745089 - Move all single-precision floating point operations into FloatingPoint.h

Status: RESOLVED FIXED

(Core :: MFBT, defect)

Description

[Jeff Walden [:Waldo]]

I moved all the double-precision stuff I could find into FloatingPoint.h in bug 714260, but I didn't do anything about single-precision stuff.  I'm not sure how much we actually have; searches for mant(issa), significand, mask, expmask, and so on didn't return anything on-point that I could see.  (I skimmed MXR results, so I could have missed something.)  I do know that gfx/src/nsCoord.h has a little bit of this stuff; there might be more.  We should probably add the corresponding methods and such to be complete here, but given the comparatively few users, it doesn't seem like a priority.

Comment 1

[Jeff Walden [:Waldo]]

Attachment: Replace nscoord-specific positive-infinity/is-nan primitives with FloatingPoint.h uses

Years later I don't have much idea what other code I might have meant by this bug.  If there's more beyond this nscoord stuff, I think it's time to leave it to new bugs and finish up what's been specifically identified here.

Comment 2

[Jeff Walden [:Waldo]]

Tryserver run indicated Mac builds are fine now with that bit of volatile ad-hoc-ery removed, FYI.

Comment 3

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Comment on attachment 8982979 [details] [diff] [review]
Replace nscoord-specific positive-infinity/is-nan primitives with FloatingPoint.h uses

Looks fine to me.

The only interesting bit is the removal of https://hg.mozilla.org/mozilla-central/rev/a1ab5a4ed1f37b8cafee829bcac891c1be9f1409 , but if the tree is fine with it, then I'm fine with it.

r=dbaron

Comment 4

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/11c5f8019bca
Replace nscoord-specific positive-infinity/is-nan primitives with FloatingPoint.h uses.  r=dbaron

Comment 5

[Arthur Iakab [arthur_iakab]]

https://hg.mozilla.org/mozilla-central/rev/11c5f8019bca