Commits · e81f54bbb9d124e2be97baeb92b0054628aabcb4 · gaoqiong / composable_kernel · GitLab

14 Sep, 2022 20 commits
- Remove no-longer-used 'using' directives · e81f54bb
  Po-Yen, Chen authored Sep 14, 2022
  
  e81f54bb
- Use AsSpan() to shorten check_err() calls · 0fa35b29
  Po-Yen, Chen authored Sep 14, 2022
  
  0fa35b29
- Add Tensor<>::AsSpan<>() to create view of tensor values · c0c1d247
  Po-Yen, Chen authored Sep 14, 2022
  
  c0c1d247
- Add to_array() conversion tool to eliminate more variables · b85e8611
  Po-Yen, Chen authored Sep 14, 2022
  
  b85e8611
- Use function return value directly to eliminate variables · fb05bd38
  Po-Yen, Chen authored Sep 14, 2022
  
  fb05bd38
- Use rangified copy() to copy elements · 097506c3
  Po-Yen, Chen authored Sep 14, 2022
  
  097506c3
- Use more meaningful names in permute element examples · ff6a04fd
  Po-Yen, Chen authored Sep 14, 2022
  
  ff6a04fd
- Use more meaningful names in permute bundle example · d53443d5
  Po-Yen, Chen authored Sep 14, 2022
  
  d53443d5
- Add Tensor<>::size() · 7b135645
  Po-Yen, Chen authored Sep 14, 2022
  
  7b135645
- Intialize random values in range · 4643f0e6
  Po-Yen, Chen authored Sep 14, 2022
  
  4643f0e6
- Allow FillUniformDistribution accept range arugment · 0d5a0bc9
  Po-Yen, Chen authored Sep 14, 2022
  
  0d5a0bc9
- Use existing method to shorten code · ee119d9e
  Po-Yen, Chen authored Sep 14, 2022
  
  ee119d9e
- Use type alias to replace duplicated code · 9e03c383
  Po-Yen, Chen authored Sep 14, 2022
  
  9e03c383
- Use new methods to simplify code · 95f21ea5
  Po-Yen, Chen authored Sep 14, 2022
  
  95f21ea5
- Add Tensor<>::data() · 0cfbefce
  Po-Yen, Chen authored Sep 14, 2022
  
  0cfbefce
- Add Tensor<>::GetElementSpaceSizeInBytes() · 31875367
  Po-Yen, Chen authored Sep 14, 2022
  
  31875367
- Simplify permute bundle example · dacc5119
  Po-Yen, Chen authored Sep 14, 2022
  
  dacc5119
- Extract bundle permutation logic out · aef15d2e
  Po-Yen, Chen authored Sep 14, 2022
  
  aef15d2e
- Rename functions · 72f90d4f
  Po-Yen, Chen authored Sep 14, 2022
  
  72f90d4f
- Remove cmd arg parsing logics · 67e84aed
  Po-Yen, Chen authored Sep 14, 2022
  
  67e84aed
13 Sep, 2022 11 commits
- Use faster device op config for example 'HxWx4_fp16' · 234c0580
  Po-Yen, Chen authored Sep 13, 2022
  
  234c0580
- Use faster device op config for example '1xHxW_fp16' · cc8aee3b
  Po-Yen, Chen authored Sep 13, 2022
  
  cc8aee3b
- Use faster device op config for example 'NxHxW_fp16' · c1f605b9
  Po-Yen, Chen authored Sep 13, 2022
  
  c1f605b9
- Measure execution time by default · 65adf21b
  Po-Yen, Chen authored Sep 13, 2022
  
  65adf21b
- Remove global load/store loop in kernel code · f650c019
  Po-Yen, Chen authored Sep 13, 2022
  
  f650c019
- Use more stricter error threshold · 684e012e
  Po-Yen, Chen authored Sep 12, 2022
  
  684e012e
- Add check for the bundled type in example · 73c3a98f
  Po-Yen, Chen authored Sep 12, 2022
  
  73c3a98f
- Enalrge N-dim in example · 7f303b3c
  Po-Yen, Chen authored Sep 12, 2022
  
  7f303b3c
- Enlarge tensor size in example · f4d18f41
  Po-Yen, Chen authored Sep 12, 2022
  
  f4d18f41
- Use fp16 as data type in all examples · 0fc7dc6a
  Po-Yen, Chen authored Sep 12, 2022
  
  0fc7dc6a
- Simply initialize tensor with floating point values · 8789fec4
  Po-Yen, Chen authored Sep 12, 2022
  
  8789fec4
12 Sep, 2022 9 commits
- Add const to the span<> type arguement · 77f87d3a
  Po-Yen, Chen authored Sep 12, 2022
  
  77f87d3a
- Usw switch() to group similar codes · 4defb148
  Po-Yen, Chen authored Sep 12, 2022
  
  4defb148
- Remove redundant example · 3cbf5d96
  Po-Yen, Chen authored Sep 12, 2022
  
  3cbf5d96
- Avoid using too-large LDS in kernel code · e8b0e222
  Po-Yen, Chen authored Sep 12, 2022
  
  e8b0e222
- Merge branch 'develop' into feature/add-permute-device-op · 0a2131e9
  Po-Yen, Chen authored Sep 12, 2022
  
  0a2131e9
- Rename example folder '36_permute' into '37_permute' · e294c675
  Po-Yen, Chen authored Sep 12, 2022
  
  e294c675
- Embed shape info in name of descriptor constructor · 5502bac2
  Po-Yen, Chen authored Sep 12, 2022
  
  5502bac2
- Use fixed 'VectorDim' & 'ScalarPerVector' for LDS · 5a7c738d
  Po-Yen, Chen authored Sep 12, 2022
  
  5a7c738d
- Use more verbose name to avoid name collision · 23c3a395
  Po-Yen, Chen authored Sep 12, 2022
  
  23c3a395