Peter Nimmervoll

IIRC this patch did not break any tests. There must be something else going on.

Yes, this should be good, @dylanmckay just need to verify it and upstream it.

> Also if I understand the meaning of earclyclobber well then there is a disagreement between LDWRdPtr being marked as earlycloober and the corresponding expansion template specialization which checks and...

If you compile llvm with asserts you will hit that and don't run into OOM error. I debugged this in the past and IIRC the problem is because 2 nodes...

There are glue nodes which glues 2 (or more?) nodes together While the IR code does not have a cycle, llvm wrongly fuses 2 instructions into 1 node which leads...

Imho, and i might be wrong, this is flaw in llvm. Either way I think you should discuss this with the llvm developers. Also, when I looked up the code...

I don't have any plans to implement this feature myself. As far as I understand this feature, it is not much work to implement it but a lot of trouble...

I can reproduce this, looking into it.

After playing around a bit the problem seems to be that the messages are too small. So the server basically waits to get enough data that compress actually has an...

We use Resteasy and from my experience the JAX-RS filters are always called if the resource method is called. Resteasy has 'unhandled' exceptions which would increment this metric, but I...