b7803eb0c6
static inline in a header, just like excess sugar in a diet, causes bloat (#1692 )
...
[ROCm/hip commit: be70b9f7e7
2019-12-23 19:09:38 +05:30
ef3ee5908f
Fix for windows dtest build failure ( #1742 )
...
[ROCm/hip commit: e2fc00da65
2019-12-19 13:10:43 -08:00
ded857e2b9
[hip] Add macro guarding the enum conversion for scalar accessor. ( #1748 )
...
- That's a high overhead part, which needs enabling ONLY if necessary.
[ROCm/hip commit: 99a3b66110
2019-12-19 10:08:37 -08:00
ca99107c7b
[HIP] Unify hipError_t (Step 1)
...
Step 1. Set the same values for RT error codes as for analogous Driver's
[Reason] RT's and Driver's error codes unification in CUDA 10.2
[ROCm/hip commit: d8737ba50c
2019-12-13 19:40:16 +03:00
e1f55b0cee
Only add hipExtLaunchMultiKernelMultiDevice for non-HCC compilers. ( #1729 )
...
[ROCm/hip commit: 444c931641
2019-12-10 10:32:25 -08:00
65f80f80a1
fix linking of vector types with gcc ( #1690 )
...
* fix linking of vector types when linking hipcc objects with gcc
* use __atribute__((vector_size)) with both clang and gcc
and reinstate nonaligned n=3 vector type
* use implicit conversion to value and ext_vector_type when available
* Alternate formulation for GCC compatibility
* Built-in arrays don't mix well with placement new
* Fix typo
* Add conversions to enum
* Fix Scalar_accessor assignment.
* Update hip_vector_types.h
* stir up the underlying_type hideous mess
This fixes the HIP build issue "error: only enumeration types have underlying types".
[ROCm/hip commit: 00d735cdc9
2019-12-10 09:40:15 +05:30
7703031d33
Fix hipExtLaunchMultiKernelMultiDevice refactoring. ( #1714 )
...
- Use the correct condition for HIP VDI runtime.
[ROCm/hip commit: e9da934ac6
2019-12-06 09:49:17 -08:00
ebdd811632
Revert - Changes related to hipMemcpyWithStream ( #1718 )
...
Reverting #1673 , #1697 and #1707 .
Support for hipMemcpyWithStream and memcpy optimizations, will be brought in again once issues seen with these are resolved independently.
[ROCm/hip commit: e53fc316f1
2019-12-06 09:51:53 +05:30
abe7531676
[hip] refactoring hipExtLaunchMultiKernelMultiDevice API ( #1698 )
...
[Background] it was found that if lazy linking used for a library that calls hipExtLaunchMultiKernelMultiDevice API then this API can get the wrong program_state object for looking up device kernels leading to a "No device code available" error in this API.
To fix this issue, the API was refactored to be inline and get and pass the correct program_state to an internal hip API to request a multi-device kernel launch.
[ROCm/hip commit: 68cc787781
2019-12-04 11:50:51 +05:30
4e564f783e
Revert changes for atomic FADD support when address is in LDS ( #1701 )
...
This reverts PR #1591 and follow-on PR #1695
[ROCm/hip commit: 32442c6506
2019-11-29 11:58:12 +05:30
dc9bc8ee86
Uniform is_shared query. ( #1695 )
...
[ROCm/hip commit: b6514fffb9
2019-11-28 13:39:05 +05:30
f220a7592e
Unary operators were too restrictive in the type of their argument. ( #1683 )
...
[ROCm/hip commit: aaf31b6b96
2019-11-22 07:54:53 +05:30
92837d4b4d
Fix rocBLAS compilation failure ( #1677 )
...
SWDEV-212749:
o Recent changes to “add support for extended launch” require hip_runtime.h to be include in hip_ext.h
o Order in which external applications include hip_hcc.h/hip_runtime.h causes compilation failure
[ROCm/hip commit: e60dec51da
2019-11-22 07:54:17 +05:30
fd08d068c7
Use native support for atomic FADD when address is in LDS ( #1591 )
...
[ROCm/hip commit: d597e7ca20
2019-11-22 07:53:48 +05:30
3848cd97ea
fixed directed tests fail when hcc bumped to 3.0 ( #1678 )
...
Handled the HCC version check appropriately as few of the directed tests (SWDEV-212161) were failing when hcc was bumped to 3.0.
[ROCm/hip commit: 6b06911ef1
2019-11-20 21:37:52 +05:30
1baa84eae4
General sync memcpy improvements. Add hipMemcpyWithStream ( #1673 )
...
* General sync memcpy improvements. Add `hipMemcpyWithStream`
* Update hip_memory.cpp
[ROCm/hip commit: 5a1f823739
2019-11-20 21:36:37 +05:30
7e75cf12df
Update error codes for hipGetDevice for doxygen and move up null check ( #1668 )
...
* [docs] Update error codes for hipGetDevice
* Move up out ptr check
[ROCm/hip commit: b3161e9fa0
2019-11-20 21:35:27 +05:30
27112bbdf0
Fix helper header when using c++17 ( #1666 )
...
This will fix issue #1621 . It also adds tests for is_callable with c++11, c++14, and c++17.
The fallback implementation was completely broken so I rewrote it so it pass the tests as well. This should be used instead of PR #1631 .
[ROCm/hip commit: 8519a1411c
2019-11-20 21:33:42 +05:30
4ec70a7e03
Extend vector type capabilities and add tests to reflect it. ( #1656 )
...
[ROCm/hip commit: b5b3d1bbaa
2019-11-20 21:32:32 +05:30
97646b52f1
Fix mathlib and app builds with hip-clang. ( #1665 )
...
[ROCm/hip commit: ebe0c56f4f
2019-11-18 08:18:20 -08:00
ecdba45d29
Fix gcc build on NVCC path ( #1661 )
...
* Fix gcc build on NVCC path
* Fix CI build errors
* [dtest] Fix texture and surface obj2D tests
[ROCm/hip commit: e39d7497ec
2019-11-18 12:19:22 +05:30
05ef9bd386
fix complex conjugate for double-complex ( #1659 )
...
The sign in the y component returned from hipConj incorrect for double-complex. Fix to match as in hipConjf above.
[ROCm/hip commit: 3f2316086f
2019-11-18 12:19:12 +05:30
a4f2c7717f
Added null check in hipEventSynchronize ( #1627 )
...
* Added missing null check in hipEventSynchronize
* Minor correction in the Event API description
[ROCm/hip commit: b865a50e44
2019-11-18 12:18:55 +05:30
3ea404a8ba
Add support for extended launch syntax. ( #1530 )
...
* Add support for extended launch syntax.
* Add unit test.
* Fix typo
* hipExtLaunchKernelGGL lives in hip_ext.h
Change-Id: Ice32dab0d43475fda65c6a910c11416871a8f2ff
* [dtest] remove redundant include from hipModuleGetGlobal dtest
[ROCm/hip commit: 69e74c3e96
2019-11-16 22:24:07 -08:00
6318875ebc
Remove redundant declarations.
...
- The revised `hip/hip_ext.h` have that declarations.
[ROCm/hip commit: d6ff22510e
2019-11-07 10:11:22 -05:00
68b4fbd043
Remove native vector support from the GCC case, since it never worked ( #1637 )
...
[ROCm/hip commit: 5530c15cc3
2019-11-07 13:19:14 +05:30
92dcba11ac
Fixed texture 2D mapping for pitched arrays & 3D Texture read ( #1415 )
...
Texture 2D image mapping for pitched arrays:
github issue: Texture Object's Buffer seems to be Misaligned #886
JIRA ticket: SWDEV-199313
SWDEV-151670 : Fixed issue with 3D texture with 4 components
SWDEV-151671 : Issue with 2D layered texture with 4 components
[ROCm/hip commit: e07926ce0f
2019-11-07 13:17:46 +05:30
6968362d99
Rename hip/hip_hcc.h to hip/hip_ext.h ( #1341 )
...
* Rename hip/hip_hcc.h to hip/hip_ext.h
* Deprecate hip_hcc.h
[ROCm/hip commit: 579a4f36fa
2019-11-07 13:17:10 +05:30
f8506bb669
Remove leftover noise.
...
[ROCm/hip commit: b9faa9f8ae
2019-11-06 02:46:21 +02:00
1baaf44179
__half2 should walk like CUDA and talk like CUDA
...
[ROCm/hip commit: e5bd00d06b
2019-11-06 02:43:04 +02:00
e250f76183
Use portable macro for deprecation message.
...
[ROCm/hip commit: a7f311cc14
2019-11-05 11:51:00 -05:00
1b8973d6be
Deprecate HIP Markers ( #1622 )
...
* Deprecate HIP markers
* Deprecate profiler start/stop
[ROCm/hip commit: 54fab7c35c
2019-11-05 12:32:59 +05:30
e87f9b1944
Separate volatile for clarity. Handle assignment.
...
[ROCm/hip commit: 99b9d5449f
2019-11-02 22:02:08 +02:00
b30a42f678
Accessors should work even when oddly volatile.
...
[ROCm/hip commit: ee5097f2c2
2019-11-01 22:18:01 +02:00
7e742b1216
Merge pull request #1515 from ansurya/tex_unbind_issue_fix
...
Fix undefined ref to hipUnbindTexture for texture types
[ROCm/hip commit: ba8105e0cd
2019-10-30 17:54:15 -07:00
2bff0748cd
[HIP] Correct headers and add missing function templates for hip-clang.
...
- Fix 2 runtime API prototypes
`hipOccupancyMaxActiveBlocksPerMultiprocessor` and
`hipOccupancyMaxActiveBlocksPerMultiprocessorWithFlags`
- Add missing function templates of them in hip-clang.
[ROCm/hip commit: 5c8a7521f4
2019-10-29 22:00:11 -04:00
70449cfa92
Revert "Fix occupany APIs ( #1560 )"
...
This reverts commit 4f23f9cb18e4a1e44162
2019-10-29 11:41:08 -07:00
5c77b7d19a
Merge branch 'master' into tex_unbind_issue_fix
...
[ROCm/hip commit: 03623cc3f1
2019-10-25 15:54:25 +05:30
196bdea9c0
merge from master
...
[ROCm/hip commit: 5f47e99ffe
2019-10-25 15:52:09 +05:30
8460793117
Add missing operators, fix GCC compilation. ( #1589 )
...
[ROCm/hip commit: dabd939048
2019-10-25 15:44:24 +05:30
2e9868d597
Fix deadlock, remove old __sync_* use. ( #1584 )
...
This fixes a deadlock introduced by the switch to TTAS loops, and is therefore mildly urgent (to prevent the CI from hoovering in the broken code).
[ROCm/hip commit: a855a13c22
2019-10-25 15:44:17 +05:30
c315da2028
Add hipMemcpy2DfromArray ( #1510 )
...
Adds hipMemcpy2DFromArray and hipMemcpy2DFromArrayAsync equivalent to cudaMemcpy2DFromArray and cudaMemcpy2DFromArrayAsync.
[ROCm/hip commit: 356765a223
2019-10-25 15:43:33 +05:30
3007505d30
Merge branch 'master' into tex_unbind_issue_fix
...
[ROCm/hip commit: 259d8b4cdf
2019-10-25 15:36:55 +05:30
f51eeeb5de
Fixed CI build failure
...
[ROCm/hip commit: ce04bdaa1a
2019-10-25 12:21:41 +05:30
4a7884105f
Hip vdi profiling header ( #1577 )
...
Add HIP-VDI profiling interface for GPU timing collection.
[ROCm/hip commit: 81952ce5a7
2019-10-24 17:45:42 +05:30
8f020907c7
Make CAS loops use the TTAS idiom. ( #1573 )
...
* Make CAS loops use the TTAS idiom.
* More efficient re-formulation of TTAS.
* Fix typo.
* The typo was not quite a typo
[ROCm/hip commit: 9ba25b42c8
2019-10-24 17:45:20 +05:30
4f23f9cb18
Fix occupany APIs ( #1560 )
...
Addresses SWDEV-205006
[ROCm/hip commit: af351d7e1b
2019-10-24 17:44:47 +05:30
4d668d5a52
Improve performance of v2 arg handling ( #1539 )
...
* Improve performance of v2 arg handling
* Missing change to `std::string`
[ROCm/hip commit: c4a51f3679
2019-10-24 17:44:05 +05:30
52a5380263
Improve scalar access into vector types. ( #1531 )
...
The improvement is based on the ideas here: https://t0rakka.silvrback.com/simd-scalar-accessor . It yields significantly better ISA when the base's .xyzw members are used.
[ROCm/hip commit: 4a635add45
2019-10-24 17:43:49 +05:30
370d039d24
Update the declarations of hipMemsetD8, hipMemsetD8Async, hipMemsetD16, hipMemsetD16Async. These functions are type aware and take in as their third argument the number of elements in the buffer, not the buffer size. Change the name of this argument from sizeBytes to count to align with the above description.
...
[ROCm/hip commit: 0b52c1d9d8
2019-10-15 14:18:42 -04:00
Alex Voicu
Sarbojit2019
mhbliao
Evgeny Mankov
jglaser
Rahul Garg
Aryan Salmanpour
Maneesh Gupta
ansurya
satyanveshd
Paul Fultz II
Nick Curtis
Michael LIAO
Anusha Godavarthy Surya
gandryey
searlmc1
Vladislav Sytchenko