skia /
skia /
f4a214ecf1974eb3289564d5728b3fa8d8f3c079 Switch std::isfinite to SkIsFinite throughout Skia.
On Windows (clang-cl), using std::isfinite appears to be slower
than our old bit-manipulation code. This CL switches to SkIsFinite
roughly everywhere (modulo SkSL and a few other low-level places)
and replaces the SkIsFinite implementation with `(x - x) == 0`,
which appears to generate great code in Clang.
If metrics don't improve, we can go back to the bit masking
approach again.
Change-Id: I61dbf4545ef53031fe5ea492aa8c2f0a8c30c5cf
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/842240
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
102 files changed