mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-25 04:02:41 +01:00
96 lines
3.0 KiB
LLVM
96 lines
3.0 KiB
LLVM
; RUN: llvm-as < %s | llc -march=arm -mattr=+neon > %t
|
|
; RUN: grep {vceq\\.i8} %t | count 2
|
|
; RUN: grep {vceq\\.i16} %t | count 2
|
|
; RUN: grep {vceq\\.i32} %t | count 2
|
|
; RUN: grep vmvn %t | count 6
|
|
; RUN: grep {vcgt\\.s8} %t | count 1
|
|
; RUN: grep {vcge\\.s16} %t | count 1
|
|
; RUN: grep {vcgt\\.u16} %t | count 1
|
|
; RUN: grep {vcge\\.u32} %t | count 1
|
|
|
|
; This tests icmp operations that do not map directly to NEON instructions.
|
|
; Not-equal (ne) operations are implemented by VCEQ/VMVN. Less-than (lt/ult)
|
|
; and less-than-or-equal (le/ule) are implemented by swapping the arguments
|
|
; to VCGT and VCGE. Test all the operand types for not-equal but only sample
|
|
; the other operations.
|
|
|
|
define <8 x i8> @vcnei8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
|
|
%tmp1 = load <8 x i8>* %A
|
|
%tmp2 = load <8 x i8>* %B
|
|
%tmp3 = icmp ne <8 x i8> %tmp1, %tmp2
|
|
%tmp4 = sext <8 x i1> %tmp3 to <8 x i8>
|
|
ret <8 x i8> %tmp4
|
|
}
|
|
|
|
define <4 x i16> @vcnei16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
|
|
%tmp1 = load <4 x i16>* %A
|
|
%tmp2 = load <4 x i16>* %B
|
|
%tmp3 = icmp ne <4 x i16> %tmp1, %tmp2
|
|
%tmp4 = sext <4 x i1> %tmp3 to <4 x i16>
|
|
ret <4 x i16> %tmp4
|
|
}
|
|
|
|
define <2 x i32> @vcnei32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
|
|
%tmp1 = load <2 x i32>* %A
|
|
%tmp2 = load <2 x i32>* %B
|
|
%tmp3 = icmp ne <2 x i32> %tmp1, %tmp2
|
|
%tmp4 = sext <2 x i1> %tmp3 to <2 x i32>
|
|
ret <2 x i32> %tmp4
|
|
}
|
|
|
|
define <16 x i8> @vcneQi8(<16 x i8>* %A, <16 x i8>* %B) nounwind {
|
|
%tmp1 = load <16 x i8>* %A
|
|
%tmp2 = load <16 x i8>* %B
|
|
%tmp3 = icmp ne <16 x i8> %tmp1, %tmp2
|
|
%tmp4 = sext <16 x i1> %tmp3 to <16 x i8>
|
|
ret <16 x i8> %tmp4
|
|
}
|
|
|
|
define <8 x i16> @vcneQi16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
|
|
%tmp1 = load <8 x i16>* %A
|
|
%tmp2 = load <8 x i16>* %B
|
|
%tmp3 = icmp ne <8 x i16> %tmp1, %tmp2
|
|
%tmp4 = sext <8 x i1> %tmp3 to <8 x i16>
|
|
ret <8 x i16> %tmp4
|
|
}
|
|
|
|
define <4 x i32> @vcneQi32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
|
|
%tmp1 = load <4 x i32>* %A
|
|
%tmp2 = load <4 x i32>* %B
|
|
%tmp3 = icmp ne <4 x i32> %tmp1, %tmp2
|
|
%tmp4 = sext <4 x i1> %tmp3 to <4 x i32>
|
|
ret <4 x i32> %tmp4
|
|
}
|
|
|
|
define <16 x i8> @vcltQs8(<16 x i8>* %A, <16 x i8>* %B) nounwind {
|
|
%tmp1 = load <16 x i8>* %A
|
|
%tmp2 = load <16 x i8>* %B
|
|
%tmp3 = icmp slt <16 x i8> %tmp1, %tmp2
|
|
%tmp4 = sext <16 x i1> %tmp3 to <16 x i8>
|
|
ret <16 x i8> %tmp4
|
|
}
|
|
|
|
define <4 x i16> @vcles16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
|
|
%tmp1 = load <4 x i16>* %A
|
|
%tmp2 = load <4 x i16>* %B
|
|
%tmp3 = icmp sle <4 x i16> %tmp1, %tmp2
|
|
%tmp4 = sext <4 x i1> %tmp3 to <4 x i16>
|
|
ret <4 x i16> %tmp4
|
|
}
|
|
|
|
define <4 x i16> @vcltu16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
|
|
%tmp1 = load <4 x i16>* %A
|
|
%tmp2 = load <4 x i16>* %B
|
|
%tmp3 = icmp ult <4 x i16> %tmp1, %tmp2
|
|
%tmp4 = sext <4 x i1> %tmp3 to <4 x i16>
|
|
ret <4 x i16> %tmp4
|
|
}
|
|
|
|
define <4 x i32> @vcleQu32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
|
|
%tmp1 = load <4 x i32>* %A
|
|
%tmp2 = load <4 x i32>* %B
|
|
%tmp3 = icmp ule <4 x i32> %tmp1, %tmp2
|
|
%tmp4 = sext <4 x i1> %tmp3 to <4 x i32>
|
|
ret <4 x i32> %tmp4
|
|
}
|