[Mlir-commits] [mlir] [mlir] Change default NVVM compilation to `fatbin` from `bin` (PR #70052)

llvmlistbot at llvm.org llvmlistbot at llvm.org
Tue Oct 24 07:42:55 PDT 2023


llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-mlir-gpu

Author: Guray Ozen (grypp)

<details>
<summary>Changes</summary>

Change the NVVM assembly to `fatbin` as it is executable for multiple architectures. Using `bin` caused test errors at runtime in the test systems.

---
Full diff: https://github.com/llvm/llvm-project/pull/70052.diff


1 Files Affected:

- (modified) mlir/test/lib/Dialect/GPU/TestLowerToNVVM.cpp (+2-2) 


``````````diff
diff --git a/mlir/test/lib/Dialect/GPU/TestLowerToNVVM.cpp b/mlir/test/lib/Dialect/GPU/TestLowerToNVVM.cpp
index c4cc0d5ae38d9be..ed7634fbecf49fd 100644
--- a/mlir/test/lib/Dialect/GPU/TestLowerToNVVM.cpp
+++ b/mlir/test/lib/Dialect/GPU/TestLowerToNVVM.cpp
@@ -60,7 +60,7 @@ struct TestLowerToNVVMOptions
   PassOptions::Option<std::string> cubinFormat{
       *this, "cubin-format",
       llvm::cl::desc("Compilation format to use to serialize to cubin."),
-      llvm::cl::init("bin")};
+      llvm::cl::init("fatbin")};
   PassOptions::Option<int> optLevel{
       *this, "opt-level",
       llvm::cl::desc("Optimization level for NVVM compilation"),
@@ -156,4 +156,4 @@ void registerTestLowerToNVVM() {
 }
 } // namespace test
 } // namespace mlir
-#endif // MLIR_CUDA_CONVERSIONS_ENABLED
+#endif // MLIR_CUDA_CONVERSIONS_ENABLED
\ No newline at end of file

``````````

</details>


https://github.com/llvm/llvm-project/pull/70052


More information about the Mlir-commits mailing list