| Summary: | ASTERISK-06925: [patch] no need in -fomit-frame-pointer | ||
| Reporter: | Andrey S Pankov (casper) | Labels: | |
| Date Opened: | 2006-05-08 13:08:23 | Date Closed: | 2006-05-19 14:10:26 |
| Priority: | Minor | Regression? | No |
| Status: | Closed/Complete | Components: | Core/General |
| Versions: | Frequency of Occurrence | ||
| Related Issues: | |||
| Environment: | Attachments: | ( 0) Makefile.r25682.diff | |
| Description: | It's enabled at levels -O, -O2, -O3, -Os by default. Let's let gcc control that... :) | ||
| Comments: | By: Jeffrey C. Ollie (jcollie) 2006-05-08 13:36:46 Sounds like a good idea to me... Same could probably be said of the use of "-fsigned-char" and "-mcpu=X" options. There are other places that -fomit-frame-pointer shows up as well. By: Jeffrey C. Ollie (jcollie) 2006-05-08 13:42:26 Also, gcc does not appear to do anything extra for optimization levels above -O3. Also, above -O2 gcc will do optimizations that increase code size. Even with the larger L1 and L2 caches available in CPUs today increasing the size of the code can have a negative effect on performance because of mempory cache misses that outweighs any gains from code optimization. By: Andrey S Pankov (casper) 2006-05-08 13:51:42 > There are other places that -fomit-frame-pointer shows up as well. They are there for platform-specific cases. And I don't know how that should be handled on those platforms. >Also, gcc does not appear to do anything extra for optimization levels above -O3. Please discuss this on the dev@ or file a _separate_ bugreport for that. By: Serge Vecher (serge-v) 2006-05-19 13:27:22 *ping* By: Russell Bryant (russell) 2006-05-19 14:10:25 added to the trunk in revision 28789 | ||