Add x86 SIMD immediate shifts

cranelift/filetests/filetests/isa/x86/simd-bitwise-binemit.clif

@@ -49,3 +49,57 @@ ebb0(v0: i32x4 [%xmm4], v1: i64x2 [%xmm0]):
 [-, %xmm4]  v2 = x86_psra v0, v1      ; bin: 66 0f e2 e0
             return v2
 }
+
+function %ishl_imm_i16x8(i16x8) -> i16x8 {
+ebb0(v0: i16x8 [%xmm2]):
+[-, %xmm2]  v2 = ishl_imm v0, 3      ; bin: 66 0f 71 f2 03
+            return v2
+}
+
+function %ishl_imm_i32x4(i32x4) -> i32x4 {
+ebb0(v0: i32x4 [%xmm4]):
+[-, %xmm4]  v2 = ishl_imm v0, 10     ; bin: 66 0f 72 f4 0a
+            return v2
+}
+
+function %ishl_imm_i64x2(i64x2) -> i64x2 {
+ebb0(v0: i64x2 [%xmm6]):
+[-, %xmm6]  v2 = ishl_imm v0, 42     ; bin: 66 0f 73 f6 2a
+            return v2
+}
+
+function %ushr_imm_i16x8(i16x8) -> i16x8 {
+ebb0(v0: i16x8 [%xmm2]):
+[-, %xmm2]  v2 = ushr_imm v0, 3      ; bin: 66 0f 71 d2 03
+            return v2
+}
+
+function %ushr_imm_i32x4(i32x4) -> i32x4 {
+ebb0(v0: i32x4 [%xmm4]):
+[-, %xmm4]  v2 = ushr_imm v0, 10     ; bin: 66 0f 72 d4 0a
+            return v2
+}
+
+function %ushr_imm_i64x2(i64x2) -> i64x2 {
+ebb0(v0: i64x2 [%xmm6]):
+[-, %xmm6]  v2 = ushr_imm v0, 42     ; bin: 66 0f 73 d6 2a
+            return v2
+}
+
+function %sshr_imm_i16x8(i16x8) -> i16x8 {
+ebb0(v0: i16x8 [%xmm2]):
+[-, %xmm2]  v2 = sshr_imm v0, 3      ; bin: 66 0f 71 e2 03
+            return v2
+}
+
+function %sshr_imm_i32x4(i32x4) -> i32x4 {
+ebb0(v0: i32x4 [%xmm4]):
+[-, %xmm4]  v2 = sshr_imm v0, 10     ; bin: 66 0f 72 e4 0a
+            return v2
+}
+
+function %sshr_imm_i64x2(i64x2) -> i64x2 {
+ebb0(v0: i64x2 [%xmm6]):
+[-, %xmm6]  v2 = sshr_imm v0, 42     ; bin: 66 0f 73 e6 2a
+            return v2
+}

cranelift/filetests/filetests/isa/x86/simd-bitwise-run.clif

@@ -127,3 +127,39 @@ ebb0:
     return v11
 }
 ; run
+
+function %sshr_imm_i32x4() -> b1 {
+ebb0:
+    v1 = vconst.i32x4 [1 2 4 -8]
+    v2 = sshr_imm v1, 1
+
+    v3 = vconst.i32x4 [0 1 2 -4]
+    v4 = icmp eq v2, v3
+    v5 = vall_true v4
+    return v5
+}
+; run
+
+function %sshr_imm_i16x8() -> b1 {
+ebb0:
+    v1 = vconst.i16x8 [1 2 4 -8 0 0 0 0]
+    v2 = ushr_imm v1, 1
+
+    v3 = vconst.i16x8 [0 1 2 32764 0 0 0 0] ; -4 with MSB unset == 32764
+    v4 = icmp eq v2, v3
+    v5 = vall_true v4
+    return v5
+}
+; run
+
+function %ishl_imm_i64x2() -> b1 {
+ebb0:
+    v1 = vconst.i64x2 [1 0]
+    v2 = ishl_imm v1, 1
+
+    v3 = vconst.i64x2 [2 0]
+    v4 = icmp eq v2, v3
+    v5 = vall_true v4
+    return v5
+}
+; run