| Title | AMCHeaderFix is an unnecessary copy. |
| Status | closed |
| Priority | nice |
| Assigned user | Richard Brooksby |
| Organization | Ravenbrook |
| Description | AMCHeaderFix is an unnecessary copy. It's clear that at some point this function was created by copying AMCFix and editing it. Bad bad bad. |
| Analysis | Probably AMCHeaderFix should do something like: adjust, call AMCFix, adjust. Possibly this should be a generic so that any pool can (with a small cost in speed) have the ability to support headered objects. See also job001789. The AMC design says that the copy was made deliberately: "There are two versions of the fix method, due to its criticality" (see design.mps.poolamc.header.fix [1]). That's bogus with modern compilers. GDR 2014-09-26: All test cases pass if we always set pool->fix = AMCHeaderFix in amcInitComm. Here are some benchmark results for gcbench in HOT variety with seed 1937632292 (best of 3 runs): xci6ll lii6ll w3i6mv AMCFix: 61.94 35.56 8.01 AMCHeaderFix: 59.35 35.69 8.03 It looks as though the difference, if any, is within the normal range of variation. |
| How found | inspection |
| Evidence | See poolamc.c [1] < http://www.ravenbrook.com/project/mps/...lamc.html#design.mps.poolamc.header.fix> |
| Observed in | 1.108.0 |
| Created by | David Jones |
| Created on | 2007-07-12 15:35:00 |
| Last modified by | Gareth Rees |
| Last modified on | 2014-10-20 16:38:54 |
| History | 2007-07-12 DRJ Created 2013-03-19 GDR Assigned to RB. 2013-05-23 GDR Added reference to design. 2013-06-16 RB Downgraded to "nice" as this isn't affecting anyone (see job003499). |
| Change | Effect | Date | User | Description |
|---|---|---|---|---|
| 187293 | closed | 2014-10-20 16:38:54 | Gareth Rees | Remove duplicate function AMCFix (and rename AMCHeaderFix to AMCFix) so that we don't have the burden of maintaining two copies of this function. |