Very cool!
The performance culprit is most likely the fact that Cygwin is ass-slow
sometimes.
C-Reduce is fork-intensive and the impedance mismatch between UNIX and
MS semantics is probably at its worst right there. Think about how slow
a configure script is on Cygwin.
John
On 6/20/12 11:48 PM, Eric Eide wrote:
FWIW, tonight I built the current C-Reduce on a Windows 7 box with Cygwin.
(This exercise was really about testing Emulab's prototype Windows 7 disk
image, newly created by Kirk Webb (cc'ed), but I digress....)
Building and running C-Reduce worked, but it runs rrreeaaalllyyy slowly.
For example, in `run_tests 1', each execution of pass_lines is taking 10-30
seconds. This is in contrast to running on my Mac laptop, say, where each
pass_lines execution takes a second or less.
`run_tests 1' has been running for a while now on my Windows7+Cygwin box, and
it still hasn't finished with pass_lines executions :-(. It is, however,
doing the right thing, producing the same reduction steps as on my Mac.
Maybe I screwed something up; I'm no Windows/Cygwin expert. Maybe my clang is
really really slow, even though I --enable-optimized. "time clang --version"
consistently reports 1.5s on my Windows box.
But anyway: the result of my exerice tonight is a working C-Reduce on Cygwin
(yay!) that is really really slow (boo!).
Eric.
