[Mlir-commits] [mlir] faf697e - [mlir] Add support for f8 data types to LLVM dialect types

[Eugene Zhulenev]

Author: Eugene Zhulenev
Date: 2023-03-09T18:52:36-08:00
New Revision: faf697e49bad099b36770c7738f36917e3d90d2d

URL: https://github.com/llvm/llvm-project/commit/faf697e49bad099b36770c7738f36917e3d90d2d
DIFF: https://github.com/llvm/llvm-project/commit/faf697e49bad099b36770c7738f36917e3d90d2d.diff

LOG: [mlir] Add support for f8 data types to LLVM dialect types

This change allows using fp8 pointers when exporting to LLVM, because we anyway export them as opaque pointers, however full support of fp8 types is not yet implemented on the LLVM side.

Differential Revision: https://reviews.llvm.org/D143008

Added: 
    

Modified: 
    mlir/lib/Dialect/LLVMIR/IR/LLVMTypes.cpp
    mlir/test/Target/LLVMIR/llvmir.mlir

Removed: 
    


################################################################################
diff  --git a/mlir/lib/Dialect/LLVMIR/IR/LLVMTypes.cpp b/mlir/lib/Dialect/LLVMIR/IR/LLVMTypes.cpp
index 2e6bf86022fa3..4b3b1e183903f 100644
--- a/mlir/lib/Dialect/LLVMIR/IR/LLVMTypes.cpp
+++ b/mlir/lib/Dialect/LLVMIR/IR/LLVMTypes.cpp
@@ -731,6 +731,8 @@ bool mlir::LLVM::isCompatibleOuterType(Type type) {
       Float64Type,
       Float80Type,
       Float128Type,
+      Float8E4M3FNType,
+      Float8E5M2Type,
       LLVMArrayType,
       LLVMFunctionType,
       LLVMLabelType,
@@ -801,6 +803,8 @@ static bool isCompatibleImpl(Type type, DenseSet<Type> &compatibleTypes) {
             Float64Type,
             Float80Type,
             Float128Type,
+            Float8E4M3FNType,
+            Float8E5M2Type,
             LLVMLabelType,
             LLVMMetadataType,
             LLVMPPCFP128Type,

diff  --git a/mlir/test/Target/LLVMIR/llvmir.mlir b/mlir/test/Target/LLVMIR/llvmir.mlir
index 82d05dbf0015a..d19f49d4b2938 100644
--- a/mlir/test/Target/LLVMIR/llvmir.mlir
+++ b/mlir/test/Target/LLVMIR/llvmir.mlir
@@ -944,6 +944,11 @@ llvm.func @vector_splat_nonzero_scalable() -> vector<[4]xf32> {
   llvm.return %0 : vector<[4]xf32>
 }
 
+// CHECK-LABEL: @f8_ptrs(ptr {{%.*}}, ptr {{%.*}})
+llvm.func @f8_ptrs(%arg0: !llvm.ptr<f8E5M2>, %arg1: !llvm.ptr<f8E4M3FN>) {
+  llvm.return
+}
+
 // CHECK-LABEL: @ops
 llvm.func @ops(%arg0: f32, %arg1: f32, %arg2: i32, %arg3: i32) -> !llvm.struct<(f32, i32)> {
 // CHECK-NEXT: fsub float %0, %1