1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/media/libvpx/vp8/common/x86/sad_sse4.asm Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,353 @@
1.4 +;
1.5 +; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
1.6 +;
1.7 +; Use of this source code is governed by a BSD-style license
1.8 +; that can be found in the LICENSE file in the root of the source
1.9 +; tree. An additional intellectual property rights grant can be found
1.10 +; in the file PATENTS. All contributing project authors may
1.11 +; be found in the AUTHORS file in the root of the source tree.
1.12 +;
1.13 +
1.14 +
1.15 +%include "vpx_ports/x86_abi_support.asm"
1.16 +
1.17 +%macro PROCESS_16X2X8 1
1.18 +%if %1
1.19 + movdqa xmm0, XMMWORD PTR [rsi]
1.20 + movq xmm1, MMWORD PTR [rdi]
1.21 + movq xmm3, MMWORD PTR [rdi+8]
1.22 + movq xmm2, MMWORD PTR [rdi+16]
1.23 + punpcklqdq xmm1, xmm3
1.24 + punpcklqdq xmm3, xmm2
1.25 +
1.26 + movdqa xmm2, xmm1
1.27 + mpsadbw xmm1, xmm0, 0x0
1.28 + mpsadbw xmm2, xmm0, 0x5
1.29 +
1.30 + psrldq xmm0, 8
1.31 +
1.32 + movdqa xmm4, xmm3
1.33 + mpsadbw xmm3, xmm0, 0x0
1.34 + mpsadbw xmm4, xmm0, 0x5
1.35 +
1.36 + paddw xmm1, xmm2
1.37 + paddw xmm1, xmm3
1.38 + paddw xmm1, xmm4
1.39 +%else
1.40 + movdqa xmm0, XMMWORD PTR [rsi]
1.41 + movq xmm5, MMWORD PTR [rdi]
1.42 + movq xmm3, MMWORD PTR [rdi+8]
1.43 + movq xmm2, MMWORD PTR [rdi+16]
1.44 + punpcklqdq xmm5, xmm3
1.45 + punpcklqdq xmm3, xmm2
1.46 +
1.47 + movdqa xmm2, xmm5
1.48 + mpsadbw xmm5, xmm0, 0x0
1.49 + mpsadbw xmm2, xmm0, 0x5
1.50 +
1.51 + psrldq xmm0, 8
1.52 +
1.53 + movdqa xmm4, xmm3
1.54 + mpsadbw xmm3, xmm0, 0x0
1.55 + mpsadbw xmm4, xmm0, 0x5
1.56 +
1.57 + paddw xmm5, xmm2
1.58 + paddw xmm5, xmm3
1.59 + paddw xmm5, xmm4
1.60 +
1.61 + paddw xmm1, xmm5
1.62 +%endif
1.63 + movdqa xmm0, XMMWORD PTR [rsi + rax]
1.64 + movq xmm5, MMWORD PTR [rdi+ rdx]
1.65 + movq xmm3, MMWORD PTR [rdi+ rdx+8]
1.66 + movq xmm2, MMWORD PTR [rdi+ rdx+16]
1.67 + punpcklqdq xmm5, xmm3
1.68 + punpcklqdq xmm3, xmm2
1.69 +
1.70 + lea rsi, [rsi+rax*2]
1.71 + lea rdi, [rdi+rdx*2]
1.72 +
1.73 + movdqa xmm2, xmm5
1.74 + mpsadbw xmm5, xmm0, 0x0
1.75 + mpsadbw xmm2, xmm0, 0x5
1.76 +
1.77 + psrldq xmm0, 8
1.78 + movdqa xmm4, xmm3
1.79 + mpsadbw xmm3, xmm0, 0x0
1.80 + mpsadbw xmm4, xmm0, 0x5
1.81 +
1.82 + paddw xmm5, xmm2
1.83 + paddw xmm5, xmm3
1.84 + paddw xmm5, xmm4
1.85 +
1.86 + paddw xmm1, xmm5
1.87 +%endmacro
1.88 +
1.89 +%macro PROCESS_8X2X8 1
1.90 +%if %1
1.91 + movq xmm0, MMWORD PTR [rsi]
1.92 + movq xmm1, MMWORD PTR [rdi]
1.93 + movq xmm3, MMWORD PTR [rdi+8]
1.94 + punpcklqdq xmm1, xmm3
1.95 +
1.96 + movdqa xmm2, xmm1
1.97 + mpsadbw xmm1, xmm0, 0x0
1.98 + mpsadbw xmm2, xmm0, 0x5
1.99 + paddw xmm1, xmm2
1.100 +%else
1.101 + movq xmm0, MMWORD PTR [rsi]
1.102 + movq xmm5, MMWORD PTR [rdi]
1.103 + movq xmm3, MMWORD PTR [rdi+8]
1.104 + punpcklqdq xmm5, xmm3
1.105 +
1.106 + movdqa xmm2, xmm5
1.107 + mpsadbw xmm5, xmm0, 0x0
1.108 + mpsadbw xmm2, xmm0, 0x5
1.109 + paddw xmm5, xmm2
1.110 +
1.111 + paddw xmm1, xmm5
1.112 +%endif
1.113 + movq xmm0, MMWORD PTR [rsi + rax]
1.114 + movq xmm5, MMWORD PTR [rdi+ rdx]
1.115 + movq xmm3, MMWORD PTR [rdi+ rdx+8]
1.116 + punpcklqdq xmm5, xmm3
1.117 +
1.118 + lea rsi, [rsi+rax*2]
1.119 + lea rdi, [rdi+rdx*2]
1.120 +
1.121 + movdqa xmm2, xmm5
1.122 + mpsadbw xmm5, xmm0, 0x0
1.123 + mpsadbw xmm2, xmm0, 0x5
1.124 + paddw xmm5, xmm2
1.125 +
1.126 + paddw xmm1, xmm5
1.127 +%endmacro
1.128 +
1.129 +%macro PROCESS_4X2X8 1
1.130 +%if %1
1.131 + movd xmm0, [rsi]
1.132 + movq xmm1, MMWORD PTR [rdi]
1.133 + movq xmm3, MMWORD PTR [rdi+8]
1.134 + punpcklqdq xmm1, xmm3
1.135 +
1.136 + mpsadbw xmm1, xmm0, 0x0
1.137 +%else
1.138 + movd xmm0, [rsi]
1.139 + movq xmm5, MMWORD PTR [rdi]
1.140 + movq xmm3, MMWORD PTR [rdi+8]
1.141 + punpcklqdq xmm5, xmm3
1.142 +
1.143 + mpsadbw xmm5, xmm0, 0x0
1.144 +
1.145 + paddw xmm1, xmm5
1.146 +%endif
1.147 + movd xmm0, [rsi + rax]
1.148 + movq xmm5, MMWORD PTR [rdi+ rdx]
1.149 + movq xmm3, MMWORD PTR [rdi+ rdx+8]
1.150 + punpcklqdq xmm5, xmm3
1.151 +
1.152 + lea rsi, [rsi+rax*2]
1.153 + lea rdi, [rdi+rdx*2]
1.154 +
1.155 + mpsadbw xmm5, xmm0, 0x0
1.156 +
1.157 + paddw xmm1, xmm5
1.158 +%endmacro
1.159 +
1.160 +
1.161 +;void vp8_sad16x16x8_sse4(
1.162 +; const unsigned char *src_ptr,
1.163 +; int src_stride,
1.164 +; const unsigned char *ref_ptr,
1.165 +; int ref_stride,
1.166 +; unsigned short *sad_array);
1.167 +global sym(vp8_sad16x16x8_sse4) PRIVATE
1.168 +sym(vp8_sad16x16x8_sse4):
1.169 + push rbp
1.170 + mov rbp, rsp
1.171 + SHADOW_ARGS_TO_STACK 5
1.172 + push rsi
1.173 + push rdi
1.174 + ; end prolog
1.175 +
1.176 + mov rsi, arg(0) ;src_ptr
1.177 + mov rdi, arg(2) ;ref_ptr
1.178 +
1.179 + movsxd rax, dword ptr arg(1) ;src_stride
1.180 + movsxd rdx, dword ptr arg(3) ;ref_stride
1.181 +
1.182 + PROCESS_16X2X8 1
1.183 + PROCESS_16X2X8 0
1.184 + PROCESS_16X2X8 0
1.185 + PROCESS_16X2X8 0
1.186 + PROCESS_16X2X8 0
1.187 + PROCESS_16X2X8 0
1.188 + PROCESS_16X2X8 0
1.189 + PROCESS_16X2X8 0
1.190 +
1.191 + mov rdi, arg(4) ;Results
1.192 + movdqa XMMWORD PTR [rdi], xmm1
1.193 +
1.194 + ; begin epilog
1.195 + pop rdi
1.196 + pop rsi
1.197 + UNSHADOW_ARGS
1.198 + pop rbp
1.199 + ret
1.200 +
1.201 +
1.202 +;void vp8_sad16x8x8_sse4(
1.203 +; const unsigned char *src_ptr,
1.204 +; int src_stride,
1.205 +; const unsigned char *ref_ptr,
1.206 +; int ref_stride,
1.207 +; unsigned short *sad_array
1.208 +;);
1.209 +global sym(vp8_sad16x8x8_sse4) PRIVATE
1.210 +sym(vp8_sad16x8x8_sse4):
1.211 + push rbp
1.212 + mov rbp, rsp
1.213 + SHADOW_ARGS_TO_STACK 5
1.214 + push rsi
1.215 + push rdi
1.216 + ; end prolog
1.217 +
1.218 + mov rsi, arg(0) ;src_ptr
1.219 + mov rdi, arg(2) ;ref_ptr
1.220 +
1.221 + movsxd rax, dword ptr arg(1) ;src_stride
1.222 + movsxd rdx, dword ptr arg(3) ;ref_stride
1.223 +
1.224 + PROCESS_16X2X8 1
1.225 + PROCESS_16X2X8 0
1.226 + PROCESS_16X2X8 0
1.227 + PROCESS_16X2X8 0
1.228 +
1.229 + mov rdi, arg(4) ;Results
1.230 + movdqa XMMWORD PTR [rdi], xmm1
1.231 +
1.232 + ; begin epilog
1.233 + pop rdi
1.234 + pop rsi
1.235 + UNSHADOW_ARGS
1.236 + pop rbp
1.237 + ret
1.238 +
1.239 +
1.240 +;void vp8_sad8x8x8_sse4(
1.241 +; const unsigned char *src_ptr,
1.242 +; int src_stride,
1.243 +; const unsigned char *ref_ptr,
1.244 +; int ref_stride,
1.245 +; unsigned short *sad_array
1.246 +;);
1.247 +global sym(vp8_sad8x8x8_sse4) PRIVATE
1.248 +sym(vp8_sad8x8x8_sse4):
1.249 + push rbp
1.250 + mov rbp, rsp
1.251 + SHADOW_ARGS_TO_STACK 5
1.252 + push rsi
1.253 + push rdi
1.254 + ; end prolog
1.255 +
1.256 + mov rsi, arg(0) ;src_ptr
1.257 + mov rdi, arg(2) ;ref_ptr
1.258 +
1.259 + movsxd rax, dword ptr arg(1) ;src_stride
1.260 + movsxd rdx, dword ptr arg(3) ;ref_stride
1.261 +
1.262 + PROCESS_8X2X8 1
1.263 + PROCESS_8X2X8 0
1.264 + PROCESS_8X2X8 0
1.265 + PROCESS_8X2X8 0
1.266 +
1.267 + mov rdi, arg(4) ;Results
1.268 + movdqa XMMWORD PTR [rdi], xmm1
1.269 +
1.270 + ; begin epilog
1.271 + pop rdi
1.272 + pop rsi
1.273 + UNSHADOW_ARGS
1.274 + pop rbp
1.275 + ret
1.276 +
1.277 +
1.278 +;void vp8_sad8x16x8_sse4(
1.279 +; const unsigned char *src_ptr,
1.280 +; int src_stride,
1.281 +; const unsigned char *ref_ptr,
1.282 +; int ref_stride,
1.283 +; unsigned short *sad_array
1.284 +;);
1.285 +global sym(vp8_sad8x16x8_sse4) PRIVATE
1.286 +sym(vp8_sad8x16x8_sse4):
1.287 + push rbp
1.288 + mov rbp, rsp
1.289 + SHADOW_ARGS_TO_STACK 5
1.290 + push rsi
1.291 + push rdi
1.292 + ; end prolog
1.293 +
1.294 + mov rsi, arg(0) ;src_ptr
1.295 + mov rdi, arg(2) ;ref_ptr
1.296 +
1.297 + movsxd rax, dword ptr arg(1) ;src_stride
1.298 + movsxd rdx, dword ptr arg(3) ;ref_stride
1.299 +
1.300 + PROCESS_8X2X8 1
1.301 + PROCESS_8X2X8 0
1.302 + PROCESS_8X2X8 0
1.303 + PROCESS_8X2X8 0
1.304 + PROCESS_8X2X8 0
1.305 + PROCESS_8X2X8 0
1.306 + PROCESS_8X2X8 0
1.307 + PROCESS_8X2X8 0
1.308 + mov rdi, arg(4) ;Results
1.309 + movdqa XMMWORD PTR [rdi], xmm1
1.310 +
1.311 + ; begin epilog
1.312 + pop rdi
1.313 + pop rsi
1.314 + UNSHADOW_ARGS
1.315 + pop rbp
1.316 + ret
1.317 +
1.318 +
1.319 +;void vp8_sad4x4x8_c(
1.320 +; const unsigned char *src_ptr,
1.321 +; int src_stride,
1.322 +; const unsigned char *ref_ptr,
1.323 +; int ref_stride,
1.324 +; unsigned short *sad_array
1.325 +;);
1.326 +global sym(vp8_sad4x4x8_sse4) PRIVATE
1.327 +sym(vp8_sad4x4x8_sse4):
1.328 + push rbp
1.329 + mov rbp, rsp
1.330 + SHADOW_ARGS_TO_STACK 5
1.331 + push rsi
1.332 + push rdi
1.333 + ; end prolog
1.334 +
1.335 + mov rsi, arg(0) ;src_ptr
1.336 + mov rdi, arg(2) ;ref_ptr
1.337 +
1.338 + movsxd rax, dword ptr arg(1) ;src_stride
1.339 + movsxd rdx, dword ptr arg(3) ;ref_stride
1.340 +
1.341 + PROCESS_4X2X8 1
1.342 + PROCESS_4X2X8 0
1.343 +
1.344 + mov rdi, arg(4) ;Results
1.345 + movdqa XMMWORD PTR [rdi], xmm1
1.346 +
1.347 + ; begin epilog
1.348 + pop rdi
1.349 + pop rsi
1.350 + UNSHADOW_ARGS
1.351 + pop rbp
1.352 + ret
1.353 +
1.354 +
1.355 +
1.356 +
