[mlir] Link mlir_runner_utils statically into cuda/rocm-runtime-wrappers.

The runtime-wrappers depend on LLVMSupport, pulling in static initialization code (e.g. command line arguments). Dynamically loading multiple such libraries results in ODR violoations.

So far this has not been an issue, but in D94421, I would like to load both the async-runtime and the cuda-runtime-wrappers as part of a cuda-runner integration test. When doing this, code that asserts that an option category is only registered once fails (note that I've only experienced this in Google's bazel where the async-runtime depends on LLVMSupport, but a similar issue would happen in cmake if more than one runtime-wrapper starts to depend on LLVMSupport).

The underlying issue is that we have a mix of static and dynamic linking. If all dependencies were loaded as shared objects (i.e. if LLVMSupport was linked dynamically to the runtime wrappers), each dependency would only get loaded once. However, linking dependencies dynamically would require special attention to paths (one could dynamically load the dependencies first given explicit paths). The simpler approach seems to be to link all dependencies statically into a single shared object.

This change basically applies the same logic that we have in the c_runner_utils: we have a shared object target that can be loaded dynamically, and we have a static library target that can be linked to other runtime-wrapper shared object targets.

Reviewed By: herhut

Differential Revision: https://reviews.llvm.org/D94399

diff --git a/mlir/include/mlir/ExecutionEngine/CRunnerUtils.h b/mlir/include/mlir/ExecutionEngine/CRunnerUtils.h
index 2d0608a..272058d 100644 (file)
--- a/mlir/include/mlir/ExecutionEngine/CRunnerUtils.h
+++ b/mlir/include/mlir/ExecutionEngine/CRunnerUtils.h
@@ -26,7 +26,7 @@
 #define MLIR_CRUNNERUTILS_EXPORT __declspec(dllimport)
 #endif // mlir_c_runner_utils_EXPORTS
 #endif // MLIR_CRUNNERUTILS_EXPORT
-#else
+#else  // _WIN32
 #define MLIR_CRUNNERUTILS_EXPORT
 #define MLIR_CRUNNERUTILS_DEFINE_FUNCTIONS
 #endif // _WIN32

diff --git a/mlir/lib/ExecutionEngine/CMakeLists.txt b/mlir/lib/ExecutionEngine/CMakeLists.txt
index 7d86811..f995ad0 100644 (file)
--- a/mlir/lib/ExecutionEngine/CMakeLists.txt
+++ b/mlir/lib/ExecutionEngine/CMakeLists.txt
@@ -80,6 +80,7 @@ add_mlir_library(mlir_c_runner_utils
   EXCLUDE_FROM_LIBMLIR
   )
 set_property(TARGET mlir_c_runner_utils PROPERTY CXX_STANDARD 11)
+target_compile_definitions(mlir_c_runner_utils PRIVATE mlir_c_runner_utils_EXPORTS)
 
 add_mlir_library(mlir_c_runner_utils_static
   CRunnerUtils.cpp
@@ -88,7 +89,6 @@ add_mlir_library(mlir_c_runner_utils_static
   EXCLUDE_FROM_LIBMLIR
   )
 set_property(TARGET mlir_c_runner_utils_static PROPERTY CXX_STANDARD 11)
-target_compile_definitions(mlir_c_runner_utils PRIVATE mlir_c_runner_utils_EXPORTS)
 
 add_mlir_library(mlir_runner_utils
   SHARED
@@ -101,6 +101,15 @@ add_mlir_library(mlir_runner_utils
 )
 target_compile_definitions(mlir_runner_utils PRIVATE mlir_runner_utils_EXPORTS)
 
+add_mlir_library(mlir_runner_utils_static
+  RunnerUtils.cpp
+
+  EXCLUDE_FROM_LIBMLIR
+
+  LINK_LIBS PUBLIC
+  mlir_c_runner_utils_static
+)
+
 add_mlir_library(mlir_async_runtime
   SHARED
   AsyncRuntime.cpp
@@ -113,3 +122,13 @@ add_mlir_library(mlir_async_runtime
 )
 set_property(TARGET mlir_async_runtime PROPERTY CXX_VISIBILITY_PRESET hidden)
 target_compile_definitions(mlir_async_runtime PRIVATE mlir_async_runtime_EXPORTS)
+
+add_mlir_library(mlir_async_runtime_static
+  AsyncRuntime.cpp
+
+  EXCLUDE_FROM_LIBMLIR
+
+  LINK_LIBS PUBLIC
+  mlir_c_runner_utils_static
+  ${LLVM_PTHREAD_LIB}
+)

diff --git a/mlir/test/mlir-cuda-runner/all-reduce-and.mlir b/mlir/test/mlir-cuda-runner/all-reduce-and.mlir
index 9096cfe..205f4e7 100644 (file)
--- a/mlir/test/mlir-cuda-runner/all-reduce-and.mlir
+++ b/mlir/test/mlir-cuda-runner/all-reduce-and.mlir
@@ -1,4 +1,7 @@
-// RUN: mlir-cuda-runner %s --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext,%linalg_test_lib_dir/libmlir_runner_utils%shlibext --entry-point-result=void | FileCheck %s
+// RUN: mlir-cuda-runner %s \
+// RUN:   --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext
+// RUN:   --entry-point-result=void \
+// RUN: | FileCheck %s
 
 func @main() {
   %data = alloc() : memref<2x6xi32>

diff --git a/mlir/test/mlir-cuda-runner/all-reduce-max.mlir b/mlir/test/mlir-cuda-runner/all-reduce-max.mlir
index 68dbf04..32f4511 100644 (file)
--- a/mlir/test/mlir-cuda-runner/all-reduce-max.mlir
+++ b/mlir/test/mlir-cuda-runner/all-reduce-max.mlir
@@ -1,4 +1,7 @@
-// RUN: mlir-cuda-runner %s --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext,%linalg_test_lib_dir/libmlir_runner_utils%shlibext --entry-point-result=void | FileCheck %s
+// RUN: mlir-cuda-runner %s \
+// RUN:   --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext
+// RUN:   --entry-point-result=void \
+// RUN: | FileCheck %s
 
 func @main() {
   %data = alloc() : memref<2x6xi32>

diff --git a/mlir/test/mlir-cuda-runner/all-reduce-min.mlir b/mlir/test/mlir-cuda-runner/all-reduce-min.mlir
index d078d92..39c62b0 100644 (file)
--- a/mlir/test/mlir-cuda-runner/all-reduce-min.mlir
+++ b/mlir/test/mlir-cuda-runner/all-reduce-min.mlir
@@ -1,4 +1,7 @@
-// RUN: mlir-cuda-runner %s --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext,%linalg_test_lib_dir/libmlir_runner_utils%shlibext --entry-point-result=void | FileCheck %s
+// RUN: mlir-cuda-runner %s \
+// RUN:   --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext
+// RUN:   --entry-point-result=void \
+// RUN: | FileCheck %s
 
 func @main() {
   %data = alloc() : memref<2x6xi32>

diff --git a/mlir/test/mlir-cuda-runner/all-reduce-op.mlir b/mlir/test/mlir-cuda-runner/all-reduce-op.mlir
index 247d2c5..e20e386 100644 (file)
--- a/mlir/test/mlir-cuda-runner/all-reduce-op.mlir
+++ b/mlir/test/mlir-cuda-runner/all-reduce-op.mlir
@@ -1,4 +1,7 @@
-// RUN: mlir-cuda-runner %s --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext,%linalg_test_lib_dir/libmlir_runner_utils%shlibext --entry-point-result=void | FileCheck %s
+// RUN: mlir-cuda-runner %s \
+// RUN:   --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext
+// RUN:   --entry-point-result=void \
+// RUN: | FileCheck %s
 
 // CHECK-COUNT-8: [{{(5356, ){12}5356}}]
 func @main() {

diff --git a/mlir/test/mlir-cuda-runner/all-reduce-or.mlir b/mlir/test/mlir-cuda-runner/all-reduce-or.mlir
index c6a5ee5..175fda6 100644 (file)
--- a/mlir/test/mlir-cuda-runner/all-reduce-or.mlir
+++ b/mlir/test/mlir-cuda-runner/all-reduce-or.mlir
@@ -1,4 +1,7 @@
-// RUN: mlir-cuda-runner %s --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext,%linalg_test_lib_dir/libmlir_runner_utils%shlibext --entry-point-result=void | FileCheck %s
+// RUN: mlir-cuda-runner %s \
+// RUN:   --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext
+// RUN:   --entry-point-result=void \
+// RUN: | FileCheck %s
 
 func @main() {
   %data = alloc() : memref<2x6xi32>

diff --git a/mlir/test/mlir-cuda-runner/all-reduce-region.mlir b/mlir/test/mlir-cuda-runner/all-reduce-region.mlir
index dcff9d0..dd70a5b 100644 (file)
--- a/mlir/test/mlir-cuda-runner/all-reduce-region.mlir
+++ b/mlir/test/mlir-cuda-runner/all-reduce-region.mlir
@@ -1,4 +1,7 @@
-// RUN: mlir-cuda-runner %s --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext,%linalg_test_lib_dir/libmlir_runner_utils%shlibext --entry-point-result=void | FileCheck %s
+// RUN: mlir-cuda-runner %s \
+// RUN:   --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext
+// RUN:   --entry-point-result=void \
+// RUN: | FileCheck %s
 
 // CHECK: [{{(35, ){34}35}}]
 func @main() {

diff --git a/mlir/test/mlir-cuda-runner/all-reduce-xor.mlir b/mlir/test/mlir-cuda-runner/all-reduce-xor.mlir
index 9ca81df..cf0d409 100644 (file)
--- a/mlir/test/mlir-cuda-runner/all-reduce-xor.mlir
+++ b/mlir/test/mlir-cuda-runner/all-reduce-xor.mlir
@@ -1,4 +1,7 @@
-// RUN: mlir-cuda-runner %s --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext,%linalg_test_lib_dir/libmlir_runner_utils%shlibext --entry-point-result=void | FileCheck %s
+// RUN: mlir-cuda-runner %s \
+// RUN:   --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext
+// RUN:   --entry-point-result=void \
+// RUN: | FileCheck %s
 
 func @main() {
   %data = alloc() : memref<2x6xi32>

diff --git a/mlir/test/mlir-cuda-runner/gpu-to-cubin.mlir b/mlir/test/mlir-cuda-runner/gpu-to-cubin.mlir
index e2340bd..8480053 100644 (file)
--- a/mlir/test/mlir-cuda-runner/gpu-to-cubin.mlir
+++ b/mlir/test/mlir-cuda-runner/gpu-to-cubin.mlir
@@ -1,4 +1,7 @@
-// RUN: mlir-cuda-runner %s --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext,%linalg_test_lib_dir/libmlir_runner_utils%shlibext --entry-point-result=void | FileCheck %s
+// RUN: mlir-cuda-runner %s \
+// RUN:   --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext
+// RUN:   --entry-point-result=void \
+// RUN: | FileCheck %s
 
 func @other_func(%arg0 : f32, %arg1 : memref<?xf32>) {
   %cst = constant 1 : index

diff --git a/mlir/test/mlir-cuda-runner/multiple-all-reduce.mlir b/mlir/test/mlir-cuda-runner/multiple-all-reduce.mlir
index 4427879..c4b8683 100644 (file)
--- a/mlir/test/mlir-cuda-runner/multiple-all-reduce.mlir
+++ b/mlir/test/mlir-cuda-runner/multiple-all-reduce.mlir
@@ -1,4 +1,7 @@
-// RUN: mlir-cuda-runner %s --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext,%linalg_test_lib_dir/libmlir_runner_utils%shlibext --entry-point-result=void | FileCheck %s
+// RUN: mlir-cuda-runner %s \
+// RUN:   --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext
+// RUN:   --entry-point-result=void \
+// RUN: | FileCheck %s
 
 func @main() {
   %data = alloc() : memref<2x6xf32>

diff --git a/mlir/test/mlir-cuda-runner/shuffle.mlir b/mlir/test/mlir-cuda-runner/shuffle.mlir
index b1fde24..9219bd0 100644 (file)
--- a/mlir/test/mlir-cuda-runner/shuffle.mlir
+++ b/mlir/test/mlir-cuda-runner/shuffle.mlir
@@ -1,4 +1,7 @@
-// RUN: mlir-cuda-runner %s --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext,%linalg_test_lib_dir/libmlir_runner_utils%shlibext --entry-point-result=void | FileCheck %s
+// RUN: mlir-cuda-runner %s \
+// RUN:   --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext
+// RUN:   --entry-point-result=void \
+// RUN: | FileCheck %s
 
 // CHECK: [4, 5, 6, 7, 0, 1, 2, 3, 12, -1, -1, -1, 8]
 func @main() {

diff --git a/mlir/test/mlir-cuda-runner/two-modules.mlir b/mlir/test/mlir-cuda-runner/two-modules.mlir
index 982e8aa..2f3111d 100644 (file)
--- a/mlir/test/mlir-cuda-runner/two-modules.mlir
+++ b/mlir/test/mlir-cuda-runner/two-modules.mlir
@@ -1,4 +1,7 @@
-// RUN: mlir-cuda-runner %s --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext,%linalg_test_lib_dir/libmlir_runner_utils%shlibext --entry-point-result=void | FileCheck %s
+// RUN: mlir-cuda-runner %s \
+// RUN:   --shared-libs=%cuda_wrapper_library_dir/libcuda-runtime-wrappers%shlibext
+// RUN:   --entry-point-result=void \
+// RUN: | FileCheck %s
 
 // CHECK: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]
 func @main() {

diff --git a/mlir/test/mlir-rocm-runner/gpu-to-hsaco.mlir b/mlir/test/mlir-rocm-runner/gpu-to-hsaco.mlir
index 1444774..89f4ef8 100644 (file)
--- a/mlir/test/mlir-rocm-runner/gpu-to-hsaco.mlir
+++ b/mlir/test/mlir-rocm-runner/gpu-to-hsaco.mlir
@@ -1,4 +1,7 @@
-// RUN: mlir-rocm-runner %s --shared-libs=%rocm_wrapper_library_dir/librocm-runtime-wrappers%shlibext,%linalg_test_lib_dir/libmlir_runner_utils%shlibext --entry-point-result=void | FileCheck %s
+// RUN: mlir-rocm-runner %s \
+// RUN:   --shared-libs=%rocm_wrapper_library_dir/librocm-runtime-wrappers%shlibext
+// RUN:   --entry-point-result=void \
+// RUN: | FileCheck %s
 
 func @other_func(%arg0 : f32, %arg1 : memref<?xf32>) {
   %c0 = constant 0 : index

diff --git a/mlir/test/mlir-rocm-runner/two-modules.mlir b/mlir/test/mlir-rocm-runner/two-modules.mlir
index f196c8e..95a45da 100644 (file)
--- a/mlir/test/mlir-rocm-runner/two-modules.mlir
+++ b/mlir/test/mlir-rocm-runner/two-modules.mlir
@@ -1,4 +1,7 @@
-// RUN: mlir-rocm-runner %s --shared-libs=%rocm_wrapper_library_dir/librocm-runtime-wrappers%shlibext,%linalg_test_lib_dir/libmlir_runner_utils%shlibext --entry-point-result=void | FileCheck %s
+// RUN: mlir-rocm-runner %s \
+// RUN:   --shared-libs=%rocm_wrapper_library_dir/librocm-runtime-wrappers%shlibext
+// RUN:   --entry-point-result=void \
+// RUN: | FileCheck %s
 
 // CHECK: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]
 func @main() {

diff --git a/mlir/test/mlir-rocm-runner/vecadd.mlir b/mlir/test/mlir-rocm-runner/vecadd.mlir
index 43c5e92..5975f0c 100644 (file)
--- a/mlir/test/mlir-rocm-runner/vecadd.mlir
+++ b/mlir/test/mlir-rocm-runner/vecadd.mlir
@@ -1,4 +1,7 @@
-// RUN: mlir-rocm-runner %s --shared-libs=%rocm_wrapper_library_dir/librocm-runtime-wrappers%shlibext,%linalg_test_lib_dir/libmlir_runner_utils%shlibext --entry-point-result=void | FileCheck %s
+// RUN: mlir-rocm-runner %s \
+// RUN:   --shared-libs=%rocm_wrapper_library_dir/librocm-runtime-wrappers%shlibext
+// RUN:   --entry-point-result=void \
+// RUN: | FileCheck %s
 
 func @vecadd(%arg0 : memref<?xf32>, %arg1 : memref<?xf32>, %arg2 : memref<?xf32>) {
   %c0 = constant 0 : index

diff --git a/mlir/test/mlir-rocm-runner/vector-transferops.mlir b/mlir/test/mlir-rocm-runner/vector-transferops.mlir
index 87a05c5..91e2108 100644 (file)
--- a/mlir/test/mlir-rocm-runner/vector-transferops.mlir
+++ b/mlir/test/mlir-rocm-runner/vector-transferops.mlir
@@ -1,4 +1,7 @@
-// RUN: mlir-rocm-runner %s --shared-libs=%rocm_wrapper_library_dir/librocm-runtime-wrappers%shlibext,%linalg_test_lib_dir/libmlir_runner_utils%shlibext --entry-point-result=void | FileCheck %s
+// RUN: mlir-rocm-runner %s \
+// RUN:   --shared-libs=%rocm_wrapper_library_dir/librocm-runtime-wrappers%shlibext
+// RUN:   --entry-point-result=void \
+// RUN: | FileCheck %s
 
 func @vectransferx2(%arg0 : memref<?xf32>, %arg1 : memref<?xf32>) {
   %cst = constant 1 : index

diff --git a/mlir/tools/mlir-cuda-runner/CMakeLists.txt b/mlir/tools/mlir-cuda-runner/CMakeLists.txt
index 8a99418..9f00c26 100644 (file)
--- a/mlir/tools/mlir-cuda-runner/CMakeLists.txt
+++ b/mlir/tools/mlir-cuda-runner/CMakeLists.txt
@@ -37,6 +37,7 @@ if(MLIR_CUDA_RUNNER_ENABLED)
   target_link_libraries(cuda-runtime-wrappers
     PUBLIC
     LLVMSupport
+    mlir_runner_utils_static
     ${CUDA_RUNTIME_LIBRARY}
   )
 

diff --git a/mlir/tools/mlir-rocm-runner/CMakeLists.txt b/mlir/tools/mlir-rocm-runner/CMakeLists.txt
index a4da496..910a494 100644 (file)
--- a/mlir/tools/mlir-rocm-runner/CMakeLists.txt
+++ b/mlir/tools/mlir-rocm-runner/CMakeLists.txt
@@ -61,6 +61,7 @@ if(MLIR_ROCM_RUNNER_ENABLED)
   target_link_libraries(rocm-runtime-wrappers
     PUBLIC
     LLVMSupport
+    mlir_runner_utils_static
     ${ROCM_RUNTIME_LIBRARY}
   )