blog post on parallel threads #368
Conversation
mflatt
force-pushed
the
parallel-blog
branch
from
c5233e2 to
38c2000
Compare
…together and smooth things a little bit
|
Line 413: |
Thanks! |
| has concurrency, and the expectation for libraries to work with | ||
| threads does not change with the introduction of parallel threads. | ||
| Racket's semaphores, channels, and other synchronization constructs | ||
| work the same will parallel threads as concurrent threads. Even |
There was a problem hiding this comment.
Should be work the same with (not “will ”)
|
Should the attached code files be included in this PR (as opposed to keeping them on the utah site)? |
That would be better, but I didn't figure out how to get |
| absolute time for the coroutine column in this table is more | ||
| than double the absolute time in the previous table, meaning | ||
| that we're not learning anything about parallelism with the | ||
| speed up in this table, but instead the cost of consulting |
There was a problem hiding this comment.
| speed up in this table, but instead the cost of consulting | |
| speed up in this column, but instead the cost of consulting |
The x1.3 in this column is not interesting for parallelism, but the x4.0 in the next column is interesting for parallelism.
| parameters in smaller continuations vs the cost in larger | ||
| ones. | ||
|
|
||
| And, as expected futures now provide no speedup at all. |
There was a problem hiding this comment.
| And, as expected futures now provide no speedup at all. | |
| And, as expected futures now provide no speedup at all. |
Not very important: No text change, but I'd move it after the "parallel" column discussion, between lines 462 and 463.
Try to draw back out that parallel threads are robust to language features like parameters, particularly compared to futures, and try to dwell less on the performance of parameters themselves. Also, some commas and spacing.
|
It looks like the thing done elsewhere on our blog is to put things in the |
5 participants
No description provided.