Fixing permlanex16 intrinsic failure

Signed-off-by: Muzammiluddin Syed <[email protected]>

Author: Muzammiluddin-Syed-ECE

mlir/include/mlir/Dialect/LLVMIR/ROCDLOps.td

@@ -670,13 +670,13 @@ def ROCDL_DPPUpdateOp : ROCDL_IntrOp<"update.dpp", [], [0],
 
 // PermLaneX16 intrinsic operation
 def ROCDL_PermlaneX16Op : ROCDL_IntrOp<"permlanex16", [], [0],
-    [AllTypesMatch<["res", "old", "src0", "src1", "src2"]>], 1, 0, 0,
+    [AllTypesMatch<["res", "old", "src0"]>, AllTypesMatch<["src1", "src2"]>], 1, 0, 0,
     [4, 5], ["fi", "boundControl"]>,
   Arguments<(ins LLVM_Type:$old, LLVM_Type:$src0, LLVM_Type:$src1, LLVM_Type:$src2,
              I1Attr:$fi, I1Attr:$boundControl)> {
   let results = (outs LLVM_Type:$res);
   let assemblyFormat = [{
-    attr-dict $old `,` $src0 `,` $src1 `,` $src2 `,` $fi `,` $boundControl `:` type($src0)
+    attr-dict $old `,` $src0 `,` $src1 `,` $src2 `,` $fi `,` $boundControl `:` type($src0) `,` type($src1)
   }];
   let description = [{
     Performs a `permlanex16` operation with the given operands, applying the

mlir/lib/Dialect/GPU/Transforms/SubgroupReduceLowering.cpp

@@ -405,14 +405,9 @@ Value createSubgroupDPPReduction(OpBuilder &b, Location loc, Value input,
 
   const int allRows = 0xf;
   const int allBanks = 0xf;
-  auto uint32Type = b.getIntegerType(32, false);
   if (ci.clusterSize >= 32) {
-    // auto permArg = b.getI32IntegerAttr(15);
-    // Value dppResult = b.create<amdgpu::DPPOp>(
-    //     loc, result.getType(), result, result, amdgpu::DPPPerm::row_bcast_15,
-    //     b.getUnitAttr(), 0xa, allBanks, false);
     auto uIntMax = llvm::APInt::getMaxValue(32u);
-    Value uIntMaxConst = b.create<LLVM::ConstantOp>(loc, uint32Type, uIntMax);
+    Value uIntMaxConst = b.create<LLVM::ConstantOp>(loc, b.getI32Type(), uIntMax);
     Value dppResult = b.create<ROCDL::PermlaneX16Op>(loc, input.getType(), result, result, uIntMaxConst, uIntMaxConst, true, false);
     result = vector::makeArithReduction(b, loc, gpu::convertReductionKind(mode),
                                         result, dppResult);