forked from len0rd/rockbox
Optimise EQ coef calculation routines for both speed and size. Move now unneeded fsqrt function to plugin fixed point library in case it'll be needed. Move all fixed point helper macros to dsp.h. Added FRACMUL_SHL macro to facilitate high-precision shifting of 64 bit multiplies and remove rounding from macsr in main thread to make this work as intended.
Tested quite thorougly, but as always, be careful with your ears. git-svn-id: svn://svn.rockbox.org/rockbox/trunk@12203 a1c6a512-1295-4272-9138-f99709370657
This commit is contained in:
parent
7170a00daa
commit
5f48e1590f
6 changed files with 252 additions and 234 deletions
|
@ -240,10 +240,10 @@ void system_init(void)
|
|||
"movclr.l %%acc2, %%d0\n\t"
|
||||
"movclr.l %%acc3, %%d0\n\t"
|
||||
: : : "d0");
|
||||
/* Set EMAC unit to saturating and rounding fractional mode, since that's
|
||||
/* Set EMAC unit to fractional mode with saturation, since that's
|
||||
what'll be the most useful for most things which the main thread
|
||||
will do. */
|
||||
coldfire_set_macsr(EMAC_FRACTIONAL | EMAC_SATURATE | EMAC_ROUND);
|
||||
coldfire_set_macsr(EMAC_FRACTIONAL | EMAC_SATURATE);
|
||||
|
||||
/* Set INTBASE and SPURVEC */
|
||||
INTBASE = 64;
|
||||
|
|
Loading…
Add table
Add a link
Reference in a new issue