Skip to content

use half of the available cores for hipBLASLt #281

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

Open
bedroge wants to merge 8 commits into
base: main
Choose a base branch
from
Open

use half of the available cores for hipBLASLt #281

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
d253e69
use half of the cores for rocBLAS
bedroge Aug 6, 2026
efc4394
add easystack
bedroge Aug 6, 2026
ea3ce6d
Merge branch 'main' of github.com:EESSI/software-layer-scripts into r…
bedroge Aug 12, 2026
8447f74
use half of the cores for hipBLASLt instead of rocBLAS
bedroge Aug 12, 2026
1e9c891
use 1/4th of the cores for hipBLASLt
bedroge Aug 12, 2026
823f389
use 2 cores for hipBLASLt
bedroge Aug 12, 2026
0ce79fb
use single core for hipBLASLt
bedroge Aug 12, 2026
c5e929a
use hook to set -DTensile_CPU_THREADS=1 for hipBLASLt
bedroge Aug 17, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions easystacks/software.eessi.io/2025.06/accel/amd/eessi-2025.06-eb-5.3.1-2025a.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,2 @@
easyconfigs:
- rocBLAS-4.4.0-rocm-compilers-19.0.0-ROCm-6.4.1.eb
11 changes: 11 additions & 0 deletions eb_hooks.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -587,6 +587,13 @@ def parse_hook_cgal_toolchainopts_precise(ec, eprefix):
raise EasyBuildError("CGAL-specific hook triggered for non-CGAL easyconfig?!")


def parse_hook_hipblaslt_single_core_build(ec, eprefix):
"""Inject -DTensile_CPU_THREADS=1 to force Tensile to use a single core."""
# todo: if this works, only do it for certain gfx targets
if ec.name == 'hipBLASLt':
ec.update('configopts', '-DTensile_CPU_THREADS=1')


def parse_hook_fontconfig_add_fonts(ec, eprefix):
"""Inject --with-add-fonts configure option for fontconfig."""
if ec.name == 'fontconfig':
Expand Down Expand Up @@ -2272,6 +2279,7 @@ def post_easyblock_hook(self, *args, **kwargs):
'fontconfig': parse_hook_fontconfig_add_fonts,
'FreeImage': parse_hook_freeimage_aarch64,
'grpcio': parse_hook_grpcio_zlib,
'hipBLASLt': parse_hook_hipblaslt_single_core_build,
'maturin': parse_hook_maturin,
'Mesa': parse_hook_mesa_use_llvm_minimal,
'OpenBLAS': parse_hook_openblas_relax_lapack_tests_num_errors,
Expand Down Expand Up @@ -2380,6 +2388,9 @@ def set_maximum(parallel, max_value):
# Deucalion has 32GB HBM for 48 cores per node
CPU_TARGET_A64FX: (divide_by_factor, 4),
# software-specific limits
'hipBLASLt': {
'*': (set_maximum, 1), #(divide_by_factor, 4),
},
'libxc': {
'*': (divide_by_factor, 2),
CPU_TARGET_A64FX: (set_maximum, 6),
Expand Down
Loading