mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-10-30 15:32:52 +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
200 lines
6.1 KiB
LLVM
200 lines
6.1 KiB
LLVM
; RUN: llc < %s -march=arm -mattr=+neon,+fp16 | FileCheck %s
|
|
|
|
define <2 x i32> @vcvt_f32tos32(<2 x float>* %A) nounwind {
|
|
;CHECK-LABEL: vcvt_f32tos32:
|
|
;CHECK: vcvt.s32.f32
|
|
%tmp1 = load <2 x float>* %A
|
|
%tmp2 = fptosi <2 x float> %tmp1 to <2 x i32>
|
|
ret <2 x i32> %tmp2
|
|
}
|
|
|
|
define <2 x i32> @vcvt_f32tou32(<2 x float>* %A) nounwind {
|
|
;CHECK-LABEL: vcvt_f32tou32:
|
|
;CHECK: vcvt.u32.f32
|
|
%tmp1 = load <2 x float>* %A
|
|
%tmp2 = fptoui <2 x float> %tmp1 to <2 x i32>
|
|
ret <2 x i32> %tmp2
|
|
}
|
|
|
|
define <2 x float> @vcvt_s32tof32(<2 x i32>* %A) nounwind {
|
|
;CHECK-LABEL: vcvt_s32tof32:
|
|
;CHECK: vcvt.f32.s32
|
|
%tmp1 = load <2 x i32>* %A
|
|
%tmp2 = sitofp <2 x i32> %tmp1 to <2 x float>
|
|
ret <2 x float> %tmp2
|
|
}
|
|
|
|
define <2 x float> @vcvt_u32tof32(<2 x i32>* %A) nounwind {
|
|
;CHECK-LABEL: vcvt_u32tof32:
|
|
;CHECK: vcvt.f32.u32
|
|
%tmp1 = load <2 x i32>* %A
|
|
%tmp2 = uitofp <2 x i32> %tmp1 to <2 x float>
|
|
ret <2 x float> %tmp2
|
|
}
|
|
|
|
define <4 x i32> @vcvtQ_f32tos32(<4 x float>* %A) nounwind {
|
|
;CHECK-LABEL: vcvtQ_f32tos32:
|
|
;CHECK: vcvt.s32.f32
|
|
%tmp1 = load <4 x float>* %A
|
|
%tmp2 = fptosi <4 x float> %tmp1 to <4 x i32>
|
|
ret <4 x i32> %tmp2
|
|
}
|
|
|
|
define <4 x i32> @vcvtQ_f32tou32(<4 x float>* %A) nounwind {
|
|
;CHECK-LABEL: vcvtQ_f32tou32:
|
|
;CHECK: vcvt.u32.f32
|
|
%tmp1 = load <4 x float>* %A
|
|
%tmp2 = fptoui <4 x float> %tmp1 to <4 x i32>
|
|
ret <4 x i32> %tmp2
|
|
}
|
|
|
|
define <4 x float> @vcvtQ_s32tof32(<4 x i32>* %A) nounwind {
|
|
;CHECK-LABEL: vcvtQ_s32tof32:
|
|
;CHECK: vcvt.f32.s32
|
|
%tmp1 = load <4 x i32>* %A
|
|
%tmp2 = sitofp <4 x i32> %tmp1 to <4 x float>
|
|
ret <4 x float> %tmp2
|
|
}
|
|
|
|
define <4 x float> @vcvtQ_u32tof32(<4 x i32>* %A) nounwind {
|
|
;CHECK-LABEL: vcvtQ_u32tof32:
|
|
;CHECK: vcvt.f32.u32
|
|
%tmp1 = load <4 x i32>* %A
|
|
%tmp2 = uitofp <4 x i32> %tmp1 to <4 x float>
|
|
ret <4 x float> %tmp2
|
|
}
|
|
|
|
define <2 x i32> @vcvt_n_f32tos32(<2 x float>* %A) nounwind {
|
|
;CHECK-LABEL: vcvt_n_f32tos32:
|
|
;CHECK: vcvt.s32.f32
|
|
%tmp1 = load <2 x float>* %A
|
|
%tmp2 = call <2 x i32> @llvm.arm.neon.vcvtfp2fxs.v2i32.v2f32(<2 x float> %tmp1, i32 1)
|
|
ret <2 x i32> %tmp2
|
|
}
|
|
|
|
define <2 x i32> @vcvt_n_f32tou32(<2 x float>* %A) nounwind {
|
|
;CHECK-LABEL: vcvt_n_f32tou32:
|
|
;CHECK: vcvt.u32.f32
|
|
%tmp1 = load <2 x float>* %A
|
|
%tmp2 = call <2 x i32> @llvm.arm.neon.vcvtfp2fxu.v2i32.v2f32(<2 x float> %tmp1, i32 1)
|
|
ret <2 x i32> %tmp2
|
|
}
|
|
|
|
define <2 x float> @vcvt_n_s32tof32(<2 x i32>* %A) nounwind {
|
|
;CHECK-LABEL: vcvt_n_s32tof32:
|
|
;CHECK: vcvt.f32.s32
|
|
%tmp1 = load <2 x i32>* %A
|
|
%tmp2 = call <2 x float> @llvm.arm.neon.vcvtfxs2fp.v2f32.v2i32(<2 x i32> %tmp1, i32 1)
|
|
ret <2 x float> %tmp2
|
|
}
|
|
|
|
define <2 x float> @vcvt_n_u32tof32(<2 x i32>* %A) nounwind {
|
|
;CHECK-LABEL: vcvt_n_u32tof32:
|
|
;CHECK: vcvt.f32.u32
|
|
%tmp1 = load <2 x i32>* %A
|
|
%tmp2 = call <2 x float> @llvm.arm.neon.vcvtfxu2fp.v2f32.v2i32(<2 x i32> %tmp1, i32 1)
|
|
ret <2 x float> %tmp2
|
|
}
|
|
|
|
declare <2 x i32> @llvm.arm.neon.vcvtfp2fxs.v2i32.v2f32(<2 x float>, i32) nounwind readnone
|
|
declare <2 x i32> @llvm.arm.neon.vcvtfp2fxu.v2i32.v2f32(<2 x float>, i32) nounwind readnone
|
|
declare <2 x float> @llvm.arm.neon.vcvtfxs2fp.v2f32.v2i32(<2 x i32>, i32) nounwind readnone
|
|
declare <2 x float> @llvm.arm.neon.vcvtfxu2fp.v2f32.v2i32(<2 x i32>, i32) nounwind readnone
|
|
|
|
define <4 x i32> @vcvtQ_n_f32tos32(<4 x float>* %A) nounwind {
|
|
;CHECK-LABEL: vcvtQ_n_f32tos32:
|
|
;CHECK: vcvt.s32.f32
|
|
%tmp1 = load <4 x float>* %A
|
|
%tmp2 = call <4 x i32> @llvm.arm.neon.vcvtfp2fxs.v4i32.v4f32(<4 x float> %tmp1, i32 1)
|
|
ret <4 x i32> %tmp2
|
|
}
|
|
|
|
define <4 x i32> @vcvtQ_n_f32tou32(<4 x float>* %A) nounwind {
|
|
;CHECK-LABEL: vcvtQ_n_f32tou32:
|
|
;CHECK: vcvt.u32.f32
|
|
%tmp1 = load <4 x float>* %A
|
|
%tmp2 = call <4 x i32> @llvm.arm.neon.vcvtfp2fxu.v4i32.v4f32(<4 x float> %tmp1, i32 1)
|
|
ret <4 x i32> %tmp2
|
|
}
|
|
|
|
define <4 x float> @vcvtQ_n_s32tof32(<4 x i32>* %A) nounwind {
|
|
;CHECK-LABEL: vcvtQ_n_s32tof32:
|
|
;CHECK: vcvt.f32.s32
|
|
%tmp1 = load <4 x i32>* %A
|
|
%tmp2 = call <4 x float> @llvm.arm.neon.vcvtfxs2fp.v4f32.v4i32(<4 x i32> %tmp1, i32 1)
|
|
ret <4 x float> %tmp2
|
|
}
|
|
|
|
define <4 x float> @vcvtQ_n_u32tof32(<4 x i32>* %A) nounwind {
|
|
;CHECK-LABEL: vcvtQ_n_u32tof32:
|
|
;CHECK: vcvt.f32.u32
|
|
%tmp1 = load <4 x i32>* %A
|
|
%tmp2 = call <4 x float> @llvm.arm.neon.vcvtfxu2fp.v4f32.v4i32(<4 x i32> %tmp1, i32 1)
|
|
ret <4 x float> %tmp2
|
|
}
|
|
|
|
declare <4 x i32> @llvm.arm.neon.vcvtfp2fxs.v4i32.v4f32(<4 x float>, i32) nounwind readnone
|
|
declare <4 x i32> @llvm.arm.neon.vcvtfp2fxu.v4i32.v4f32(<4 x float>, i32) nounwind readnone
|
|
declare <4 x float> @llvm.arm.neon.vcvtfxs2fp.v4f32.v4i32(<4 x i32>, i32) nounwind readnone
|
|
declare <4 x float> @llvm.arm.neon.vcvtfxu2fp.v4f32.v4i32(<4 x i32>, i32) nounwind readnone
|
|
|
|
define <4 x float> @vcvt_f16tof32(<4 x i16>* %A) nounwind {
|
|
;CHECK-LABEL: vcvt_f16tof32:
|
|
;CHECK: vcvt.f32.f16
|
|
%tmp1 = load <4 x i16>* %A
|
|
%tmp2 = call <4 x float> @llvm.arm.neon.vcvthf2fp(<4 x i16> %tmp1)
|
|
ret <4 x float> %tmp2
|
|
}
|
|
|
|
define <4 x i16> @vcvt_f32tof16(<4 x float>* %A) nounwind {
|
|
;CHECK-LABEL: vcvt_f32tof16:
|
|
;CHECK: vcvt.f16.f32
|
|
%tmp1 = load <4 x float>* %A
|
|
%tmp2 = call <4 x i16> @llvm.arm.neon.vcvtfp2hf(<4 x float> %tmp1)
|
|
ret <4 x i16> %tmp2
|
|
}
|
|
|
|
declare <4 x float> @llvm.arm.neon.vcvthf2fp(<4 x i16>) nounwind readnone
|
|
declare <4 x i16> @llvm.arm.neon.vcvtfp2hf(<4 x float>) nounwind readnone
|
|
|
|
|
|
define <4 x i16> @fix_float_to_i16(<4 x float> %in) {
|
|
; CHECK-LABEL: fix_float_to_i16:
|
|
; CHECK: vcvt.u32.f32 [[TMP:q[0-9]+]], {{q[0-9]+}}, #1
|
|
; CHECK: vmovn.i32 {{d[0-9]+}}, [[TMP]]
|
|
|
|
%scale = fmul <4 x float> %in, <float 2.0, float 2.0, float 2.0, float 2.0>
|
|
%conv = fptoui <4 x float> %scale to <4 x i16>
|
|
ret <4 x i16> %conv
|
|
}
|
|
|
|
define <2 x i64> @fix_float_to_i64(<2 x float> %in) {
|
|
; CHECK-LABEL: fix_float_to_i64:
|
|
; CHECK: bl
|
|
; CHECK: bl
|
|
|
|
%scale = fmul <2 x float> %in, <float 2.0, float 2.0>
|
|
%conv = fptoui <2 x float> %scale to <2 x i64>
|
|
ret <2 x i64> %conv
|
|
}
|
|
|
|
define <4 x i16> @fix_double_to_i16(<4 x double> %in) {
|
|
; CHECK-LABEL: fix_double_to_i16:
|
|
; CHECK: vcvt.s32.f64
|
|
; CHECK: vcvt.s32.f64
|
|
|
|
%scale = fmul <4 x double> %in, <double 2.0, double 2.0, double 2.0, double 2.0>
|
|
%conv = fptoui <4 x double> %scale to <4 x i16>
|
|
ret <4 x i16> %conv
|
|
}
|
|
|
|
define <2 x i64> @fix_double_to_i64(<2 x double> %in) {
|
|
; CHECK-LABEL: fix_double_to_i64:
|
|
; CHECK: bl
|
|
; CHECK: bl
|
|
%scale = fmul <2 x double> %in, <double 2.0, double 2.0>
|
|
%conv = fptoui <2 x double> %scale to <2 x i64>
|
|
ret <2 x i64> %conv
|
|
}
|
|
|