You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi, first of all thanks for the awesome project! I really enjoy the self-managed CPython versions.
I noticed that the versions downloaded do not take advantage of CPU features like AVX (SIMD), because rye always installs the x86_64-* variants and not the x86_64_v{2,3,4}-* variants, even if applicable (doc). The new variants might give slight performance improvements. However, I have to note that the linked doc is not overly optimistic about the achieved performance gains, although slight gains are to be expected.
I would prefer if rye would detect the supported CPU features and download the corresponding Python version for me, as nowadays almost all common devices support at least v3 with AVX support. I took a short look at the code, but the way rye fetches its CPython builds seems rather complex to me.
I think that this does not affect 3rd-party libraries like NumPy which ship their own AVX-supporting binaries.
The text was updated successfully, but these errors were encountered:
Hi, first of all thanks for the awesome project! I really enjoy the self-managed CPython versions.
I noticed that the versions downloaded do not take advantage of CPU features like AVX (SIMD), because rye always installs the
x86_64-*
variants and not thex86_64_v{2,3,4}-*
variants, even if applicable (doc). The new variants might give slight performance improvements. However, I have to note that the linked doc is not overly optimistic about the achieved performance gains, although slight gains are to be expected.I would prefer if rye would detect the supported CPU features and download the corresponding Python version for me, as nowadays almost all common devices support at least
v3
with AVX support. I took a short look at the code, but the way rye fetches its CPython builds seems rather complex to me.I think that this does not affect 3rd-party libraries like NumPy which ship their own AVX-supporting binaries.
The text was updated successfully, but these errors were encountered: