libvpx: Pull from upstream

source/libvpx/vpx_dsp/mips/vpx_convolve8_avg_horiz_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_hz_8t_and_aver_dst_4x4_msa(const uint8_t *src,
                                               int32_t src_stride,
                                               uint8_t *dst,
                                               int32_t dst_stride,
                                               int8_t *filter) {
   v16i8 src0, src1, src2, src3, filt0, filt1, filt2, filt3;
   v16u8 dst0, dst1, dst2, dst3, res2, res3;
@@ skipping 295 matching lines @@
   }
 }
 
 static void common_hz_2t_and_aver_dst_4x4_msa(const uint8_t *src,
                                               int32_t src_stride,
                                               uint8_t *dst,
                                               int32_t dst_stride,
                                               int8_t *filter) {
   v16i8 src0, src1, src2, src3, mask;
   v16u8 filt0, dst0, dst1, dst2, dst3, vec0, vec1, res0, res1;
-  v8u16 vec2, vec3, const255, filt;
+  v8u16 vec2, vec3, filt;
 
   mask = LD_SB(&mc_filt_mask_arr[16]);
 
   /* rearranging filter */
   filt = LD_UH(filter);
   filt0 = (v16u8)__msa_splati_h((v8i16)filt, 0);
 
-  const255 = (v8u16)__msa_ldi_h(255);
-
   LD_SB4(src, src_stride, src0, src1, src2, src3);
   LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
   VSHF_B2_UB(src0, src1, src2, src3, mask, mask, vec0, vec1);
   DOTP_UB2_UH(vec0, vec1, filt0, filt0, vec2, vec3);
   SRARI_H2_UH(vec2, vec3, FILTER_BITS);
-  MIN_UH2_UH(vec2, vec3, const255);
   PCKEV_B2_UB(vec2, vec2, vec3, vec3, res0, res1);
   ILVR_W2_UB(dst1, dst0, dst3, dst2, dst0, dst2);
   AVER_UB2_UB(res0, dst0, res1, dst2, res0, res1);
   ST4x4_UB(res0, res1, 0, 1, 0, 1, dst, dst_stride);
 }
 
 static void common_hz_2t_and_aver_dst_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, mask;
   v16u8 filt0, vec0, vec1, vec2, vec3, res0, res1, res2, res3;
   v16u8 dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7;
-  v8u16 vec4, vec5, vec6, vec7, const255, filt;
+  v8u16 vec4, vec5, vec6, vec7, filt;
 
   mask = LD_SB(&mc_filt_mask_arr[16]);
 
   /* rearranging filter */
   filt = LD_UH(filter);
   filt0 = (v16u8)__msa_splati_h((v8i16)filt, 0);
 
-  const255 = (v8u16)__msa_ldi_h(255);
-
   LD_SB8(src, src_stride, src0, src1, src2, src3, src4, src5, src6, src7);
   LD_UB8(dst, dst_stride, dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7);
   VSHF_B2_UB(src0, src1, src2, src3, mask, mask, vec0, vec1);
   VSHF_B2_UB(src4, src5, src6, src7, mask, mask, vec2, vec3);
   DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, vec4, vec5,
               vec6, vec7);
   SRARI_H4_UH(vec4, vec5, vec6, vec7, FILTER_BITS);
-  MIN_UH4_UH(vec4, vec5, vec6, vec7, const255);
   PCKEV_B4_UB(vec4, vec4, vec5, vec5, vec6, vec6, vec7, vec7, res0, res1, res2,
               res3);
   ILVR_W4_UB(dst1, dst0, dst3, dst2, dst5, dst4, dst7, dst6, dst0, dst2, dst4,
              dst6);
   AVER_UB4_UB(res0, dst0, res1, dst2, res2, dst4, res3, dst6, 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);
 }
@@ skipping 11 matching lines @@
   }
 }
 
 static void common_hz_2t_and_aver_dst_8x4_msa(const uint8_t *src,
                                               int32_t src_stride,
                                               uint8_t *dst,
                                               int32_t dst_stride,
                                               int8_t *filter) {
   v16i8 src0, src1, src2, src3, mask;
   v16u8 filt0, dst0, dst1, dst2, dst3;
-  v8u16 vec0, vec1, vec2, vec3, const255, filt;
+  v8u16 vec0, vec1, vec2, vec3, filt;
 
   mask = LD_SB(&mc_filt_mask_arr[0]);
 
   /* rearranging filter */
   filt = LD_UH(filter);
   filt0 = (v16u8)__msa_splati_h((v8i16)filt, 0);
 
-  const255 = (v8u16)__msa_ldi_h(255);
-
   LD_SB4(src, src_stride, src0, src1, src2, src3);
   VSHF_B2_UH(src0, src0, src1, src1, mask, mask, vec0, vec1);
   VSHF_B2_UH(src2, src2, src3, src3, mask, mask, vec2, vec3);
   DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, vec0, vec1,
               vec2, vec3);
   SRARI_H4_UH(vec0, vec1, vec2, vec3, FILTER_BITS);
   LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
-  MIN_UH4_UH(vec0, vec1, vec2, vec3, const255);
   PCKEV_AVG_ST8x4_UB(vec0, dst0, vec1, dst1, vec2, dst2, vec3, dst3,
                      dst, dst_stride);
 }
 
 static void common_hz_2t_and_aver_dst_8x8mult_msa(const uint8_t *src,
                                                   int32_t src_stride,
                                                   uint8_t *dst,
                                                   int32_t dst_stride,
                                                   int8_t *filter,
                                                   int32_t height) {
   v16i8 src0, src1, src2, src3, mask;
   v16u8 filt0, dst0, dst1, dst2, dst3;
-  v8u16 vec0, vec1, vec2, vec3, const255, filt;
+  v8u16 vec0, vec1, vec2, vec3, filt;
 
   mask = LD_SB(&mc_filt_mask_arr[0]);
 
   /* rearranging filter */
   filt = LD_UH(filter);
   filt0 = (v16u8)__msa_splati_h((v8i16)filt, 0);
 
-  const255 = (v8u16)__msa_ldi_h(255);
-
   LD_SB4(src, src_stride, src0, src1, src2, src3);
   src += (4 * src_stride);
   VSHF_B2_UH(src0, src0, src1, src1, mask, mask, vec0, vec1);
   VSHF_B2_UH(src2, src2, src3, src3, mask, mask, vec2, vec3);
   DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, vec0, vec1,
               vec2, vec3);
   SRARI_H4_UH(vec0, vec1, vec2, vec3, FILTER_BITS);
   LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
-  MIN_UH4_UH(vec0, vec1, vec2, vec3, const255);
   LD_SB4(src, src_stride, src0, src1, src2, src3);
   src += (4 * src_stride);
   PCKEV_AVG_ST8x4_UB(vec0, dst0, vec1, dst1, vec2, dst2, vec3, dst3,
                      dst, dst_stride);
   dst += (4 * dst_stride);
 
   VSHF_B2_UH(src0, src0, src1, src1, mask, mask, vec0, vec1);
   VSHF_B2_UH(src2, src2, src3, src3, mask, mask, vec2, vec3);
   DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, vec0, vec1,
               vec2, vec3);
   SRARI_H4_UH(vec0, vec1, vec2, vec3, FILTER_BITS);
   LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
-  MIN_UH4_UH(vec0, vec1, vec2, vec3, const255);
   PCKEV_AVG_ST8x4_UB(vec0, dst0, vec1, dst1, vec2, dst2, vec3, dst3,
                      dst, dst_stride);
   dst += (4 * dst_stride);
 
   if (16 == height) {
     LD_SB4(src, src_stride, src0, src1, src2, src3);
     src += (4 * src_stride);
 
     VSHF_B2_UH(src0, src0, src1, src1, mask, mask, vec0, vec1);
     VSHF_B2_UH(src2, src2, src3, src3, mask, mask, vec2, vec3);
     DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, vec0, vec1,
                 vec2, vec3);
     SRARI_H4_UH(vec0, vec1, vec2, vec3, FILTER_BITS);
     LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
-    MIN_UH4_UH(vec0, vec1, vec2, vec3, const255);
     LD_SB4(src, src_stride, src0, src1, src2, src3);
     PCKEV_AVG_ST8x4_UB(vec0, dst0, vec1, dst1, vec2, dst2, vec3, dst3,
                        dst, dst_stride);
     dst += (4 * dst_stride);
 
     VSHF_B2_UH(src0, src0, src1, src1, mask, mask, vec0, vec1);
     VSHF_B2_UH(src2, src2, src3, src3, mask, mask, vec2, vec3);
     DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, vec0, vec1,
                 vec2, vec3);
     SRARI_H4_UH(vec0, vec1, vec2, vec3, FILTER_BITS);
     LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
-    MIN_UH4_UH(vec0, vec1, vec2, vec3, const255);
     PCKEV_AVG_ST8x4_UB(vec0, dst0, vec1, dst1, vec2, dst2, vec3, dst3,
                        dst, dst_stride);
   }
 }
 
 static void common_hz_2t_and_aver_dst_8w_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_hz_2t_and_aver_dst_8x4_msa(src, src_stride, dst, dst_stride, filter);
   } else {
     common_hz_2t_and_aver_dst_8x8mult_msa(src, src_stride, dst, dst_stride,
                                           filter, height);
   }
 }
 
 static void common_hz_2t_and_aver_dst_16w_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, mask;
   v16u8 filt0, dst0, dst1, dst2, dst3;
   v16u8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
-  v8u16 res0, res1, res2, res3, res4, res5, res6, res7, const255, filt;
+  v8u16 res0, res1, res2, res3, res4, res5, res6, res7, filt;
 
   mask = LD_SB(&mc_filt_mask_arr[0]);
 
   /* rearranging filter */
   filt = LD_UH(filter);
   filt0 = (v16u8)__msa_splati_h((v8i16)filt, 0);
 
-  const255 = (v8u16)__msa_ldi_h(255);
-
   LD_SB4(src, src_stride, src0, src2, src4, src6);
   LD_SB4(src + 8, src_stride, src1, src3, src5, src7);
   src += (4 * src_stride);
 
   VSHF_B2_UB(src0, src0, src1, src1, mask, mask, vec0, vec1);
   VSHF_B2_UB(src2, src2, src3, src3, mask, mask, vec2, vec3);
   VSHF_B2_UB(src4, src4, src5, src5, mask, mask, vec4, vec5);
   VSHF_B2_UB(src6, src6, src7, src7, mask, mask, vec6, vec7);
   DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, res0, res1,
               res2, res3);
   DOTP_UB4_UH(vec4, vec5, vec6, vec7, filt0, filt0, filt0, filt0, res4, res5,
               res6, res7);
   SRARI_H4_UH(res0, res1, res2, res3, FILTER_BITS);
   SRARI_H4_UH(res4, res5, res6, res7, FILTER_BITS);
   LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
-  MIN_UH4_UH(res0, res1, res2, res3, const255);
-  MIN_UH4_UH(res4, res5, res6, res7, const255);
   PCKEV_AVG_ST_UB(res1, res0, dst0, dst);
   dst += dst_stride;
   PCKEV_AVG_ST_UB(res3, res2, dst1, dst);
   dst += dst_stride;
   PCKEV_AVG_ST_UB(res5, res4, dst2, dst);
   dst += dst_stride;
   PCKEV_AVG_ST_UB(res7, res6, dst3, dst);
   dst += dst_stride;
 
   for (loop_cnt = (height >> 2) - 1; 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);
 
     VSHF_B2_UB(src0, src0, src1, src1, mask, mask, vec0, vec1);
     VSHF_B2_UB(src2, src2, src3, src3, mask, mask, vec2, vec3);
     VSHF_B2_UB(src4, src4, src5, src5, mask, mask, vec4, vec5);
     VSHF_B2_UB(src6, src6, src7, src7, mask, mask, vec6, vec7);
     DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, res0, res1,
                 res2, res3);
     DOTP_UB4_UH(vec4, vec5, vec6, vec7, filt0, filt0, filt0, filt0, res4, res5,
                 res6, res7);
     SRARI_H4_UH(res0, res1, res2, res3, FILTER_BITS);
     SRARI_H4_UH(res4, res5, res6, res7, FILTER_BITS);
     LD_UB4(dst, dst_stride, dst0, dst1, dst2, dst3);
-    MIN_UH4_UH(res0, res1, res2, res3, const255);
-    MIN_UH4_UH(res4, res5, res6, res7, const255);
     PCKEV_AVG_ST_UB(res1, res0, dst0, dst);
     dst += dst_stride;
     PCKEV_AVG_ST_UB(res3, res2, dst1, dst);
     dst += dst_stride;
     PCKEV_AVG_ST_UB(res5, res4, dst2, dst);
     dst += dst_stride;
     PCKEV_AVG_ST_UB(res7, res6, dst3, dst);
     dst += dst_stride;
   }
 }
 
 static void common_hz_2t_and_aver_dst_32w_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, mask;
   v16u8 filt0, dst0, dst1, dst2, dst3;
   v16u8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
-  v8u16 res0, res1, res2, res3, res4, res5, res6, res7, const255, filt;
+  v8u16 res0, res1, res2, res3, res4, res5, res6, res7, filt;
 
   mask = LD_SB(&mc_filt_mask_arr[0]);
 
   /* rearranging filter */
   filt = LD_UH(filter);
   filt0 = (v16u8)__msa_splati_h((v8i16)filt, 0);
 
-  const255 = (v8u16)__msa_ldi_h(255);
-
   for (loop_cnt = (height >> 1); loop_cnt--;) {
     src0 = LD_SB(src);
     src2 = LD_SB(src + 16);
     src3 = LD_SB(src + 24);
     src1 = __msa_sldi_b(src2, src0, 8);
     src += src_stride;
     src4 = LD_SB(src);
     src6 = LD_SB(src + 16);
     src7 = LD_SB(src + 24);
     src5 = __msa_sldi_b(src6, src4, 8);
     src += src_stride;
 
     VSHF_B2_UB(src0, src0, src1, src1, mask, mask, vec0, vec1);
     VSHF_B2_UB(src2, src2, src3, src3, mask, mask, vec2, vec3);
     VSHF_B2_UB(src4, src4, src5, src5, mask, mask, vec4, vec5);
     VSHF_B2_UB(src6, src6, src7, src7, mask, mask, vec6, vec7);
     DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, res0, res1,
                 res2, res3);
     DOTP_UB4_UH(vec4, vec5, vec6, vec7, filt0, filt0, filt0, filt0, res4, res5,
                 res6, res7);
     SRARI_H4_UH(res0, res1, res2, res3, FILTER_BITS);
     SRARI_H4_UH(res4, res5, res6, res7, FILTER_BITS);
-    MIN_UH4_UH(res0, res1, res2, res3, const255);
-    MIN_UH4_UH(res4, res5, res6, res7, const255);
     LD_UB2(dst, 16, dst0, dst1);
     PCKEV_AVG_ST_UB(res1, res0, dst0, dst);
     PCKEV_AVG_ST_UB(res3, res2, dst1, (dst + 16));
     dst += dst_stride;
     LD_UB2(dst, 16, dst2, dst3);
     PCKEV_AVG_ST_UB(res5, res4, dst2, dst);
     PCKEV_AVG_ST_UB(res7, res6, dst3, (dst + 16));
     dst += dst_stride;
   }
 }
 
 static void common_hz_2t_and_aver_dst_64w_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, mask;
   v16u8 filt0, dst0, dst1, dst2, dst3;
   v16u8 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
-  v8u16 out0, out1, out2, out3, out4, out5, out6, out7, const255, filt;
+  v8u16 out0, out1, out2, out3, out4, out5, out6, out7, filt;
 
   mask = LD_SB(&mc_filt_mask_arr[0]);
 
   /* rearranging filter */
   filt = LD_UH(filter);
   filt0 = (v16u8)__msa_splati_h((v8i16)filt, 0);
 
-  const255 = (v8u16)__msa_ldi_h(255);
-
   for (loop_cnt = height; loop_cnt--;) {
     LD_SB4(src, 16, src0, src2, src4, src6);
     src7 = LD_SB(src + 56);
     SLDI_B3_SB(src2, src4, src6, src0, src2, src4, src1, src3, src5, 8);
     src += src_stride;
 
     VSHF_B2_UB(src0, src0, src1, src1, mask, mask, vec0, vec1);
     VSHF_B2_UB(src2, src2, src3, src3, mask, mask, vec2, vec3);
     VSHF_B2_UB(src4, src4, src5, src5, mask, mask, vec4, vec5);
     VSHF_B2_UB(src6, src6, src7, src7, mask, mask, vec6, vec7);
     DOTP_UB4_UH(vec0, vec1, vec2, vec3, filt0, filt0, filt0, filt0, out0, out1,
                 out2, out3);
     DOTP_UB4_UH(vec4, vec5, vec6, vec7, filt0, filt0, filt0, filt0, out4, out5,
                 out6, out7);
     SRARI_H4_UH(out0, out1, out2, out3, FILTER_BITS);
     SRARI_H4_UH(out4, out5, out6, out7, FILTER_BITS);
     LD_UB4(dst, 16, dst0, dst1, dst2, dst3);
-    MIN_UH4_UH(out0, out1, out2, out3, const255);
-    MIN_UH4_UH(out4, out5, out6, out7, const255);
     PCKEV_AVG_ST_UB(out1, out0, dst0, dst);
     PCKEV_AVG_ST_UB(out3, out2, dst1, dst + 16);
     PCKEV_AVG_ST_UB(out5, out4, dst2, dst + 32);
     PCKEV_AVG_ST_UB(out7, out6, dst3, dst + 48);
     dst += dst_stride;
   }
 }
 
 void vpx_convolve8_avg_horiz_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_hor[8];
 
-  if (16 != x_step_q4) {
-    vpx_convolve8_avg_horiz_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) {
-    vpx_convolve_avg(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(((const int32_t *)filter_x)[1] != 0x800000);
 
   for (cnt = 0; cnt < 8; ++cnt) {
     filt_hor[cnt] = filter_x[cnt];
   }
 
   if (((const int32_t *)filter_x)[0] == 0) {
     switch (w) {
       case 4:
         common_hz_2t_and_aver_dst_4w_msa(src, (int32_t)src_stride,
                                          dst, (int32_t)dst_stride,
@@ skipping 53 matching lines @@
                                           filt_hor, h);
         break;
       default:
         vpx_convolve8_avg_horiz_c(src, src_stride, dst, dst_stride,
                                   filter_x, x_step_q4, filter_y, y_step_q4,
                                   w, h);
         break;
     }
   }
 }