Math tests: if fma is not available, relax some tests from exact equality to approximate equality #48102
Conversation
…lity to approximate equality
DilumAluthge
added
test
Co-authored-by: Mosè Giordano <giordano@users.noreply.github.com>
|
IIRC julia> f() = Core.Compiler.have_fma(Float64)
f (generic function with 1 method)
julia> f()
true
julia> Core.Compiler.have_fma(Float64)
false |
| # If the machine supports fma (fused multiply add), we require exact equality. | ||
| # Otherwise, we only require approximate equality. | ||
| if has_fma[T] | ||
| my_eq = (==) |
There was a problem hiding this comment.
Could use a different infix comparison operator to make this look nicer.
|
Would it be slicker to have and overloaded infix operator here that chooses |
|
This appears to fix one Windows, and I just triggered a re-run for the other given it appeared to be a httpbingo blip. If that succeeds and the suggested change above isn't imminent, it might be worth merging this as-is to get CI green again? |
|
httpbingo fail is still there, plus a few others. Also, windows tests reports are so noisy.. |
|
At least it fixes Let's merge this so at least that platform will be green on master. And we can implement the above-discussed improvements in a follow-up PR. |
…lity to approximate equality (#48102) * Math tests: if fma is not available, relax some tests from exact equality to approximate equality * Apply suggestions from code review Co-authored-by: Mosè Giordano <giordano@users.noreply.github.com> * `has_fma` has no runtime support * Add `Rational{Int}` * Put the FMA support info in the testset context * Fix whitespace * Remove inaccurate testset name Co-authored-by: Mosè Giordano <giordano@users.noreply.github.com> (cherry picked from commit 6d14b0f)
Ref #48101