FEAT: Missing vecdot, matvec and vecmat ufunc implementation #71

Activity

1. SwayamInSync

   added the

   Feature

   issue type on Feb 6, 2026
2. SwayamInSync

   changed the title [-]FEAT: Missing `vecdot`, `matvec` and `vetmac` ufunc implementation[/-] [+]FEAT: Missing `vecdot`, `matvec` and `vecmat` ufunc implementation[/+] on Feb 6, 2026
3. dlakaplan commented on May 26, 2026

   

   dlakaplan

   on May 26, 2026

   More actions

   Hi, @SwayamInSync I am trying to use np.dot but it's failing. I can't tell if that is implemented yet or not?

   import numpy_quaddtype
   import numpy as np
   
   x = numpy_quaddtype.QuadPrecision(np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]]))
   y = np.dot(x, x.T)
   

   (numpy 2.4.6, numpy_quaddtype from GitHub). Is this still being worked on?
4. SwayamInSync commented on May 26, 2026

   

   SwayamInSync

   on May 26, 2026

   MemberAuthor

   More actions

   Yes, IIRC dot needs to be re-wired from NumPy side. This is under radar and will ship asap
5. Brian-Castle commented on Jun 19, 2026

   

   Brian-Castle

   on Jun 19, 2026

   More actions

   Hi all, may I ask where this stands? I need this too! The dot is my most frequent request. I'm failing to understand the issue with blas, I thought linalg was built "on top of" the data type.
6. SwayamInSync commented on Jun 19, 2026

   

   SwayamInSync

   on Jun 19, 2026

   MemberAuthor

   More actions

   dot integration support for user-dtypes needs to land in NumPy first, PR is here numpy/numpy#31574

   This will be soon be added. And quaddtype uses its own custom BLAS ops but they should be accessible via np.linalg.[operation]
7. SwayamInSync

   self-assigned this

   on Jul 23, 2026
8. SwayamInSync

   added

   v1.1

   on Sep 16, 2026