File: arm-neon-misc.c

package info (click to toggle)
llvm-toolchain-17 1%3A17.0.6-22
  • links: PTS, VCS
  • area: main
  • in suites: forky, sid, trixie
  • size: 1,799,624 kB
  • sloc: cpp: 6,428,607; ansic: 1,383,196; asm: 793,408; python: 223,504; objc: 75,364; f90: 60,502; lisp: 33,869; pascal: 15,282; sh: 9,684; perl: 7,453; ml: 4,937; awk: 3,523; makefile: 2,889; javascript: 2,149; xml: 888; fortran: 619; cs: 573
file content (33 lines) | stat: -rw-r--r-- 1,096 bytes parent folder | download | duplicates (4)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
// REQUIRES: arm-registered-target
// RUN: %clang_cc1 -triple thumbv7-apple-darwin \
// RUN:   -target-abi apcs-gnu \
// RUN:   -target-cpu cortex-a8 \
// RUN:   -mfloat-abi soft \
// RUN:   -target-feature +soft-float-abi \
// RUN:   -ffreestanding \
// RUN:   -emit-llvm -w -o - %s | FileCheck %s

#include <arm_neon.h>

// Radar 11998303: Avoid using i64 types for vld1q_lane and vst1q_lane Neon
// intrinsics with <2 x i64> vectors to avoid poor code for i64 in the backend.
void t1(uint64_t *src, uint8_t *dst) {
// CHECK: @t1
  uint64x2_t q = vld1q_u64(src);
// CHECK: call <2 x i64> @llvm.arm.neon.vld1.v2i64.p0
  vst1q_lane_u64(dst, q, 1);
// CHECK: shufflevector <2 x i64>
// CHECK: call void @llvm.arm.neon.vst1.p0.v1i64
}

void t2(uint64_t *src1, uint8_t *src2, uint64x2_t *dst) {
// CHECK: @t2
    uint64x2_t q = vld1q_u64(src1);
// CHECK: call <2 x i64> @llvm.arm.neon.vld1.v2i64.p0
    q = vld1q_lane_u64(src2, q, 0);
// CHECK: shufflevector <2 x i64>
// CHECK: call <1 x i64> @llvm.arm.neon.vld1.v1i64.p0
// CHECK: shufflevector <1 x i64>
    *dst = q;
// CHECK: store <2 x i64>
}