Add Expm1 op to the math dialect.

[akuegel]

Summary

Add Expm1 to the math dialect.
On GPU, there is a dedicated intrinsic for this,
and we want to be able to lower to it.

mlir/include/mlir/Dialect/Math/IR/MathOps.td

let description = [{
 
     // Tensor element-wise natural exponential.
     %x = math.exp2 %y : tensor<4x?xf8>
     ```
   }];
 }
 
 //===----------------------------------------------------------------------===//
+// ExpM1Op
+//===----------------------------------------------------------------------===//
+
+def ExpM1Op : FloatUnaryOp<"expm1"> {
+  let summary = "base-e exponential of the specified value minus 1";
+  let description = [{
+    Syntax:
+
+    ```
+    operation ::= ssa-id `=` `math.expm1` ssa-use `:` type
+    ```
+
+    expm1(x) := exp(x) - 1
+
+    The `expm1` operation takes one operand and returns one result of the same
+    type. This type may be a float scalar type, a vector whose element type is
+    float, or a tensor of floats. It has no standard attributes.
+
+    Example:
+
+    ```mlir
+    // Scalar natural exponential minus 1.
+    %a = math.expm1 %b : f64
+
+    // SIMD vector element-wise natural exponential minus 1.
+    %f = math.expm1 %g : vector<4xf32>
+
+    // Tensor element-wise natural exponential minus 1.
+    %x = math.expm1 %y : tensor<4x?xf8>
+    ```
+  }];
+}
+
+//===----------------------------------------------------------------------===//
 // LogOp
 //===----------------------------------------------------------------------===//
 
 def LogOp : FloatUnaryOp<"log"> {
   let summary = "base-e logarithm of the specified value";
 
   let description = [{
     Computes the base-e logarithm of the given value. It takes one operand and

mlir/test/Dialect/Math/ops.mlir

func @exp2(%f: f32, %v: vector<4xf32>, %t: tensor<4x4x?xf32>) {
   %0 = math.exp2 %f : f32
   // CHECK: %{{.*}} = math.exp2 %[[V]] : vector<4xf32>
   %1 = math.exp2 %v : vector<4xf32>
   // CHECK: %{{.*}} = math.exp2 %[[T]] : tensor<4x4x?xf32>
   %2 = math.exp2 %t : tensor<4x4x?xf32>
   return
 }
 
+// CHECK-LABEL: func @expm1(
+// CHECK-SAME:            %[[F:.*]]: f32, %[[V:.*]]: vector<4xf32>, %[[T:.*]]: tensor<4x4x?xf32>)
+func @expm1(%f: f32, %v: vector<4xf32>, %t: tensor<4x4x?xf32>) {
+  // CHECK: %{{.*}} = math.expm1 %[[F]] : f32
+  %0 = math.expm1 %f : f32
+  // CHECK: %{{.*}} = math.expm1 %[[V]] : vector<4xf32>
+  %1 = math.expm1 %v : vector<4xf32>
+  // CHECK: %{{.*}} = math.expm1 %[[T]] : tensor<4x4x?xf32>
+  %2 = math.expm1 %t : tensor<4x4x?xf32>
+  return
+}
+
 // CHECK-LABEL: func @log(
 // CHECK-SAME:            %[[F:.*]]: f32, %[[V:.*]]: vector<4xf32>, %[[T:.*]]: tensor<4x4x?xf32>)
 func @log(%f: f32, %v: vector<4xf32>, %t: tensor<4x4x?xf32>) {
   // CHECK: %{{.*}} = math.log %[[F]] : f32
   %0 = math.log %f : f32
   // CHECK: %{{.*}} = math.log %[[V]] : vector<4xf32>
   %1 = math.log %v : vector<4xf32>
   // CHECK: %{{.*}} = math.log %[[T]] : tensor<4x4x?xf32>