Home Explore Help
Register Sign In
lloda
Repositories Public Activity

lloda pushed to main at lloda/ra-ra

  • e801b4666d Guard nan/inf checks in test suite * ra/test.hh (TestRecorder::amax_strict, test_rel, test_abs): As stated, since --ffast-math breaks them otherwise.

1 week ago

lloda pushed to main at lloda/guile-ffi-blis

  • dcf55ec877 Bindings for bli_?amaxv
  • 724351d7f3 Bindings for bli-?norm1v bli-?normfv bli-?normiv
  • 7c2e475b40 Bindings for bli_?swapv
  • 814a00b9f4 Fix docstring for blis-?gemv
  • 110c9426e0 Typed constants for trans_t conj_t side_t uplo_t diag_t These have hyphened names (BLIS-TRANSPOSE) and are used in the typed and functional interfaces. The raw interface still uses numeric constants which keep the C underscored names (BLIS_TRANSPOSE).

1 week ago

lloda created new branch main at lloda/guile-ffi-blis

1 week ago

lloda pushed to main at lloda/guile-ffi-cblas

  • 2c0f73ab32 Remove srfi-4-type-size Use instead srfi-4-vector-type-size from (srfi srfi-4 gnu).
  • 4458d50f84 Generalize srfi4-type-size in (ffi cblas arrays)
  • 9956d613c5 Prefix all typed exports with cblas-
  • 82d3d3ab10 Remove BLIS bindings and fix references to guile-ffi-blis
  • b1c60a322f Simplify scalar->arg

1 week ago

lloda created new branch main at lloda/guile-ffi-cblas

1 week ago

lloda pushed to main at lloda/guile-ffi-fftw

  • 90d5da6583 Have fftw-dft! return the output array Also fixes to the docstrings.
  • b2cda149f9 Fix test for srfi-64 fix in Guile
  • 294ad9e749 Fix license mismatches
  • 95d7ffb558 Add license
  • 3d42ab7f08 Test instructions don't need a file This is parallel to the presentation of guile-ffi-cblas. Also, mention FFTW-FORWARD and FFTW-BACKWARD in the docstrings.

1 week ago

lloda created new branch main at lloda/guile-ffi-fftw

1 week ago

lloda pushed to main at lloda/ra-ra

  • a2292645c4 Add options --no-sanitize, -DSANITIZE=0 to the test suite To fix a problem where the sanitizers couldn't be completely disabled by appending -fno-sanitize=all to CXXFLAGS. * SConstruct, */SConstruct: New scons options --blas, --sanitize, --no-sanitize. * CMakeLists, */CMakeLists: New cmake variable SANITIZE (as in cmake ... -DSANITIZE=0). * vasu.texi.in (Building the test suite): Explain the new options. (Using the library): Move the doc on #defines here.

1 week ago

lloda pushed to main at lloda/slatec-bessel-cpp

  • 86d3d0794b Fixes for gcc 13.1 Old fortran spaghetti provoked gcc into bogus maybe-used-uninitialized warnings.
  • 845feba253 Simplify 'constexpr inline' to just 'constexpr'
  • 326c2456f7 Fixes for gcc 12.2 * Makefile: More strict flags. * f2c/*: Fix maybe-uninitialized warnings.
  • 1140750c3b Move folders
  • f00c6bf720 Initial commit

1 week ago

lloda created new branch main at lloda/slatec-bessel-cpp

1 week ago

lloda pushed to main at lloda/guile-raw-strings

  • fd299e3469 Fix license in test.scm
  • ce36d59356 Create LICENSE
  • c8f752159d Add GH CI
  • 47feacdc8c Make configuration options more explicit * mod/raw-strings.scm: Offer configuration in variables at the top of the file. * README.md: Explain the configuration options.
  • d8d0d985ce Remove .travis.yml That hasn't worked for a while.

1 week ago

lloda created new branch main at lloda/guile-raw-strings

1 week ago

lloda pushed to main at lloda/guile-newra

  • 1439f5b089 Relax FindGuile.cmake * config/FindGuile.cmake: Do not require GUILE_ROOT_DIR since I'm not using it. Even though guile-config is available, the pkg-config line with --variable=prefix fails somehow...
  • 02fd1de263 Add global CMakeLists.txt and run maxwell.scm as part of the test suite * CMakeLists.txt: As stated. * test/CMakeLists.txt: Fix to work as main or sub. Run maxwell.scm as a test. * examples/maxwell.scm: Exit as in a test. * test/maxwell.scm: Link.
  • 8338db587e Make test/format.scm into an actual test * test/CMakeLists.txt: Test driver. * README.md: Describe test suite operation. * cofig/FindGuile.cmake: Needed for test/CMakelists.txt. * test/format.scm: As stated; check the output sizes. * test/everything.scm: Renamed from test/test.scm.
  • c8d50089fc Options #:compact 0 1 2 for ra-format * mod/newra/print.scm (ra-format): Handle compact 0/1/2. (*ra-print*): New options. test/format.scm: Exercise. Elsewhere fix users.
  • c407a3fb12 Disable CI until guile-3.0.8 is available

1 week ago

lloda created new branch main at lloda/guile-newra

1 week ago

lloda pushed to main at lloda/ra-ra

  • db86177fc9 Replace macros in bench/bench-gemm.cc
  • 4a1d23933a Fix for io segfault under gcc11 The segfault only happened with sanitizers off, -O3, and -DRA_DO_CHECK=1, and only on gcc11; gcc12 and gcc13 work fine, or any -O<3 in gcc11. The error went away when i removed random combinations of lines in test/io.cc and I couldn't reduce it. Go figure. * ra/ply.hh (operator<<): As stated.
  • 9e7778a3ce Default RA_DO_CHECK to 1 * ra/expr.hh: As stated. * docs/ra-ra.texi: Reword explanation.
  • fa08040100 Option for simplified error reporting I've found that the printing of informative args might slow down checks, even though those arguments aren't evaluated when the check passes :-/ So, provide the option to ignore those args but still do the checks. * ra/base.hh (VERSION): Bump to 28. * ra/expr.hh (RA_DO_CHECK): Provide variants of RA_ASSERT depending on value, if not 0. * docs/ra-ra.texi (Error handling): Explain the new options.
  • 44b9ffcafd Clean up checks * ra/big.hh (ViewBig): Fix typo in conversion to scalar. * ra/expr.hh (RA_ASSERT): Use __VA_OPT__ instead of ##__VA_ARGS__. (Match): Print full shapes on match error. * test/checks.cc, test/throw.cc: Use __VA_OPT__ instead of ##__VA_ARGS__. * test/SConstruct, test/CMakeLists.txt: Don't run ra-16.cc, which causes gcc11 to hang or give spurious errors. Tbf.

1 week ago

lloda created new branch main at lloda/ra-ra

1 week ago

lloda pushed to master at lloda/ra-ra

  • 4a1d23933a Fix for io segfault under gcc11 The segfault only happened with sanitizers off, -O3, and -DRA_DO_CHECK=1, and only on gcc11; gcc12 and gcc13 work fine, or any -O<3 in gcc11. The error went away when i removed random combinations of lines in test/io.cc and I couldn't reduce it. Go figure. * ra/ply.hh (operator<<): As stated.

3 weeks ago

lloda pushed to master at lloda/ra-ra

  • 9e7778a3ce Default RA_DO_CHECK to 1 * ra/expr.hh: As stated. * docs/ra-ra.texi: Reword explanation.

1 month ago

lloda pushed tag v19b to lloda/ra-ra

1 month ago

lloda pushed to master at lloda/ra-ra

  • fa08040100 Option for simplified error reporting I've found that the printing of informative args might slow down checks, even though those arguments aren't evaluated when the check passes :-/ So, provide the option to ignore those args but still do the checks. * ra/base.hh (VERSION): Bump to 28. * ra/expr.hh (RA_DO_CHECK): Provide variants of RA_ASSERT depending on value, if not 0. * docs/ra-ra.texi (Error handling): Explain the new options.
  • 44b9ffcafd Clean up checks * ra/big.hh (ViewBig): Fix typo in conversion to scalar. * ra/expr.hh (RA_ASSERT): Use __VA_OPT__ instead of ##__VA_ARGS__. (Match): Print full shapes on match error. * test/checks.cc, test/throw.cc: Use __VA_OPT__ instead of ##__VA_ARGS__. * test/SConstruct, test/CMakeLists.txt: Don't run ra-16.cc, which causes gcc11 to hang or give spurious errors. Tbf.
  • 3e77557e8c Some checks for conversion to scalar
  • 0888466a37 Always define RA_DO_CHECK * ra/expr.hh: As stated. * docs/ra-ra.texi: Fix typo.
  • 4266050494 Rename ra/bootstrap.hh -> ra/base.hh, ra::xI -> ra::xi
  • View comparison for these 5 commits »

1 month ago