ref: 4da2a8b7187618b1d2daf3e56e9d8eacc266fa30
parent: 248cf6f69f73d581595ccdb92f86c8f6b16381b4
parent: 1530a6b77fadc975f9839316222bf2f28e760bf7
author: Johann <johannkoenig@google.com>
date: Wed Nov 13 04:00:09 EST 2013
Merge "mips dsp-ase r2 vp9 decoder intra module optimizations (rebase)"
--- /dev/null
+++ b/vp9/common/mips/dspr2/vp9_intrapred16_dspr2.c
@@ -1,0 +1,332 @@
+/*
+ * 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 <stdlib.h>
+
+#include "./vp9_rtcd.h"
+#include "vp9/common/vp9_common.h"
+#include "vp9/common/mips/dspr2/vp9_common_dspr2.h"
+
+#if HAVE_DSPR2
+void vp9_h_predictor_16x16_dspr2(uint8_t *dst, ptrdiff_t stride,
+ const uint8_t *above, const uint8_t *left) {+ int32_t tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp8;
+ int32_t tmp9, tmp10, tmp11, tmp12, tmp13, tmp14, tmp15, tmp16;
+
+ __asm__ __volatile__ (
+ "lb %[tmp1], (%[left]) \n\t"
+ "lb %[tmp2], 1(%[left]) \n\t"
+ "lb %[tmp3], 2(%[left]) \n\t"
+ "lb %[tmp4], 3(%[left]) \n\t"
+ "lb %[tmp5], 4(%[left]) \n\t"
+ "lb %[tmp6], 5(%[left]) \n\t"
+ "lb %[tmp7], 6(%[left]) \n\t"
+ "lb %[tmp8], 7(%[left]) \n\t"
+ "lb %[tmp9], 8(%[left]) \n\t"
+ "lb %[tmp10], 9(%[left]) \n\t"
+ "lb %[tmp11], 10(%[left]) \n\t"
+ "lb %[tmp12], 11(%[left]) \n\t"
+ "lb %[tmp13], 12(%[left]) \n\t"
+ "lb %[tmp14], 13(%[left]) \n\t"
+ "lb %[tmp15], 14(%[left]) \n\t"
+ "lb %[tmp16], 15(%[left]) \n\t"
+
+ "replv.qb %[tmp1], %[tmp1] \n\t"
+ "replv.qb %[tmp2], %[tmp2] \n\t"
+ "replv.qb %[tmp3], %[tmp3] \n\t"
+ "replv.qb %[tmp4], %[tmp4] \n\t"
+ "replv.qb %[tmp5], %[tmp5] \n\t"
+ "replv.qb %[tmp6], %[tmp6] \n\t"
+ "replv.qb %[tmp7], %[tmp7] \n\t"
+ "replv.qb %[tmp8], %[tmp8] \n\t"
+ "replv.qb %[tmp9], %[tmp9] \n\t"
+ "replv.qb %[tmp10], %[tmp10] \n\t"
+ "replv.qb %[tmp11], %[tmp11] \n\t"
+ "replv.qb %[tmp12], %[tmp12] \n\t"
+ "replv.qb %[tmp13], %[tmp13] \n\t"
+ "replv.qb %[tmp14], %[tmp14] \n\t"
+ "replv.qb %[tmp15], %[tmp15] \n\t"
+ "replv.qb %[tmp16], %[tmp16] \n\t"
+
+ "sw %[tmp1], (%[dst]) \n\t"
+ "sw %[tmp1], 4(%[dst]) \n\t"
+ "sw %[tmp1], 8(%[dst]) \n\t"
+ "sw %[tmp1], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp2], (%[dst]) \n\t"
+ "sw %[tmp2], 4(%[dst]) \n\t"
+ "sw %[tmp2], 8(%[dst]) \n\t"
+ "sw %[tmp2], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp3], (%[dst]) \n\t"
+ "sw %[tmp3], 4(%[dst]) \n\t"
+ "sw %[tmp3], 8(%[dst]) \n\t"
+ "sw %[tmp3], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp4], (%[dst]) \n\t"
+ "sw %[tmp4], 4(%[dst]) \n\t"
+ "sw %[tmp4], 8(%[dst]) \n\t"
+ "sw %[tmp4], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp5], (%[dst]) \n\t"
+ "sw %[tmp5], 4(%[dst]) \n\t"
+ "sw %[tmp5], 8(%[dst]) \n\t"
+ "sw %[tmp5], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp6], (%[dst]) \n\t"
+ "sw %[tmp6], 4(%[dst]) \n\t"
+ "sw %[tmp6], 8(%[dst]) \n\t"
+ "sw %[tmp6], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp7], (%[dst]) \n\t"
+ "sw %[tmp7], 4(%[dst]) \n\t"
+ "sw %[tmp7], 8(%[dst]) \n\t"
+ "sw %[tmp7], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp8], (%[dst]) \n\t"
+ "sw %[tmp8], 4(%[dst]) \n\t"
+ "sw %[tmp8], 8(%[dst]) \n\t"
+ "sw %[tmp8], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp9], (%[dst]) \n\t"
+ "sw %[tmp9], 4(%[dst]) \n\t"
+ "sw %[tmp9], 8(%[dst]) \n\t"
+ "sw %[tmp9], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp10], (%[dst]) \n\t"
+ "sw %[tmp10], 4(%[dst]) \n\t"
+ "sw %[tmp10], 8(%[dst]) \n\t"
+ "sw %[tmp10], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp11], (%[dst]) \n\t"
+ "sw %[tmp11], 4(%[dst]) \n\t"
+ "sw %[tmp11], 8(%[dst]) \n\t"
+ "sw %[tmp11], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp12], (%[dst]) \n\t"
+ "sw %[tmp12], 4(%[dst]) \n\t"
+ "sw %[tmp12], 8(%[dst]) \n\t"
+ "sw %[tmp12], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp13], (%[dst]) \n\t"
+ "sw %[tmp13], 4(%[dst]) \n\t"
+ "sw %[tmp13], 8(%[dst]) \n\t"
+ "sw %[tmp13], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp14], (%[dst]) \n\t"
+ "sw %[tmp14], 4(%[dst]) \n\t"
+ "sw %[tmp14], 8(%[dst]) \n\t"
+ "sw %[tmp14], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp15], (%[dst]) \n\t"
+ "sw %[tmp15], 4(%[dst]) \n\t"
+ "sw %[tmp15], 8(%[dst]) \n\t"
+ "sw %[tmp15], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp16], (%[dst]) \n\t"
+ "sw %[tmp16], 4(%[dst]) \n\t"
+ "sw %[tmp16], 8(%[dst]) \n\t"
+ "sw %[tmp16], 12(%[dst]) \n\t"
+
+ : [tmp1] "=&r" (tmp1), [tmp2] "=&r" (tmp2),
+ [tmp3] "=&r" (tmp3), [tmp4] "=&r" (tmp4),
+ [tmp5] "=&r" (tmp5), [tmp7] "=&r" (tmp7),
+ [tmp6] "=&r" (tmp6), [tmp8] "=&r" (tmp8),
+ [tmp9] "=&r" (tmp9), [tmp10] "=&r" (tmp10),
+ [tmp11] "=&r" (tmp11), [tmp12] "=&r" (tmp12),
+ [tmp13] "=&r" (tmp13), [tmp14] "=&r" (tmp14),
+ [tmp15] "=&r" (tmp15), [tmp16] "=&r" (tmp16)
+ : [left] "r" (left), [dst] "r" (dst), [stride] "r" (stride)
+ );
+}
+
+void vp9_dc_predictor_16x16_dspr2(uint8_t *dst, ptrdiff_t stride,
+ const uint8_t *above, const uint8_t *left) {+ int32_t expected_dc;
+ int32_t average;
+ int32_t tmp, above1, above_l1, above_r1, left1, left_r1, left_l1;
+ int32_t above2, left2;
+
+ __asm__ __volatile__ (
+ "lw %[above1], (%[above]) \n\t"
+ "lw %[above2], 4(%[above]) \n\t"
+ "lw %[left1], (%[left]) \n\t"
+ "lw %[left2], 4(%[left]) \n\t"
+
+ "preceu.ph.qbl %[above_l1], %[above1] \n\t"
+ "preceu.ph.qbr %[above_r1], %[above1] \n\t"
+ "preceu.ph.qbl %[left_l1], %[left1] \n\t"
+ "preceu.ph.qbr %[left_r1], %[left1] \n\t"
+
+ "addu.ph %[average], %[above_r1], %[above_l1] \n\t"
+ "addu.ph %[average], %[average], %[left_l1] \n\t"
+ "addu.ph %[average], %[average], %[left_r1] \n\t"
+
+ "preceu.ph.qbl %[above_l1], %[above2] \n\t"
+ "preceu.ph.qbr %[above_r1], %[above2] \n\t"
+ "preceu.ph.qbl %[left_l1], %[left2] \n\t"
+ "preceu.ph.qbr %[left_r1], %[left2] \n\t"
+
+ "addu.ph %[average], %[average], %[above_l1] \n\t"
+ "addu.ph %[average], %[average], %[above_r1] \n\t"
+ "addu.ph %[average], %[average], %[left_l1] \n\t"
+ "addu.ph %[average], %[average], %[left_r1] \n\t"
+
+ "lw %[above1], 8(%[above]) \n\t"
+ "lw %[above2], 12(%[above]) \n\t"
+ "lw %[left1], 8(%[left]) \n\t"
+ "lw %[left2], 12(%[left]) \n\t"
+
+ "preceu.ph.qbl %[above_l1], %[above1] \n\t"
+ "preceu.ph.qbr %[above_r1], %[above1] \n\t"
+ "preceu.ph.qbl %[left_l1], %[left1] \n\t"
+ "preceu.ph.qbr %[left_r1], %[left1] \n\t"
+
+ "addu.ph %[average], %[average], %[above_l1] \n\t"
+ "addu.ph %[average], %[average], %[above_r1] \n\t"
+ "addu.ph %[average], %[average], %[left_l1] \n\t"
+ "addu.ph %[average], %[average], %[left_r1] \n\t"
+
+ "preceu.ph.qbl %[above_l1], %[above2] \n\t"
+ "preceu.ph.qbr %[above_r1], %[above2] \n\t"
+ "preceu.ph.qbl %[left_l1], %[left2] \n\t"
+ "preceu.ph.qbr %[left_r1], %[left2] \n\t"
+
+ "addu.ph %[average], %[average], %[above_l1] \n\t"
+ "addu.ph %[average], %[average], %[above_r1] \n\t"
+ "addu.ph %[average], %[average], %[left_l1] \n\t"
+ "addu.ph %[average], %[average], %[left_r1] \n\t"
+
+ "addiu %[average], %[average], 16 \n\t"
+ "srl %[tmp], %[average], 16 \n\t"
+ "addu.ph %[average], %[tmp], %[average] \n\t"
+ "srl %[expected_dc], %[average], 5 \n\t"
+ "replv.qb %[expected_dc], %[expected_dc] \n\t"
+
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+ "sw %[expected_dc], 8(%[dst]) \n\t"
+ "sw %[expected_dc], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+ "sw %[expected_dc], 8(%[dst]) \n\t"
+ "sw %[expected_dc], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+ "sw %[expected_dc], 8(%[dst]) \n\t"
+ "sw %[expected_dc], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+ "sw %[expected_dc], 8(%[dst]) \n\t"
+ "sw %[expected_dc], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+ "sw %[expected_dc], 8(%[dst]) \n\t"
+ "sw %[expected_dc], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+ "sw %[expected_dc], 8(%[dst]) \n\t"
+ "sw %[expected_dc], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+ "sw %[expected_dc], 8(%[dst]) \n\t"
+ "sw %[expected_dc], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+ "sw %[expected_dc], 8(%[dst]) \n\t"
+ "sw %[expected_dc], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+ "sw %[expected_dc], 8(%[dst]) \n\t"
+ "sw %[expected_dc], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+ "sw %[expected_dc], 8(%[dst]) \n\t"
+ "sw %[expected_dc], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+ "sw %[expected_dc], 8(%[dst]) \n\t"
+ "sw %[expected_dc], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+ "sw %[expected_dc], 8(%[dst]) \n\t"
+ "sw %[expected_dc], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+ "sw %[expected_dc], 8(%[dst]) \n\t"
+ "sw %[expected_dc], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+ "sw %[expected_dc], 8(%[dst]) \n\t"
+ "sw %[expected_dc], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+ "sw %[expected_dc], 8(%[dst]) \n\t"
+ "sw %[expected_dc], 12(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+ "sw %[expected_dc], 8(%[dst]) \n\t"
+ "sw %[expected_dc], 12(%[dst]) \n\t"
+
+ : [left1] "=&r" (left1), [above1] "=&r" (above1),
+ [left_l1] "=&r" (left_l1), [above_l1] "=&r" (above_l1),
+ [left_r1] "=&r" (left_r1), [above_r1] "=&r" (above_r1),
+ [above2] "=&r" (above2), [left2] "=&r" (left2),
+ [average] "=&r" (average), [tmp] "=&r" (tmp),
+ [expected_dc] "=&r" (expected_dc)
+ : [above] "r" (above), [left] "r" (left),
+ [dst] "r" (dst), [stride] "r" (stride)
+ );
+}
+#endif // #if HAVE_DSPR2
--- /dev/null
+++ b/vp9/common/mips/dspr2/vp9_intrapred4_dspr2.c
@@ -1,0 +1,232 @@
+/*
+ * 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 <stdlib.h>
+
+#include "./vp9_rtcd.h"
+#include "vp9/common/vp9_common.h"
+#include "vp9/common/mips/dspr2/vp9_common_dspr2.h"
+
+#if HAVE_DSPR2
+void vp9_h_predictor_4x4_dspr2(uint8_t *dst, ptrdiff_t stride,
+ const uint8_t *above, const uint8_t *left) {+ int32_t tmp1, tmp2, tmp3, tmp4;
+
+ __asm__ __volatile__ (
+ "lb %[tmp1], (%[left]) \n\t"
+ "lb %[tmp2], 1(%[left]) \n\t"
+ "lb %[tmp3], 2(%[left]) \n\t"
+ "lb %[tmp4], 3(%[left]) \n\t"
+ "replv.qb %[tmp1], %[tmp1] \n\t"
+ "replv.qb %[tmp2], %[tmp2] \n\t"
+ "replv.qb %[tmp3], %[tmp3] \n\t"
+ "replv.qb %[tmp4], %[tmp4] \n\t"
+ "sw %[tmp1], (%[dst]) \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp2], (%[dst]) \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp3], (%[dst]) \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp4], (%[dst]) \n\t"
+
+ : [tmp1] "=&r" (tmp1), [tmp2] "=&r" (tmp2),
+ [tmp3] "=&r" (tmp3), [tmp4] "=&r" (tmp4)
+ : [left] "r" (left), [dst] "r" (dst), [stride] "r" (stride)
+ );
+}
+
+void vp9_dc_predictor_4x4_dspr2(uint8_t *dst, ptrdiff_t stride,
+ const uint8_t *above, const uint8_t *left) {+ int32_t expected_dc;
+ int32_t average;
+ int32_t tmp, above_c, above_l, above_r, left_c, left_r, left_l;
+
+ __asm__ __volatile__ (
+ "lw %[above_c], (%[above]) \n\t"
+ "lw %[left_c], (%[left]) \n\t"
+
+ "preceu.ph.qbl %[above_l], %[above_c] \n\t"
+ "preceu.ph.qbr %[above_r], %[above_c] \n\t"
+ "preceu.ph.qbl %[left_l], %[left_c] \n\t"
+ "preceu.ph.qbr %[left_r], %[left_c] \n\t"
+
+ "addu.ph %[average], %[above_r], %[above_l] \n\t"
+ "addu.ph %[average], %[average], %[left_l] \n\t"
+ "addu.ph %[average], %[average], %[left_r] \n\t"
+ "addiu %[average], %[average], 4 \n\t"
+ "srl %[tmp], %[average], 16 \n\t"
+ "addu.ph %[average], %[tmp], %[average] \n\t"
+ "srl %[expected_dc], %[average], 3 \n\t"
+ "replv.qb %[expected_dc], %[expected_dc] \n\t"
+
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+
+ : [above_c] "=&r" (above_c), [above_l] "=&r" (above_l),
+ [above_r] "=&r" (above_r), [left_c] "=&r" (left_c),
+ [left_l] "=&r" (left_l), [left_r] "=&r" (left_r),
+ [average] "=&r" (average), [tmp] "=&r" (tmp),
+ [expected_dc] "=&r" (expected_dc)
+ : [above] "r" (above), [left] "r" (left),
+ [dst] "r" (dst), [stride] "r" (stride)
+ );
+}
+
+void vp9_tm_predictor_4x4_dspr2(uint8_t *dst, ptrdiff_t stride,
+ const uint8_t *above, const uint8_t *left) {+ int32_t abovel, abover;
+ int32_t left0, left1, left2, left3;
+ int32_t res0, res1;
+ int32_t resl;
+ int32_t resr;
+ int32_t top_left;
+ uint8_t *cm = vp9_ff_cropTbl;
+
+ __asm__ __volatile__ (
+ "ulw %[resl], (%[above]) \n\t"
+
+ "lbu %[left0], (%[left]) \n\t"
+ "lbu %[left1], 1(%[left]) \n\t"
+ "lbu %[left2], 2(%[left]) \n\t"
+ "lbu %[left3], 3(%[left]) \n\t"
+
+ "lbu %[top_left], -1(%[above]) \n\t"
+
+ "preceu.ph.qbl %[abovel], %[resl] \n\t"
+ "preceu.ph.qbr %[abover], %[resl] \n\t"
+
+ "replv.ph %[left0], %[left0] \n\t"
+ "replv.ph %[left1], %[left1] \n\t"
+ "replv.ph %[left2], %[left2] \n\t"
+ "replv.ph %[left3], %[left3] \n\t"
+
+ "replv.ph %[top_left], %[top_left] \n\t"
+
+ "addu.ph %[resl], %[abovel], %[left0] \n\t"
+ "subu.ph %[resl], %[resl], %[top_left] \n\t"
+
+ "addu.ph %[resr], %[abover], %[left0] \n\t"
+ "subu.ph %[resr], %[resr], %[top_left] \n\t"
+
+ "sll %[res0], %[resr], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+
+ "sra %[res1], %[resr], 16 \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "sb %[res0], (%[dst]) \n\t"
+
+ "sll %[res0], %[resl], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+ "sb %[res1], 1(%[dst]) \n\t"
+
+ "sra %[res1], %[resl], 16 \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+
+ "addu.ph %[resl], %[abovel], %[left1] \n\t"
+ "subu.ph %[resl], %[resl], %[top_left] \n\t"
+
+ "addu.ph %[resr], %[abover], %[left1] \n\t"
+ "subu.ph %[resr], %[resr], %[top_left] \n\t"
+
+ "sb %[res0], 2(%[dst]) \n\t"
+ "sb %[res1], 3(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+
+ "sll %[res0], %[resr], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+
+ "sra %[res1], %[resr], 16 \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "sb %[res0], (%[dst]) \n\t"
+
+ "sll %[res0], %[resl], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+
+ "sb %[res1], 1(%[dst]) \n\t"
+ "sra %[res1], %[resl], 16 \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+
+ "addu.ph %[resl], %[abovel], %[left2] \n\t"
+ "subu.ph %[resl], %[resl], %[top_left] \n\t"
+
+ "addu.ph %[resr], %[abover], %[left2] \n\t"
+ "subu.ph %[resr], %[resr], %[top_left] \n\t"
+
+ "sb %[res0], 2(%[dst]) \n\t"
+ "sb %[res1], 3(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+
+ "sll %[res0], %[resr], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+
+
+ "sra %[res1], %[resr], 16 \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "sb %[res0], (%[dst]) \n\t"
+
+ "sll %[res0], %[resl], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+
+
+ "sb %[res1], 1(%[dst]) \n\t"
+ "sra %[res1], %[resl], 16 \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+
+ "addu.ph %[resl], %[abovel], %[left3] \n\t"
+ "subu.ph %[resl], %[resl], %[top_left] \n\t"
+
+ "addu.ph %[resr], %[abover], %[left3] \n\t"
+ "subu.ph %[resr], %[resr], %[top_left] \n\t"
+
+ "sb %[res0], 2(%[dst]) \n\t"
+ "sb %[res1], 3(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+
+ "sll %[res0], %[resr], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+
+ "sra %[res1], %[resr], 16 \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "sb %[res0], (%[dst]) \n\t"
+
+ "sll %[res0], %[resl], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+ "sb %[res1], 1(%[dst]) \n\t"
+
+ "sra %[res1], %[resl], 16 \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+
+ "sb %[res0], 2(%[dst]) \n\t"
+ "sb %[res1], 3(%[dst]) \n\t"
+
+ : [abovel] "=&r" (abovel), [abover] "=&r" (abover),
+ [left0] "=&r" (left0), [left1] "=&r" (left1), [left2] "=&r" (left2),
+ [res0] "=&r" (res0), [res1] "=&r" (res1), [left3] "=&r" (left3),
+ [resl] "=&r" (resl), [resr] "=&r" (resr), [top_left] "=&r" (top_left)
+ : [above] "r" (above), [left] "r" (left),
+ [dst] "r" (dst), [stride] "r" (stride), [cm] "r" (cm)
+ );
+}
+#endif // #if HAVE_DSPR2
--- /dev/null
+++ b/vp9/common/mips/dspr2/vp9_intrapred8_dspr2.c
@@ -1,0 +1,610 @@
+/*
+ * 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 <stdlib.h>
+
+#include "./vp9_rtcd.h"
+#include "vp9/common/vp9_common.h"
+#include "vp9/common/mips/dspr2/vp9_common_dspr2.h"
+
+#if HAVE_DSPR2
+void vp9_h_predictor_8x8_dspr2(uint8_t *dst, ptrdiff_t stride,
+ const uint8_t *above, const uint8_t *left) {+ int32_t tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp8;
+
+ __asm__ __volatile__ (
+ "lb %[tmp1], (%[left]) \n\t"
+ "lb %[tmp2], 1(%[left]) \n\t"
+ "lb %[tmp3], 2(%[left]) \n\t"
+ "lb %[tmp4], 3(%[left]) \n\t"
+ "lb %[tmp5], 4(%[left]) \n\t"
+ "lb %[tmp6], 5(%[left]) \n\t"
+ "lb %[tmp7], 6(%[left]) \n\t"
+ "lb %[tmp8], 7(%[left]) \n\t"
+
+ "replv.qb %[tmp1], %[tmp1] \n\t"
+ "replv.qb %[tmp2], %[tmp2] \n\t"
+ "replv.qb %[tmp3], %[tmp3] \n\t"
+ "replv.qb %[tmp4], %[tmp4] \n\t"
+ "replv.qb %[tmp5], %[tmp5] \n\t"
+ "replv.qb %[tmp6], %[tmp6] \n\t"
+ "replv.qb %[tmp7], %[tmp7] \n\t"
+ "replv.qb %[tmp8], %[tmp8] \n\t"
+
+ "sw %[tmp1], (%[dst]) \n\t"
+ "sw %[tmp1], 4(%[dst]) \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp2], (%[dst]) \n\t"
+ "sw %[tmp2], 4(%[dst]) \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp3], (%[dst]) \n\t"
+ "sw %[tmp3], 4(%[dst]) \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp4], (%[dst]) \n\t"
+ "sw %[tmp4], 4(%[dst]) \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp5], (%[dst]) \n\t"
+ "sw %[tmp5], 4(%[dst]) \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp6], (%[dst]) \n\t"
+ "sw %[tmp6], 4(%[dst]) \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp7], (%[dst]) \n\t"
+ "sw %[tmp7], 4(%[dst]) \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[tmp8], (%[dst]) \n\t"
+ "sw %[tmp8], 4(%[dst]) \n\t"
+
+ : [tmp1] "=&r" (tmp1), [tmp2] "=&r" (tmp2),
+ [tmp3] "=&r" (tmp3), [tmp4] "=&r" (tmp4),
+ [tmp5] "=&r" (tmp5), [tmp7] "=&r" (tmp7),
+ [tmp6] "=&r" (tmp6), [tmp8] "=&r" (tmp8)
+ : [left] "r" (left), [dst] "r" (dst),
+ [stride] "r" (stride)
+ );
+}
+
+void vp9_dc_predictor_8x8_dspr2(uint8_t *dst, ptrdiff_t stride,
+ const uint8_t *above, const uint8_t *left) {+ int32_t expected_dc;
+ int32_t average;
+ int32_t tmp, above1, above_l1, above_r1, left1, left_r1, left_l1;
+ int32_t above2, above_l2, above_r2, left2, left_r2, left_l2;
+
+ __asm__ __volatile__ (
+ "lw %[above1], (%[above]) \n\t"
+ "lw %[above2], 4(%[above]) \n\t"
+ "lw %[left1], (%[left]) \n\t"
+ "lw %[left2], 4(%[left]) \n\t"
+
+ "preceu.ph.qbl %[above_l1], %[above1] \n\t"
+ "preceu.ph.qbr %[above_r1], %[above1] \n\t"
+ "preceu.ph.qbl %[left_l1], %[left1] \n\t"
+ "preceu.ph.qbr %[left_r1], %[left1] \n\t"
+
+ "preceu.ph.qbl %[above_l2], %[above2] \n\t"
+ "preceu.ph.qbr %[above_r2], %[above2] \n\t"
+ "preceu.ph.qbl %[left_l2], %[left2] \n\t"
+ "preceu.ph.qbr %[left_r2], %[left2] \n\t"
+
+ "addu.ph %[average], %[above_r1], %[above_l1] \n\t"
+ "addu.ph %[average], %[average], %[left_l1] \n\t"
+ "addu.ph %[average], %[average], %[left_r1] \n\t"
+
+ "addu.ph %[average], %[average], %[above_l2] \n\t"
+ "addu.ph %[average], %[average], %[above_r2] \n\t"
+ "addu.ph %[average], %[average], %[left_l2] \n\t"
+ "addu.ph %[average], %[average], %[left_r2] \n\t"
+
+ "addiu %[average], %[average], 8 \n\t"
+
+ "srl %[tmp], %[average], 16 \n\t"
+ "addu.ph %[average], %[tmp], %[average] \n\t"
+ "srl %[expected_dc], %[average], 4 \n\t"
+ "replv.qb %[expected_dc], %[expected_dc] \n\t"
+
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+
+ "add %[dst], %[dst], %[stride] \n\t"
+ "sw %[expected_dc], (%[dst]) \n\t"
+ "sw %[expected_dc], 4(%[dst]) \n\t"
+
+ : [above1] "=&r" (above1), [above_l1] "=&r" (above_l1),
+ [above_r1] "=&r" (above_r1), [left1] "=&r" (left1),
+ [left_l1] "=&r" (left_l1), [left_r1] "=&r" (left_r1),
+ [above2] "=&r" (above2), [above_l2] "=&r" (above_l2),
+ [above_r2] "=&r" (above_r2), [left2] "=&r" (left2),
+ [left_l2] "=&r" (left_l2), [left_r2] "=&r" (left_r2),
+ [average] "=&r" (average), [tmp] "=&r" (tmp),
+ [expected_dc] "=&r" (expected_dc)
+ : [above] "r" (above), [left] "r" (left), [dst] "r" (dst),
+ [stride] "r" (stride)
+ );
+}
+
+void vp9_tm_predictor_8x8_dspr2(uint8_t *dst, ptrdiff_t stride,
+ const uint8_t *above, const uint8_t *left) {+ int32_t abovel, abover;
+ int32_t abovel_1, abover_1;
+ int32_t left0;
+ int32_t res0, res1, res2, res3;
+ int32_t reshw;
+ int32_t top_left;
+ uint8_t *cm = vp9_ff_cropTbl;
+
+ __asm__ __volatile__ (
+ "ulw %[reshw], (%[above]) \n\t"
+ "ulw %[top_left], 4(%[above]) \n\t"
+
+ "lbu %[left0], (%[left]) \n\t"
+
+ "preceu.ph.qbl %[abovel], %[reshw] \n\t"
+ "preceu.ph.qbr %[abover], %[reshw] \n\t"
+ "preceu.ph.qbl %[abovel_1], %[top_left] \n\t"
+ "preceu.ph.qbr %[abover_1], %[top_left] \n\t"
+
+ "lbu %[top_left], -1(%[above]) \n\t"
+ "replv.ph %[left0], %[left0] \n\t"
+
+ "replv.ph %[top_left], %[top_left] \n\t"
+
+ "addu.ph %[reshw], %[abovel], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res2], %[reshw], 16 \n\t"
+ "sra %[res2], %[res2], 16 \n\t"
+ "sra %[res3], %[reshw], 16 \n\t"
+
+ "addu.ph %[reshw], %[abover], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res0], %[reshw], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "sra %[res1], %[reshw], 16 \n\t"
+
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "lbux %[res2], %[res2](%[cm]) \n\t"
+ "lbux %[res3], %[res3](%[cm]) \n\t"
+
+ "sb %[res0], (%[dst]) \n\t"
+ "sb %[res1], 1(%[dst]) \n\t"
+ "sb %[res2], 2(%[dst]) \n\t"
+ "sb %[res3], 3(%[dst]) \n\t"
+
+ "addu.ph %[reshw], %[abovel_1], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res2], %[reshw], 16 \n\t"
+ "sra %[res2], %[res2], 16 \n\t"
+ "sra %[res3], %[reshw], 16 \n\t"
+
+ "addu.ph %[reshw], %[abover_1], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res0], %[reshw], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "sra %[res1], %[reshw], 16 \n\t"
+
+ "lbu %[left0], 1(%[left]) \n\t"
+
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "lbux %[res2], %[res2](%[cm]) \n\t"
+ "lbux %[res3], %[res3](%[cm]) \n\t"
+
+ "sb %[res0], 4(%[dst]) \n\t"
+ "sb %[res1], 5(%[dst]) \n\t"
+ "sb %[res2], 6(%[dst]) \n\t"
+ "sb %[res3], 7(%[dst]) \n\t"
+
+ "replv.ph %[left0], %[left0] \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+
+ "addu.ph %[reshw], %[abovel], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res2], %[reshw], 16 \n\t"
+ "sra %[res2], %[res2], 16 \n\t"
+ "sra %[res3], %[reshw], 16 \n\t"
+
+ "addu.ph %[reshw], %[abover], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res0], %[reshw], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "sra %[res1], %[reshw], 16 \n\t"
+
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "lbux %[res2], %[res2](%[cm]) \n\t"
+ "lbux %[res3], %[res3](%[cm]) \n\t"
+
+ "sb %[res0], (%[dst]) \n\t"
+ "sb %[res1], 1(%[dst]) \n\t"
+ "sb %[res2], 2(%[dst]) \n\t"
+ "sb %[res3], 3(%[dst]) \n\t"
+
+ "addu.ph %[reshw], %[abovel_1], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res2], %[reshw], 16 \n\t"
+ "sra %[res2], %[res2], 16 \n\t"
+ "sra %[res3], %[reshw], 16 \n\t"
+
+ "addu.ph %[reshw], %[abover_1], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res0], %[reshw], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "sra %[res1], %[reshw], 16 \n\t"
+
+ "lbu %[left0], 2(%[left]) \n\t"
+
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "lbux %[res2], %[res2](%[cm]) \n\t"
+ "lbux %[res3], %[res3](%[cm]) \n\t"
+
+ "sb %[res0], 4(%[dst]) \n\t"
+ "sb %[res1], 5(%[dst]) \n\t"
+ "sb %[res2], 6(%[dst]) \n\t"
+ "sb %[res3], 7(%[dst]) \n\t"
+
+ "replv.ph %[left0], %[left0] \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+
+ "addu.ph %[reshw], %[abovel], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res2], %[reshw], 16 \n\t"
+ "sra %[res2], %[res2], 16 \n\t"
+ "sra %[res3], %[reshw], 16 \n\t"
+
+ "addu.ph %[reshw], %[abover], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res0], %[reshw], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "sra %[res1], %[reshw], 16 \n\t"
+
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "lbux %[res2], %[res2](%[cm]) \n\t"
+ "lbux %[res3], %[res3](%[cm]) \n\t"
+
+ "sb %[res0], (%[dst]) \n\t"
+ "sb %[res1], 1(%[dst]) \n\t"
+ "sb %[res2], 2(%[dst]) \n\t"
+ "sb %[res3], 3(%[dst]) \n\t"
+
+ "addu.ph %[reshw], %[abovel_1], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res2], %[reshw], 16 \n\t"
+ "sra %[res2], %[res2], 16 \n\t"
+ "sra %[res3], %[reshw], 16 \n\t"
+
+ "addu.ph %[reshw], %[abover_1], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res0], %[reshw], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "sra %[res1], %[reshw], 16 \n\t"
+
+ "lbu %[left0], 3(%[left]) \n\t"
+
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "lbux %[res2], %[res2](%[cm]) \n\t"
+ "lbux %[res3], %[res3](%[cm]) \n\t"
+
+ "sb %[res0], 4(%[dst]) \n\t"
+ "sb %[res1], 5(%[dst]) \n\t"
+ "sb %[res2], 6(%[dst]) \n\t"
+ "sb %[res3], 7(%[dst]) \n\t"
+
+ "replv.ph %[left0], %[left0] \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+
+ "addu.ph %[reshw], %[abovel], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res2], %[reshw], 16 \n\t"
+ "sra %[res2], %[res2], 16 \n\t"
+ "sra %[res3], %[reshw], 16 \n\t"
+
+ "addu.ph %[reshw], %[abover], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res0], %[reshw], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "sra %[res1], %[reshw], 16 \n\t"
+
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "lbux %[res2], %[res2](%[cm]) \n\t"
+ "lbux %[res3], %[res3](%[cm]) \n\t"
+
+ "sb %[res0], (%[dst]) \n\t"
+ "sb %[res1], 1(%[dst]) \n\t"
+ "sb %[res2], 2(%[dst]) \n\t"
+ "sb %[res3], 3(%[dst]) \n\t"
+
+ "addu.ph %[reshw], %[abovel_1], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res2], %[reshw], 16 \n\t"
+ "sra %[res2], %[res2], 16 \n\t"
+ "sra %[res3], %[reshw], 16 \n\t"
+
+ "addu.ph %[reshw], %[abover_1], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res0], %[reshw], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "sra %[res1], %[reshw], 16 \n\t"
+
+ "lbu %[left0], 4(%[left]) \n\t"
+
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "lbux %[res2], %[res2](%[cm]) \n\t"
+ "lbux %[res3], %[res3](%[cm]) \n\t"
+
+ "sb %[res0], 4(%[dst]) \n\t"
+ "sb %[res1], 5(%[dst]) \n\t"
+ "sb %[res2], 6(%[dst]) \n\t"
+ "sb %[res3], 7(%[dst]) \n\t"
+
+ "replv.ph %[left0], %[left0] \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+
+ "addu.ph %[reshw], %[abovel], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res2], %[reshw], 16 \n\t"
+ "sra %[res2], %[res2], 16 \n\t"
+ "sra %[res3], %[reshw], 16 \n\t"
+
+ "addu.ph %[reshw], %[abover], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res0], %[reshw], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "sra %[res1], %[reshw], 16 \n\t"
+
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "lbux %[res2], %[res2](%[cm]) \n\t"
+ "lbux %[res3], %[res3](%[cm]) \n\t"
+
+ "sb %[res0], (%[dst]) \n\t"
+ "sb %[res1], 1(%[dst]) \n\t"
+ "sb %[res2], 2(%[dst]) \n\t"
+ "sb %[res3], 3(%[dst]) \n\t"
+
+ "addu.ph %[reshw], %[abovel_1], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res2], %[reshw], 16 \n\t"
+ "sra %[res2], %[res2], 16 \n\t"
+ "sra %[res3], %[reshw], 16 \n\t"
+
+ "addu.ph %[reshw], %[abover_1], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res0], %[reshw], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "sra %[res1], %[reshw], 16 \n\t"
+
+ "lbu %[left0], 5(%[left]) \n\t"
+
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "lbux %[res2], %[res2](%[cm]) \n\t"
+ "lbux %[res3], %[res3](%[cm]) \n\t"
+
+ "sb %[res0], 4(%[dst]) \n\t"
+ "sb %[res1], 5(%[dst]) \n\t"
+ "sb %[res2], 6(%[dst]) \n\t"
+ "sb %[res3], 7(%[dst]) \n\t"
+
+ "replv.ph %[left0], %[left0] \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+
+ "addu.ph %[reshw], %[abovel], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res2], %[reshw], 16 \n\t"
+ "sra %[res2], %[res2], 16 \n\t"
+ "sra %[res3], %[reshw], 16 \n\t"
+
+ "addu.ph %[reshw], %[abover], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res0], %[reshw], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "sra %[res1], %[reshw], 16 \n\t"
+
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "lbux %[res2], %[res2](%[cm]) \n\t"
+ "lbux %[res3], %[res3](%[cm]) \n\t"
+
+ "sb %[res0], (%[dst]) \n\t"
+ "sb %[res1], 1(%[dst]) \n\t"
+ "sb %[res2], 2(%[dst]) \n\t"
+ "sb %[res3], 3(%[dst]) \n\t"
+
+ "addu.ph %[reshw], %[abovel_1], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res2], %[reshw], 16 \n\t"
+ "sra %[res2], %[res2], 16 \n\t"
+ "sra %[res3], %[reshw], 16 \n\t"
+
+ "addu.ph %[reshw], %[abover_1], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res0], %[reshw], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "sra %[res1], %[reshw], 16 \n\t"
+
+ "lbu %[left0], 6(%[left]) \n\t"
+
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "lbux %[res2], %[res2](%[cm]) \n\t"
+ "lbux %[res3], %[res3](%[cm]) \n\t"
+
+ "sb %[res0], 4(%[dst]) \n\t"
+ "sb %[res1], 5(%[dst]) \n\t"
+ "sb %[res2], 6(%[dst]) \n\t"
+ "sb %[res3], 7(%[dst]) \n\t"
+
+ "replv.ph %[left0], %[left0] \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+
+ "addu.ph %[reshw], %[abovel], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res2], %[reshw], 16 \n\t"
+ "sra %[res2], %[res2], 16 \n\t"
+ "sra %[res3], %[reshw], 16 \n\t"
+
+ "addu.ph %[reshw], %[abover], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res0], %[reshw], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "sra %[res1], %[reshw], 16 \n\t"
+
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "lbux %[res2], %[res2](%[cm]) \n\t"
+ "lbux %[res3], %[res3](%[cm]) \n\t"
+
+ "sb %[res0], (%[dst]) \n\t"
+ "sb %[res1], 1(%[dst]) \n\t"
+ "sb %[res2], 2(%[dst]) \n\t"
+ "sb %[res3], 3(%[dst]) \n\t"
+
+ "addu.ph %[reshw], %[abovel_1], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res2], %[reshw], 16 \n\t"
+ "sra %[res2], %[res2], 16 \n\t"
+ "sra %[res3], %[reshw], 16 \n\t"
+
+ "addu.ph %[reshw], %[abover_1], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res0], %[reshw], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "sra %[res1], %[reshw], 16 \n\t"
+
+ "lbu %[left0], 7(%[left]) \n\t"
+
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "lbux %[res2], %[res2](%[cm]) \n\t"
+ "lbux %[res3], %[res3](%[cm]) \n\t"
+
+ "sb %[res0], 4(%[dst]) \n\t"
+ "sb %[res1], 5(%[dst]) \n\t"
+ "sb %[res2], 6(%[dst]) \n\t"
+ "sb %[res3], 7(%[dst]) \n\t"
+
+ "replv.ph %[left0], %[left0] \n\t"
+ "add %[dst], %[dst], %[stride] \n\t"
+
+ "addu.ph %[reshw], %[abovel], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res2], %[reshw], 16 \n\t"
+ "sra %[res2], %[res2], 16 \n\t"
+ "sra %[res3], %[reshw], 16 \n\t"
+
+ "addu.ph %[reshw], %[abover], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res0], %[reshw], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "sra %[res1], %[reshw], 16 \n\t"
+
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "lbux %[res2], %[res2](%[cm]) \n\t"
+ "lbux %[res3], %[res3](%[cm]) \n\t"
+
+ "sb %[res0], (%[dst]) \n\t"
+ "sb %[res1], 1(%[dst]) \n\t"
+ "sb %[res2], 2(%[dst]) \n\t"
+ "sb %[res3], 3(%[dst]) \n\t"
+
+ "addu.ph %[reshw], %[abovel_1], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res2], %[reshw], 16 \n\t"
+ "sra %[res2], %[res2], 16 \n\t"
+ "sra %[res3], %[reshw], 16 \n\t"
+
+ "addu.ph %[reshw], %[abover_1], %[left0] \n\t"
+ "subu.ph %[reshw], %[reshw], %[top_left] \n\t"
+
+ "sll %[res0], %[reshw], 16 \n\t"
+ "sra %[res0], %[res0], 16 \n\t"
+ "sra %[res1], %[reshw], 16 \n\t"
+
+ "lbux %[res0], %[res0](%[cm]) \n\t"
+ "lbux %[res1], %[res1](%[cm]) \n\t"
+ "lbux %[res2], %[res2](%[cm]) \n\t"
+ "lbux %[res3], %[res3](%[cm]) \n\t"
+
+ "sb %[res0], 4(%[dst]) \n\t"
+ "sb %[res1], 5(%[dst]) \n\t"
+ "sb %[res2], 6(%[dst]) \n\t"
+ "sb %[res3], 7(%[dst]) \n\t"
+
+ : [abovel] "=&r" (abovel), [abover] "=&r" (abover),
+ [abovel_1] "=&r" (abovel_1), [abover_1] "=&r" (abover_1),
+ [left0] "=&r" (left0), [res2] "=&r" (res2), [res3] "=&r" (res3),
+ [res0] "=&r" (res0), [res1] "=&r" (res1),
+ [reshw] "=&r" (reshw), [top_left] "=&r" (top_left)
+ : [above] "r" (above), [left] "r" (left),
+ [dst] "r" (dst), [stride] "r" (stride), [cm] "r" (cm)
+ );
+}
+#endif // #if HAVE_DSPR2
--- a/vp9/common/vp9_rtcd_defs.sh
+++ b/vp9/common/vp9_rtcd_defs.sh
@@ -41,7 +41,7 @@
specialize vp9_d63_predictor_4x4 $ssse3_x86inc
prototype void vp9_h_predictor_4x4 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_h_predictor_4x4 $ssse3_x86inc
+specialize vp9_h_predictor_4x4 $ssse3_x86inc dspr2
prototype void vp9_d117_predictor_4x4 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
specialize vp9_d117_predictor_4x4
@@ -56,10 +56,10 @@
specialize vp9_v_predictor_4x4 $sse_x86inc
prototype void vp9_tm_predictor_4x4 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_tm_predictor_4x4 $sse_x86inc
+specialize vp9_tm_predictor_4x4 $sse_x86inc dspr2
prototype void vp9_dc_predictor_4x4 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_predictor_4x4 $sse_x86inc
+specialize vp9_dc_predictor_4x4 $sse_x86inc dspr2
prototype void vp9_dc_top_predictor_4x4 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
specialize vp9_dc_top_predictor_4x4
@@ -80,7 +80,7 @@
specialize vp9_d63_predictor_8x8 $ssse3_x86inc
prototype void vp9_h_predictor_8x8 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_h_predictor_8x8 $ssse3_x86inc
+specialize vp9_h_predictor_8x8 $ssse3_x86inc dspr2
prototype void vp9_d117_predictor_8x8 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
specialize vp9_d117_predictor_8x8
@@ -95,10 +95,10 @@
specialize vp9_v_predictor_8x8 $sse_x86inc
prototype void vp9_tm_predictor_8x8 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_tm_predictor_8x8 $sse2_x86inc
+specialize vp9_tm_predictor_8x8 $sse2_x86inc dspr2
prototype void vp9_dc_predictor_8x8 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_predictor_8x8 $sse_x86inc
+specialize vp9_dc_predictor_8x8 $sse_x86inc dspr2
prototype void vp9_dc_top_predictor_8x8 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
specialize vp9_dc_top_predictor_8x8
@@ -119,7 +119,7 @@
specialize vp9_d63_predictor_16x16 $ssse3_x86inc
prototype void vp9_h_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_h_predictor_16x16 $ssse3_x86inc
+specialize vp9_h_predictor_16x16 $ssse3_x86inc dspr2
prototype void vp9_d117_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
specialize vp9_d117_predictor_16x16
@@ -137,7 +137,7 @@
specialize vp9_tm_predictor_16x16 $sse2_x86inc
prototype void vp9_dc_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_dc_predictor_16x16 $sse2_x86inc
+specialize vp9_dc_predictor_16x16 $sse2_x86inc dspr2
prototype void vp9_dc_top_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
specialize vp9_dc_top_predictor_16x16
--- a/vp9/vp9_common.mk
+++ b/vp9/vp9_common.mk
@@ -103,6 +103,9 @@
VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_convolve8_dspr2.c
VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_convolve8_horiz_dspr2.c
VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_convolve8_vert_dspr2.c
+VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_intrapred4_dspr2.c
+VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_intrapred8_dspr2.c
+VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_intrapred16_dspr2.c
VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_itrans4_dspr2.c
VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_itrans8_dspr2.c
VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_itrans16_dspr2.c