| Anonymous | Login | Signup for a new account | 2013-05-25 17:47 CEST | ![]() |
| Main | My View | View Issues | Change Log | Roadmap |
| View Issue Details [ Jump to Notes ] | [ Issue History ] [ Print ] | |||||||
| ID | Project | Category | View Status | Date Submitted | Last Update | |||
| 0005017 | OCaml | OCaml general | public | 2010-04-06 04:24 | 2012-03-24 15:01 | |||
| Reporter | yoiwa | |||||||
| Assigned To | xleroy | |||||||
| Priority | normal | Severity | feature | Reproducibility | always | |||
| Status | closed | Resolution | fixed | |||||
| Platform | OS | OS Version | ||||||
| Product Version | ||||||||
| Target Version | Fixed in Version | 3.12.0+dev | ||||||
| Summary | 0005017: Negative number support for big_int bit manipulations | |||||||
| Description | The issue 0004641 has added a basic bit manipulation functions to big_int library; very nice. However, I have a concern about handling of the negative number on those, especially for shift operations. I feel that the current behavior of the shift operations for negative numbers are confusing; it seems to be better consistent with x * (2 ** y) and x / (2 ** y), or two's complement. I wish to have those implemented those in consistent with two's complement (or at least reject them). Also, other bit manipulation functions for negative numbers can also be implemented in 2's complement. | |||||||
| Additional Information | I have an implementation using public interfaces of Nat and Big_int; it is available at < http://www.oiwa.jp/~yutaka/temp/big_int2.ml [^] >. It may better be revised using internal functions (which will reduce copying of Nat's); If required, I will implement/submit a revision. | |||||||
| Tags | No tags attached. | |||||||
| Attached Files | ||||||||
Notes |
|
|
(0005413) xleroy (administrator) 2010-04-29 16:03 |
After thinking about it, I agree that it is better for right shifts on negative numbers to behave like Euclidean division by 2^n, i.e. round towards minus infinity. Adjusted the behavior of shift_right_big_int accordingly. The old behavior is available as function shift_right_towards_zero_big_int. All this will be in release 3.12.0. Concerning and, or and xor on negative numbers, an efficient implementation that minimizes allocation of temporary nats is quite complex indeed, so I'd rather not do it immediately. I'll keep this in mind for later. |
Issue History |
|||
| Date Modified | Username | Field | Change |
| 2010-04-06 04:24 | yoiwa | New Issue | |
| 2010-04-29 16:03 | xleroy | Note Added: 0005413 | |
| 2010-04-29 16:03 | xleroy | Status | new => resolved |
| 2010-04-29 16:03 | xleroy | Resolution | open => fixed |
| 2010-04-29 16:03 | xleroy | Fixed in Version | => 3.12.0+dev |
| 2010-04-29 17:11 | xleroy | Assigned To | => xleroy |
| 2012-03-24 15:01 | xleroy | Status | resolved => closed |
| Copyright © 2000 - 2011 MantisBT Group |