Evan Weinberg

Just an update, as a quick test I saw if I could reproduce the behavior on a smaller lattice (64^3x96) and I was unsuccessful, so now I'm going to try...

I may have found the issue, will post back soon.

I've reproduced the behavior on a 144^3 configuration, for both an odd _and_ even source. It seems like the logic for heavy quark residual reliable updates is breaking down in...

I've put together a fix for the heavy quark convergence issue that at least works in the cases where I've been able to trigger the issue. Can you please test...

> I guess the main difference make using `__noinline__` vs `inline`? The overall difference is keeping the compiler from wasting time on code it can never reach, which both `[[noreturn]]`...

> When you say partial fix, are you referring to the 25% increase in compile time that we still observe when `Kokkos_ENABLE_DEBUG_BOUNDS_CHECK` is `ON` ? At a high level, yes....

> Blocking until we tested on pre-volta It's a WIP. There's a test when I do a full compile that's taking "forever" to compile for `sm_60` with debug bounds enabled...

The following `diff` fixes the compile time issues specifically in `Kokkos_CoreUnitTest_Default` . However, it looks like there are still major slowdowns in compiling other unit tests (so I imagine, by...

I'll open a PR with the patch above today so we can ideally get that in.

Thanks @paboyle, I appreciate the references!