/* { dg-do compile } */ | |
/* { dg-options "-O2 -march=armv8-a+simd -mcpu=generic" } */ | |
/* Using a SIMD intrinsic from a function tagged with nosimd should fail | |
due to inlining rules. */ | |
#include "arm_neon.h" | |
__attribute__ ((target ("+nosimd"))) | |
uint8x16_t | |
foo (uint8x16_t a, uint8x16_t b, uint8x16_t c) | |
{ | |
return vbslq_u8 (a, b, c); /* { dg-message "called from here" } */ | |
} | |
/* { dg-error "inlining failed in call to 'always_inline'" "" { target *-*-* } 0 } */ |