From a05990485c7000036ab34fe93a5ed0cbeb93b997 Mon Sep 17 00:00:00 2001
From: Whitney Tsang <whitney.tsang@intel.com>
Date: Thu, 25 Apr 2024 16:04:47 +0000
Subject: [PATCH] Revert "Replace GenISA ftof usage with arith::TruncFOp"

This reverts commit 548612eddd9a56c1d9477b2f6c5ed97cee79f3a9.
---
 .../ElementwiseOpToLLVM.cpp                   | 32 +++++++++++++++++--
 1 file changed, 30 insertions(+), 2 deletions(-)

diff --git a/third_party/intel/lib/TritonIntelGPUToLLVM/ElementwiseOpToLLVM.cpp b/third_party/intel/lib/TritonIntelGPUToLLVM/ElementwiseOpToLLVM.cpp
index 6d79c9606c..67d6da8aef 100644
--- a/third_party/intel/lib/TritonIntelGPUToLLVM/ElementwiseOpToLLVM.cpp
+++ b/third_party/intel/lib/TritonIntelGPUToLLVM/ElementwiseOpToLLVM.cpp
@@ -1,4 +1,5 @@
 #include "PatternTritonGPUOpToLLVM.h"
+#include "mlir/Conversion/ArithCommon/AttrToLLVMConverter.h"
 #include "mlir/Dialect/LLVMIR/LLVMDialect.h"
 #include "mlir/IR/MLIRContext.h"
 #include "triton/Conversion/TritonGPUToLLVM/TargetInfoBase.h"
@@ -1329,6 +1330,33 @@ struct FpToFpOpConversion
                                  ConversionPatternRewriter &rewriter,
                                  const Value &v, const RoundingMode rounding) {
     MLIRContext *ctx = rewriter.getContext();
+
+    LLVM::RoundingMode roundingMode;
+    switch (rounding) {
+    case RoundingMode::RTNE:
+      roundingMode = LLVM::RoundingMode::NearestTiesToEven;
+      break;
+    case RoundingMode::RTZ:
+      roundingMode = LLVM::RoundingMode::TowardZero;
+      break;
+    default:
+      llvm::errs() << "WARNING: unsupported rounding mode for f32->f16 "
+                      "conversion: "
+                   << stringifyRoundingMode(rounding) << "\n";
+      llvm_unreachable("");
+    }
+
+    NamedAttrList convertedAttr;
+    convertedAttr.set(LLVM::ConstrainedFPTruncIntr::getRoundingModeAttrName(),
+                      LLVM::RoundingModeAttr::get(ctx, roundingMode));
+    convertedAttr.set(
+        LLVM::ConstrainedFPTruncIntr::getFPExceptionBehaviorAttrName(),
+        arith::getLLVMDefaultFPExceptionBehavior(*ctx));
+    return rewriter.create<LLVM::ConstrainedFPTruncIntr>(loc, f16_ty, v,
+                                                         convertedAttr);
+
+#if 0
+    // FIXME: test_typeconvert_downcast fails when lower to arith::TruncFOp.
     arith::RoundingMode roundingMode;
     switch (rounding) {
     case RoundingMode::RTNE:
@@ -1345,6 +1373,7 @@ struct FpToFpOpConversion
     }
     return rewriter.create<arith::TruncFOp>(
         loc, f16_ty, v, arith::RoundingModeAttr::get(ctx, roundingMode));
+#endif
   }
 
   std::pair<ConverterT, size_t>
@@ -1994,10 +2023,9 @@ struct TruncFOpConversion
       return {// Trunc uses the default rounding mode: RTNE
               FpToFpOpConversion::convertFp32ToBf16(
                   loc, rewriter, operands[0][0], RoundingMode::RTNE)};
-    } else if (!op.getRoundingModeAttr()) {
+    } else {
       return {rewriter.create<LLVM::FPTruncOp>(loc, elemTy, operands[0][0])};
     }
-    return {};
   }
 };