Ye Luo

In QE, EPW code needs wannier90 as a library right now internally. When it is used as an internal library, QE just download/clone and build wannier90 inside QE. This works...

I checked out 183bf9b34b7a and found compileDeviceRTLLibrary is not at line 252. Could you check your source code?

https://github.com/llvm/llvm-project/commit/053e894106beb534ec27010b960ee58031a2fdb8 should have fixed the regression caused by https://github.com/llvm/llvm-project/commit/92233159035d1b50face95d886901cf99035bd99 at least in my use case. Since I changed add_custom_command to add_custom_target, it is likely your patch needs to be updated...

I saw `-- LIBOMPTARGET: Not building DeviceRTL. Missing clang: /build/llvm-toolchain-snapshot-16~++20221014091649+ae27323428ff/build-llvm/bin/clang, llvm-link: /build/llvm-toolchain-snapshot-16~++20221014091649+ae27323428ff/build-llvm/bin/llvm-link, opt: /build/llvm-toolchain-snapshot-16~++20221014091649+ae27323428ff/build-llvm/bin/opt, llvm-extract: EXTRACT_TOOL-NOTFOUND, or clang-offload-packager: /build/llvm-toolchain-snapshot-16~++20221014091649+ae27323428ff/build-llvm/bin/clang-offload-packager` Probably you need to first figure out why `llvm-extract: EXTRACT_TOOL-NOTFOUND,` Try...

What is the error message in stderr file?

After writing this, I feel more comfortable to make `samples` interacting with `blocks` rather than `steps`

@lshulen from my perspective, substeps are much cheaper to decorrelation samples and this feature is kept in batched VMC/WFOpt drivers. That was the motivation to remove `steps_between_samples`. Do you still...

Here is `substeps` from the manual https://qmcpack.readthedocs.io/en/develop/methods.html#batched-vmc-driver-experimental `substeps` For each substep, an attempt is made to move each of the electrons once only by either particle-by-particle or an all-electron move....

I consider "samples", "total_samples" interchangeable words in this discussion. It is physically related to sampling accuracy and can be directly translated from the latter if we switch to an input...

Probably I gave the wrong impression. What I meant was that `walker_per_rank` or `total_walkers` were predetermined before handing sample input and they should not be part of the interaction with...