Tom Lin
41f1767365
Pause GC during benchmark to reduce noise
2021-08-28 11:16:19 +01:00
Tom Lin
13cb8ffced
Use custom static reduction for CPU
2021-08-28 11:10:49 +01:00
Tom Lin
4853457dca
Add type annotation for all kernels
...
Update dependencies
2021-08-27 14:04:58 +01:00
Tom Lin
c445b64690
Address CUDA comments
...
Drop soft=false for AMDGPU as this option was removed
Update dependencies
2021-08-18 02:00:50 +01:00
Tom Lin
bb271dd046
Update PlainStream with context
2021-08-18 01:59:06 +01:00
Tom Lin
a26699c5b5
Add oneAPI and KA implementation
...
Isolate projects to avoid transitive dependency
Add parameter for passing devices
Incorporate further reviews
Update all dependencies
2021-08-17 14:28:47 +01:00
Tom Lin
7c1e04a42b
Add comment about blockIdx/workgroupIdx in Julia
2021-06-30 19:31:42 +01:00
Tom Lin
2e957d3f60
Inline blocks in CUDAStream
2021-06-30 19:20:37 +01:00
Tom Lin
418315543c
Use -p 2 and no arg for JuliaStream in CI
2021-06-30 19:09:37 +01:00
Tom Lin
d675875dcd
Switch back to -p for DistributedStream
2021-06-30 19:03:39 +01:00
Tom Lin
fe180656d1
Merge branch 'main' into julia
2021-06-30 18:44:17 +01:00
Tom Lin
4e6c56729b
Inline AMDGPU's hard_wait
...
Show the selected implementation and not a constant "threaded"
2021-06-30 18:09:54 +01:00
Tom Lin
6fe81e1955
Update CUDA to 11.3 for CI script
2021-06-30 16:31:14 +01:00
Tom Lin
ce7f013a8e
Update NVHPC to 2.15 w/ CUDA 11.3
2021-06-30 16:04:27 +01:00
Tom Lin
cd367c7da3
Mirror Fujitsu flags for CMake
2021-06-29 17:53:32 +01:00
Tom Deakin
fa6433bab1
update changelog
2021-06-25 09:45:38 -05:00
Tom Deakin
eba2e79eab
[OpenMP] Add Fujitsu compiler flags
...
For best performance on the A64FX with the Fujitsu compiler,
the array pointers also need to be labeled __restrict and const
as appropriate.
Closes #94 .
2021-06-25 09:44:16 -05:00
Tom Lin
d799535c96
Larger arraysize for CI
2021-06-10 05:06:48 +01:00
Tom Lin
c5ad3f34d9
Drop -p N for DistributedStream.jl CI
2021-06-10 05:01:24 +01:00
Tom Lin
2cf8ca5f8c
Use addprocs() for DistributedStream
2021-06-10 04:57:52 +01:00
Tom Lin
63f471f880
set pwd to JuliaStream.jl for CI run
2021-06-10 04:33:12 +01:00
Tom Lin
b3efa6af67
Initial Julia implementation
2021-06-10 04:20:40 +01:00
Tom Deakin
25e021caa3
Update CHANGELOG.md
2021-06-03 16:08:14 +01:00
Tom Deakin
dd90598e20
Merge pull request #105 from UoB-HPC/tbb
...
Initial TBB implementation
2021-06-03 16:07:38 +01:00
Tom Lin
0e3727d8f8
Make partitioner a compile option
...
Inline all abstractions
Add intel compilers for Make
2021-06-03 13:43:12 +01:00
Tom Lin
0867115d8d
Remove references to oneapi/tbb.h
2021-05-27 10:51:45 +01:00
Tom Lin
7a130a59bc
Don't tie implementation to oneTBB specific headers
...
Fix wrong TBB_ROOT detection
2021-05-27 10:23:06 +01:00
Tom Lin
742f0629be
Initial TBB implementation
2021-05-27 09:28:40 +01:00
Tom Deakin
2ab68ab39e
Update changelog
2021-05-18 16:47:56 +01:00
Tom Deakin
82dedad676
[SYCL 1.2.1] Add check for FP64 support
...
Fixes #98
2021-05-18 16:47:36 +01:00
Tom Deakin
d3269ece59
Merge pull request #101 from UoB-HPC/revert-100-issue-98
...
Revert "Add check for FP64 support"
2021-05-18 16:44:20 +01:00
Tom Deakin
b772d00fe4
Revert "Add check for FP64 support"
2021-05-18 16:44:06 +01:00
Tom Deakin
a209f16c16
Merge pull request #100 from UoB-HPC/issue-98
...
Add check for FP64 support
2021-05-18 11:31:36 +01:00
Tom Lin
0616068ade
Merge branch 'main' into issue-98
2021-05-17 20:16:50 +01:00
Tom Lin
45ebd09ef2
Don't use hardcoded rocm path
2021-05-17 20:00:00 +01:00
Tom Deakin
6581ee63b8
Update changelog
2021-05-17 15:33:54 +01:00
Tom Deakin
e20aecd845
[SYCL 1.2.1] Add check for FP64 support
...
Fixes #98
2021-05-17 15:25:43 +01:00
Tom Lin
cc16547e4d
Add install target for CMake
2021-05-10 17:50:36 +01:00
Tom Deakin
e5c9f313cb
Merge pull request #96 from tob2/stdlib_fix
...
Include stdlib.h for aligned_alloc
2021-04-28 11:19:27 +01:00
Tobias Burnus
75a4394830
Include stdlib.h for aligned_alloc
...
Silence "error: there are no arguments to 'aligned_alloc' that depend
on a template parameter, so a declaration of 'aligned_alloc' must be
available"
* OMPStream.cpp: #include <cstdlib>.
* RAJAStream.cpp: Likewise.
2021-04-22 09:00:55 +02:00
Tom Lin
2eca3974e6
Disable CI for RAJA on gcc-10+CUDA due to ICE
...
Update changelog to include RAJA 0.13.x
2021-04-21 16:28:12 +01:00
Tom Deakin
6ecd10bf98
Deprecate HC
2021-04-20 16:16:56 +01:00
Tom Deakin
a94d678b34
Update CHANGELOG.md
2021-04-08 17:06:43 +01:00
Tom Deakin
9642821afa
Merge pull request #91 from UoB-HPC/cmake
...
CMake+CI integration
2021-04-08 16:18:38 +01:00
Andrei Poenaru
f279f58a72
Improve CMake messages
2021-03-30 17:08:03 +03:00
Andrei Poenaru
4d83bf4b18
Fix some README typos
2021-03-30 16:54:48 +03:00
Tom Lin
7ec28e55f1
CMake: Update CI rocm to 4.1.0
2021-03-30 13:02:48 +01:00
Tom Lin
23f343fa15
Default to C++11
2021-03-24 17:20:11 +00:00
Tom Lin
fc6d032d7f
Use model name as exe prefix
...
Drop C++17 requirement for OMP
2021-03-23 18:16:42 +00:00
Tom Lin
14aefecc57
Re-add all compile and arch dependent flags
...
Fix ACC not linking on CMake < 3.16
Fix CUDA warnings for CMP0104 and avoid repeated -O[n] flags
Fix ComputeCpp not picking up custom flags
[CI] Highlight compiler warnings
[CI] Don't skip remaining tests when one fails
[CI] Add CMake 3.13, 3.15, 3.18 checks
2021-03-11 15:46:23 +00:00