Add MSA optimized TransposeWx8_MSA and TransposeUVWx8_MSA functions

include/libyuv/macros_msa.h

 /*
  *  Copyright 2016 The LibYuv 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.
  */
 
@@ skipping 32 matching lines @@
     uint32 val0_m, val1_m;                                             \
     uint64 val_m = 0;                                                  \
     val0_m = LW(psrc_ld_m);                                            \
     val1_m = LW(psrc_ld_m + 4);                                        \
     val_m = (uint64)(val1_m);                             /* NOLINT */ \
     val_m = (uint64)((val_m << 32) & 0xFFFFFFFF00000000); /* NOLINT */ \
     val_m = (uint64)(val_m | (uint64)val0_m);             /* NOLINT */ \
     val_m;                                                             \
   })
 #endif  // (__mips == 64)
+
+#define SW(val, pdst)                                \
+  ({                                                 \
+    uint8_t* pdst_sw_m = (uint8_t*)(pdst);           \
+    uint32_t val_m = (val);                          \
+                                                     \

[fbarchard1, 2016/12/12 19:59:18]

remove blank lines from macros.

[manojkumar.bhosale, 2016/12/13 08:52:22]

Done.

+    asm volatile("sw  %[val_m],  %[pdst_sw_m]  \n\t" \
+                                                     \
+                 : [pdst_sw_m] "=m"(*pdst_sw_m)      \
+                 : [val_m] "r"(val_m));              \
+  })
+
+#if (__mips == 64)
+#define SD(val, pdst)                                \
+  ({                                                 \
+    uint8_t* pdst_sd_m = (uint8_t*)(pdst);           \
+    uint64_t val_m = (val);                          \
+                                                     \
+    asm volatile("sd  %[val_m],  %[pdst_sd_m]  \n\t" \
+                                                     \
+                 : [pdst_sd_m] "=m"(*pdst_sd_m)      \
+                 : [val_m] "r"(val_m));              \
+  })
+#else  // !(__mips == 64)
+#define SD(val, pdst)                                        \
+  ({                                                         \
+    uint8_t* pdst_sd_m = (uint8_t*)(pdst);                   \
+    uint32_t val0_m, val1_m;                                 \
+                                                             \
+    val0_m = (uint32_t)((val)&0x00000000FFFFFFFF);           \
+    val1_m = (uint32_t)(((val) >> 32) & 0x00000000FFFFFFFF); \
+                                                             \
+    SW(val0_m, pdst_sd_m);                                   \
+    SW(val1_m, pdst_sd_m + 4);                               \
+  })
+#endif  // !(__mips == 64)
 #else   // !(__mips_isa_rev >= 6)
 #define LW(psrc)                                      \
   ({                                                  \
     uint8* psrc_lw_m = (uint8*)(psrc); /* NOLINT */   \
     uint32 val_m;                                     \
     asm volatile("ulw  %[val_m],  %[psrc_lw_m]  \n\t" \
                  : [val_m] "=r"(val_m)                \
                  : [psrc_lw_m] "m"(*psrc_lw_m));      \
     val_m;                                            \
   })
@@ skipping 15 matching lines @@
     uint32 val0_m, val1_m;                                             \
     uint64 val_m = 0;                                                  \
     val0_m = LW(psrc_ld_m);                                            \
     val1_m = LW(psrc_ld_m + 4);                                        \
     val_m = (uint64)(val1_m);                             /* NOLINT */ \
     val_m = (uint64)((val_m << 32) & 0xFFFFFFFF00000000); /* NOLINT */ \
     val_m = (uint64)(val_m | (uint64)val0_m);             /* NOLINT */ \
     val_m;                                                             \
   })
 #endif  // (__mips == 64)
+
+#define SW(val, pdst)                                 \
+  ({                                                  \
+    uint8_t* pdst_sw_m = (uint8_t*)(pdst);            \
+    uint32_t val_m = (val);                           \
+                                                      \
+    asm volatile("usw  %[val_m],  %[pdst_sw_m]  \n\t" \

[fbarchard1, 2016/12/12 19:59:18]

remove \t ?  I've never found it necessary.

[manojkumar.bhosale, 2016/12/13 08:52:22]

Done.

+                                                      \
+                 : [pdst_sw_m] "=m"(*pdst_sw_m)       \
+                 : [val_m] "r"(val_m));               \
+  })
+
+#define SD(val, pdst)                                        \
+  ({                                                         \
+    uint8_t* pdst_sd_m = (uint8_t*)(pdst);                   \
+    uint32_t val0_m, val1_m;                                 \
+                                                             \
+    val0_m = (uint32_t)((val)&0x00000000FFFFFFFF);           \
+    val1_m = (uint32_t)(((val) >> 32) & 0x00000000FFFFFFFF); \
+                                                             \
+    SW(val0_m, pdst_sd_m);                                   \
+    SW(val1_m, pdst_sd_m + 4);                               \
+  })
 #endif  // (__mips_isa_rev >= 6)
 
 // TODO(fbarchard): Consider removing __VAR_ARGS versions.
 #define LD_B(RTYPE, psrc) *((RTYPE*)(psrc)) /* NOLINT */
 #define LD_UB(...) LD_B(v16u8, __VA_ARGS__)
 
 #define ST_B(RTYPE, in, pdst) *((RTYPE*)(pdst)) = (in) /* NOLINT */
 #define ST_UB(...) ST_B(v16u8, __VA_ARGS__)
 
 /* Description : Load two vectors with 16 'byte' sized elements
@@ skipping 62 matching lines @@
 #define ILVRL_B2(RTYPE, in0, in1, out0, out1)           \
   {                                                     \
     out0 = (RTYPE)__msa_ilvr_b((v16i8)in0, (v16i8)in1); \
     out1 = (RTYPE)__msa_ilvl_b((v16i8)in0, (v16i8)in1); \
   }
 #define ILVRL_B2_UB(...) ILVRL_B2(v16u8, __VA_ARGS__)
 
 #endif /* !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa) */
 
 #endif  // INCLUDE_LIBYUV_MACROS_MSA_H_