Re: Are f.p. manipulation functions only used in initialization?

"Steven G. Kargl" <kargl@xxxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:fv086u$roe$1@xxxxxxxxxxxxxxxxxxxxxxxxxx

Appears to work for me. Perhaps, it's an OS problem.

Thank you for testing. Maybe an OS problem or a regression. What
version are you using? I'm on 20080421.

I ran into another problem that may be with the documentation. In:

http://gcc.gnu.org/onlinedocs/gcc/i386-and-x86_002d64-Options.html#i386-and-x86_002d64-Options

It says:

"-mpc32
-mpc64
-mpc80
Set 80387 floating-point precision to 32, 64 or 80 bits. When
-mpc32 is specified, the significands of results of floating-point
operations are rounded to 24 bits (single precision); -mpc64 rounds
the the significands of results of floating-point operations to 53
bits (double precision) and -mpc80 rounds the significands of results
of floating-point operations to 64 bits (extended double precision),
which is the default. When this option is used, floating-point
operations in higher precisions are not available to the programmer
without setting the FPU control word explicitly.

Setting the rounding of floating-point operations to less than the
default 80 bits can speed some programs by 2% or more. Note that some
mathematical libraries assume that extended precision (80 bit)
floating-point operations are enabled by default; routines in such
libraries could suffer significant loss of accuracy, typically through
so-called "catastrophic cancellation", when this option is used to set
the precision to less than extended precision."

My reading of this is that precision control should be set to
b'11' = Extended Precision, but I must be reading the wrong part
of the manual because I get:

C:\gfortran\james\archpi>c:\gfortran\win64\bin\x86_64-pc-mingw32-gfortran -v
Using built-in specs.
Target: x86_64-pc-mingw32
Configured with:
.../../trunk/configure --prefix=/home/FX/irun64 --build=i586-pc-
mingw32 --target=x86_64-pc-mingw32 --with-gmp=/home/FX/local --enable-languages=
c,fortran --disable-werror --disable-nls --enable-threads
Thread model: win32
gcc version 4.4.0 20080421 (experimental) [trunk revision 134506] (GCC)

C:\gfortran\james\archpi>type test_status.f90
program test_status
implicit none
interface
function status()
integer(selected_int_kind(4)) status
end function status
subroutine control(cw)
integer(selected_int_kind(4)) cw
end subroutine control
end interface
real(10) x
real(10) y
integer(selected_int_kind(2)) array(10)
integer(selected_int_kind(4)) old_control

x = 1
x = 4*atan(x)
y = x-nearest(x,-1.0_10)
y = x+y
array = transfer(x,array)
write(*,'(10z2.2)') array(size(array):1:-1)
array = transfer(y,array)
write(*,'(10z2.2)') array(size(array):1:-1)
old_control = status()
write(*,'(z4.4)') old_control
call control(ior(old_control,int(z'0300',kind(old_control))))
x = 1
x = 4*atan(x)
y = x-nearest(x,-1.0_10)
y = x+y
array = transfer(x,array)
write(*,'(10z2.2)') array(size(array):1:-1)
array = transfer(y,array)
write(*,'(10z2.2)') array(size(array):1:-1)
write(*,'(z4.4)') status()
call control(old_control)
end program test_status

C:\gfortran\james\archpi>type status.s
.text
..globl _status_
.def _status_; .scl 2; .type 32; .endef
_status_:
fstcw 8(%rsp)
movzwl 8(%rsp), %eax
ret
..globl _control_
.def _control_; .scl 2; .type 32; .endef
_control_:
fldcw (%rcx)
ret

C:\gfortran\james\archpi>c:\gfortran\win64\bin\x86_64-pc-mingw32-gfortran -mpc80
test_status.f90 status.s -otest_status

C:\gfortran\james\archpi>test_status
4000C90FDAA22168C000
4000C90FDAA22168C000
027F
4000C90FDAA22168C235
4000C90FDAA22168C236
037F

So you can see that gfortran isn't setting precision control to b'11'
even with the -mpc80 switch in effect so that real(10) arithmetic
isn't working until I set the x87 control word by hand. Probably I'm
just missing something in the documentation somewhere but it can sure
be hard to find it from a standing start.

--
write(*,*) transfer((/17.392111325966148d0,6.5794487871554595D-85, &
6.0134700243160014d-154/),(/'x'/)); end


.

Relevant Pages

  • Re: Are f.p. manipulation functions only used in initialization?
    ... the significands of results of floating-point ... bits (double precision) and -mpc80 rounds the significands of results ... of floating-point operations to 64 bits, ... operations in higher precisions are not available to the programmer ...
    (comp.lang.fortran)
  • Re: Are f.p. manipulation functions only used in initialization?
    ... Easier IMO to look at Intels documentation such as ... as cos, sin, exp, log) if you change the FPU precision. ... You know that precision control only applies to addition, subtraction, ... having the FPU in double precision, ...
    (comp.lang.fortran)
  • Re: How to use FPU 80bit FP
    ... the precision bits in this newsgroup. ... INTEL FPU FLOATING POINT FORMATS ... FPU CONTROL WORD ... because stack faults should never be masked, ...
    (comp.lang.asm.x86)
  • Re: most shock absorbing inserts?
    ... > means that actually precise control *IS* important. ... Barry Publow working on technique doing lots of exercises, ... angle of the skate to anything like a precision of plus-minus two degrees. ...
    (rec.sport.skating.inline)
  • Re: Call Array valued Fortran function from C
    ... documentation about gfortran array descriptors for you. ... double precision, dimension:: array ... type, bind:: descr ... integerunknown1 ...
    (comp.lang.fortran)