[mlir][nvgpu] Add nvgpu.rcp OP (#100965)

This PR introduces a new OP for reciprocal calculation for `vector`
types using `nvvm.rcp` OPs. Currently, it supports only f32 types

---------

Co-authored-by: jingzec <[email protected]>

Author: Observer007

mlir/include/mlir/Dialect/NVGPU/IR/NVGPU.td

@@ -20,6 +20,7 @@
 #ifndef NVGPU
 #define NVGPU
 
+include "mlir/Interfaces/InferTypeOpInterface.td"
 include "mlir/Interfaces/SideEffectInterfaces.td"
 include "mlir/IR/AttrTypeBase.td"
 include "mlir/IR/OpBase.td"
@@ -109,10 +110,22 @@ def TensorMapInterleaveKind : I32EnumAttr<"TensorMapInterleaveKind",
   let cppNamespace = "::mlir::nvgpu";
 }
 
+def RcpApprox : I32EnumAttrCase<"APPROX", 0, "approx">;
+def RcpRN     : I32EnumAttrCase<"RN", 1, "rn">;
+def RcpRZ     : I32EnumAttrCase<"RZ", 2, "rz">;
+def RcpRM     : I32EnumAttrCase<"RM", 3, "rm">;
+def RcpRP     : I32EnumAttrCase<"RP", 4, "rp">;
+def RcpRoundingMode   : I32EnumAttr<"RcpRoundingMode", "Rounding mode of rcp",
+  [RcpApprox, RcpRN, RcpRZ, RcpRM, RcpRP]> {
+  let genSpecializedAttr = 0;
+  let cppNamespace = "::mlir::nvgpu";
+}
+
 def TensorMapSwizzleAttr : EnumAttr<NVGPU_Dialect, TensorMapSwizzleKind, "swizzle">;
 def TensorMapL2PromoAttr : EnumAttr<NVGPU_Dialect, TensorMapL2PromoKind, "l2promo">;
 def TensorMapOOBAttr : EnumAttr<NVGPU_Dialect, TensorMapOOBKind, "oob">;
 def TensorMapInterleaveAttr : EnumAttr<NVGPU_Dialect, TensorMapInterleaveKind, "interleave">;
+def RcpRoundingModeAttr : EnumAttr<NVGPU_Dialect, RcpRoundingMode, "rcp_rounding_mode">;
 
 //===----------------------------------------------------------------------===//
 // NVGPU Type Definitions
@@ -802,4 +815,24 @@ def NVGPU_WarpgroupMmaInitAccumulatorOp : NVGPU_Op<"warpgroup.mma.init.accumulat
   let hasVerifier = 1;
 }
 
+def NVGPU_RcpOp : NVGPU_Op<"rcp", [Pure,
+                                   SameOperandsAndResultType]> {
+  let summary = "The reciprocal calculation for vector types";
+  let description = [{
+    Reciprocal calculation for `vector` types using `nvvm.rcp` OPs.
+
+    Currently, only the `approx` rounding mode and `ftz` are supported, and only for the `f32` type.
+
+    The input and output must be of the same vector type and shape.
+  }];
+  let arguments = (ins VectorOf<[F32]>:$in,
+                       DefaultValuedAttr<RcpRoundingModeAttr, "RcpRoundingMode::APPROX">:$rounding,
+                       UnitAttr:$ftz);
+  let results = (outs VectorOf<[F32]>:$out);
+  let assemblyFormat = [{
+    $in `{` `rounding` `=` $rounding (`,` `ftz` $ftz^)? `}` 
+    attr-dict `:` type($out)
+  }];
+  let hasVerifier = 1;
+}
 #endif // NVGPU

mlir/include/mlir/Dialect/NVGPU/IR/NVGPUDialect.h

@@ -17,6 +17,7 @@
 #include "mlir/IR/BuiltinTypes.h"
 #include "mlir/IR/Dialect.h"
 #include "mlir/IR/OpDefinition.h"
+#include "mlir/Interfaces/InferTypeOpInterface.h"
 #include "mlir/Interfaces/SideEffectInterfaces.h"
 
 #include "mlir/Dialect/NVGPU/IR/NVGPUEnums.h.inc"

mlir/lib/Conversion/NVGPUToNVVM/NVGPUToNVVM.cpp

@@ -11,6 +11,7 @@
 #include "mlir/Conversion/GPUCommon/GPUCommonPass.h"
 #include "mlir/Conversion/LLVMCommon/ConversionTarget.h"
 #include "mlir/Conversion/LLVMCommon/Pattern.h"
+#include "mlir/Conversion/LLVMCommon/VectorPattern.h"
 #include "mlir/Dialect/Arith/IR/Arith.h"
 #include "mlir/Dialect/GPU/IR/GPUDialect.h"
 #include "mlir/Dialect/LLVMIR/LLVMDialect.h"
@@ -1666,6 +1667,40 @@ struct NVGPUTmaPrefetchOpLowering
   }
 };
 
+struct NVGPURcpOpLowering : public ConvertOpToLLVMPattern<nvgpu::RcpOp> {
+  using ConvertOpToLLVMPattern<nvgpu::RcpOp>::ConvertOpToLLVMPattern;
+  LogicalResult
+  matchAndRewrite(nvgpu::RcpOp op, OpAdaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const override {
+    ImplicitLocOpBuilder b(op->getLoc(), rewriter);
+    auto i64Ty = b.getI64Type();
+    auto f32Ty = b.getF32Type();
+    VectorType inTy = op.getIn().getType();
+    // apply rcp.approx.ftz.f on each element in vector.
+    auto convert1DVec = [&](Type llvm1DVectorTy, Value inVec) {
+      Value ret1DVec = b.create<LLVM::UndefOp>(llvm1DVectorTy);
+      int numElems = llvm::cast<VectorType>(llvm1DVectorTy).getNumElements();
+      for (int i = 0; i < numElems; i++) {
+        Value idx = b.create<LLVM::ConstantOp>(i64Ty, b.getI64IntegerAttr(i));
+        Value elem = b.create<LLVM::ExtractElementOp>(inVec, idx);
+        Value dst = b.create<NVVM::RcpApproxFtzF32Op>(f32Ty, elem);
+        ret1DVec = b.create<LLVM::InsertElementOp>(ret1DVec, dst, idx);
+      }
+      return ret1DVec;
+    };
+    if (inTy.getRank() == 1) {
+      rewriter.replaceOp(op, convert1DVec(inTy, adaptor.getIn()));
+      return success();
+    }
+    return LLVM::detail::handleMultidimensionalVectors(
+        op.getOperation(), adaptor.getOperands(), *(this->getTypeConverter()),
+        [&](Type llvm1DVectorTy, ValueRange operands) -> Value {
+          OpAdaptor adaptor(operands);
+          return convert1DVec(llvm1DVectorTy, adaptor.getIn());
+        },
+        rewriter);
+  }
+};
 } // namespace
 
 void mlir::populateNVGPUToNVVMConversionPatterns(LLVMTypeConverter &converter,
@@ -1688,5 +1723,5 @@ void mlir::populateNVGPUToNVVMConversionPatterns(LLVMTypeConverter &converter,
       NVGPUWarpgroupMmaInitAccumulatorOpLowering, // nvgpu.warpgroup.mma.init.accumulator
       MmaSyncOptoNVVM, MmaLdMatrixOpToNVVM, NVGPUAsyncCopyLowering,
       NVGPUAsyncCreateGroupLowering, NVGPUAsyncWaitLowering,
-      NVGPUMmaSparseSyncLowering>(converter);
+      NVGPUMmaSparseSyncLowering, NVGPURcpOpLowering>(converter);
 }

mlir/lib/Dialect/NVGPU/IR/NVGPUDialect.cpp

@@ -644,6 +644,21 @@ LogicalResult WarpgroupMmaInitAccumulatorOp::verify() {
   return success();
 }
 
+//===----------------------------------------------------------------------===//
+// RcpOp
+//===----------------------------------------------------------------------===//
+
+LogicalResult RcpOp::verify() {
+  RcpRoundingModeAttr rounding = getRoundingAttr();
+  bool ftz = getFtz();
+  // Currently, only `rcp_approx` and `ftz` is supported.
+  if (rounding.getValue() != RcpRoundingMode::APPROX || !ftz) {
+    return emitOpError() << "has a limitation. " << rounding
+                         << " or non-ftz is not supported yet.";
+  }
+  return success();
+}
+
 //===----------------------------------------------------------------------===//
 // TableGen'd dialect, type, and op definitions
 //===----------------------------------------------------------------------===//

mlir/test/Conversion/NVGPUToNVVM/nvgpu-to-nvvm.mlir

@@ -1339,3 +1339,18 @@ module attributes {transform.with_named_sequence} {
     transform.yield
   }
 }
+
+// CHECK-LABEL: @rcp_approx_ftz_f32
+// CHECK-SAME:  %[[IN:.*]]: vector<32x16xf32>
+func.func @rcp_approx_ftz_f32(%in: vector<32x16xf32>) {
+  // CHECK: %[[IN_LLVM:.*]] = builtin.unrealized_conversion_cast %[[IN]] : vector<32x16xf32> to !llvm.array<32 x vector<16xf32>>
+  // CHECK: %[[IN1DVEC:.*]] = llvm.extractvalue %[[IN_LLVM]][0] : !llvm.array<32 x vector<16xf32>>
+  // CHECK: %[[OUT1DVEC:.*]] = llvm.mlir.undef : vector<16xf32>
+  // CHECK: %[[IDX_0:.+]] = llvm.mlir.constant(0 : i64) : i64
+  // CHECK: %[[ELEM_0:.*]] = llvm.extractelement %[[IN1DVEC]][%[[IDX_0]] : i64]
+  // CHECK: %[[ELEM_RCP0:.*]] = nvvm.rcp.approx.ftz.f %[[ELEM_0]] : f32
+  // CHECK: llvm.insertelement %[[ELEM_RCP0]], %[[OUT1DVEC]][%[[IDX_0]] : i64] : vector<16xf32>
+  // CHECK-COUNT-511: nvvm.rcp.approx.ftz.f
+  %out = nvgpu.rcp %in {rounding = approx, ftz} : vector<32x16xf32>
+  return
+}

mlir/test/Dialect/NVGPU/invalid.mlir

@@ -336,3 +336,21 @@ func.func @tma_generate_descriptor_incorrect_last_dim(%desc: !desc,  %buffer2: m
   nvgpu.tma.async.load %desc[%c0, %c0], %mbarrier[%c0] to %buffer2 : !desc, !mbarrier -> memref<64x128xf32,3>
   return
 }
+// -----
+
+func.func @rcp_unsupported_rounding_0(%in : vector<16xf32>) {
+  // expected-error @+1 {{'nvgpu.rcp' op has a limitation. #nvgpu<rcp_rounding_mode rn> or non-ftz is not supported yet.}}
+  %out = nvgpu.rcp %in {rounding = rn, ftz} : vector<16xf32>
+}
+// -----
+
+func.func @rcp_unsupported_rounding_1(%in : vector<16xf32>) {
+  // expected-error @+1 {{'nvgpu.rcp' op has a limitation. #nvgpu<rcp_rounding_mode rz> or non-ftz is not supported yet.}}
+  %out = nvgpu.rcp %in {rounding = rz} : vector<16xf32>
+}
+// -----
+
+func.func @rcp_unsupported_ftz(%in : vector<16xf32>) {
+  // expected-error @+1 {{'nvgpu.rcp' op has a limitation. #nvgpu<rcp_rounding_mode approx> or non-ftz is not supported yet.}}
+  %out = nvgpu.rcp %in {rounding = approx} : vector<16xf32>
+}