Code

Feature/quda interface optimize#15

Code

Merged

detar merged 25 commits intomilc-qcd:developmilc-qcd/milc_qcd:developfrom lattice:feature/quda-interface-optimizelattice/milc_qcd:feature/quda-interface-optimizeCopy head branch name to clipboardOct 4, 2017

+414-177

Commits

Commits on Aug 13, 2015

• Merge branch 'feature/no_cpu_refine' into develop

  maddyscientist

  committed
  9edc3cd

Commits on Oct 5, 2016

• Updated Makefile for QUDA 0.9

  Show description for f957855
  maddyscientist

  authored
  f957855

Commits on Oct 11, 2016

• Merge pull request #7 from maddyscientist/patch-1

  Show description for 6798ef4
  detar

  authored
  6798ef4

Commits on Dec 1, 2016

• Merge pull request #2 from milc-qcd/master

  Show description for ae975d9
  maddyscientist

  authored
  ae975d9

• Merge branch 'master' into develop

  maddyscientist

  committed
  4a90232

• Merge pull request #3 from milc-qcd/develop

  Show description for a67218e
  maddyscientist

  authored
  a67218e

Commits on Dec 9, 2016

• Merge pull request #4 from milc-qcd/develop

  Show description for ab52202
  maddyscientist

  authored
  ab52202

Commits on Dec 12, 2016

• Merge pull request #5 from milc-qcd/develop

  Show description for 5a17bcc
  maddyscientist

  authored
  5a17bcc

Commits on Sep 19, 2017

• Merge pull request #6 from milc-qcd/develop

  Show description for f611390
  maddyscientist

  authored
  f611390

Commits on Oct 1, 2017

• Enable -Wall (with some opt outs), and fix all generated warnings

  maddyscientist

  committed
  7825122

• Fix couple of things in last commit

  maddyscientist

  committed
  188464f

Commits on Oct 2, 2017

• Add QUDA specific routines for creating gauge and momentum fields, and for extracting and inserting to/from the site array

  maddyscientist

  committed
  88e0340

• Cleanup of QUDA gauge force and link update routines to use helper functions introduced in 88e0340

  maddyscientist

  committed
  d7ed3b4

• Offload reunitarization to QUDA if gauge-force offload is enabled. Fallback to CPU if Shcroedinger functional is set

  maddyscientist

  committed
  6f23a19

• Remove legacy use of use_pinned_memory member in QudaFatLinkArgs_t

  maddyscientist

  committed
  06aa7b6

• Use pinned memory for ASQTAD GPU outer product vectors

  maddyscientist

  committed
  a399776

• When creating/restoring fermion links on GPU, use QUDA helpers for extracting the gauge-field matrices

  maddyscientist

  committed
  814f145

• Use pinned memory allocation for the auxilary and fermion link-field arrays when USE_FL_GPU is enabled

  maddyscientist

  committed
  ef7caee

• Cleanup: put load_hisq_aux_links_gpu into fermion_links_fn_load_gpu.c and remove fermion_links_hisq_load_gpu.c. Fix flop counter when USE_FL_GPU is enabled.

  maddyscientist

  committed
  44ea6d9

• Fix bug in gauge force introduced in prior cleanup in d7ed3b4

  maddyscientist

  committed
  e05bd59

• Minor cleanup of generic_quda.h

  maddyscientist

  committed
  2e773a9

• Fix warning in dslash_fn.c

  maddyscientist

  committed
  34c0500

Commits on Oct 3, 2017

• If using QUDA, do not use double-store dslash since this adds unnecessary overhead in the fermion-link construction

  maddyscientist

  committed
  d24fd72

• QUDA-accelerated multi-shift and regular CG HISQ solvers now share the same fn_last cache to avoid unnecessary invalidation when switching between CG and multi-shift CG solvers

  maddyscientist

  committed
  c50ef50

• When compiling with HISQ CG GPU offload, also offload all calls to dslash to QUDA

  maddyscientist

  committed
  a7654a0