mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-10-31 16:02:52 +01:00
916e01c917
Additional fixes: Do something reasonable for subtargets with generic itineraries by handle node latency the same as for an empty itinerary. Now nodes default to unit latency unless an itinerary explicitly specifies a zero cycle stage or it is a TokenFactor chain. Original fixes: UnitsSharePred was a source of randomness in the scheduler: node priority depended on the queue data structure. I rewrote the recent VRegCycle heuristics to completely replace the old heuristic without any randomness. To make the ndoe latency adjustments work, I also needed to do something a little more reasonable with TokenFactor. I gave it zero latency to its consumers and always schedule it as low as possible. llvm-svn: 129421
49 lines
1.1 KiB
LLVM
49 lines
1.1 KiB
LLVM
; RUN: llc < %s -march=arm -mattr=+neon -pre-RA-sched=source | FileCheck %s
|
|
|
|
define <8 x i8> @sdivi8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
|
|
;CHECK: vrecpe.f32
|
|
;CHECK: vrecpe.f32
|
|
;CHECK: vmovn.i32
|
|
;CHECK: vmovn.i32
|
|
;CHECK: vmovn.i16
|
|
%tmp1 = load <8 x i8>* %A
|
|
%tmp2 = load <8 x i8>* %B
|
|
%tmp3 = sdiv <8 x i8> %tmp1, %tmp2
|
|
ret <8 x i8> %tmp3
|
|
}
|
|
|
|
define <8 x i8> @udivi8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
|
|
;CHECK: vrecpe.f32
|
|
;CHECK: vrecps.f32
|
|
;CHECK: vrecpe.f32
|
|
;CHECK: vrecps.f32
|
|
;CHECK: vmovn.i32
|
|
;CHECK: vmovn.i32
|
|
;CHECK: vqmovun.s16
|
|
%tmp1 = load <8 x i8>* %A
|
|
%tmp2 = load <8 x i8>* %B
|
|
%tmp3 = udiv <8 x i8> %tmp1, %tmp2
|
|
ret <8 x i8> %tmp3
|
|
}
|
|
|
|
define <4 x i16> @sdivi16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
|
|
;CHECK: vrecpe.f32
|
|
;CHECK: vrecps.f32
|
|
;CHECK: vmovn.i32
|
|
%tmp1 = load <4 x i16>* %A
|
|
%tmp2 = load <4 x i16>* %B
|
|
%tmp3 = sdiv <4 x i16> %tmp1, %tmp2
|
|
ret <4 x i16> %tmp3
|
|
}
|
|
|
|
define <4 x i16> @udivi16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
|
|
;CHECK: vrecpe.f32
|
|
;CHECK: vrecps.f32
|
|
;CHECK: vrecps.f32
|
|
;CHECK: vmovn.i32
|
|
%tmp1 = load <4 x i16>* %A
|
|
%tmp2 = load <4 x i16>* %B
|
|
%tmp3 = udiv <4 x i16> %tmp1, %tmp2
|
|
ret <4 x i16> %tmp3
|
|
}
|