Mercurial > octave

comparison m4/acinclude.m4 @ 21302:8376de2eaf00

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
define OCTAVE_HAVE_FAST_INT_OPS in octave-config.h * acinclude.m4 (OCTAVE_FAST_INT_OPS): Define OCTAVE_HAVE_FAST_INT_OPS instead of HAVE_FAST_INT_OPS. * oct-inttypes.cc, oct-inttypes.h: Use new macro name. * mk-octave-config-h.sh: Emit definition of OCTAVE_HAVE_FAST_INT_OPS. * build-env-features.sh: Also include symbols beginning with "OCTAVE_HAVE_".
author John W. Eaton <jwe@octave.org>
date Thu, 18 Feb 2016 16:42:42 -0500
parents 6c4f41a9439d
children 06d15e4e611a
comparison
equal deleted inserted replaced
21301:40de9f8f23a6 21302:8376de2eaf00
1493 AC_MSG_RESULT([no]) 1493 AC_MSG_RESULT([no])
1494 ifelse([$3], , , [$3]) 1494 ifelse([$3], , , [$3])
1495 fi 1495 fi
1496 ]) 1496 ])
1497 dnl 1497 dnl
1498 dnl Check whether fast signed integer arithmetics using bit tricks 1498 dnl Check whether fast signed integer arithmetic using bit tricks
1499 dnl can be used in oct-inttypes.h. Defines HAVE_FAST_INT_OPS if 1499 dnl can be used in oct-inttypes.h.
1500 dnl the following conditions hold: 1500 dnl
1501 dnl 1. Signed numbers are represented by twos complement 1501 dnl Defines OCTAVE_HAVE_FAST_INT_OPS if the following conditions hold:
1502 dnl (see <http://en.wikipedia.org/wiki/Two%27s_complement>) 1502 dnl
1503 dnl 2. static_cast to unsigned int counterpart works like interpreting 1503 dnl 1. Signed numbers are represented by twos complement (see
1504 dnl the signed bit pattern as unsigned (and is thus zero-cost). 1504 dnl <http://en.wikipedia.org/wiki/Two%27s_complement>)
1505 dnl 3. Signed addition and subtraction yield the same bit results as unsigned. 1505 dnl
1506 dnl (We use casts to prevent optimization interference, so there is no 1506 dnl 2. static_cast to unsigned int counterpart works like
1507 dnl need for things like -ftrapv). 1507 dnl interpreting the signed bit pattern as unsigned (and is thus
1508 dnl 4. Bit operations on signed integers work like on unsigned integers, 1508 dnl zero-cost).
1509 dnl except for the shifts. Shifts are arithmetic. 1509 dnl
1510 dnl 3. Signed addition and subtraction yield the same bit results
1511 dnl as unsigned. (We use casts to prevent optimization
1512 dnl interference, so there is no need for things like -ftrapv).
1513 dnl
1514 dnl 4. Bit operations on signed integers work like on unsigned
1515 dnl integers, except for the shifts. Shifts are arithmetic.
1510 dnl 1516 dnl
1511 AC_DEFUN([OCTAVE_FAST_INT_OPS], [ 1517 AC_DEFUN([OCTAVE_FAST_INT_OPS], [
1512 AC_CACHE_CHECK([whether fast integer arithmetics is usable], 1518 AC_CACHE_CHECK([whether fast integer arithmetics is usable],
1513 [octave_cv_fast_int_ops], 1519 [octave_cv_fast_int_ops],
1514 [AC_LANG_PUSH(C++) 1520 [AC_LANG_PUSH(C++)
1561 octave_cv_fast_int_ops=no, 1567 octave_cv_fast_int_ops=no,
1562 octave_cv_fast_int_ops=yes) 1568 octave_cv_fast_int_ops=yes)
1563 AC_LANG_POP(C++) 1569 AC_LANG_POP(C++)
1564 ]) 1570 ])
1565 if test $octave_cv_fast_int_ops = yes; then 1571 if test $octave_cv_fast_int_ops = yes; then
1566 AC_DEFINE(HAVE_FAST_INT_OPS, 1, 1572 AC_DEFINE(OCTAVE_HAVE_FAST_INT_OPS, 1,
1567 [Define to 1 if signed integers use two's complement.]) 1573 [Define to 1 if signed integers use two's complement.])
1568 fi 1574 fi
1569 ]) 1575 ])
1570 dnl 1576 dnl
1571 dnl Check to see if the compiler and the linker can handle the flags 1577 dnl Check to see if the compiler and the linker can handle the flags