On Wed, 5 Aug 2015 21:12:33 -0700 Jim Meyering <j...@meyering.net> wrote:
> This sounds useful, but I discovered that the new code is not > triggered even once by any of our tests. As such, I suspect that > it is not justified to add these new conditionals in code that is > often inlined. Do you have experiments that demonstrate how > that final patch helps? Thanks for reviewing. It is effective in only special case. Could you try following test for not patched and patched? $ printf 'a\nb\n' >pat $ yes $(printf %040d 0) | head -10000000 >in $ time -p src/grep -Ff pat in In my machine, patched version is about 10% faster than not patched.
