Commits

Breadcrumbs

History for

onv0.9.0

User selector

All users

Datepicker

All time

Commit History

Commits on Jul 6, 2018

• Add functions for returning peak memory usage for the different memory spaces

  maddyscientist

  committed
  fc74876

  Copy full SHA for fc74876

Commits on Jul 4, 2018

• Merge pull request #705 from lattice/feature/jenkins_ctest_gtest

  Show description for 445cb6b
  maddyscientist

  authored
  445cb6b

  Copy full SHA for 445cb6b

Commits on Jun 29, 2018

• fixed typos, clean up, change volumes for ctest

  mathiaswagner

  committed
  d5ec6c0

  Copy full SHA for d5ec6c0

• add some doxygen comments and allow ctest to be called from the build dir directly

  mathiaswagner

  committed
  9d55828

  Copy full SHA for 9d55828

Commits on Jun 28, 2018

• needs some cleanup but seems to work mostly ok

  mathiaswagner

  committed
  c742716

  Copy full SHA for c742716

Commits on Jun 26, 2018

• Remove legacy HISQ and ASQTAD force routines from QUDA.

  maddyscientist

  committed
  43c6db3

  Copy full SHA for 43c6db3

• Merge branch 'feature/dslash_constant_cleanup' of github.com:lattice/quda into feature/milc-optimize

  maddyscientist

  committed
  8ba4303

  Copy full SHA for 8ba4303

Commits on Jun 25, 2018

• Only include cudaFuncSetAttribute wrapper when compiling on >= CUDA 9.0

  maddyscientist

  committed
  24594d2

  Copy full SHA for 24594d2

Commits on Jun 23, 2018

• dslashParam is no longer static, and now a member of DslashCuda

  maddyscientist

  committed
  12f438e

  Copy full SHA for 12f438e

Commits on Jun 22, 2018

• Only call ColorSpinorField::createGhostZone when needed

  maddyscientist

  committed
  6468afe

  Copy full SHA for 6468afe

• Precompute the dslash constants in ColorSpinorField::createGhostZone and store in DslashConstant struct. Use a copy of same struct in DslashParam and PackParam for rapid setting of these parameters.

  maddyscientist

  committed
  bb33c83

  Copy full SHA for bb33c83

Commits on Jun 21, 2018

• Check error state on call to cudaLaunchKernel when not kernel tuning. This allows us to remove the separate check at the end of each dslash

  maddyscientist

  committed
  d5c7da7

  Copy full SHA for d5c7da7

• Added API wrappers for cudaLaunchKernel and cudaFuncSetAttribute

  maddyscientist

  committed
  13cfc3b

  Copy full SHA for 13cfc3b

Commits on Jun 20, 2018

• Moved Mobius variables into DslashParam and removed initMDWFConstants

  maddyscientist

  committed
  afaae3f

  Copy full SHA for afaae3f

Commits on Jun 12, 2018

• Fix some minor issues with CUDA API wrapper - missing D2D async wrapper, and unnecessary pass by value which could increase wrapper overhead

  maddyscientist

  committed
  8f26733

  Copy full SHA for 8f26733

Commits on Jun 7, 2018

• Merge branch 'feature/p2p-write-optimize' of github.com:lattice/quda into feature/milc-optimize

  maddyscientist

  committed
  db55d59

  Copy full SHA for db55d59

Commits on Jun 6, 2018

• Extended grid-size tuner to expose a minimum grid size (used for quality of service when running concurrent kernels)

  maddyscientist

  committed
  8ca3c55

  Copy full SHA for 8ca3c55

• Only set ghost parameters for dslash kernel in first (interior) application - fixes a performance regression

  maddyscientist

  committed
  8d7d394

  Copy full SHA for 8d7d394

Commits on Jun 3, 2018

• Add qudaDeviceSynchronize wrapper to cudaDeviceSynchronize to faciltate API profiling and to switch to driver API equivalent cuCtxSynchronize

  maddyscientist

  committed
  359ed54

  Copy full SHA for 359ed54

• Add CUDA API profiling. Not enabled by default (enabled with API_PROFILE macro in quda_cuda_api.cpp). When enabled, the API profile will be printed during endQuda

  maddyscientist

  committed
  5fe1d2a

  Copy full SHA for 5fe1d2a

Commits on Jun 1, 2018

• Add new include file

  maddyscientist

  committed
  6d570b4

  Copy full SHA for 6d570b4

• Renamed quda_memcpy.cpp into quda_cuda_api.cpp and added wrappers to commonly used stream and event API calls to allow easy redirection to using driver or runtime API. By default we now use driver …

  Show description for 5fc8ef8
  maddyscientist

  committed
  5fc8ef8

  Copy full SHA for 5fc8ef8

Commits on May 9, 2018

• QUDA_ENABLE_P2P promoted to 2-bit number, used for enabling dslash p2p policies - first bit to enable copy engines, second bit for enabling load-store access

  maddyscientist

  committed
  3c3b075

  Copy full SHA for 3c3b075

Commits on May 7, 2018

• Fix potential race condition with remote halo writing: p2p event should be posted to packing stream not the direction streams

  maddyscientist

  committed
  9a26d1f

  Copy full SHA for 9a26d1f

Commits on May 4, 2018

• Add virtual method Tuneable::maxGridSize()

  maddyscientist

  committed
  39c95a1

  Copy full SHA for 39c95a1

Commits on May 3, 2018

• First attempt at direct remote p2p writing of halo buffers for dslash kernels

  maddyscientist

  committed
  cbd08fb

  Copy full SHA for cbd08fb

Commits on Mar 26, 2018

• Disable multi-blas support when running without texture objects - fix for Fermi

  maddyscientist

  committed
  6103a2d

  Copy full SHA for 6103a2d

Commits on Mar 23, 2018

• Fixes for Fermi multi-GPU with texture references

  maddyscientist

  committed
  dd7621f

  Copy full SHA for dd7621f

Commits on Mar 10, 2018

• Remove redundant parameter from computeHISQForceQuda interface

  maddyscientist

  committed
  16de6b0

  Copy full SHA for 16de6b0

Commits on Mar 9, 2018

• Fusion of outer-product into the computeHISQForceQuda function - this reduces memory traffic significantly (requires update MILC code)

  maddyscientist

  committed
  2d28842

  Copy full SHA for 2d28842

• Add ax function for rescaling gauge fields. Added overloaded cudaGaugeField::loadCPUField and cudaGaugeField::saveCPUField with built in profiling. Fill in some missing cases for copying between ex…

  Show description for 84413a7
  maddyscientist

  committed
  84413a7

  Copy full SHA for 84413a7

Commits on Mar 7, 2018

• Add doxygen for ks_improved_force.h

  maddyscientist

  committed
  751e400

  Copy full SHA for 751e400

Commits on Mar 3, 2018

• Unify the HISQ fat-link force Tunable class launchers

  maddyscientist

  committed
  2e1b87e

  Copy full SHA for 2e1b87e

Commits on Mar 1, 2018

• More clean up of HISQ force

  maddyscientist

  committed
  a4061c4

  Copy full SHA for a4061c4

Commits on Feb 28, 2018

• Removal of legacy ASQTAD and HISQ force kernels and associated testing

  maddyscientist

  committed
  ef2bc21

  Copy full SHA for ef2bc21

Pagination

Previous

Next