core: fix missing vec_cvfo on IBM POWER9 due to unavailable VSX float64 conversion #27633
+28
−2
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
4 tasks
asmorkalov
reviewed
Aug 7, 2025
Comment on lines
280
to
286
| static inline vec_double2 vec_cvfo(const vec_float4& a) | ||
| { | ||
| double r0 = static_cast<double>(reinterpret_cast<const float*>(&a)[0]); | ||
| double r1 = static_cast<double>(reinterpret_cast<const float*>(&a)[1]); | ||
| return (vec_double2){r0, r1}; | ||
| } |
There was a problem hiding this comment.
Documentation for vec_floate says: "Converts an input vector to a vector of single-precision numbers. The even-numbered target elements are obtained by converting the source elements to single-precision numbers." So the replacement is not equivalent.
There was a problem hiding this comment.
Sorry, how about this?
static inline vec_double2 vec_cvfo(const vec_float4& a)
{
double r0 = static_cast<double>(reinterpret_cast<const float*>(&a)[0]);
double r1 = static_cast<double>(reinterpret_cast<const float*>(&a)[2]);
return (vec_double2){r0, r1};
}
…onversion Some POWER9 systems fail to build downstream projects (e.g. ncnn) due to undefined references to `vec_cvfo`. This happens because POWER9 does not support the VSX float64 conversion instructions introduced in POWER10. This patch introduces a fallback for `vec_cvfo` using scalar operations when building on POWER9. All other architectures (including POWER10, x86, ARM) will continue using the optimized VSX conversions. Fixes: undefined symbol vec_cvfo on POWER9 with OpenCV 4.12.0 Tested on: - POWER9 (fallback used, links correctly) - POWER10 and x86 (optimized path used as expected)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
#27635
Some POWER9 systems fail to build downstream projects (e.g. ncnn) due to undefined references to
vec_cvfo. This happens because POWER9 does not support the VSX float64 conversion instructions introduced in POWER10.This patch introduces a fallback for
vec_cvfousing scalar operations when building on POWER9. All other architectures (including POWER10, x86, ARM) will continue using the optimized VSX conversions.Fixes: undefined symbol vec_cvfo on POWER9 with OpenCV 4.12.0
Tested on:
Pull Request Readiness Checklist
See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request
Patch to opencv_extra has the same branch name.