Closed zaunere closed 1 month ago
Hi @zaunere
Are you using make install-cuda
instead of make install
? Make sure the compilation used the following commands:
phpize
./configure --with-cuda
make install-cuda
Hello,
I wasn't using make install-cuda
:(
I always do a `make && make install
Using install-cuda seems to have worked.
Thank you,
Hans
No problem, let me know if you have any further questions/problems. I will close this issue.
Hello,
I'm guessing I can't use CUDA with an AMD CPU?
/home/working/build/numpower/src/ndarray.c: In function 'NDArray_ToGD': /home/working/build/numpower/src/ndarray.c:188:13: warning: AVX vector return without AVX enabled changes the ABI [-Wpsabi] 188 | __m256i alpha_mask = _mm256_set_epi32(0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF); | ^
~~~~~ In file included from /usr/lib/gcc/x86_64-redhat-linux/13/include/immintrin.h:43, from /home/working/build/numpower/src/ndarray.c:14: /usr/lib/gcc/x86_64-redhat-linux/13/include/avxintrin.h:1274:1: error: inlining failed in call to 'always_inline' '_mm256_set_epi32': target specific option mismatchTo note, the Nvidia/nvcc/etc tools are installed and seem to work normally with pytorch/etc.
Thanks,
Hans