-; RUN: llc < %s -march=x86 -mattr=+sse42 -disable-mmx -o %t
-; RUN: grep paddw %t | count 1
-; RUN: grep movd %t | count 2
-; RUN: grep pextrw %t | count 1
+; RUN: llc < %s -mcpu=generic -march=x86 -mattr=+sse4.2 -post-RA-scheduler=true | FileCheck %s
+; CHECK: paddd
; Widen a v3i16 to v8i16 to do a vector add
-target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
-target triple = "i686-apple-darwin10.0.0d2"
@.str = internal constant [4 x i8] c"%d \00" ; <[4 x i8]*> [#uses=1]
@.str1 = internal constant [2 x i8] c"\0A\00" ; <[2 x i8]*> [#uses=1]
br label %forcond
forcond: ; preds = %forinc, %entry
- %tmp = load i32* %i ; <i32> [#uses=1]
- %tmp1 = load i32* %n.addr ; <i32> [#uses=1]
+ %tmp = load i32, i32* %i ; <i32> [#uses=1]
+ %tmp1 = load i32, i32* %n.addr ; <i32> [#uses=1]
%cmp = icmp slt i32 %tmp, %tmp1 ; <i1> [#uses=1]
br i1 %cmp, label %forbody, label %afterfor
forbody: ; preds = %forcond
- %tmp2 = load i32* %i ; <i32> [#uses=1]
- %tmp3 = load <3 x i16>** %dst.addr ; <<3 x i16>*> [#uses=1]
- %arrayidx = getelementptr <3 x i16>* %tmp3, i32 %tmp2 ; <<3 x i16>*> [#uses=1]
- %tmp4 = load i32* %i ; <i32> [#uses=1]
- %tmp5 = load <3 x i16>** %src.addr ; <<3 x i16>*> [#uses=1]
- %arrayidx6 = getelementptr <3 x i16>* %tmp5, i32 %tmp4 ; <<3 x i16>*> [#uses=1]
- %tmp7 = load <3 x i16>* %arrayidx6 ; <<3 x i16>> [#uses=1]
+ %tmp2 = load i32, i32* %i ; <i32> [#uses=1]
+ %tmp3 = load <3 x i16>*, <3 x i16>** %dst.addr ; <<3 x i16>*> [#uses=1]
+ %arrayidx = getelementptr <3 x i16>, <3 x i16>* %tmp3, i32 %tmp2 ; <<3 x i16>*> [#uses=1]
+ %tmp4 = load i32, i32* %i ; <i32> [#uses=1]
+ %tmp5 = load <3 x i16>*, <3 x i16>** %src.addr ; <<3 x i16>*> [#uses=1]
+ %arrayidx6 = getelementptr <3 x i16>, <3 x i16>* %tmp5, i32 %tmp4 ; <<3 x i16>*> [#uses=1]
+ %tmp7 = load <3 x i16>, <3 x i16>* %arrayidx6 ; <<3 x i16>> [#uses=1]
%add = add <3 x i16> %tmp7, < i16 1, i16 1, i16 1 > ; <<3 x i16>> [#uses=1]
store <3 x i16> %add, <3 x i16>* %arrayidx
br label %forinc
forinc: ; preds = %forbody
- %tmp8 = load i32* %i ; <i32> [#uses=1]
+ %tmp8 = load i32, i32* %i ; <i32> [#uses=1]
%inc = add i32 %tmp8, 1 ; <i32> [#uses=1]
store i32 %inc, i32* %i
br label %forcond