Code

merge CUDA and ROCm codes in op#2847

Code

Merged

wanghan-iapcm merged 25 commits intodeepmodeling:develdeepmodeling/deepmd-kit:develfrom njzjz:merge-cuda-rocm-opnjzjz/deepmd-kit:merge-cuda-rocm-opCopy head branch name to clipboardSep 26, 2023

+66-608

Commits

Commits on Sep 19, 2023

• use the same macro between CUDA and ROCm

  Show description for 2fa777f
  njzjz

  committed
  2fa777f

  Copy full SHA for 2fa777f

• checkpoint

  Show description for 3f8d05d
  njzjz

  committed
  3f8d05d

  Copy full SHA for 3f8d05d

• gpuMemset

  Show description for 6ee416e
  njzjz

  committed
  6ee416e

  Copy full SHA for 6ee416e

• checkpoint

  Show description for 79b31ca
  njzjz

  committed
  79b31ca

  Copy full SHA for 79b31ca

• merge locate_xx_se_a and locate_xx_se_r: the same function

  Show description for 972919f
  njzjz

  committed
  972919f

  Copy full SHA for 972919f

• merge cmake files

  Show description for 797dad2
  njzjz

  committed
  797dad2

  Copy full SHA for 797dad2

• renaming

  Show description for 6ef697f
  njzjz

  committed
  6ef697f

  Copy full SHA for 6ef697f

• Revert "merge locate_xx_se_a and locate_xx_se_r: the same function"

  Show description for c32407c
  njzjz

  committed
  c32407c

  Copy full SHA for c32407c

• fix typo

  Show description for a1cdc20
  njzjz

  committed
  a1cdc20

  Copy full SHA for a1cdc20

• fix trap

  Show description for f7bfadb
  njzjz

  committed
  f7bfadb

  Copy full SHA for f7bfadb

• typo

  Show description for 245964c
  njzjz

  committed
  245964c

  Copy full SHA for 245964c

• fix tabulate_fusion_se_a_fifth_order_polynomial for rocm

  Show description for 15f5729
  njzjz

  committed
  15f5729

  Copy full SHA for 15f5729

• use GPU_DYNAMIC_SHARED_MEM_DECL

  Show description for f36ecc6
  njzjz

  committed
  f36ecc6

  Copy full SHA for f36ecc6

• clean never used shared memory decalare

  Show description for 074bbdc
  njzjz

  committed
  074bbdc

  Copy full SHA for 074bbdc

• fix submodule

  Show description for 4edce5f
  njzjz

  committed
  4edce5f

  Copy full SHA for 4edce5f

Commits on Sep 20, 2023

• merge CUDA and ROCm codes in op

  Show description for 30b61e7
  njzjz

  committed
  30b61e7

  Copy full SHA for 30b61e7

• address comments

  Show description for 28dc52d
  njzjz

  committed
  28dc52d

  Copy full SHA for 28dc52d

• add GpuSyncThreads

  Show description for 1f2e049
  njzjz

  committed
  1f2e049

  Copy full SHA for 1f2e049

Commits on Sep 21, 2023

• Merge branch 'devel' into merge-cuda-rocm

  njzjz

  committed
  e3edda1

  Copy full SHA for e3edda1

• fix offset in warp_reduce

  Show description for 06c3eb6
  njzjz

  committed
  06c3eb6

  Copy full SHA for 06c3eb6

• Merge remote-tracking branch 'origin/devel' into merge-cuda-rocm-op

  Show description for 2049ac4
  njzjz

  committed
  2049ac4

  Copy full SHA for 2049ac4

• cuda -> gpu

  Show description for 7313d73
  njzjz

  committed
  7313d73

  Copy full SHA for 7313d73

• Merge branch 'merge-cuda-rocm' into merge-cuda-rocm-op

  njzjz

  committed
  357417f

  Copy full SHA for 357417f

• define gpuMemcpyDeviceToDevice

  Show description for b2ee402
  njzjz

  committed
  b2ee402

  Copy full SHA for b2ee402

Commits on Sep 22, 2023

• Merge branch 'devel' into merge-cuda-rocm-op

  Show description for b627c86
  njzjz

  authored
  b627c86

  Copy full SHA for b627c86