libvpx: Pull from upstream

source/libvpx/vpx_dsp/mips/vpx_convolve8_vert_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"
 
 static void common_vt_8t_4w_msa(const uint8_t *src, int32_t src_stride,
                                 uint8_t *dst, int32_t dst_stride,
                                 int8_t *filter, int32_t height) {
   uint32_t loop_cnt;
   v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8, src9, src10;
   v16i8 src10_r, src32_r, src54_r, src76_r, src98_r, src21_r, src43_r;
   v16i8 src65_r, src87_r, src109_r, src2110, src4332, src6554, src8776;
@@ skipping 288 matching lines @@
   filt0 = (v16u8)__msa_splati_h(filt, 0);
 
   LD_SB5(src, src_stride, src0, src1, src2, src3, src4);
   src += (5 * src_stride);
 
   ILVR_B4_SB(src1, src0, src2, src1, src3, src2, src4, src3, src10_r, src21_r,
              src32_r, src43_r);
   ILVR_D2_SB(src21_r, src10_r, src43_r, src32_r, src2110, src4332);
   DOTP_UB2_UH(src2110, src4332, filt0, filt0, tmp0, tmp1);
   SRARI_H2_UH(tmp0, tmp1, FILTER_BITS);
-  SAT_UH2_UH(tmp0, tmp1, 7);
   src2110 = __msa_pckev_b((v16i8)tmp1, (v16i8)tmp0);
   ST4x4_UB(src2110, src2110, 0, 1, 2, 3, dst, dst_stride);
 }
 
 static void common_vt_2t_4x8_msa(const uint8_t *src, int32_t src_stride,
                                  uint8_t *dst, int32_t dst_stride,
                                  int8_t *filter) {
   v16i8 src0, src1, src2, src3, src4, src5, src6, src7, src8;
   v16i8 src10_r, src32_r, src54_r, src76_r, src21_r, src43_r;
   v16i8 src65_r, src87_r, src2110, src4332, src6554, src8776;
@@ skipping 12 matching lines @@
 
   ILVR_B4_SB(src1, src0, src2, src1, src3, src2, src4, src3, src10_r, src21_r,
              src32_r, src43_r);
   ILVR_B4_SB(src5, src4, src6, src5, src7, src6, src8, src7, src54_r, src65_r,
              src76_r, src87_r);
   ILVR_D4_SB(src21_r, src10_r, src43_r, src32_r, src65_r, src54_r,
              src87_r, src76_r, src2110, src4332, src6554, src8776);
   DOTP_UB4_UH(src2110, src4332, src6554, src8776, filt0, filt0, filt0, filt0,
               tmp0, tmp1, tmp2, tmp3);
   SRARI_H4_UH(tmp0, tmp1, tmp2, tmp3, FILTER_BITS);
-  SAT_UH4_UH(tmp0, tmp1, tmp2, tmp3, 7);
   PCKEV_B2_SB(tmp1, tmp0, tmp3, tmp2, src2110, src4332);
   ST4x4_UB(src2110, src2110, 0, 1, 2, 3, dst, dst_stride);
   ST4x4_UB(src4332, src4332, 0, 1, 2, 3, dst + 4 * dst_stride, dst_stride);
 }
 
 static void common_vt_2t_4w_msa(const uint8_t *src, int32_t src_stride,
                                 uint8_t *dst, int32_t dst_stride,
                                 int8_t *filter, int32_t height) {
   if (4 == height) {
     common_vt_2t_4x4_msa(src, src_stride, dst, dst_stride, filter);
@@ skipping 13 matching lines @@
   /* rearranging filter_y */
   filt = LD_SH(filter);
   filt0 = (v16u8)__msa_splati_h(filt, 0);
 
   LD_UB5(src, src_stride, src0, src1, src2, src3, src4);
   ILVR_B2_UB(src1, src0, src2, src1, vec0, vec1);
   ILVR_B2_UB(src3, src2, src4, src3, vec2, vec3);
   DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, tmp0, tmp1,
               tmp2, tmp3);
   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_vt_2t_8x8mult_msa(const uint8_t *src, int32_t src_stride,
                                      uint8_t *dst, int32_t dst_stride,
                                      int8_t *filter, int32_t height) {
   uint32_t loop_cnt;
   v16u8 src0, src1, src2, src3, src4, src5, src6, src7, src8;
   v16u8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7, filt0;
@@ skipping 12 matching lines @@
     LD_UB8(src, src_stride, src1, src2, src3, src4, src5, src6, src7, src8);
     src += (8 * src_stride);
 
     ILVR_B4_UB(src1, src0, src2, src1, src3, src2, src4, src3, vec0, vec1,
                vec2, vec3);
     ILVR_B4_UB(src5, src4, src6, src5, src7, src6, src8, src7, vec4, vec5,
                vec6, vec7);
     DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, tmp0, tmp1,
                 tmp2, tmp3);
     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);
     dst += (4 * dst_stride);
 
     DOTP_UB4_UH(vec4, vec5, vec6, vec7, filt0, filt0, filt0, filt0, tmp0, tmp1,
                 tmp2, tmp3);
     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);
     dst += (4 * dst_stride);
 
     src0 = src8;
   }
 }
 
 static void common_vt_2t_8w_msa(const uint8_t *src, int32_t src_stride,
                                 uint8_t *dst, int32_t dst_stride,
@@ skipping 22 matching lines @@
   src += src_stride;
 
   for (loop_cnt = (height >> 2); loop_cnt--;) {
     LD_UB4(src, src_stride, src1, src2, src3, src4);
     src += (4 * src_stride);
 
     ILVR_B2_UB(src1, src0, src2, src1, vec0, vec2);
     ILVL_B2_UB(src1, src0, src2, src1, vec1, vec3);
     DOTP_UB2_UH(vec0, vec1, filt0, filt0, tmp0, tmp1);
     SRARI_H2_UH(tmp0, tmp1, FILTER_BITS);
-    SAT_UH2_UH(tmp0, tmp1, 7);
     PCKEV_ST_SB(tmp0, tmp1, dst);
     dst += dst_stride;
 
     ILVR_B2_UB(src3, src2, src4, src3, vec4, vec6);
     ILVL_B2_UB(src3, src2, src4, src3, vec5, vec7);
     DOTP_UB2_UH(vec2, vec3, filt0, filt0, tmp2, tmp3);
     SRARI_H2_UH(tmp2, tmp3, FILTER_BITS);
-    SAT_UH2_UH(tmp2, tmp3, 7);
     PCKEV_ST_SB(tmp2, tmp3, dst);
     dst += dst_stride;
 
     DOTP_UB2_UH(vec4, vec5, filt0, filt0, tmp0, tmp1);
     SRARI_H2_UH(tmp0, tmp1, FILTER_BITS);
-    SAT_UH2_UH(tmp0, tmp1, 7);
     PCKEV_ST_SB(tmp0, tmp1, dst);
     dst += dst_stride;
 
     DOTP_UB2_UH(vec6, vec7, filt0, filt0, tmp2, tmp3);
     SRARI_H2_UH(tmp2, tmp3, FILTER_BITS);
-    SAT_UH2_UH(tmp2, tmp3, 7);
     PCKEV_ST_SB(tmp2, tmp3, dst);
     dst += dst_stride;
 
     src0 = src4;
   }
 }
 
 static void common_vt_2t_32w_msa(const uint8_t *src, int32_t src_stride,
                                  uint8_t *dst, int32_t dst_stride,
                                  int8_t *filter, int32_t height) {
@@ skipping 14 matching lines @@
   for (loop_cnt = (height >> 2); loop_cnt--;) {
     LD_UB4(src, src_stride, src1, src2, src3, src4);
     ILVR_B2_UB(src1, src0, src2, src1, vec0, vec2);
     ILVL_B2_UB(src1, src0, src2, src1, vec1, vec3);
 
     LD_UB4(src + 16, src_stride, src6, src7, src8, src9);
     src += (4 * src_stride);
 
     DOTP_UB2_UH(vec0, vec1, filt0, filt0, tmp0, tmp1);
     SRARI_H2_UH(tmp0, tmp1, FILTER_BITS);
-    SAT_UH2_UH(tmp0, tmp1, 7);
     PCKEV_ST_SB(tmp0, tmp1, dst);
     DOTP_UB2_UH(vec2, vec3, filt0, filt0, tmp2, tmp3);
     SRARI_H2_UH(tmp2, tmp3, FILTER_BITS);
-    SAT_UH2_UH(tmp2, tmp3, 7);
     PCKEV_ST_SB(tmp2, tmp3, dst + dst_stride);
 
     ILVR_B2_UB(src3, src2, src4, src3, vec4, vec6);
     ILVL_B2_UB(src3, src2, src4, src3, vec5, vec7);
     DOTP_UB2_UH(vec4, vec5, filt0, filt0, tmp0, tmp1);
     SRARI_H2_UH(tmp0, tmp1, FILTER_BITS);
-    SAT_UH2_UH(tmp0, tmp1, 7);
     PCKEV_ST_SB(tmp0, tmp1, dst + 2 * dst_stride);
 
     DOTP_UB2_UH(vec6, vec7, filt0, filt0, tmp2, tmp3);
     SRARI_H2_UH(tmp2, tmp3, FILTER_BITS);
-    SAT_UH2_UH(tmp2, tmp3, 7);
     PCKEV_ST_SB(tmp2, tmp3, dst + 3 * dst_stride);
 
     ILVR_B2_UB(src6, src5, src7, src6, vec0, vec2);
     ILVL_B2_UB(src6, src5, src7, src6, vec1, vec3);
     DOTP_UB2_UH(vec0, vec1, filt0, filt0, tmp0, tmp1);
     SRARI_H2_UH(tmp0, tmp1, FILTER_BITS);
-    SAT_UH2_UH(tmp0, tmp1, 7);
     PCKEV_ST_SB(tmp0, tmp1, dst + 16);
 
     DOTP_UB2_UH(vec2, vec3, filt0, filt0, tmp2, tmp3);
     SRARI_H2_UH(tmp2, tmp3, FILTER_BITS);
-    SAT_UH2_UH(tmp2, tmp3, 7);
     PCKEV_ST_SB(tmp2, tmp3, dst + 16 + dst_stride);
 
     ILVR_B2_UB(src8, src7, src9, src8, vec4, vec6);
     ILVL_B2_UB(src8, src7, src9, src8, vec5, vec7);
     DOTP_UB2_UH(vec4, vec5, filt0, filt0, tmp0, tmp1);
     SRARI_H2_UH(tmp0, tmp1, FILTER_BITS);
-    SAT_UH2_UH(tmp0, tmp1, 7);
     PCKEV_ST_SB(tmp0, tmp1, dst + 16 + 2 * dst_stride);
 
     DOTP_UB2_UH(vec6, vec7, filt0, filt0, tmp2, tmp3);
     SRARI_H2_UH(tmp2, tmp3, FILTER_BITS);
-    SAT_UH2_UH(tmp2, tmp3, 7);
     PCKEV_ST_SB(tmp2, tmp3, dst + 16 + 3 * dst_stride);
     dst += (4 * dst_stride);
 
     src0 = src4;
     src5 = src9;
   }
 }
 
 static void common_vt_2t_64w_msa(const uint8_t *src, int32_t src_stride,
                                  uint8_t *dst, int32_t dst_stride,
@@ skipping 15 matching lines @@
     LD_UB2(src, src_stride, src1, src2);
     LD_UB2(src + 16, src_stride, src4, src5);
     LD_UB2(src + 32, src_stride, src7, src8);
     LD_UB2(src + 48, src_stride, src10, src11);
     src += (2 * src_stride);
 
     ILVR_B2_UB(src1, src0, src2, src1, vec0, vec2);
     ILVL_B2_UB(src1, src0, src2, src1, vec1, vec3);
     DOTP_UB2_UH(vec0, vec1, filt0, filt0, tmp0, tmp1);
     SRARI_H2_UH(tmp0, tmp1, FILTER_BITS);
-    SAT_UH2_UH(tmp0, tmp1, 7);
     PCKEV_ST_SB(tmp0, tmp1, dst);
 
     DOTP_UB2_UH(vec2, vec3, filt0, filt0, tmp2, tmp3);
     SRARI_H2_UH(tmp2, tmp3, FILTER_BITS);
-    SAT_UH2_UH(tmp2, tmp3, 7);
     PCKEV_ST_SB(tmp2, tmp3, dst + dst_stride);
 
     ILVR_B2_UB(src4, src3, src5, src4, vec4, vec6);
     ILVL_B2_UB(src4, src3, src5, src4, vec5, vec7);
     DOTP_UB2_UH(vec4, vec5, filt0, filt0, tmp4, tmp5);
     SRARI_H2_UH(tmp4, tmp5, FILTER_BITS);
-    SAT_UH2_UH(tmp4, tmp5, 7);
     PCKEV_ST_SB(tmp4, tmp5, dst + 16);
 
     DOTP_UB2_UH(vec6, vec7, filt0, filt0, tmp6, tmp7);
     SRARI_H2_UH(tmp6, tmp7, FILTER_BITS);
-    SAT_UH2_UH(tmp6, tmp7, 7);
     PCKEV_ST_SB(tmp6, tmp7, dst + 16 + dst_stride);
 
     ILVR_B2_UB(src7, src6, src8, src7, vec0, vec2);
     ILVL_B2_UB(src7, src6, src8, src7, vec1, vec3);
     DOTP_UB2_UH(vec0, vec1, filt0, filt0, tmp0, tmp1);
     SRARI_H2_UH(tmp0, tmp1, FILTER_BITS);
-    SAT_UH2_UH(tmp0, tmp1, 7);
     PCKEV_ST_SB(tmp0, tmp1, dst + 32);
 
     DOTP_UB2_UH(vec2, vec3, filt0, filt0, tmp2, tmp3);
     SRARI_H2_UH(tmp2, tmp3, FILTER_BITS);
-    SAT_UH2_UH(tmp2, tmp3, 7);
     PCKEV_ST_SB(tmp2, tmp3, dst + 32 + dst_stride);
 
     ILVR_B2_UB(src10, src9, src11, src10, vec4, vec6);
     ILVL_B2_UB(src10, src9, src11, src10, vec5, vec7);
     DOTP_UB2_UH(vec4, vec5, filt0, filt0, tmp4, tmp5);
     SRARI_H2_UH(tmp4, tmp5, FILTER_BITS);
-    SAT_UH2_UH(tmp4, tmp5, 7);
     PCKEV_ST_SB(tmp4, tmp5, dst + 48);
 
     DOTP_UB2_UH(vec6, vec7, filt0, filt0, tmp6, tmp7);
     SRARI_H2_UH(tmp6, tmp7, FILTER_BITS);
-    SAT_UH2_UH(tmp6, tmp7, 7);
     PCKEV_ST_SB(tmp6, tmp7, dst + 48 + dst_stride);
     dst += (2 * dst_stride);
 
     src0 = src2;
     src3 = src5;
     src6 = src8;
     src9 = src11;
   }
 }
 
 void vpx_convolve8_vert_msa(const uint8_t *src, ptrdiff_t src_stride,
                             uint8_t *dst, ptrdiff_t dst_stride,
                             const int16_t *filter_x, int x_step_q4,
                             const int16_t *filter_y, int y_step_q4,
                             int w, int h) {
   int8_t cnt, filt_ver[8];
 
-  if (16 != y_step_q4) {
-    vpx_convolve8_vert_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_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(y_step_q4 == 16);
+  assert(((const int32_t *)filter_y)[1] != 0x800000);
 
   for (cnt = 8; cnt--;) {
     filt_ver[cnt] = filter_y[cnt];
   }
 
   if (((const int32_t *)filter_y)[0] == 0) {
     switch (w) {
       case 4:
         common_vt_2t_4w_msa(src, (int32_t)src_stride,
                             dst, (int32_t)dst_stride,
@@ skipping 53 matching lines @@
                              filt_ver, h);
         break;
       default:
         vpx_convolve8_vert_c(src, src_stride, dst, dst_stride,
                              filter_x, x_step_q4, filter_y, y_step_q4,
                              w, h);
         break;
     }
   }
 }