Ensure that ArrayLS and ArrayLSLS really are LinearSlow #16
Conversation
| B = similar(A, r+1, c+1) | ||
| B[1:r, 1:c] = A | ||
| AS = ArrayLS(B) | ||
| ASS = ArrayLSLS(B) |
There was a problem hiding this comment.
Hmm, maybe we should change those names 😄
There was a problem hiding this comment.
Yeah, I didn't notice it before...maybe it's now an immortal piece of Julia benchmarking history? I'll leave it up to you to decide. 😁
|
The removal of UTF16String from base seems to have broken JLD (I see what you mean...), so this won't pass on 0.5. |
|
Since JLD was tagged and all of this passes, I'm going to merge (I want to test a PR 😉). |
|
Out of curiosity, did https://github.com/JuliaCI/BaseBenchmarkReports/blob/9dcaa7b11fc1ab036c06ef34d488d55838e31d40/019ba50_vs_f881efb/report.md run with this, or without it? It's exactly the pattern I would have predicted if this was not included. |
|
Ah, sorry. Execution parameters have to be retuned before new benchmarks (or changes) land in the |
|
No hurry, I've checked locally and there doesn't seem to be a problem with that PR. Thanks for explaining, it's nice to know the inner mysteries of the universe 🎱 |
RFC: Remove command-line Git from Mac and Windows binaries
Our inlining and codegen are getting so dang awesome on julia 0.5 that recently LLVM has begun to notice that it can skip the
divwhen performing linear indexingArrayLSandArrayLSLS. Since testing that is not really the intention of these tests, I modified the tests to defeat it. Ref JuliaLang/julia#17355 (comment).