QUDA 0.5.0 is now live.

[Clark, Michael]

The QUDA developers are pleased to announce release 0.5.0, available at http://attice.github.com/quda.

It has been almost a year since QUDA 0.4.0 was released, and many new features have been added in the interim, including multi-GPU domain wall support, much improved mixed-precision CG and multi-shift CG solvers and general library improvements.  Please report any issues at http://github.com/lattice/quda/issues (requires a GitHub account) or by emailing quda-de...@googlegroups.com.

Enjoy!

QUDA: A library for QCD on GPUs

QUDA is a library for performing calculations in lattice QCD on graphics processing units (GPUs), leveraging NVIDIA's CUDA platform. The current release includes optimized Dirac operators and solvers for the following fermion actions:

• Wilson
• Clover-improved Wilson
• Twisted mass
• Improved staggered (asqtad or HISQ)
• Domain wall

Implementations of CG, multi-shift CG, BiCGstab, and DD-preconditioned GCR are provided, including robust mixed-precision variants supporting combinations of double, single, and half (16-bit "block floating point") precision. The library also includes routines for HISQ link fattening and force terms for the HISQ fermion action and one-loop improved Symanzik gauge action. Use of many GPUs in parallel is supported throughout, with communication handled by QMP or MPI.

Several commonly-used packages integrate support for QUDA as a compile-time option, including Chroma, MILC, CPS, and BQCD (in a specific branch available here). 

[Clark, Michael]

Please note the there was a typo in the original email.  The correct quda home page is of course http://lattice.github.com/quda.

--
You received this message because you are subscribed to the Google Groups "QUDA announce" group.
To unsubscribe from this group and stop receiving emails from it, send an email to quda-announc...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.