a99cf56c7d
the arm compiler can lift long->vlong casts on multiplcation and convert 64x64->64 multiplication into a 32x32->64 one with optional 64 bit accumulate.
36 lines
859 B
C
36 lines
859 B
C
#include <u.h>
|
|
|
|
/* mul64fract(uvlong*r, uvlong a, uvlong b)
|
|
*
|
|
* Multiply two 64 numbers and return the middle 64 bits of the 128 bit result.
|
|
*
|
|
* The assumption is that one of the numbers is a
|
|
* fixed point number with the integer portion in the
|
|
* high word and the fraction in the low word.
|
|
*
|
|
* There should be an assembler version of this routine
|
|
* for each architecture. This one is intended to
|
|
* make ports easier.
|
|
*
|
|
* ignored r0 = lo(a0*b0)
|
|
* lsw of result r1 = hi(a0*b0) +lo(a0*b1) +lo(a1*b0)
|
|
* msw of result r2 = hi(a0*b1) +hi(a1*b0) +lo(a1*b1)
|
|
* ignored r3 = hi(a1*b1)
|
|
*/
|
|
|
|
void
|
|
mul64fract(uvlong *r, uvlong a, uvlong b)
|
|
{
|
|
ulong bh, bl, ah, al;
|
|
|
|
bl = b;
|
|
bh = b >> 32;
|
|
al = a;
|
|
ah = a >> 32;
|
|
|
|
*r = (((uvlong)al*(uvlong)bl)>>32)
|
|
+ ((uvlong)al*(uvlong)bh)
|
|
+ ((uvlong)ah*(uvlong)bl)
|
|
+ (((uvlong)ah*(uvlong)bh)<<32);
|
|
}
|