Code

Prepare 0.3 release#10

Code

Merged

stijnh merged 26 commits intomainKernelTuner/kernel_float:mainfrom

devKernelTuner/kernel_float:devCopy head branch name to clipboard

Apr 11, 2025

Commits

Commits on Nov 1, 2024

• Make code compile with HIP on Lumi

  stijnh

  committed
  5490ea7

  Copy full SHA for 5490ea7

Commits on Nov 18, 2024

• Changes to make code compile under HIPRTC

  stijnh

  committed
  9385655

  Copy full SHA for 9385655

• Merge branch 'hip-support' into dev

  stijnh

  committed
  de62ad0

  Copy full SHA for de62ad0

• Add approximation functions

  stijnh

  committed
  4fca9cb

  Copy full SHA for 4fca9cb

• Simplify how policies are implemented internally

  stijnh

  committed
  ae0e6b1

  Copy full SHA for ae0e6b1

• Rename FP16 primitive names: `__half` to `half_t` and `__nv_bfloat16` `to bfloat16_t`

  stijnh

  committed
  f89cf98

  Copy full SHA for f89cf98

• Implement approximation for `pow`

  stijnh

  committed
  014e32f

  Copy full SHA for 014e32f

• Add `apply_fallback_impl` struct

  stijnh

  committed
  003ce36

  Copy full SHA for 003ce36

• Add `approx_*` functions

  stijnh

  committed
  76501fd

  Copy full SHA for 76501fd

Commits on Nov 20, 2024

• Fix incorrect definition of `KERNEL_FLOAT_POLICY`

  stijnh

  committed
  ba7356a

  Copy full SHA for ba7356a

• Add `Accuracy` parameter to `zip_common`

  stijnh

  committed
  4231f44

  Copy full SHA for 4231f44

• Add `add_mul` to `vector`

  stijnh

  committed
  f5edbc8

  Copy full SHA for f5edbc8

• Overwrite `fast_policy` for FP16 and BF16

  stijnh

  committed
  e6c8a7c

  Copy full SHA for e6c8a7c

• `kernel_float::approx::sqrt(0)` now returns 0

  stijnh

  committed
  5c859b9

  Copy full SHA for 5c859b9

Commits on Nov 26, 2024

• Fix compilation error on HIP due to `KERNEL_FLOAT_FAST_F32_MAP`

  stijnh

  committed
  76c695a

  Copy full SHA for 76c695a

• Remove call to `__exp2f` since it does not exist

  stijnh

  committed
  d8a53a3

  Copy full SHA for d8a53a3

• Change github workflow to compile for all architectures

  stijnh

  committed
  a2b08a5

  Copy full SHA for a2b08a5

Commits on Dec 2, 2024

• Fix bug in `approx::exp(bfloat16)` for HIP

  stijnh

  committed
  846de1f

  Copy full SHA for 846de1f

• Fix several issues related to HIP compilation for bfloat16

  stijnh

  committed
  f94bd10

  Copy full SHA for f94bd10

Commits on Jan 27, 2025

• `vector_ptr` now requires alignment in bytes instead of elements

  stijnh

  committed
  2730789

  Copy full SHA for 2730789

• Remove `apply_fallback_impl`

  stijnh

  committed
  1611258

  Copy full SHA for 1611258

• Fix incorrect type name in `approx.h`

  stijnh

  committed
  c44c6ed

  Copy full SHA for c44c6ed

• Fix incorrect type name in `binops.h`

  stijnh

  committed
  212efee

  Copy full SHA for 212efee

• Change `AssignConversionProxy` to also accept rvalues

  stijnh

  committed
  09dc820

  Copy full SHA for 09dc820

• Change `ops::cast` to get rid of `cast_float_fallback`

  stijnh

  committed
  126737c

  Copy full SHA for 126737c

Commits on Apr 11, 2025

• Update `Jimver/cuda-toolkit` workflow action to newer version

  stijnh

  committed
  4d18563

  Copy full SHA for 4d18563