Re: gcc handling of the sin, cos functions

[dvv at egcs dot dvv dot org (Dima Volodin)]

On 19 Nov 1999 01:00:22 -0200, you wrote:

>On Nov 18, 1999, Jim Carrig <jcarrig@bellatlantic.net> wrote:
>
>> I get the feeling that the gcc C compiler (math library) for Intel
>> platforms does not take advantage of the floating point unit in
>> calculating sin, cos, exp etc.  Is this correct?
>
>I believe it is, as this is a C library issue.  On GNU/Linux, this is
>handled by the GNU C Library, which is a separate project: the math
>functions are defined as inline assembly code, making use of the
>special instructions.  But gcc might be taught to expand these
>particular functions by itself, just like it currently does with some
>string and memory functions.

Isn't it what -ffast-math does? Check it out:

       .file   "00.c"
        .version        "01.01"
gcc2_compiled.:
.section        .rodata
.LC0:
        .string "%f\n"
.text
        .align 4
.globl main
        .type    main,@function
main:
        fldz
        pushl %ebp
        movl %esp,%ebp
        subl $8,%esp
        addl $-4,%esp
        fsin
        subl $8,%esp
        fstpl (%esp)
        pushl $.LC0
        call printf
        movl %ebp,%esp
        popl %ebp
        ret
.Lfe1:
        .size    main,.Lfe1-main
        .ident  "GCC: (GNU) 2.95.1 19990816 (release)"

>Alexandre Oliva http://www.ic.unicamp.br/~oliva IC-Unicamp, Bra[sz]il

Dima