Build AllReduce only mode (#443)

* Initial commit of all_reduce_only support

* Working AllReduce only build

* Removing printfs and restoring release build

* Restore P2P index

* Updates to build_allreduce_only mode.

* cleaning up macro ifdefs

[ROCm/rccl commit: 7e55b211c5]

projects/rccl/test/CMakeLists.txt

@@ -12,38 +12,51 @@ if(BUILD_TESTS)
 
   include_directories(${GTEST_INCLUDE_DIRS})
 
-  # Collect source files for tests
-  set(TEST_SOURCES_SINGLE_PROCESS
-    test_AllGather.cpp
-    test_AllReduce.cpp
-    test_AllReduceGroup.cpp
-    test_Broadcast.cpp
-    test_Reduce.cpp
-    test_ReduceScatter.cpp
-    test_GroupCalls.cpp
-    test_CombinedCalls.cpp
-    test_AllReduceAbort.cpp
-    test_BroadcastAbort.cpp
-    test_Scatter.cpp
-    test_Gather.cpp
-    test_AllToAll.cpp
-    test_AllToAllv.cpp
-  )
-
-  set(TEST_SOURCES_MULTI_PROCESS
-    test_AllGatherMultiProcess.cpp
-    test_AllReduceMultiProcess.cpp
-    test_AllReduceGroupMultiProcess.cpp
-    test_AllToAllMultiProcess.cpp
-    test_BroadcastMultiProcess.cpp
-    test_CombinedCallsMultiProcess.cpp
-    test_GatherMultiProcess.cpp
-    test_GroupCallsMultiProcess.cpp
-    test_ReduceMultiProcess.cpp
-    test_ReduceScatterMultiProcess.cpp
-    test_ScatterMultiProcess.cpp
-  )
-
+  if(BUILD_ALLREDUCE_ONLY)
+    set(TEST_SOURCES_SINGLE_PROCESS
+      test_AllReduce.cpp
+      test_AllReduceAbort.cpp
+      test_AllReduceGroup.cpp
+    )
+  else()
+    # Collect source files for tests
+    set(TEST_SOURCES_SINGLE_PROCESS
+      test_AllGather.cpp
+      test_AllReduce.cpp
+      test_AllReduceGroup.cpp
+      test_Broadcast.cpp
+      test_Reduce.cpp
+      test_ReduceScatter.cpp
+      test_GroupCalls.cpp
+      test_CombinedCalls.cpp
+      test_AllReduceAbort.cpp
+      test_BroadcastAbort.cpp
+      test_Scatter.cpp
+      test_Gather.cpp
+      test_AllToAll.cpp
+      test_AllToAllv.cpp
+    )
+  endif()
+  if(BUILD_ALLREDUCE_ONLY)
+    set(TEST_SOURCES_MULTI_PROCESS
+      test_AllReduceMultiProcess.cpp
+      test_AllReduceGroupMultiProcess.cpp
+    )
+  else()
+    set(TEST_SOURCES_MULTI_PROCESS
+      test_AllGatherMultiProcess.cpp
+      test_AllReduceMultiProcess.cpp
+      test_AllReduceGroupMultiProcess.cpp
+      test_AllToAllMultiProcess.cpp
+      test_BroadcastMultiProcess.cpp
+      test_CombinedCallsMultiProcess.cpp
+      test_GatherMultiProcess.cpp
+      test_GroupCallsMultiProcess.cpp
+      test_ReduceMultiProcess.cpp
+      test_ReduceScatterMultiProcess.cpp
+      test_ScatterMultiProcess.cpp
+    )
+  endif()
   add_executable(UnitTests ${TEST_SOURCES_SINGLE_PROCESS})
   target_include_directories(UnitTests PRIVATE ${ROCM_PATH} ${GTEST_INCLUDE_DIRS})
   target_link_libraries(UnitTests PRIVATE ${GTEST_BOTH_LIBRARIES})

projects/rccl/test/test_AllReduce.cpp

@@ -35,7 +35,23 @@ namespace CorrectnessTests
 
     dataset.Release();
   }
-
+#if defined(BUILD_ALLREDUCE_ONLY)
+  INSTANTIATE_TEST_SUITE_P(AllReduceCorrectnessSweep,
+                           AllReduceCorrectnessTest,
+                           testing::Combine(
+                             // Reduction operator
+                             testing::Values(ncclSum),
+                             // Data types
+                             testing::Values(ncclFloat32),
+                             // Number of elements
+                             testing::Values(1024, 1048576),
+                             // Number of devices
+                             testing::Values(2,3,4,5,6,7,8),
+                             // In-place or not
+                             testing::Values(false, true),
+                             testing::Values("RCCL_ENABLE_CLIQUE=0", "RCCL_ENABLE_CLIQUE=1")),
+                           CorrectnessTest::PrintToStringParamName());
+#else
   INSTANTIATE_TEST_SUITE_P(AllReduceCorrectnessSweep,
                            AllReduceCorrectnessTest,
                            testing::Combine(
@@ -60,4 +76,5 @@ namespace CorrectnessTests
                              testing::Values(false, true),
                              testing::Values("RCCL_ENABLE_CLIQUE=0", "RCCL_ENABLE_CLIQUE=1")),
                            CorrectnessTest::PrintToStringParamName());
+#endif
 } // namespace

projects/rccl/test/test_AllReduceGroup.cpp

@@ -47,7 +47,23 @@ namespace CorrectnessTests
         dataset2.Release();
         dataset3.Release();
   }
-
+#if defined(BUILD_ALLREDUCE_ONLY)
+  INSTANTIATE_TEST_SUITE_P(AllReduceGroupCorrectnessSweep,
+                           AllReduceGroupCorrectnessTest,
+                           testing::Combine(
+                             // Reduction operator
+                             testing::Values(ncclSum),
+                             // Data types
+                             testing::Values(ncclFloat32),
+                             // Number of elements
+                             testing::Values(1024, 1048576),
+                             // Number of devices
+                             testing::Values(2,3,4,5,6,7,8),
+                             // In-place or not
+                             testing::Values(false, true),
+                             testing::Values("RCCL_ENABLE_CLIQUE=0", "RCCL_ENABLE_CLIQUE=1")),
+                           CorrectnessTest::PrintToStringParamName());
+#else
   INSTANTIATE_TEST_SUITE_P(AllReduceGroupCorrectnessSweep,
                            AllReduceGroupCorrectnessTest,
                            testing::Combine(
@@ -63,4 +79,5 @@ namespace CorrectnessTests
                              testing::Values(false, true),
                              testing::Values("RCCL_ENABLE_CLIQUE=0", "RCCL_ENABLE_CLIQUE=1")),
                            CorrectnessTest::PrintToStringParamName());
+#endif
 } // namespace