ATLAS Changelog

What's new in ATLAS 3.11.23

Feb 5, 2014
  • Got basic cross-compilation working when host/target share filesystem and you can ssh w/o passwd to target: --rtarg=
  • Added basic support for XeonPHI working: --accel=2 --rtarg=mic0 - Expects icc in path and ready for use
  • Added SIMD vect support for MIC's AVX-512, called AVXZ in ATLAS
  • Added basic AMM AVXZ code generator -> presently disabled in search due to errors.
  • Fixed bugs in ammsearch where very large mu cause M/NB=0
  • Added PHI-specific assembly kernels

New in ATLAS 3.11.22 (Jan 25, 2014)

  • Changed it so complex block-major gemm installed for non-default installs
  • Changed it so ARM block-major gemm kernels default to HARDFP ABI
  • Added NB tuning for complex access-major gemm
  • Uglied up atlas_install to avoid gcc's unalterable BS warnings
  • Updated archdefs for Corei364AVXMAC
  • Plugged several one-time mem leaks in lanbsrch
  • Added basic config support for cross-compilation
  • Updated complex cmat2blk to correct prototype & type def for complex
  • Rakib wrote cmat2blk complex
  • Changed emit_uamm to handle multiple installs
  • Boatload of TI_C99_BM accelerator patches from Tony Castaldo

New in ATLAS 3.10.0 (Jul 12, 2012)

  • This version fixes rare bugs in the ATL_syr, ATL_her, ATL_syr2, and ATL_her2 L2 BLAS functions, adds GEMV accelerators for TRMV, TRSV, SYMV, and HEMV for large problems, adds code to recognize IvyBridge as Corei2, applies fixes for ARM GEMM kernels, updates documentation, and tweaks the Windows DLL build.

New in ATLAS 3.9.81 (Jun 30, 2012)

  • Dynamic libraries now build correctly for all supported configurations for Windows, OS/X, and Linux.

New in ATLAS 3.9.80 (Jun 25, 2012)

  • This release builds under mingw for the Windows 32bit platform, adds assorted bugfixes, and removes unsupported archdefs.

New in ATLAS 3.9.78 (Jun 12, 2012)

  • This version added the ability to specify which processors to use when tuning and fixes issue with virtual processors.
  • Bugs in POWER7 support were fixed, and all compliance checks now pass.
  • GCC 4.7.0 support was improved.
  • ATLAS can now be compiled with MSVC++.
  • ARM assembly was cleaned and ARM NEON support was improved.
  • Basic support was provided for Intel's compilers.
  • Threads and 64-bit support in Windows are now both native and longer need Cygwin.

New in ATLAS 3.9.63 (Feb 1, 2012)

  • This version updates the API of the test harness to work with current LAPACK.
  • It reworks LAPACK bindings accordingly (which breaks compatibility with versions of LAPACK prior to 3.4.0).
  • There are assorted bugfixes.

New in ATLAS 3.9.56 (Dec 9, 2011)

  • Fixed to work with LAPack 3.4.0.
  • Fixes for buffer overruns in configuration.
  • The addition of the AVX cgemvT kernel.
  • New Corei264AVX architecture definitions for gcc 4.6.2.

New in ATLAS 3.9.54 (Nov 18, 2011)

  • This version improves automatic selection of compilers and libraries, to produce the best results.
  • It adds further development and bugfixes to the multithreading support.

New in ATLAS 3.9.52 (Oct 4, 2011)

  • Improved complex TRSM performance, particularly for small L/U, large RHS
  • Fixed bug in complex ATLAS/tune/blas/level3/invtrsm.c
  • Accepted series of patches & arch defs to add ATLAS support for IBM Z9, Z10, and z196 mainframe computers. Patches submitted by Christian Borntraeger of IBM.

New in ATLAS 3.9.48 (Sep 2, 2011)

  • Performance improvements to Real TRSM.
  • Architectural defaults for Atom64SSE3.
  • A bug in gerk assembly has been fixed.

New in ATLAS 3.9.47 (Aug 14, 2011)

  • QR-related routines were cleaned up.
  • Improved performance for parallel LU and QR and for serial LQ and RQ.
  • Numerous new platform defaults.
  • Many bugfixes.

New in ATLAS 3.9.40 (Apr 28, 2011)

  • This version adds beta versions of simple threaded GEMV & GER.
  • It adds threaded L2 testing to the tester.
  • It fixes a bug in axpby where it called SCAL with alpha=0, which fixes GEMM error for BETA=0 case.
  • It fixes several simple buffer overruns in the full tester.
  • It adds dynamically scheduled tgemm, which is used whenever all dimensions are large.
  • It adds support for complex types for both dynamic cases (rank-K, large).
  • It fixes several errors in GEMM that occurred when K dim was cut.

New in ATLAS 3.9.39 (Mar 21, 2011)

  • Basic AVX GEMM kernels and new Corei264AVX arch defs.
  • Now use dynamically scheduled parallel rank-K updates
  • Complete rewrite of all threaded routines to use goparallel, and thus dynamic spawn.
  • OpenMP now uses same codebase as windows & pthreads forall threading.
  • Thread tune now creates atlas_tsumm.h for summation of threaded tuning
  • Added ATL_thread_yield function
  • If affinity is not set, dynamic funcs now yield thread execution when waiting for their peers to signal completion of a stage
  • Otherwise, active poller prevents thread running on same core from exec

New in ATLAS 3.9.35 (Feb 10, 2011)

  • Improved affinity setting.
  • Fixes for numerous bugs.

New in ATLAS 3.9.33 (Jan 26, 2011)

  • Extensive bugfixes were made.
  • Preliminary support was added for threaded/parallel lapack.
  • The architecture definitions were updated for P4ESSE3, PPCG564AltiVec, Core264SSE, and AMD64K10h64SSE3.
  • PCA codes were added for LU and QR.

New in ATLAS 3.9.15 (Oct 13, 2009)

  • A new tuning mechanism tunes GER for in-L1, in-L2, and out-of-cache.
  • The C2F LAPACK interface is now only built if F77 LAPACK is provided.
  • Chad Zalkin's SSE GEMM generator has been added to ATLAS.
  • There are numerous bugfixes, including off-by-one bugs.

New in ATLAS 3.9.10 (Mar 13, 2009)

  • Bugs were fixed in Xtsyrk's ATL_tsyrkdecomp_K, EnforceNonPwr2LO, and lanbtst.
  • atlas_install was changed to have a final library build phase.
  • Missing lapack arch defs were added for Corei764 and MIPSICE9.
  • tgemm's combine routine was rewritten to work on arbitrary partitionings combined in arbitrary orders.
  • A restricted fix for SYRK was made.
  • A makefile problem that caused ATLAS to re-archive the L3BLAS for every tester compile was fixed.

New in ATLAS 3.9.8 (Feb 24, 2009)

  • This release fixes a major bug involving systems where the core count is not a power of 2.
  • It fixes a bug in ATL_tgemm that caused segmentation faults for some small-M tGEMMs.
  • It introduces some fixes and workarounds for SiCortex/MIPSICE9.
  • It adds architectural defaults for K7323DNow.

New in ATLAS 3.9.5 (Dec 23, 2008)

  • A complete rewrite of the ATLAS threading system so as to include support native Windows threads and processor affinity, along with other code speedups.
  • Fixes for building errors under Windows.
  • Dependency info for ATL_ilaenv so that it is recompiled once lapack tuning is complete.
  • A fix for an error in configure where it issued commands in the wrong directory when the user built lapack directly from a tar file.
  • Detection of kernels that do well in parallel operations.
  • Assorted bugfixes.

New in ATLAS 3.9.4 (Oct 19, 2008)

  • Improved Windows/Cygwin configure.
  • Basic support for Windows/interix.
  • Special "no-need-to-copy" cases have been removed from ATLmm_JIK/IJK.c, since they occasionally seem to cause large performance drops.
  • JIK matmul is always called for rank-K update, in order to reduce access costs on C.
  • Several errors have been fixed in ATLAS's ILAENV.
  • Several errors in configure have been fixed.