in reply to Re^12: Perl 5 Optimizing Compiler, Part 4: LLVM Backend? in thread Perl 5 Optimizing Compiler, Part 4: LLVM Backend?
Maybe that should change. Someone should compare Perl 5.10.1 against blead for time per opcode type on same machine on same Perl code. I've thought and tried in the past to modifying NYTProf to profile all opcodes, not just entersub and a couple POSIX C Lib named opcodes. Other projects like Mozilla and Microsoft routinely have performance teams that track degradation in speed. While Perl doesn't have anywhere close to corporate funding or salaried programmers working on it, some kind of crude report for finding order of magnitude slow downs between releases. Currently it takes individual programmers benchmarking their scripts on old and new perls to trace down performance problems. Most programmers just update their Perls and *assume* P5P did their diligence and didn't create a Vista. My ActivePerl 5.10.0 DLL is 873KB, Blead, same config (O1, 32bit Win32, Visual C, no DEBUGGING) is 1056KB. About a 20% difference. Why? I dont know. Is all that growth healthy? I dont know.
|