|
From: Donal K. F. <don...@ma...> - 2017-05-29 14:57:15
|
On 29/05/2017 04:51, Kevin Kenny wrote: > A great many tests are slower than uncompiled, probably because it takes > us longer to construct and push a callframe than it does the Core. I'm still at the “get it working at all” stage with callframes. The expensive part right now is that we're building a hashtable of variables on entry instead of using a LVT. The LVT will be (much!) cheaper, but needs me to finish off doing the metadata or it will just blow up in our faces. The hashtable approach was simpler to get started with. So yes, slower is expected. Donal. |