[NFC][Codegen] Add/sub constant-folding: add scalar tests too

Just for completeness.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@362208 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/test/CodeGen/AArch64/addsub-constant-folding.ll b/test/CodeGen/AArch64/addsub-constant-folding.ll
index 47a236e373be2c33dfc053d4c83452f44698cdfc..bb3b2ebe8ecd1578a0b7924ab756e37da5c4f2f3 100644 (file)
--- a/test/CodeGen/AArch64/addsub-constant-folding.ll
+++ b/test/CodeGen/AArch64/addsub-constant-folding.ll
@@ -1,13 +1,43 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=aarch64-unknown-unknown | FileCheck %s
 
-declare void @use(<4 x i32> %arg)
+declare void @use(i32 %arg)
+declare void @vec_use(<4 x i32> %arg)
 
 ; (x+c1)+c2
 
-define <4 x i32> @add_const_add_const(<4 x i32> %arg) {
+define i32 @add_const_add_const(i32 %arg) {
 ; CHECK-LABEL: add_const_add_const:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    add w0, w0, #10 // =10
+; CHECK-NEXT:    ret
+  %t0 = add i32 %arg, 8
+  %t1 = add i32 %t0, 2
+  ret i32 %t1
+}
+
+define i32 @add_const_add_const_extrause(i32 %arg) {
+; CHECK-LABEL: add_const_add_const_extrause:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    stp x19, x30, [sp, #-16]! // 16-byte Folded Spill
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset w30, -8
+; CHECK-NEXT:    .cfi_offset w19, -16
+; CHECK-NEXT:    mov w19, w0
+; CHECK-NEXT:    add w0, w0, #8 // =8
+; CHECK-NEXT:    bl use
+; CHECK-NEXT:    add w0, w19, #10 // =10
+; CHECK-NEXT:    ldp x19, x30, [sp], #16 // 16-byte Folded Reload
+; CHECK-NEXT:    ret
+  %t0 = add i32 %arg, 8
+  call void @use(i32 %t0)
+  %t1 = add i32 %t0, 2
+  ret i32 %t1
+}
+
+define <4 x i32> @vec_add_const_add_const(<4 x i32> %arg) {
+; CHECK-LABEL: vec_add_const_add_const:
+; CHECK:       // %bb.0:
 ; CHECK-NEXT:    movi v1.4s, #10
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
@@ -16,8 +46,8 @@ define <4 x i32> @add_const_add_const(<4 x i32> %arg) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @add_const_add_const_extrause(<4 x i32> %arg) {
-; CHECK-LABEL: add_const_add_const_extrause:
+define <4 x i32> @vec_add_const_add_const_extrause(<4 x i32> %arg) {
+; CHECK-LABEL: vec_add_const_add_const_extrause:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #32 // =32
 ; CHECK-NEXT:    str x30, [sp, #16] // 8-byte Folded Spill
@@ -26,7 +56,7 @@ define <4 x i32> @add_const_add_const_extrause(<4 x i32> %arg) {
 ; CHECK-NEXT:    movi v1.4s, #8
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
-; CHECK-NEXT:    bl use
+; CHECK-NEXT:    bl vec_use
 ; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr x30, [sp, #16] // 8-byte Folded Reload
 ; CHECK-NEXT:    movi v0.4s, #10
@@ -34,16 +64,16 @@ define <4 x i32> @add_const_add_const_extrause(<4 x i32> %arg) {
 ; CHECK-NEXT:    add sp, sp, #32 // =32
 ; CHECK-NEXT:    ret
   %t0 = add <4 x i32> %arg, <i32 8, i32 8, i32 8, i32 8>
-  call void @use(<4 x i32> %t0)
+  call void @vec_use(<4 x i32> %t0)
   %t1 = add <4 x i32> %t0, <i32 2, i32 2, i32 2, i32 2>
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @add_const_add_const_nonsplat(<4 x i32> %arg) {
-; CHECK-LABEL: add_const_add_const_nonsplat:
+define <4 x i32> @vec_add_const_add_const_nonsplat(<4 x i32> %arg) {
+; CHECK-LABEL: vec_add_const_add_const_nonsplat:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    adrp x8, .LCPI2_0
-; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI2_0]
+; CHECK-NEXT:    adrp x8, .LCPI4_0
+; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI4_0]
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
   %t0 = add <4 x i32> %arg, <i32 21, i32 undef, i32 8, i32 8>
@@ -53,9 +83,38 @@ define <4 x i32> @add_const_add_const_nonsplat(<4 x i32> %arg) {
 
 ; (x+c1)-c2
 
-define <4 x i32> @add_const_sub_const(<4 x i32> %arg) {
+define i32 @add_const_sub_const(i32 %arg) {
 ; CHECK-LABEL: add_const_sub_const:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    add w0, w0, #6 // =6
+; CHECK-NEXT:    ret
+  %t0 = add i32 %arg, 8
+  %t1 = sub i32 %t0, 2
+  ret i32 %t1
+}
+
+define i32 @add_const_sub_const_extrause(i32 %arg) {
+; CHECK-LABEL: add_const_sub_const_extrause:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    stp x19, x30, [sp, #-16]! // 16-byte Folded Spill
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset w30, -8
+; CHECK-NEXT:    .cfi_offset w19, -16
+; CHECK-NEXT:    mov w19, w0
+; CHECK-NEXT:    add w0, w0, #8 // =8
+; CHECK-NEXT:    bl use
+; CHECK-NEXT:    add w0, w19, #6 // =6
+; CHECK-NEXT:    ldp x19, x30, [sp], #16 // 16-byte Folded Reload
+; CHECK-NEXT:    ret
+  %t0 = add i32 %arg, 8
+  call void @use(i32 %t0)
+  %t1 = sub i32 %t0, 2
+  ret i32 %t1
+}
+
+define <4 x i32> @vec_add_const_sub_const(<4 x i32> %arg) {
+; CHECK-LABEL: vec_add_const_sub_const:
+; CHECK:       // %bb.0:
 ; CHECK-NEXT:    movi v1.4s, #6
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
@@ -64,8 +123,8 @@ define <4 x i32> @add_const_sub_const(<4 x i32> %arg) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @add_const_sub_const_extrause(<4 x i32> %arg) {
-; CHECK-LABEL: add_const_sub_const_extrause:
+define <4 x i32> @vec_add_const_sub_const_extrause(<4 x i32> %arg) {
+; CHECK-LABEL: vec_add_const_sub_const_extrause:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #32 // =32
 ; CHECK-NEXT:    str x30, [sp, #16] // 8-byte Folded Spill
@@ -74,7 +133,7 @@ define <4 x i32> @add_const_sub_const_extrause(<4 x i32> %arg) {
 ; CHECK-NEXT:    movi v1.4s, #8
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
-; CHECK-NEXT:    bl use
+; CHECK-NEXT:    bl vec_use
 ; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr x30, [sp, #16] // 8-byte Folded Reload
 ; CHECK-NEXT:    movi v0.4s, #6
@@ -82,16 +141,16 @@ define <4 x i32> @add_const_sub_const_extrause(<4 x i32> %arg) {
 ; CHECK-NEXT:    add sp, sp, #32 // =32
 ; CHECK-NEXT:    ret
   %t0 = add <4 x i32> %arg, <i32 8, i32 8, i32 8, i32 8>
-  call void @use(<4 x i32> %t0)
+  call void @vec_use(<4 x i32> %t0)
   %t1 = sub <4 x i32> %t0, <i32 2, i32 2, i32 2, i32 2>
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @add_const_sub_const_nonsplat(<4 x i32> %arg) {
-; CHECK-LABEL: add_const_sub_const_nonsplat:
+define <4 x i32> @vec_add_const_sub_const_nonsplat(<4 x i32> %arg) {
+; CHECK-LABEL: vec_add_const_sub_const_nonsplat:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    adrp x8, .LCPI5_0
-; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI5_0]
+; CHECK-NEXT:    adrp x8, .LCPI9_0
+; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI9_0]
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
   %t0 = add <4 x i32> %arg, <i32 21, i32 undef, i32 8, i32 8>
@@ -101,9 +160,40 @@ define <4 x i32> @add_const_sub_const_nonsplat(<4 x i32> %arg) {
 
 ; c2-(x+c1)
 
-define <4 x i32> @add_const_const_sub(<4 x i32> %arg) {
+define i32 @add_const_const_sub(i32 %arg) {
 ; CHECK-LABEL: add_const_const_sub:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #-6
+; CHECK-NEXT:    sub w0, w8, w0
+; CHECK-NEXT:    ret
+  %t0 = add i32 %arg, 8
+  %t1 = sub i32 2, %t0
+  ret i32 %t1
+}
+
+define i32 @add_const_const_sub_extrause(i32 %arg) {
+; CHECK-LABEL: add_const_const_sub_extrause:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    stp x19, x30, [sp, #-16]! // 16-byte Folded Spill
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset w30, -8
+; CHECK-NEXT:    .cfi_offset w19, -16
+; CHECK-NEXT:    mov w19, w0
+; CHECK-NEXT:    add w0, w0, #8 // =8
+; CHECK-NEXT:    bl use
+; CHECK-NEXT:    mov w8, #-6
+; CHECK-NEXT:    sub w0, w8, w19
+; CHECK-NEXT:    ldp x19, x30, [sp], #16 // 16-byte Folded Reload
+; CHECK-NEXT:    ret
+  %t0 = add i32 %arg, 8
+  call void @use(i32 %t0)
+  %t1 = sub i32 2, %t0
+  ret i32 %t1
+}
+
+define <4 x i32> @vec_add_const_const_sub(<4 x i32> %arg) {
+; CHECK-LABEL: vec_add_const_const_sub:
+; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mvni v1.4s, #5
 ; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    ret
@@ -112,8 +202,8 @@ define <4 x i32> @add_const_const_sub(<4 x i32> %arg) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @add_const_const_sub_extrause(<4 x i32> %arg) {
-; CHECK-LABEL: add_const_const_sub_extrause:
+define <4 x i32> @vec_add_const_const_sub_extrause(<4 x i32> %arg) {
+; CHECK-LABEL: vec_add_const_const_sub_extrause:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #32 // =32
 ; CHECK-NEXT:    str x30, [sp, #16] // 8-byte Folded Spill
@@ -122,7 +212,7 @@ define <4 x i32> @add_const_const_sub_extrause(<4 x i32> %arg) {
 ; CHECK-NEXT:    movi v1.4s, #8
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
-; CHECK-NEXT:    bl use
+; CHECK-NEXT:    bl vec_use
 ; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr x30, [sp, #16] // 8-byte Folded Reload
 ; CHECK-NEXT:    mvni v0.4s, #5
@@ -130,16 +220,16 @@ define <4 x i32> @add_const_const_sub_extrause(<4 x i32> %arg) {
 ; CHECK-NEXT:    add sp, sp, #32 // =32
 ; CHECK-NEXT:    ret
   %t0 = add <4 x i32> %arg, <i32 8, i32 8, i32 8, i32 8>
-  call void @use(<4 x i32> %t0)
+  call void @vec_use(<4 x i32> %t0)
   %t1 = sub <4 x i32> <i32 2, i32 2, i32 2, i32 2>, %t0
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @add_const_const_sub_nonsplat(<4 x i32> %arg) {
-; CHECK-LABEL: add_const_const_sub_nonsplat:
+define <4 x i32> @vec_add_const_const_sub_nonsplat(<4 x i32> %arg) {
+; CHECK-LABEL: vec_add_const_const_sub_nonsplat:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    adrp x8, .LCPI8_0
-; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI8_0]
+; CHECK-NEXT:    adrp x8, .LCPI14_0
+; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI14_0]
 ; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    ret
   %t0 = add <4 x i32> %arg, <i32 21, i32 undef, i32 8, i32 8>
@@ -149,9 +239,38 @@ define <4 x i32> @add_const_const_sub_nonsplat(<4 x i32> %arg) {
 
 ; (x-c1)+c2
 
-define <4 x i32> @sub_const_add_const(<4 x i32> %arg) {
+define i32 @sub_const_add_const(i32 %arg) {
 ; CHECK-LABEL: sub_const_add_const:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    sub w0, w0, #6 // =6
+; CHECK-NEXT:    ret
+  %t0 = sub i32 %arg, 8
+  %t1 = add i32 %t0, 2
+  ret i32 %t1
+}
+
+define i32 @sub_const_add_const_extrause(i32 %arg) {
+; CHECK-LABEL: sub_const_add_const_extrause:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    stp x19, x30, [sp, #-16]! // 16-byte Folded Spill
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset w30, -8
+; CHECK-NEXT:    .cfi_offset w19, -16
+; CHECK-NEXT:    mov w19, w0
+; CHECK-NEXT:    sub w0, w0, #8 // =8
+; CHECK-NEXT:    bl use
+; CHECK-NEXT:    sub w0, w19, #6 // =6
+; CHECK-NEXT:    ldp x19, x30, [sp], #16 // 16-byte Folded Reload
+; CHECK-NEXT:    ret
+  %t0 = sub i32 %arg, 8
+  call void @use(i32 %t0)
+  %t1 = add i32 %t0, 2
+  ret i32 %t1
+}
+
+define <4 x i32> @vec_sub_const_add_const(<4 x i32> %arg) {
+; CHECK-LABEL: vec_sub_const_add_const:
+; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mvni v1.4s, #5
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
@@ -160,8 +279,8 @@ define <4 x i32> @sub_const_add_const(<4 x i32> %arg) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @sub_const_add_const_extrause(<4 x i32> %arg) {
-; CHECK-LABEL: sub_const_add_const_extrause:
+define <4 x i32> @vec_sub_const_add_const_extrause(<4 x i32> %arg) {
+; CHECK-LABEL: vec_sub_const_add_const_extrause:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #32 // =32
 ; CHECK-NEXT:    str x30, [sp, #16] // 8-byte Folded Spill
@@ -170,7 +289,7 @@ define <4 x i32> @sub_const_add_const_extrause(<4 x i32> %arg) {
 ; CHECK-NEXT:    movi v1.4s, #8
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
 ; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
-; CHECK-NEXT:    bl use
+; CHECK-NEXT:    bl vec_use
 ; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr x30, [sp, #16] // 8-byte Folded Reload
 ; CHECK-NEXT:    mvni v0.4s, #5
@@ -178,16 +297,16 @@ define <4 x i32> @sub_const_add_const_extrause(<4 x i32> %arg) {
 ; CHECK-NEXT:    add sp, sp, #32 // =32
 ; CHECK-NEXT:    ret
   %t0 = sub <4 x i32> %arg, <i32 8, i32 8, i32 8, i32 8>
-  call void @use(<4 x i32> %t0)
+  call void @vec_use(<4 x i32> %t0)
   %t1 = add <4 x i32> %t0, <i32 2, i32 2, i32 2, i32 2>
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @sub_const_add_const_nonsplat(<4 x i32> %arg) {
-; CHECK-LABEL: sub_const_add_const_nonsplat:
+define <4 x i32> @vec_sub_const_add_const_nonsplat(<4 x i32> %arg) {
+; CHECK-LABEL: vec_sub_const_add_const_nonsplat:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    adrp x8, .LCPI11_0
-; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI11_0]
+; CHECK-NEXT:    adrp x8, .LCPI19_0
+; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI19_0]
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
   %t0 = sub <4 x i32> %arg, <i32 21, i32 undef, i32 8, i32 8>
@@ -197,9 +316,38 @@ define <4 x i32> @sub_const_add_const_nonsplat(<4 x i32> %arg) {
 
 ; (x-c1)-c2
 
-define <4 x i32> @sub_const_sub_const(<4 x i32> %arg) {
+define i32 @sub_const_sub_const(i32 %arg) {
 ; CHECK-LABEL: sub_const_sub_const:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    sub w0, w0, #10 // =10
+; CHECK-NEXT:    ret
+  %t0 = sub i32 %arg, 8
+  %t1 = sub i32 %t0, 2
+  ret i32 %t1
+}
+
+define i32 @sub_const_sub_const_extrause(i32 %arg) {
+; CHECK-LABEL: sub_const_sub_const_extrause:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    stp x19, x30, [sp, #-16]! // 16-byte Folded Spill
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset w30, -8
+; CHECK-NEXT:    .cfi_offset w19, -16
+; CHECK-NEXT:    mov w19, w0
+; CHECK-NEXT:    sub w0, w0, #8 // =8
+; CHECK-NEXT:    bl use
+; CHECK-NEXT:    sub w0, w19, #10 // =10
+; CHECK-NEXT:    ldp x19, x30, [sp], #16 // 16-byte Folded Reload
+; CHECK-NEXT:    ret
+  %t0 = sub i32 %arg, 8
+  call void @use(i32 %t0)
+  %t1 = sub i32 %t0, 2
+  ret i32 %t1
+}
+
+define <4 x i32> @vec_sub_const_sub_const(<4 x i32> %arg) {
+; CHECK-LABEL: vec_sub_const_sub_const:
+; CHECK:       // %bb.0:
 ; CHECK-NEXT:    movi v1.4s, #10
 ; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
@@ -208,8 +356,8 @@ define <4 x i32> @sub_const_sub_const(<4 x i32> %arg) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @sub_const_sub_const_extrause(<4 x i32> %arg) {
-; CHECK-LABEL: sub_const_sub_const_extrause:
+define <4 x i32> @vec_sub_const_sub_const_extrause(<4 x i32> %arg) {
+; CHECK-LABEL: vec_sub_const_sub_const_extrause:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #32 // =32
 ; CHECK-NEXT:    str x30, [sp, #16] // 8-byte Folded Spill
@@ -218,7 +366,7 @@ define <4 x i32> @sub_const_sub_const_extrause(<4 x i32> %arg) {
 ; CHECK-NEXT:    movi v1.4s, #8
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
 ; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
-; CHECK-NEXT:    bl use
+; CHECK-NEXT:    bl vec_use
 ; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr x30, [sp, #16] // 8-byte Folded Reload
 ; CHECK-NEXT:    movi v0.4s, #10
@@ -226,16 +374,16 @@ define <4 x i32> @sub_const_sub_const_extrause(<4 x i32> %arg) {
 ; CHECK-NEXT:    add sp, sp, #32 // =32
 ; CHECK-NEXT:    ret
   %t0 = sub <4 x i32> %arg, <i32 8, i32 8, i32 8, i32 8>
-  call void @use(<4 x i32> %t0)
+  call void @vec_use(<4 x i32> %t0)
   %t1 = sub <4 x i32> %t0, <i32 2, i32 2, i32 2, i32 2>
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @sub_const_sub_const_nonsplat(<4 x i32> %arg) {
-; CHECK-LABEL: sub_const_sub_const_nonsplat:
+define <4 x i32> @vec_sub_const_sub_const_nonsplat(<4 x i32> %arg) {
+; CHECK-LABEL: vec_sub_const_sub_const_nonsplat:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    adrp x8, .LCPI14_0
-; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI14_0]
+; CHECK-NEXT:    adrp x8, .LCPI24_0
+; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI24_0]
 ; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
   %t0 = sub <4 x i32> %arg, <i32 21, i32 undef, i32 8, i32 8>
@@ -245,9 +393,40 @@ define <4 x i32> @sub_const_sub_const_nonsplat(<4 x i32> %arg) {
 
 ; c2-(x-c1)
 
-define <4 x i32> @sub_const_const_sub(<4 x i32> %arg) {
+define i32 @sub_const_const_sub(i32 %arg) {
 ; CHECK-LABEL: sub_const_const_sub:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #10
+; CHECK-NEXT:    sub w0, w8, w0
+; CHECK-NEXT:    ret
+  %t0 = sub i32 %arg, 8
+  %t1 = sub i32 2, %t0
+  ret i32 %t1
+}
+
+define i32 @sub_const_const_sub_extrause(i32 %arg) {
+; CHECK-LABEL: sub_const_const_sub_extrause:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    stp x19, x30, [sp, #-16]! // 16-byte Folded Spill
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset w30, -8
+; CHECK-NEXT:    .cfi_offset w19, -16
+; CHECK-NEXT:    mov w19, w0
+; CHECK-NEXT:    sub w0, w0, #8 // =8
+; CHECK-NEXT:    bl use
+; CHECK-NEXT:    mov w8, #10
+; CHECK-NEXT:    sub w0, w8, w19
+; CHECK-NEXT:    ldp x19, x30, [sp], #16 // 16-byte Folded Reload
+; CHECK-NEXT:    ret
+  %t0 = sub i32 %arg, 8
+  call void @use(i32 %t0)
+  %t1 = sub i32 2, %t0
+  ret i32 %t1
+}
+
+define <4 x i32> @vec_sub_const_const_sub(<4 x i32> %arg) {
+; CHECK-LABEL: vec_sub_const_const_sub:
+; CHECK:       // %bb.0:
 ; CHECK-NEXT:    movi v1.4s, #10
 ; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    ret
@@ -256,8 +435,8 @@ define <4 x i32> @sub_const_const_sub(<4 x i32> %arg) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @sub_const_const_sub_extrause(<4 x i32> %arg) {
-; CHECK-LABEL: sub_const_const_sub_extrause:
+define <4 x i32> @vec_sub_const_const_sub_extrause(<4 x i32> %arg) {
+; CHECK-LABEL: vec_sub_const_const_sub_extrause:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #32 // =32
 ; CHECK-NEXT:    str x30, [sp, #16] // 8-byte Folded Spill
@@ -266,7 +445,7 @@ define <4 x i32> @sub_const_const_sub_extrause(<4 x i32> %arg) {
 ; CHECK-NEXT:    movi v1.4s, #8
 ; CHECK-NEXT:    sub v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
-; CHECK-NEXT:    bl use
+; CHECK-NEXT:    bl vec_use
 ; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr x30, [sp, #16] // 8-byte Folded Reload
 ; CHECK-NEXT:    movi v0.4s, #2
@@ -274,16 +453,16 @@ define <4 x i32> @sub_const_const_sub_extrause(<4 x i32> %arg) {
 ; CHECK-NEXT:    add sp, sp, #32 // =32
 ; CHECK-NEXT:    ret
   %t0 = sub <4 x i32> %arg, <i32 8, i32 8, i32 8, i32 8>
-  call void @use(<4 x i32> %t0)
+  call void @vec_use(<4 x i32> %t0)
   %t1 = sub <4 x i32> <i32 2, i32 2, i32 2, i32 2>, %t0
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @sub_const_const_sub_nonsplat(<4 x i32> %arg) {
-; CHECK-LABEL: sub_const_const_sub_nonsplat:
+define <4 x i32> @vec_sub_const_const_sub_nonsplat(<4 x i32> %arg) {
+; CHECK-LABEL: vec_sub_const_const_sub_nonsplat:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    adrp x8, .LCPI17_0
-; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI17_0]
+; CHECK-NEXT:    adrp x8, .LCPI29_0
+; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI29_0]
 ; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    ret
   %t0 = sub <4 x i32> %arg, <i32 21, i32 undef, i32 8, i32 8>
@@ -293,9 +472,41 @@ define <4 x i32> @sub_const_const_sub_nonsplat(<4 x i32> %arg) {
 
 ; (c1-x)+c2
 
-define <4 x i32> @const_sub_add_const(<4 x i32> %arg) {
+define i32 @const_sub_add_const(i32 %arg) {
 ; CHECK-LABEL: const_sub_add_const:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #10
+; CHECK-NEXT:    sub w0, w8, w0
+; CHECK-NEXT:    ret
+  %t0 = sub i32 8, %arg
+  %t1 = add i32 %t0, 2
+  ret i32 %t1
+}
+
+define i32 @const_sub_add_const_extrause(i32 %arg) {
+; CHECK-LABEL: const_sub_add_const_extrause:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    stp x19, x30, [sp, #-16]! // 16-byte Folded Spill
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset w30, -8
+; CHECK-NEXT:    .cfi_offset w19, -16
+; CHECK-NEXT:    mov w8, #8
+; CHECK-NEXT:    mov w19, w0
+; CHECK-NEXT:    sub w0, w8, w0
+; CHECK-NEXT:    bl use
+; CHECK-NEXT:    mov w8, #10
+; CHECK-NEXT:    sub w0, w8, w19
+; CHECK-NEXT:    ldp x19, x30, [sp], #16 // 16-byte Folded Reload
+; CHECK-NEXT:    ret
+  %t0 = sub i32 8, %arg
+  call void @use(i32 %t0)
+  %t1 = add i32 %t0, 2
+  ret i32 %t1
+}
+
+define <4 x i32> @vec_const_sub_add_const(<4 x i32> %arg) {
+; CHECK-LABEL: vec_const_sub_add_const:
+; CHECK:       // %bb.0:
 ; CHECK-NEXT:    movi v1.4s, #10
 ; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    ret
@@ -304,8 +515,8 @@ define <4 x i32> @const_sub_add_const(<4 x i32> %arg) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @const_sub_add_const_extrause(<4 x i32> %arg) {
-; CHECK-LABEL: const_sub_add_const_extrause:
+define <4 x i32> @vec_const_sub_add_const_extrause(<4 x i32> %arg) {
+; CHECK-LABEL: vec_const_sub_add_const_extrause:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #32 // =32
 ; CHECK-NEXT:    str x30, [sp, #16] // 8-byte Folded Spill
@@ -314,7 +525,7 @@ define <4 x i32> @const_sub_add_const_extrause(<4 x i32> %arg) {
 ; CHECK-NEXT:    movi v1.4s, #8
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
 ; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    bl use
+; CHECK-NEXT:    bl vec_use
 ; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr x30, [sp, #16] // 8-byte Folded Reload
 ; CHECK-NEXT:    movi v0.4s, #10
@@ -322,16 +533,16 @@ define <4 x i32> @const_sub_add_const_extrause(<4 x i32> %arg) {
 ; CHECK-NEXT:    add sp, sp, #32 // =32
 ; CHECK-NEXT:    ret
   %t0 = sub <4 x i32> <i32 8, i32 8, i32 8, i32 8>, %arg
-  call void @use(<4 x i32> %t0)
+  call void @vec_use(<4 x i32> %t0)
   %t1 = add <4 x i32> %t0, <i32 2, i32 2, i32 2, i32 2>
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @const_sub_add_const_nonsplat(<4 x i32> %arg) {
-; CHECK-LABEL: const_sub_add_const_nonsplat:
+define <4 x i32> @vec_const_sub_add_const_nonsplat(<4 x i32> %arg) {
+; CHECK-LABEL: vec_const_sub_add_const_nonsplat:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    adrp x8, .LCPI20_0
-; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI20_0]
+; CHECK-NEXT:    adrp x8, .LCPI34_0
+; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI34_0]
 ; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    ret
   %t0 = sub <4 x i32> <i32 21, i32 undef, i32 8, i32 8>, %arg
@@ -341,9 +552,41 @@ define <4 x i32> @const_sub_add_const_nonsplat(<4 x i32> %arg) {
 
 ; (c1-x)-c2
 
-define <4 x i32> @const_sub_sub_const(<4 x i32> %arg) {
+define i32 @const_sub_sub_const(i32 %arg) {
 ; CHECK-LABEL: const_sub_sub_const:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    mov w8, #6
+; CHECK-NEXT:    sub w0, w8, w0
+; CHECK-NEXT:    ret
+  %t0 = sub i32 8, %arg
+  %t1 = sub i32 %t0, 2
+  ret i32 %t1
+}
+
+define i32 @const_sub_sub_const_extrause(i32 %arg) {
+; CHECK-LABEL: const_sub_sub_const_extrause:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    stp x19, x30, [sp, #-16]! // 16-byte Folded Spill
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset w30, -8
+; CHECK-NEXT:    .cfi_offset w19, -16
+; CHECK-NEXT:    mov w8, #8
+; CHECK-NEXT:    mov w19, w0
+; CHECK-NEXT:    sub w0, w8, w0
+; CHECK-NEXT:    bl use
+; CHECK-NEXT:    mov w8, #6
+; CHECK-NEXT:    sub w0, w8, w19
+; CHECK-NEXT:    ldp x19, x30, [sp], #16 // 16-byte Folded Reload
+; CHECK-NEXT:    ret
+  %t0 = sub i32 8, %arg
+  call void @use(i32 %t0)
+  %t1 = sub i32 %t0, 2
+  ret i32 %t1
+}
+
+define <4 x i32> @vec_const_sub_sub_const(<4 x i32> %arg) {
+; CHECK-LABEL: vec_const_sub_sub_const:
+; CHECK:       // %bb.0:
 ; CHECK-NEXT:    movi v1.4s, #6
 ; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    ret
@@ -352,8 +595,8 @@ define <4 x i32> @const_sub_sub_const(<4 x i32> %arg) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @const_sub_sub_const_extrause(<4 x i32> %arg) {
-; CHECK-LABEL: const_sub_sub_const_extrause:
+define <4 x i32> @vec_const_sub_sub_const_extrause(<4 x i32> %arg) {
+; CHECK-LABEL: vec_const_sub_sub_const_extrause:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #32 // =32
 ; CHECK-NEXT:    str x30, [sp, #16] // 8-byte Folded Spill
@@ -362,7 +605,7 @@ define <4 x i32> @const_sub_sub_const_extrause(<4 x i32> %arg) {
 ; CHECK-NEXT:    movi v1.4s, #8
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
 ; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
-; CHECK-NEXT:    bl use
+; CHECK-NEXT:    bl vec_use
 ; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr x30, [sp, #16] // 8-byte Folded Reload
 ; CHECK-NEXT:    movi v0.4s, #6
@@ -370,16 +613,16 @@ define <4 x i32> @const_sub_sub_const_extrause(<4 x i32> %arg) {
 ; CHECK-NEXT:    add sp, sp, #32 // =32
 ; CHECK-NEXT:    ret
   %t0 = sub <4 x i32> <i32 8, i32 8, i32 8, i32 8>, %arg
-  call void @use(<4 x i32> %t0)
+  call void @vec_use(<4 x i32> %t0)
   %t1 = sub <4 x i32> %t0, <i32 2, i32 2, i32 2, i32 2>
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @const_sub_sub_const_nonsplat(<4 x i32> %arg) {
-; CHECK-LABEL: const_sub_sub_const_nonsplat:
+define <4 x i32> @vec_const_sub_sub_const_nonsplat(<4 x i32> %arg) {
+; CHECK-LABEL: vec_const_sub_sub_const_nonsplat:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    adrp x8, .LCPI23_0
-; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI23_0]
+; CHECK-NEXT:    adrp x8, .LCPI39_0
+; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI39_0]
 ; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    ret
   %t0 = sub <4 x i32> <i32 21, i32 undef, i32 8, i32 8>, %arg
@@ -389,9 +632,40 @@ define <4 x i32> @const_sub_sub_const_nonsplat(<4 x i32> %arg) {
 
 ; c2-(c1-x)
 
-define <4 x i32> @const_sub_const_sub(<4 x i32> %arg) {
+define i32 @const_sub_const_sub(i32 %arg) {
 ; CHECK-LABEL: const_sub_const_sub:
 ; CHECK:       // %bb.0:
+; CHECK-NEXT:    sub w0, w0, #6 // =6
+; CHECK-NEXT:    ret
+  %t0 = sub i32 8, %arg
+  %t1 = sub i32 2, %t0
+  ret i32 %t1
+}
+
+define i32 @const_sub_const_sub_extrause(i32 %arg) {
+; CHECK-LABEL: const_sub_const_sub_extrause:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    stp x19, x30, [sp, #-16]! // 16-byte Folded Spill
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset w30, -8
+; CHECK-NEXT:    .cfi_offset w19, -16
+; CHECK-NEXT:    mov w8, #8
+; CHECK-NEXT:    sub w19, w8, w0
+; CHECK-NEXT:    mov w0, w19
+; CHECK-NEXT:    bl use
+; CHECK-NEXT:    mov w8, #2
+; CHECK-NEXT:    sub w0, w8, w19
+; CHECK-NEXT:    ldp x19, x30, [sp], #16 // 16-byte Folded Reload
+; CHECK-NEXT:    ret
+  %t0 = sub i32 8, %arg
+  call void @use(i32 %t0)
+  %t1 = sub i32 2, %t0
+  ret i32 %t1
+}
+
+define <4 x i32> @vec_const_sub_const_sub(<4 x i32> %arg) {
+; CHECK-LABEL: vec_const_sub_const_sub:
+; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mvni v1.4s, #5
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
@@ -400,8 +674,8 @@ define <4 x i32> @const_sub_const_sub(<4 x i32> %arg) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @const_sub_const_sub_extrause(<4 x i32> %arg) {
-; CHECK-LABEL: const_sub_const_sub_extrause:
+define <4 x i32> @vec_const_sub_const_sub_extrause(<4 x i32> %arg) {
+; CHECK-LABEL: vec_const_sub_const_sub_extrause:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #32 // =32
 ; CHECK-NEXT:    str x30, [sp, #16] // 8-byte Folded Spill
@@ -410,7 +684,7 @@ define <4 x i32> @const_sub_const_sub_extrause(<4 x i32> %arg) {
 ; CHECK-NEXT:    movi v1.4s, #8
 ; CHECK-NEXT:    sub v0.4s, v1.4s, v0.4s
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
-; CHECK-NEXT:    bl use
+; CHECK-NEXT:    bl vec_use
 ; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr x30, [sp, #16] // 8-byte Folded Reload
 ; CHECK-NEXT:    movi v0.4s, #2
@@ -418,16 +692,16 @@ define <4 x i32> @const_sub_const_sub_extrause(<4 x i32> %arg) {
 ; CHECK-NEXT:    add sp, sp, #32 // =32
 ; CHECK-NEXT:    ret
   %t0 = sub <4 x i32> <i32 8, i32 8, i32 8, i32 8>, %arg
-  call void @use(<4 x i32> %t0)
+  call void @vec_use(<4 x i32> %t0)
   %t1 = sub <4 x i32> <i32 2, i32 2, i32 2, i32 2>, %t0
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @const_sub_const_sub_nonsplat(<4 x i32> %arg) {
-; CHECK-LABEL: const_sub_const_sub_nonsplat:
+define <4 x i32> @vec_const_sub_const_sub_nonsplat(<4 x i32> %arg) {
+; CHECK-LABEL: vec_const_sub_const_sub_nonsplat:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    adrp x8, .LCPI26_0
-; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI26_0]
+; CHECK-NEXT:    adrp x8, .LCPI44_0
+; CHECK-NEXT:    ldr q1, [x8, :lo12:.LCPI44_0]
 ; CHECK-NEXT:    add v0.4s, v0.4s, v1.4s
 ; CHECK-NEXT:    ret
   %t0 = sub <4 x i32> <i32 21, i32 undef, i32 8, i32 8>, %arg

diff --git a/test/CodeGen/X86/addsub-constant-folding.ll b/test/CodeGen/X86/addsub-constant-folding.ll
index 3c48494ae39de97f03e372a7f66e43550ccc28c1..09cd2bcea962dd1287eaa2075634896ef6385eb6 100644 (file)
--- a/test/CodeGen/X86/addsub-constant-folding.ll
+++ b/test/CodeGen/X86/addsub-constant-folding.ll
@@ -2,18 +2,73 @@
 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefixes=X86,SSE,X86-SSE
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefixes=X64,SSE,X64-SSE
 
-declare void @use(<4 x i32> %arg)
+declare void @use(i32 %arg)
+declare void @vec_use(<4 x i32> %arg)
 
 ; (x+c1)+c2
 
-define <4 x i32> @add_const_add_const(<4 x i32> %arg) {
+define i32 @add_const_add_const(i32 %arg) {
 ; X86-LABEL: add_const_add_const:
 ; X86:       # %bb.0:
-; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm0
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    addl $10, %eax
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: add_const_add_const:
 ; X64:       # %bb.0:
+; X64-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-NEXT:    leal 10(%rdi), %eax
+; X64-NEXT:    retq
+  %t0 = add i32 %arg, 8
+  %t1 = add i32 %t0, 2
+  ret i32 %t1
+}
+
+define i32 @add_const_add_const_extrause(i32 %arg) {
+; X86-LABEL: add_const_add_const_extrause:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    .cfi_offset %esi, -8
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    leal 8(%esi), %eax
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    .cfi_adjust_cfa_offset 4
+; X86-NEXT:    calll use
+; X86-NEXT:    addl $4, %esp
+; X86-NEXT:    .cfi_adjust_cfa_offset -4
+; X86-NEXT:    addl $10, %esi
+; X86-NEXT:    movl %esi, %eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 4
+; X86-NEXT:    retl
+;
+; X64-LABEL: add_const_add_const_extrause:
+; X64:       # %bb.0:
+; X64-NEXT:    pushq %rbx
+; X64-NEXT:    .cfi_def_cfa_offset 16
+; X64-NEXT:    .cfi_offset %rbx, -16
+; X64-NEXT:    movl %edi, %ebx
+; X64-NEXT:    leal 8(%rbx), %edi
+; X64-NEXT:    callq use
+; X64-NEXT:    leal 10(%rbx), %eax
+; X64-NEXT:    popq %rbx
+; X64-NEXT:    .cfi_def_cfa_offset 8
+; X64-NEXT:    retq
+  %t0 = add i32 %arg, 8
+  call void @use(i32 %t0)
+  %t1 = add i32 %t0, 2
+  ret i32 %t1
+}
+
+define <4 x i32> @vec_add_const_add_const(<4 x i32> %arg) {
+; X86-LABEL: vec_add_const_add_const:
+; X86:       # %bb.0:
+; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm0
+; X86-NEXT:    retl
+;
+; X64-LABEL: vec_add_const_add_const:
+; X64:       # %bb.0:
 ; X64-NEXT:    paddd {{.*}}(%rip), %xmm0
 ; X64-NEXT:    retq
   %t0 = add <4 x i32> %arg, <i32 8, i32 8, i32 8, i32 8>
@@ -21,8 +76,8 @@ define <4 x i32> @add_const_add_const(<4 x i32> %arg) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @add_const_add_const_extrause(<4 x i32> %arg) {
-; X86-LABEL: add_const_add_const_extrause:
+define <4 x i32> @vec_add_const_add_const_extrause(<4 x i32> %arg) {
+; X86-LABEL: vec_add_const_add_const_extrause:
 ; X86:       # %bb.0:
 ; X86-NEXT:    subl $28, %esp
 ; X86-NEXT:    .cfi_def_cfa_offset 32
@@ -30,14 +85,14 @@ define <4 x i32> @add_const_add_const_extrause(<4 x i32> %arg) {
 ; X86-NEXT:    movdqu %xmm0, (%esp) # 16-byte Spill
 ; X86-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8]
 ; X86-NEXT:    paddd %xmm1, %xmm0
-; X86-NEXT:    calll use
+; X86-NEXT:    calll vec_use
 ; X86-NEXT:    movdqu (%esp), %xmm0 # 16-byte Reload
 ; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm0
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    .cfi_def_cfa_offset 4
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: add_const_add_const_extrause:
+; X64-LABEL: vec_add_const_add_const_extrause:
 ; X64:       # %bb.0:
 ; X64-NEXT:    subq $24, %rsp
 ; X64-NEXT:    .cfi_def_cfa_offset 32
@@ -45,25 +100,25 @@ define <4 x i32> @add_const_add_const_extrause(<4 x i32> %arg) {
 ; X64-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
 ; X64-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8]
 ; X64-NEXT:    paddd %xmm1, %xmm0
-; X64-NEXT:    callq use
+; X64-NEXT:    callq vec_use
 ; X64-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
 ; X64-NEXT:    paddd {{.*}}(%rip), %xmm0
 ; X64-NEXT:    addq $24, %rsp
 ; X64-NEXT:    .cfi_def_cfa_offset 8
 ; X64-NEXT:    retq
   %t0 = add <4 x i32> %arg, <i32 8, i32 8, i32 8, i32 8>
-  call void @use(<4 x i32> %t0)
+  call void @vec_use(<4 x i32> %t0)
   %t1 = add <4 x i32> %t0, <i32 2, i32 2, i32 2, i32 2>
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @add_const_add_const_nonsplat(<4 x i32> %arg) {
-; X86-LABEL: add_const_add_const_nonsplat:
+define <4 x i32> @vec_add_const_add_const_nonsplat(<4 x i32> %arg) {
+; X86-LABEL: vec_add_const_add_const_nonsplat:
 ; X86:       # %bb.0:
 ; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm0
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: add_const_add_const_nonsplat:
+; X64-LABEL: vec_add_const_add_const_nonsplat:
 ; X64:       # %bb.0:
 ; X64-NEXT:    paddd {{.*}}(%rip), %xmm0
 ; X64-NEXT:    retq
@@ -74,14 +129,68 @@ define <4 x i32> @add_const_add_const_nonsplat(<4 x i32> %arg) {
 
 ; (x+c1)-c2
 
-define <4 x i32> @add_const_sub_const(<4 x i32> %arg) {
+define i32 @add_const_sub_const(i32 %arg) {
 ; X86-LABEL: add_const_sub_const:
 ; X86:       # %bb.0:
-; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm0
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    addl $6, %eax
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: add_const_sub_const:
 ; X64:       # %bb.0:
+; X64-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-NEXT:    leal 6(%rdi), %eax
+; X64-NEXT:    retq
+  %t0 = add i32 %arg, 8
+  %t1 = sub i32 %t0, 2
+  ret i32 %t1
+}
+
+define i32 @add_const_sub_const_extrause(i32 %arg) {
+; X86-LABEL: add_const_sub_const_extrause:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    .cfi_offset %esi, -8
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    leal 8(%esi), %eax
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    .cfi_adjust_cfa_offset 4
+; X86-NEXT:    calll use
+; X86-NEXT:    addl $4, %esp
+; X86-NEXT:    .cfi_adjust_cfa_offset -4
+; X86-NEXT:    addl $6, %esi
+; X86-NEXT:    movl %esi, %eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 4
+; X86-NEXT:    retl
+;
+; X64-LABEL: add_const_sub_const_extrause:
+; X64:       # %bb.0:
+; X64-NEXT:    pushq %rbx
+; X64-NEXT:    .cfi_def_cfa_offset 16
+; X64-NEXT:    .cfi_offset %rbx, -16
+; X64-NEXT:    movl %edi, %ebx
+; X64-NEXT:    leal 8(%rbx), %edi
+; X64-NEXT:    callq use
+; X64-NEXT:    leal 6(%rbx), %eax
+; X64-NEXT:    popq %rbx
+; X64-NEXT:    .cfi_def_cfa_offset 8
+; X64-NEXT:    retq
+  %t0 = add i32 %arg, 8
+  call void @use(i32 %t0)
+  %t1 = sub i32 %t0, 2
+  ret i32 %t1
+}
+
+define <4 x i32> @vec_add_const_sub_const(<4 x i32> %arg) {
+; X86-LABEL: vec_add_const_sub_const:
+; X86:       # %bb.0:
+; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm0
+; X86-NEXT:    retl
+;
+; X64-LABEL: vec_add_const_sub_const:
+; X64:       # %bb.0:
 ; X64-NEXT:    paddd {{.*}}(%rip), %xmm0
 ; X64-NEXT:    retq
   %t0 = add <4 x i32> %arg, <i32 8, i32 8, i32 8, i32 8>
@@ -89,8 +198,8 @@ define <4 x i32> @add_const_sub_const(<4 x i32> %arg) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @add_const_sub_const_extrause(<4 x i32> %arg) {
-; X86-LABEL: add_const_sub_const_extrause:
+define <4 x i32> @vec_add_const_sub_const_extrause(<4 x i32> %arg) {
+; X86-LABEL: vec_add_const_sub_const_extrause:
 ; X86:       # %bb.0:
 ; X86-NEXT:    subl $28, %esp
 ; X86-NEXT:    .cfi_def_cfa_offset 32
@@ -98,14 +207,14 @@ define <4 x i32> @add_const_sub_const_extrause(<4 x i32> %arg) {
 ; X86-NEXT:    movdqu %xmm0, (%esp) # 16-byte Spill
 ; X86-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8]
 ; X86-NEXT:    paddd %xmm1, %xmm0
-; X86-NEXT:    calll use
+; X86-NEXT:    calll vec_use
 ; X86-NEXT:    movdqu (%esp), %xmm0 # 16-byte Reload
 ; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm0
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    .cfi_def_cfa_offset 4
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: add_const_sub_const_extrause:
+; X64-LABEL: vec_add_const_sub_const_extrause:
 ; X64:       # %bb.0:
 ; X64-NEXT:    subq $24, %rsp
 ; X64-NEXT:    .cfi_def_cfa_offset 32
@@ -113,25 +222,25 @@ define <4 x i32> @add_const_sub_const_extrause(<4 x i32> %arg) {
 ; X64-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
 ; X64-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8]
 ; X64-NEXT:    paddd %xmm1, %xmm0
-; X64-NEXT:    callq use
+; X64-NEXT:    callq vec_use
 ; X64-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
 ; X64-NEXT:    paddd {{.*}}(%rip), %xmm0
 ; X64-NEXT:    addq $24, %rsp
 ; X64-NEXT:    .cfi_def_cfa_offset 8
 ; X64-NEXT:    retq
   %t0 = add <4 x i32> %arg, <i32 8, i32 8, i32 8, i32 8>
-  call void @use(<4 x i32> %t0)
+  call void @vec_use(<4 x i32> %t0)
   %t1 = sub <4 x i32> %t0, <i32 2, i32 2, i32 2, i32 2>
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @add_const_sub_const_nonsplat(<4 x i32> %arg) {
-; X86-LABEL: add_const_sub_const_nonsplat:
+define <4 x i32> @vec_add_const_sub_const_nonsplat(<4 x i32> %arg) {
+; X86-LABEL: vec_add_const_sub_const_nonsplat:
 ; X86:       # %bb.0:
 ; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm0
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: add_const_sub_const_nonsplat:
+; X64-LABEL: vec_add_const_sub_const_nonsplat:
 ; X64:       # %bb.0:
 ; X64-NEXT:    paddd {{.*}}(%rip), %xmm0
 ; X64-NEXT:    retq
@@ -142,15 +251,70 @@ define <4 x i32> @add_const_sub_const_nonsplat(<4 x i32> %arg) {
 
 ; c2-(x+c1)
 
-define <4 x i32> @add_const_const_sub(<4 x i32> %arg) {
+define i32 @add_const_const_sub(i32 %arg) {
 ; X86-LABEL: add_const_const_sub:
 ; X86:       # %bb.0:
+; X86-NEXT:    movl $-6, %eax
+; X86-NEXT:    subl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    retl
+;
+; X64-LABEL: add_const_const_sub:
+; X64:       # %bb.0:
+; X64-NEXT:    movl $-6, %eax
+; X64-NEXT:    subl %edi, %eax
+; X64-NEXT:    retq
+  %t0 = add i32 %arg, 8
+  %t1 = sub i32 2, %t0
+  ret i32 %t1
+}
+
+define i32 @add_const_const_sub_extrause(i32 %arg) {
+; X86-LABEL: add_const_const_sub_extrause:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    .cfi_offset %esi, -8
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    leal 8(%esi), %eax
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    .cfi_adjust_cfa_offset 4
+; X86-NEXT:    calll use
+; X86-NEXT:    addl $4, %esp
+; X86-NEXT:    .cfi_adjust_cfa_offset -4
+; X86-NEXT:    movl $-6, %eax
+; X86-NEXT:    subl %esi, %eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 4
+; X86-NEXT:    retl
+;
+; X64-LABEL: add_const_const_sub_extrause:
+; X64:       # %bb.0:
+; X64-NEXT:    pushq %rbx
+; X64-NEXT:    .cfi_def_cfa_offset 16
+; X64-NEXT:    .cfi_offset %rbx, -16
+; X64-NEXT:    movl %edi, %ebx
+; X64-NEXT:    leal 8(%rbx), %edi
+; X64-NEXT:    callq use
+; X64-NEXT:    movl $-6, %eax
+; X64-NEXT:    subl %ebx, %eax
+; X64-NEXT:    popq %rbx
+; X64-NEXT:    .cfi_def_cfa_offset 8
+; X64-NEXT:    retq
+  %t0 = add i32 %arg, 8
+  call void @use(i32 %t0)
+  %t1 = sub i32 2, %t0
+  ret i32 %t1
+}
+
+define <4 x i32> @vec_add_const_const_sub(<4 x i32> %arg) {
+; X86-LABEL: vec_add_const_const_sub:
+; X86:       # %bb.0:
 ; X86-NEXT:    movdqa {{.*#+}} xmm1 = [4294967290,4294967290,4294967290,4294967290]
 ; X86-NEXT:    psubd %xmm0, %xmm1
 ; X86-NEXT:    movdqa %xmm1, %xmm0
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: add_const_const_sub:
+; X64-LABEL: vec_add_const_const_sub:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movdqa {{.*#+}} xmm1 = [4294967290,4294967290,4294967290,4294967290]
 ; X64-NEXT:    psubd %xmm0, %xmm1
@@ -161,8 +325,8 @@ define <4 x i32> @add_const_const_sub(<4 x i32> %arg) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @add_const_const_sub_extrause(<4 x i32> %arg) {
-; X86-LABEL: add_const_const_sub_extrause:
+define <4 x i32> @vec_add_const_const_sub_extrause(<4 x i32> %arg) {
+; X86-LABEL: vec_add_const_const_sub_extrause:
 ; X86:       # %bb.0:
 ; X86-NEXT:    subl $28, %esp
 ; X86-NEXT:    .cfi_def_cfa_offset 32
@@ -170,7 +334,7 @@ define <4 x i32> @add_const_const_sub_extrause(<4 x i32> %arg) {
 ; X86-NEXT:    movdqu %xmm0, (%esp) # 16-byte Spill
 ; X86-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8]
 ; X86-NEXT:    paddd %xmm1, %xmm0
-; X86-NEXT:    calll use
+; X86-NEXT:    calll vec_use
 ; X86-NEXT:    movdqa {{.*#+}} xmm0 = [4294967290,4294967290,4294967290,4294967290]
 ; X86-NEXT:    movdqu (%esp), %xmm1 # 16-byte Reload
 ; X86-NEXT:    psubd %xmm1, %xmm0
@@ -178,7 +342,7 @@ define <4 x i32> @add_const_const_sub_extrause(<4 x i32> %arg) {
 ; X86-NEXT:    .cfi_def_cfa_offset 4
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: add_const_const_sub_extrause:
+; X64-LABEL: vec_add_const_const_sub_extrause:
 ; X64:       # %bb.0:
 ; X64-NEXT:    subq $24, %rsp
 ; X64-NEXT:    .cfi_def_cfa_offset 32
@@ -186,27 +350,27 @@ define <4 x i32> @add_const_const_sub_extrause(<4 x i32> %arg) {
 ; X64-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
 ; X64-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8]
 ; X64-NEXT:    paddd %xmm1, %xmm0
-; X64-NEXT:    callq use
+; X64-NEXT:    callq vec_use
 ; X64-NEXT:    movdqa {{.*#+}} xmm0 = [4294967290,4294967290,4294967290,4294967290]
 ; X64-NEXT:    psubd (%rsp), %xmm0 # 16-byte Folded Reload
 ; X64-NEXT:    addq $24, %rsp
 ; X64-NEXT:    .cfi_def_cfa_offset 8
 ; X64-NEXT:    retq
   %t0 = add <4 x i32> %arg, <i32 8, i32 8, i32 8, i32 8>
-  call void @use(<4 x i32> %t0)
+  call void @vec_use(<4 x i32> %t0)
   %t1 = sub <4 x i32> <i32 2, i32 2, i32 2, i32 2>, %t0
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @add_const_const_sub_nonsplat(<4 x i32> %arg) {
-; X86-LABEL: add_const_const_sub_nonsplat:
+define <4 x i32> @vec_add_const_const_sub_nonsplat(<4 x i32> %arg) {
+; X86-LABEL: vec_add_const_const_sub_nonsplat:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movdqa {{.*#+}} xmm1 = <4294967277,u,u,4294967290>
 ; X86-NEXT:    psubd %xmm0, %xmm1
 ; X86-NEXT:    movdqa %xmm1, %xmm0
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: add_const_const_sub_nonsplat:
+; X64-LABEL: vec_add_const_const_sub_nonsplat:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movdqa {{.*#+}} xmm1 = <4294967277,u,u,4294967290>
 ; X64-NEXT:    psubd %xmm0, %xmm1
@@ -219,14 +383,68 @@ define <4 x i32> @add_const_const_sub_nonsplat(<4 x i32> %arg) {
 
 ; (x-c1)+c2
 
-define <4 x i32> @sub_const_add_const(<4 x i32> %arg) {
+define i32 @sub_const_add_const(i32 %arg) {
 ; X86-LABEL: sub_const_add_const:
 ; X86:       # %bb.0:
-; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm0
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    addl $-6, %eax
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: sub_const_add_const:
 ; X64:       # %bb.0:
+; X64-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-NEXT:    leal -6(%rdi), %eax
+; X64-NEXT:    retq
+  %t0 = sub i32 %arg, 8
+  %t1 = add i32 %t0, 2
+  ret i32 %t1
+}
+
+define i32 @sub_const_add_const_extrause(i32 %arg) {
+; X86-LABEL: sub_const_add_const_extrause:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    .cfi_offset %esi, -8
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    leal -8(%esi), %eax
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    .cfi_adjust_cfa_offset 4
+; X86-NEXT:    calll use
+; X86-NEXT:    addl $4, %esp
+; X86-NEXT:    .cfi_adjust_cfa_offset -4
+; X86-NEXT:    addl $-6, %esi
+; X86-NEXT:    movl %esi, %eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 4
+; X86-NEXT:    retl
+;
+; X64-LABEL: sub_const_add_const_extrause:
+; X64:       # %bb.0:
+; X64-NEXT:    pushq %rbx
+; X64-NEXT:    .cfi_def_cfa_offset 16
+; X64-NEXT:    .cfi_offset %rbx, -16
+; X64-NEXT:    movl %edi, %ebx
+; X64-NEXT:    leal -8(%rbx), %edi
+; X64-NEXT:    callq use
+; X64-NEXT:    leal -6(%rbx), %eax
+; X64-NEXT:    popq %rbx
+; X64-NEXT:    .cfi_def_cfa_offset 8
+; X64-NEXT:    retq
+  %t0 = sub i32 %arg, 8
+  call void @use(i32 %t0)
+  %t1 = add i32 %t0, 2
+  ret i32 %t1
+}
+
+define <4 x i32> @vec_sub_const_add_const(<4 x i32> %arg) {
+; X86-LABEL: vec_sub_const_add_const:
+; X86:       # %bb.0:
+; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm0
+; X86-NEXT:    retl
+;
+; X64-LABEL: vec_sub_const_add_const:
+; X64:       # %bb.0:
 ; X64-NEXT:    paddd {{.*}}(%rip), %xmm0
 ; X64-NEXT:    retq
   %t0 = sub <4 x i32> %arg, <i32 8, i32 8, i32 8, i32 8>
@@ -234,45 +452,45 @@ define <4 x i32> @sub_const_add_const(<4 x i32> %arg) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @sub_const_add_const_extrause(<4 x i32> %arg) {
-; X86-LABEL: sub_const_add_const_extrause:
+define <4 x i32> @vec_sub_const_add_const_extrause(<4 x i32> %arg) {
+; X86-LABEL: vec_sub_const_add_const_extrause:
 ; X86:       # %bb.0:
 ; X86-NEXT:    subl $28, %esp
 ; X86-NEXT:    .cfi_def_cfa_offset 32
 ; X86-NEXT:    movdqu %xmm0, (%esp) # 16-byte Spill
 ; X86-NEXT:    psubd {{\.LCPI.*}}, %xmm0
-; X86-NEXT:    calll use
+; X86-NEXT:    calll vec_use
 ; X86-NEXT:    movdqu (%esp), %xmm0 # 16-byte Reload
 ; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm0
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    .cfi_def_cfa_offset 4
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: sub_const_add_const_extrause:
+; X64-LABEL: vec_sub_const_add_const_extrause:
 ; X64:       # %bb.0:
 ; X64-NEXT:    subq $24, %rsp
 ; X64-NEXT:    .cfi_def_cfa_offset 32
 ; X64-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
 ; X64-NEXT:    psubd {{.*}}(%rip), %xmm0
-; X64-NEXT:    callq use
+; X64-NEXT:    callq vec_use
 ; X64-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
 ; X64-NEXT:    paddd {{.*}}(%rip), %xmm0
 ; X64-NEXT:    addq $24, %rsp
 ; X64-NEXT:    .cfi_def_cfa_offset 8
 ; X64-NEXT:    retq
   %t0 = sub <4 x i32> %arg, <i32 8, i32 8, i32 8, i32 8>
-  call void @use(<4 x i32> %t0)
+  call void @vec_use(<4 x i32> %t0)
   %t1 = add <4 x i32> %t0, <i32 2, i32 2, i32 2, i32 2>
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @sub_const_add_const_nonsplat(<4 x i32> %arg) {
-; X86-LABEL: sub_const_add_const_nonsplat:
+define <4 x i32> @vec_sub_const_add_const_nonsplat(<4 x i32> %arg) {
+; X86-LABEL: vec_sub_const_add_const_nonsplat:
 ; X86:       # %bb.0:
 ; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm0
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: sub_const_add_const_nonsplat:
+; X64-LABEL: vec_sub_const_add_const_nonsplat:
 ; X64:       # %bb.0:
 ; X64-NEXT:    paddd {{.*}}(%rip), %xmm0
 ; X64-NEXT:    retq
@@ -283,14 +501,68 @@ define <4 x i32> @sub_const_add_const_nonsplat(<4 x i32> %arg) {
 
 ; (x-c1)-c2
 
-define <4 x i32> @sub_const_sub_const(<4 x i32> %arg) {
+define i32 @sub_const_sub_const(i32 %arg) {
 ; X86-LABEL: sub_const_sub_const:
 ; X86:       # %bb.0:
-; X86-NEXT:    psubd {{\.LCPI.*}}, %xmm0
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    addl $-10, %eax
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: sub_const_sub_const:
 ; X64:       # %bb.0:
+; X64-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-NEXT:    leal -10(%rdi), %eax
+; X64-NEXT:    retq
+  %t0 = sub i32 %arg, 8
+  %t1 = sub i32 %t0, 2
+  ret i32 %t1
+}
+
+define i32 @sub_const_sub_const_extrause(i32 %arg) {
+; X86-LABEL: sub_const_sub_const_extrause:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    .cfi_offset %esi, -8
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    leal -8(%esi), %eax
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    .cfi_adjust_cfa_offset 4
+; X86-NEXT:    calll use
+; X86-NEXT:    addl $4, %esp
+; X86-NEXT:    .cfi_adjust_cfa_offset -4
+; X86-NEXT:    addl $-10, %esi
+; X86-NEXT:    movl %esi, %eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 4
+; X86-NEXT:    retl
+;
+; X64-LABEL: sub_const_sub_const_extrause:
+; X64:       # %bb.0:
+; X64-NEXT:    pushq %rbx
+; X64-NEXT:    .cfi_def_cfa_offset 16
+; X64-NEXT:    .cfi_offset %rbx, -16
+; X64-NEXT:    movl %edi, %ebx
+; X64-NEXT:    leal -8(%rbx), %edi
+; X64-NEXT:    callq use
+; X64-NEXT:    leal -10(%rbx), %eax
+; X64-NEXT:    popq %rbx
+; X64-NEXT:    .cfi_def_cfa_offset 8
+; X64-NEXT:    retq
+  %t0 = sub i32 %arg, 8
+  call void @use(i32 %t0)
+  %t1 = sub i32 %t0, 2
+  ret i32 %t1
+}
+
+define <4 x i32> @vec_sub_const_sub_const(<4 x i32> %arg) {
+; X86-LABEL: vec_sub_const_sub_const:
+; X86:       # %bb.0:
+; X86-NEXT:    psubd {{\.LCPI.*}}, %xmm0
+; X86-NEXT:    retl
+;
+; X64-LABEL: vec_sub_const_sub_const:
+; X64:       # %bb.0:
 ; X64-NEXT:    psubd {{.*}}(%rip), %xmm0
 ; X64-NEXT:    retq
   %t0 = sub <4 x i32> %arg, <i32 8, i32 8, i32 8, i32 8>
@@ -298,45 +570,45 @@ define <4 x i32> @sub_const_sub_const(<4 x i32> %arg) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @sub_const_sub_const_extrause(<4 x i32> %arg) {
-; X86-LABEL: sub_const_sub_const_extrause:
+define <4 x i32> @vec_sub_const_sub_const_extrause(<4 x i32> %arg) {
+; X86-LABEL: vec_sub_const_sub_const_extrause:
 ; X86:       # %bb.0:
 ; X86-NEXT:    subl $28, %esp
 ; X86-NEXT:    .cfi_def_cfa_offset 32
 ; X86-NEXT:    movdqu %xmm0, (%esp) # 16-byte Spill
 ; X86-NEXT:    psubd {{\.LCPI.*}}, %xmm0
-; X86-NEXT:    calll use
+; X86-NEXT:    calll vec_use
 ; X86-NEXT:    movdqu (%esp), %xmm0 # 16-byte Reload
 ; X86-NEXT:    psubd {{\.LCPI.*}}, %xmm0
 ; X86-NEXT:    addl $28, %esp
 ; X86-NEXT:    .cfi_def_cfa_offset 4
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: sub_const_sub_const_extrause:
+; X64-LABEL: vec_sub_const_sub_const_extrause:
 ; X64:       # %bb.0:
 ; X64-NEXT:    subq $24, %rsp
 ; X64-NEXT:    .cfi_def_cfa_offset 32
 ; X64-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
 ; X64-NEXT:    psubd {{.*}}(%rip), %xmm0
-; X64-NEXT:    callq use
+; X64-NEXT:    callq vec_use
 ; X64-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
 ; X64-NEXT:    psubd {{.*}}(%rip), %xmm0
 ; X64-NEXT:    addq $24, %rsp
 ; X64-NEXT:    .cfi_def_cfa_offset 8
 ; X64-NEXT:    retq
   %t0 = sub <4 x i32> %arg, <i32 8, i32 8, i32 8, i32 8>
-  call void @use(<4 x i32> %t0)
+  call void @vec_use(<4 x i32> %t0)
   %t1 = sub <4 x i32> %t0, <i32 2, i32 2, i32 2, i32 2>
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @sub_const_sub_const_nonsplat(<4 x i32> %arg) {
-; X86-LABEL: sub_const_sub_const_nonsplat:
+define <4 x i32> @vec_sub_const_sub_const_nonsplat(<4 x i32> %arg) {
+; X86-LABEL: vec_sub_const_sub_const_nonsplat:
 ; X86:       # %bb.0:
 ; X86-NEXT:    psubd {{\.LCPI.*}}, %xmm0
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: sub_const_sub_const_nonsplat:
+; X64-LABEL: vec_sub_const_sub_const_nonsplat:
 ; X64:       # %bb.0:
 ; X64-NEXT:    psubd {{.*}}(%rip), %xmm0
 ; X64-NEXT:    retq
@@ -347,15 +619,70 @@ define <4 x i32> @sub_const_sub_const_nonsplat(<4 x i32> %arg) {
 
 ; c2-(x-c1)
 
-define <4 x i32> @sub_const_const_sub(<4 x i32> %arg) {
+define i32 @sub_const_const_sub(i32 %arg) {
 ; X86-LABEL: sub_const_const_sub:
 ; X86:       # %bb.0:
+; X86-NEXT:    movl $10, %eax
+; X86-NEXT:    subl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    retl
+;
+; X64-LABEL: sub_const_const_sub:
+; X64:       # %bb.0:
+; X64-NEXT:    movl $10, %eax
+; X64-NEXT:    subl %edi, %eax
+; X64-NEXT:    retq
+  %t0 = sub i32 %arg, 8
+  %t1 = sub i32 2, %t0
+  ret i32 %t1
+}
+
+define i32 @sub_const_const_sub_extrause(i32 %arg) {
+; X86-LABEL: sub_const_const_sub_extrause:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    .cfi_offset %esi, -8
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    leal -8(%esi), %eax
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    .cfi_adjust_cfa_offset 4
+; X86-NEXT:    calll use
+; X86-NEXT:    addl $4, %esp
+; X86-NEXT:    .cfi_adjust_cfa_offset -4
+; X86-NEXT:    movl $10, %eax
+; X86-NEXT:    subl %esi, %eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 4
+; X86-NEXT:    retl
+;
+; X64-LABEL: sub_const_const_sub_extrause:
+; X64:       # %bb.0:
+; X64-NEXT:    pushq %rbx
+; X64-NEXT:    .cfi_def_cfa_offset 16
+; X64-NEXT:    .cfi_offset %rbx, -16
+; X64-NEXT:    movl %edi, %ebx
+; X64-NEXT:    leal -8(%rbx), %edi
+; X64-NEXT:    callq use
+; X64-NEXT:    movl $10, %eax
+; X64-NEXT:    subl %ebx, %eax
+; X64-NEXT:    popq %rbx
+; X64-NEXT:    .cfi_def_cfa_offset 8
+; X64-NEXT:    retq
+  %t0 = sub i32 %arg, 8
+  call void @use(i32 %t0)
+  %t1 = sub i32 2, %t0
+  ret i32 %t1
+}
+
+define <4 x i32> @vec_sub_const_const_sub(<4 x i32> %arg) {
+; X86-LABEL: vec_sub_const_const_sub:
+; X86:       # %bb.0:
 ; X86-NEXT:    movdqa {{.*#+}} xmm1 = [10,10,10,10]
 ; X86-NEXT:    psubd %xmm0, %xmm1
 ; X86-NEXT:    movdqa %xmm1, %xmm0
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: sub_const_const_sub:
+; X64-LABEL: vec_sub_const_const_sub:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movdqa {{.*#+}} xmm1 = [10,10,10,10]
 ; X64-NEXT:    psubd %xmm0, %xmm1
@@ -366,14 +693,14 @@ define <4 x i32> @sub_const_const_sub(<4 x i32> %arg) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @sub_const_const_sub_extrause(<4 x i32> %arg) {
-; X86-LABEL: sub_const_const_sub_extrause:
+define <4 x i32> @vec_sub_const_const_sub_extrause(<4 x i32> %arg) {
+; X86-LABEL: vec_sub_const_const_sub_extrause:
 ; X86:       # %bb.0:
 ; X86-NEXT:    subl $28, %esp
 ; X86-NEXT:    .cfi_def_cfa_offset 32
 ; X86-NEXT:    psubd {{\.LCPI.*}}, %xmm0
 ; X86-NEXT:    movdqu %xmm0, (%esp) # 16-byte Spill
-; X86-NEXT:    calll use
+; X86-NEXT:    calll vec_use
 ; X86-NEXT:    movdqa {{.*#+}} xmm0 = [2,2,2,2]
 ; X86-NEXT:    movdqu (%esp), %xmm1 # 16-byte Reload
 ; X86-NEXT:    psubd %xmm1, %xmm0
@@ -381,33 +708,33 @@ define <4 x i32> @sub_const_const_sub_extrause(<4 x i32> %arg) {
 ; X86-NEXT:    .cfi_def_cfa_offset 4
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: sub_const_const_sub_extrause:
+; X64-LABEL: vec_sub_const_const_sub_extrause:
 ; X64:       # %bb.0:
 ; X64-NEXT:    subq $24, %rsp
 ; X64-NEXT:    .cfi_def_cfa_offset 32
 ; X64-NEXT:    psubd {{.*}}(%rip), %xmm0
 ; X64-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
-; X64-NEXT:    callq use
+; X64-NEXT:    callq vec_use
 ; X64-NEXT:    movdqa {{.*#+}} xmm0 = [2,2,2,2]
 ; X64-NEXT:    psubd (%rsp), %xmm0 # 16-byte Folded Reload
 ; X64-NEXT:    addq $24, %rsp
 ; X64-NEXT:    .cfi_def_cfa_offset 8
 ; X64-NEXT:    retq
   %t0 = sub <4 x i32> %arg, <i32 8, i32 8, i32 8, i32 8>
-  call void @use(<4 x i32> %t0)
+  call void @vec_use(<4 x i32> %t0)
   %t1 = sub <4 x i32> <i32 2, i32 2, i32 2, i32 2>, %t0
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @sub_const_const_sub_nonsplat(<4 x i32> %arg) {
-; X86-LABEL: sub_const_const_sub_nonsplat:
+define <4 x i32> @vec_sub_const_const_sub_nonsplat(<4 x i32> %arg) {
+; X86-LABEL: vec_sub_const_const_sub_nonsplat:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movdqa {{.*#+}} xmm1 = <23,u,u,10>
 ; X86-NEXT:    psubd %xmm0, %xmm1
 ; X86-NEXT:    movdqa %xmm1, %xmm0
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: sub_const_const_sub_nonsplat:
+; X64-LABEL: vec_sub_const_const_sub_nonsplat:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movdqa {{.*#+}} xmm1 = <23,u,u,10>
 ; X64-NEXT:    psubd %xmm0, %xmm1
@@ -420,15 +747,72 @@ define <4 x i32> @sub_const_const_sub_nonsplat(<4 x i32> %arg) {
 
 ; (c1-x)+c2
 
-define <4 x i32> @const_sub_add_const(<4 x i32> %arg) {
+define i32 @const_sub_add_const(i32 %arg) {
 ; X86-LABEL: const_sub_add_const:
 ; X86:       # %bb.0:
+; X86-NEXT:    movl $10, %eax
+; X86-NEXT:    subl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    retl
+;
+; X64-LABEL: const_sub_add_const:
+; X64:       # %bb.0:
+; X64-NEXT:    movl $10, %eax
+; X64-NEXT:    subl %edi, %eax
+; X64-NEXT:    retq
+  %t0 = sub i32 8, %arg
+  %t1 = add i32 %t0, 2
+  ret i32 %t1
+}
+
+define i32 @const_sub_add_const_extrause(i32 %arg) {
+; X86-LABEL: const_sub_add_const_extrause:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    .cfi_offset %esi, -8
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    movl $8, %eax
+; X86-NEXT:    subl %esi, %eax
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    .cfi_adjust_cfa_offset 4
+; X86-NEXT:    calll use
+; X86-NEXT:    addl $4, %esp
+; X86-NEXT:    .cfi_adjust_cfa_offset -4
+; X86-NEXT:    movl $10, %eax
+; X86-NEXT:    subl %esi, %eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 4
+; X86-NEXT:    retl
+;
+; X64-LABEL: const_sub_add_const_extrause:
+; X64:       # %bb.0:
+; X64-NEXT:    pushq %rbx
+; X64-NEXT:    .cfi_def_cfa_offset 16
+; X64-NEXT:    .cfi_offset %rbx, -16
+; X64-NEXT:    movl %edi, %ebx
+; X64-NEXT:    movl $8, %edi
+; X64-NEXT:    subl %ebx, %edi
+; X64-NEXT:    callq use
+; X64-NEXT:    movl $10, %eax
+; X64-NEXT:    subl %ebx, %eax
+; X64-NEXT:    popq %rbx
+; X64-NEXT:    .cfi_def_cfa_offset 8
+; X64-NEXT:    retq
+  %t0 = sub i32 8, %arg
+  call void @use(i32 %t0)
+  %t1 = add i32 %t0, 2
+  ret i32 %t1
+}
+
+define <4 x i32> @vec_const_sub_add_const(<4 x i32> %arg) {
+; X86-LABEL: vec_const_sub_add_const:
+; X86:       # %bb.0:
 ; X86-NEXT:    movdqa {{.*#+}} xmm1 = [10,10,10,10]
 ; X86-NEXT:    psubd %xmm0, %xmm1
 ; X86-NEXT:    movdqa %xmm1, %xmm0
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: const_sub_add_const:
+; X64-LABEL: vec_const_sub_add_const:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movdqa {{.*#+}} xmm1 = [10,10,10,10]
 ; X64-NEXT:    psubd %xmm0, %xmm1
@@ -439,8 +823,8 @@ define <4 x i32> @const_sub_add_const(<4 x i32> %arg) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @const_sub_add_const_extrause(<4 x i32> %arg) {
-; X86-LABEL: const_sub_add_const_extrause:
+define <4 x i32> @vec_const_sub_add_const_extrause(<4 x i32> %arg) {
+; X86-LABEL: vec_const_sub_add_const_extrause:
 ; X86:       # %bb.0:
 ; X86-NEXT:    subl $28, %esp
 ; X86-NEXT:    .cfi_def_cfa_offset 32
@@ -448,7 +832,7 @@ define <4 x i32> @const_sub_add_const_extrause(<4 x i32> %arg) {
 ; X86-NEXT:    movdqu %xmm0, (%esp) # 16-byte Spill
 ; X86-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8]
 ; X86-NEXT:    psubd %xmm1, %xmm0
-; X86-NEXT:    calll use
+; X86-NEXT:    calll vec_use
 ; X86-NEXT:    movdqa {{.*#+}} xmm0 = [10,10,10,10]
 ; X86-NEXT:    movdqu (%esp), %xmm1 # 16-byte Reload
 ; X86-NEXT:    psubd %xmm1, %xmm0
@@ -456,7 +840,7 @@ define <4 x i32> @const_sub_add_const_extrause(<4 x i32> %arg) {
 ; X86-NEXT:    .cfi_def_cfa_offset 4
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: const_sub_add_const_extrause:
+; X64-LABEL: vec_const_sub_add_const_extrause:
 ; X64:       # %bb.0:
 ; X64-NEXT:    subq $24, %rsp
 ; X64-NEXT:    .cfi_def_cfa_offset 32
@@ -464,27 +848,27 @@ define <4 x i32> @const_sub_add_const_extrause(<4 x i32> %arg) {
 ; X64-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
 ; X64-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8]
 ; X64-NEXT:    psubd %xmm1, %xmm0
-; X64-NEXT:    callq use
+; X64-NEXT:    callq vec_use
 ; X64-NEXT:    movdqa {{.*#+}} xmm0 = [10,10,10,10]
 ; X64-NEXT:    psubd (%rsp), %xmm0 # 16-byte Folded Reload
 ; X64-NEXT:    addq $24, %rsp
 ; X64-NEXT:    .cfi_def_cfa_offset 8
 ; X64-NEXT:    retq
   %t0 = sub <4 x i32> <i32 8, i32 8, i32 8, i32 8>, %arg
-  call void @use(<4 x i32> %t0)
+  call void @vec_use(<4 x i32> %t0)
   %t1 = add <4 x i32> %t0, <i32 2, i32 2, i32 2, i32 2>
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @const_sub_add_const_nonsplat(<4 x i32> %arg) {
-; X86-LABEL: const_sub_add_const_nonsplat:
+define <4 x i32> @vec_const_sub_add_const_nonsplat(<4 x i32> %arg) {
+; X86-LABEL: vec_const_sub_add_const_nonsplat:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movdqa {{.*#+}} xmm1 = <23,u,u,10>
 ; X86-NEXT:    psubd %xmm0, %xmm1
 ; X86-NEXT:    movdqa %xmm1, %xmm0
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: const_sub_add_const_nonsplat:
+; X64-LABEL: vec_const_sub_add_const_nonsplat:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movdqa {{.*#+}} xmm1 = <23,u,u,10>
 ; X64-NEXT:    psubd %xmm0, %xmm1
@@ -497,15 +881,72 @@ define <4 x i32> @const_sub_add_const_nonsplat(<4 x i32> %arg) {
 
 ; (c1-x)-c2
 
-define <4 x i32> @const_sub_sub_const(<4 x i32> %arg) {
+define i32 @const_sub_sub_const(i32 %arg) {
 ; X86-LABEL: const_sub_sub_const:
 ; X86:       # %bb.0:
+; X86-NEXT:    movl $6, %eax
+; X86-NEXT:    subl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    retl
+;
+; X64-LABEL: const_sub_sub_const:
+; X64:       # %bb.0:
+; X64-NEXT:    movl $6, %eax
+; X64-NEXT:    subl %edi, %eax
+; X64-NEXT:    retq
+  %t0 = sub i32 8, %arg
+  %t1 = sub i32 %t0, 2
+  ret i32 %t1
+}
+
+define i32 @const_sub_sub_const_extrause(i32 %arg) {
+; X86-LABEL: const_sub_sub_const_extrause:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    .cfi_offset %esi, -8
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    movl $8, %eax
+; X86-NEXT:    subl %esi, %eax
+; X86-NEXT:    pushl %eax
+; X86-NEXT:    .cfi_adjust_cfa_offset 4
+; X86-NEXT:    calll use
+; X86-NEXT:    addl $4, %esp
+; X86-NEXT:    .cfi_adjust_cfa_offset -4
+; X86-NEXT:    movl $6, %eax
+; X86-NEXT:    subl %esi, %eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 4
+; X86-NEXT:    retl
+;
+; X64-LABEL: const_sub_sub_const_extrause:
+; X64:       # %bb.0:
+; X64-NEXT:    pushq %rbx
+; X64-NEXT:    .cfi_def_cfa_offset 16
+; X64-NEXT:    .cfi_offset %rbx, -16
+; X64-NEXT:    movl %edi, %ebx
+; X64-NEXT:    movl $8, %edi
+; X64-NEXT:    subl %ebx, %edi
+; X64-NEXT:    callq use
+; X64-NEXT:    movl $6, %eax
+; X64-NEXT:    subl %ebx, %eax
+; X64-NEXT:    popq %rbx
+; X64-NEXT:    .cfi_def_cfa_offset 8
+; X64-NEXT:    retq
+  %t0 = sub i32 8, %arg
+  call void @use(i32 %t0)
+  %t1 = sub i32 %t0, 2
+  ret i32 %t1
+}
+
+define <4 x i32> @vec_const_sub_sub_const(<4 x i32> %arg) {
+; X86-LABEL: vec_const_sub_sub_const:
+; X86:       # %bb.0:
 ; X86-NEXT:    movdqa {{.*#+}} xmm1 = [6,6,6,6]
 ; X86-NEXT:    psubd %xmm0, %xmm1
 ; X86-NEXT:    movdqa %xmm1, %xmm0
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: const_sub_sub_const:
+; X64-LABEL: vec_const_sub_sub_const:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movdqa {{.*#+}} xmm1 = [6,6,6,6]
 ; X64-NEXT:    psubd %xmm0, %xmm1
@@ -516,8 +957,8 @@ define <4 x i32> @const_sub_sub_const(<4 x i32> %arg) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @const_sub_sub_const_extrause(<4 x i32> %arg) {
-; X86-LABEL: const_sub_sub_const_extrause:
+define <4 x i32> @vec_const_sub_sub_const_extrause(<4 x i32> %arg) {
+; X86-LABEL: vec_const_sub_sub_const_extrause:
 ; X86:       # %bb.0:
 ; X86-NEXT:    subl $28, %esp
 ; X86-NEXT:    .cfi_def_cfa_offset 32
@@ -525,7 +966,7 @@ define <4 x i32> @const_sub_sub_const_extrause(<4 x i32> %arg) {
 ; X86-NEXT:    movdqu %xmm0, (%esp) # 16-byte Spill
 ; X86-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8]
 ; X86-NEXT:    psubd %xmm1, %xmm0
-; X86-NEXT:    calll use
+; X86-NEXT:    calll vec_use
 ; X86-NEXT:    movdqa {{.*#+}} xmm0 = [6,6,6,6]
 ; X86-NEXT:    movdqu (%esp), %xmm1 # 16-byte Reload
 ; X86-NEXT:    psubd %xmm1, %xmm0
@@ -533,7 +974,7 @@ define <4 x i32> @const_sub_sub_const_extrause(<4 x i32> %arg) {
 ; X86-NEXT:    .cfi_def_cfa_offset 4
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: const_sub_sub_const_extrause:
+; X64-LABEL: vec_const_sub_sub_const_extrause:
 ; X64:       # %bb.0:
 ; X64-NEXT:    subq $24, %rsp
 ; X64-NEXT:    .cfi_def_cfa_offset 32
@@ -541,27 +982,27 @@ define <4 x i32> @const_sub_sub_const_extrause(<4 x i32> %arg) {
 ; X64-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
 ; X64-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8]
 ; X64-NEXT:    psubd %xmm1, %xmm0
-; X64-NEXT:    callq use
+; X64-NEXT:    callq vec_use
 ; X64-NEXT:    movdqa {{.*#+}} xmm0 = [6,6,6,6]
 ; X64-NEXT:    psubd (%rsp), %xmm0 # 16-byte Folded Reload
 ; X64-NEXT:    addq $24, %rsp
 ; X64-NEXT:    .cfi_def_cfa_offset 8
 ; X64-NEXT:    retq
   %t0 = sub <4 x i32> <i32 8, i32 8, i32 8, i32 8>, %arg
-  call void @use(<4 x i32> %t0)
+  call void @vec_use(<4 x i32> %t0)
   %t1 = sub <4 x i32> %t0, <i32 2, i32 2, i32 2, i32 2>
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @const_sub_sub_const_nonsplat(<4 x i32> %arg) {
-; X86-LABEL: const_sub_sub_const_nonsplat:
+define <4 x i32> @vec_const_sub_sub_const_nonsplat(<4 x i32> %arg) {
+; X86-LABEL: vec_const_sub_sub_const_nonsplat:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movdqa {{.*#+}} xmm1 = <19,u,u,6>
 ; X86-NEXT:    psubd %xmm0, %xmm1
 ; X86-NEXT:    movdqa %xmm1, %xmm0
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: const_sub_sub_const_nonsplat:
+; X64-LABEL: vec_const_sub_sub_const_nonsplat:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movdqa {{.*#+}} xmm1 = <19,u,u,6>
 ; X64-NEXT:    psubd %xmm0, %xmm1
@@ -574,14 +1015,70 @@ define <4 x i32> @const_sub_sub_const_nonsplat(<4 x i32> %arg) {
 
 ; c2-(c1-x)
 
-define <4 x i32> @const_sub_const_sub(<4 x i32> %arg) {
+define i32 @const_sub_const_sub(i32 %arg) {
 ; X86-LABEL: const_sub_const_sub:
 ; X86:       # %bb.0:
-; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm0
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    addl $-6, %eax
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: const_sub_const_sub:
 ; X64:       # %bb.0:
+; X64-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-NEXT:    leal -6(%rdi), %eax
+; X64-NEXT:    retq
+  %t0 = sub i32 8, %arg
+  %t1 = sub i32 2, %t0
+  ret i32 %t1
+}
+
+define i32 @const_sub_const_sub_extrause(i32 %arg) {
+; X86-LABEL: const_sub_const_sub_extrause:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    .cfi_offset %esi, -8
+; X86-NEXT:    movl $8, %esi
+; X86-NEXT:    subl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    .cfi_adjust_cfa_offset 4
+; X86-NEXT:    calll use
+; X86-NEXT:    addl $4, %esp
+; X86-NEXT:    .cfi_adjust_cfa_offset -4
+; X86-NEXT:    movl $2, %eax
+; X86-NEXT:    subl %esi, %eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 4
+; X86-NEXT:    retl
+;
+; X64-LABEL: const_sub_const_sub_extrause:
+; X64:       # %bb.0:
+; X64-NEXT:    pushq %rbx
+; X64-NEXT:    .cfi_def_cfa_offset 16
+; X64-NEXT:    .cfi_offset %rbx, -16
+; X64-NEXT:    movl $8, %ebx
+; X64-NEXT:    subl %edi, %ebx
+; X64-NEXT:    movl %ebx, %edi
+; X64-NEXT:    callq use
+; X64-NEXT:    movl $2, %eax
+; X64-NEXT:    subl %ebx, %eax
+; X64-NEXT:    popq %rbx
+; X64-NEXT:    .cfi_def_cfa_offset 8
+; X64-NEXT:    retq
+  %t0 = sub i32 8, %arg
+  call void @use(i32 %t0)
+  %t1 = sub i32 2, %t0
+  ret i32 %t1
+}
+
+define <4 x i32> @vec_const_sub_const_sub(<4 x i32> %arg) {
+; X86-LABEL: vec_const_sub_const_sub:
+; X86:       # %bb.0:
+; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm0
+; X86-NEXT:    retl
+;
+; X64-LABEL: vec_const_sub_const_sub:
+; X64:       # %bb.0:
 ; X64-NEXT:    paddd {{.*}}(%rip), %xmm0
 ; X64-NEXT:    retq
   %t0 = sub <4 x i32> <i32 8, i32 8, i32 8, i32 8>, %arg
@@ -589,8 +1086,8 @@ define <4 x i32> @const_sub_const_sub(<4 x i32> %arg) {
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @const_sub_const_sub_extrause(<4 x i32> %arg) {
-; X86-LABEL: const_sub_const_sub_extrause:
+define <4 x i32> @vec_const_sub_const_sub_extrause(<4 x i32> %arg) {
+; X86-LABEL: vec_const_sub_const_sub_extrause:
 ; X86:       # %bb.0:
 ; X86-NEXT:    subl $28, %esp
 ; X86-NEXT:    .cfi_def_cfa_offset 32
@@ -598,7 +1095,7 @@ define <4 x i32> @const_sub_const_sub_extrause(<4 x i32> %arg) {
 ; X86-NEXT:    psubd %xmm0, %xmm1
 ; X86-NEXT:    movdqu %xmm1, (%esp) # 16-byte Spill
 ; X86-NEXT:    movdqa %xmm1, %xmm0
-; X86-NEXT:    calll use
+; X86-NEXT:    calll vec_use
 ; X86-NEXT:    movdqa {{.*#+}} xmm0 = [2,2,2,2]
 ; X86-NEXT:    movdqu (%esp), %xmm1 # 16-byte Reload
 ; X86-NEXT:    psubd %xmm1, %xmm0
@@ -606,7 +1103,7 @@ define <4 x i32> @const_sub_const_sub_extrause(<4 x i32> %arg) {
 ; X86-NEXT:    .cfi_def_cfa_offset 4
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: const_sub_const_sub_extrause:
+; X64-LABEL: vec_const_sub_const_sub_extrause:
 ; X64:       # %bb.0:
 ; X64-NEXT:    subq $24, %rsp
 ; X64-NEXT:    .cfi_def_cfa_offset 32
@@ -614,25 +1111,25 @@ define <4 x i32> @const_sub_const_sub_extrause(<4 x i32> %arg) {
 ; X64-NEXT:    psubd %xmm0, %xmm1
 ; X64-NEXT:    movdqa %xmm1, (%rsp) # 16-byte Spill
 ; X64-NEXT:    movdqa %xmm1, %xmm0
-; X64-NEXT:    callq use
+; X64-NEXT:    callq vec_use
 ; X64-NEXT:    movdqa {{.*#+}} xmm0 = [2,2,2,2]
 ; X64-NEXT:    psubd (%rsp), %xmm0 # 16-byte Folded Reload
 ; X64-NEXT:    addq $24, %rsp
 ; X64-NEXT:    .cfi_def_cfa_offset 8
 ; X64-NEXT:    retq
   %t0 = sub <4 x i32> <i32 8, i32 8, i32 8, i32 8>, %arg
-  call void @use(<4 x i32> %t0)
+  call void @vec_use(<4 x i32> %t0)
   %t1 = sub <4 x i32> <i32 2, i32 2, i32 2, i32 2>, %t0
   ret <4 x i32> %t1
 }
 
-define <4 x i32> @const_sub_const_sub_nonsplat(<4 x i32> %arg) {
-; X86-LABEL: const_sub_const_sub_nonsplat:
+define <4 x i32> @vec_const_sub_const_sub_nonsplat(<4 x i32> %arg) {
+; X86-LABEL: vec_const_sub_const_sub_nonsplat:
 ; X86:       # %bb.0:
 ; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm0
 ; X86-NEXT:    retl
 ;
-; X64-LABEL: const_sub_const_sub_nonsplat:
+; X64-LABEL: vec_const_sub_const_sub_nonsplat:
 ; X64:       # %bb.0:
 ; X64-NEXT:    paddd {{.*}}(%rip), %xmm0
 ; X64-NEXT:    retq