diff options
Diffstat (limited to 'vp8/encoder/x86/sad_sse3.asm')
-rw-r--r-- | vp8/encoder/x86/sad_sse3.asm | 66 |
1 files changed, 66 insertions, 0 deletions
diff --git a/vp8/encoder/x86/sad_sse3.asm b/vp8/encoder/x86/sad_sse3.asm index 2dbcc7dc9..1c41c322a 100644 --- a/vp8/encoder/x86/sad_sse3.asm +++ b/vp8/encoder/x86/sad_sse3.asm @@ -20,6 +20,7 @@ %define ret_var rbx %define result_ptr arg(4) %define max_err arg(4) + %define height dword ptr arg(4) push rbp mov rbp, rsp push rsi @@ -42,6 +43,7 @@ %define ret_var r11 %define result_ptr [rsp+xmm_stack_space+8+4*8] %define max_err [rsp+xmm_stack_space+8+4*8] + %define height [rsp+xmm_stack_space+8+4*8] %else %define src_ptr rdi %define src_stride rsi @@ -51,6 +53,7 @@ %define ret_var r10 %define result_ptr r8 %define max_err r8 + %define height r8 %endif %endif @@ -65,6 +68,7 @@ %define ret_var %define result_ptr %define max_err + %define height %if ABI_IS_32BIT pop rbx @@ -632,6 +636,67 @@ sym(vp8_sad16x16_sse3): STACK_FRAME_DESTROY_X3 +;void vp8_copy32xn_sse3( +; unsigned char *src_ptr, +; int src_stride, +; unsigned char *dst_ptr, +; int dst_stride, +; int height); +global sym(vp8_copy32xn_sse3) +sym(vp8_copy32xn_sse3): + + STACK_FRAME_CREATE_X3 + +block_copy_sse3_loopx4: + lea end_ptr, [src_ptr+src_stride*2] + + movdqu xmm0, XMMWORD PTR [src_ptr] + movdqu xmm1, XMMWORD PTR [src_ptr + 16] + movdqu xmm2, XMMWORD PTR [src_ptr + src_stride] + movdqu xmm3, XMMWORD PTR [src_ptr + src_stride + 16] + movdqu xmm4, XMMWORD PTR [end_ptr] + movdqu xmm5, XMMWORD PTR [end_ptr + 16] + movdqu xmm6, XMMWORD PTR [end_ptr + src_stride] + movdqu xmm7, XMMWORD PTR [end_ptr + src_stride + 16] + + lea src_ptr, [src_ptr+src_stride*4] + + lea end_ptr, [ref_ptr+ref_stride*2] + + movdqa XMMWORD PTR [ref_ptr], xmm0 + movdqa XMMWORD PTR [ref_ptr + 16], xmm1 + movdqa XMMWORD PTR [ref_ptr + ref_stride], xmm2 + movdqa XMMWORD PTR [ref_ptr + ref_stride + 16], xmm3 + movdqa XMMWORD PTR [end_ptr], xmm4 + movdqa XMMWORD PTR [end_ptr + 16], xmm5 + movdqa XMMWORD PTR [end_ptr + ref_stride], xmm6 + movdqa XMMWORD PTR [end_ptr + ref_stride + 16], xmm7 + + lea ref_ptr, [ref_ptr+ref_stride*4] + + sub height, 4 + cmp height, 4 + jge block_copy_sse3_loopx4 + + ;Check to see if there is more rows need to be copied. + cmp height, 0 + je copy_is_done + +block_copy_sse3_loop: + movdqu xmm0, XMMWORD PTR [src_ptr] + movdqu xmm1, XMMWORD PTR [src_ptr + 16] + lea src_ptr, [src_ptr+src_stride] + + movdqa XMMWORD PTR [ref_ptr], xmm0 + movdqa XMMWORD PTR [ref_ptr + 16], xmm1 + lea ref_ptr, [ref_ptr+ref_stride] + + sub height, 1 + jne block_copy_sse3_loop + +copy_is_done: + STACK_FRAME_DESTROY_X3 + ;void vp8_sad16x16x4d_sse3( ; unsigned char *src_ptr, ; int src_stride, @@ -892,3 +957,4 @@ sym(vp8_sad4x4x4d_sse3): STACK_FRAME_DESTROY_X4 + |