[Hadstruct] Wrong results with redstar meta branch

[Eloy Romero]

Hi all,

We have solved all known issues for the meta branch of redstar. And all 
the new data generated by me needs to be read with that branch.

The "next" branch has been archived. And the newest branch that Robert, 
Jie, and me work on is meta.

We aren't happy about the gpu performance of redstar, but we are working 
on that.

For chromaform:
./chromaform --mpi=no redstar # ok
./chromaform --mpi=no redstar --pdf # ok
./chromaform --mpi=no redstar --next # new installations will fail
./chromaform --mpi=no redstar --meta # ok

Cheers,
Eloy

On 3/28/25 10:00, Eloy Romero wrote:
> Hi all,
> 
> We have detected that the new meta branch of redstar is giving wrong 
> results. For now, please DO NOT compute correlation functions with the 
> meta branch of redstar. adat and its tools (eg dbutil) are ok to use. 
> The recommended branches for adat, colorvec and redstar are devel, 
> devel-pdf and eloy/mix-phasings-with-momclass, and devel for hadron.
> 
> For chromaform:
> ./chromaform --mpi=no redstar # ok
> ./chromaform --mpi=no redstar --pdf # ok
> ./chromaform --mpi=no redstar --next # ok
> 
> ./chromaform --mpi=no redstar --meta # don't do it for now
> 
> I'll keep you posted.
> 
> @David, you don't have a choice for now, mix-phasing support for mesons 
> is only implemented in the new meta branch. But the cpu version seems to 
> be correct.
> 
> Cheers,
> Eloy