[mlir][VectorOps] Generalized vector.print to i32/i64

[aartbik]

Summary

Lowering to LLVM IR was restricted to float/double.
This CL also adds the integral values.

[ftynse]

mlir/test/Conversion/VectorToLLVM/vector-to-llvm.mlir

403

Nit: let's drop the arg[0-9]+ pattern and just use .*, we don't guarantee anything on SSA names

mlir/test/mlir-cpu-runner/mlir_runner_utils.cpp

81

Cool, I did not know these macros existed!

[nicolasvasilache]

Thanks Aart!

[aartbik]

addressed comments

[aartbik]

mlir/test/Conversion/VectorToLLVM/vector-to-llvm.mlir

403

Done, also for the other occurrences for consistency.

mlir/test/mlir-cpu-runner/mlir_runner_utils.cpp

81

A great way to keep our code portable!

mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVM.cpp

matchAndRewrite(Operation *op, ArrayRef<Value> operands,
     if (lowering.convertType(printType) == nullptr)
       return matchFailure();
 
     // Make sure element type has runtime support (currently just Float/Double).
     VectorType vectorType = printType.dyn_cast<VectorType>();
     Type eltType = vectorType ? vectorType.getElementType() : printType;
     int64_t rank = vectorType ? vectorType.getRank() : 0;
     Operation *printer;
-    if (eltType.isF32())
+    if (eltType.isInteger(32))
+      printer = getPrintI32(op);
+    else if (eltType.isInteger(64))
+      printer = getPrintI64(op);
+    else if (eltType.isF32())
       printer = getPrintFloat(op);
     else if (eltType.isF64())
       printer = getPrintDouble(op);
     else
       return matchFailure();
 
     // Unroll vector into elementary print calls.
     emitRanks(rewriter, op, adaptor.source(), vectorType, printer, rank);
static Operation *getPrint(Operation *op, LLVM::LLVMDialect *dialect,
     OpBuilder moduleBuilder(module.getBodyRegion());
     return moduleBuilder.create<LLVM::LLVMFuncOp>(
         op->getLoc(), name,
         LLVM::LLVMType::getFunctionTy(LLVM::LLVMType::getVoidTy(dialect),
                                       params, /*isVarArg=*/false));
   }
 
   // Helpers for method names.
+  Operation *getPrintI32(Operation *op) const {
+    LLVM::LLVMDialect *dialect = lowering.getDialect();
+    return getPrint(op, dialect, "print_i32",
+                    LLVM::LLVMType::getInt32Ty(dialect));
+  }
+  Operation *getPrintI64(Operation *op) const {
+    LLVM::LLVMDialect *dialect = lowering.getDialect();
+    return getPrint(op, dialect, "print_i64",
+                    LLVM::LLVMType::getInt64Ty(dialect));
+  }
   Operation *getPrintFloat(Operation *op) const {
     LLVM::LLVMDialect *dialect = lowering.getDialect();
     return getPrint(op, dialect, "print_f32",
                     LLVM::LLVMType::getFloatTy(dialect));
   }
   Operation *getPrintDouble(Operation *op) const {
     LLVM::LLVMDialect *dialect = lowering.getDialect();
     return getPrint(op, dialect, "print_f64",

mlir/test/Conversion/VectorToLLVM/vector-to-llvm.mlir

 //       CHECK:   llvm.insertvalue {{.*}}[1] : !llvm<"[2 x <3 x float>]">
 //       CHECK:   llvm.return {{.*}} : !llvm<"[2 x <3 x float>]">
 
 func @shuffle_1D_direct(%arg0: vector<2xf32>, %arg1: vector<2xf32>) -> vector<2xf32> {
   %1 = vector.shuffle %arg0, %arg1 [0, 1] : vector<2xf32>, vector<2xf32>
   return %1 : vector<2xf32>
 }
 // CHECK-LABEL: llvm.func @shuffle_1D_direct
-// CHECK-SAME: %[[A:arg[0-9]+]]: !llvm<"<2 x float>">
-// CHECK-SAME: %[[B:arg[0-9]+]]: !llvm<"<2 x float>">
+// CHECK-SAME: %[[A:.*]]: !llvm<"<2 x float>">,
+// CHECK-SAME: %[[B:.*]]: !llvm<"<2 x float>">
 //       CHECK:   %[[s:.*]] = llvm.shufflevector %[[A]], %[[B]] [0, 1] : !llvm<"<2 x float>">, !llvm<"<2 x float>">
 //       CHECK:   llvm.return %[[s]] : !llvm<"<2 x float>">
 
 func @shuffle_1D(%arg0: vector<2xf32>, %arg1: vector<3xf32>) -> vector<5xf32> {
   %1 = vector.shuffle %arg0, %arg1 [4, 3, 2, 1, 0] : vector<2xf32>, vector<3xf32>
   return %1 : vector<5xf32>
 }
 // CHECK-LABEL: llvm.func @shuffle_1D
-// CHECK-SAME: %[[A:arg[0-9]+]]: !llvm<"<2 x float>">
-// CHECK-SAME: %[[B:arg[0-9]+]]: !llvm<"<3 x float>">
+// CHECK-SAME: %[[A:.*]]: !llvm<"<2 x float>">,
+// CHECK-SAME: %[[B:.*]]: !llvm<"<3 x float>">
 //       CHECK:   %[[u0:.*]] = llvm.mlir.undef : !llvm<"<5 x float>">
 //       CHECK:   %[[c2:.*]] = llvm.mlir.constant(2 : index) : !llvm.i64
 //       CHECK:   %[[e1:.*]] = llvm.extractelement %[[B]][%[[c2]] : !llvm.i64] : !llvm<"<3 x float>">
 //       CHECK:   %[[c0:.*]] = llvm.mlir.constant(0 : index) : !llvm.i64
 //       CHECK:   %[[i1:.*]] = llvm.insertelement %[[e1]], %[[u0]][%[[c0]] : !llvm.i64] : !llvm<"<5 x float>">
 //       CHECK:   %[[c1:.*]] = llvm.mlir.constant(1 : index) : !llvm.i64
 //       CHECK:   %[[e2:.*]] = llvm.extractelement %[[B]][%[[c1]] : !llvm.i64] : !llvm<"<3 x float>">
 //       CHECK:   %[[c1:.*]] = llvm.mlir.constant(1 : index) : !llvm.i64

 //       CHECK:   %[[i5:.*]] = llvm.insertelement %[[e5]], %[[i4]][%[[c4]] : !llvm.i64] : !llvm<"<5 x float>">
 //       CHECK:   llvm.return %[[i5]] : !llvm<"<5 x float>">
 
 func @shuffle_2D(%a: vector<1x4xf32>, %b: vector<2x4xf32>) -> vector<3x4xf32> {
   %1 = vector.shuffle %a, %b[1, 0, 2] : vector<1x4xf32>, vector<2x4xf32>
   return %1 : vector<3x4xf32>
 }
 // CHECK-LABEL: llvm.func @shuffle_2D
-// CHECK-SAME: %[[A:arg[0-9]+]]: !llvm<"[1 x <4 x float>]">
-// CHECK-SAME: %[[B:arg[0-9]+]]: !llvm<"[2 x <4 x float>]">
+// CHECK-SAME: %[[A:.*]]: !llvm<"[1 x <4 x float>]">,
+// CHECK-SAME: %[[B:.*]]: !llvm<"[2 x <4 x float>]">
 //       CHECK:   %[[u0:.*]] = llvm.mlir.undef : !llvm<"[3 x <4 x float>]">
 //       CHECK:   %[[e1:.*]] = llvm.extractvalue %[[B]][0] : !llvm<"[2 x <4 x float>]">
 //       CHECK:   %[[i1:.*]] = llvm.insertvalue %[[e1]], %[[u0]][0] : !llvm<"[3 x <4 x float>]">
 //       CHECK:   %[[e2:.*]] = llvm.extractvalue %[[A]][0] : !llvm<"[1 x <4 x float>]">
 //       CHECK:   %[[i2:.*]] = llvm.insertvalue %[[e2]], %[[i1]][1] : !llvm<"[3 x <4 x float>]">
 //       CHECK:   %[[e3:.*]] = llvm.extractvalue %[[B]][1] : !llvm<"[2 x <4 x float>]">
 //       CHECK:   %[[i3:.*]] = llvm.insertvalue %[[e3]], %[[i2]][2] : !llvm<"[3 x <4 x float>]">
 //       CHECK:   llvm.return %[[i3]] : !llvm<"[3 x <4 x float>]">
 
 func @extract_element(%arg0: vector<16xf32>) -> f32 {
   %0 = constant 15 : i32
   %1 = vector.extractelement %arg0[%0 : i32]: vector<16xf32>
   return %1 : f32
 }
 // CHECK-LABEL: llvm.func @extract_element
-// CHECK-SAME: %[[A:arg[0-9]+]]: !llvm<"<16 x float>">
+// CHECK-SAME: %[[A:.*]]: !llvm<"<16 x float>">
 //       CHECK:   %[[c:.*]] = llvm.mlir.constant(15 : i32) : !llvm.i32
 //       CHECK:   %[[x:.*]] = llvm.extractelement %[[A]][%[[c]] : !llvm.i32] : !llvm<"<16 x float>">
 //       CHECK:   llvm.return %[[x]] : !llvm.float
 
 func @extract_element_from_vec_1d(%arg0: vector<16xf32>) -> f32 {
   %0 = vector.extract %arg0[15]: vector<16xf32>
   return %0 : f32
 }

 //       CHECK:   llvm.return {{.*}} : !llvm.float
 
 func @insert_element(%arg0: f32, %arg1: vector<4xf32>) -> vector<4xf32> {
   %0 = constant 3 : i32
   %1 = vector.insertelement %arg0, %arg1[%0 : i32] : vector<4xf32>
   return %1 : vector<4xf32>
 }
 // CHECK-LABEL: llvm.func @insert_element
-// CHECK-SAME: %[[A:arg[0-9]+]]: !llvm.float
-// CHECK-SAME: %[[B:arg[0-9]+]]: !llvm<"<4 x float>">
+// CHECK-SAME: %[[A:.*]]: !llvm.float,
+// CHECK-SAME: %[[B:.*]]: !llvm<"<4 x float>">
 //       CHECK:   %[[c:.*]] = llvm.mlir.constant(3 : i32) : !llvm.i32
 //       CHECK:   %[[x:.*]] = llvm.insertelement %[[A]], %[[B]][%[[c]] : !llvm.i32] : !llvm<"<4 x float>">
 //       CHECK:   llvm.return %[[x]] : !llvm<"<4 x float>">
 
 func @insert_element_into_vec_1d(%arg0: f32, %arg1: vector<4xf32>) -> vector<4xf32> {
   %0 = vector.insert %arg0, %arg1[3] : f32 into vector<4xf32>
   return %0 : vector<4xf32>
 }

 //       CHECK:   %[[allocatedBit:.*]] = llvm.bitcast %[[allocated]] : !llvm<"float*"> to !llvm<"[8 x [8 x <8 x float>]]*">
 //       CHECK:   llvm.insertvalue %[[allocatedBit]], {{.*}}[0] : !llvm<"{ [8 x [8 x <8 x float>]]*, [8 x [8 x <8 x float>]]*, i64 }">
 //       CHECK:   %[[aligned:.*]] = llvm.extractvalue {{.*}}[1] : !llvm<"{ float*, float*, i64, [3 x i64], [3 x i64] }">
 //       CHECK:   %[[alignedBit:.*]] = llvm.bitcast %[[aligned]] : !llvm<"float*"> to !llvm<"[8 x [8 x <8 x float>]]*">
 //       CHECK:   llvm.insertvalue %[[alignedBit]], {{.*}}[1] : !llvm<"{ [8 x [8 x <8 x float>]]*, [8 x [8 x <8 x float>]]*, i64 }">
 //       CHECK:   llvm.mlir.constant(0 : index
 //       CHECK:   llvm.insertvalue {{.*}}[2] : !llvm<"{ [8 x [8 x <8 x float>]]*, [8 x [8 x <8 x float>]]*, i64 }">
 
-func @vector_print_scalar(%arg0: f32) {
+func @vector_print_scalar_i32(%arg0: i32) {
+  vector.print %arg0 : i32
+  return
+}
+// CHECK-LABEL: llvm.func @vector_print_scalar_i32
+// CHECK-SAME: %[[A:.*]]: !llvm.i32
+//       CHECK:    llvm.call @print_i32(%[[A]]) : (!llvm.i32) -> ()
+//       CHECK:    llvm.call @print_newline() : () -> ()
+
+func @vector_print_scalar_i64(%arg0: i64) {
+  vector.print %arg0 : i64
+  return
+}
+// CHECK-LABEL: llvm.func @vector_print_scalar_i64
+// CHECK-SAME: %[[A:.*]]: !llvm.i64
+//       CHECK:    llvm.call @print_i64(%[[A]]) : (!llvm.i64) -> ()
+//       CHECK:    llvm.call @print_newline() : () -> ()
+
+func @vector_print_scalar_f32(%arg0: f32) {
   vector.print %arg0 : f32
   return
 }
-// CHECK-LABEL: llvm.func @vector_print_scalar
-// CHECK-SAME: %[[A:arg[0-9]+]]: !llvm.float
+// CHECK-LABEL: llvm.func @vector_print_scalar_f32
+// CHECK-SAME: %[[A:.*]]: !llvm.float
 //       CHECK:    llvm.call @print_f32(%[[A]]) : (!llvm.float) -> ()
 //       CHECK:    llvm.call @print_newline() : () -> ()
 
+func @vector_print_scalar_f64(%arg0: f64) {
+  vector.print %arg0 : f64
+  return
+}
+// CHECK-LABEL: llvm.func @vector_print_scalar_f64
+// CHECK-SAME: %[[A:.*]]: !llvm.double
+//       CHECK:    llvm.call @print_f64(%[[A]]) : (!llvm.double) -> ()
+//       CHECK:    llvm.call @print_newline() : () -> ()
+
 func @vector_print_vector(%arg0: vector<2x2xf32>) {
   vector.print %arg0 : vector<2x2xf32>
   return
 }
 // CHECK-LABEL: llvm.func @vector_print_vector
-// CHECK-SAME: %[[A:arg[0-9]+]]: !llvm<"[2 x <2 x float>]">
+// CHECK-SAME: %[[A:.*]]: !llvm<"[2 x <2 x float>]">
 //       CHECK:    llvm.call @print_open() : () -> ()
 //       CHECK:    %[[x0:.*]] = llvm.extractvalue %[[A]][0] : !llvm<"[2 x <2 x float>]">
 //       CHECK:    llvm.call @print_open() : () -> ()
 //       CHECK:    %[[x1:.*]] = llvm.mlir.constant(0 : index) : !llvm.i64
 //       CHECK:    %[[x2:.*]] = llvm.extractelement %[[x0]][%[[x1]] : !llvm.i64] : !llvm<"<2 x float>">
 //       CHECK:    llvm.call @print_f32(%[[x2]]) : (!llvm.float) -> ()
 //       CHECK:    llvm.call @print_comma() : () -> ()
 //       CHECK:    %[[x3:.*]] = llvm.mlir.constant(1 : index) : !llvm.i64

 //  CHECK-NEXT:    llvm.insertvalue {{.*}}, {{.*}}[3] : !llvm<"[4 x <4 x float>]">
 
 func @insert_strided_slice3(%arg0: vector<2x4xf32>, %arg1: vector<16x4x8xf32>) -> vector<16x4x8xf32> {
   %0 = vector.insert_strided_slice %arg0, %arg1 {offsets = [0, 0, 2], strides = [1, 1]}:
         vector<2x4xf32> into vector<16x4x8xf32>
   return %0 : vector<16x4x8xf32>
 }
 // CHECK-LABEL: llvm.func @insert_strided_slice3
-// CHECK-SAME: %[[A:arg[0-9]+]]: !llvm<"[2 x <4 x float>]">
-// CHECK-SAME: %[[B:arg[0-9]+]]: !llvm<"[16 x [4 x <8 x float>]]">
+// CHECK-SAME: %[[A:.*]]: !llvm<"[2 x <4 x float>]">,
+// CHECK-SAME: %[[B:.*]]: !llvm<"[16 x [4 x <8 x float>]]">
 //      CHECK: %[[s0:.*]] = llvm.extractvalue %[[B]][0] : !llvm<"[16 x [4 x <8 x float>]]">
 //      CHECK: %[[s1:.*]] = llvm.extractvalue %[[A]][0] : !llvm<"[2 x <4 x float>]">
 //      CHECK: %[[s2:.*]] = llvm.extractvalue %[[s0]][0] : !llvm<"[4 x <8 x float>]">
 //      CHECK: %[[s3:.*]] = llvm.mlir.constant(0 : index) : !llvm.i64
 //      CHECK: %[[s4:.*]] = llvm.extractelement %[[s1]][%[[s3]] : !llvm.i64] : !llvm<"<4 x float>">
 //      CHECK: %[[s5:.*]] = llvm.mlir.constant(2 : index) : !llvm.i64
 //      CHECK: %[[s6:.*]] = llvm.insertelement %[[s4]], %[[s2]][%[[s5]] : !llvm.i64] : !llvm<"<8 x float>">
 //      CHECK: %[[s7:.*]] = llvm.mlir.constant(1 : index) : !llvm.i64

 
 func @extract_strides(%arg0: vector<3x3xf32>) -> vector<1x1xf32> {
   %0 = vector.extract_slices %arg0, [2, 2], [1, 1]
     : vector<3x3xf32> into tuple<vector<2x2xf32>, vector<2x1xf32>, vector<1x2xf32>, vector<1x1xf32>>
   %1 = vector.tuple_get %0, 3 : tuple<vector<2x2xf32>, vector<2x1xf32>, vector<1x2xf32>, vector<1x1xf32>>
   return %1 : vector<1x1xf32>
 }
 // CHECK-LABEL: llvm.func @extract_strides
-// CHECK-SAME: %[[A:arg[0-9]+]]: !llvm<"[3 x <3 x float>]">
+// CHECK-SAME: %[[A:.*]]: !llvm<"[3 x <3 x float>]">
 //      CHECK: %[[s0:.*]] = llvm.mlir.constant(dense<0.000000e+00> : vector<1x1xf32>) : !llvm<"[1 x <1 x float>]">
 //      CHECK: %[[s1:.*]] = llvm.extractvalue %[[A]][2] : !llvm<"[3 x <3 x float>]">
 //      CHECK: %[[s3:.*]] = llvm.mlir.constant(dense<0.000000e+00> : vector<1xf32>) : !llvm<"<1 x float>">
 //      CHECK: %[[s4:.*]] = llvm.mlir.constant(2 : index) : !llvm.i64
 //      CHECK: %[[s5:.*]] = llvm.extractelement %[[s1]][%[[s4]] : !llvm.i64] : !llvm<"<3 x float>">
 //      CHECK: %[[s6:.*]] = llvm.mlir.constant(0 : index) : !llvm.i64
 //      CHECK: %[[s7:.*]] = llvm.insertelement %[[s5]], %[[s3]][%[[s6]] : !llvm.i64] : !llvm<"<1 x float>">
 //      CHECK: %[[s8:.*]] = llvm.insertvalue %[[s7]], %[[s0]][0] : !llvm<"[1 x <1 x float>]">
 //      CHECK: llvm.return %[[s8]] : !llvm<"[1 x <1 x float>]">

mlir/test/mlir-cpu-runner/mlir_runner_utils.cpp

 //===- mlir_runner_utils.cpp - Utils for MLIR CPU execution ---------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
 // Utilities for interfacing MLIR types with C code as well as printing,
 // debugging etc.
 //
 //===----------------------------------------------------------------------===//
 
 #include "include/mlir_runner_utils.h"
 
+#include <cinttypes>
 #include <cstdio>
 
 extern "C" void
 print_memref_vector_4x4xf32(StridedMemRefType<Vector2D<4, 4, float>, 2> *M) {
   impl::printMemRef(*M);
 }
 
 #define MEMREF_CASE(TYPE, RANK)                                                \

 }
 extern "C" void print_memref_4d_f32(StridedMemRefType<float, 4> *M) {
   impl::printMemRef(*M);
 }
 
 // Small runtime support "lib" for vector.print lowering.
 // By providing elementary printing methods only, this
 // library can remain fully unaware of low-level implementation
-// details of our vectors.
+// details of our vectors. Also useful for direct LLVM IR output.
+extern "C" void print_i32(int32_t i) { fprintf(stdout, "%" PRId32, i); }
+extern "C" void print_i64(int64_t l) { fprintf(stdout, "%" PRId64, l); }
 extern "C" void print_f32(float f) { fprintf(stdout, "%g", f); }
 extern "C" void print_f64(double d) { fprintf(stdout, "%lg", d); }
 extern "C" void print_open() { fputs("( ", stdout); }
 extern "C" void print_close() { fputs(" )", stdout); }
 extern "C" void print_comma() { fputs(", ", stdout); }
 extern "C" void print_newline() { fputc('\n', stdout); }