`tmEval lazy` is slow

[annenkov]

I've noticed that even with the fastest reduction strategy available - lazy - running tmEval within the TemplateMonad is still much slower than vm_compute. I have an example for which tmEval lazy takes around 22 sec, and vm_compute finishes in 4.5 sec.

Is there a way of using vm_compute for reduction?

[mattam82]

That's to be expected indeed. We should indeed have a way to call vm_compute and not just vm_cast.