76 lines
2.6 KiB
C
76 lines
2.6 KiB
C
/******************************************************************************\
|
|
* Authors: Iconoclast *
|
|
* Release: 2013.11.26 *
|
|
* License: CC0 Public Domain Dedication *
|
|
* *
|
|
* To the extent possible under law, the author(s) have dedicated all copyright *
|
|
* and related and neighboring rights to this software to the public domain *
|
|
* worldwide. This software is distributed without any warranty. *
|
|
* *
|
|
* You should have received a copy of the CC0 Public Domain Dedication along *
|
|
* with this software. *
|
|
* If not, see <http://creativecommons.org/publicdomain/zero/1.0/>. *
|
|
\******************************************************************************/
|
|
#include "vu.h"
|
|
|
|
INLINE static void do_madh(usf_state_t * state, short* VD, short* VS, short* VT)
|
|
{
|
|
|
|
#ifdef ARCH_MIN_ARM_NEON
|
|
|
|
int16x8_t vs,vt, vaccm, vacch, vacc_h, vacc_m,prod_low,prod_high,one,cond;
|
|
uint16x8_t cond_u,res;
|
|
|
|
one = vdupq_n_s16(1);
|
|
vs = vld1q_s16((const int16_t *)VS);
|
|
vt = vld1q_s16((const int16_t *)VT);
|
|
vaccm = vld1q_s16((const int16_t *)VACC_M);
|
|
vacch = vld1q_s16((const int16_t *)VACC_H);
|
|
|
|
prod_low = vmulq_s16(vs,vt);
|
|
vacc_m = vaddq_s16(vaccm,prod_low);
|
|
|
|
prod_high = vqdmulhq_s16(vs,vt);
|
|
prod_high = vshrq_n_s16(prod_high, 1);
|
|
|
|
res = vqaddq_u16((uint16x8_t)vaccm, (uint16x8_t)prod_low);
|
|
cond_u = vceqq_s16((int16x8_t)res,vacc_m);
|
|
cond_u = vaddq_u16(cond_u, (uint16x8_t)one);
|
|
|
|
vacc_h = vaddq_s16(prod_high, vacch);
|
|
vacc_h = vaddq_s16((int16x8_t)cond_u, vacc_h);
|
|
|
|
vst1q_s16(VACC_M, vacc_m);
|
|
vst1q_s16(VACC_H, vacc_h);
|
|
|
|
SIGNED_CLAMP_AM(state, VD);
|
|
return;
|
|
|
|
#else
|
|
|
|
ALIGNED int32_t product[N];
|
|
ALIGNED uint32_t addend[N];
|
|
register int i;
|
|
|
|
for (i = 0; i < N; i++)
|
|
product[i] = (signed short)(VS[i]) * (signed short)(VT[i]);
|
|
for (i = 0; i < N; i++)
|
|
addend[i] = (unsigned short)(VACC_M[i]) + (unsigned short)(product[i]);
|
|
for (i = 0; i < N; i++)
|
|
VACC_M[i] += (short)(VS[i] * VT[i]);
|
|
for (i = 0; i < N; i++)
|
|
VACC_H[i] += (addend[i] >> 16) + (product[i] >> 16);
|
|
SIGNED_CLAMP_AM(state, VD);
|
|
return;
|
|
|
|
#endif
|
|
}
|
|
|
|
static void VMADH(usf_state_t * state, int vd, int vs, int vt, int e)
|
|
{
|
|
ALIGNED short ST[N];
|
|
|
|
SHUFFLE_VECTOR(ST, state->VR[vt], e);
|
|
do_madh(state, state->VR[vd], state->VR[vs], ST);
|
|
return;
|
|
}
|