benchmarks - testing and newfound failures

[Justin DeVuyst]

A few days ago I noticed that the vpm benchmark was failing. Then
today, I saw that the majority of the OO benchmarks were failing.
I'd like to propose including testbench under fulltest so that the
benchmarks can be more readily tested. Maybe that will help keep
the benchmarks from getting broken so easily.

I've included a patch that fixes the vpm bench. Seems the recent
changes to split broke it.

I didn't work out the OO bench fixes yet. Anyone else wanna take a
stab at it? Its most likely ridiculously simple. Actually, I have
a related question. Since the current object isn't in P2 anymore
and "self" in PIR replaces it, what is the equivalent in pasm? That
looks like what's breaking the OO benches right now.

Thanks,
Justin DeVuyst

[Leopold Toetsch]

Justin DeVuyst <jus...@devuyst.com> wrote:

> A few days ago I noticed that the vpm benchmark was failing. Then
> today, I saw that the majority of the OO benchmarks were failing.

Yeah. Fixed.

> I'd like to propose including testbench under fulltest so that the
> benchmarks can be more readily tested. Maybe that will help keep
> the benchmarks from getting broken so easily.

Yep. But first I've removed a bunch of benchmarks: hash-utf8 used
non-existent opcodes, b6* used python methods which aren't in core.

WRT including under fulltest: the benchmarks.t should patch the files so
that they don't run too long.

> I've included a patch that fixes the vpm bench. Seems the recent
> changes to split broke it.

Thanks, applied
leo