Home | History | Annotate | Line # | Download | only in notes 
      1 $NetBSD: README.NetBSD,v 1.3 2003/07/26 19:38:47 salo Exp $
      2 
      3 
      4 Changes at initial import...ross (a] NetBSD.org
      5 
      6   o  WARNS=2 fixes
      7   o  source reorganized with arch/
      8   o  <sys/endian.h> conversion
      9   o  <inttypes.h> conversion
     10   o  <ieeefp.h> conversion
     11   o  <asm.h> conversion
     12   o  the previously target-specific softfloat.h and milieu.h made mostly
     13      generic. Still some work to do with default NaN bitpatterns, endian,
     14      and arm/fpa-DEMANGLE issues.
     15   o  arch/i386/systfloat.S extended to handle:
     16 	int32_t syst_floatx80_to_int32_round_to_zero(floatx80)
     17 	int64_t syst_floatx80_to_int64_round_to_zero(floatx80)
     18   o  LONG_DOUBLE_IS_FLOATX80 was used to enable both the C module system
     19      access ops and the FLOATX80 tests, all of which also required FLOATX80.
     20      Besides being rundundant, this made it impossible to use the asm module
     21      for i386 which the package actually comes with and also made it
     22      impossible to test FLOATX80 ops without compiler support even if the
     23      machine actually does them.  While this is arguably OK for a regression
     24      test, the two cases are obviously different. Now, the tests (which
     25      don't actually require any compiler-understood extended type) are
     26      always run just by defining FLOATX80. If LONG_DOUBLE_IS_FLOATX80 is
     27      also defined, then the C system ops are also enabled. To switch back
     28      and forth, you modify only the arch/${MACHINE_ARCH}/Makefile.inc to
     29      do the cpp def and add or remove systfloat.S. For regression testing,
     30      it's better in C, but for testing the test itself or verifying a
     31      regression, the .S module is a better reference case.
     32 
     33      A similar change should probably be made for LONG_DOUBLE_IS_FLOAT128,
     34      but we can't test that yet.
     35 
     36      The basic rules:
     37 	LONG_DOUBLE_IS_FLOAT{X80,128}
     38 		define if you do not have a .S file and do have compiler
     39 		support, or have both and want to use the compiled version
     40 		(but then it may be necessary to remove the .S file or add
     41 		an ifndef to it)
     42 	FLOAT{X80,128}
     43 		define if you have either HW or SW support and want it tested
     44 
     45     And to confuse things further: the .S files for i386 appear to have been
     46     originally created by compiling C code with -S, at least as a start.
     47 
     48     You can see some of the price of ANSI+IEEE in these files: a simple
     49     conversion from floatxx to intxx requires storing, modifying, loading,
     50     and restoring  the modes on each conversion in order to get a specific
     51     round, thanks to the four IEEE-mandated options. A pipeline/peephole
     52     pass can no doubt fix this up within one function, but not if other
     53     arithmetic is done in between.
     54