On Tue, Oct 5, 2010 at 5:43 PM, Allan McRae <[email protected]> wrote: > On 06/10/10 08:37, Dan McGee wrote: >> >> On Tue, Oct 5, 2010 at 5:13 PM, Dan McGee<[email protected]> wrote: >>> >>> And looking closer at my numbers, did we not lose time here? If you >>> through the read() calls out (which I left in only because that was >>> the top time user), I should do some new timings or someone else >>> should as well... >> >> Slower it is! Non-truncated test output below after two warm up runs of >> each. >> > > I just replicated the same result. It is even worse if you drop all caches > before running the strace.
Off to the trash bin for this patch then? Performance improvements that don't really improve performance are kind of a downer, unless someone shows me results on OS X or Cygwin that this is a huge win and worth keeping around. -Dan
