Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Dec 8, 2016

• aligned_malloc() added for systems that do not have aligned_malloc in stdlib.h

  ap-hynninen

  committed
  4c251c6

Commits on Dec 7, 2016

• Altivec vectorization implemented

  ap-hynninen

  committed
  ce1c74c

Commits on Dec 2, 2016

• Faster CPU algorithms. Easier vector library. Cache for frequent CUDA API

  Show description for acb7958
  ap-hynninen

  committed
  acb7958

Commits on Nov 24, 2016

• Merge branch 'master' into dev

  ap-hynninen

  committed
  2ac2096

• Fixed bugs in vectorized CPU code

  ap-hynninen

  committed
  0d0a767

Commits on Nov 23, 2016

• increased data buffer size

  ap-hynninen

  committed
  96f4d72

• Optimized plan creation

  ap-hynninen

  committed
  b5c834d

Commits on Nov 19, 2016

• Update README.md

  ap-hynninen

  authored
  2c74e86

• Add files via upload

  ap-hynninen

  authored
  0fb6e6a

Commits on Oct 18, 2016

• 2.5 -> 2.8

  ap-hynninen

  committed
  188ca3b

• Pascal GPU model implemented. L1 switched on in Makefile

  ap-hynninen

  committed
  791dec0

Commits on Oct 12, 2016

• Added support for Pascal in Makefile. Changes to bench_memcpy

  ap-hynninen

  committed
  604887c

Commits on Oct 6, 2016

• Removed MATLAB printout

  ap-hynninen

  committed
  cbbd73a

• Maxwell GPU performance model parameters updated

  ap-hynninen

  committed
  d508aaf

Commits on Oct 5, 2016

• New TiledCopy kernel. Improved performance model

  ap-hynninen

  committed
  3252865

Commits on Oct 3, 2016

• Added single precision versions to many benchmarks. sizeofType now only input once in cuttplant_t -class.

  ap-hynninen

  committed
  93ee5f9

Commits on Oct 2, 2016

• Changed method names to Tiled, "TiledCopy, Packed, and PackedSplit. Changed to using only cycles in heuristic selection

  ap-hynninen

  committed
  6f4d0aa

Commits on Sep 30, 2016

• Increased dataSize to fit bench 5 into memory

  ap-hynninen

  committed
  9a577af

• New bench5: 500 iterations on every rank

  ap-hynninen

  committed
  ba93fda

Commits on Sep 15, 2016

• Major new version: new GPU model, GPU transaction counting

  ap-hynninen

  committed
  443bc62

Commits on Aug 25, 2016

• Shared memory bank config added to device info

  aphynninen

  committed
  819419d

Commits on Aug 24, 2016

• Changed GPU model

  aphynninen

  committed
  cc8209a

• Device info added

  aphynninen

  committed
  6a5cf0b

• Duplicate plan-generation removed. GeneralSplit created in General -subroutine

  aphynninen

  committed
  7c1cd19

Commits on Aug 22, 2016

• New algorithms. New MWP GPU Model that works.

  aphynninen

  committed
  80e6667

Commits on Aug 12, 2016

• Benchmarks changed. cuttTimer: now stores all bandwidth data to vector. Added "getData()"

  aphynninen

  committed
  b609edb

Commits on Aug 11, 2016

• Disabled "mixed" algorithms. Switched to timer based on std::chrono

  aphynninen

  committed
  2f5f0be

Commits on Aug 9, 2016

• Added median calculation and some missing #includes

  aphynninen

  committed
  bbe9551

• CUDA compiler flag added for Ubuntu 16.04

  aphynninen

  committed
  dba8966

Commits on Jul 25, 2016

• CUDA stream moved to planning phase

  aphynninen

  committed
  dd5f5ac

Commits on Jul 22, 2016

• Tried rank reduction, did improve performance. tiled kernels blockIdx.y limitation lifted.

  aphynninen

  committed
  2c77bbd

Commits on Jul 21, 2016

• Better heuristics

  aphynninen

  committed
  3668850

Commits on Jul 13, 2016

• New transpose methods added: TiledSingleInRank, TiledSingleOutRank, GeneralSplitInRank, GeneralSplitOutRank.

  aphynninen

  committed
  6fee322

Commits on Jul 5, 2016

• Fixed memory leak

  aphynninen

  committed
  ff6e6e6

Commits on Jul 1, 2016

• Updated docs and benchmarks

  aphynninen

  committed
  73e3367

Pagination

PreviousNext