Libdfp Versions Save

Decimal Floating Point C Library

1.0.16

2 years ago

1.0.16 - 2021-07-14

  • Release 1.0.16

  • Run tests on GCC 11

  • Run tests on gcc-9 and gcc-10

  • Simplify even further the example configure call

  • Remove references to --with-backend and --enable-decimal-float

  • libbid: sync to gcc commit 99dee82307f

  • Ignore PLT references from GCC intrinsic functions

  • Fix installation permissions

  • Add TS18661-2 strfromdN

  • Enable travis CI for libdfp.

  • powerpc: use __builtin_fabsdN for fabsdN

  • Remove malloc/realloc checks in configury

  • Do not disambiguate configury top_srcdir

  • Cleanup libdecnumber config.h

  • Cleanup new configure generated files in distclean

  • Cleanup *.d files

  • Remove dfp_backend variable

  • Fix .pc permissions

1.0.15

3 years ago

ChangeLog

1.0.15 - 2020-07-13

  • Release 1.0.15

  • Detect srcdir with a file not removed in the release tarball

  • Add a pkgconfig file for libdecnumber

  • Use python3 when running bare scripts

  • libdecnumber: fix srcdir==builddir builds

  • libdecnumber: fixup mainter uri

  • Workaround broken autotools cross-compile tests

  • libdecnumber: build and install static libdecnumber

  • S390: Fix left_justify as also done for other architectures.

  • Creating new nextup and nextdown functions

  • Decouple submachine and detect soft/hard dfp

  • remove powerpc fenvinline.h and imply hard-fp

  • cleanup and remove ldouble Implies

  • Improve configure

  • Test if compiler and libraries support _Float128

  • Add a simple build testing script

  • powerpc64le: add missing GCC 9 symbols for KFmode

  • Explicitly require python, and favor python3

  • Add fminmagN function as TS18661 recommends

  • test-bfp-conversions: don't expose _Float128 if not supported

  • configure: fix compiler/glibc check

  • Remove unusable __decrm_location private symbol

  • dpd: add missing redirects

  • Makefile.in: disambiguate top_srcdir with realpath

  • Makefile.in: add datarootdir variable

  • x86_64/i386: remove unused sysdep dirs

  • configure.ac: remove AC_FUNC_MALLOC check

  • test-bfp-conversions: refactorwrite support for _Float128

  • Refactor dfp/math.h and built with C11

  • Add float128 support

  • S390: Mark internal quantized function as hidden.

  • Fix pkg-config.

  • Add pkg-config file

  • Remove mini-gmp

  • bid: use decNumber for decoded128

  • bid: remove u128_from_str and related bits

  • expd64: correctly check for range

  • libbid: fixup PLT hacks for TFmode

  • floatti*d: Ensure PLT hacks are included

  • test-isnan-d*: Use boolean test for success

  • test-fenv: Don't test non-standard modes on BID

  • libbid: wireup exceptions to libm

  • libbid: use __set_status_flags macro in bfp conversions

  • libbid: use bfp rounding mode when converting to bfp

  • libbid: integration part 2

  • libbid: hacks to support shared library

  • libbid: Import from GCC

  • Generic fabs function without falling back to libdecnumber

  • Fix TabError in tests/gen-libdfp-tests.py. #121

  • libdecnumber: inline host-ieeeN.c files for BID

  • libdecnumber: sync with GCC

  • logd128: remove printf debug noise

  • powerpc: Make dfp_inline.h functions static too

  • powd64: cleanup with 'indent'

  • Fix inline round/restore macro

  • gen-auto-dfp-tests: Fix small value rounding

  • powd64: move powerpc64 version to common.

  • powerpc: remove processor specific sysdep dirs

  • powerpc: cleanup __dpd_trunctdsd2

  • powerpc, expd128: fix power9 compilation

  • Add native logd128

  • s390: update ulps

  • Add fmaxmagN function as TS18661 recommends

  • powerpc: remove unused soft-dfp ulps file

  • powerpc: better inline isfinite checks

  • Add rounding mode enforcement to expd128

  • powerpc: inline quantized128 with dfp

  • soft-dfp, powerpc64: update ulps

  • Add native expd128

  • Fix and test ldexp

  • Move ppc expd64/logd64 to common ieee754r

  • Move test rules into test/Makefile

  • Remove unused or fixed configury

  • sin/cos: more correctly compute the requested value

  • Fix spurious integer overflow warnings

  • Add a framework to verify DFP math functions

  • powerpc: fix expd{32,64} array out of bounds bug

  • Replace the install_root with DESTDIR

  • Re-add test output

  • Use GCC builtins for fpclassify tests

  • s390: fix warning about missing prototype

  • isunordered*: use GCC builtin

  • unord{s,d,t}d2: use __builtin_isunordered instead

  • base-math: remove unnecessary d32 -> d64 promotions

  • powerpc: remove float{uns,}disd assembly routines

  • powerpc: remove remaining assembly files

  • powerpc: remove some assembly overrides

  • Remove source from root directory

  • Rename CFLAGS-sysdeps to sysdeps-CFLAGS

  • Contain with_dfp to autoconf

  • Remove --enable-decimal-float option.

  • Remove --with-backend option

  • Use sysdeps to infer bid/dpd sysdep dirs

  • Remove _{DPD,BID}_BACKEND compiler flag

  • Fix GCC 9 warnings for hidden definitions

  • Workaround rounding failures in libdfp conversion

  • Integrate libdecnumber into build

  • Remove implicit append of -mno-hard-dfp

  • Refactor and remove mzarch configure variable

  • Simplify and mandate __int128 support for 64b targets

  • Fixup x86 conversion routines

  • S390: Add support for z15/arch13 s390 cpu.

  • powerpc: Ignore tests that are known to fail on soft-dfp and POWER hard-dfp

  • Show details about test fails in output of make check.

  • Install COPYING.txt in inst_docdir.

1.0.14

4 years ago

ChangeLog

1.0.14 - 2019-07-10

  • Release 1.0.14

  • Fix exponent range check on Decimal64 to Float conversion.

  • Fix incorrect conversions from Decimal128 to IBM long double.

  • Fix the printing format of some test results in test-bfp-conversions

  • Fix overflow test failures for float conversion.

  • Remove unnecessary messages from find

  • Enable POWER9 build

  • Validate successful return code from *printf

  • Fix comment in samequantumd32.c.

  • S390: Add hardware support for islessgreaterd32|64|128 functions.

  • Add islessgreater tests.

  • S390: Add hardware support for isunorderedd32|64|128 functions.

  • Add isunordered tests.

  • S390: Add hardware support for islessequald32|64|128 functions.

  • S390: Add hardware support for islessd32|64|128 functions.

  • S390: Add hardware support for isgreaterequald32|64|128 functions.

  • S390: Add hardware support for isgreaterd32|64|128 functions.

  • S390: Add hardware support for fdimd32|64|128 functions.

  • Add fdim tests.

  • S390: Add hardware support for fmaxd32|64|128 functions.

  • Add fmax tests.

  • S390: Add hardware support for fmind32|64|128 functions.

  • Add fmin tests.

  • S390: Add hardware support for signbitd32|64|128 functions.

  • Add __signbit tests.

  • S390: Add hardware support for llquantexpd32|64|128 functions.

  • S390: Add hardware support for samequantumd32|64|128 functions.

  • S390: Add hardware support for quantumd32|64|128 functions.

  • S390: Add hardware support for issignalingd32|64|128 functions.

  • Enhance issignaling tests.

  • S390: Add hardware support for isinfd32|64|128 functions.

  • Enhance isinf tests.

  • S390: Add hardware support for isnand32|64|128 functions.

  • Enhance isnan tests.

  • S390: Add hardware support for isnormald32|64|128 functions.

  • Add isnormal tests.

  • S390: Add hardware support for isfinited32|64|128 functions.

  • Enhance isfinite tests.

  • S390: Add hardware support for fpclassifyd32|64|128 functions.

  • Enhance fpclassify tests.

  • S390: Add hardware support for copysignd32|64|128 functions.

  • S390: Add hardware support for fabsd32|64|128 functions.

  • S390: Add hardware support for llrintd32|64|128 functions.

  • Add llrint tests and check invalid/inexact exceptions.

  • S390: Add hardware support for lrintd32|64|128 functions.

  • Add lrint tests and check invalid/inexact exception.

  • S390: Add hardware support for llround32|64|128 functions.

  • Add llround tests.

  • S390: Add hardware support for lroundd32|64|128 functions.

  • Add lround tests and extend testframework.

  • S390: Add hardware support for roundevend32|64|128 functions.

  • Enhance roundeven tests.

  • S390: Add hardware support for roundd32|64|128 functions.

  • Enhance round tests.

  • S390: Add hardware support for truncd32|64|128 functions.

  • Enhance trunc tests.

  • S390: Add hardware support for floord32|64|128 functions.

  • Enhance floor tests.

  • S390: Add hardware support for ceild32|64|128 functions.

  • Enhance ceil tests.

  • S390: Add hardware support for rintd32|64|128

  • Add inexact-exception tests for rint/nearbyint tests.

  • Add rint test with test-framework.

  • S390: Add hardware support for nearbyintd[32|64|128] functions.

  • S390: Add hardware support for quantized32|64|128 functions.

  • Return the number of chars output in printf

  • Make test-cast-to-overflow.c more readable.

  • S390: Add configure check in order to determine if gcc supports.

  • Add more under/overflow tests for dfp to bfp conversion.

  • S390: Raise inexact for d128 -> __int128_t conversions.

  • S390: Use correct rounding mode for d128 to d32 rounds.

  • S390: Add support for z14/arch12 s390 cpu.

  • Adapt the code for glibc 2.28

  • Cleanup files for GCC 8

  • Clean unused variable in configure test

  • Clean 'misleading indentation' warning

  • Clean 'unused const variable' warnings

  • Fix raised inexact exception wrong cleanup.

  • Fix subnormals for decimal to binary conversions.

  • Fix incorrect conversion on underflow.

  • Fix incorrect conversion on overflow.

  • Fix -Wunused-result warning message in test-amort.c

1.0.13

7 years ago

ChangeLog

1.0.13 - 2016-07-01

  • Release 1.0.13
  • libdecnumber: Fix broken magic constant
  • Update maintainers and contributors
  • s390: Fix soft-dfp TFmode conversions
  • s390: Properly override long double functions
  • s390: Fix left_justify
  • Fix tests when long double is IEEE128
  • S390: Add new libdfp-no-ti.abilist for 31bit sw-/hw-dfp.
  • libdfp: Use correct Macros for Decimal 32|64 in test-setexp.
  • S390: Add new libdfp.abilist for sw-dfp.
  • S390: Add support for newer s390 cpus.
  • S390: Get rid of compile warnings with hw-dfp.
  • S390: Add new libdfp.abilist for hw-dfp.
  • libdfp: Add hidden definition for base_math symbols.
  • S390: Fix casting from _Decimal32|128 to _Decimal128|32 with hw-dfp.
  • libdfp: Use correct variant d32|64|128 of __isfinite instead of always __isfinited32 in nearbyint.
  • S390: Fix "Missing required PLT reference: libdfp.so: register_printf_dfp".
  • S390: Use direct calls of __isnand64|128 instead of extra PLT references.
  • S390: New libdfp-test-ulps file with hw-dfp.
  • S390: Add convert_helpers.h implementation for hw-dfp.
  • Avoid dscliq builtin
  • Optimize 128b conversions
  • Add mantissa extraction functions
  • First round of conversion optimizations
  • Ensure optimized exception tests for ppc
  • Add hack to fix tests building with older GCC
  • Synchronize plt checking scripts
  • Fix double rounding bug
  • Implement remaining decimal ti conversions.
  • Fix incremental builds
  • Prevent config.h misconfigurations
  • Address missed comments from PR #24
  • Support older GCC variants
  • Fix return type of comparison operators
  • Remove archaic inline asm output modifier
  • Add most ti <-> Decimal conversions

1.0.12

8 years ago

ChangeLog

1.0.12 - 2015-10-27

  • Release 1.0.12
  • Fix log10 expected test result
  • Print correct number of digits testing ulps
  • Fix quantize test inputs
  • Remove a workaround from make-release.sh
  • Minor fixes to make-release.sh
  • Ensure hardware dfp is disabled with sw only
  • Fix configure when --with-cpu not supplied
  • Fix edge case with strtod*
  • libdfp: powerpc: Simplify _Decimal128 implementation
  • libdfp: powerpc: Fix fabsd128 implementation
  • Update the list of contributors
  • Harden check for the sign bit in the tests
  • Fix the usage of mfcr extended mnemonic

1.0.11

8 years ago

ChangeLog

1.0.11 - 2015-07-13

  • Release 1.0.11
  • Install ChangeLog.md
  • Fix handle of tags in generate-changelog.sh
  • Add a bootstrap script and fix warnings
  • Add release script
  • Add ChangeLog generator script
  • Add a printf test for 2 variables
  • Honor value passed through --docdir
  • Move system header inclusion
  • Automatically detect hardware DFP support
  • Do not clobber configure output
  • Stop using ldconfig to create symbolic links
  • Ignore automatically generated files
  • Cleanup all files after running 'make clean'
  • Add removing of libdfp-test-ulps.h
  • Adding support for LDFLAGS passed from env
  • Allow building from the source code directory
  • Fix cross compilation

1.0.10

9 years ago

The "Decimal Floating Point C Library" is an implementation of ISO/IEC Technical report "ISO/IEC TR 24732" which describes the C-Language library routines necessary to provide the C library runtime support for decimal floating point data types introduced in IEEE 754-2008, namely _Decimal32, _Decimal64, and _Decimal128.

New in version 1.0.10:

  • Added BID support for x86_64 and x86.
  • Fixes to avoid external PLT calls to internal supplied symbols.
  • Add optimized fabsd[32|64|128], and copysignd[32|64|128] for PowerPC with DFP hardware support.

1.0.9-1

10 years ago

This is a bugfix update for version 1.0.9. It includes the following fixes:

  • Fix build configuration when setting compiler with flags (for instance, CC="gcc -m32").
  • Fix build configuration when building for POWER8 (--with-cpu=power8).
  • Fix issignaling for PowerPC DFP for zero inputs.
  • Add missing fpclassify prototypes.

1.0.9

10 years ago

The "Decimal Floating Point C Library" is an implementation of ISO/IEC Technical report "ISO/IEC TR 24732" which describes the C-Language library routines necessary to provide the C library runtime support for decimal floating point data types introduced in IEEE 754-2008, namely _Decimal32, _Decimal64, and _Decimal128.

New in version 1.0.9

  • Support for powerpc64le has beed added.
  • Various fixes for LE build.
  • Support for configure with --with-cpu=power8 has been added.
  • Fix for isinfd[32|64|128]: it now follow GLIBC convention and returns 1 for plus infinity and -1 for minus infinity.
  • Fix finited result to return a positive argument for any normal number.
  • New <math.h> macro named issignaling to check for a signaling NaN (sNaN). It is based on draft TS 18661.
  • New llogbd[32|64|128] implementation to return the logb result as long int.
  • New roundevend[32|64|128] implementation that rounds its argument to the nearest integer value in floating-point format, rounding halfway cases to even (that is, to the nearest value whose least significant bit 0), regardless of the current rounding direction. It is based on draft 18661-2.
  • Add optimized ceild[32|64|128], truncd[32|64|128], floord[32|64|128], nearbyintd[32|64|128], roundd[32|64|128], and roundevend[32|64|128] for PowerPC with DFP hardware support.
  • New quantumd[32|64|128] implementation that compute the quantum of a finite argument. It is based on draft 18661-2.
  • Add optimized quantumd[32|64|128] and samequantumd[32|64|128] implementation for PowerPC with DFP hardware support.