| View Issue Details [ Jump to Notes ] | [ Issue History ] [ Print ] | ||||||||||||
| ID | Project | Category | View Status | Date Submitted | Last Update | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0000069 | mercury | Bug | public | 2008-07-23 20:02 | 2008-07-23 21:06 | ||||||||
| Reporter | pbone | ||||||||||||
| Assigned To | |||||||||||||
| Priority | low | Severity | minor | Reproducibility | always | ||||||||
| Status | new | Resolution | open | ||||||||||
| Product Version | |||||||||||||
| Target Version | Fixed in Version | ||||||||||||
| Summary | 0000069: Deep profiling tools can have integer overflows. | ||||||||||||
| Description | The deep profiler and it's tools use machine word sized integers. (the int type in mercury). On 32 bit machines these values can be overflowed by a long-enough profiling run. Since the int type is a signed integer this can result in negative numbers. This can be seen with the icfp2000 program in the 'bytes allocated' statistic of the float.+ and float.* functions. | ||||||||||||
| Tags | No tags attached. | ||||||||||||
| Attached Files |
| ||||||||||||
 Relationships |
|
 Relationships |
| Notes |
|
|
juliensf (administrator) 2008-07-23 21:06 |
This has come up before. IIRC, there is not a good solution (other than to use a 64-bit machine) since the Mercury code in the deep profiler also needs to manipulate these values. One thing which might alleviate the problem somewhat is to support unsigned integers as atomic types in Mercury. |
| Notes |
| Issue History |
|||
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2008-07-23 20:02 | pbone | New Issue | |
| 2008-07-23 21:06 | juliensf | Note Added: 0000106 | |
| Issue History |