1. compile for your particular processor, generating assembly files. Just use your normal compile command line and add -S and -march=
g++ -g -Wall -march=pentium4 -O2 -S -I.. *.cc
2. search through the .s files for the new instructions and xmm registers listed on this page: http://en.wikipedia.org/wiki/Streaming_SIMD_Extensions
eg.
fgrep "movss" *.s | grep -v "\.string" -
fgrep "xmm" *.s | grep -v "\.string" -
I found that it used xmm0 now and then but only used a couple of the new instructions a couple of times.
It makes me wonder if this will make it slower because it has to save the extra registers on to the stack every time. I can only assume GCC is smart enough so it knows it doesn't have to do this.
All of the interesting technological, artistic or just plain fun subjects I'd investigate if I had an infinite number of lifetimes. In other words, a dumping ground...
Subscribe to:
Post Comments (Atom)
tim's shared items
Blog Archive
-
▼
2008
(150)
-
▼
May
(22)
- Sunlight & GNU Scientific library
- Google AJAX library hosting
- Authentication, passwords, hashing and salts for w...
- Secure Programming for Linux and Unix
- Finance application - Buddi
- Git to CVS and back again
- Google Doctype & Compiere open source ERP & CRM
- DNS
- GCC FORTIFY_SOURCE
- IPv6
- Beer Brewing Software
- PostgreSQL & REST & Adobe Flex
- Book - C++ Templates: The Complete Guide
- Python & Vim & C++ templates book
- client side XSL and SVG
- more xsl
- XSL
- GPS units, botnet archives and Trygve Reenskaug
- more NTP scripts
- How to determine if you program is using MMX, SSE ...
- mapping #include files using dot and graphviz
- Gmail IMAP and virtual box
-
▼
May
(22)
No comments:
Post a Comment