aboutsummaryrefslogtreecommitdiffstats
path: root/test/CodeGen/R600/llvm.AMDGPU.rcp.f64.ll
diff options
context:
space:
mode:
authorStephen Hines <srhines@google.com>2014-12-01 14:51:49 -0800
committerStephen Hines <srhines@google.com>2014-12-02 16:08:10 -0800
commit37ed9c199ca639565f6ce88105f9e39e898d82d0 (patch)
tree8fb36d3910e3ee4c4e1b7422f4f017108efc52f5 /test/CodeGen/R600/llvm.AMDGPU.rcp.f64.ll
parentd2327b22152ced7bc46dc629fc908959e8a52d03 (diff)
downloadexternal_llvm-37ed9c199ca639565f6ce88105f9e39e898d82d0.zip
external_llvm-37ed9c199ca639565f6ce88105f9e39e898d82d0.tar.gz
external_llvm-37ed9c199ca639565f6ce88105f9e39e898d82d0.tar.bz2
Update aosp/master LLVM for rebase to r222494.
Change-Id: Ic787f5e0124df789bd26f3f24680f45e678eef2d
Diffstat (limited to 'test/CodeGen/R600/llvm.AMDGPU.rcp.f64.ll')
-rw-r--r--test/CodeGen/R600/llvm.AMDGPU.rcp.f64.ll30
1 files changed, 30 insertions, 0 deletions
diff --git a/test/CodeGen/R600/llvm.AMDGPU.rcp.f64.ll b/test/CodeGen/R600/llvm.AMDGPU.rcp.f64.ll
new file mode 100644
index 0000000..c4b04c5
--- /dev/null
+++ b/test/CodeGen/R600/llvm.AMDGPU.rcp.f64.ll
@@ -0,0 +1,30 @@
+; RUN: llc -march=r600 -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+
+declare double @llvm.AMDGPU.rcp.f64(double) nounwind readnone
+declare double @llvm.sqrt.f64(double) nounwind readnone
+
+; FUNC-LABEL: {{^}}rcp_f64:
+; SI: v_rcp_f64_e32
+define void @rcp_f64(double addrspace(1)* %out, double %src) nounwind {
+ %rcp = call double @llvm.AMDGPU.rcp.f64(double %src) nounwind readnone
+ store double %rcp, double addrspace(1)* %out, align 8
+ ret void
+}
+
+; FUNC-LABEL: {{^}}rcp_pat_f64:
+; SI: v_rcp_f64_e32
+define void @rcp_pat_f64(double addrspace(1)* %out, double %src) nounwind {
+ %rcp = fdiv double 1.0, %src
+ store double %rcp, double addrspace(1)* %out, align 8
+ ret void
+}
+
+; FUNC-LABEL: {{^}}rsq_rcp_pat_f64:
+; SI-UNSAFE: v_rsq_f64_e32
+; SI-SAFE-NOT: v_rsq_f64_e32
+define void @rsq_rcp_pat_f64(double addrspace(1)* %out, double %src) nounwind {
+ %sqrt = call double @llvm.sqrt.f64(double %src) nounwind readnone
+ %rcp = call double @llvm.AMDGPU.rcp.f64(double %sqrt) nounwind readnone
+ store double %rcp, double addrspace(1)* %out, align 8
+ ret void
+}