Build failure with GCC14 #7650
Description
To reproduce the issue:
- Download AlmaLinux release 8.10
- Install GCC 14. ( I installed gcc-toolset-14-gcc-14.2.1-1.1.el8_10.aarch64 and gcc-toolset-14-gcc-c++-14.2.1-1.1.el8_10.aarch64 RPM packages)
- Checkout xnnpack code and build it
git clone https://github.com/google/XNNPACK.git
cd XNNPACK/
mkdir b
cd b
cmake .. -DCMAKE_BUILD_TYPE=Debug -G Ninja
ninjaYou will see:
/tmp/XNNPACK/src/f16-raddstoreexpminusmax/gen/f16-raddstoreexpminusmax-neonfp16arith-rr2-p2-u32.c: In function ‘xnn_f16_raddstoreexpminusmax_ukernel__neonfp16arith_rr2_p2_u32’:
/tmp/XNNPACK/src/f16-raddstoreexpminusmax/gen/f16-raddstoreexpminusmax-neonfp16arith-rr2-p2-u32.c:189:17: error: passing argument 1 of ‘vst1_lane_u16’ from incompatible pointer type [-Wincompatible-pointer-types]
189 | vst1_lane_u16(sum, vreinterpret_u16_f16(vacc_lo), 0);
| ^~~
| |
| xnn_float16 * {aka _Float16 *}
In file included from /tmp/XNNPACK/src/f16-raddstoreexpminusmax/gen/f16-raddstoreexpminusmax-neonfp16arith-rr2-p2-u32.c:12:
/opt/rh/gcc-toolset-14/root/usr/lib/gcc/aarch64-redhat-linux/14/include/arm_neon.h:22267:26: note: expected ‘uint16_t *’ {aka ‘short unsigned int *’} but argument is of type ‘xnn_float16 *’ {aka ‘_Float16 *’}
22267 | vst1_lane_u16 (uint16_t *__a, uint16x4_t __b, const int __lane)
| ~~~~~~~~~~^~~