Skip to content

Conversation

@jishnub
Copy link
Member

@jishnub jishnub commented Apr 10, 2025

This improves performance

julia> B = Bidiagonal(fill(10.0,400), fill(2.0,399), :L);

julia> x = UpperTriangular(fill(400.0, size(B)));

julia> @btime $x/$B;
  193.799 μs (3 allocations: 1.22 MiB) # nightly
  163.389 μs (3 allocations: 1.22 MiB) # this PR

Also, seems to reduce TTFX slightly, but at the expense of allocations:

julia> T = Bidiagonal(ones(4), ones(3), :U); TM = Matrix(T); x = UpperTriangular(TM);

julia> @time x/T;
  0.125422 seconds (104.07 k allocations: 5.261 MiB, 99.95% compilation time) # nightly
  0.099386 seconds (113.82 k allocations: 5.748 MiB, 99.94% compilation time) # this PR

@jishnub jishnub added the performance Must go faster label Apr 10, 2025
@codecov
Copy link

codecov bot commented Apr 10, 2025

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.00%. Comparing base (84fd21b) to head (13aeddb).
Report is 1 commits behind head on master.

Additional details and impacted files
@@           Coverage Diff           @@
##           master    #1281   +/-   ##
=======================================
  Coverage   92.00%   92.00%           
=======================================
  Files          34       34           
  Lines       15488    15488           
=======================================
  Hits        14249    14249           
  Misses       1239     1239           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

@jishnub jishnub merged commit 6e2de14 into master Apr 10, 2025
4 checks passed
@jishnub jishnub deleted the jishnub/bidiag_rdiv_inbounds branch April 10, 2025 11:18
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

performance Must go faster

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants