Difference between revisions of "Talk:Haskell for multicores"
Jump to navigation
Jump to search
(Answers) |
|||
| (3 intermediate revisions by one other user not shown) | |||
| Line 1: | Line 1: | ||
Quick questions: |
Quick questions: |
||
| + | |||
| − | 1. Is the BangPatterns pragma strictly necessary? GHC compiles them without complaining. |
||
| − | + | 1. Is the BangPatterns pragma strictly necessary? If it's omitted, GHC compiles the MD5 examples without complaining. |
|
| + | |||
| + | TomMD: Odd. It _should_ be needed. I'd at least leave it till someone tests with GHC 6.10.1 - if it still isn't needed then sure, lets ditch it. |
||
| + | |||
| + | 2. Is pure timing really the best indicator of parallelism or concurrency for the disk-heavy MD5 examples? |
||
| + | |||
| + | TomMD: On all my systems this is still computationally bound. Feel free to replace it if you write a better example - I considered doing just that, but didn't want to bother writing something new. |
||
Latest revision as of 23:04, 23 October 2008
Quick questions:
1. Is the BangPatterns pragma strictly necessary? If it's omitted, GHC compiles the MD5 examples without complaining.
TomMD: Odd. It _should_ be needed. I'd at least leave it till someone tests with GHC 6.10.1 - if it still isn't needed then sure, lets ditch it.
2. Is pure timing really the best indicator of parallelism or concurrency for the disk-heavy MD5 examples?
TomMD: On all my systems this is still computationally bound. Feel free to replace it if you write a better example - I considered doing just that, but didn't want to bother writing something new.