Skip to content

add gpu free shared region concurrency regression test #244

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
iemAnshuman wants to merge 5 commits into
base: main
Choose a base branch
from
Open

add gpu free shared region concurrency regression test #244

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
7981a60
add gpu free shared region concurrency regression test
iemAnshuman Jul 29, 2026
7a6a66c
raise ctest timeout above the test's own deadline budget
iemAnshuman Jul 30, 2026
063db85
keep the first lock attempt timestamp so wait time counts retries
iemAnshuman Jul 30, 2026
24d3d31
resolve main conflict
iemAnshuman Aug 3, 2026
016c74f
fix: drop the duplicate enable_testing() call
iemAnshuman Aug 26, 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
1 change: 1 addition & 0 deletions src/multiprocess/multiprocess_memory_limit.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -853,6 +853,7 @@ void lock_shrreg() {
struct timespec sem_ts;
get_timespec(SEM_WAIT_TIME, &sem_ts);

SEQ_POINT_MARK(SEQ_BEFORE_ACQUIRE_SEMLOCK);
int status = sem_timedwait(&region->sem, &sem_ts);
SEQ_POINT_MARK(SEQ_ACQUIRE_SEMLOCK_OK);

Expand Down
6 changes: 5 additions & 1 deletion src/multiprocess/multiprocess_memory_limit.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -53,8 +53,12 @@

#define SEQ_AFTER_INC 8
#define SEQ_AFTER_DEC 9
#define SEQ_BEFORE_ACQUIRE_SEMLOCK 10

#ifndef SEQ_POINT_MARK
#ifdef SHRREG_SEQUENCE_POINT_CALLBACK
void SHRREG_SEQUENCE_POINT_CALLBACK(int sequence);
#define SEQ_POINT_MARK(s) SHRREG_SEQUENCE_POINT_CALLBACK(s)
#elif !defined(SEQ_POINT_MARK)
#define SEQ_POINT_MARK(s)
#endif

Expand Down
22 changes: 20 additions & 2 deletions test/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,17 @@ foreach(TEST_SCRIPT ${TEST_SCRIPTS})
get_filename_component(TEST_TARGET_NAME ${RELATIVE_TEST_PATH} NAME_WE)
set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/${TEST_TARGET_DIR})

if (TEST_TARGET_NAME STREQUAL "test_postinit_owner_death")
if (TEST_SCRIPT MATCHES "test_shared_region_concurrency.c")
add_executable(${TEST_TARGET_NAME} ${TEST_SCRIPT}
${CMAKE_CURRENT_SOURCE_DIR}/../src/multiprocess/multiprocess_memory_limit.c
${CMAKE_CURRENT_SOURCE_DIR}/../src/log_utils.c)
target_compile_definitions(${TEST_TARGET_NAME} PRIVATE
SHRREG_SEQUENCE_POINT_CALLBACK=shrreg_test_sequence_point)
target_compile_options(${TEST_TARGET_NAME} PRIVATE
-D_GNU_SOURCE -ffunction-sections -fdata-sections)
set_target_properties(${TEST_TARGET_NAME} PROPERTIES
LINK_FLAGS "-Wl,--no-export-dynamic,--gc-sections")
elseif (TEST_TARGET_NAME STREQUAL "test_postinit_owner_death")
# Build this focused regression test with the production shared-region
# implementation. It does not invoke any CUDA/NVML entry point at
# runtime. Section garbage collection drops unrelated GPU-facing
Expand All @@ -34,7 +44,8 @@ foreach(TEST_SCRIPT ${TEST_SCRIPTS})
endif()

list(APPEND TEST_TARGET_NAMES_LIST ${TEST_TARGET_NAME})
if (TEST_TARGET_NAME STREQUAL "test_postinit_owner_death")
if (TEST_SCRIPT MATCHES "test_shared_region_concurrency.c" OR
TEST_TARGET_NAME STREQUAL "test_postinit_owner_death")
target_link_libraries(${TEST_TARGET_NAME} -lrt -lpthread)
else()
target_link_libraries(${TEST_TARGET_NAME} -lrt -lpthread
Expand All @@ -48,6 +59,13 @@ foreach(TEST_SCRIPT ${TEST_SCRIPTS})
endforeach()
set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR})

add_test(NAME shared_region_concurrency
COMMAND test_shared_region_concurrency)
# The test waits on the start barrier and on registration in sequence, each
# bounded by SHRREG_TEST_TIMEOUT_MS (20s default), so it can spend 40s before
# printing which stage stalled. Keep this above that budget or ctest kills it
# first and the diagnostic is lost.
set_tests_properties(shared_region_concurrency PROPERTIES TIMEOUT 90)
add_test(NAME postinit_owner_death
COMMAND test_postinit_owner_death)
set_tests_properties(postinit_owner_death PROPERTIES TIMEOUT 20)
Expand Down
Loading
Loading