Skip to content

[hipBLASLt] Fix shard overlay convergence #5354

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
davidd-amd wants to merge 3 commits into
base: develop
Choose a base branch
from
+11 −4
Open

[hipBLASLt] Fix shard overlay convergence #5354

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
5fae749
[hipBLASLt] Fix shard overlay convergence for metadata artifacts
davidd-amd Mar 11, 2026
927de9d
Update comment
davidd-amd Mar 11, 2026
94b3508
Merge branch 'develop' into users/davidd-amd/hipblaslt-convergence
davidd-amd Mar 18, 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
7 changes: 7 additions & 0 deletions projects/hipblaslt/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -105,6 +105,13 @@ else()
endif()
message(STATUS "Building for GPU targets: ${GPU_TARGETS}")

set(HIPBLASLT_DIST_TARGETS "" CACHE STRING
"Full distribution GPU target list. When set by TheRock, all shards produce artifacts for each target in distribution list. Defaults to GPU_TARGETS when unset.")
if(NOT HIPBLASLT_DIST_TARGETS)
set(HIPBLASLT_DIST_TARGETS "${GPU_TARGETS}")
endif()
message(STATUS "Distribution GPU targets for metadata: ${HIPBLASLT_DIST_TARGETS}")

find_package(hip REQUIRED)
if(HIPBLASLT_ENABLE_HOST)
if(HIPBLASLT_ENABLE_HIPBLAS_DIRECT)
Expand Down
4 changes: 2 additions & 2 deletions projects/hipblaslt/device-library/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,9 +13,9 @@ set(HIPBLASLT_LIBLOGIC_PATH "" CACHE STRING "Path to library logic files (will u
set(HIPBLASLT_TENSILE_LIBPATH "${hipblaslt_BINARY_DIR}/Tensile" CACHE STRING "Path to output the device gemm libraries.")
file(MAKE_DIRECTORY "${HIPBLASLT_TENSILE_LIBPATH}/library")

list(JOIN GPU_TARGETS "$<SEMICOLON>" TENSILELITE_GPU_TARGETS_SEMI_ESCAPED)
list(JOIN HIPBLASLT_DIST_TARGETS "$<SEMICOLON>" TENSILELITE_DIST_TARGETS_SEMI_ESCAPED)

set(TENSILELITE_BUILD_OPTS ${TENSILELITE_BUILD_OPTS} "--architecture=${TENSILELITE_GPU_TARGETS_SEMI_ESCAPED}")
set(TENSILELITE_BUILD_OPTS ${TENSILELITE_BUILD_OPTS} "--architecture=${TENSILELITE_DIST_TARGETS_SEMI_ESCAPED}")
set(TENSILELITE_BUILD_OPTS ${TENSILELITE_BUILD_OPTS} "--cxx-compiler=${CMAKE_CXX_COMPILER}")
if(HIPBLASLT_ENABLE_ASAN)
set(TENSILELITE_BUILD_OPTS ${TENSILELITE_BUILD_OPTS} "--address-sanitizer")
Expand Down
2 changes: 1 addition & 1 deletion projects/hipblaslt/device-library/extops/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,7 @@

set(ops_dir "${CMAKE_CURRENT_SOURCE_DIR}/../../tensilelite")
set(output_dir "${CMAKE_CURRENT_BINARY_DIR}")
string(REGEX MATCHALL "gfx[a-z0-9]+" archs "${GPU_TARGETS}")
string(REGEX MATCHALL "gfx[a-z0-9]+" archs "${HIPBLASLT_DIST_TARGETS}")
list(REMOVE_DUPLICATES archs)
message(STATUS "Architectures to process: ${archs}")

Expand Down
2 changes: 1 addition & 1 deletion projects/hipblaslt/device-library/matrix-transform/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,7 +23,7 @@

# Need to support Release/Debug

string(REGEX MATCHALL "gfx[a-z0-9]+" archs "${GPU_TARGETS}")
string(REGEX MATCHALL "gfx[a-z0-9]+" archs "${HIPBLASLT_DIST_TARGETS}")
list(REMOVE_DUPLICATES archs)

# Create separate --offload-arch flags for each architecture
Expand Down
Loading