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-vcvt.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

405 lines
14 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,+fullfp16 -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-MVE
; RUN: llc -mtriple=thumbv8.1m.main-none-none-eabi -mattr=+mve.fp -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-MVEFP
define arm_aapcs_vfpcc <4 x float> @foo_float_int32(<4 x i32> %src) {
; CHECK-MVE-LABEL: foo_float_int32:
; CHECK-MVE: @ %bb.0: @ %entry
; CHECK-MVE-NEXT: vcvt.f32.s32 s7, s3
; CHECK-MVE-NEXT: vcvt.f32.s32 s6, s2
; CHECK-MVE-NEXT: vcvt.f32.s32 s5, s1
; CHECK-MVE-NEXT: vcvt.f32.s32 s4, s0
; CHECK-MVE-NEXT: vmov q0, q1
; CHECK-MVE-NEXT: bx lr
;
; CHECK-MVEFP-LABEL: foo_float_int32:
; CHECK-MVEFP: @ %bb.0: @ %entry
; CHECK-MVEFP-NEXT: vcvt.f32.s32 q0, q0
; CHECK-MVEFP-NEXT: bx lr
entry:
%out = sitofp <4 x i32> %src to <4 x float>
ret <4 x float> %out
}
define arm_aapcs_vfpcc <4 x float> @foo_float_uint32(<4 x i32> %src) {
; CHECK-MVE-LABEL: foo_float_uint32:
; CHECK-MVE: @ %bb.0: @ %entry
; CHECK-MVE-NEXT: vcvt.f32.u32 s7, s3
; CHECK-MVE-NEXT: vcvt.f32.u32 s6, s2
; CHECK-MVE-NEXT: vcvt.f32.u32 s5, s1
; CHECK-MVE-NEXT: vcvt.f32.u32 s4, s0
; CHECK-MVE-NEXT: vmov q0, q1
; CHECK-MVE-NEXT: bx lr
;
; CHECK-MVEFP-LABEL: foo_float_uint32:
; CHECK-MVEFP: @ %bb.0: @ %entry
; CHECK-MVEFP-NEXT: vcvt.f32.u32 q0, q0
; CHECK-MVEFP-NEXT: bx lr
entry:
%out = uitofp <4 x i32> %src to <4 x float>
ret <4 x float> %out
}
define arm_aapcs_vfpcc <4 x i32> @foo_int32_float(<4 x float> %src) {
; CHECK-MVE-LABEL: foo_int32_float:
; CHECK-MVE: @ %bb.0: @ %entry
; CHECK-MVE-NEXT: vcvt.s32.f32 s4, s2
; CHECK-MVE-NEXT: vcvt.s32.f32 s6, s0
; CHECK-MVE-NEXT: vcvt.s32.f32 s8, s3
; CHECK-MVE-NEXT: vcvt.s32.f32 s10, s1
; CHECK-MVE-NEXT: vmov r0, s4
; CHECK-MVE-NEXT: vmov r1, s6
; CHECK-MVE-NEXT: vmov q0[2], q0[0], r1, r0
; CHECK-MVE-NEXT: vmov r0, s8
; CHECK-MVE-NEXT: vmov r1, s10
; CHECK-MVE-NEXT: vmov q0[3], q0[1], r1, r0
; CHECK-MVE-NEXT: bx lr
;
; CHECK-MVEFP-LABEL: foo_int32_float:
; CHECK-MVEFP: @ %bb.0: @ %entry
; CHECK-MVEFP-NEXT: vcvt.s32.f32 q0, q0
; CHECK-MVEFP-NEXT: bx lr
entry:
%out = fptosi <4 x float> %src to <4 x i32>
ret <4 x i32> %out
}
define arm_aapcs_vfpcc <4 x i32> @foo_uint32_float(<4 x float> %src) {
; CHECK-MVE-LABEL: foo_uint32_float:
; CHECK-MVE: @ %bb.0: @ %entry
; CHECK-MVE-NEXT: vcvt.u32.f32 s4, s2
; CHECK-MVE-NEXT: vcvt.u32.f32 s6, s0
; CHECK-MVE-NEXT: vcvt.u32.f32 s8, s3
; CHECK-MVE-NEXT: vcvt.u32.f32 s10, s1
; CHECK-MVE-NEXT: vmov r0, s4
; CHECK-MVE-NEXT: vmov r1, s6
; CHECK-MVE-NEXT: vmov q0[2], q0[0], r1, r0
; CHECK-MVE-NEXT: vmov r0, s8
; CHECK-MVE-NEXT: vmov r1, s10
; CHECK-MVE-NEXT: vmov q0[3], q0[1], r1, r0
; CHECK-MVE-NEXT: bx lr
;
; CHECK-MVEFP-LABEL: foo_uint32_float:
; CHECK-MVEFP: @ %bb.0: @ %entry
; CHECK-MVEFP-NEXT: vcvt.u32.f32 q0, q0
; CHECK-MVEFP-NEXT: bx lr
entry:
%out = fptoui <4 x float> %src to <4 x i32>
ret <4 x i32> %out
}
define arm_aapcs_vfpcc <8 x half> @foo_half_int16(<8 x i16> %src) {
; CHECK-MVE-LABEL: foo_half_int16:
; CHECK-MVE: @ %bb.0: @ %entry
; CHECK-MVE-NEXT: vmov q1, q0
; CHECK-MVE-NEXT: vmov.s16 r0, q0[0]
; CHECK-MVE-NEXT: vmov s0, r0
; CHECK-MVE-NEXT: vmov.s16 r0, q1[1]
; CHECK-MVE-NEXT: vmov s2, r0
; CHECK-MVE-NEXT: vmov.s16 r0, q1[3]
; CHECK-MVE-NEXT: vcvt.f16.s32 s8, s2
; CHECK-MVE-NEXT: vcvt.f16.s32 s0, s0
; CHECK-MVE-NEXT: vins.f16 s0, s8
; CHECK-MVE-NEXT: vmov s8, r0
; CHECK-MVE-NEXT: vmov.s16 r0, q1[2]
; CHECK-MVE-NEXT: vcvt.f16.s32 s8, s8
; CHECK-MVE-NEXT: vmov s10, r0
; CHECK-MVE-NEXT: vmov.s16 r0, q1[4]
; CHECK-MVE-NEXT: vcvt.f16.s32 s1, s10
; CHECK-MVE-NEXT: vins.f16 s1, s8
; CHECK-MVE-NEXT: vmov s8, r0
; CHECK-MVE-NEXT: vmov.s16 r0, q1[5]
; CHECK-MVE-NEXT: vcvt.f16.s32 s2, s8
; CHECK-MVE-NEXT: vmov s10, r0
; CHECK-MVE-NEXT: vmov.s16 r0, q1[7]
; CHECK-MVE-NEXT: vmov s8, r0
; CHECK-MVE-NEXT: vmov.s16 r0, q1[6]
; CHECK-MVE-NEXT: vcvt.f16.s32 s10, s10
; CHECK-MVE-NEXT: vmov s4, r0
; CHECK-MVE-NEXT: vins.f16 s2, s10
; CHECK-MVE-NEXT: vcvt.f16.s32 s8, s8
; CHECK-MVE-NEXT: vcvt.f16.s32 s3, s4
; CHECK-MVE-NEXT: vins.f16 s3, s8
; CHECK-MVE-NEXT: bx lr
;
; CHECK-MVEFP-LABEL: foo_half_int16:
; CHECK-MVEFP: @ %bb.0: @ %entry
; CHECK-MVEFP-NEXT: vcvt.f16.s16 q0, q0
; CHECK-MVEFP-NEXT: bx lr
entry:
%out = sitofp <8 x i16> %src to <8 x half>
ret <8 x half> %out
}
define arm_aapcs_vfpcc <8 x half> @foo_half_uint16(<8 x i16> %src) {
; CHECK-MVE-LABEL: foo_half_uint16:
; CHECK-MVE: @ %bb.0: @ %entry
; CHECK-MVE-NEXT: vmov q1, q0
; CHECK-MVE-NEXT: vmov.u16 r0, q0[0]
; CHECK-MVE-NEXT: vmov s0, r0
; CHECK-MVE-NEXT: vmov.u16 r0, q1[1]
; CHECK-MVE-NEXT: vmov s2, r0
; CHECK-MVE-NEXT: vmov.u16 r0, q1[3]
; CHECK-MVE-NEXT: vcvt.f16.u32 s8, s2
; CHECK-MVE-NEXT: vcvt.f16.u32 s0, s0
; CHECK-MVE-NEXT: vins.f16 s0, s8
; CHECK-MVE-NEXT: vmov s8, r0
; CHECK-MVE-NEXT: vmov.u16 r0, q1[2]
; CHECK-MVE-NEXT: vcvt.f16.u32 s8, s8
; CHECK-MVE-NEXT: vmov s10, r0
; CHECK-MVE-NEXT: vmov.u16 r0, q1[4]
; CHECK-MVE-NEXT: vcvt.f16.u32 s1, s10
; CHECK-MVE-NEXT: vins.f16 s1, s8
; CHECK-MVE-NEXT: vmov s8, r0
; CHECK-MVE-NEXT: vmov.u16 r0, q1[5]
; CHECK-MVE-NEXT: vcvt.f16.u32 s2, s8
; CHECK-MVE-NEXT: vmov s10, r0
; CHECK-MVE-NEXT: vmov.u16 r0, q1[7]
; CHECK-MVE-NEXT: vmov s8, r0
; CHECK-MVE-NEXT: vmov.u16 r0, q1[6]
; CHECK-MVE-NEXT: vcvt.f16.u32 s10, s10
; CHECK-MVE-NEXT: vmov s4, r0
; CHECK-MVE-NEXT: vins.f16 s2, s10
; CHECK-MVE-NEXT: vcvt.f16.u32 s8, s8
; CHECK-MVE-NEXT: vcvt.f16.u32 s3, s4
; CHECK-MVE-NEXT: vins.f16 s3, s8
; CHECK-MVE-NEXT: bx lr
;
; CHECK-MVEFP-LABEL: foo_half_uint16:
; CHECK-MVEFP: @ %bb.0: @ %entry
; CHECK-MVEFP-NEXT: vcvt.f16.u16 q0, q0
; CHECK-MVEFP-NEXT: bx lr
entry:
%out = uitofp <8 x i16> %src to <8 x half>
ret <8 x half> %out
}
define arm_aapcs_vfpcc <8 x i16> @foo_int16_half(<8 x half> %src) {
; CHECK-MVE-LABEL: foo_int16_half:
; CHECK-MVE: @ %bb.0: @ %entry
; CHECK-MVE-NEXT: vmovx.f16 s14, s0
; CHECK-MVE-NEXT: vcvt.s32.f16 s0, s0
; CHECK-MVE-NEXT: vcvt.s32.f16 s14, s14
; CHECK-MVE-NEXT: vmov r0, s0
; CHECK-MVE-NEXT: vmovx.f16 s4, s3
; CHECK-MVE-NEXT: vmovx.f16 s6, s2
; CHECK-MVE-NEXT: vmovx.f16 s10, s1
; CHECK-MVE-NEXT: vcvt.s32.f16 s8, s3
; CHECK-MVE-NEXT: vcvt.s32.f16 s12, s2
; CHECK-MVE-NEXT: vcvt.s32.f16 s5, s1
; CHECK-MVE-NEXT: vmov.16 q0[0], r0
; CHECK-MVE-NEXT: vmov r0, s14
; CHECK-MVE-NEXT: vmov.16 q0[1], r0
; CHECK-MVE-NEXT: vmov r0, s5
; CHECK-MVE-NEXT: vcvt.s32.f16 s10, s10
; CHECK-MVE-NEXT: vmov.16 q0[2], r0
; CHECK-MVE-NEXT: vmov r0, s10
; CHECK-MVE-NEXT: vcvt.s32.f16 s6, s6
; CHECK-MVE-NEXT: vmov.16 q0[3], r0
; CHECK-MVE-NEXT: vmov r0, s12
; CHECK-MVE-NEXT: vmov.16 q0[4], r0
; CHECK-MVE-NEXT: vmov r0, s6
; CHECK-MVE-NEXT: vmov.16 q0[5], r0
; CHECK-MVE-NEXT: vmov r0, s8
; CHECK-MVE-NEXT: vcvt.s32.f16 s4, s4
; CHECK-MVE-NEXT: vmov.16 q0[6], r0
; CHECK-MVE-NEXT: vmov r0, s4
; CHECK-MVE-NEXT: vmov.16 q0[7], r0
; CHECK-MVE-NEXT: bx lr
;
; CHECK-MVEFP-LABEL: foo_int16_half:
; CHECK-MVEFP: @ %bb.0: @ %entry
; CHECK-MVEFP-NEXT: vcvt.s16.f16 q0, q0
; CHECK-MVEFP-NEXT: bx lr
entry:
%out = fptosi <8 x half> %src to <8 x i16>
ret <8 x i16> %out
}
define arm_aapcs_vfpcc <8 x i16> @foo_uint16_half(<8 x half> %src) {
; CHECK-MVE-LABEL: foo_uint16_half:
; CHECK-MVE: @ %bb.0: @ %entry
; CHECK-MVE-NEXT: vmovx.f16 s14, s0
; CHECK-MVE-NEXT: vcvt.s32.f16 s0, s0
; CHECK-MVE-NEXT: vcvt.s32.f16 s14, s14
; CHECK-MVE-NEXT: vmov r0, s0
; CHECK-MVE-NEXT: vmovx.f16 s4, s3
; CHECK-MVE-NEXT: vmovx.f16 s6, s2
; CHECK-MVE-NEXT: vmovx.f16 s10, s1
; CHECK-MVE-NEXT: vcvt.s32.f16 s8, s3
; CHECK-MVE-NEXT: vcvt.s32.f16 s12, s2
; CHECK-MVE-NEXT: vcvt.s32.f16 s5, s1
; CHECK-MVE-NEXT: vmov.16 q0[0], r0
; CHECK-MVE-NEXT: vmov r0, s14
; CHECK-MVE-NEXT: vmov.16 q0[1], r0
; CHECK-MVE-NEXT: vmov r0, s5
; CHECK-MVE-NEXT: vcvt.s32.f16 s10, s10
; CHECK-MVE-NEXT: vmov.16 q0[2], r0
; CHECK-MVE-NEXT: vmov r0, s10
; CHECK-MVE-NEXT: vcvt.s32.f16 s6, s6
; CHECK-MVE-NEXT: vmov.16 q0[3], r0
; CHECK-MVE-NEXT: vmov r0, s12
; CHECK-MVE-NEXT: vmov.16 q0[4], r0
; CHECK-MVE-NEXT: vmov r0, s6
; CHECK-MVE-NEXT: vmov.16 q0[5], r0
; CHECK-MVE-NEXT: vmov r0, s8
; CHECK-MVE-NEXT: vcvt.s32.f16 s4, s4
; CHECK-MVE-NEXT: vmov.16 q0[6], r0
; CHECK-MVE-NEXT: vmov r0, s4
; CHECK-MVE-NEXT: vmov.16 q0[7], r0
; CHECK-MVE-NEXT: bx lr
;
; CHECK-MVEFP-LABEL: foo_uint16_half:
; CHECK-MVEFP: @ %bb.0: @ %entry
; CHECK-MVEFP-NEXT: vcvt.u16.f16 q0, q0
; CHECK-MVEFP-NEXT: bx lr
entry:
%out = fptoui <8 x half> %src to <8 x i16>
ret <8 x i16> %out
}
define arm_aapcs_vfpcc <2 x double> @foo_float_int64(<2 x i64> %src) {
; CHECK-LABEL: foo_float_int64:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: .save {r7, lr}
; CHECK-NEXT: push {r7, lr}
; CHECK-NEXT: .vsave {d8, d9}
; CHECK-NEXT: vpush {d8, d9}
; CHECK-NEXT: vmov q4, q0
; CHECK-NEXT: vmov r0, r1, d9
; CHECK-NEXT: bl __aeabi_l2d
; CHECK-NEXT: vmov r2, r3, d8
; CHECK-NEXT: vmov d9, r0, r1
; CHECK-NEXT: mov r0, r2
; CHECK-NEXT: mov r1, r3
; CHECK-NEXT: bl __aeabi_l2d
; CHECK-NEXT: vmov d8, r0, r1
; CHECK-NEXT: vmov q0, q4
; CHECK-NEXT: vpop {d8, d9}
; CHECK-NEXT: pop {r7, pc}
entry:
%out = sitofp <2 x i64> %src to <2 x double>
ret <2 x double> %out
}
define arm_aapcs_vfpcc <2 x double> @foo_float_uint64(<2 x i64> %src) {
; CHECK-LABEL: foo_float_uint64:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: .save {r7, lr}
; CHECK-NEXT: push {r7, lr}
; CHECK-NEXT: .vsave {d8, d9}
; CHECK-NEXT: vpush {d8, d9}
; CHECK-NEXT: vmov q4, q0
; CHECK-NEXT: vmov r0, r1, d9
; CHECK-NEXT: bl __aeabi_ul2d
; CHECK-NEXT: vmov r2, r3, d8
; CHECK-NEXT: vmov d9, r0, r1
; CHECK-NEXT: mov r0, r2
; CHECK-NEXT: mov r1, r3
; CHECK-NEXT: bl __aeabi_ul2d
; CHECK-NEXT: vmov d8, r0, r1
; CHECK-NEXT: vmov q0, q4
; CHECK-NEXT: vpop {d8, d9}
; CHECK-NEXT: pop {r7, pc}
entry:
%out = uitofp <2 x i64> %src to <2 x double>
ret <2 x double> %out
}
define arm_aapcs_vfpcc <2 x i64> @foo_int64_float(<2 x double> %src) {
; CHECK-LABEL: foo_int64_float:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: .save {r4, r5, r7, lr}
; CHECK-NEXT: push {r4, r5, r7, lr}
; CHECK-NEXT: .vsave {d8, d9}
; CHECK-NEXT: vpush {d8, d9}
; CHECK-NEXT: vmov q4, q0
; CHECK-NEXT: vmov r0, r1, d9
; CHECK-NEXT: bl __aeabi_d2lz
; CHECK-NEXT: mov r4, r0
; CHECK-NEXT: mov r5, r1
; CHECK-NEXT: vmov r0, r1, d8
; CHECK-NEXT: bl __aeabi_d2lz
; CHECK-NEXT: vmov q0[2], q0[0], r0, r4
; CHECK-NEXT: vmov q0[3], q0[1], r1, r5
; CHECK-NEXT: vpop {d8, d9}
; CHECK-NEXT: pop {r4, r5, r7, pc}
entry:
%out = fptosi <2 x double> %src to <2 x i64>
ret <2 x i64> %out
}
define arm_aapcs_vfpcc <2 x i64> @foo_uint64_float(<2 x double> %src) {
; CHECK-LABEL: foo_uint64_float:
; CHECK: @ %bb.0: @ %entry
; CHECK-NEXT: .save {r4, r5, r7, lr}
; CHECK-NEXT: push {r4, r5, r7, lr}
; CHECK-NEXT: .vsave {d8, d9}
; CHECK-NEXT: vpush {d8, d9}
; CHECK-NEXT: vmov q4, q0
; CHECK-NEXT: vmov r0, r1, d9
; CHECK-NEXT: bl __aeabi_d2ulz
; CHECK-NEXT: mov r4, r0
; CHECK-NEXT: mov r5, r1
; CHECK-NEXT: vmov r0, r1, d8
; CHECK-NEXT: bl __aeabi_d2ulz
; CHECK-NEXT: vmov q0[2], q0[0], r0, r4
; CHECK-NEXT: vmov q0[3], q0[1], r1, r5
; CHECK-NEXT: vpop {d8, d9}
; CHECK-NEXT: pop {r4, r5, r7, pc}
entry:
%out = fptoui <2 x double> %src to <2 x i64>
ret <2 x i64> %out
}
define arm_aapcs_vfpcc <8 x half> @vmovn32_trunc1(<4 x float> %src1, <4 x float> %src2) {
; CHECK-MVE-LABEL: vmovn32_trunc1:
; CHECK-MVE: @ %bb.0: @ %entry
; CHECK-MVE-NEXT: vmov q2, q0
; CHECK-MVE-NEXT: vcvtb.f16.f32 s0, s8
; CHECK-MVE-NEXT: vcvtt.f16.f32 s0, s4
; CHECK-MVE-NEXT: vcvtb.f16.f32 s1, s9
; CHECK-MVE-NEXT: vcvtt.f16.f32 s1, s5
; CHECK-MVE-NEXT: vcvtb.f16.f32 s2, s10
; CHECK-MVE-NEXT: vcvtt.f16.f32 s2, s6
; CHECK-MVE-NEXT: vcvtb.f16.f32 s3, s11
; CHECK-MVE-NEXT: vcvtt.f16.f32 s3, s7
; CHECK-MVE-NEXT: bx lr
;
; CHECK-MVEFP-LABEL: vmovn32_trunc1:
; CHECK-MVEFP: @ %bb.0: @ %entry
; CHECK-MVEFP-NEXT: vcvtb.f16.f32 q0, q0
; CHECK-MVEFP-NEXT: vcvtt.f16.f32 q0, q1
; CHECK-MVEFP-NEXT: bx lr
entry:
%strided.vec = shufflevector <4 x float> %src1, <4 x float> %src2, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
%out = fptrunc <8 x float> %strided.vec to <8 x half>
ret <8 x half> %out
}
define arm_aapcs_vfpcc <8 x half> @vmovn32_trunc2(<4 x float> %src1, <4 x float> %src2) {
; CHECK-MVE-LABEL: vmovn32_trunc2:
; CHECK-MVE: @ %bb.0: @ %entry
; CHECK-MVE-NEXT: vmov q2, q0
; CHECK-MVE-NEXT: vcvtb.f16.f32 s0, s4
; CHECK-MVE-NEXT: vcvtt.f16.f32 s0, s8
; CHECK-MVE-NEXT: vcvtb.f16.f32 s1, s5
; CHECK-MVE-NEXT: vcvtt.f16.f32 s1, s9
; CHECK-MVE-NEXT: vcvtb.f16.f32 s2, s6
; CHECK-MVE-NEXT: vcvtt.f16.f32 s2, s10
; CHECK-MVE-NEXT: vcvtb.f16.f32 s3, s7
; CHECK-MVE-NEXT: vcvtt.f16.f32 s3, s11
; CHECK-MVE-NEXT: bx lr
;
; CHECK-MVEFP-LABEL: vmovn32_trunc2:
; CHECK-MVEFP: @ %bb.0: @ %entry
; CHECK-MVEFP-NEXT: vcvtb.f16.f32 q1, q1
; CHECK-MVEFP-NEXT: vcvtt.f16.f32 q1, q0
; CHECK-MVEFP-NEXT: vmov q0, q1
; CHECK-MVEFP-NEXT: bx lr
entry:
%strided.vec = shufflevector <4 x float> %src1, <4 x float> %src2, <8 x i32> <i32 4, i32 0, i32 5, i32 1, i32 6, i32 2, i32 7, i32 3>
%out = fptrunc <8 x float> %strided.vec to <8 x half>
ret <8 x half> %out
}