2013-09-13 11:48:32 +02:00
|
|
|
/*
|
|
|
|
|
* Copyright (c) 2013 The WebM project authors. All Rights Reserved.
|
|
|
|
|
*
|
|
|
|
|
* Use of this source code is governed by a BSD-style license
|
|
|
|
|
* that can be found in the LICENSE file in the root of the source
|
|
|
|
|
* tree. An additional intellectual property rights grant can be found
|
|
|
|
|
* in the file PATENTS. All contributing project authors may
|
|
|
|
|
* be found in the AUTHORS file in the root of the source tree.
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
#include <assert.h>
|
|
|
|
|
#include <stdio.h>
|
|
|
|
|
|
2015-07-28 19:52:24 +02:00
|
|
|
#include "./vpx_dsp_rtcd.h"
|
2015-08-03 18:54:13 +02:00
|
|
|
#include "vpx_dsp/mips/convolve_common_dspr2.h"
|
2015-07-28 19:52:24 +02:00
|
|
|
#include "vpx_dsp/vpx_dsp_common.h"
|
|
|
|
|
#include "vpx_dsp/vpx_filter.h"
|
2013-09-13 11:48:32 +02:00
|
|
|
#include "vpx_ports/mem.h"
|
|
|
|
|
|
|
|
|
|
#if HAVE_DSPR2
|
|
|
|
|
static void convolve_vert_4_dspr2(const uint8_t *src,
|
|
|
|
|
int32_t src_stride,
|
|
|
|
|
uint8_t *dst,
|
|
|
|
|
int32_t dst_stride,
|
|
|
|
|
const int16_t *filter_y,
|
|
|
|
|
int32_t w,
|
|
|
|
|
int32_t h) {
|
|
|
|
|
int32_t x, y;
|
|
|
|
|
const uint8_t *src_ptr;
|
|
|
|
|
uint8_t *dst_ptr;
|
2015-07-28 19:52:24 +02:00
|
|
|
uint8_t *cm = vpx_ff_cropTbl;
|
2013-09-13 11:48:32 +02:00
|
|
|
uint32_t vector4a = 64;
|
|
|
|
|
uint32_t load1, load2, load3, load4;
|
|
|
|
|
uint32_t p1, p2;
|
|
|
|
|
uint32_t n1, n2;
|
|
|
|
|
uint32_t scratch1, scratch2;
|
|
|
|
|
uint32_t store1, store2;
|
|
|
|
|
int32_t vector1b, vector2b, vector3b, vector4b;
|
|
|
|
|
int32_t Temp1, Temp2;
|
|
|
|
|
|
|
|
|
|
vector1b = ((const int32_t *)filter_y)[0];
|
|
|
|
|
vector2b = ((const int32_t *)filter_y)[1];
|
|
|
|
|
vector3b = ((const int32_t *)filter_y)[2];
|
|
|
|
|
vector4b = ((const int32_t *)filter_y)[3];
|
|
|
|
|
|
|
|
|
|
src -= 3 * src_stride;
|
|
|
|
|
|
|
|
|
|
for (y = h; y--;) {
|
|
|
|
|
/* prefetch data to cache memory */
|
2015-07-17 21:31:53 +02:00
|
|
|
prefetch_store(dst + dst_stride);
|
2013-09-13 11:48:32 +02:00
|
|
|
|
|
|
|
|
for (x = 0; x < w; x += 4) {
|
|
|
|
|
src_ptr = src + x;
|
|
|
|
|
dst_ptr = dst + x;
|
|
|
|
|
|
|
|
|
|
__asm__ __volatile__ (
|
|
|
|
|
"ulw %[load1], 0(%[src_ptr]) \n\t"
|
|
|
|
|
"add %[src_ptr], %[src_ptr], %[src_stride] \n\t"
|
|
|
|
|
"ulw %[load2], 0(%[src_ptr]) \n\t"
|
|
|
|
|
"add %[src_ptr], %[src_ptr], %[src_stride] \n\t"
|
|
|
|
|
"ulw %[load3], 0(%[src_ptr]) \n\t"
|
|
|
|
|
"add %[src_ptr], %[src_ptr], %[src_stride] \n\t"
|
|
|
|
|
"ulw %[load4], 0(%[src_ptr]) \n\t"
|
|
|
|
|
|
|
|
|
|
"mtlo %[vector4a], $ac0 \n\t"
|
|
|
|
|
"mtlo %[vector4a], $ac1 \n\t"
|
|
|
|
|
"mtlo %[vector4a], $ac2 \n\t"
|
|
|
|
|
"mtlo %[vector4a], $ac3 \n\t"
|
|
|
|
|
"mthi $zero, $ac0 \n\t"
|
|
|
|
|
"mthi $zero, $ac1 \n\t"
|
|
|
|
|
"mthi $zero, $ac2 \n\t"
|
|
|
|
|
"mthi $zero, $ac3 \n\t"
|
|
|
|
|
|
|
|
|
|
"preceu.ph.qbr %[scratch1], %[load1] \n\t"
|
|
|
|
|
"preceu.ph.qbr %[p1], %[load2] \n\t"
|
|
|
|
|
"precrq.ph.w %[n1], %[p1], %[scratch1] \n\t" /* pixel 2 */
|
|
|
|
|
"append %[p1], %[scratch1], 16 \n\t" /* pixel 1 */
|
|
|
|
|
"preceu.ph.qbr %[scratch2], %[load3] \n\t"
|
|
|
|
|
"preceu.ph.qbr %[p2], %[load4] \n\t"
|
|
|
|
|
"precrq.ph.w %[n2], %[p2], %[scratch2] \n\t" /* pixel 2 */
|
|
|
|
|
"append %[p2], %[scratch2], 16 \n\t" /* pixel 1 */
|
|
|
|
|
|
|
|
|
|
"dpa.w.ph $ac0, %[p1], %[vector1b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac0, %[p2], %[vector2b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac1, %[n1], %[vector1b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac1, %[n2], %[vector2b] \n\t"
|
|
|
|
|
|
|
|
|
|
"preceu.ph.qbl %[scratch1], %[load1] \n\t"
|
|
|
|
|
"preceu.ph.qbl %[p1], %[load2] \n\t"
|
|
|
|
|
"precrq.ph.w %[n1], %[p1], %[scratch1] \n\t" /* pixel 2 */
|
|
|
|
|
"append %[p1], %[scratch1], 16 \n\t" /* pixel 1 */
|
|
|
|
|
"preceu.ph.qbl %[scratch2], %[load3] \n\t"
|
|
|
|
|
"preceu.ph.qbl %[p2], %[load4] \n\t"
|
|
|
|
|
"precrq.ph.w %[n2], %[p2], %[scratch2] \n\t" /* pixel 2 */
|
|
|
|
|
"append %[p2], %[scratch2], 16 \n\t" /* pixel 1 */
|
|
|
|
|
|
|
|
|
|
"dpa.w.ph $ac2, %[p1], %[vector1b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac2, %[p2], %[vector2b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac3, %[n1], %[vector1b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac3, %[n2], %[vector2b] \n\t"
|
|
|
|
|
|
|
|
|
|
"add %[src_ptr], %[src_ptr], %[src_stride] \n\t"
|
|
|
|
|
"ulw %[load1], 0(%[src_ptr]) \n\t"
|
|
|
|
|
"add %[src_ptr], %[src_ptr], %[src_stride] \n\t"
|
|
|
|
|
"ulw %[load2], 0(%[src_ptr]) \n\t"
|
|
|
|
|
"add %[src_ptr], %[src_ptr], %[src_stride] \n\t"
|
|
|
|
|
"ulw %[load3], 0(%[src_ptr]) \n\t"
|
|
|
|
|
"add %[src_ptr], %[src_ptr], %[src_stride] \n\t"
|
|
|
|
|
"ulw %[load4], 0(%[src_ptr]) \n\t"
|
|
|
|
|
|
|
|
|
|
"preceu.ph.qbr %[scratch1], %[load1] \n\t"
|
|
|
|
|
"preceu.ph.qbr %[p1], %[load2] \n\t"
|
|
|
|
|
"precrq.ph.w %[n1], %[p1], %[scratch1] \n\t" /* pixel 2 */
|
|
|
|
|
"append %[p1], %[scratch1], 16 \n\t" /* pixel 1 */
|
|
|
|
|
"preceu.ph.qbr %[scratch2], %[load3] \n\t"
|
|
|
|
|
"preceu.ph.qbr %[p2], %[load4] \n\t"
|
|
|
|
|
"precrq.ph.w %[n2], %[p2], %[scratch2] \n\t" /* pixel 2 */
|
|
|
|
|
"append %[p2], %[scratch2], 16 \n\t" /* pixel 1 */
|
|
|
|
|
|
|
|
|
|
"dpa.w.ph $ac0, %[p1], %[vector3b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac0, %[p2], %[vector4b] \n\t"
|
|
|
|
|
"extp %[Temp1], $ac0, 31 \n\t"
|
|
|
|
|
"dpa.w.ph $ac1, %[n1], %[vector3b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac1, %[n2], %[vector4b] \n\t"
|
|
|
|
|
"extp %[Temp2], $ac1, 31 \n\t"
|
|
|
|
|
|
|
|
|
|
"preceu.ph.qbl %[scratch1], %[load1] \n\t"
|
|
|
|
|
"preceu.ph.qbl %[p1], %[load2] \n\t"
|
|
|
|
|
"precrq.ph.w %[n1], %[p1], %[scratch1] \n\t" /* pixel 2 */
|
|
|
|
|
"append %[p1], %[scratch1], 16 \n\t" /* pixel 1 */
|
|
|
|
|
"preceu.ph.qbl %[scratch2], %[load3] \n\t"
|
|
|
|
|
"preceu.ph.qbl %[p2], %[load4] \n\t"
|
|
|
|
|
"precrq.ph.w %[n2], %[p2], %[scratch2] \n\t" /* pixel 2 */
|
|
|
|
|
"append %[p2], %[scratch2], 16 \n\t" /* pixel 1 */
|
|
|
|
|
|
|
|
|
|
"lbux %[store1], %[Temp1](%[cm]) \n\t"
|
|
|
|
|
"dpa.w.ph $ac2, %[p1], %[vector3b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac2, %[p2], %[vector4b] \n\t"
|
|
|
|
|
"extp %[Temp1], $ac2, 31 \n\t"
|
|
|
|
|
|
|
|
|
|
"lbux %[store2], %[Temp2](%[cm]) \n\t"
|
|
|
|
|
"dpa.w.ph $ac3, %[n1], %[vector3b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac3, %[n2], %[vector4b] \n\t"
|
|
|
|
|
"extp %[Temp2], $ac3, 31 \n\t"
|
|
|
|
|
|
|
|
|
|
"sb %[store1], 0(%[dst_ptr]) \n\t"
|
|
|
|
|
"sb %[store2], 1(%[dst_ptr]) \n\t"
|
|
|
|
|
|
|
|
|
|
"lbux %[store1], %[Temp1](%[cm]) \n\t"
|
|
|
|
|
"lbux %[store2], %[Temp2](%[cm]) \n\t"
|
|
|
|
|
|
|
|
|
|
"sb %[store1], 2(%[dst_ptr]) \n\t"
|
|
|
|
|
"sb %[store2], 3(%[dst_ptr]) \n\t"
|
|
|
|
|
|
|
|
|
|
: [load1] "=&r" (load1), [load2] "=&r" (load2),
|
|
|
|
|
[load3] "=&r" (load3), [load4] "=&r" (load4),
|
|
|
|
|
[p1] "=&r" (p1), [p2] "=&r" (p2),
|
|
|
|
|
[n1] "=&r" (n1), [n2] "=&r" (n2),
|
|
|
|
|
[scratch1] "=&r" (scratch1), [scratch2] "=&r" (scratch2),
|
|
|
|
|
[Temp1] "=&r" (Temp1), [Temp2] "=&r" (Temp2),
|
|
|
|
|
[store1] "=&r" (store1), [store2] "=&r" (store2),
|
|
|
|
|
[src_ptr] "+r" (src_ptr)
|
|
|
|
|
: [vector1b] "r" (vector1b), [vector2b] "r" (vector2b),
|
|
|
|
|
[vector3b] "r" (vector3b), [vector4b] "r" (vector4b),
|
|
|
|
|
[vector4a] "r" (vector4a), [src_stride] "r" (src_stride),
|
|
|
|
|
[cm] "r" (cm), [dst_ptr] "r" (dst_ptr)
|
|
|
|
|
);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Next row... */
|
|
|
|
|
src += src_stride;
|
|
|
|
|
dst += dst_stride;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void convolve_vert_64_dspr2(const uint8_t *src,
|
|
|
|
|
int32_t src_stride,
|
|
|
|
|
uint8_t *dst,
|
|
|
|
|
int32_t dst_stride,
|
|
|
|
|
const int16_t *filter_y,
|
|
|
|
|
int32_t h) {
|
|
|
|
|
int32_t x, y;
|
|
|
|
|
const uint8_t *src_ptr;
|
|
|
|
|
uint8_t *dst_ptr;
|
2015-07-28 19:52:24 +02:00
|
|
|
uint8_t *cm = vpx_ff_cropTbl;
|
2013-09-13 11:48:32 +02:00
|
|
|
uint32_t vector4a = 64;
|
|
|
|
|
uint32_t load1, load2, load3, load4;
|
|
|
|
|
uint32_t p1, p2;
|
|
|
|
|
uint32_t n1, n2;
|
|
|
|
|
uint32_t scratch1, scratch2;
|
|
|
|
|
uint32_t store1, store2;
|
|
|
|
|
int32_t vector1b, vector2b, vector3b, vector4b;
|
|
|
|
|
int32_t Temp1, Temp2;
|
|
|
|
|
|
|
|
|
|
vector1b = ((const int32_t *)filter_y)[0];
|
|
|
|
|
vector2b = ((const int32_t *)filter_y)[1];
|
|
|
|
|
vector3b = ((const int32_t *)filter_y)[2];
|
|
|
|
|
vector4b = ((const int32_t *)filter_y)[3];
|
|
|
|
|
|
|
|
|
|
src -= 3 * src_stride;
|
|
|
|
|
|
|
|
|
|
for (y = h; y--;) {
|
|
|
|
|
/* prefetch data to cache memory */
|
2015-07-17 21:31:53 +02:00
|
|
|
prefetch_store(dst + dst_stride);
|
|
|
|
|
prefetch_store(dst + dst_stride + 32);
|
2013-09-13 11:48:32 +02:00
|
|
|
|
|
|
|
|
for (x = 0; x < 64; x += 4) {
|
|
|
|
|
src_ptr = src + x;
|
|
|
|
|
dst_ptr = dst + x;
|
|
|
|
|
|
|
|
|
|
__asm__ __volatile__ (
|
|
|
|
|
"ulw %[load1], 0(%[src_ptr]) \n\t"
|
|
|
|
|
"add %[src_ptr], %[src_ptr], %[src_stride] \n\t"
|
|
|
|
|
"ulw %[load2], 0(%[src_ptr]) \n\t"
|
|
|
|
|
"add %[src_ptr], %[src_ptr], %[src_stride] \n\t"
|
|
|
|
|
"ulw %[load3], 0(%[src_ptr]) \n\t"
|
|
|
|
|
"add %[src_ptr], %[src_ptr], %[src_stride] \n\t"
|
|
|
|
|
"ulw %[load4], 0(%[src_ptr]) \n\t"
|
|
|
|
|
|
|
|
|
|
"mtlo %[vector4a], $ac0 \n\t"
|
|
|
|
|
"mtlo %[vector4a], $ac1 \n\t"
|
|
|
|
|
"mtlo %[vector4a], $ac2 \n\t"
|
|
|
|
|
"mtlo %[vector4a], $ac3 \n\t"
|
|
|
|
|
"mthi $zero, $ac0 \n\t"
|
|
|
|
|
"mthi $zero, $ac1 \n\t"
|
|
|
|
|
"mthi $zero, $ac2 \n\t"
|
|
|
|
|
"mthi $zero, $ac3 \n\t"
|
|
|
|
|
|
|
|
|
|
"preceu.ph.qbr %[scratch1], %[load1] \n\t"
|
|
|
|
|
"preceu.ph.qbr %[p1], %[load2] \n\t"
|
|
|
|
|
"precrq.ph.w %[n1], %[p1], %[scratch1] \n\t" /* pixel 2 */
|
|
|
|
|
"append %[p1], %[scratch1], 16 \n\t" /* pixel 1 */
|
|
|
|
|
"preceu.ph.qbr %[scratch2], %[load3] \n\t"
|
|
|
|
|
"preceu.ph.qbr %[p2], %[load4] \n\t"
|
|
|
|
|
"precrq.ph.w %[n2], %[p2], %[scratch2] \n\t" /* pixel 2 */
|
|
|
|
|
"append %[p2], %[scratch2], 16 \n\t" /* pixel 1 */
|
|
|
|
|
|
|
|
|
|
"dpa.w.ph $ac0, %[p1], %[vector1b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac0, %[p2], %[vector2b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac1, %[n1], %[vector1b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac1, %[n2], %[vector2b] \n\t"
|
|
|
|
|
|
|
|
|
|
"preceu.ph.qbl %[scratch1], %[load1] \n\t"
|
|
|
|
|
"preceu.ph.qbl %[p1], %[load2] \n\t"
|
|
|
|
|
"precrq.ph.w %[n1], %[p1], %[scratch1] \n\t" /* pixel 2 */
|
|
|
|
|
"append %[p1], %[scratch1], 16 \n\t" /* pixel 1 */
|
|
|
|
|
"preceu.ph.qbl %[scratch2], %[load3] \n\t"
|
|
|
|
|
"preceu.ph.qbl %[p2], %[load4] \n\t"
|
|
|
|
|
"precrq.ph.w %[n2], %[p2], %[scratch2] \n\t" /* pixel 2 */
|
|
|
|
|
"append %[p2], %[scratch2], 16 \n\t" /* pixel 1 */
|
|
|
|
|
|
|
|
|
|
"dpa.w.ph $ac2, %[p1], %[vector1b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac2, %[p2], %[vector2b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac3, %[n1], %[vector1b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac3, %[n2], %[vector2b] \n\t"
|
|
|
|
|
|
|
|
|
|
"add %[src_ptr], %[src_ptr], %[src_stride] \n\t"
|
|
|
|
|
"ulw %[load1], 0(%[src_ptr]) \n\t"
|
|
|
|
|
"add %[src_ptr], %[src_ptr], %[src_stride] \n\t"
|
|
|
|
|
"ulw %[load2], 0(%[src_ptr]) \n\t"
|
|
|
|
|
"add %[src_ptr], %[src_ptr], %[src_stride] \n\t"
|
|
|
|
|
"ulw %[load3], 0(%[src_ptr]) \n\t"
|
|
|
|
|
"add %[src_ptr], %[src_ptr], %[src_stride] \n\t"
|
|
|
|
|
"ulw %[load4], 0(%[src_ptr]) \n\t"
|
|
|
|
|
|
|
|
|
|
"preceu.ph.qbr %[scratch1], %[load1] \n\t"
|
|
|
|
|
"preceu.ph.qbr %[p1], %[load2] \n\t"
|
|
|
|
|
"precrq.ph.w %[n1], %[p1], %[scratch1] \n\t" /* pixel 2 */
|
|
|
|
|
"append %[p1], %[scratch1], 16 \n\t" /* pixel 1 */
|
|
|
|
|
"preceu.ph.qbr %[scratch2], %[load3] \n\t"
|
|
|
|
|
"preceu.ph.qbr %[p2], %[load4] \n\t"
|
|
|
|
|
"precrq.ph.w %[n2], %[p2], %[scratch2] \n\t" /* pixel 2 */
|
|
|
|
|
"append %[p2], %[scratch2], 16 \n\t" /* pixel 1 */
|
|
|
|
|
|
|
|
|
|
"dpa.w.ph $ac0, %[p1], %[vector3b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac0, %[p2], %[vector4b] \n\t"
|
|
|
|
|
"extp %[Temp1], $ac0, 31 \n\t"
|
|
|
|
|
"dpa.w.ph $ac1, %[n1], %[vector3b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac1, %[n2], %[vector4b] \n\t"
|
|
|
|
|
"extp %[Temp2], $ac1, 31 \n\t"
|
|
|
|
|
|
|
|
|
|
"preceu.ph.qbl %[scratch1], %[load1] \n\t"
|
|
|
|
|
"preceu.ph.qbl %[p1], %[load2] \n\t"
|
|
|
|
|
"precrq.ph.w %[n1], %[p1], %[scratch1] \n\t" /* pixel 2 */
|
|
|
|
|
"append %[p1], %[scratch1], 16 \n\t" /* pixel 1 */
|
|
|
|
|
"preceu.ph.qbl %[scratch2], %[load3] \n\t"
|
|
|
|
|
"preceu.ph.qbl %[p2], %[load4] \n\t"
|
|
|
|
|
"precrq.ph.w %[n2], %[p2], %[scratch2] \n\t" /* pixel 2 */
|
|
|
|
|
"append %[p2], %[scratch2], 16 \n\t" /* pixel 1 */
|
|
|
|
|
|
|
|
|
|
"lbux %[store1], %[Temp1](%[cm]) \n\t"
|
|
|
|
|
"dpa.w.ph $ac2, %[p1], %[vector3b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac2, %[p2], %[vector4b] \n\t"
|
|
|
|
|
"extp %[Temp1], $ac2, 31 \n\t"
|
|
|
|
|
|
|
|
|
|
"lbux %[store2], %[Temp2](%[cm]) \n\t"
|
|
|
|
|
"dpa.w.ph $ac3, %[n1], %[vector3b] \n\t"
|
|
|
|
|
"dpa.w.ph $ac3, %[n2], %[vector4b] \n\t"
|
|
|
|
|
"extp %[Temp2], $ac3, 31 \n\t"
|
|
|
|
|
|
|
|
|
|
"sb %[store1], 0(%[dst_ptr]) \n\t"
|
|
|
|
|
"sb %[store2], 1(%[dst_ptr]) \n\t"
|
|
|
|
|
|
|
|
|
|
"lbux %[store1], %[Temp1](%[cm]) \n\t"
|
|
|
|
|
"lbux %[store2], %[Temp2](%[cm]) \n\t"
|
|
|
|
|
|
|
|
|
|
"sb %[store1], 2(%[dst_ptr]) \n\t"
|
|
|
|
|
"sb %[store2], 3(%[dst_ptr]) \n\t"
|
|
|
|
|
|
|
|
|
|
: [load1] "=&r" (load1), [load2] "=&r" (load2),
|
|
|
|
|
[load3] "=&r" (load3), [load4] "=&r" (load4),
|
|
|
|
|
[p1] "=&r" (p1), [p2] "=&r" (p2),
|
|
|
|
|
[n1] "=&r" (n1), [n2] "=&r" (n2),
|
|
|
|
|
[scratch1] "=&r" (scratch1), [scratch2] "=&r" (scratch2),
|
|
|
|
|
[Temp1] "=&r" (Temp1), [Temp2] "=&r" (Temp2),
|
|
|
|
|
[store1] "=&r" (store1), [store2] "=&r" (store2),
|
|
|
|
|
[src_ptr] "+r" (src_ptr)
|
|
|
|
|
: [vector1b] "r" (vector1b), [vector2b] "r" (vector2b),
|
|
|
|
|
[vector3b] "r" (vector3b), [vector4b] "r" (vector4b),
|
|
|
|
|
[vector4a] "r" (vector4a), [src_stride] "r" (src_stride),
|
|
|
|
|
[cm] "r" (cm), [dst_ptr] "r" (dst_ptr)
|
|
|
|
|
);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* Next row... */
|
|
|
|
|
src += src_stride;
|
|
|
|
|
dst += dst_stride;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2015-07-28 19:52:24 +02:00
|
|
|
void vpx_convolve8_vert_dspr2(const uint8_t *src, ptrdiff_t src_stride,
|
2013-09-13 11:48:32 +02:00
|
|
|
uint8_t *dst, ptrdiff_t dst_stride,
|
|
|
|
|
const int16_t *filter_x, int x_step_q4,
|
|
|
|
|
const int16_t *filter_y, int y_step_q4,
|
|
|
|
|
int w, int h) {
|
2015-08-13 20:27:04 +02:00
|
|
|
assert(y_step_q4 == 16);
|
2015-08-13 21:57:01 +02:00
|
|
|
assert(((const int32_t *)filter_y)[1] != 0x800000);
|
|
|
|
|
|
|
|
|
|
if (((const int32_t *)filter_y)[0] == 0) {
|
2015-07-28 19:52:24 +02:00
|
|
|
vpx_convolve2_vert_dspr2(src, src_stride,
|
2013-10-09 14:35:27 +02:00
|
|
|
dst, dst_stride,
|
|
|
|
|
filter_x, x_step_q4,
|
|
|
|
|
filter_y, y_step_q4,
|
|
|
|
|
w, h);
|
2013-09-13 11:48:32 +02:00
|
|
|
} else {
|
2015-08-13 20:27:04 +02:00
|
|
|
uint32_t pos = 38;
|
|
|
|
|
|
|
|
|
|
/* bit positon for extract from acc */
|
|
|
|
|
__asm__ __volatile__ (
|
|
|
|
|
"wrdsp %[pos], 1 \n\t"
|
|
|
|
|
:
|
|
|
|
|
: [pos] "r" (pos)
|
|
|
|
|
);
|
|
|
|
|
|
|
|
|
|
prefetch_store(dst);
|
|
|
|
|
|
|
|
|
|
switch (w) {
|
|
|
|
|
case 4 :
|
|
|
|
|
case 8 :
|
|
|
|
|
case 16 :
|
|
|
|
|
case 32 :
|
|
|
|
|
convolve_vert_4_dspr2(src, src_stride,
|
|
|
|
|
dst, dst_stride,
|
|
|
|
|
filter_y, w, h);
|
|
|
|
|
break;
|
|
|
|
|
case 64 :
|
|
|
|
|
prefetch_store(dst + 32);
|
|
|
|
|
convolve_vert_64_dspr2(src, src_stride,
|
2013-09-13 11:48:32 +02:00
|
|
|
dst, dst_stride,
|
2015-08-13 20:27:04 +02:00
|
|
|
filter_y, h);
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
vpx_convolve8_vert_c(src, src_stride,
|
|
|
|
|
dst, dst_stride,
|
|
|
|
|
filter_x, x_step_q4,
|
|
|
|
|
filter_y, y_step_q4,
|
|
|
|
|
w, h);
|
|
|
|
|
break;
|
2013-09-13 11:48:32 +02:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
#endif
|
