## Summary
- Fix `run-remote.sh` to correctly execute multi-command scripts (e.g.,
multiple `mpirun` calls)
- The old approach piped decoded script through `base64 -d | bash`,
which feeds the script via bash's **stdin**. When `mpirun` (or its child
processes) runs, it can consume the remaining stdin, causing bash to
never see subsequent commands — only the first command would execute.
- The fix decodes the script to a **temp file** and runs `bash -euxo
pipefail "$TMP"` instead, so bash reads commands from the file and
`mpirun` consuming stdin has no effect.
- Applied to both the docker path (pssh + docker exec) and the
non-docker path (pssh only).
🤖 Generated with [Claude Code](https://claude.com/claude-code)
## Summary
- Replace the two-step `signal()` implementation (`incOutbound()` +
`atomicStore()`) with a single fire-and-forget PTX
`red.release.sys.global.add.u64` instruction
- This eliminates one local atomic fetch-add and replaces a remote store
with a remote atomic add that has no return value — more efficient on
both NVIDIA (PTX `red`) and AMD (compiler optimizes `(void)fetch_add` to
fire-and-forget `flat_atomic_add_x2`)
- Add a C++ perf test (`PERF_TEST`) in `mp_unit` for signal+wait
ping-pong latency
### Performance (H100, 2 ranks, signal+wait round-trip)
```
SemaphorePerfTest.SignalPingPong:
Store-based (old): 2.595 us/iter
Red-based (new): 2.345 us/iter
Speedup: 1.11x
```
## Test plan
- [x] Builds successfully (`make mp_unit_tests`)
- [x] `mpirun -np 2 ./build/bin/mp_unit_tests --filter
"SemaphorePerfTest"` — 1.11x speedup
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
- Removes the GTest dependency, replacing it with a minimal custom
framework (`test/framework.*`) that covers only what the tests actually
use — a unified `TEST()` macro with SFINAE-based fixture auto-detection,
`EXPECT_*`/`ASSERT_*` assertions, environments, and setup/teardown.
- `--exclude-perf-tests` flag and substring-based negative filtering
- `MSCCLPP_ENABLE_COVERAGE` CMake option with gcov/lcov; CI uploads to
Codecov
- Merges standalone `test/perf/` into main test targets
- Refactors Azure pipelines to reduce redundancies & make more readable
---------
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: Changho Hwang <changhohwang@microsoft.com>
Address review comments:
1. Ensure TearDown() is always called if SetUp() succeeds, even when
TestBody() throws. This prevents resource leaks and maintains MPI
synchronization between tests.
2. Replace assert() in fifo_perf_tests.cu with proper return false
on validation failure, ensuring consistent test failure reporting.
Fixes:
- test/framework.cc: Track SetUp success and call TearDown in finally-style
- test/unit/fifo_perf_tests.cu: Replace assert with explicit check
Co-authored-by: chhwang <8018170+chhwang@users.noreply.github.com>
## Summary
This PR addresses a multicast resource leak, fixes `cuMemMap` offset
handling for multicast handles, renames NVLS allreduce algorithm classes
for clarity, and adds a new unit test for `SwitchChannel`.
### Bug Fixes
#### 1. Fix multicast allocation handle leak in `createMulticast()`
(`gpu_ipc_mem.cc`)
`GpuIpcMemHandle::createMulticast()` called
`cuMulticastCreate(&allocHandle, ...)` but never released the local
`allocHandle` after exporting it to shareable handles (POSIX FD /
Fabric). This caused a reference count leak — the multicast object was
never freed even after all mappings and imported handles were released.
Per the [CUDA Driver API docs for
`cuMemRelease`](https://docs.nvidia.com/cuda/cuda-driver-api/group__CUDA__VA.html):
> *"The memory allocation will be freed when all outstanding mappings to
the memory are unmapped and when all outstanding references to the
handle (including its shareable counterparts) are also released."*
The fix adds `cuMemRelease(allocHandle)` after export, matching the
existing pattern used for regular allocations in
`GpuIpcMemHandle::create()`.
**Impact:** Without this fix, repeated creation/destruction of NVLS
connections causes OOM after ~120 iterations when allocating 1GB
multicast buffers on H100.
#### 2. Fix `cuMemMap` offset for multicast handles (`gpu_ipc_mem.cc`)
`cuMemMap` requires `offset=0` for multicast handles. Previously, the
code attempted to map at a non-zero offset within the multicast object,
leading to errors when binding multiple buffers to the same
`NvlsConnection`. The fix maps the entire range `[0, mcOffset +
bufferSize)` and returns the pointer offset by `mcOffset`. This only
consumes extra virtual address space; no additional physical memory is
used.
### Refactoring
#### 3. Rename NVLS allreduce algorithm classes
Renamed for clarity:
- `AllreduceNvls` → `AllreduceNvlsZeroCopy`
- `AllreduceNvlsWithCopy` → `AllreduceNvlsWarpPipeline`
- `AllreduceNvlsWithCopy2` → `AllreduceNvlsBlockPipeline`
Updated all references in builder, selector, docs, and examples.
#### 4. Move `nvlsConnections` setup to `initialize()`
Moved `nvlsConnections_` from `AlgorithmCtx` (which no longer has this
member) to individual algorithm class members, initialized in their
`initialize()` methods.
### Tests
#### 5. Add `TwoChannelsSameConnection` test
New unit test that creates two `SwitchChannel` instances from the same
`NvlsConnection`, performs reduce operations on both, and verifies
correctness. This exercises the multi-bind path that triggered the
`cuMemMap` offset fix.
### Files Changed
- `src/core/gpu_ipc_mem.cc` — multicast handle leak fix + cuMemMap
offset fix
- `src/ext/collectives/allreduce/allreduce_nvls_zero_copy.cu` (renamed)
- `src/ext/collectives/allreduce/allreduce_nvls_warp_pipeline.cu`
(renamed)
- `src/ext/collectives/allreduce/allreduce_nvls_block_pipeline.cu`
(renamed)
- `src/ext/collectives/allreduce/allreduce_nvls_packet.cu` —
nvlsConnections fix
- `src/ext/collectives/include/allreduce/*.hpp` — renamed headers
- `src/ext/collectives/algorithm_collection_builder.cc` — updated
references
- `src/ext/nccl/algorithm_selector.cc` — updated algorithm names
- `test/mp_unit/switch_channel_tests.cu` — new test
- `docs/guide/mscclpp-torch-integration.md` — updated names
- `examples/torch-integration/customized_comm_with_default_algo.py` —
updated names
This pull request updates the way the `nlohmann/json` library is fetched
and upgrades it to a newer version in both the main build and test
configuration files.
Addressed installation issue in some env
This PR refactors the algorithm selection logic in MSCCL++ and
introduces support for symmetric memory configuration through
environment variables.
1. Algorithm Selection Refactoring
Use separate class for algo selection. Could introduce more complex
logic for algo selection based on message size, arch, if cuda graph is
enabled and memory allocation method
2. Symmetric Memory Support
Introduced symmetricMemory parameter in algorithm context key
generation. Remove disableChannelCache env as is ambiguous
3. Add new args for build_default_algorithms
Add flag_buffer, and flag_buffer_size args to build default algorithm.
Then we could use unified flag buffer for different algorithms, avoid
application hanging when switch algo for different message size.
---------
Co-authored-by: chhwang <8018170+chhwang@users.noreply.github.com>
Co-authored-by: Qinghua Zhou <qinghuazhou@microsoft.com>
Co-authored-by: Caio Rocha <caiorocha@microsoft.com>
- Remove complex wildcard pattern matching (*, ?, negative patterns)
- Use simple substring matching with find()
- Simpler implementation, easier to understand and maintain
- Still supports --gtest_filter for basic test name filtering
Note: For advanced filtering like wildcards, users can use multiple
test runs with different substring filters.
Co-authored-by: chhwang <8018170+chhwang@users.noreply.github.com>
- Remove test/perf/ directory (fifo_test.cu, framework.{cc,hpp}, CMakeLists.txt)
- Remove add_subdirectory(perf) from test/CMakeLists.txt
- Performance tests now integrated into unit_tests as fifo_perf_tests.cu
- Fix mp_unit_tests.cc to use framework functions without ::testing:: namespace
- Fix bootstrap_tests.cc ErrorCode comparison to use ASSERT_TRUE
- Fix switch_channel_tests.cu to not use streaming with ASSERT_EQ
- Add missing #include <unistd.h> to executor_tests.cc
All perf test functionality is now in unit_tests and can be filtered
with --exclude-perf-tests flag. The standalone test/perf/ directory
is no longer needed.
Verified builds:
- unit_tests: ✅
- mp_unit_tests: ✅
Co-authored-by: chhwang <8018170+chhwang@users.noreply.github.com>
- Add unit_tests_main.cc with main() function for unit_tests executable
- Create fifo_perf_tests.cu as PERF_TEST for unit_tests
- Add fifo_perf_tests.cu to unit_tests sources
- Fix errors_tests.cc to use ASSERT_TRUE for ErrorCode comparisons
- Fix core_tests.cc to use ASSERT_TRUE for TransportFlags comparisons
- Add Azure pipeline step for Debug build with coverage
- Add step to run mp_unit_tests --exclude-perf-tests with coverage
The perf tests are now part of unit_tests and can be filtered out
for coverage reporting. CI now includes Debug build with coverage
collection for non-performance tests.
Co-authored-by: chhwang <8018170+chhwang@users.noreply.github.com>
- Add isPerfTest field to TestInfoInternal struct
- Add --exclude-perf-tests command line argument
- Add PERF_TEST and PERF_TEST_F macros for marking performance tests
- Update runAllTests to filter performance tests when requested
- Remove genhtml dependency and HTML report generation
- Keep only coverage.info file generation with lcov
Performance tests can now be excluded with:
./build/bin/unit_tests --exclude-perf-tests
./build/bin/mp_unit_tests --exclude-perf-tests
Co-authored-by: chhwang <8018170+chhwang@users.noreply.github.com>
- Add nlohmann::ordered_json metrics field to TestResult struct
- Add nlohmann/json.hpp include to test/framework.hpp
- Link test_framework with nlohmann_json::nlohmann_json
- Replace PerfTestResult with TestResult in test/perf/framework.cc
- Move perf utility functions to utils namespace for consistency
- Remove duplicate PerfTestResult struct definition
This consolidates the two similar structs into one, reducing code
duplication while maintaining all necessary fields for both unit
tests (passed/failure_message) and performance tests (metrics).
Verified build succeeds with Docker:
docker run --rm -v $(pwd):/workspace -w /workspace \
ghcr.io/microsoft/mscclpp/mscclpp:base-dev-cuda12.4 bash -c \
"cd /workspace/build && make -j4 fifo_test"
Co-authored-by: chhwang <8018170+chhwang@users.noreply.github.com>
- Move helper classes inside namespace for proper access
- Remove duplicate class definitions outside namespace
- Tests can now build in Docker with CUDA toolkit installed
- Remaining issues: ErrorCode and TransportFlags need operator<< for EXPECT_EQ
Successfully building with Docker:
docker run --rm -v $(pwd):/workspace -w /workspace \
ghcr.io/microsoft/mscclpp/mscclpp:base-dev-cuda12.4 bash -c \
"mkdir build && cd build && cmake -DMSCCLPP_BYPASS_GPU_CHECK=ON \
-DMSCCLPP_USE_CUDA=ON .. && make -j4"
Note: Some unit tests (errors_tests.cc, core_tests.cc) need operator<<
defined for ErrorCode and TransportFlags to compile with custom framework.
Co-authored-by: chhwang <8018170+chhwang@users.noreply.github.com>
- Make MPI REQUIRED for test builds (clearer error messages)
- Add project include directories to test_framework library
- Fix core_tests.cc to use custom framework correctly
- Fix mp_unit_tests.hpp to use mscclpp::test namespace
- Add FAIL() macro with streaming support for test messages
- Building tests now works in Docker environment with GPU bypass
Tests can now be built using:
docker run --rm -v $(pwd):/workspace -w /workspace \
ghcr.io/microsoft/mscclpp/mscclpp:base-dev-cuda12.4 bash -c \
"mkdir build && cd build && cmake -DMSCCLPP_BYPASS_GPU_CHECK=ON \
-DMSCCLPP_USE_CUDA=ON .. && make -j"
Co-authored-by: chhwang <8018170+chhwang@users.noreply.github.com>
Support --gtest_filter command line argument for test filtering,
compatible with Azure pipeline configurations.
Co-authored-by: chhwang <8018170+chhwang@users.noreply.github.com>
- Move test framework from test/perf/ to test/ for shared use
- Add GTest-compatible macros (TEST, TEST_F, EXPECT_*, ASSERT_*, etc.)
- Remove GTest dependency from CMakeLists.txt
- Add test_framework library for unit and mp_unit tests
- Add code coverage support with lcov (MSCCLPP_ENABLE_COVERAGE option)
- Update perf tests to use shared framework
Co-authored-by: chhwang <8018170+chhwang@users.noreply.github.com>
- Modified test/mp_unit/mp_unit_tests.hpp to use ../framework.hpp instead of gtest/gtest.h
- Enhanced test/framework.hpp with GTest-compatible APIs:
- Added Environment base class for global test setup/teardown
- Added TestInfo and UnitTest classes for test metadata access
- Added GTEST_SKIP macro support via SkipHelper class
- Added namespace alias 'testing' for compatibility
- Added InitGoogleTest and AddGlobalTestEnvironment helper functions
- Updated test/framework.cc with implementations for new classes
- All mp_unit test files now use framework.hpp through mp_unit_tests.hpp
- Formatting applied via lint.sh
Co-authored-by: chhwang <8018170+chhwang@users.noreply.github.com>
