| Matt Arsenault | e93d06a | 2015-01-13 20:53:18 +0000 | [diff] [blame] | 1 | ; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck %s |
| 2 | ; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck %s |
| Marek Olsak | 7517077 | 2015-01-27 17:27:15 +0000 | [diff] [blame] | 3 | ; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck %s |
| 4 | ; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck %s |
| Matt Arsenault | e93d06a | 2015-01-13 20:53:18 +0000 | [diff] [blame] | 5 | |
| 6 | ; Run with unsafe-fp-math to make sure nothing tries to turn this into 1 / rsqrt(x) |
| Tom Stellard | 8ed7b45 | 2013-07-12 18:15:13 +0000 | [diff] [blame] | 7 | |
| Tom Stellard | 79243d9 | 2014-10-01 17:15:17 +0000 | [diff] [blame] | 8 | ; CHECK: {{^}}fsqrt_f32: |
| Tom Stellard | 326d6ec | 2014-11-05 14:50:53 +0000 | [diff] [blame] | 9 | ; CHECK: v_sqrt_f32_e32 {{v[0-9]+, v[0-9]+}} |
| Tom Stellard | 8ed7b45 | 2013-07-12 18:15:13 +0000 | [diff] [blame] | 10 | |
| 11 | define void @fsqrt_f32(float addrspace(1)* %out, float addrspace(1)* %in) { |
| 12 | %r0 = load float addrspace(1)* %in |
| 13 | %r1 = call float @llvm.sqrt.f32(float %r0) |
| 14 | store float %r1, float addrspace(1)* %out |
| 15 | ret void |
| 16 | } |
| 17 | |
| Tom Stellard | 79243d9 | 2014-10-01 17:15:17 +0000 | [diff] [blame] | 18 | ; CHECK: {{^}}fsqrt_f64: |
| Tom Stellard | 326d6ec | 2014-11-05 14:50:53 +0000 | [diff] [blame] | 19 | ; CHECK: v_sqrt_f64_e32 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\]}} |
| Tom Stellard | 8ed7b45 | 2013-07-12 18:15:13 +0000 | [diff] [blame] | 20 | |
| 21 | define void @fsqrt_f64(double addrspace(1)* %out, double addrspace(1)* %in) { |
| 22 | %r0 = load double addrspace(1)* %in |
| 23 | %r1 = call double @llvm.sqrt.f64(double %r0) |
| 24 | store double %r1, double addrspace(1)* %out |
| 25 | ret void |
| 26 | } |
| 27 | |
| 28 | declare float @llvm.sqrt.f32(float %Val) |
| 29 | declare double @llvm.sqrt.f64(double %Val) |