890beb81d6
Guard rcp rounded implementation as well
...
Since rcp implementations of non-default rounded versions are not correct or supported in OCML, guard them using the same macro OCML_BASIC_ROUNDED_OPERATIONS. Also update the docs and tests.
[ROCm/clr commit: 7b3bbc85c5
2018-11-06 19:53:28 +00:00
4480bb6d06
Guard the OCML rounded operations instead
...
Instead of commenting all these functions out, guard the functions with a macro OCML_BASIC_ROUNDED_OPERATIONS.
[ROCm/clr commit: 9aa92238ab
2018-11-06 16:32:14 +00:00
5c1dc7a071
Remove non-working non-default-rounded math apis
...
In ROCm-Device-Libs, they have dropped the non-default-rounded versions of add, sub, mul, div, sqrt and fma. Therefore, ocml has removed the rte, rtp, rtn, and rtz counterparts. This will remove the same math APIs in HIP for _ru, _rd, _rn, and _rz.
[ROCm/clr commit: cef6e8ef1f
2018-11-05 22:34:16 +00:00
e26a728173
Move the global arrays for hip malloc/free
...
from a header into a source file such that
there's only an unique copy in an executable
and prevent wasting static memory on the host
Change-Id: Id5b62766f77809c8d7b47892cb7149c490dcbdb9
[ROCm/clr commit: cdd0109e70
2018-11-01 16:20:35 -04:00
34c5d69f7c
typedef struct hipFuncAttributes now C compatible.
...
Fixes #591 .
Fixes #694 .
[ROCm/clr commit: 6eac038694
2018-10-31 14:12:45 +05:30
130edf2c43
Merge pull request #714 from aaronenyeshi/remove-static-hip-device
...
Remove libhip_device.a static library
[ROCm/clr commit: c7f2198c6e
2018-10-31 14:08:29 +05:30
58aec2b859
Merge pull request #713 from scchan/master-hcc_early_finalize_support
...
Support more than one device code blobs in a single .kernel section
[ROCm/clr commit: 8f667984af
2018-10-31 14:04:06 +05:30
1c05b06bcd
Remove libhip_device.a static library
...
Move remaining function definitions from device_util.cpp to hip_runtime.h header. We can now remove the static library completely as device_util.cpp was the last file part of target hip::hip_device .
[ROCm/clr commit: ba43d1ce1b
2018-10-25 19:52:35 +00:00
5a6601c62a
hipLaunchKernel, hipLaunchParm are deprecated, and shall be removed.
...
[ROCm/clr commit: a743939900
2018-10-25 13:32:17 +01:00
bd0e1c87b7
Merge pull request #723 from mangupta/fix_double_shfl
...
Fix logic for double variants of __shfl*
[ROCm/clr commit: 05beeec84d
2018-10-25 06:01:38 +05:30
b6f6734285
Fix logic for double variants of __shfl*
...
Change-Id: I604f00b54cf4bd9c5f26ca6fa680fca5e9629417
[ROCm/clr commit: 7b6d232672
2018-10-24 12:39:09 +05:30
127f7d0320
Make HIP functional again with HCC from ROCm 1.9.x
...
Change-Id: I214acdfd0b79dcf783993e44fe31baee64fd4dc3
[ROCm/clr commit: b01ac26948
2018-10-24 10:41:56 +05:30
b859ab46df
Merge pull request #705 from ROCm-Developer-Tools/feature_minimal_changes_for_hc_next
...
Feature minimal changes for hc next
[ROCm/clr commit: 407e092a13
2018-10-19 06:58:31 +05:30
5d1a265efc
Support more than one bundles in a single .kernel section
...
When compiling with Early Finalization enabled in HCC,
the resulting .kernel section of the host object now may
contain more than one device code bundles. This is to
teach the HIP runtime to correctly extract all the
bundles from the .kernel section.
[ROCm/clr commit: f39873c584
2018-10-18 17:00:27 -04:00
35e9dfc593
Dumb workaround is still needed, so add it back.
...
[ROCm/clr commit: 3678063598
2018-10-18 15:33:46 +01:00
c505c2be38
Re-sync with upstream.
...
[ROCm/clr commit: 602280eb8b
2018-10-18 12:27:03 +01:00
5070b00c83
Merge pull request #688 from aaronenyeshi/fix-sinf-cosf-ocml
...
Use sinf and cosf from ocml device libs
[ROCm/clr commit: 30523b72a2
2018-10-18 16:39:20 +05:30
cf95358003
Merge pull request #692 from whchung/hip-reinit-take2
...
HIP program state re-initialization logic (take 2)
[ROCm/clr commit: 9143ae6bdb
2018-10-18 12:06:41 +05:30
bb3706b6f6
Merge pull request #703 from mangupta/stream_create_with_priority
...
Implementation for stream priority
[ROCm/clr commit: 95be669f4a
2018-10-17 10:53:43 +05:30
62e4c224c7
Merge pull request #702 from aaronenyeshi/fix-missing-irif-lib
...
Replace IRIF fences with atomic_work_item_fence
[ROCm/clr commit: 3485d86746
2018-10-17 10:53:27 +05:30
8175561a42
Add missing hipHostRegister flags on nvcc path
...
Change-Id: I69f09204d9c544935104d4168ab8d3626666a623
[ROCm/clr commit: d71006eb99
2018-10-15 15:30:24 +05:30
f3495964ea
Minimal should mean minimal.
...
[ROCm/clr commit: 217fe345ca
2018-10-11 00:21:41 +01:00
21997797ff
Address Aaron's comments
...
[ROCm/clr commit: b2b482c78e
2018-10-11 00:03:01 +01:00
93e6d2f730
Merge branch 'master' of https://github.com/ROCm-Developer-Tools/HIP into feature_minimal_changes_for_hc_next
...
[ROCm/clr commit: c564ccbffc
2018-10-10 11:44:09 +01:00
50388a3bd1
Re-sync with upstream.
...
[ROCm/clr commit: 9ec697c620
2018-10-10 11:43:49 +01:00
c4ed6b18fb
Implementation for stream priority
...
- Requires ROCm 1.9.x or higher
- Requires HCC with PR#886 merged
Change-Id: Id7c95ea091ee610e80c9ad815f1cb989cba570ca
[ROCm/clr commit: 07ee1f07d8
2018-10-05 16:27:46 +05:30
1835636df3
Replace IRIF fences with atomic_work_item_fence
...
[ROCm/clr commit: 03822afaa9
2018-10-04 21:47:28 +00:00
766ce1b4ff
Fix hip_vector_types.h for long long vectors
...
There was a missing long in the declaration for [u]longlongN types.
[ROCm/clr commit: 6a71cf17a1
2018-10-03 13:57:52 -04:00
4e0f0b3e89
HIP program state re-initialization logic
...
This commit is to support kernels dynamically loaded thru means such as
dlopen() after HIP runtime initializes.
[ROCm/clr commit: 319f007bf1
2018-09-26 19:48:47 +00:00
4d59fbbcbe
Use sinf and cosf from ocml device libs
...
Using llvm_amdgcn builtin fails to produce accurate values, we should move to using the ocml device library versions.
[ROCm/clr commit: eeb6c11050
2018-09-25 19:31:39 +00:00
f648d28bcb
Merge pull request #614 from ROCm-Developer-Tools/fma
...
Add overloading resolution functions for fma
[ROCm/clr commit: 7cec85e1ae
2018-09-20 13:38:03 +05:30
2f6bee46f9
Silent warnings about duplicate static keyword
...
static is already in __DEVICE__, so should be removed.
[ROCm/clr commit: cfa71293ce
2018-09-19 10:39:45 -04:00
a7b03bee08
Add fma function with float and _Float16 arguments
...
[ROCm/clr commit: a8bed200b7
2018-09-19 09:59:33 -04:00
445697cfe3
Fix build failure of hipTestHalf and hipTestIncludeMath for hip-clang
...
[ROCm/clr commit: 781f0d97b5
2018-09-18 21:00:15 -04:00
814ac65527
Merge pull request #672 from iotamudelta/fp16_fix
...
Only LLVM6 and higher contain the necessary intrinsics.
[ROCm/clr commit: d44f7401f0
2018-09-18 08:43:33 +05:30
d7c37a2743
Merge pull request #674 from mangupta/fix_dtests_on_nvcc
...
[dtests] Fix hipTestClock, hipTestNew, hipTestGlobalVariable, hipSimpleAtomicsTest & hipTestIncludeMath tests on nvcc path
[ROCm/clr commit: 77ed6e2c46
2018-09-18 07:50:52 +05:30
0e8815b3ad
Merge pull request #677 from yxsamliu/fix-launch-decay
...
Fix hipLaunchKernelGGL for hip-clang
[ROCm/clr commit: dd0373dc02
2018-09-18 07:50:37 +05:30
c0592f16e6
Disable device code for gcc in hip_memory.h
...
These device code should only be seen by HCC or hip-clang. They causd build failure
for HIP-VDI runtime and should be disabled for gcc.
[ROCm/clr commit: b23db4e452
2018-09-17 16:50:42 -04:00
eb9c852dc5
Fix hipLaunchKernelGGL for hip-clang
...
Do not decay function pointer type of the kernel argument passed to hipLaunchKernelGGL
and hipLaunchKernel, otherwise some type information is lost which may cause
type inference failure for the template.
This issue caused compilation error of FeatureLPPooling in Caffe2/PyTorch and this patch
fixes that.
[ROCm/clr commit: 0471ce2da6
2018-09-17 11:20:41 -04:00
9623689bc7
Add mappings for __clock* in nvcc_detail/hip_runtime.h
...
Change-Id: Ibcecf52f3e69298268d921efc036090544fa0ed0
[ROCm/clr commit: 933c13f003
2018-09-17 15:23:30 +05:30
f5d204450c
Align with HC Next.
...
[ROCm/clr commit: 26c26f63cb
2018-09-17 11:50:29 +03:00
20abaa15c4
Merge branch 'master' into support-malloc
...
[ROCm/clr commit: b23609ad42
2018-09-17 10:17:25 +05:30
191a752474
Merge pull request #650 from ROCm-Developer-Tools/hip-clang-new
...
Support placement new in hip-clang
[ROCm/clr commit: 9f553c0b40
2018-09-15 11:21:01 +05:30
24868e6605
Merge pull request #665 from aaronenyeshi/fix-min-funcs
...
Use templates for min to prevent ambiguity
[ROCm/clr commit: 17d12bbfef
2018-09-14 13:21:38 +05:30
70d1143f62
Fix Tensorflow ambiguous min issue
...
[ROCm/clr commit: 6fa3ba3e12
2018-09-13 23:16:20 +00:00
1d0e3c3ece
Only LLVM6 and higher contain the necessary intrinsics.
...
[ROCm/clr commit: dbe0cbcb93
2018-09-13 13:55:43 -05:00
2bca65a8a3
Merge pull request #669 from ROCm-Developer-Tools/feature_automatic_cast
...
Remove potential for mismatch between runtime passed actuals and defined formals
[ROCm/clr commit: d7a832bce1
2018-09-13 07:54:22 +05:30
fbcf531e47
Merge pull request #661 from yxsamliu/add-empty-printf
...
Add empty printf for hip-clang
[ROCm/clr commit: 2081cc9bd4
2018-09-13 07:54:03 +05:30
96fe74cd2b
Avoid AMP-retrict call to CPU-restrict
...
[ROCm/clr commit: ebf565819e
2018-09-12 14:54:31 +00:00
1292ba70a7
Remove potential for mismatch between runtime passed actuals and defined formals.
...
[ROCm/clr commit: 5a8eb9eba5
2018-09-12 10:30:48 +01:00
Aaron Enye Shi
Siu Chi Chan
Jeff Daily
Maneesh Gupta
Alex Voicu
Wen-Heng (Jack) Chung
Yaxun Sam Liu
Johannes M Dieterich