On Sat, 19 Sep 1998 17:53:20 EDT, you said: > Therefore, any comparision is not truely fair since I assume the Base64 code > I am using has be optimized to its fullest potential by Microsoft whereas > the LZJU90 code is not. It is always bad to assume that production code has been optimized. Do you have *any* reason to assume that Microsoft optimized this code? Do you have any reason to even assume that the Microsoft code is in fact a totally correct and bug-free Base64 encoder? Have you stress-tested it with various inputs? Does it Do The Right Thing in case of memory/disk space exhaustion? (Try attaching an uncompressed GCC source tree tarball and see what happens?) Have you hand-checked the code for bogosity such as un-needed debugging code inside the central loop? Was the Microsoft code either compiled with an optimizer that did CSE and code hoisting, or was the code previously hand-tuned for this? About all you've actually proven so far is that there's reason to suspect that Microsoft has, in fact, NOT optimized this code - presumably because they felt it was "fast enough" already, and they had other problems to deal with, like making sure the program didn't induce a Blue Screen of Death.... -- Valdis Kletnieks Computer Systems Senior Engineer Virginia Tech
Attachment:
pgp00022.pgp
Description: PGP signature