diff --git a/vp9/common/mips/dspr2/vp9_convolve2_dspr2.c b/vp9/common/mips/dspr2/vp9_convolve2_dspr2.c
index bc422bcf1083eb905a64adc0891c13e79e7b106b..4eefe39677cd81b66e5650f09918333d66aa0bd3 100644
--- a/vp9/common/mips/dspr2/vp9_convolve2_dspr2.c
+++ b/vp9/common/mips/dspr2/vp9_convolve2_dspr2.c
@@ -20,698 +20,698 @@
#include "vp9/common/mips/dspr2/vp9_common_dspr2.h"
#if HAVE_DSPR2
-static void convolve_bi_horiz_4_transposed_dspr2(const uint8_t *src,
- int32_t src_stride,
- uint8_t *dst,
- int32_t dst_stride,
- const int16_t *filter_x0,
- int32_t h) {
- int32_t y;
- uint8_t *cm = vp9_ff_cropTbl;
- uint8_t *dst_ptr;
- int32_t Temp1, Temp2;
- uint32_t vector4a = 64;
- uint32_t tp1, tp2;
- uint32_t p1, p2;
- const int16_t *filter = &filter_x0[3];
- uint32_t filter45;
-
- filter45 = ((const int32_t *)filter)[0];
-
- for (y = h; y--;) {
- dst_ptr = dst;
- /* prefetch data to cache memory */
- vp9_prefetch_load(src + src_stride);
- vp9_prefetch_load(src + src_stride + 32);
-
- __asm__ __volatile__ (
- "ulw %[tp1], 0(%[src]) \n\t"
- "ulw %[tp2], 4(%[src]) \n\t"
-
- /* even 1. pixel */
- "mtlo %[vector4a], $ac3 \n\t"
- "mthi $zero, $ac3 \n\t"
- "preceu.ph.qbr %[p1], %[tp1] \n\t"
- "preceu.ph.qbl %[p2], %[tp1] \n\t"
- "dpa.w.ph $ac3, %[p1], %[filter45] \n\t"
- "extp %[Temp1], $ac3, 31 \n\t"
-
- /* even 2. pixel */
- "mtlo %[vector4a], $ac2 \n\t"
- "mthi $zero, $ac2 \n\t"
- "balign %[tp2], %[tp1], 3 \n\t"
- "dpa.w.ph $ac2, %[p2], %[filter45] \n\t"
- "extp %[Temp2], $ac2, 31 \n\t"
-
- /* odd 1. pixel */
- "lbux %[tp1], %[Temp1](%[cm]) \n\t"
- "mtlo %[vector4a], $ac3 \n\t"
- "mthi $zero, $ac3 \n\t"
- "preceu.ph.qbr %[p1], %[tp2] \n\t"
- "preceu.ph.qbl %[p2], %[tp2] \n\t"
- "dpa.w.ph $ac3, %[p1], %[filter45] \n\t"
- "extp %[Temp1], $ac3, 31 \n\t"
-
- /* odd 2. pixel */
- "lbux %[tp2], %[Temp2](%[cm]) \n\t"
- "mtlo %[vector4a], $ac2 \n\t"
- "mthi $zero, $ac2 \n\t"
- "dpa.w.ph $ac2, %[p2], %[filter45] \n\t"
- "extp %[Temp2], $ac2, 31 \n\t"
-
- /* clamp */
- "lbux %[p1], %[Temp1](%[cm]) \n\t"
- "lbux %[p2], %[Temp2](%[cm]) \n\t"
-
- /* store bytes */
- "sb %[tp1], 0(%[dst_ptr]) \n\t"
- "addu %[dst_ptr], %[dst_ptr], %[dst_stride] \n\t"
-
- "sb %[p1], 0(%[dst_ptr]) \n\t"
- "addu %[dst_ptr], %[dst_ptr], %[dst_stride] \n\t"
-
- "sb %[tp2], 0(%[dst_ptr]) \n\t"
- "addu %[dst_ptr], %[dst_ptr], %[dst_stride] \n\t"
-
- "sb %[p2], 0(%[dst_ptr]) \n\t"
- "addu %[dst_ptr], %[dst_ptr], %[dst_stride] \n\t"
-
- : [tp1] "=&r" (tp1), [tp2] "=&r" (tp2),
- [p1] "=&r" (p1), [p2] "=&r" (p2),
- [Temp1] "=&r" (Temp1), [Temp2] "=&r" (Temp2),
- [dst_ptr] "+r" (dst_ptr)
- : [filter45] "r" (filter45),[vector4a] "r" (vector4a),
- [cm] "r" (cm), [src] "r" (src), [dst_stride] "r" (dst_stride)
- );
-
- /* Next row... */
- src += src_stride;
- dst += 1;
- }
+static void convolve_bi_horiz_4_transposed_dspr2(const uint8_t *src,
+ int32_t src_stride,
+ uint8_t *dst,
+ int32_t dst_stride,
+ const int16_t *filter_x0,
+ int32_t h) {
+ int32_t y;
+ uint8_t *cm = vp9_ff_cropTbl;
+ uint8_t *dst_ptr;
+ int32_t Temp1, Temp2;
+ uint32_t vector4a = 64;
+ uint32_t tp1, tp2;
+ uint32_t p1, p2;
+ const int16_t *filter = &filter_x0[3];
+ uint32_t filter45;
+
+ filter45 = ((const int32_t *)filter)[0];
+
+ for (y = h; y--;) {
+ dst_ptr = dst;
+ /* prefetch data to cache memory */
+ vp9_prefetch_load(src + src_stride);
+ vp9_prefetch_load(src + src_stride + 32);
+
+ __asm__ __volatile__ (
+ "ulw %[tp1], 0(%[src]) \n\t"
+ "ulw %[tp2], 4(%[src]) \n\t"
+
+ /* even 1. pixel */
+ "mtlo %[vector4a], $ac3 \n\t"
+ "mthi $zero, $ac3 \n\t"
+ "preceu.ph.qbr %[p1], %[tp1] \n\t"
+ "preceu.ph.qbl %[p2], %[tp1] \n\t"
+ "dpa.w.ph $ac3, %[p1], %[filter45] \n\t"
+ "extp %[Temp1], $ac3, 31 \n\t"
+
+ /* even 2. pixel */
+ "mtlo %[vector4a], $ac2 \n\t"
+ "mthi $zero, $ac2 \n\t"
+ "balign %[tp2], %[tp1], 3 \n\t"
+ "dpa.w.ph $ac2, %[p2], %[filter45] \n\t"
+ "extp %[Temp2], $ac2, 31 \n\t"
+
+ /* odd 1. pixel */
+ "lbux %[tp1], %[Temp1](%[cm]) \n\t"
+ "mtlo %[vector4a], $ac3 \n\t"
+ "mthi $zero, $ac3 \n\t"
+ "preceu.ph.qbr %[p1], %[tp2] \n\t"
+ "preceu.ph.qbl %[p2], %[tp2] \n\t"
+ "dpa.w.ph $ac3, %[p1], %[filter45] \n\t"
+ "extp %[Temp1], $ac3, 31 \n\t"
+
+ /* odd 2. pixel */
+ "lbux %[tp2], %[Temp2](%[cm]) \n\t"
+ "mtlo %[vector4a], $ac2 \n\t"
+ "mthi $zero, $ac2 \n\t"
+ "dpa.w.ph $ac2, %[p2], %[filter45] \n\t"
+ "extp %[Temp2], $ac2, 31 \n\t"
+
+ /* clamp */
+ "lbux %[p1], %[Temp1](%[cm]) \n\t"
+ "lbux %[p2], %[Temp2](%[cm]) \n\t"
+
+ /* store bytes */
+ "sb %[tp1], 0(%[dst_ptr]) \n\t"
+ "addu %[dst_ptr], %[dst_ptr], %[dst_stride] \n\t"
+
+ "sb %[p1], 0(%[dst_ptr]) \n\t"
+ "addu %[dst_ptr], %[dst_ptr], %[dst_stride] \n\t"
+
+ "sb %[tp2], 0(%[dst_ptr]) \n\t"
+ "addu %[dst_ptr], %[dst_ptr], %[dst_stride] \n\t"
+
+ "sb %[p2], 0(%[dst_ptr]) \n\t"
+ "addu %[dst_ptr], %[dst_ptr], %[dst_stride] \n\t"
+
+ : [tp1] "=&r" (tp1), [tp2] "=&r" (tp2),
+ [p1] "=&r" (p1), [p2] "=&r" (p2),
+ [Temp1] "=&r" (Temp1), [Temp2] "=&r" (Temp2),
+ [dst_ptr] "+r" (dst_ptr)
+ : [filter45] "r" (filter45),[vector4a] "r" (vector4a),
+ [cm] "r" (cm), [src] "r" (src), [dst_stride] "r" (dst_stride)
+ );
+
+ /* Next row... */
+ src += src_stride;
+ dst += 1;
+ }
}
-static void convolve_bi_horiz_8_transposed_dspr2(const uint8_t *src,
- int32_t src_stride,
- uint8_t *dst,
- int32_t dst_stride,
- const int16_t *filter_x0,
- int32_t h) {
- int32_t y;
- uint8_t *cm = vp9_ff_cropTbl;
- uint8_t *dst_ptr;
- uint32_t vector4a = 64;
- int32_t Temp1, Temp2, Temp3;
- uint32_t tp1, tp2, tp3;
- uint32_t p1, p2, p3, p4;
- uint8_t *odd_dst;
- uint32_t dst_pitch_2 = (dst_stride << 1);
- const int16_t *filter = &filter_x0[3];
- uint32_t filter45;
-
- filter45 = ((const int32_t *)filter)[0];
-
- for (y = h; y--;) {
- /* prefetch data to cache memory */
- vp9_prefetch_load(src + src_stride);
- vp9_prefetch_load(src + src_stride + 32);
-
- dst_ptr = dst;
- odd_dst = (dst_ptr + dst_stride);
-
- __asm__ __volatile__ (
- "ulw %[tp1], 0(%[src]) \n\t"
- "ulw %[tp2], 4(%[src]) \n\t"
-
- /* even 1. pixel */
- "mtlo %[vector4a], $ac3 \n\t"
- "mthi $zero, $ac3 \n\t"
- "mtlo %[vector4a], $ac2 \n\t"
- "mthi $zero, $ac2 \n\t"
- "preceu.ph.qbr %[p1], %[tp1] \n\t"
- "preceu.ph.qbl %[p2], %[tp1] \n\t"
- "preceu.ph.qbr %[p3], %[tp2] \n\t"
- "preceu.ph.qbl %[p4], %[tp2] \n\t"
- "ulw %[tp3], 8(%[src]) \n\t"
- "dpa.w.ph $ac3, %[p1], %[filter45] \n\t"
- "extp %[Temp1], $ac3, 31 \n\t"
-
- /* even 2. pixel */
- "dpa.w.ph $ac2, %[p2], %[filter45] \n\t"
- "extp %[Temp3], $ac2, 31 \n\t"
-
- /* even 3. pixel */
- "lbux %[Temp2], %[Temp1](%[cm]) \n\t"
- "mtlo %[vector4a], $ac1 \n\t"
- "mthi $zero, $ac1 \n\t"
- "balign %[tp3], %[tp2], 3 \n\t"
- "balign %[tp2], %[tp1], 3 \n\t"
- "dpa.w.ph $ac1, %[p3], %[filter45] \n\t"
- "lbux %[tp1], %[Temp3](%[cm]) \n\t"
- "extp %[p3], $ac1, 31 \n\t"
-
- /* even 4. pixel */
- "mtlo %[vector4a], $ac2 \n\t"
- "mthi $zero, $ac2 \n\t"
- "mtlo %[vector4a], $ac3 \n\t"
- "mthi $zero, $ac3 \n\t"
- "sb %[Temp2], 0(%[dst_ptr]) \n\t"
- "addu %[dst_ptr], %[dst_ptr], %[dst_pitch_2] \n\t"
- "sb %[tp1], 0(%[dst_ptr]) \n\t"
- "addu %[dst_ptr], %[dst_ptr], %[dst_pitch_2] \n\t"
-
- "dpa.w.ph $ac2, %[p4], %[filter45] \n\t"
- "extp %[Temp3], $ac2, 31 \n\t"
-
- "lbux %[Temp1], %[p3](%[cm]) \n\t"
-
- /* odd 1. pixel */
- "mtlo %[vector4a], $ac1 \n\t"
- "mthi $zero, $ac1 \n\t"
- "preceu.ph.qbr %[p1], %[tp2] \n\t"
- "preceu.ph.qbl %[p2], %[tp2] \n\t"
- "preceu.ph.qbr %[p3], %[tp3] \n\t"
- "preceu.ph.qbl %[p4], %[tp3] \n\t"
- "sb %[Temp1], 0(%[dst_ptr]) \n\t"
- "addu %[dst_ptr], %[dst_ptr], %[dst_pitch_2] \n\t"
-
- "dpa.w.ph $ac3, %[p1], %[filter45] \n\t"
- "extp %[Temp2], $ac3, 31 \n\t"
-
- /* odd 2. pixel */
- "lbux %[tp1], %[Temp3](%[cm]) \n\t"
- "mtlo %[vector4a], $ac3 \n\t"
- "mthi $zero, $ac3 \n\t"
- "mtlo %[vector4a], $ac2 \n\t"
- "mthi $zero, $ac2 \n\t"
- "dpa.w.ph $ac1, %[p2], %[filter45] \n\t"
- "sb %[tp1], 0(%[dst_ptr]) \n\t"
- "addu %[dst_ptr], %[dst_ptr], %[dst_pitch_2] \n\t"
- "extp %[Temp3], $ac1, 31 \n\t"
-
- /* odd 3. pixel */
- "lbux %[tp3], %[Temp2](%[cm]) \n\t"
- "dpa.w.ph $ac3, %[p3], %[filter45] \n\t"
- "extp %[Temp2], $ac3, 31 \n\t"
-
- /* odd 4. pixel */
- "sb %[tp3], 0(%[odd_dst]) \n\t"
- "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
- "dpa.w.ph $ac2, %[p4], %[filter45] \n\t"
- "extp %[Temp1], $ac2, 31 \n\t"
-
- /* clamp */
- "lbux %[p4], %[Temp3](%[cm]) \n\t"
- "lbux %[p2], %[Temp2](%[cm]) \n\t"
- "lbux %[p1], %[Temp1](%[cm]) \n\t"
-
- /* store bytes */
- "sb %[p4], 0(%[odd_dst]) \n\t"
- "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
-
- "sb %[p2], 0(%[odd_dst]) \n\t"
- "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
-
- "sb %[p1], 0(%[odd_dst]) \n\t"
-
- : [tp1] "=&r" (tp1), [tp2] "=&r" (tp2), [tp3] "=&r" (tp3),
- [p1] "=&r" (p1), [p2] "=&r" (p2), [p3] "=&r" (p3), [p4] "=&r" (p4),
- [Temp1] "=&r" (Temp1), [Temp2] "=&r" (Temp2), [Temp3] "=&r" (Temp3),
- [dst_ptr] "+r" (dst_ptr), [odd_dst] "+r" (odd_dst)
- : [filter45] "r" (filter45),[vector4a] "r" (vector4a), [cm] "r" (cm),
- [src] "r" (src), [dst_pitch_2] "r" (dst_pitch_2)
- );
-
- /* Next row... */
- src += src_stride;
- dst += 1;
- }
+static void convolve_bi_horiz_8_transposed_dspr2(const uint8_t *src,
+ int32_t src_stride,
+ uint8_t *dst,
+ int32_t dst_stride,
+ const int16_t *filter_x0,
+ int32_t h) {
+ int32_t y;
+ uint8_t *cm = vp9_ff_cropTbl;
+ uint8_t *dst_ptr;
+ uint32_t vector4a = 64;
+ int32_t Temp1, Temp2, Temp3;
+ uint32_t tp1, tp2, tp3;
+ uint32_t p1, p2, p3, p4;
+ uint8_t *odd_dst;
+ uint32_t dst_pitch_2 = (dst_stride << 1);
+ const int16_t *filter = &filter_x0[3];
+ uint32_t filter45;
+
+ filter45 = ((const int32_t *)filter)[0];
+
+ for (y = h; y--;) {
+ /* prefetch data to cache memory */
+ vp9_prefetch_load(src + src_stride);
+ vp9_prefetch_load(src + src_stride + 32);
+
+ dst_ptr = dst;
+ odd_dst = (dst_ptr + dst_stride);
+
+ __asm__ __volatile__ (
+ "ulw %[tp1], 0(%[src]) \n\t"
+ "ulw %[tp2], 4(%[src]) \n\t"
+
+ /* even 1. pixel */
+ "mtlo %[vector4a], $ac3 \n\t"
+ "mthi $zero, $ac3 \n\t"
+ "mtlo %[vector4a], $ac2 \n\t"
+ "mthi $zero, $ac2 \n\t"
+ "preceu.ph.qbr %[p1], %[tp1] \n\t"
+ "preceu.ph.qbl %[p2], %[tp1] \n\t"
+ "preceu.ph.qbr %[p3], %[tp2] \n\t"
+ "preceu.ph.qbl %[p4], %[tp2] \n\t"
+ "ulw %[tp3], 8(%[src]) \n\t"
+ "dpa.w.ph $ac3, %[p1], %[filter45] \n\t"
+ "extp %[Temp1], $ac3, 31 \n\t"
+
+ /* even 2. pixel */
+ "dpa.w.ph $ac2, %[p2], %[filter45] \n\t"
+ "extp %[Temp3], $ac2, 31 \n\t"
+
+ /* even 3. pixel */
+ "lbux %[Temp2], %[Temp1](%[cm]) \n\t"
+ "mtlo %[vector4a], $ac1 \n\t"
+ "mthi $zero, $ac1 \n\t"
+ "balign %[tp3], %[tp2], 3 \n\t"
+ "balign %[tp2], %[tp1], 3 \n\t"
+ "dpa.w.ph $ac1, %[p3], %[filter45] \n\t"
+ "lbux %[tp1], %[Temp3](%[cm]) \n\t"
+ "extp %[p3], $ac1, 31 \n\t"
+
+ /* even 4. pixel */
+ "mtlo %[vector4a], $ac2 \n\t"
+ "mthi $zero, $ac2 \n\t"
+ "mtlo %[vector4a], $ac3 \n\t"
+ "mthi $zero, $ac3 \n\t"
+ "sb %[Temp2], 0(%[dst_ptr]) \n\t"
+ "addu %[dst_ptr], %[dst_ptr], %[dst_pitch_2] \n\t"
+ "sb %[tp1], 0(%[dst_ptr]) \n\t"
+ "addu %[dst_ptr], %[dst_ptr], %[dst_pitch_2] \n\t"
+
+ "dpa.w.ph $ac2, %[p4], %[filter45] \n\t"
+ "extp %[Temp3], $ac2, 31 \n\t"
+
+ "lbux %[Temp1], %[p3](%[cm]) \n\t"
+
+ /* odd 1. pixel */
+ "mtlo %[vector4a], $ac1 \n\t"
+ "mthi $zero, $ac1 \n\t"
+ "preceu.ph.qbr %[p1], %[tp2] \n\t"
+ "preceu.ph.qbl %[p2], %[tp2] \n\t"
+ "preceu.ph.qbr %[p3], %[tp3] \n\t"
+ "preceu.ph.qbl %[p4], %[tp3] \n\t"
+ "sb %[Temp1], 0(%[dst_ptr]) \n\t"
+ "addu %[dst_ptr], %[dst_ptr], %[dst_pitch_2] \n\t"
+
+ "dpa.w.ph $ac3, %[p1], %[filter45] \n\t"
+ "extp %[Temp2], $ac3, 31 \n\t"
+
+ /* odd 2. pixel */
+ "lbux %[tp1], %[Temp3](%[cm]) \n\t"
+ "mtlo %[vector4a], $ac3 \n\t"
+ "mthi $zero, $ac3 \n\t"
+ "mtlo %[vector4a], $ac2 \n\t"
+ "mthi $zero, $ac2 \n\t"
+ "dpa.w.ph $ac1, %[p2], %[filter45] \n\t"
+ "sb %[tp1], 0(%[dst_ptr]) \n\t"
+ "addu %[dst_ptr], %[dst_ptr], %[dst_pitch_2] \n\t"
+ "extp %[Temp3], $ac1, 31 \n\t"
+
+ /* odd 3. pixel */
+ "lbux %[tp3], %[Temp2](%[cm]) \n\t"
+ "dpa.w.ph $ac3, %[p3], %[filter45] \n\t"
+ "extp %[Temp2], $ac3, 31 \n\t"
+
+ /* odd 4. pixel */
+ "sb %[tp3], 0(%[odd_dst]) \n\t"
+ "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
+ "dpa.w.ph $ac2, %[p4], %[filter45] \n\t"
+ "extp %[Temp1], $ac2, 31 \n\t"
+
+ /* clamp */
+ "lbux %[p4], %[Temp3](%[cm]) \n\t"
+ "lbux %[p2], %[Temp2](%[cm]) \n\t"
+ "lbux %[p1], %[Temp1](%[cm]) \n\t"
+
+ /* store bytes */
+ "sb %[p4], 0(%[odd_dst]) \n\t"
+ "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
+
+ "sb %[p2], 0(%[odd_dst]) \n\t"
+ "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
+
+ "sb %[p1], 0(%[odd_dst]) \n\t"
+
+ : [tp1] "=&r" (tp1), [tp2] "=&r" (tp2), [tp3] "=&r" (tp3),
+ [p1] "=&r" (p1), [p2] "=&r" (p2), [p3] "=&r" (p3), [p4] "=&r" (p4),
+ [Temp1] "=&r" (Temp1), [Temp2] "=&r" (Temp2), [Temp3] "=&r" (Temp3),
+ [dst_ptr] "+r" (dst_ptr), [odd_dst] "+r" (odd_dst)
+ : [filter45] "r" (filter45),[vector4a] "r" (vector4a), [cm] "r" (cm),
+ [src] "r" (src), [dst_pitch_2] "r" (dst_pitch_2)
+ );
+
+ /* Next row... */
+ src += src_stride;
+ dst += 1;
+ }
+}
+
+static void convolve_bi_horiz_16_transposed_dspr2(const uint8_t *src_ptr,
+ int32_t src_stride,
+ uint8_t *dst_ptr,
+ int32_t dst_stride,
+ const int16_t *filter_x0,
+ int32_t h,
+ int32_t count) {
+ int32_t c, y;
+ const uint8_t *src;
+ uint8_t *dst;
+ uint8_t *cm = vp9_ff_cropTbl;
+ uint32_t vector_64 = 64;
+ int32_t Temp1, Temp2, Temp3;
+ uint32_t qload1, qload2;
+ uint32_t p1, p2, p3, p4, p5;
+ uint32_t st1, st2, st3;
+ uint32_t dst_pitch_2 = (dst_stride << 1);
+ uint8_t *odd_dst;
+ const int16_t *filter = &filter_x0[3];
+ uint32_t filter45;
+
+ filter45 = ((const int32_t *)filter)[0];
+
+ for (y = h; y--;) {
+ /* prefetch data to cache memory */
+ vp9_prefetch_load(src_ptr + src_stride);
+ vp9_prefetch_load(src_ptr + src_stride + 32);
+
+ src = src_ptr;
+ dst = dst_ptr;
+
+ odd_dst = (dst + dst_stride);
+
+ for (c = 0; c < count; c++) {
+ __asm__ __volatile__ (
+ "ulw %[qload1], 0(%[src]) \n\t"
+ "ulw %[qload2], 4(%[src]) \n\t"
+
+ /* even 1. pixel */
+ "mtlo %[vector_64], $ac1 \n\t" /* even 1 */
+ "mthi $zero, $ac1 \n\t"
+ "mtlo %[vector_64], $ac2 \n\t" /* even 2 */
+ "mthi $zero, $ac2 \n\t"
+ "preceu.ph.qbr %[p1], %[qload1] \n\t"
+ "preceu.ph.qbl %[p2], %[qload1] \n\t"
+ "preceu.ph.qbr %[p3], %[qload2] \n\t"
+ "preceu.ph.qbl %[p4], %[qload2] \n\t"
+ "ulw %[qload1], 8(%[src]) \n\t"
+ "dpa.w.ph $ac1, %[p1], %[filter45] \n\t" /* even 1 */
+ "extp %[Temp1], $ac1, 31 \n\t" /* even 1 */
+
+ /* even 2. pixel */
+ "mtlo %[vector_64], $ac3 \n\t" /* even 3 */
+ "mthi $zero, $ac3 \n\t"
+ "preceu.ph.qbr %[p1], %[qload1] \n\t"
+ "preceu.ph.qbl %[p5], %[qload1] \n\t"
+ "ulw %[qload2], 12(%[src]) \n\t"
+ "dpa.w.ph $ac2, %[p2], %[filter45] \n\t" /* even 1 */
+ "lbux %[st1], %[Temp1](%[cm]) \n\t" /* even 1 */
+ "extp %[Temp2], $ac2, 31 \n\t" /* even 1 */
+
+ /* even 3. pixel */
+ "mtlo %[vector_64], $ac1 \n\t" /* even 4 */
+ "mthi $zero, $ac1 \n\t"
+ "preceu.ph.qbr %[p2], %[qload2] \n\t"
+ "sb %[st1], 0(%[dst]) \n\t" /* even 1 */
+ "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
+ "dpa.w.ph $ac3, %[p3], %[filter45] \n\t" /* even 3 */
+ "extp %[Temp3], $ac3, 31 \n\t" /* even 3 */
+ "lbux %[st2], %[Temp2](%[cm]) \n\t" /* even 1 */
+
+ /* even 4. pixel */
+ "mtlo %[vector_64], $ac2 \n\t" /* even 5 */
+ "mthi $zero, $ac2 \n\t"
+ "preceu.ph.qbl %[p3], %[qload2] \n\t"
+ "sb %[st2], 0(%[dst]) \n\t" /* even 2 */
+ "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
+ "dpa.w.ph $ac1, %[p4], %[filter45] \n\t" /* even 4 */
+ "extp %[Temp1], $ac1, 31 \n\t" /* even 4 */
+ "lbux %[st3], %[Temp3](%[cm]) \n\t" /* even 3 */
+
+ /* even 5. pixel */
+ "mtlo %[vector_64], $ac3 \n\t" /* even 6 */
+ "mthi $zero, $ac3 \n\t"
+ "sb %[st3], 0(%[dst]) \n\t" /* even 3 */
+ "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
+ "dpa.w.ph $ac2, %[p1], %[filter45] \n\t" /* even 5 */
+ "extp %[Temp2], $ac2, 31 \n\t" /* even 5 */
+ "lbux %[st1], %[Temp1](%[cm]) \n\t" /* even 4 */
+
+ /* even 6. pixel */
+ "mtlo %[vector_64], $ac1 \n\t" /* even 7 */
+ "mthi $zero, $ac1 \n\t"
+ "sb %[st1], 0(%[dst]) \n\t" /* even 4 */
+ "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
+ "ulw %[qload1], 20(%[src]) \n\t"
+ "dpa.w.ph $ac3, %[p5], %[filter45] \n\t" /* even 6 */
+ "extp %[Temp3], $ac3, 31 \n\t" /* even 6 */
+ "lbux %[st2], %[Temp2](%[cm]) \n\t" /* even 5 */
+
+ /* even 7. pixel */
+ "mtlo %[vector_64], $ac2 \n\t" /* even 8 */
+ "mthi $zero, $ac2 \n\t"
+ "preceu.ph.qbr %[p5], %[qload1] \n\t"
+ "sb %[st2], 0(%[dst]) \n\t" /* even 5 */
+ "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
+ "dpa.w.ph $ac1, %[p2], %[filter45] \n\t" /* even 7 */
+ "extp %[Temp1], $ac1, 31 \n\t" /* even 7 */
+ "lbux %[st3], %[Temp3](%[cm]) \n\t" /* even 6 */
+
+ /* even 8. pixel */
+ "mtlo %[vector_64], $ac3 \n\t" /* odd 1 */
+ "mthi $zero, $ac3 \n\t"
+ "dpa.w.ph $ac2, %[p3], %[filter45] \n\t" /* even 8 */
+ "sb %[st3], 0(%[dst]) \n\t" /* even 6 */
+ "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
+ "extp %[Temp2], $ac2, 31 \n\t" /* even 8 */
+ "lbux %[st1], %[Temp1](%[cm]) \n\t" /* even 7 */
+
+ /* ODD pixels */
+ "ulw %[qload1], 1(%[src]) \n\t"
+ "ulw %[qload2], 5(%[src]) \n\t"
+
+ /* odd 1. pixel */
+ "mtlo %[vector_64], $ac1 \n\t" /* odd 2 */
+ "mthi $zero, $ac1 \n\t"
+ "preceu.ph.qbr %[p1], %[qload1] \n\t"
+ "preceu.ph.qbl %[p2], %[qload1] \n\t"
+ "preceu.ph.qbr %[p3], %[qload2] \n\t"
+ "preceu.ph.qbl %[p4], %[qload2] \n\t"
+ "sb %[st1], 0(%[dst]) \n\t" /* even 7 */
+ "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
+ "ulw %[qload2], 9(%[src]) \n\t"
+ "dpa.w.ph $ac3, %[p1], %[filter45] \n\t" /* odd 1 */
+ "extp %[Temp3], $ac3, 31 \n\t" /* odd 1 */
+ "lbux %[st2], %[Temp2](%[cm]) \n\t" /* even 8 */
+
+ /* odd 2. pixel */
+ "mtlo %[vector_64], $ac2 \n\t" /* odd 3 */
+ "mthi $zero, $ac2 \n\t"
+ "preceu.ph.qbr %[p1], %[qload2] \n\t"
+ "preceu.ph.qbl %[p5], %[qload2] \n\t"
+ "sb %[st2], 0(%[dst]) \n\t" /* even 8 */
+ "ulw %[qload1], 13(%[src]) \n\t"
+ "dpa.w.ph $ac1, %[p2], %[filter45] \n\t" /* odd 2 */
+ "extp %[Temp1], $ac1, 31 \n\t" /* odd 2 */
+ "lbux %[st3], %[Temp3](%[cm]) \n\t" /* odd 1 */
+
+ /* odd 3. pixel */
+ "mtlo %[vector_64], $ac3 \n\t" /* odd 4 */
+ "mthi $zero, $ac3 \n\t"
+ "preceu.ph.qbr %[p2], %[qload1] \n\t"
+ "sb %[st3], 0(%[odd_dst]) \n\t" /* odd 1 */
+ "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
+ "dpa.w.ph $ac2, %[p3], %[filter45] \n\t" /* odd 3 */
+ "extp %[Temp2], $ac2, 31 \n\t" /* odd 3 */
+ "lbux %[st1], %[Temp1](%[cm]) \n\t" /* odd 2 */
+
+ /* odd 4. pixel */
+ "mtlo %[vector_64], $ac1 \n\t" /* odd 5 */
+ "mthi $zero, $ac1 \n\t"
+ "preceu.ph.qbl %[p3], %[qload1] \n\t"
+ "sb %[st1], 0(%[odd_dst]) \n\t" /* odd 2 */
+ "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
+ "dpa.w.ph $ac3, %[p4], %[filter45] \n\t" /* odd 4 */
+ "extp %[Temp3], $ac3, 31 \n\t" /* odd 4 */
+ "lbux %[st2], %[Temp2](%[cm]) \n\t" /* odd 3 */
+
+ /* odd 5. pixel */
+ "mtlo %[vector_64], $ac2 \n\t" /* odd 6 */
+ "mthi $zero, $ac2 \n\t"
+ "sb %[st2], 0(%[odd_dst]) \n\t" /* odd 3 */
+ "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
+ "dpa.w.ph $ac1, %[p1], %[filter45] \n\t" /* odd 5 */
+ "extp %[Temp1], $ac1, 31 \n\t" /* odd 5 */
+ "lbux %[st3], %[Temp3](%[cm]) \n\t" /* odd 4 */
+
+ /* odd 6. pixel */
+ "mtlo %[vector_64], $ac3 \n\t" /* odd 7 */
+ "mthi $zero, $ac3 \n\t"
+ "sb %[st3], 0(%[odd_dst]) \n\t" /* odd 4 */
+ "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
+ "ulw %[qload1], 21(%[src]) \n\t"
+ "dpa.w.ph $ac2, %[p5], %[filter45] \n\t" /* odd 6 */
+ "extp %[Temp2], $ac2, 31 \n\t" /* odd 6 */
+ "lbux %[st1], %[Temp1](%[cm]) \n\t" /* odd 5 */
+
+ /* odd 7. pixel */
+ "mtlo %[vector_64], $ac1 \n\t" /* odd 8 */
+ "mthi $zero, $ac1 \n\t"
+ "preceu.ph.qbr %[p5], %[qload1] \n\t"
+ "sb %[st1], 0(%[odd_dst]) \n\t" /* odd 5 */
+ "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
+ "dpa.w.ph $ac3, %[p2], %[filter45] \n\t" /* odd 7 */
+ "extp %[Temp3], $ac3, 31 \n\t" /* odd 7 */
+
+ /* odd 8. pixel */
+ "dpa.w.ph $ac1, %[p3], %[filter45] \n\t" /* odd 8 */
+ "extp %[Temp1], $ac1, 31 \n\t" /* odd 8 */
+
+ "lbux %[st2], %[Temp2](%[cm]) \n\t" /* odd 6 */
+ "lbux %[st3], %[Temp3](%[cm]) \n\t" /* odd 7 */
+ "lbux %[st1], %[Temp1](%[cm]) \n\t" /* odd 8 */
+
+ "sb %[st2], 0(%[odd_dst]) \n\t" /* odd 6 */
+ "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
+
+ "sb %[st3], 0(%[odd_dst]) \n\t" /* odd 7 */
+ "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
+
+ "sb %[st1], 0(%[odd_dst]) \n\t" /* odd 8 */
+
+ : [qload1] "=&r" (qload1), [qload2] "=&r" (qload2), [p5] "=&r" (p5),
+ [st1] "=&r" (st1), [st2] "=&r" (st2), [st3] "=&r" (st3),
+ [p1] "=&r" (p1), [p2] "=&r" (p2), [p3] "=&r" (p3), [p4] "=&r" (p4),
+ [Temp1] "=&r" (Temp1), [Temp2] "=&r" (Temp2), [Temp3] "=&r" (Temp3),
+ [dst] "+r" (dst), [odd_dst] "+r" (odd_dst)
+ : [filter45] "r" (filter45), [vector_64] "r" (vector_64),
+ [cm] "r" (cm),
+ [src] "r" (src), [dst_pitch_2] "r" (dst_pitch_2)
+ );
+
+ src += 16;
+ dst = (dst_ptr + ((c + 1) * 16 * dst_stride));
+ odd_dst = (dst + dst_stride);
+ }
+
+ /* Next row... */
+ src_ptr += src_stride;
+ dst_ptr += 1;
+ }
}
-static void convolve_bi_horiz_16_transposed_dspr2(const uint8_t *src_ptr,
- int32_t src_stride,
- uint8_t *dst_ptr,
- int32_t dst_stride,
- const int16_t *filter_x0,
- int32_t h,
- int32_t count) {
- int32_t c, y;
- const uint8_t *src;
- uint8_t *dst;
- uint8_t *cm = vp9_ff_cropTbl;
- uint32_t vector_64 = 64;
- int32_t Temp1, Temp2, Temp3;
- uint32_t qload1, qload2;
- uint32_t p1, p2, p3, p4, p5;
- uint32_t st1, st2, st3;
- uint32_t dst_pitch_2 = (dst_stride << 1);
- uint8_t *odd_dst;
- const int16_t *filter = &filter_x0[3];
- uint32_t filter45;
-
- filter45 = ((const int32_t *)filter)[0];
-
- for (y = h; y--;) {
- /* prefetch data to cache memory */
- vp9_prefetch_load(src_ptr + src_stride);
- vp9_prefetch_load(src_ptr + src_stride + 32);
-
- src = src_ptr;
- dst = dst_ptr;
-
- odd_dst = (dst + dst_stride);
-
- for (c = 0; c < count; c++) {
- __asm__ __volatile__ (
- "ulw %[qload1], 0(%[src]) \n\t"
- "ulw %[qload2], 4(%[src]) \n\t"
-
- /* even 1. pixel */
- "mtlo %[vector_64], $ac1 \n\t" /* even 1 */
- "mthi $zero, $ac1 \n\t"
- "mtlo %[vector_64], $ac2 \n\t" /* even 2 */
- "mthi $zero, $ac2 \n\t"
- "preceu.ph.qbr %[p1], %[qload1] \n\t"
- "preceu.ph.qbl %[p2], %[qload1] \n\t"
- "preceu.ph.qbr %[p3], %[qload2] \n\t"
- "preceu.ph.qbl %[p4], %[qload2] \n\t"
- "ulw %[qload1], 8(%[src]) \n\t"
- "dpa.w.ph $ac1, %[p1], %[filter45] \n\t" /* even 1 */
- "extp %[Temp1], $ac1, 31 \n\t" /* even 1 */
-
- /* even 2. pixel */
- "mtlo %[vector_64], $ac3 \n\t" /* even 3 */
- "mthi $zero, $ac3 \n\t"
- "preceu.ph.qbr %[p1], %[qload1] \n\t"
- "preceu.ph.qbl %[p5], %[qload1] \n\t"
- "ulw %[qload2], 12(%[src]) \n\t"
- "dpa.w.ph $ac2, %[p2], %[filter45] \n\t" /* even 1 */
- "lbux %[st1], %[Temp1](%[cm]) \n\t" /* even 1 */
- "extp %[Temp2], $ac2, 31 \n\t" /* even 1 */
-
- /* even 3. pixel */
- "mtlo %[vector_64], $ac1 \n\t" /* even 4 */
- "mthi $zero, $ac1 \n\t"
- "preceu.ph.qbr %[p2], %[qload2] \n\t"
- "sb %[st1], 0(%[dst]) \n\t" /* even 1 */
- "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
- "dpa.w.ph $ac3, %[p3], %[filter45] \n\t" /* even 3 */
- "extp %[Temp3], $ac3, 31 \n\t" /* even 3 */
- "lbux %[st2], %[Temp2](%[cm]) \n\t" /* even 1 */
-
- /* even 4. pixel */
- "mtlo %[vector_64], $ac2 \n\t" /* even 5 */
- "mthi $zero, $ac2 \n\t"
- "preceu.ph.qbl %[p3], %[qload2] \n\t"
- "sb %[st2], 0(%[dst]) \n\t" /* even 2 */
- "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
- "dpa.w.ph $ac1, %[p4], %[filter45] \n\t" /* even 4 */
- "extp %[Temp1], $ac1, 31 \n\t" /* even 4 */
- "lbux %[st3], %[Temp3](%[cm]) \n\t" /* even 3 */
-
- /* even 5. pixel */
- "mtlo %[vector_64], $ac3 \n\t" /* even 6 */
- "mthi $zero, $ac3 \n\t"
- "sb %[st3], 0(%[dst]) \n\t" /* even 3 */
- "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
- "dpa.w.ph $ac2, %[p1], %[filter45] \n\t" /* even 5 */
- "extp %[Temp2], $ac2, 31 \n\t" /* even 5 */
- "lbux %[st1], %[Temp1](%[cm]) \n\t" /* even 4 */
-
- /* even 6. pixel */
- "mtlo %[vector_64], $ac1 \n\t" /* even 7 */
- "mthi $zero, $ac1 \n\t"
- "sb %[st1], 0(%[dst]) \n\t" /* even 4 */
- "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
- "ulw %[qload1], 20(%[src]) \n\t"
- "dpa.w.ph $ac3, %[p5], %[filter45] \n\t" /* even 6 */
- "extp %[Temp3], $ac3, 31 \n\t" /* even 6 */
- "lbux %[st2], %[Temp2](%[cm]) \n\t" /* even 5 */
-
- /* even 7. pixel */
- "mtlo %[vector_64], $ac2 \n\t" /* even 8 */
- "mthi $zero, $ac2 \n\t"
- "preceu.ph.qbr %[p5], %[qload1] \n\t"
- "sb %[st2], 0(%[dst]) \n\t" /* even 5 */
- "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
- "dpa.w.ph $ac1, %[p2], %[filter45] \n\t" /* even 7 */
- "extp %[Temp1], $ac1, 31 \n\t" /* even 7 */
- "lbux %[st3], %[Temp3](%[cm]) \n\t" /* even 6 */
-
- /* even 8. pixel */
- "mtlo %[vector_64], $ac3 \n\t" /* odd 1 */
- "mthi $zero, $ac3 \n\t"
- "dpa.w.ph $ac2, %[p3], %[filter45] \n\t" /* even 8 */
- "sb %[st3], 0(%[dst]) \n\t" /* even 6 */
- "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
- "extp %[Temp2], $ac2, 31 \n\t" /* even 8 */
- "lbux %[st1], %[Temp1](%[cm]) \n\t" /* even 7 */
-
- /* ODD pixels */
- "ulw %[qload1], 1(%[src]) \n\t"
- "ulw %[qload2], 5(%[src]) \n\t"
-
- /* odd 1. pixel */
- "mtlo %[vector_64], $ac1 \n\t" /* odd 2 */
- "mthi $zero, $ac1 \n\t"
- "preceu.ph.qbr %[p1], %[qload1] \n\t"
- "preceu.ph.qbl %[p2], %[qload1] \n\t"
- "preceu.ph.qbr %[p3], %[qload2] \n\t"
- "preceu.ph.qbl %[p4], %[qload2] \n\t"
- "sb %[st1], 0(%[dst]) \n\t" /* even 7 */
- "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
- "ulw %[qload2], 9(%[src]) \n\t"
- "dpa.w.ph $ac3, %[p1], %[filter45] \n\t" /* odd 1 */
- "extp %[Temp3], $ac3, 31 \n\t" /* odd 1 */
- "lbux %[st2], %[Temp2](%[cm]) \n\t" /* even 8 */
-
- /* odd 2. pixel */
- "mtlo %[vector_64], $ac2 \n\t" /* odd 3 */
- "mthi $zero, $ac2 \n\t"
- "preceu.ph.qbr %[p1], %[qload2] \n\t"
- "preceu.ph.qbl %[p5], %[qload2] \n\t"
- "sb %[st2], 0(%[dst]) \n\t" /* even 8 */
- "ulw %[qload1], 13(%[src]) \n\t"
- "dpa.w.ph $ac1, %[p2], %[filter45] \n\t" /* odd 2 */
- "extp %[Temp1], $ac1, 31 \n\t" /* odd 2 */
- "lbux %[st3], %[Temp3](%[cm]) \n\t" /* odd 1 */
-
- /* odd 3. pixel */
- "mtlo %[vector_64], $ac3 \n\t" /* odd 4 */
- "mthi $zero, $ac3 \n\t"
- "preceu.ph.qbr %[p2], %[qload1] \n\t"
- "sb %[st3], 0(%[odd_dst]) \n\t" /* odd 1 */
- "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
- "dpa.w.ph $ac2, %[p3], %[filter45] \n\t" /* odd 3 */
- "extp %[Temp2], $ac2, 31 \n\t" /* odd 3 */
- "lbux %[st1], %[Temp1](%[cm]) \n\t" /* odd 2 */
-
- /* odd 4. pixel */
- "mtlo %[vector_64], $ac1 \n\t" /* odd 5 */
- "mthi $zero, $ac1 \n\t"
- "preceu.ph.qbl %[p3], %[qload1] \n\t"
- "sb %[st1], 0(%[odd_dst]) \n\t" /* odd 2 */
- "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
- "dpa.w.ph $ac3, %[p4], %[filter45] \n\t" /* odd 4 */
- "extp %[Temp3], $ac3, 31 \n\t" /* odd 4 */
- "lbux %[st2], %[Temp2](%[cm]) \n\t" /* odd 3 */
-
- /* odd 5. pixel */
- "mtlo %[vector_64], $ac2 \n\t" /* odd 6 */
- "mthi $zero, $ac2 \n\t"
- "sb %[st2], 0(%[odd_dst]) \n\t" /* odd 3 */
- "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
- "dpa.w.ph $ac1, %[p1], %[filter45] \n\t" /* odd 5 */
- "extp %[Temp1], $ac1, 31 \n\t" /* odd 5 */
- "lbux %[st3], %[Temp3](%[cm]) \n\t" /* odd 4 */
-
- /* odd 6. pixel */
- "mtlo %[vector_64], $ac3 \n\t" /* odd 7 */
- "mthi $zero, $ac3 \n\t"
- "sb %[st3], 0(%[odd_dst]) \n\t" /* odd 4 */
- "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
- "ulw %[qload1], 21(%[src]) \n\t"
- "dpa.w.ph $ac2, %[p5], %[filter45] \n\t" /* odd 6 */
- "extp %[Temp2], $ac2, 31 \n\t" /* odd 6 */
- "lbux %[st1], %[Temp1](%[cm]) \n\t" /* odd 5 */
-
- /* odd 7. pixel */
- "mtlo %[vector_64], $ac1 \n\t" /* odd 8 */
- "mthi $zero, $ac1 \n\t"
- "preceu.ph.qbr %[p5], %[qload1] \n\t"
- "sb %[st1], 0(%[odd_dst]) \n\t" /* odd 5 */
- "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
- "dpa.w.ph $ac3, %[p2], %[filter45] \n\t" /* odd 7 */
- "extp %[Temp3], $ac3, 31 \n\t" /* odd 7 */
-
- /* odd 8. pixel */
- "dpa.w.ph $ac1, %[p3], %[filter45] \n\t" /* odd 8 */
- "extp %[Temp1], $ac1, 31 \n\t" /* odd 8 */
-
- "lbux %[st2], %[Temp2](%[cm]) \n\t" /* odd 6 */
- "lbux %[st3], %[Temp3](%[cm]) \n\t" /* odd 7 */
- "lbux %[st1], %[Temp1](%[cm]) \n\t" /* odd 8 */
-
- "sb %[st2], 0(%[odd_dst]) \n\t" /* odd 6 */
- "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
-
- "sb %[st3], 0(%[odd_dst]) \n\t" /* odd 7 */
- "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
-
- "sb %[st1], 0(%[odd_dst]) \n\t" /* odd 8 */
-
- : [qload1] "=&r" (qload1), [qload2] "=&r" (qload2), [p5] "=&r" (p5),
- [st1] "=&r" (st1), [st2] "=&r" (st2), [st3] "=&r" (st3),
- [p1] "=&r" (p1), [p2] "=&r" (p2), [p3] "=&r" (p3), [p4] "=&r" (p4),
- [Temp1] "=&r" (Temp1), [Temp2] "=&r" (Temp2), [Temp3] "=&r" (Temp3),
- [dst] "+r" (dst), [odd_dst] "+r" (odd_dst)
- : [filter45] "r" (filter45), [vector_64] "r" (vector_64),
- [cm] "r" (cm),
- [src] "r" (src), [dst_pitch_2] "r" (dst_pitch_2)
- );
-
- src += 16;
- dst = (dst_ptr + ((c + 1) * 16 * dst_stride));
- odd_dst = (dst + dst_stride);
- }
-
- /* Next row... */
- src_ptr += src_stride;
- dst_ptr += 1;
- }
-}
-
-static void convolve_bi_horiz_64_transposed_dspr2(const uint8_t *src_ptr,
- int32_t src_stride,
- uint8_t *dst_ptr,
- int32_t dst_stride,
- const int16_t *filter_x0,
- int32_t h) {
- int32_t c, y;
- const uint8_t *src;
- uint8_t *dst;
- uint8_t *cm = vp9_ff_cropTbl;
- uint32_t vector_64 = 64;
- int32_t Temp1, Temp2, Temp3;
- uint32_t qload1, qload2;
- uint32_t p1, p2, p3, p4, p5;
- uint32_t st1, st2, st3;
- uint32_t dst_pitch_2 = (dst_stride << 1);
- uint8_t *odd_dst;
- const int16_t *filter = &filter_x0[3];
- uint32_t filter45;
-
- filter45 = ((const int32_t *)filter)[0];
-
- for (y = h; y--;) {
- /* prefetch data to cache memory */
- vp9_prefetch_load(src_ptr + src_stride);
- vp9_prefetch_load(src_ptr + src_stride + 32);
- vp9_prefetch_load(src_ptr + src_stride + 64);
-
- src = src_ptr;
- dst = dst_ptr;
-
- odd_dst = (dst + dst_stride);
-
- for (c = 0; c < 4; c++) {
- __asm__ __volatile__ (
- "ulw %[qload1], 0(%[src]) \n\t"
- "ulw %[qload2], 4(%[src]) \n\t"
-
- /* even 1. pixel */
- "mtlo %[vector_64], $ac1 \n\t" /* even 1 */
- "mthi $zero, $ac1 \n\t"
- "mtlo %[vector_64], $ac2 \n\t" /* even 2 */
- "mthi $zero, $ac2 \n\t"
- "preceu.ph.qbr %[p1], %[qload1] \n\t"
- "preceu.ph.qbl %[p2], %[qload1] \n\t"
- "preceu.ph.qbr %[p3], %[qload2] \n\t"
- "preceu.ph.qbl %[p4], %[qload2] \n\t"
- "ulw %[qload1], 8(%[src]) \n\t"
- "dpa.w.ph $ac1, %[p1], %[filter45] \n\t" /* even 1 */
- "extp %[Temp1], $ac1, 31 \n\t" /* even 1 */
-
- /* even 2. pixel */
- "mtlo %[vector_64], $ac3 \n\t" /* even 3 */
- "mthi $zero, $ac3 \n\t"
- "preceu.ph.qbr %[p1], %[qload1] \n\t"
- "preceu.ph.qbl %[p5], %[qload1] \n\t"
- "ulw %[qload2], 12(%[src]) \n\t"
- "dpa.w.ph $ac2, %[p2], %[filter45] \n\t" /* even 1 */
- "lbux %[st1], %[Temp1](%[cm]) \n\t" /* even 1 */
- "extp %[Temp2], $ac2, 31 \n\t" /* even 1 */
-
- /* even 3. pixel */
- "mtlo %[vector_64], $ac1 \n\t" /* even 4 */
- "mthi $zero, $ac1 \n\t"
- "preceu.ph.qbr %[p2], %[qload2] \n\t"
- "sb %[st1], 0(%[dst]) \n\t" /* even 1 */
- "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
- "dpa.w.ph $ac3, %[p3], %[filter45] \n\t" /* even 3 */
- "extp %[Temp3], $ac3, 31 \n\t" /* even 3 */
- "lbux %[st2], %[Temp2](%[cm]) \n\t" /* even 1 */
-
- /* even 4. pixel */
- "mtlo %[vector_64], $ac2 \n\t" /* even 5 */
- "mthi $zero, $ac2 \n\t"
- "preceu.ph.qbl %[p3], %[qload2] \n\t"
- "sb %[st2], 0(%[dst]) \n\t" /* even 2 */
- "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
- "dpa.w.ph $ac1, %[p4], %[filter45] \n\t" /* even 4 */
- "extp %[Temp1], $ac1, 31 \n\t" /* even 4 */
- "lbux %[st3], %[Temp3](%[cm]) \n\t" /* even 3 */
-
- /* even 5. pixel */
- "mtlo %[vector_64], $ac3 \n\t" /* even 6 */
- "mthi $zero, $ac3 \n\t"
- "sb %[st3], 0(%[dst]) \n\t" /* even 3 */
- "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
- "dpa.w.ph $ac2, %[p1], %[filter45] \n\t" /* even 5 */
- "extp %[Temp2], $ac2, 31 \n\t" /* even 5 */
- "lbux %[st1], %[Temp1](%[cm]) \n\t" /* even 4 */
-
- /* even 6. pixel */
- "mtlo %[vector_64], $ac1 \n\t" /* even 7 */
- "mthi $zero, $ac1 \n\t"
- "sb %[st1], 0(%[dst]) \n\t" /* even 4 */
- "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
- "ulw %[qload1], 20(%[src]) \n\t"
- "dpa.w.ph $ac3, %[p5], %[filter45] \n\t" /* even 6 */
- "extp %[Temp3], $ac3, 31 \n\t" /* even 6 */
- "lbux %[st2], %[Temp2](%[cm]) \n\t" /* even 5 */
-
- /* even 7. pixel */
- "mtlo %[vector_64], $ac2 \n\t" /* even 8 */
- "mthi $zero, $ac2 \n\t"
- "preceu.ph.qbr %[p5], %[qload1] \n\t"
- "sb %[st2], 0(%[dst]) \n\t" /* even 5 */
- "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
- "dpa.w.ph $ac1, %[p2], %[filter45] \n\t" /* even 7 */
- "extp %[Temp1], $ac1, 31 \n\t" /* even 7 */
- "lbux %[st3], %[Temp3](%[cm]) \n\t" /* even 6 */
-
- /* even 8. pixel */
- "mtlo %[vector_64], $ac3 \n\t" /* odd 1 */
- "mthi $zero, $ac3 \n\t"
- "dpa.w.ph $ac2, %[p3], %[filter45] \n\t" /* even 8 */
- "sb %[st3], 0(%[dst]) \n\t" /* even 6 */
- "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
- "extp %[Temp2], $ac2, 31 \n\t" /* even 8 */
- "lbux %[st1], %[Temp1](%[cm]) \n\t" /* even 7 */
-
- /* ODD pixels */
- "ulw %[qload1], 1(%[src]) \n\t"
- "ulw %[qload2], 5(%[src]) \n\t"
-
- /* odd 1. pixel */
- "mtlo %[vector_64], $ac1 \n\t" /* odd 2 */
- "mthi $zero, $ac1 \n\t"
- "preceu.ph.qbr %[p1], %[qload1] \n\t"
- "preceu.ph.qbl %[p2], %[qload1] \n\t"
- "preceu.ph.qbr %[p3], %[qload2] \n\t"
- "preceu.ph.qbl %[p4], %[qload2] \n\t"
- "sb %[st1], 0(%[dst]) \n\t" /* even 7 */
- "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
- "ulw %[qload2], 9(%[src]) \n\t"
- "dpa.w.ph $ac3, %[p1], %[filter45] \n\t" /* odd 1 */
- "extp %[Temp3], $ac3, 31 \n\t" /* odd 1 */
- "lbux %[st2], %[Temp2](%[cm]) \n\t" /* even 8 */
-
- /* odd 2. pixel */
- "mtlo %[vector_64], $ac2 \n\t" /* odd 3 */
- "mthi $zero, $ac2 \n\t"
- "preceu.ph.qbr %[p1], %[qload2] \n\t"
- "preceu.ph.qbl %[p5], %[qload2] \n\t"
- "sb %[st2], 0(%[dst]) \n\t" /* even 8 */
- "ulw %[qload1], 13(%[src]) \n\t"
- "dpa.w.ph $ac1, %[p2], %[filter45] \n\t" /* odd 2 */
- "extp %[Temp1], $ac1, 31 \n\t" /* odd 2 */
- "lbux %[st3], %[Temp3](%[cm]) \n\t" /* odd 1 */
-
- /* odd 3. pixel */
- "mtlo %[vector_64], $ac3 \n\t" /* odd 4 */
- "mthi $zero, $ac3 \n\t"
- "preceu.ph.qbr %[p2], %[qload1] \n\t"
- "sb %[st3], 0(%[odd_dst]) \n\t" /* odd 1 */
- "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
- "dpa.w.ph $ac2, %[p3], %[filter45] \n\t" /* odd 3 */
- "extp %[Temp2], $ac2, 31 \n\t" /* odd 3 */
- "lbux %[st1], %[Temp1](%[cm]) \n\t" /* odd 2 */
-
- /* odd 4. pixel */
- "mtlo %[vector_64], $ac1 \n\t" /* odd 5 */
- "mthi $zero, $ac1 \n\t"
- "preceu.ph.qbl %[p3], %[qload1] \n\t"
- "sb %[st1], 0(%[odd_dst]) \n\t" /* odd 2 */
- "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
- "dpa.w.ph $ac3, %[p4], %[filter45] \n\t" /* odd 4 */
- "extp %[Temp3], $ac3, 31 \n\t" /* odd 4 */
- "lbux %[st2], %[Temp2](%[cm]) \n\t" /* odd 3 */
-
- /* odd 5. pixel */
- "mtlo %[vector_64], $ac2 \n\t" /* odd 6 */
- "mthi $zero, $ac2 \n\t"
- "sb %[st2], 0(%[odd_dst]) \n\t" /* odd 3 */
- "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
- "dpa.w.ph $ac1, %[p1], %[filter45] \n\t" /* odd 5 */
- "extp %[Temp1], $ac1, 31 \n\t" /* odd 5 */
- "lbux %[st3], %[Temp3](%[cm]) \n\t" /* odd 4 */
-
- /* odd 6. pixel */
- "mtlo %[vector_64], $ac3 \n\t" /* odd 7 */
- "mthi $zero, $ac3 \n\t"
- "sb %[st3], 0(%[odd_dst]) \n\t" /* odd 4 */
- "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
- "ulw %[qload1], 21(%[src]) \n\t"
- "dpa.w.ph $ac2, %[p5], %[filter45] \n\t" /* odd 6 */
- "extp %[Temp2], $ac2, 31 \n\t" /* odd 6 */
- "lbux %[st1], %[Temp1](%[cm]) \n\t" /* odd 5 */
-
- /* odd 7. pixel */
- "mtlo %[vector_64], $ac1 \n\t" /* odd 8 */
- "mthi $zero, $ac1 \n\t"
- "preceu.ph.qbr %[p5], %[qload1] \n\t"
- "sb %[st1], 0(%[odd_dst]) \n\t" /* odd 5 */
- "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
- "dpa.w.ph $ac3, %[p2], %[filter45] \n\t" /* odd 7 */
- "extp %[Temp3], $ac3, 31 \n\t" /* odd 7 */
-
- /* odd 8. pixel */
- "dpa.w.ph $ac1, %[p3], %[filter45] \n\t" /* odd 8 */
- "extp %[Temp1], $ac1, 31 \n\t" /* odd 8 */
-
- "lbux %[st2], %[Temp2](%[cm]) \n\t" /* odd 6 */
- "lbux %[st3], %[Temp3](%[cm]) \n\t" /* odd 7 */
- "lbux %[st1], %[Temp1](%[cm]) \n\t" /* odd 8 */
-
- "sb %[st2], 0(%[odd_dst]) \n\t" /* odd 6 */
- "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
-
- "sb %[st3], 0(%[odd_dst]) \n\t" /* odd 7 */
- "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
-
- "sb %[st1], 0(%[odd_dst]) \n\t" /* odd 8 */
-
- : [qload1] "=&r" (qload1), [qload2] "=&r" (qload2), [p5] "=&r" (p5),
- [st1] "=&r" (st1), [st2] "=&r" (st2), [st3] "=&r" (st3),
- [p1] "=&r" (p1), [p2] "=&r" (p2), [p3] "=&r" (p3), [p4] "=&r" (p4),
- [Temp1] "=&r" (Temp1), [Temp2] "=&r" (Temp2), [Temp3] "=&r" (Temp3),
- [dst] "+r" (dst), [odd_dst] "+r" (odd_dst)
- : [filter45] "r" (filter45), [vector_64] "r" (vector_64),
- [cm] "r" (cm),
- [src] "r" (src), [dst_pitch_2] "r" (dst_pitch_2)
- );
-
- src += 16;
- dst = (dst_ptr + ((c + 1) * 16 * dst_stride));
- odd_dst = (dst + dst_stride);
- }
-
- /* Next row... */
- src_ptr += src_stride;
- dst_ptr += 1;
- }
-}
+static void convolve_bi_horiz_64_transposed_dspr2(const uint8_t *src_ptr,
+ int32_t src_stride,
+ uint8_t *dst_ptr,
+ int32_t dst_stride,
+ const int16_t *filter_x0,
+ int32_t h) {
+ int32_t c, y;
+ const uint8_t *src;
+ uint8_t *dst;
+ uint8_t *cm = vp9_ff_cropTbl;
+ uint32_t vector_64 = 64;
+ int32_t Temp1, Temp2, Temp3;
+ uint32_t qload1, qload2;
+ uint32_t p1, p2, p3, p4, p5;
+ uint32_t st1, st2, st3;
+ uint32_t dst_pitch_2 = (dst_stride << 1);
+ uint8_t *odd_dst;
+ const int16_t *filter = &filter_x0[3];
+ uint32_t filter45;
+
+ filter45 = ((const int32_t *)filter)[0];
+
+ for (y = h; y--;) {
+ /* prefetch data to cache memory */
+ vp9_prefetch_load(src_ptr + src_stride);
+ vp9_prefetch_load(src_ptr + src_stride + 32);
+ vp9_prefetch_load(src_ptr + src_stride + 64);
+
+ src = src_ptr;
+ dst = dst_ptr;
+
+ odd_dst = (dst + dst_stride);
+
+ for (c = 0; c < 4; c++) {
+ __asm__ __volatile__ (
+ "ulw %[qload1], 0(%[src]) \n\t"
+ "ulw %[qload2], 4(%[src]) \n\t"
+
+ /* even 1. pixel */
+ "mtlo %[vector_64], $ac1 \n\t" /* even 1 */
+ "mthi $zero, $ac1 \n\t"
+ "mtlo %[vector_64], $ac2 \n\t" /* even 2 */
+ "mthi $zero, $ac2 \n\t"
+ "preceu.ph.qbr %[p1], %[qload1] \n\t"
+ "preceu.ph.qbl %[p2], %[qload1] \n\t"
+ "preceu.ph.qbr %[p3], %[qload2] \n\t"
+ "preceu.ph.qbl %[p4], %[qload2] \n\t"
+ "ulw %[qload1], 8(%[src]) \n\t"
+ "dpa.w.ph $ac1, %[p1], %[filter45] \n\t" /* even 1 */
+ "extp %[Temp1], $ac1, 31 \n\t" /* even 1 */
+
+ /* even 2. pixel */
+ "mtlo %[vector_64], $ac3 \n\t" /* even 3 */
+ "mthi $zero, $ac3 \n\t"
+ "preceu.ph.qbr %[p1], %[qload1] \n\t"
+ "preceu.ph.qbl %[p5], %[qload1] \n\t"
+ "ulw %[qload2], 12(%[src]) \n\t"
+ "dpa.w.ph $ac2, %[p2], %[filter45] \n\t" /* even 1 */
+ "lbux %[st1], %[Temp1](%[cm]) \n\t" /* even 1 */
+ "extp %[Temp2], $ac2, 31 \n\t" /* even 1 */
+
+ /* even 3. pixel */
+ "mtlo %[vector_64], $ac1 \n\t" /* even 4 */
+ "mthi $zero, $ac1 \n\t"
+ "preceu.ph.qbr %[p2], %[qload2] \n\t"
+ "sb %[st1], 0(%[dst]) \n\t" /* even 1 */
+ "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
+ "dpa.w.ph $ac3, %[p3], %[filter45] \n\t" /* even 3 */
+ "extp %[Temp3], $ac3, 31 \n\t" /* even 3 */
+ "lbux %[st2], %[Temp2](%[cm]) \n\t" /* even 1 */
+
+ /* even 4. pixel */
+ "mtlo %[vector_64], $ac2 \n\t" /* even 5 */
+ "mthi $zero, $ac2 \n\t"
+ "preceu.ph.qbl %[p3], %[qload2] \n\t"
+ "sb %[st2], 0(%[dst]) \n\t" /* even 2 */
+ "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
+ "dpa.w.ph $ac1, %[p4], %[filter45] \n\t" /* even 4 */
+ "extp %[Temp1], $ac1, 31 \n\t" /* even 4 */
+ "lbux %[st3], %[Temp3](%[cm]) \n\t" /* even 3 */
+
+ /* even 5. pixel */
+ "mtlo %[vector_64], $ac3 \n\t" /* even 6 */
+ "mthi $zero, $ac3 \n\t"
+ "sb %[st3], 0(%[dst]) \n\t" /* even 3 */
+ "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
+ "dpa.w.ph $ac2, %[p1], %[filter45] \n\t" /* even 5 */
+ "extp %[Temp2], $ac2, 31 \n\t" /* even 5 */
+ "lbux %[st1], %[Temp1](%[cm]) \n\t" /* even 4 */
+
+ /* even 6. pixel */
+ "mtlo %[vector_64], $ac1 \n\t" /* even 7 */
+ "mthi $zero, $ac1 \n\t"
+ "sb %[st1], 0(%[dst]) \n\t" /* even 4 */
+ "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
+ "ulw %[qload1], 20(%[src]) \n\t"
+ "dpa.w.ph $ac3, %[p5], %[filter45] \n\t" /* even 6 */
+ "extp %[Temp3], $ac3, 31 \n\t" /* even 6 */
+ "lbux %[st2], %[Temp2](%[cm]) \n\t" /* even 5 */
+
+ /* even 7. pixel */
+ "mtlo %[vector_64], $ac2 \n\t" /* even 8 */
+ "mthi $zero, $ac2 \n\t"
+ "preceu.ph.qbr %[p5], %[qload1] \n\t"
+ "sb %[st2], 0(%[dst]) \n\t" /* even 5 */
+ "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
+ "dpa.w.ph $ac1, %[p2], %[filter45] \n\t" /* even 7 */
+ "extp %[Temp1], $ac1, 31 \n\t" /* even 7 */
+ "lbux %[st3], %[Temp3](%[cm]) \n\t" /* even 6 */
+
+ /* even 8. pixel */
+ "mtlo %[vector_64], $ac3 \n\t" /* odd 1 */
+ "mthi $zero, $ac3 \n\t"
+ "dpa.w.ph $ac2, %[p3], %[filter45] \n\t" /* even 8 */
+ "sb %[st3], 0(%[dst]) \n\t" /* even 6 */
+ "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
+ "extp %[Temp2], $ac2, 31 \n\t" /* even 8 */
+ "lbux %[st1], %[Temp1](%[cm]) \n\t" /* even 7 */
+
+ /* ODD pixels */
+ "ulw %[qload1], 1(%[src]) \n\t"
+ "ulw %[qload2], 5(%[src]) \n\t"
+
+ /* odd 1. pixel */
+ "mtlo %[vector_64], $ac1 \n\t" /* odd 2 */
+ "mthi $zero, $ac1 \n\t"
+ "preceu.ph.qbr %[p1], %[qload1] \n\t"
+ "preceu.ph.qbl %[p2], %[qload1] \n\t"
+ "preceu.ph.qbr %[p3], %[qload2] \n\t"
+ "preceu.ph.qbl %[p4], %[qload2] \n\t"
+ "sb %[st1], 0(%[dst]) \n\t" /* even 7 */
+ "addu %[dst], %[dst], %[dst_pitch_2] \n\t"
+ "ulw %[qload2], 9(%[src]) \n\t"
+ "dpa.w.ph $ac3, %[p1], %[filter45] \n\t" /* odd 1 */
+ "extp %[Temp3], $ac3, 31 \n\t" /* odd 1 */
+ "lbux %[st2], %[Temp2](%[cm]) \n\t" /* even 8 */
+
+ /* odd 2. pixel */
+ "mtlo %[vector_64], $ac2 \n\t" /* odd 3 */
+ "mthi $zero, $ac2 \n\t"
+ "preceu.ph.qbr %[p1], %[qload2] \n\t"
+ "preceu.ph.qbl %[p5], %[qload2] \n\t"
+ "sb %[st2], 0(%[dst]) \n\t" /* even 8 */
+ "ulw %[qload1], 13(%[src]) \n\t"
+ "dpa.w.ph $ac1, %[p2], %[filter45] \n\t" /* odd 2 */
+ "extp %[Temp1], $ac1, 31 \n\t" /* odd 2 */
+ "lbux %[st3], %[Temp3](%[cm]) \n\t" /* odd 1 */
+
+ /* odd 3. pixel */
+ "mtlo %[vector_64], $ac3 \n\t" /* odd 4 */
+ "mthi $zero, $ac3 \n\t"
+ "preceu.ph.qbr %[p2], %[qload1] \n\t"
+ "sb %[st3], 0(%[odd_dst]) \n\t" /* odd 1 */
+ "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
+ "dpa.w.ph $ac2, %[p3], %[filter45] \n\t" /* odd 3 */
+ "extp %[Temp2], $ac2, 31 \n\t" /* odd 3 */
+ "lbux %[st1], %[Temp1](%[cm]) \n\t" /* odd 2 */
+
+ /* odd 4. pixel */
+ "mtlo %[vector_64], $ac1 \n\t" /* odd 5 */
+ "mthi $zero, $ac1 \n\t"
+ "preceu.ph.qbl %[p3], %[qload1] \n\t"
+ "sb %[st1], 0(%[odd_dst]) \n\t" /* odd 2 */
+ "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
+ "dpa.w.ph $ac3, %[p4], %[filter45] \n\t" /* odd 4 */
+ "extp %[Temp3], $ac3, 31 \n\t" /* odd 4 */
+ "lbux %[st2], %[Temp2](%[cm]) \n\t" /* odd 3 */
+
+ /* odd 5. pixel */
+ "mtlo %[vector_64], $ac2 \n\t" /* odd 6 */
+ "mthi $zero, $ac2 \n\t"
+ "sb %[st2], 0(%[odd_dst]) \n\t" /* odd 3 */
+ "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
+ "dpa.w.ph $ac1, %[p1], %[filter45] \n\t" /* odd 5 */
+ "extp %[Temp1], $ac1, 31 \n\t" /* odd 5 */
+ "lbux %[st3], %[Temp3](%[cm]) \n\t" /* odd 4 */
+
+ /* odd 6. pixel */
+ "mtlo %[vector_64], $ac3 \n\t" /* odd 7 */
+ "mthi $zero, $ac3 \n\t"
+ "sb %[st3], 0(%[odd_dst]) \n\t" /* odd 4 */
+ "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
+ "ulw %[qload1], 21(%[src]) \n\t"
+ "dpa.w.ph $ac2, %[p5], %[filter45] \n\t" /* odd 6 */
+ "extp %[Temp2], $ac2, 31 \n\t" /* odd 6 */
+ "lbux %[st1], %[Temp1](%[cm]) \n\t" /* odd 5 */
+
+ /* odd 7. pixel */
+ "mtlo %[vector_64], $ac1 \n\t" /* odd 8 */
+ "mthi $zero, $ac1 \n\t"
+ "preceu.ph.qbr %[p5], %[qload1] \n\t"
+ "sb %[st1], 0(%[odd_dst]) \n\t" /* odd 5 */
+ "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
+ "dpa.w.ph $ac3, %[p2], %[filter45] \n\t" /* odd 7 */
+ "extp %[Temp3], $ac3, 31 \n\t" /* odd 7 */
+
+ /* odd 8. pixel */
+ "dpa.w.ph $ac1, %[p3], %[filter45] \n\t" /* odd 8 */
+ "extp %[Temp1], $ac1, 31 \n\t" /* odd 8 */
+
+ "lbux %[st2], %[Temp2](%[cm]) \n\t" /* odd 6 */
+ "lbux %[st3], %[Temp3](%[cm]) \n\t" /* odd 7 */
+ "lbux %[st1], %[Temp1](%[cm]) \n\t" /* odd 8 */
+
+ "sb %[st2], 0(%[odd_dst]) \n\t" /* odd 6 */
+ "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
+
+ "sb %[st3], 0(%[odd_dst]) \n\t" /* odd 7 */
+ "addu %[odd_dst], %[odd_dst], %[dst_pitch_2] \n\t"
+
+ "sb %[st1], 0(%[odd_dst]) \n\t" /* odd 8 */
+
+ : [qload1] "=&r" (qload1), [qload2] "=&r" (qload2), [p5] "=&r" (p5),
+ [st1] "=&r" (st1), [st2] "=&r" (st2), [st3] "=&r" (st3),
+ [p1] "=&r" (p1), [p2] "=&r" (p2), [p3] "=&r" (p3), [p4] "=&r" (p4),
+ [Temp1] "=&r" (Temp1), [Temp2] "=&r" (Temp2), [Temp3] "=&r" (Temp3),
+ [dst] "+r" (dst), [odd_dst] "+r" (odd_dst)
+ : [filter45] "r" (filter45), [vector_64] "r" (vector_64),
+ [cm] "r" (cm),
+ [src] "r" (src), [dst_pitch_2] "r" (dst_pitch_2)
+ );
+
+ src += 16;
+ dst = (dst_ptr + ((c + 1) * 16 * dst_stride));
+ odd_dst = (dst + dst_stride);
+ }
+
+ /* Next row... */
+ src_ptr += src_stride;
+ dst_ptr += 1;
+ }
+}
void convolve_bi_horiz_transposed(const uint8_t *src, ptrdiff_t src_stride,
uint8_t *dst, ptrdiff_t dst_stride,
diff --git a/vp9/common/mips/dspr2/vp9_convolve2_vert_dspr2.c b/vp9/common/mips/dspr2/vp9_convolve2_vert_dspr2.c
index 8eb105c0c9104f4027f785fe1155dc9c0bc51898..bf01f11541d0bb22559da6ee9485d4fd2f165526 100644
--- a/vp9/common/mips/dspr2/vp9_convolve2_vert_dspr2.c
+++ b/vp9/common/mips/dspr2/vp9_convolve2_vert_dspr2.c
@@ -20,103 +20,103 @@
#include "vp9/common/mips/dspr2/vp9_common_dspr2.h"
#if HAVE_DSPR2
-static void convolve_bi_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;
- uint8_t *cm = vp9_ff_cropTbl;
- uint32_t vector4a = 64;
- uint32_t load1, load2;
- uint32_t p1, p2;
- uint32_t scratch1;
- uint32_t store1, store2;
- int32_t Temp1, Temp2;
- const int16_t *filter = &filter_y[3];
- uint32_t filter45;
-
- filter45 = ((const int32_t *)filter)[0];
-
- for (y = h; y--;) {
- /* prefetch data to cache memory */
- vp9_prefetch_store(dst + dst_stride);
-
- 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"
-
- "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 %[p2], %[p1], %[scratch1] \n\t" /* pixel 2 */
- "append %[p1], %[scratch1], 16 \n\t" /* pixel 1 */
-
- "dpa.w.ph $ac0, %[p1], %[filter45] \n\t"
- "dpa.w.ph $ac1, %[p2], %[filter45] \n\t"
-
- "preceu.ph.qbl %[scratch1], %[load1] \n\t"
- "preceu.ph.qbl %[p1], %[load2] \n\t"
-
- "precrq.ph.w %[p2], %[p1], %[scratch1] \n\t" /* pixel 2 */
- "append %[p1], %[scratch1], 16 \n\t" /* pixel 1 */
-
- "dpa.w.ph $ac2, %[p1], %[filter45] \n\t"
- "dpa.w.ph $ac3, %[p2], %[filter45] \n\t"
-
- "extp %[Temp1], $ac0, 31 \n\t"
- "extp %[Temp2], $ac1, 31 \n\t"
-
- "lbux %[store1], %[Temp1](%[cm]) \n\t"
- "extp %[Temp1], $ac2, 31 \n\t"
-
- "lbux %[store2], %[Temp2](%[cm]) \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),
- [p1] "=&r" (p1), [p2] "=&r" (p2),
- [scratch1] "=&r" (scratch1),
- [Temp1] "=&r" (Temp1), [Temp2] "=&r" (Temp2),
- [store1] "=&r" (store1), [store2] "=&r" (store2),
- [src_ptr] "+r" (src_ptr)
- : [filter45] "r" (filter45),[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_bi_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;
+ uint8_t *cm = vp9_ff_cropTbl;
+ uint32_t vector4a = 64;
+ uint32_t load1, load2;
+ uint32_t p1, p2;
+ uint32_t scratch1;
+ uint32_t store1, store2;
+ int32_t Temp1, Temp2;
+ const int16_t *filter = &filter_y[3];
+ uint32_t filter45;
+
+ filter45 = ((const int32_t *)filter)[0];
+
+ for (y = h; y--;) {
+ /* prefetch data to cache memory */
+ vp9_prefetch_store(dst + dst_stride);
+
+ 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"
+
+ "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 %[p2], %[p1], %[scratch1] \n\t" /* pixel 2 */
+ "append %[p1], %[scratch1], 16 \n\t" /* pixel 1 */
+
+ "dpa.w.ph $ac0, %[p1], %[filter45] \n\t"
+ "dpa.w.ph $ac1, %[p2], %[filter45] \n\t"
+
+ "preceu.ph.qbl %[scratch1], %[load1] \n\t"
+ "preceu.ph.qbl %[p1], %[load2] \n\t"
+
+ "precrq.ph.w %[p2], %[p1], %[scratch1] \n\t" /* pixel 2 */
+ "append %[p1], %[scratch1], 16 \n\t" /* pixel 1 */
+
+ "dpa.w.ph $ac2, %[p1], %[filter45] \n\t"
+ "dpa.w.ph $ac3, %[p2], %[filter45] \n\t"
+
+ "extp %[Temp1], $ac0, 31 \n\t"
+ "extp %[Temp2], $ac1, 31 \n\t"
+
+ "lbux %[store1], %[Temp1](%[cm]) \n\t"
+ "extp %[Temp1], $ac2, 31 \n\t"
+
+ "lbux %[store2], %[Temp2](%[cm]) \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),
+ [p1] "=&r" (p1), [p2] "=&r" (p2),
+ [scratch1] "=&r" (scratch1),
+ [Temp1] "=&r" (Temp1), [Temp2] "=&r" (Temp2),
+ [store1] "=&r" (store1), [store2] "=&r" (store2),
+ [src_ptr] "+r" (src_ptr)
+ : [filter45] "r" (filter45),[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_bi_vert_64_dspr2(const uint8_t *src,
int32_t src_stride,
@@ -124,95 +124,95 @@ static void convolve_bi_vert_64_dspr2(const uint8_t *src,
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;
- uint8_t *cm = vp9_ff_cropTbl;
- uint32_t vector4a = 64;
- uint32_t load1, load2;
- uint32_t p1, p2;
- uint32_t scratch1;
- uint32_t store1, store2;
- int32_t Temp1, Temp2;
- const int16_t *filter = &filter_y[3];
- uint32_t filter45;
-
- filter45 = ((const int32_t *)filter)[0];
-
- for (y = h; y--;) {
- /* prefetch data to cache memory */
- vp9_prefetch_store(dst + dst_stride);
-
- 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"
-
- "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 %[p2], %[p1], %[scratch1] \n\t" /* pixel 2 */
- "append %[p1], %[scratch1], 16 \n\t" /* pixel 1 */
-
- "dpa.w.ph $ac0, %[p1], %[filter45] \n\t"
- "dpa.w.ph $ac1, %[p2], %[filter45] \n\t"
-
- "preceu.ph.qbl %[scratch1], %[load1] \n\t"
- "preceu.ph.qbl %[p1], %[load2] \n\t"
-
- "precrq.ph.w %[p2], %[p1], %[scratch1] \n\t" /* pixel 2 */
- "append %[p1], %[scratch1], 16 \n\t" /* pixel 1 */
-
- "dpa.w.ph $ac2, %[p1], %[filter45] \n\t"
- "dpa.w.ph $ac3, %[p2], %[filter45] \n\t"
-
- "extp %[Temp1], $ac0, 31 \n\t"
- "extp %[Temp2], $ac1, 31 \n\t"
-
- "lbux %[store1], %[Temp1](%[cm]) \n\t"
- "extp %[Temp1], $ac2, 31 \n\t"
-
- "lbux %[store2], %[Temp2](%[cm]) \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),
- [p1] "=&r" (p1), [p2] "=&r" (p2),
- [scratch1] "=&r" (scratch1),
- [Temp1] "=&r" (Temp1), [Temp2] "=&r" (Temp2),
- [store1] "=&r" (store1), [store2] "=&r" (store2),
- [src_ptr] "+r" (src_ptr)
- : [filter45] "r" (filter45),[vector4a] "r" (vector4a),
- [src_stride] "r" (src_stride),
- [cm] "r" (cm), [dst_ptr] "r" (dst_ptr)
- );
- }
-
- /* Next row... */
- src += src_stride;
- dst += dst_stride;
- }
+ int32_t x, y;
+ const uint8_t *src_ptr;
+ uint8_t *dst_ptr;
+ uint8_t *cm = vp9_ff_cropTbl;
+ uint32_t vector4a = 64;
+ uint32_t load1, load2;
+ uint32_t p1, p2;
+ uint32_t scratch1;
+ uint32_t store1, store2;
+ int32_t Temp1, Temp2;
+ const int16_t *filter = &filter_y[3];
+ uint32_t filter45;
+
+ filter45 = ((const int32_t *)filter)[0];
+
+ for (y = h; y--;) {
+ /* prefetch data to cache memory */
+ vp9_prefetch_store(dst + dst_stride);
+
+ 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"
+
+ "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 %[p2], %[p1], %[scratch1] \n\t" /* pixel 2 */
+ "append %[p1], %[scratch1], 16 \n\t" /* pixel 1 */
+
+ "dpa.w.ph $ac0, %[p1], %[filter45] \n\t"
+ "dpa.w.ph $ac1, %[p2], %[filter45] \n\t"
+
+ "preceu.ph.qbl %[scratch1], %[load1] \n\t"
+ "preceu.ph.qbl %[p1], %[load2] \n\t"
+
+ "precrq.ph.w %[p2], %[p1], %[scratch1] \n\t" /* pixel 2 */
+ "append %[p1], %[scratch1], 16 \n\t" /* pixel 1 */
+
+ "dpa.w.ph $ac2, %[p1], %[filter45] \n\t"
+ "dpa.w.ph $ac3, %[p2], %[filter45] \n\t"
+
+ "extp %[Temp1], $ac0, 31 \n\t"
+ "extp %[Temp2], $ac1, 31 \n\t"
+
+ "lbux %[store1], %[Temp1](%[cm]) \n\t"
+ "extp %[Temp1], $ac2, 31 \n\t"
+
+ "lbux %[store2], %[Temp2](%[cm]) \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),
+ [p1] "=&r" (p1), [p2] "=&r" (p2),
+ [scratch1] "=&r" (scratch1),
+ [Temp1] "=&r" (Temp1), [Temp2] "=&r" (Temp2),
+ [store1] "=&r" (store1), [store2] "=&r" (store2),
+ [src_ptr] "+r" (src_ptr)
+ : [filter45] "r" (filter45),[vector4a] "r" (vector4a),
+ [src_stride] "r" (src_stride),
+ [cm] "r" (cm), [dst_ptr] "r" (dst_ptr)
+ );
+ }
+
+ /* Next row... */
+ src += src_stride;
+ dst += dst_stride;
+ }
}
void vp9_convolve2_vert_dspr2(const uint8_t *src, ptrdiff_t src_stride,