[Mlir-commits] [mlir] [mlir][nvvm] Fix the PTX lowering of wgmma.mma_async (PR #76150)

[Adam Paszke]

https://github.com/apaszke created https://github.com/llvm/llvm-project/pull/76150

The default layout of A and B matrices is row- and column-major respectively, meaning that the transpose flags have opposite meanings between those two operands.

>From 5eef4c04e62d937f48db5422b8e2e45cdd165a86 Mon Sep 17 00:00:00 2001
From: Adam Paszke <apaszke at google.com>
Date: Thu, 21 Dec 2023 13:05:49 +0000
Subject: [PATCH] [mlir][nvvm] Fix the PTX lowering of wgmma.mma_async

The default layout of A and B matrices is row- and column-major
respectively, meaning that the transpose flags have opposite meanings
between those two operands.
---
 mlir/lib/Dialect/LLVMIR/IR/NVVMDialect.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mlir/lib/Dialect/LLVMIR/IR/NVVMDialect.cpp b/mlir/lib/Dialect/LLVMIR/IR/NVVMDialect.cpp
index 4f5d71e10f68c1..a2117bb5b428a0 100644
--- a/mlir/lib/Dialect/LLVMIR/IR/NVVMDialect.cpp
+++ b/mlir/lib/Dialect/LLVMIR/IR/NVVMDialect.cpp
@@ -1003,7 +1003,7 @@ void NVVM::WgmmaMmaAsyncOp::getAsmValues(
         {makeConstantI32(rewriter, static_cast<int>(getLayoutA())),
          mlir::NVVM::PTXRegisterMod::Read});
     asmValues.push_back(
-        {makeConstantI32(rewriter, static_cast<int>(getLayoutB())),
+        {makeConstantI32(rewriter, static_cast<int>(1 - getLayoutB())),
          mlir::NVVM::PTXRegisterMod::Read});
   }
 }