mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-01 08:23:21 +01:00
7e501cf4c3
This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
155 lines
4.9 KiB
LLVM
155 lines
4.9 KiB
LLVM
; Test 32-bit square root.
|
|
;
|
|
; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
|
|
|
|
declare float @llvm.sqrt.f32(float %f)
|
|
|
|
; Check register square root.
|
|
define float @f1(float %val) {
|
|
; CHECK-LABEL: f1:
|
|
; CHECK: sqebr %f0, %f0
|
|
; CHECK: br %r14
|
|
%res = call float @llvm.sqrt.f32(float %val)
|
|
ret float %res
|
|
}
|
|
|
|
; Check the low end of the SQEB range.
|
|
define float @f2(float *%ptr) {
|
|
; CHECK-LABEL: f2:
|
|
; CHECK: sqeb %f0, 0(%r2)
|
|
; CHECK: br %r14
|
|
%val = load float *%ptr
|
|
%res = call float @llvm.sqrt.f32(float %val)
|
|
ret float %res
|
|
}
|
|
|
|
; Check the high end of the aligned SQEB range.
|
|
define float @f3(float *%base) {
|
|
; CHECK-LABEL: f3:
|
|
; CHECK: sqeb %f0, 4092(%r2)
|
|
; CHECK: br %r14
|
|
%ptr = getelementptr float *%base, i64 1023
|
|
%val = load float *%ptr
|
|
%res = call float @llvm.sqrt.f32(float %val)
|
|
ret float %res
|
|
}
|
|
|
|
; Check the next word up, which needs separate address logic.
|
|
; Other sequences besides this one would be OK.
|
|
define float @f4(float *%base) {
|
|
; CHECK-LABEL: f4:
|
|
; CHECK: aghi %r2, 4096
|
|
; CHECK: sqeb %f0, 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr = getelementptr float *%base, i64 1024
|
|
%val = load float *%ptr
|
|
%res = call float @llvm.sqrt.f32(float %val)
|
|
ret float %res
|
|
}
|
|
|
|
; Check negative displacements, which also need separate address logic.
|
|
define float @f5(float *%base) {
|
|
; CHECK-LABEL: f5:
|
|
; CHECK: aghi %r2, -4
|
|
; CHECK: sqeb %f0, 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr = getelementptr float *%base, i64 -1
|
|
%val = load float *%ptr
|
|
%res = call float @llvm.sqrt.f32(float %val)
|
|
ret float %res
|
|
}
|
|
|
|
; Check that SQEB allows indices.
|
|
define float @f6(float *%base, i64 %index) {
|
|
; CHECK-LABEL: f6:
|
|
; CHECK: sllg %r1, %r3, 2
|
|
; CHECK: sqeb %f0, 400(%r1,%r2)
|
|
; CHECK: br %r14
|
|
%ptr1 = getelementptr float *%base, i64 %index
|
|
%ptr2 = getelementptr float *%ptr1, i64 100
|
|
%val = load float *%ptr2
|
|
%res = call float @llvm.sqrt.f32(float %val)
|
|
ret float %res
|
|
}
|
|
|
|
; Test a case where we spill the source of at least one SQEBR. We want
|
|
; to use SQEB if possible.
|
|
define void @f7(float *%ptr) {
|
|
; CHECK-LABEL: f7:
|
|
; CHECK: sqeb {{%f[0-9]+}}, 16{{[04]}}(%r15)
|
|
; CHECK: br %r14
|
|
%val0 = load volatile float *%ptr
|
|
%val1 = load volatile float *%ptr
|
|
%val2 = load volatile float *%ptr
|
|
%val3 = load volatile float *%ptr
|
|
%val4 = load volatile float *%ptr
|
|
%val5 = load volatile float *%ptr
|
|
%val6 = load volatile float *%ptr
|
|
%val7 = load volatile float *%ptr
|
|
%val8 = load volatile float *%ptr
|
|
%val9 = load volatile float *%ptr
|
|
%val10 = load volatile float *%ptr
|
|
%val11 = load volatile float *%ptr
|
|
%val12 = load volatile float *%ptr
|
|
%val13 = load volatile float *%ptr
|
|
%val14 = load volatile float *%ptr
|
|
%val15 = load volatile float *%ptr
|
|
%val16 = load volatile float *%ptr
|
|
|
|
%sqrt0 = call float @llvm.sqrt.f32(float %val0)
|
|
%sqrt1 = call float @llvm.sqrt.f32(float %val1)
|
|
%sqrt2 = call float @llvm.sqrt.f32(float %val2)
|
|
%sqrt3 = call float @llvm.sqrt.f32(float %val3)
|
|
%sqrt4 = call float @llvm.sqrt.f32(float %val4)
|
|
%sqrt5 = call float @llvm.sqrt.f32(float %val5)
|
|
%sqrt6 = call float @llvm.sqrt.f32(float %val6)
|
|
%sqrt7 = call float @llvm.sqrt.f32(float %val7)
|
|
%sqrt8 = call float @llvm.sqrt.f32(float %val8)
|
|
%sqrt9 = call float @llvm.sqrt.f32(float %val9)
|
|
%sqrt10 = call float @llvm.sqrt.f32(float %val10)
|
|
%sqrt11 = call float @llvm.sqrt.f32(float %val11)
|
|
%sqrt12 = call float @llvm.sqrt.f32(float %val12)
|
|
%sqrt13 = call float @llvm.sqrt.f32(float %val13)
|
|
%sqrt14 = call float @llvm.sqrt.f32(float %val14)
|
|
%sqrt15 = call float @llvm.sqrt.f32(float %val15)
|
|
%sqrt16 = call float @llvm.sqrt.f32(float %val16)
|
|
|
|
store volatile float %val0, float *%ptr
|
|
store volatile float %val1, float *%ptr
|
|
store volatile float %val2, float *%ptr
|
|
store volatile float %val3, float *%ptr
|
|
store volatile float %val4, float *%ptr
|
|
store volatile float %val5, float *%ptr
|
|
store volatile float %val6, float *%ptr
|
|
store volatile float %val7, float *%ptr
|
|
store volatile float %val8, float *%ptr
|
|
store volatile float %val9, float *%ptr
|
|
store volatile float %val10, float *%ptr
|
|
store volatile float %val11, float *%ptr
|
|
store volatile float %val12, float *%ptr
|
|
store volatile float %val13, float *%ptr
|
|
store volatile float %val14, float *%ptr
|
|
store volatile float %val15, float *%ptr
|
|
store volatile float %val16, float *%ptr
|
|
|
|
store volatile float %sqrt0, float *%ptr
|
|
store volatile float %sqrt1, float *%ptr
|
|
store volatile float %sqrt2, float *%ptr
|
|
store volatile float %sqrt3, float *%ptr
|
|
store volatile float %sqrt4, float *%ptr
|
|
store volatile float %sqrt5, float *%ptr
|
|
store volatile float %sqrt6, float *%ptr
|
|
store volatile float %sqrt7, float *%ptr
|
|
store volatile float %sqrt8, float *%ptr
|
|
store volatile float %sqrt9, float *%ptr
|
|
store volatile float %sqrt10, float *%ptr
|
|
store volatile float %sqrt11, float *%ptr
|
|
store volatile float %sqrt12, float *%ptr
|
|
store volatile float %sqrt13, float *%ptr
|
|
store volatile float %sqrt14, float *%ptr
|
|
store volatile float %sqrt15, float *%ptr
|
|
store volatile float %sqrt16, float *%ptr
|
|
|
|
ret void
|
|
}
|