|
Message-ID: <20120319174100.GA17401@brightrain.aerifal.cx> Date: Mon, 19 Mar 2012 13:41:00 -0400 From: Rich Felker <dalias@...ifal.cx> To: musl@...ts.openwall.com Subject: further work on musl's math lib Hi all, Since 0.8.7, nsz and I have been working a good deal on cleaning up the math library. In particular, I noticed the size of libc.a and libc.so had shot up quite a lot with the addition of float and long double versions of each function, complex functions, previously missing special functions, and additional C99 functions for IEEE math support. Since the release, I've implemented or re-added i386(7) assembly versions of many functions, bringing the total code size of musl down by nearly 50k (around 10%) on i386. So far the same can't be said for x86_64; in order to get that covered, I need some help from x86_64 users and coders. I can port the long double functions to x86_64 (fairly trivial to do since the FPU code is basically the same) but I'm hesitant to commit anything without a volunteer ready to check/test it. I'm also quite unfamiliar with SSE math, so I'm going to need more help getting asm written for float/double functions on x86_64. Meanwhile, nsz has been testing, improving, and cleaning up or replacing parts of the C code that were imported from BSD sources. This should also result in significant size reductions and hopefully improvements to correctness and performance. Rich
Powered by blists - more mailing lists
Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.