aa21689bc343c10fd4b699546cf500ced5b2
2020-03-30 10:18:47 -04:00
Michael LIAO 3a690e960f [vdi] Fix calculation of MaxWaves
- Consider the case where `usedVGPRs` is zero.
- This fixes [SWDEV-228537](http://ontrack-internal.amd.com/browse/SWDEV-228537)

Change-Id: I8675311f5fe24fb59c5d45bada122afefb55b128
2020-03-30 09:10:16 -04:00
Sameer Sahasrabuddhe 9a0c5d0653 enable HCC printf when using hip-clang
This is cherry-picked from PR#1947 that was committed to the
github repo. It allows printf to work with hip-clang and HCC
runtime.

Change-Id: I754753250ea1e694cf3441722e2d4c9d25fa75bc
2020-03-28 00:18:21 -04:00
Saleel Kudchadker 5a8add03a5 Merge "Check event status before notify" into amd-master-next 2020-03-27 20:19:40 -04:00
Vladislav Sytchenko 6ed73f50f7 (SWDEV-228794)
Adjust the origin of the copy if the user passes a pointer that wasn't allocated by the runtime.

Change-Id: I0aeb20195ed730857a461a53f537626ec2573fd1
2020-03-27 16:33:16 -04:00
Vladislav Sytchenko fd7a8f0367 (SWDEV-228794)
Add hipMallocHost()

Change-Id: Ia3c7c5ca94b39fe30f3a51d1b60782d3472259ff
2020-03-27 15:57:48 -04:00
Vladislav Sytchenko a91b82f00e (SWDEV-228782)
The only requirment from hipMallocPitch() is that the returned pitch is aligned to the HW image pitch alignment. There is no restriction on the size of the allocation, since the memory might not be used for images.

Change-Id: I97438e5fe4012ca4721b14b85f514dbac803c17c
2020-03-27 15:52:17 -04:00
Tao Sang af7d131749 Merge "Support hipFuncGetAttributes with hip-clang+Hcc RT" into amd-master-next 2020-03-27 14:36:41 -04:00
Saleel Kudchadker 178fc17846 Check event status before notify
Change-Id: I68f6bbbf236e49b859be2d5afbe0c8282fe15dd3
2020-03-27 11:32:46 -07:00
Tao Sang d967be8eeb Support hipFuncGetAttributes with hip-clang+Hcc RT
Fix issues of missing kernel function symbols and missing argument list via
using __hipRegister* functions.
Then the following tests can pass,
directed_tests/runtimeApi/module/hipFuncGetAttributes
directed_tests/runtimeApi/module/hipExtLaunchMultiKernelMultiDevice
directed_tests/gcc/LaunchKernel

Change-Id: I52135b61e8283eb4f9f10f77895151e4e55418d9
2020-03-26 23:28:35 -04:00
Vladislav Sytchenko 2a2b9e47a0 Add support for formating hipExtent objects
Change-Id: Iea54a510e81a856c0c450305b3e5a7179ee48295
2020-03-26 14:45:20 -04:00
Vladislav Sytchenko e0187ba405 Add initial entry points for mipmapped array API
Change-Id: Icd59cc7323ddcb6773da6105260415a1e6f4cdcb
2020-03-26 14:45:20 -04:00
Vladislav Sytchenko c1475f948e Replace hip::TextureObject with __hip_texture
This avoids the use of extra casts when obtaining a texture object handle.

Change-Id: I42df22bdad0ab9ac6c33cb8b282dee65fe7cfd6e
2020-03-26 14:45:20 -04:00
Vladislav Sytchenko 2028b6eb29 Headers need to export C symbols for texture API
This also adds declarations of all the missing texture APIs.

hipTexRefSet*() functions need to take a textureReference as a ptr for type erasure to work. Runtime has been modified to accomodate this.

This change only applies to VDI.

Change-Id: Icf43cc5bd44dfc2c39084b7fe56d5a793bf7319f
2020-03-26 14:45:20 -04:00
Vladislav Sytchenko 8ddeeb4551 Enable initial sRGB support
Instead of using the sampler field force_degamma to perform sRGB->linear conversion during pixel sampling, we use an appropriate image format instead. The overhead of this is having to create an image view when creating a texture object from an array.

Change-Id: I1ca368c312c1fd4b6f784a3a1b35b5eeb28070ff
2020-03-26 14:45:20 -04:00
Vladislav Sytchenko faf3b83594 Handle offsets for dptr <-> image copies
Change-Id: I7a4a56ee07a26a741d2aac35502446d248f720ad
2020-03-26 14:45:20 -04:00
Vladislav Sytchenko fbe30090a2 Correctly format hipResourceDesc objects
The struct consists of a union - only the active object should be read.

Change-Id: I1c40965b61518acd91a2dcbae92a015ac9be346a
2020-03-26 14:45:20 -04:00
Vladislav Sytchenko ced0582a52 Set textureObject to nullptr
This avoids dangling pointers for newly initiazlied textures

Change-Id: Ia444b91fe17fd756ed583ec595ae1febbdfbd034
2020-03-26 14:45:20 -04:00
Vladislav Sytchenko 39558dc1b9 Modify formatting for textureReferences
We don't program the numChannels and format members (these are HCC specific), so printing these will only display garbage.

Change-Id: I83dc8be9a3cae2659c64f4594d07c05330d2dd14
2020-03-26 14:45:20 -04:00
Vladislav Sytchenko b09fe1280e Correct typos in texture function declarations
Change-Id: I492995e984eda2e8a5e806c5d4c9c78da09ac483
2020-03-26 12:43:17 -04:00
Vladislav Sytchenko ea701777d8 Allow creating texture from unaligned user ptr
All we have to do is align the ptr to HW requirments an if it's not zero, then return the offset to the user.

We currently don't have anywhere to store this offset, so hipGetTextureAlignmentOffset() will still always return 0.

Change-Id: If31998127d99a2a3222a026d88249519d6102505
2020-03-26 12:43:17 -04:00
Payam Ghafari 8e8b372726 Merge "updated cmake to create libamdhip64 static file as well" into amd-master-next 2020-03-26 01:10:12 -04:00
jujiang 3774a1b106 SWDEV-226708-hipMemcpyPeerAsync failed-->Update test app to create stream on proper device
Change-Id: Ia9f51435ffb5a8fbbff39e363acecabb338bf34c
2020-03-25 17:22:56 -04:00
Payam 76703acb1d updated cmake to create libamdhip64 static file as well
Change-Id: I2054b9501cefa232abbf398524ab62450ab6805d
2020-03-25 16:37:57 -04:00
Vladislav Sytchenko e7d91b8230 Disable failing tests for VDI
The hipExtModuleLaunchKernel and hipModuleLoadDataMultiThreaded tests keeps randomly failing on Jenkins.

Change-Id: I87e5d54fb7429c14ff1dcecb20e03a7816670fae
2020-03-23 17:09:57 -04:00
Christophe Paquot 3dce22538c Merge "Enable VDI Occupancy test" into amd-master-next 2020-03-23 13:38:57 -04:00
Saleel Kudchadker 91069c0de4 Merge "Sync streams when freeing or destroying mem" into amd-master-next 2020-03-21 13:33:41 -04:00
Christophe Paquot 653277bd3f hipStreamAddCallback test seg faults
Change-Id: If419d2fad490d0ed50eb1315af809fc1deda1ce3
SWDEV-227875: Add a lock in streams to lock when the callback is call so we make sure things aren't moving forward in the stream
2020-03-20 13:07:34 -07:00
Saleel Kudchadker 8b39e0b74e Sync streams when freeing or destroying mem
Change-Id: I6932f225a8b932bb2adbd5e37880f7e604496809
2020-03-20 10:53:23 -07:00
Christophe Paquot d98879a250 Merge "hipStreamAddCallback test seg faults" into amd-master-next 2020-03-20 12:05:37 -04:00
kjayapra-amd cd92bd7fee SWDEV-216213 - Lookup module functions from PlatformState::functions_.
Change-Id: I91dfe327f2ebdcf4c9b39ddd14d60aa0ce2fa9f4
2020-03-20 11:52:28 -04:00
Christophe Paquot 3dfbfc408b hipStreamAddCallback test seg faults
Change-Id: I1f107fc8a5c586cd571f0280ed8716c5f89d25b7
SWDEV-227875: Need to add a dummy marker in case the stream is empty.
2020-03-19 11:11:59 -07:00
Sarbojit Sarkar 3baff10618 Enable VDI Occupancy test
Change-Id: Ia89155ca909b9b7346584d19962fa0a94033bb17
2020-03-19 10:43:40 +05:30
Vladislav Sytchenko 4829a7c215 Add support for creating typed buffers
What Cuda refers to "linear texture memory" is the OpenCL equivalent of CL_MEM_OBJECT_IMAGE1D_BUFFER. For these types of allocations we should create a typed buffer instead of an image.

Currently there is no check in the texture fetch functions as to what kind of SRD is written into the texture object, so any kind of incorrect programming will cause the TA to hang. Fortunately for us, every one writes correct code :)

Change-Id: I80dab85a992f2c0754ebf303d40ac6b5e045c7c1
2020-03-18 18:15:17 -04:00
Vladislav Sytchenko e8fa3b2589 Program texture flags in a better way
Not sure what I was thinking when initially implementing this...

Change-Id: Ib82f0f5a86683c08823dd4b59c98259d27151822
2020-03-18 18:15:09 -04:00
Vladislav Sytchenko 2bad9e2821 Purge the use of ihip*impl() texture APIs
These are artifacts left from HIP-HCC and now are not needed by HIP-VDI.

Change-Id: Ib25a1081fe6146c8a89659395151e9d5bdaf7519
2020-03-18 18:15:01 -04:00
Vladislav Sytchenko 5429b40afe Rework the texture C++ API
Currently the texture C++ API is forwarded to the ihip*Impl() calls, which are not even a part of Cuda. These should be forwarded to their respective Cuda C APIs instead.

This change also fixes a bug with hipUnbindTexture() creating a dangling pointer.

Change-Id: Ifafc9d106855a11bec84a18ea214b3d89e39990d
2020-03-18 18:14:53 -04:00
Vladislav Sytchenko 7385a032ae Correctly infer the texture read mode
Currently we extract the read mode from the ihip*impl() calls, which is not correct. We should be getting it from the texture itself directly.

Change-Id: Idf6449fefa395a887138a252e8ea937a6897e600
2020-03-18 18:14:45 -04:00
Vladislav Sytchenko c7407a3b57 Correct the definition of ...
hipBindTextureToMipmappedArray()

The texture reference needs to be passed as a constant pointer.

Change-Id: I6d31204c7f2325a5bc1e8b6e089fd9f8d21d1d78
2020-03-18 18:14:36 -04:00
Vladislav Sytchenko 3e460ab514 Correct the declaration of hipBindTexture2D()
The texture reference needs to be passed as a constant pointer.

Change-Id: Idde461f0f328ac87ce677b6bab3203161b514cbf
2020-03-18 18:08:23 -04:00
Vladislav Sytchenko 2d77399747 Correct the declaration of hipBindTextureToArray()
The texture reference needs to be passed as a constant pointer.

Change-Id: Iff171626536071fb2020cfff7132ec930577b1b9
2020-03-18 18:08:13 -04:00
Vladislav Sytchenko 7190fa518e Correct the declaration of hipBindTexture()
The texture reference needs to be passed as a constant pointer.

Change-Id: I36ca0bddaba30becfc2ce70dd9e5b7db66c57f27
2020-03-18 18:08:01 -04:00