Code

[WiP]: Cuasr 1.2#24

Code

Open

thakkarV wants to merge 15 commits intohpcgarage:masterhpcgarage/cuASR:masterfrom thakkarV:cuasr-1.2thakkarV/cuASR:cuasr-1.2Copy head branch name to clipboard

+49,419-35,807

Commits

Commits on Oct 20, 2021

• bump up cutlass upstream to 2.7

  thakkarV

  committed
  c5866aa

  Copy full SHA for c5866aa

Commits on Oct 23, 2021

• fixes for build against cutlass 2.7 version bump

  thakkarV

  committed
  6df6200

  Copy full SHA for 6df6200

Commits on Oct 25, 2021

• nits

  thakkarV

  committed
  e968316

  Copy full SHA for e968316

Commits on Oct 28, 2021

• set and require C++14 for CUDA

  thakkarV

  committed
  4ee508e

  Copy full SHA for 4ee508e

Commits on Oct 31, 2021

• generalize test generator and rename test files in prep for SM80

  thakkarV

  committed
  55b372c

  Copy full SHA for 55b372c

• generalize bench generator and rename bench files in prep for SM80

  thakkarV

  committed
  6b7fa23

  Copy full SHA for 6b7fa23

• add benchmark generator, generated code, cmake file for default configs

  thakkarV

  committed
  a40f7bc

  Copy full SHA for a40f7bc

Commits on Nov 3, 2021

• rework DefualtSemiRingConfiguration struct API for clarity

  thakkarV

  committed
  2c6fa5e

  Copy full SHA for 2c6fa5e

• Add initial Ampere multi-stage templates

  thakkarV

  committed
  4de2c56

  Copy full SHA for 4de2c56

• add default config benchmarks for ampere

  thakkarV

  committed
  7964328

  Copy full SHA for 7964328

• add default srgemm configs for SM80 and bind to multistage kernel

  thakkarV

  committed
  bf1a4e2

  Copy full SHA for bf1a4e2

• change CMake default CUDA arch to SM80

  thakkarV

  committed
  a7ba0cc

  Copy full SHA for a7ba0cc

Commits on Nov 11, 2021

• add default test generator, SM50 defualt tests

  thakkarV

  committed
  62ae551

  Copy full SHA for 62ae551

Commits on Dec 23, 2021

• remove cuASR CUDA Archs and use CMAKE_CUDA_ARCHITECTURES

  thakkarV

  committed
  e6c8962

  Copy full SHA for e6c8962

Commits on Apr 8, 2022

• cmake: fix broken option for user defined CUDA arch

  thakkarV

  committed
  689edcf

  Copy full SHA for 689edcf