2009-09-09 02:09:15 +02:00
|
|
|
; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s
|
2009-08-06 20:47:44 +02:00
|
|
|
|
|
|
|
define void @vst4i8(i8* %A, <8 x i8>* %B) nounwind {
|
2013-07-14 08:24:09 +02:00
|
|
|
;CHECK-LABEL: vst4i8:
|
2010-09-24 01:42:37 +02:00
|
|
|
;Check the alignment value. Max for this instruction is 256 bits:
|
2013-02-22 11:01:33 +01:00
|
|
|
;CHECK: vst4.8 {d16, d17, d18, d19}, [r0:64]
|
2009-08-06 20:47:44 +02:00
|
|
|
%tmp1 = load <8 x i8>* %B
|
2010-09-24 01:42:37 +02:00
|
|
|
call void @llvm.arm.neon.vst4.v8i8(i8* %A, <8 x i8> %tmp1, <8 x i8> %tmp1, <8 x i8> %tmp1, <8 x i8> %tmp1, i32 8)
|
2009-08-06 20:47:44 +02:00
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
2011-02-07 18:43:21 +01:00
|
|
|
;Check for a post-increment updating store with register increment.
|
|
|
|
define void @vst4i8_update(i8** %ptr, <8 x i8>* %B, i32 %inc) nounwind {
|
2013-07-14 08:24:09 +02:00
|
|
|
;CHECK-LABEL: vst4i8_update:
|
2013-02-22 11:01:33 +01:00
|
|
|
;CHECK: vst4.8 {d16, d17, d18, d19}, [r1:128], r2
|
2011-02-07 18:43:21 +01:00
|
|
|
%A = load i8** %ptr
|
|
|
|
%tmp1 = load <8 x i8>* %B
|
|
|
|
call void @llvm.arm.neon.vst4.v8i8(i8* %A, <8 x i8> %tmp1, <8 x i8> %tmp1, <8 x i8> %tmp1, <8 x i8> %tmp1, i32 16)
|
|
|
|
%tmp2 = getelementptr i8* %A, i32 %inc
|
|
|
|
store i8* %tmp2, i8** %ptr
|
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
2009-08-06 20:47:44 +02:00
|
|
|
define void @vst4i16(i16* %A, <4 x i16>* %B) nounwind {
|
2013-07-14 08:24:09 +02:00
|
|
|
;CHECK-LABEL: vst4i16:
|
2010-09-24 01:42:37 +02:00
|
|
|
;Check the alignment value. Max for this instruction is 256 bits:
|
2013-02-22 11:01:33 +01:00
|
|
|
;CHECK: vst4.16 {d16, d17, d18, d19}, [r0:128]
|
2010-04-20 02:17:16 +02:00
|
|
|
%tmp0 = bitcast i16* %A to i8*
|
2009-08-06 20:47:44 +02:00
|
|
|
%tmp1 = load <4 x i16>* %B
|
2010-09-24 01:42:37 +02:00
|
|
|
call void @llvm.arm.neon.vst4.v4i16(i8* %tmp0, <4 x i16> %tmp1, <4 x i16> %tmp1, <4 x i16> %tmp1, <4 x i16> %tmp1, i32 16)
|
2009-08-06 20:47:44 +02:00
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
|
|
|
define void @vst4i32(i32* %A, <2 x i32>* %B) nounwind {
|
2013-07-14 08:24:09 +02:00
|
|
|
;CHECK-LABEL: vst4i32:
|
2010-09-24 01:42:37 +02:00
|
|
|
;Check the alignment value. Max for this instruction is 256 bits:
|
2013-02-22 11:01:33 +01:00
|
|
|
;CHECK: vst4.32 {d16, d17, d18, d19}, [r0:256]
|
2010-04-20 02:17:16 +02:00
|
|
|
%tmp0 = bitcast i32* %A to i8*
|
2009-08-06 20:47:44 +02:00
|
|
|
%tmp1 = load <2 x i32>* %B
|
2010-09-24 01:42:37 +02:00
|
|
|
call void @llvm.arm.neon.vst4.v2i32(i8* %tmp0, <2 x i32> %tmp1, <2 x i32> %tmp1, <2 x i32> %tmp1, <2 x i32> %tmp1, i32 32)
|
2009-08-06 20:47:44 +02:00
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
|
|
|
define void @vst4f(float* %A, <2 x float>* %B) nounwind {
|
2013-07-14 08:24:09 +02:00
|
|
|
;CHECK-LABEL: vst4f:
|
2009-08-06 20:47:44 +02:00
|
|
|
;CHECK: vst4.32
|
2010-04-20 02:17:16 +02:00
|
|
|
%tmp0 = bitcast float* %A to i8*
|
2009-08-06 20:47:44 +02:00
|
|
|
%tmp1 = load <2 x float>* %B
|
2010-08-27 19:13:24 +02:00
|
|
|
call void @llvm.arm.neon.vst4.v2f32(i8* %tmp0, <2 x float> %tmp1, <2 x float> %tmp1, <2 x float> %tmp1, <2 x float> %tmp1, i32 1)
|
2009-08-06 20:47:44 +02:00
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
2009-10-08 07:18:18 +02:00
|
|
|
define void @vst4i64(i64* %A, <1 x i64>* %B) nounwind {
|
2013-07-14 08:24:09 +02:00
|
|
|
;CHECK-LABEL: vst4i64:
|
2010-09-24 01:42:37 +02:00
|
|
|
;Check the alignment value. Max for this instruction is 256 bits:
|
2013-02-22 11:01:33 +01:00
|
|
|
;CHECK: vst1.64 {d16, d17, d18, d19}, [r0:256]
|
2010-04-20 02:17:16 +02:00
|
|
|
%tmp0 = bitcast i64* %A to i8*
|
2009-10-08 07:18:18 +02:00
|
|
|
%tmp1 = load <1 x i64>* %B
|
2010-09-24 01:42:37 +02:00
|
|
|
call void @llvm.arm.neon.vst4.v1i64(i8* %tmp0, <1 x i64> %tmp1, <1 x i64> %tmp1, <1 x i64> %tmp1, <1 x i64> %tmp1, i32 64)
|
2009-10-08 07:18:18 +02:00
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
2009-10-07 22:49:18 +02:00
|
|
|
define void @vst4Qi8(i8* %A, <16 x i8>* %B) nounwind {
|
2013-07-14 08:24:09 +02:00
|
|
|
;CHECK-LABEL: vst4Qi8:
|
2010-09-24 01:42:37 +02:00
|
|
|
;Check the alignment value. Max for this instruction is 256 bits:
|
2013-02-22 11:01:33 +01:00
|
|
|
;CHECK: vst4.8 {d16, d18, d20, d22}, [r0:256]!
|
|
|
|
;CHECK: vst4.8 {d17, d19, d21, d23}, [r0:256]
|
2009-10-07 22:49:18 +02:00
|
|
|
%tmp1 = load <16 x i8>* %B
|
2010-09-24 01:42:37 +02:00
|
|
|
call void @llvm.arm.neon.vst4.v16i8(i8* %A, <16 x i8> %tmp1, <16 x i8> %tmp1, <16 x i8> %tmp1, <16 x i8> %tmp1, i32 64)
|
2009-10-07 22:49:18 +02:00
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
|
|
|
define void @vst4Qi16(i16* %A, <8 x i16>* %B) nounwind {
|
2013-07-14 08:24:09 +02:00
|
|
|
;CHECK-LABEL: vst4Qi16:
|
2010-09-24 01:42:37 +02:00
|
|
|
;Check for no alignment specifier.
|
2010-10-08 08:15:13 +02:00
|
|
|
;CHECK: vst4.16 {d16, d18, d20, d22}, [r0]!
|
|
|
|
;CHECK: vst4.16 {d17, d19, d21, d23}, [r0]
|
2010-04-20 02:17:16 +02:00
|
|
|
%tmp0 = bitcast i16* %A to i8*
|
2009-10-07 22:49:18 +02:00
|
|
|
%tmp1 = load <8 x i16>* %B
|
2010-08-27 19:13:24 +02:00
|
|
|
call void @llvm.arm.neon.vst4.v8i16(i8* %tmp0, <8 x i16> %tmp1, <8 x i16> %tmp1, <8 x i16> %tmp1, <8 x i16> %tmp1, i32 1)
|
2009-10-07 22:49:18 +02:00
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
|
|
|
define void @vst4Qi32(i32* %A, <4 x i32>* %B) nounwind {
|
2013-07-14 08:24:09 +02:00
|
|
|
;CHECK-LABEL: vst4Qi32:
|
2009-10-07 22:49:18 +02:00
|
|
|
;CHECK: vst4.32
|
|
|
|
;CHECK: vst4.32
|
2010-04-20 02:17:16 +02:00
|
|
|
%tmp0 = bitcast i32* %A to i8*
|
2009-10-07 22:49:18 +02:00
|
|
|
%tmp1 = load <4 x i32>* %B
|
2010-08-27 19:13:24 +02:00
|
|
|
call void @llvm.arm.neon.vst4.v4i32(i8* %tmp0, <4 x i32> %tmp1, <4 x i32> %tmp1, <4 x i32> %tmp1, <4 x i32> %tmp1, i32 1)
|
2009-10-07 22:49:18 +02:00
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
|
|
|
define void @vst4Qf(float* %A, <4 x float>* %B) nounwind {
|
2013-07-14 08:24:09 +02:00
|
|
|
;CHECK-LABEL: vst4Qf:
|
2009-10-07 22:49:18 +02:00
|
|
|
;CHECK: vst4.32
|
|
|
|
;CHECK: vst4.32
|
2010-04-20 02:17:16 +02:00
|
|
|
%tmp0 = bitcast float* %A to i8*
|
2009-10-07 22:49:18 +02:00
|
|
|
%tmp1 = load <4 x float>* %B
|
2010-08-27 19:13:24 +02:00
|
|
|
call void @llvm.arm.neon.vst4.v4f32(i8* %tmp0, <4 x float> %tmp1, <4 x float> %tmp1, <4 x float> %tmp1, <4 x float> %tmp1, i32 1)
|
2009-10-07 22:49:18 +02:00
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
2011-02-07 18:43:21 +01:00
|
|
|
;Check for a post-increment updating store.
|
|
|
|
define void @vst4Qf_update(float** %ptr, <4 x float>* %B) nounwind {
|
2013-07-14 08:24:09 +02:00
|
|
|
;CHECK-LABEL: vst4Qf_update:
|
2011-02-07 18:43:21 +01:00
|
|
|
;CHECK: vst4.32 {d16, d18, d20, d22}, [r1]!
|
|
|
|
;CHECK: vst4.32 {d17, d19, d21, d23}, [r1]!
|
|
|
|
%A = load float** %ptr
|
|
|
|
%tmp0 = bitcast float* %A to i8*
|
|
|
|
%tmp1 = load <4 x float>* %B
|
|
|
|
call void @llvm.arm.neon.vst4.v4f32(i8* %tmp0, <4 x float> %tmp1, <4 x float> %tmp1, <4 x float> %tmp1, <4 x float> %tmp1, i32 1)
|
|
|
|
%tmp2 = getelementptr float* %A, i32 16
|
|
|
|
store float* %tmp2, float** %ptr
|
|
|
|
ret void
|
|
|
|
}
|
|
|
|
|
2010-08-27 19:13:24 +02:00
|
|
|
declare void @llvm.arm.neon.vst4.v8i8(i8*, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, i32) nounwind
|
|
|
|
declare void @llvm.arm.neon.vst4.v4i16(i8*, <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16>, i32) nounwind
|
|
|
|
declare void @llvm.arm.neon.vst4.v2i32(i8*, <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, i32) nounwind
|
|
|
|
declare void @llvm.arm.neon.vst4.v2f32(i8*, <2 x float>, <2 x float>, <2 x float>, <2 x float>, i32) nounwind
|
|
|
|
declare void @llvm.arm.neon.vst4.v1i64(i8*, <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64>, i32) nounwind
|
2009-10-07 22:49:18 +02:00
|
|
|
|
2010-08-27 19:13:24 +02:00
|
|
|
declare void @llvm.arm.neon.vst4.v16i8(i8*, <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8>, i32) nounwind
|
|
|
|
declare void @llvm.arm.neon.vst4.v8i16(i8*, <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16>, i32) nounwind
|
|
|
|
declare void @llvm.arm.neon.vst4.v4i32(i8*, <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32>, i32) nounwind
|
|
|
|
declare void @llvm.arm.neon.vst4.v4f32(i8*, <4 x float>, <4 x float>, <4 x float>, <4 x float>, i32) nounwind
|