libvpx: Pull from upstream

source/libvpx/vpx_dsp/mips/vpx_convolve8_msa.c

 /*
  *  Copyright (c) 2015 The WebM project authors. All Rights Reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
  *  tree. An additional intellectual property rights grant can be found
  *  in the file PATENTS.  All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
+#include <assert.h>
 #include "./vpx_dsp_rtcd.h"
 #include "vpx_dsp/mips/vpx_convolve_msa.h"
 
 const uint8_t mc_filt_mask_arr[16 * 3] = {
   /* 8 width cases */
   0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8,
   /* 4 width cases */
   0, 1, 1, 2, 2, 3, 3, 4, 16, 17, 17, 18, 18, 19, 19, 20,
   /* 4 width cases */
   8, 9, 9, 10, 10, 11, 11, 12, 24, 25, 25, 26, 26, 27, 27, 28
@@ skipping 228 matching lines @@
   LD_SB5(src, src_stride, src0, src1, src2, src3, src4);
   hz_out0 = HORIZ_2TAP_FILT_UH(src0, src1, mask, filt_hz, FILTER_BITS);
   hz_out2 = HORIZ_2TAP_FILT_UH(src2, src3, mask, filt_hz, FILTER_BITS);
   hz_out4 = HORIZ_2TAP_FILT_UH(src4, src4, mask, filt_hz, FILTER_BITS);
   hz_out1 = (v8u16)__msa_sldi_b((v16i8)hz_out2, (v16i8)hz_out0, 8);
   hz_out3 = (v8u16)__msa_pckod_d((v2i64)hz_out4, (v2i64)hz_out2);
 
   ILVEV_B2_UB(hz_out0, hz_out1, hz_out2, hz_out3, vec0, vec1);
   DOTP_UB2_UH(vec0, vec1, filt_vt, filt_vt, tmp0, tmp1);
   SRARI_H2_UH(tmp0, tmp1, FILTER_BITS);
-  SAT_UH2_UH(tmp0, tmp1, 7);
   PCKEV_B2_UB(tmp0, tmp0, tmp1, tmp1, res0, res1);
   ST4x4_UB(res0, res1, 0, 1, 0, 1, dst, dst_stride);
 }
 
 static void common_hv_2ht_2vt_4x8_msa(const uint8_t *src, int32_t src_stride,
                                       uint8_t *dst, int32_t dst_stride,
                                       int8_t *filter_horiz,
                                       int8_t *filter_vert) {
   v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, mask;
   v16i8 res0, res1, res2, res3;
@@ skipping 21 matching lines @@
   hz_out8 = HORIZ_2TAP_FILT_UH(src8, src8, mask, filt_hz, FILTER_BITS);
   SLDI_B3_UH(hz_out2, hz_out4, hz_out6, hz_out0, hz_out2, hz_out4, hz_out1,
              hz_out3, hz_out5, 8);
   hz_out7 = (v8u16)__msa_pckod_d((v2i64)hz_out8, (v2i64)hz_out6);
 
   ILVEV_B2_UB(hz_out0, hz_out1, hz_out2, hz_out3, vec0, vec1);
   ILVEV_B2_UB(hz_out4, hz_out5, hz_out6, hz_out7, vec2, vec3);
   DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt_vt, filt_vt, filt_vt, filt_vt,
               vec4, vec5, vec6, vec7);
   SRARI_H4_UH(vec4, vec5, vec6, vec7, FILTER_BITS);
-  SAT_UH4_UH(vec4, vec5, vec6, vec7, 7);
   PCKEV_B4_SB(vec4, vec4, vec5, vec5, vec6, vec6, vec7, vec7, res0, res1,
               res2, res3);
   ST4x4_UB(res0, res1, 0, 1, 0, 1, dst, dst_stride);
   dst += (4 * dst_stride);
   ST4x4_UB(res2, res3, 0, 1, 0, 1, dst, dst_stride);
 }
 
 static void common_hv_2ht_2vt_4w_msa(const uint8_t *src, int32_t src_stride,
                                      uint8_t *dst, int32_t dst_stride,
                                      int8_t *filter_horiz, int8_t *filter_vert,
@@ skipping 38 matching lines @@
 
   hz_out1 = HORIZ_2TAP_FILT_UH(src3, src3, mask, filt_hz, FILTER_BITS);
   vec2 = (v16u8)__msa_ilvev_b((v16i8)hz_out1, (v16i8)hz_out0);
   tmp2 = __msa_dotp_u_h(vec2, filt_vt);
 
   hz_out0 = HORIZ_2TAP_FILT_UH(src4, src4, mask, filt_hz, FILTER_BITS);
   vec3 = (v16u8)__msa_ilvev_b((v16i8)hz_out0, (v16i8)hz_out1);
   tmp3 = __msa_dotp_u_h(vec3, filt_vt);
 
   SRARI_H4_UH(tmp0, tmp1, tmp2, tmp3, FILTER_BITS);
-  SAT_UH4_UH(tmp0, tmp1, tmp2, tmp3, 7);
   PCKEV_B2_SB(tmp1, tmp0, tmp3, tmp2, out0, out1);
   ST8x4_UB(out0, out1, dst, dst_stride);
 }
 
 static void common_hv_2ht_2vt_8x8mult_msa(const uint8_t *src,
                                           int32_t src_stride,
                                           uint8_t *dst,
                                           int32_t dst_stride,
                                           int8_t *filter_horiz,
                                           int8_t *filter_vert,
@@ skipping 24 matching lines @@
 
     hz_out1 = HORIZ_2TAP_FILT_UH(src1, src1, mask, filt_hz, FILTER_BITS);
     vec0 = (v16u8)__msa_ilvev_b((v16i8)hz_out1, (v16i8)hz_out0);
     tmp1 = __msa_dotp_u_h(vec0, filt_vt);
 
     hz_out0 = HORIZ_2TAP_FILT_UH(src2, src2, mask, filt_hz, FILTER_BITS);
     vec0 = (v16u8)__msa_ilvev_b((v16i8)hz_out0, (v16i8)hz_out1);
     tmp2 = __msa_dotp_u_h(vec0, filt_vt);
 
     SRARI_H2_UH(tmp1, tmp2, FILTER_BITS);
-    SAT_UH2_UH(tmp1, tmp2, 7);
 
     hz_out1 = HORIZ_2TAP_FILT_UH(src3, src3, mask, filt_hz, FILTER_BITS);
     vec0 = (v16u8)__msa_ilvev_b((v16i8)hz_out1, (v16i8)hz_out0);
     tmp3 = __msa_dotp_u_h(vec0, filt_vt);
 
     hz_out0 = HORIZ_2TAP_FILT_UH(src4, src4, mask, filt_hz, FILTER_BITS);
     LD_SB4(src, src_stride, src1, src2, src3, src4);
     src += (4 * src_stride);
     vec0 = (v16u8)__msa_ilvev_b((v16i8)hz_out0, (v16i8)hz_out1);
     tmp4 = __msa_dotp_u_h(vec0, filt_vt);
 
     SRARI_H2_UH(tmp3, tmp4, FILTER_BITS);
-    SAT_UH2_UH(tmp3, tmp4, 7);
     PCKEV_B2_SB(tmp2, tmp1, tmp4, tmp3, out0, out1);
     ST8x4_UB(out0, out1, dst, dst_stride);
     dst += (4 * dst_stride);
 
     hz_out1 = HORIZ_2TAP_FILT_UH(src1, src1, mask, filt_hz, FILTER_BITS);
     vec0 = (v16u8)__msa_ilvev_b((v16i8)hz_out1, (v16i8)hz_out0);
     tmp5 = __msa_dotp_u_h(vec0, filt_vt);
 
     hz_out0 = HORIZ_2TAP_FILT_UH(src2, src2, mask, filt_hz, FILTER_BITS);
     vec0 = (v16u8)__msa_ilvev_b((v16i8)hz_out0, (v16i8)hz_out1);
     tmp6 = __msa_dotp_u_h(vec0, filt_vt);
 
     hz_out1 = HORIZ_2TAP_FILT_UH(src3, src3, mask, filt_hz, FILTER_BITS);
     vec0 = (v16u8)__msa_ilvev_b((v16i8)hz_out1, (v16i8)hz_out0);
     tmp7 = __msa_dotp_u_h(vec0, filt_vt);
 
     hz_out0 = HORIZ_2TAP_FILT_UH(src4, src4, mask, filt_hz, FILTER_BITS);
     vec0 = (v16u8)__msa_ilvev_b((v16i8)hz_out0, (v16i8)hz_out1);
     tmp8 = __msa_dotp_u_h(vec0, filt_vt);
 
     SRARI_H4_UH(tmp5, tmp6, tmp7, tmp8, FILTER_BITS);
-    SAT_UH4_UH(tmp5, tmp6, tmp7, tmp8, 7);
     PCKEV_B2_SB(tmp6, tmp5, tmp8, tmp7, out0, out1);
     ST8x4_UB(out0, out1, dst, dst_stride);
     dst += (4 * dst_stride);
   }
 }
 
 static void common_hv_2ht_2vt_8w_msa(const uint8_t *src, int32_t src_stride,
                                      uint8_t *dst, int32_t dst_stride,
                                      int8_t *filter_horiz, int8_t *filter_vert,
                                      int32_t height) {
@@ skipping 34 matching lines @@
   for (loop_cnt = (height >> 2); loop_cnt--;) {
     LD_SB4(src, src_stride, src0, src2, src4, src6);
     LD_SB4(src + 8, src_stride, src1, src3, src5, src7);
     src += (4 * src_stride);
 
     hz_out1 = HORIZ_2TAP_FILT_UH(src0, src0, mask, filt_hz, FILTER_BITS);
     hz_out3 = HORIZ_2TAP_FILT_UH(src1, src1, mask, filt_hz, FILTER_BITS);
     ILVEV_B2_UB(hz_out0, hz_out1, hz_out2, hz_out3, vec0, vec1);
     DOTP_UB2_UH(vec0, vec1, filt_vt, filt_vt, tmp1, tmp2);
     SRARI_H2_UH(tmp1, tmp2, FILTER_BITS);
-    SAT_UH2_UH(tmp1, tmp2, 7);
     PCKEV_ST_SB(tmp1, tmp2, dst);
     dst += dst_stride;
 
     hz_out0 = HORIZ_2TAP_FILT_UH(src2, src2, mask, filt_hz, FILTER_BITS);
     hz_out2 = HORIZ_2TAP_FILT_UH(src3, src3, mask, filt_hz, FILTER_BITS);
     ILVEV_B2_UB(hz_out1, hz_out0, hz_out3, hz_out2, vec0, vec1);
     DOTP_UB2_UH(vec0, vec1, filt_vt, filt_vt, tmp1, tmp2);
     SRARI_H2_UH(tmp1, tmp2, FILTER_BITS);
-    SAT_UH2_UH(tmp1, tmp2, 7);
     PCKEV_ST_SB(tmp1, tmp2, dst);
     dst += dst_stride;
 
     hz_out1 = HORIZ_2TAP_FILT_UH(src4, src4, mask, filt_hz, FILTER_BITS);
     hz_out3 = HORIZ_2TAP_FILT_UH(src5, src5, mask, filt_hz, FILTER_BITS);
     ILVEV_B2_UB(hz_out0, hz_out1, hz_out2, hz_out3, vec0, vec1);
     DOTP_UB2_UH(vec0, vec1, filt_vt, filt_vt, tmp1, tmp2);
     SRARI_H2_UH(tmp1, tmp2, FILTER_BITS);
-    SAT_UH2_UH(tmp1, tmp2, 7);
     PCKEV_ST_SB(tmp1, tmp2, dst);
     dst += dst_stride;
 
     hz_out0 = HORIZ_2TAP_FILT_UH(src6, src6, mask, filt_hz, FILTER_BITS);
     hz_out2 = HORIZ_2TAP_FILT_UH(src7, src7, mask, filt_hz, FILTER_BITS);
     ILVEV_B2_UB(hz_out1, hz_out0, hz_out3, hz_out2, vec0, vec1);
     DOTP_UB2_UH(vec0, vec1, filt_vt, filt_vt, tmp1, tmp2);
     SRARI_H2_UH(tmp1, tmp2, FILTER_BITS);
-    SAT_UH2_UH(tmp1, tmp2, 7);
     PCKEV_ST_SB(tmp1, tmp2, dst);
     dst += dst_stride;
   }
 }
 
 static void common_hv_2ht_2vt_32w_msa(const uint8_t *src, int32_t src_stride,
                                       uint8_t *dst, int32_t dst_stride,
                                       int8_t *filter_horiz, int8_t *filter_vert,
                                       int32_t height) {
   int32_t multiple8_cnt;
@@ skipping 18 matching lines @@
   }
 }
 
 void vpx_convolve8_msa(const uint8_t *src, ptrdiff_t src_stride,
                        uint8_t *dst, ptrdiff_t dst_stride,
                        const int16_t *filter_x, int32_t x_step_q4,
                        const int16_t *filter_y, int32_t y_step_q4,
                        int32_t w, int32_t h) {
   int8_t cnt, filt_hor[8], filt_ver[8];
 
-  if (16 != x_step_q4 || 16 != y_step_q4) {
-    vpx_convolve8_c(src, src_stride, dst, dst_stride,
-                    filter_x, x_step_q4, filter_y, y_step_q4,
-                    w, h);
-    return;
-  }
-
-  if (((const int32_t *)filter_x)[1] == 0x800000 &&
-      ((const int32_t *)filter_y)[1] == 0x800000) {
-    vpx_convolve_copy(src, src_stride, dst, dst_stride,
-                      filter_x, x_step_q4, filter_y, y_step_q4,
-                      w, h);
-    return;
-  }
+  assert(x_step_q4 == 16);
+  assert(y_step_q4 == 16);
+  assert(((const int32_t *)filter_x)[1] != 0x800000);
+  assert(((const int32_t *)filter_y)[1] != 0x800000);
 
   for (cnt = 0; cnt < 8; ++cnt) {
     filt_hor[cnt] = filter_x[cnt];
     filt_ver[cnt] = filter_y[cnt];
   }
 
   if (((const int32_t *)filter_x)[0] == 0 &&
       ((const int32_t *)filter_y)[0] == 0) {
     switch (w) {
       case 4:
@@ skipping 60 matching lines @@
                                   filt_hor, filt_ver, (int32_t)h);
         break;
       default:
         vpx_convolve8_c(src, src_stride, dst, dst_stride,
                         filter_x, x_step_q4, filter_y, y_step_q4,
                         w, h);
         break;
     }
   }
 }