Conversation
|
I see that you have changed some lazy vals to defs, could that be part of the performance regression? |
|
In elevate macros you are wrapping these everywhere (strategies, combinators): try ... catch
case _: MatchError => Failure(this)I don't think this is what we used to do? |
|
In src/main/scala/elevate/core/strategies/debug.scala, there are rule definitions which look more like strategies to me. I think we should be careful about what is a rule and what is a strategy? |
| ) `;` | ||
| reducedFusedForm `;` | ||
| normalize.apply( | ||
| normalize( |
| @@ -1,18 +1,20 @@ | |||
| package apps | |||
|
|
|||
| import apps.separableConvolution2D._ | |||
There was a problem hiding this comment.
import apps.separableConvolution2D.*, or delete?
| @@ -6,6 +6,8 @@ import shine.OpenCL._ | |||
| import shine.cuda.KernelExecutor.{KernelNoSizes, KernelWithSizes} | |||
| import util._ | |||
There was a problem hiding this comment.
what is the difference between import x._ and import x.*?
2 participants
This is a draft PR porting
shineand all dependent repositories to Scala 3.At the moment of opening this PR, there exists a large performance regression that me need to address before merging.
Here are the changes made in: