[MLIR][LLVMIR] Add {s,u}cmp intrinsics to LLVM dialect (#167870)

Author: Jezurko

mlir/include/mlir/Dialect/LLVMIR/LLVMIntrinsicOps.td

@@ -181,6 +181,18 @@ def LLVM_SMinOp : LLVM_BinarySameArgsIntrOpI<"smin">;
 def LLVM_UMaxOp : LLVM_BinarySameArgsIntrOpI<"umax">;
 def LLVM_UMinOp : LLVM_BinarySameArgsIntrOpI<"umin">;
 
+class LLVM_CmpIntrOp<string func>
+    : LLVM_OneResultIntrOp<func, [0], [0], [Pure, SameTypeOperands]> {
+  let arguments = (ins LLVM_ScalarOrVectorOf<AnySignlessInteger>:$a,
+      LLVM_ScalarOrVectorOf<AnySignlessInteger>:$b);
+  let results = (outs LLVM_ScalarOrVectorOf<AnySignlessInteger>:$res);
+  let assemblyFormat = "`(` operands `)` attr-dict `:` "
+                       "functional-type(operands, results)";
+}
+
+def LLVM_SCmpOp : LLVM_CmpIntrOp<"scmp">;
+def LLVM_UCmpOp : LLVM_CmpIntrOp<"ucmp">;
+
 def LLVM_SinOp : LLVM_UnaryIntrOpF<"sin">;
 def LLVM_CosOp : LLVM_UnaryIntrOpF<"cos">;
 def LLVM_TanOp : LLVM_UnaryIntrOpF<"tan">;

mlir/test/Target/LLVMIR/Import/intrinsic.ll

@@ -1128,6 +1128,34 @@ define void @experimental_constrained_fpext(float %s, <4 x float> %v) {
   ret void
 }
 
+; CHECK-LABEL:  llvm.func @ucmp
+define i2 @ucmp(i32 %a, i32 %b) {
+  ; CHECK: %{{.*}} = llvm.intr.ucmp(%{{.*}}, %{{.*}}) : (i32, i32) -> i2
+  %r = call i2 @llvm.ucmp.i2.i32(i32 %a, i32 %b)
+  ret i2 %r
+}
+
+; CHECK-LABEL:  llvm.func @vector_ucmp
+define <4 x i32> @vector_ucmp(<4 x i32> %a, <4 x i32> %b) {
+  ; CHECK: %{{.*}} = llvm.intr.ucmp(%{{.*}}, %{{.*}}) : (vector<4xi32>, vector<4xi32>) -> vector<4xi32>
+  %r = call <4 x i32> @llvm.ucmp.v4i32.v4i32(<4 x i32> %a, <4 x i32> %b)
+  ret <4 x i32> %r
+}
+
+; CHECK-LABEL:  llvm.func @scmp
+define i2 @scmp(i32 %a, i32 %b) {
+  ; CHECK: %{{.*}} = llvm.intr.scmp(%{{.*}}, %{{.*}}) : (i32, i32) -> i2
+  %r = call i2 @llvm.scmp.i2.i32(i32 %a, i32 %b)
+  ret i2 %r
+}
+
+; CHECK-LABEL:  llvm.func @vector_scmp
+define <4 x i32> @vector_scmp(<4 x i32> %a, <4 x i32> %b) {
+  ; CHECK: %{{.*}} = llvm.intr.scmp(%{{.*}}, %{{.*}}) : (vector<4xi32>, vector<4xi32>) -> vector<4xi32>
+  %r = call <4 x i32> @llvm.scmp.v4i32.v4i32(<4 x i32> %a, <4 x i32> %b)
+  ret <4 x i32> %r
+}
+
 declare float @llvm.fmuladd.f32(float, float, float)
 declare <8 x float> @llvm.fmuladd.v8f32(<8 x float>, <8 x float>, <8 x float>)
 declare float @llvm.fma.f32(float, float, float)
@@ -1382,3 +1410,7 @@ declare <4 x half> @llvm.experimental.constrained.fptrunc.v4f16.v4f64(<4 x doubl
 declare float @llvm.experimental.constrained.fptrunc.f32.f64(double, metadata, metadata)
 declare <4 x double> @llvm.experimental.constrained.fpext.v4f64.v4f32(<4 x float>, metadata)
 declare double @llvm.experimental.constrained.fpext.f64.f32(float, metadata)
+declare i2 @llvm.ucmp.i2.i32(i32, i32)
+declare <4 x i32> @llvm.ucmp.v4i32.v4i32(<4 x i32>, <4 x i32>)
+declare i2 @llvm.scmp.i2.i32(i32, i32)
+declare <4 x i32> @llvm.scmp.v4i32.v4i32(<4 x i32>, <4 x i32>)

mlir/test/Target/LLVMIR/llvmir-intrinsics.mlir

@@ -1276,6 +1276,34 @@ llvm.func @experimental_constrained_fpext(%s: f32, %v: vector<4xf32>) {
   llvm.return
 }
 
+// CHECK-LABEL: @ucmp
+llvm.func @ucmp(%a: i32, %b: i32) -> i2 {
+  // CHECK: call i2 @llvm.ucmp.i2.i32
+  %r = llvm.intr.ucmp(%a, %b) : (i32, i32) -> i2
+  llvm.return %r : i2
+}
+
+// CHECK-LABEL: @vector_ucmp
+llvm.func @vector_ucmp(%a: vector<4 x i32>, %b: vector<4 x i32>) -> vector<4 x i32> {
+  // CHECK: call <4 x i32> @llvm.ucmp.v4i32.v4i32
+  %0 = llvm.intr.ucmp(%a, %b) : (vector<4 x i32>, vector<4 x i32>) -> vector<4 x i32>
+  llvm.return %0 : vector<4 x i32>
+}
+
+// CHECK-LABEL: @scmp
+llvm.func @scmp(%a: i32, %b: i32) -> i2 {
+  // CHECK: call i2 @llvm.scmp.i2.i32
+  %r = llvm.intr.scmp(%a, %b) : (i32, i32) -> i2
+  llvm.return %r : i2
+}
+
+// CHECK-LABEL: @vector_scmp
+llvm.func @vector_scmp(%a: vector<4 x i32>, %b: vector<4 x i32>) -> vector<4 x i32> {
+  // CHECK: call <4 x i32> @llvm.scmp.v4i32.v4i32
+  %0 = llvm.intr.scmp(%a, %b) : (vector<4 x i32>, vector<4 x i32>) -> vector<4 x i32>
+  llvm.return %0 : vector<4 x i32>
+}
+
 // Check that intrinsics are declared with appropriate types.
 // CHECK-DAG: declare float @llvm.fma.f32(float, float, float)
 // CHECK-DAG: declare <8 x float> @llvm.fma.v8f32(<8 x float>, <8 x float>, <8 x float>) #0
@@ -1464,3 +1492,7 @@ llvm.func @experimental_constrained_fpext(%s: f32, %v: vector<4xf32>) {
 // CHECK-DAG: declare <4 x half> @llvm.experimental.constrained.fptrunc.v4f16.v4f32(<4 x float>, metadata, metadata)
 // CHECK-DAG: declare double @llvm.experimental.constrained.fpext.f64.f32(float, metadata)
 // CHECK-DAG: declare <4 x double> @llvm.experimental.constrained.fpext.v4f64.v4f32(<4 x float>, metadata)
+// CHECK-DAG: declare range(i2 -1, -2) i2 @llvm.ucmp.i2.i32(i32, i32)
+// CHECK-DAG: declare range(i32 -1, 2) <4 x i32> @llvm.ucmp.v4i32.v4i32(<4 x i32>, <4 x i32>)
+// CHECK-DAG: declare range(i2 -1, -2) i2 @llvm.scmp.i2.i32(i32, i32)
+// CHECK-DAG: declare range(i32 -1, 2) <4 x i32> @llvm.scmp.v4i32.v4i32(<4 x i32>, <4 x i32>)