With the proposed function-pointer-factory changes, I'm seeing this speedup on glibc systems:
num mbc
Before 0.056 0.152
After 0.057 0.089
-------
Speedup 1.0 1.7
factor
Find attached the profiling result.
What you can see:
- The glibc-compatible rewrite of UTF-8 mbrtowc is about 4.5 times faster
than the glibc implementation.
- The glibc wcwidth and the gnulib uc_width are approximately on par,
regarding speed. uc_width is probably slightly slower, but not
dramatically.
Bruno
callgrind.out.11605.png
Description: application/kcachegrind
