| Nadav Rotem | d88bc2a | 2011-10-18 23:05:33 +0000 | [diff] [blame^] | 1 | ; RUN: llc < %s -march=x86-64 -mcpu=corei7 | FileCheck %s |
| 2 | |
| 3 | target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i8:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128" |
| 4 | target triple = "x86_64-unknown-linux-gnu" |
| 5 | |
| 6 | |
| 7 | ; CHECK: mul_f |
| 8 | define i32 @mul_f(<4 x i8>* %A) { |
| 9 | entry: |
| 10 | ; CHECK: pmul |
| 11 | ; CHECK-NOT: mulb |
| 12 | %0 = load <4 x i8>* %A, align 8 |
| 13 | %mul = mul <4 x i8> %0, %0 |
| 14 | store <4 x i8> %mul, <4 x i8>* undef |
| 15 | ret i32 0 |
| 16 | ; CHECK: ret |
| 17 | } |
| 18 | |
| 19 | |
| 20 | ; CHECK: shuff_f |
| 21 | define i32 @shuff_f(<4 x i8>* %A) { |
| 22 | entry: |
| 23 | ; CHECK: pshufb |
| 24 | ; CHECK: paddd |
| 25 | ; CHECK: pshufb |
| 26 | %0 = load <4 x i8>* %A, align 8 |
| 27 | %add = add <4 x i8> %0, %0 |
| 28 | store <4 x i8> %add, <4 x i8>* undef |
| 29 | ret i32 0 |
| 30 | ; CHECK: ret |
| 31 | } |