1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-22 18:54:02 +01:00
llvm-mirror/test/CodeGen/Thumb2/mve-neg.ll
David Green 17e932c916 [ARM] Create VMOVRRD from adjacent vector extracts
This adds a combine for extract(x, n); extract(x, n+1)  ->
VMOVRRD(extract x, n/2). This allows two vector lanes to be moved at the
same time in a single instruction, and thanks to the other VMOVRRD folds
we have added recently can help reduce the amount of executed
instructions. Floating point types are very similar, but will include a
bitcast to an integer type.

This also adds a shouldRewriteCopySrc, to prevent copy propagation from
DPR to SPR, which can break as not all DPR regs can be extracted from
directly.  Otherwise the machine verifier is unhappy.

Differential Revision: https://reviews.llvm.org/D100244
2021-04-20 15:15:43 +01:00

52 lines
1.4 KiB
LLVM

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=thumbv8.1m.main-none-none-eabi -mattr=+mve -verify-machineinstrs %s -o - | FileCheck %s
define arm_aapcs_vfpcc <16 x i8> @neg_v16i8(<16 x i8> %s1) {
; CHECK-LABEL: neg_v16i8:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: vneg.s8 q0, q0
; CHECK-NEXT: bx lr
entry:
%0 = sub nsw <16 x i8> zeroinitializer, %s1
ret <16 x i8> %0
}
define arm_aapcs_vfpcc <8 x i16> @neg_v8i16(<8 x i16> %s1) {
; CHECK-LABEL: neg_v8i16:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: vneg.s16 q0, q0
; CHECK-NEXT: bx lr
entry:
%0 = sub nsw <8 x i16> zeroinitializer, %s1
ret <8 x i16> %0
}
define arm_aapcs_vfpcc <4 x i32> @neg_v4i32(<4 x i32> %s1) {
; CHECK-LABEL: neg_v4i32:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: vneg.s32 q0, q0
; CHECK-NEXT: bx lr
entry:
%0 = sub nsw <4 x i32> zeroinitializer, %s1
ret <4 x i32> %0
}
define arm_aapcs_vfpcc <2 x i64> @neg_v2i64(<2 x i64> %s1) {
; CHECK-LABEL: neg_v2i64:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: vmov r0, r1, d1
; CHECK-NEXT: mov.w r12, #0
; CHECK-NEXT: vmov r3, r2, d0
; CHECK-NEXT: rsbs r0, r0, #0
; CHECK-NEXT: sbc.w r1, r12, r1
; CHECK-NEXT: rsbs r3, r3, #0
; CHECK-NEXT: sbc.w r2, r12, r2
; CHECK-NEXT: vmov q0[2], q0[0], r3, r0
; CHECK-NEXT: vmov q0[3], q0[1], r2, r1
; CHECK-NEXT: bx lr
entry:
%0 = sub nsw <2 x i64> zeroinitializer, %s1
ret <2 x i64> %0
}