Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
iskander
on July 1, 2013
|
parent
|
context
|
favorite
| on:
Harlan, new Lispy language for GPU programming
How do you support higher order procedures? Do you do control flow analysis & defunctionalization or are these "real" higher order functions?
eholk
on July 1, 2013
[–]
It's basically defunctionalization. It's hard to get access to function pointers on the GPU to do "real" higher order functions.
seanmcdirmid
on July 1, 2013
|
parent
|
next
[–]
What version of CUDA are you talking about. I though the latest version supported indirect function pointers (finally).
iskander
on July 2, 2013
|
root
|
parent
|
next
[–]
That's good news! Did that change with Kepler? Any links to more info?
iskander
on July 1, 2013
|
parent
|
prev
[–]
I wish they'd at least give us a computed goto...but alas, every switch is destined for predicated branching.
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
