Commits · odd-size-image-vector-load-feature-dev · liyinrong / composable_kernel

05 Jun, 2020 1 commit
- use flag to indicate if a load access can be x4 vectorized · 71ffba65
  Jane.Zhou authored 5 years ago
  
  71ffba65
18 Feb, 2020 1 commit
- MIOpen integration (#15) · 7d09790a
  Chao Liu authored 5 years ago
```
* renaming
```
  7d09790a
17 Feb, 2020 1 commit
- MIopen integration (#13) · 1a66e35b
  Chao Liu authored 5 years ago
```
* update for miopen integration: cosmetic refactor
```
  1a66e35b
27 Jan, 2020 1 commit
- Update for recent MIOpen integration (#11) · 3406a114
  Chao Liu authored 5 years ago
```
* update for MIOpen integration
```
  3406a114
20 Jan, 2020 1 commit

Added bwd data v3r1 v4r1, tweaking v1 (#10) · c5da0377

Chao Liu authored 5 years ago

* Added bwd data v3r1: breaking down compute into a series of load balanced GEMM, and launch in a single kernel
* Added bwd data v4r1: like v3r1, but launch GEMMs in multiple kernels
* Tweaked v1r1  and v1r2 (atomic) on AMD GPU

c5da0377

05 Dec, 2019 1 commit
- update implicit GEMM forward v4r4 to use gridwise gemm (#9) · e2b4c5b4
  Chao Liu authored 5 years ago
```
* updated fwd v4r4 to use gridwise gemm
* updated gridwise gemm api calls in bwd-data v1r1 and v2r1
```
  e2b4c5b4
03 Dec, 2019 2 commits

fixed faulty padding API calls (#8) · 19a93dac
Chao Liu authored 5 years ago

19a93dac

backward data (#7) · 8f5f6496

Chao Liu authored 5 years ago

* enabled atomic add in tensor copy
* added gridwise GEMM
* added backward data conv using GEMM + atomic
* added backward data conv using GEMM, no atomic

8f5f6496

04 Nov, 2019 2 commits
- remove dead file (#6) · 31ded4ac
  Chao Liu authored 5 years ago
  
  31ded4ac
- MIOpen integration: recent bug fixes from MIOpen (#5) · 562e1e27
  Chao Liu authored 5 years ago
  
  562e1e27
11 Oct, 2019 1 commit
- Refactor for MIOpen integration (#4) · 52c3fe05
  Chao Liu authored 5 years ago
```
Refactor, so can bring multi-index transformation and padding support into MIOpen
```
  52c3fe05
30 Sep, 2019 2 commits
- Merge pull request #3 from asroy/clean_up · 9aaeacc8
  Chao Liu authored 5 years ago
```
enable type conversion in ThreadwiseGenericTensorSliceCopy_v2r1 and BlockwiseGenericTensorSliceCopy_v2
```
  9aaeacc8
- enable type conversion in blockwise copy v2 and threadwise copy v2r1 · cf218184
  Chao Liu authored 5 years ago
  
  cf218184
27 Sep, 2019 6 commits
- tweaking · 012d3a07
  Chao Liu authored 5 years ago
  
  012d3a07
- tweaking · 14315b72
  Chao Liu authored 5 years ago
  
  14315b72
- debugging · ebe38f3d
  Chao Liu authored 5 years ago
  
  ebe38f3d
- remove dead code · 9b280cc5
  Chao Liu authored 5 years ago
  
  9b280cc5
- nvidia build · 98a2cfcc
  Chao Liu authored 5 years ago
  
  98a2cfcc
- clean up · 00089cd6
  Chao Liu authored 5 years ago
  
  00089cd6
26 Sep, 2019 6 commits
- clean up · b12bbcee
  Chao Liu authored 5 years ago
  
  b12bbcee
- removing dependency on old tensor descriptor · 51a9fa1d
  Chao Liu authored 5 years ago
  
  51a9fa1d
- added type conversion in threadwise and blockwise copy · 0f52c4c0
  Chao Liu authored 5 years ago
  
  0f52c4c0
- added type conversion in threadwise and blockwise copy · b3d4595f
  Chao Liu authored 5 years ago
  
  b3d4595f
- removing old implementation of tensor descriptor · 3cb2a7d0
  Chao Liu authored 5 years ago
  
  3cb2a7d0
- removing old implementation of tensor descriptor · 39d92e7d
  Chao Liu authored 5 years ago
  
  39d92e7d
25 Sep, 2019 3 commits
- clean up · 012b5253
  Chao Liu authored 5 years ago
  
  012b5253
- added GetLinearDimensionMask · e1ae8f18
  Chao Liu authored 5 years ago
  
  e1ae8f18
- adding GetLinearDimensionMask() · 4f4aba48
  Chao Liu authored 5 years ago
  
  4f4aba48
24 Sep, 2019 1 commit
- refactor · 545d9305
  Chao Liu authored 5 years ago
  
  545d9305
22 Sep, 2019 3 commits
- nvidia build · 37f4e2b6
  Chao Liu authored 5 years ago
  
  37f4e2b6
- done: explicitly separate offset component into compile-time, block-invariant... · 6c2c50b0
  Chao Liu authored 5 years ago
```
done: explicitly separate offset component into compile-time, block-invariant and per-thread components. Experimenting
```
  6c2c50b0
- WIP: explicitly separate offset component into compile-time, block-invariant... · 51884fc2
  Chao Liu authored 5 years ago
```
WIP: explicitly separate offset component into compile-time, block-invariant and per-thread components
```
  51884fc2
21 Sep, 2019 3 commits
- refactor · 740da00a
  Chao Liu authored 5 years ago
  
  740da00a
- nvidia build · 184c6e7d
  Chao Liu authored 5 years ago
  
  184c6e7d
- adding logic to judge linear dimension · f00c1381
  Chao Liu authored 5 years ago
  
  f00c1381
20 Sep, 2019 1 commit
- refactor · bf7e7d62
  Chao Liu authored 5 years ago
  
  bf7e7d62
19 Sep, 2019 1 commit
- use buffer_load buffer_store intrinsic · b6e1c52a
  Chao Liu authored 5 years ago
  
  b6e1c52a
18 Sep, 2019 3 commits
- reduce some register usage in index · 8afbb10d
  Chao Liu authored 5 years ago
  
  8afbb10d
- refactor · 94bb1b48
  Chao Liu authored 5 years ago
  
  94bb1b48
- add global_load and buffer_load inline asm · 86cc678f
  Chao Liu authored 5 years ago
  
  86cc678f