Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

add half gemm support #68

Draft
wants to merge 25 commits into
base: master
Choose a base branch
from
Draft

add half gemm support #68

wants to merge 25 commits into from

Conversation

cayrols
Copy link
Contributor

@cayrols cayrols commented Sep 27, 2023 •
edited
Loading

This PR aims to add the support for half precision gemm.

This is a draft for now.

Tasks:

  • CMake support
  • Change the name of the config test file

@cayrols cayrols force-pushed the seb/gemm.add_hgemm_support branch from cc4396a to c3efcaa Compare September 27, 2023 20:22
mgates3
mgates3 requested changes Dec 12, 2023
View reviewed changes
config/return_float16.cc Show resolved Hide resolved
config/return_float16.cc Show resolved Hide resolved
include/blas/util.hh Show resolved Hide resolved
include/blas/util.hh Show resolved Hide resolved
include/blas/util.hh Show resolved Hide resolved
test/check_gemm.hh Show resolved Hide resolved
test/utils.cc Show resolved Hide resolved
test/cuda/utils.cu Show resolved Hide resolved
test/check_gemm.hh Show resolved Hide resolved
test/utils.hh Show resolved Hide resolved
cayrols added 25 commits December 14, 2023 13:56
@cayrols
hgemm: add half definition and interfaces
3998ebd
@cayrols
hgemm: add hgemm tester; conversion routine from and to half; Nvidia …
60afc47 
…support for now.
@cayrols
utils: move utils into test/cuda; Add missing NVCC flag; Clean utils …
f51f0c2 
…header.
@cayrols
utils: move utils into test/cuda; Add missing NVCC flag; Clean utils …
98e452b 
…header.
@cayrols
half: Add hipify from the cuda utils; update the compilation chain to…
64e6697 
… detect hip.
@cayrols
Replace the definition of half from __half to _float16; rename blas::…
c7061d0 
…half into blas::float16
@cayrols
Rename test/utils.cuh into test/utils.hh
68cb36d
@cayrols
Add hip files that got generated from cuda files.
1c99c49
@cayrols
hgemm: Add hip support for hgemm
7edd364
@cayrols
hgemm: fix compilation after cleaning.
ffac1ae
@cayrols
test: gemm change the bound by removing sqrt.
bd9511d
@cayrols
test: fix scalar type used to get the flop count in half gemm.
44a9e73
@cayrols
hgemm: Use a class float16 instead of an alias.
82752bf
@cayrols
config: fix gpu_backend name issue.
b3c836c
@cayrols
hgemm: fix compilation issue.
6a181e8
@cayrols
TMP: Add explicit compilation flag for reproducer purpose.
3ced0ce
@cayrols
hgemm: Search _Float16 support from compiler; If so, the macro BLAS_U…
d1fd85c 
…SE_ISO_FLOAT16 is defined.
@cayrols
hgemm: add CPU support through MKL.
8b8320b
@cayrols
hgemm: add CPU test with cblas wrapper; add cast_onto_device util tha…
30cd556 
…t cast from and to float16.
@cayrols
float16: update configure search and macro definition.
d686451
@cayrols
hgemm: Fake casting in cublas_wrapper through pointer casting.
5a30ad8
@cayrols
test: in hgemm, add cpu casting support, remove cast_onto_device rout…
f4fed1a 
…ine; Fix compilation issue.
@cayrols
float16: add casting routines from/to fp16 to/from fp32.
d0d2867
@cayrols
float16: add missing config file.
70d893b
@cayrols
hgemm: enable CPU hgemm only when MKL is provided.
8678c82
@cayrols cayrols force-pushed the seb/gemm.add_hgemm_support branch from 6824b9b to 8678c82 Compare December 14, 2023 19:01
@mgates3 mgates3 added the Draft label May 24, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@mgates3 mgates3 mgates3 requested changes

Requested changes must be addressed to merge this pull request.

Assignees
No one assigned
Labels
Draft
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@cayrols @mgates3