Updated vp8_build_intra_predictors_mby_s(sse2/ssse3)

to work with the latest code.

Patch Set 2: aligned the above_row buffers to fix crash

Change-Id: I7a6992a20ed079ccd302f8c26215cf3057f8b70c

5 files changed, 118 insertions, 82 deletions

diff --git a/vp8/common/reconintra.c b/vp8/common/reconintra.c
index f1cd12fc1..4067a6851 100644
--- a/vp8/common/reconintra.c
+++ b/vp8/common/reconintra.c
@@ -14,7 +14,7 @@
 #include "vpx_mem/vpx_mem.h"
 #include "blockd.h"
 
-void vp8_build_intra_predictors_mby_s(MACROBLOCKD *x,
+void vp8_build_intra_predictors_mby_s_c(MACROBLOCKD *x,
                                           unsigned char * yabove_row,
                                           unsigned char * yleft,
                                           int left_stride,
diff --git a/vp8/common/rtcd_defs.sh b/vp8/common/rtcd_defs.sh
index c069a21f1..ab9951528 100644
--- a/vp8/common/rtcd_defs.sh
+++ b/vp8/common/rtcd_defs.sh
@@ -123,7 +123,8 @@ specialize vp8_copy_mem8x4 mmx media neon
 vp8_copy_mem8x4_media=vp8_copy_mem8x4_v6
 
 prototype void vp8_build_intra_predictors_mby_s "struct macroblockd *x, unsigned char * yabove_row, unsigned char * yleft, int left_stride, unsigned char * ypred_ptr, int y_stride"
-#TODO: fix assembly --- specialize vp8_build_intra_predictors_mby_s sse2 ssse3 neon
+specialize vp8_build_intra_predictors_mby_s sse2 ssse3
+#TODO: fix assembly for neon
 
 prototype void vp8_build_intra_predictors_mbuv_s "struct macroblockd *x, unsigned char * uabove_row, unsigned char * vabove_row,  unsigned char *uleft, unsigned char *vleft, int left_stride, unsigned char * upred_ptr, unsigned char * vpred_ptr, int pred_stride"
 specialize vp8_build_intra_predictors_mbuv_s sse2 ssse3
diff --git a/vp8/common/x86/recon_sse2.asm b/vp8/common/x86/recon_sse2.asm
index d371ebd74..7b6e3cffe 100644
--- a/vp8/common/x86/recon_sse2.asm
+++ b/vp8/common/x86/recon_sse2.asm
@@ -133,22 +133,20 @@ sym(vp8_intra_pred_uv_dc_mmx2):
     ; end prolog
 
     ; from top
-    mov         rsi,        arg(2) ;above;
-    pxor        mm0,        mm0
-    movq        mm1,        [rsi]
-    psadbw      mm1,        mm0
-
-    ; from left
+    mov         rdi,        arg(2) ;above;
     mov         rsi,        arg(3) ;left;
     movsxd      rax,        dword ptr arg(4) ;left_stride;
+    pxor        mm0,        mm0
+    movq        mm1,        [rdi]
     lea         rdi,        [rax*3]
+    psadbw      mm1,        mm0
+    ; from left
     movzx       ecx,        byte [rsi]
     movzx       edx,        byte [rsi+rax*1]
     add         ecx,        edx
     movzx       edx,        byte [rsi+rax*2]
     add         ecx,        edx
 
-
     movzx       edx,        byte [rsi+rdi]
     lea         rsi,        [rsi+rax*4]
     add         ecx,        edx
@@ -166,23 +164,23 @@ sym(vp8_intra_pred_uv_dc_mmx2):
     lea         edx,        [edx+ecx+8]
     sar         edx,        4
     movd        mm1,        edx
+    movsxd      rcx,        dword ptr arg(1) ;dst_stride
     pshufw      mm1,        mm1, 0x0
+    mov         rdi,        arg(0) ;dst;
     packuswb    mm1,        mm1
 
     ; write out
-    mov         rdi,        arg(0) ;dst;
-    movsxd      rcx,        dword ptr arg(1) ;dst_stride
     lea         rax,        [rcx*3]
+    lea         rdx,        [rdi+rcx*4]
 
     movq [rdi      ],       mm1
     movq [rdi+rcx  ],       mm1
     movq [rdi+rcx*2],       mm1
     movq [rdi+rax  ],       mm1
-    lea         rdi,        [rdi+rcx*4]
-    movq [rdi      ],       mm1
-    movq [rdi+rcx  ],       mm1
-    movq [rdi+rcx*2],       mm1
-    movq [rdi+rax  ],       mm1
+    movq [rdx      ],       mm1
+    movq [rdx+rcx  ],       mm1
+    movq [rdx+rcx*2],       mm1
+    movq [rdx+rax  ],       mm1
 
     ; begin epilog
     pop         rdi
@@ -478,7 +476,7 @@ sym(vp8_intra_pred_uv_ve_mmx):
 ;    int dst_stride
 ;    unsigned char *above,
 ;    unsigned char *left,
-;    int left_stride,
+;    int left_stride
 ;    )
 %macro vp8_intra_pred_uv_ho 1
 global sym(vp8_intra_pred_uv_ho_%1)
@@ -575,38 +573,43 @@ vp8_intra_pred_uv_ho ssse3
 ;void vp8_intra_pred_y_dc_sse2(
 ;    unsigned char *dst,
 ;    int dst_stride
-;    unsigned char *src,
-;    int src_stride,
+;    unsigned char *above,
+;    unsigned char *left,
+;    int left_stride
 ;    )
 global sym(vp8_intra_pred_y_dc_sse2)
 sym(vp8_intra_pred_y_dc_sse2):
     push        rbp
     mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 4
+    SHADOW_ARGS_TO_STACK 5
     push        rsi
     push        rdi
     ; end prolog
 
     ; from top
-    mov         rsi,        arg(2) ;src;
-    movsxd      rax,        dword ptr arg(3) ;src_stride;
-    sub         rsi,        rax
+    mov         rdi,        arg(2) ;above
+    mov         rsi,        arg(3) ;left
+    movsxd      rax,        dword ptr arg(4) ;left_stride;
+
     pxor        xmm0,       xmm0
-    movdqa      xmm1,       [rsi]
+    movdqa      xmm1,       [rdi]
     psadbw      xmm1,       xmm0
     movq        xmm2,       xmm1
     punpckhqdq  xmm1,       xmm1
     paddw       xmm1,       xmm2
 
     ; from left
-    dec         rsi
     lea         rdi,        [rax*3]
-    movzx       ecx,        byte [rsi+rax]
+
+    movzx       ecx,        byte [rsi]
+    movzx       edx,        byte [rsi+rax]
+    add         ecx,        edx
     movzx       edx,        byte [rsi+rax*2]
     add         ecx,        edx
     movzx       edx,        byte [rsi+rdi]
     add         ecx,        edx
     lea         rsi,        [rsi+rax*4]
+
     movzx       edx,        byte [rsi]
     add         ecx,        edx
     movzx       edx,        byte [rsi+rax]
@@ -616,6 +619,7 @@ sym(vp8_intra_pred_y_dc_sse2):
     movzx       edx,        byte [rsi+rdi]
     add         ecx,        edx
     lea         rsi,        [rsi+rax*4]
+
     movzx       edx,        byte [rsi]
     add         ecx,        edx
     movzx       edx,        byte [rsi+rax]
@@ -625,6 +629,7 @@ sym(vp8_intra_pred_y_dc_sse2):
     movzx       edx,        byte [rsi+rdi]
     add         ecx,        edx
     lea         rsi,        [rsi+rax*4]
+
     movzx       edx,        byte [rsi]
     add         ecx,        edx
     movzx       edx,        byte [rsi+rax]
@@ -633,8 +638,6 @@ sym(vp8_intra_pred_y_dc_sse2):
     add         ecx,        edx
     movzx       edx,        byte [rsi+rdi]
     add         ecx,        edx
-    movzx       edx,        byte [rsi+rax*4]
-    add         ecx,        edx
 
     ; add up
     pextrw      edx,        xmm1, 0x0
@@ -676,22 +679,23 @@ sym(vp8_intra_pred_y_dc_sse2):
 ;void vp8_intra_pred_y_dctop_sse2(
 ;    unsigned char *dst,
 ;    int dst_stride
-;    unsigned char *src,
-;    int src_stride,
+;    unsigned char *above,
+;    unsigned char *left,
+;    int left_stride
 ;    )
 global sym(vp8_intra_pred_y_dctop_sse2)
 sym(vp8_intra_pred_y_dctop_sse2):
     push        rbp
     mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 4
+    SHADOW_ARGS_TO_STACK 5
     push        rsi
     GET_GOT     rbx
     ; end prolog
 
+    ;arg(3), arg(4) not used
+
     ; from top
-    mov         rcx,        arg(2) ;src;
-    movsxd      rax,        dword ptr arg(3) ;src_stride;
-    sub         rcx,        rax
+    mov         rcx,        arg(2) ;above;
     pxor        xmm0,       xmm0
     movdqa      xmm1,       [rcx]
     psadbw      xmm1,       xmm0
@@ -737,22 +741,25 @@ sym(vp8_intra_pred_y_dctop_sse2):
 ;void vp8_intra_pred_y_dcleft_sse2(
 ;    unsigned char *dst,
 ;    int dst_stride
-;    unsigned char *src,
-;    int src_stride,
+;    unsigned char *above,
+;    unsigned char *left,
+;    int left_stride
 ;    )
 global sym(vp8_intra_pred_y_dcleft_sse2)
 sym(vp8_intra_pred_y_dcleft_sse2):
     push        rbp
     mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 4
+    SHADOW_ARGS_TO_STACK 5
     push        rsi
     push        rdi
     ; end prolog
 
+    ;arg(2) not used
+
     ; from left
-    mov         rsi,        arg(2) ;src;
-    movsxd      rax,        dword ptr arg(3) ;src_stride;
-    dec         rsi
+    mov         rsi,        arg(3) ;left;
+    movsxd      rax,        dword ptr arg(4) ;left_stride;
+
     lea         rdi,        [rax*3]
     movzx       ecx,        byte [rsi]
     movzx       edx,        byte [rsi+rax]
@@ -827,18 +834,21 @@ sym(vp8_intra_pred_y_dcleft_sse2):
 ;void vp8_intra_pred_y_dc128_sse2(
 ;    unsigned char *dst,
 ;    int dst_stride
-;    unsigned char *src,
-;    int src_stride,
+;    unsigned char *above,
+;    unsigned char *left,
+;    int left_stride
 ;    )
 global sym(vp8_intra_pred_y_dc128_sse2)
 sym(vp8_intra_pred_y_dc128_sse2):
     push        rbp
     mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 4
+    SHADOW_ARGS_TO_STACK 5
     push        rsi
     GET_GOT     rbx
     ; end prolog
 
+    ;arg(2), arg(3), arg(4) not used
+
     ; write out
     mov         rsi,        2
     movdqa      xmm1,       [GLOBAL(dc_128)]
@@ -870,15 +880,16 @@ sym(vp8_intra_pred_y_dc128_sse2):
 ;void vp8_intra_pred_y_tm_sse2(
 ;    unsigned char *dst,
 ;    int dst_stride
-;    unsigned char *src,
-;    int src_stride,
+;    unsigned char *above,
+;    unsigned char *left,
+;    int left_stride
 ;    )
 %macro vp8_intra_pred_y_tm 1
 global sym(vp8_intra_pred_y_tm_%1)
 sym(vp8_intra_pred_y_tm_%1):
     push        rbp
     mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 4
+    SHADOW_ARGS_TO_STACK 5
     push        rsi
     push        rdi
     GET_GOT     rbx
@@ -886,9 +897,8 @@ sym(vp8_intra_pred_y_tm_%1):
 
     ; read top row
     mov         edx,        8
-    mov         rsi,        arg(2) ;src;
-    movsxd      rax,        dword ptr arg(3) ;src_stride;
-    sub         rsi,        rax
+    mov         rsi,        arg(2) ;above
+    movsxd      rax,        dword ptr arg(4) ;left_stride;
     pxor        xmm0,       xmm0
 %ifidn %1, ssse3
     movdqa      xmm3,       [GLOBAL(dc_1024)]
@@ -900,7 +910,7 @@ sym(vp8_intra_pred_y_tm_%1):
 
     ; set up left ptrs ans subtract topleft
     movd        xmm4,       [rsi-1]
-    lea         rsi,        [rsi+rax-1]
+    mov         rsi,        arg(3) ;left
 %ifidn %1, sse2
     punpcklbw   xmm4,       xmm0
     pshuflw     xmm4,       xmm4, 0x0
@@ -958,27 +968,29 @@ vp8_intra_pred_y_tm ssse3
 ;void vp8_intra_pred_y_ve_sse2(
 ;    unsigned char *dst,
 ;    int dst_stride
-;    unsigned char *src,
-;    int src_stride,
+;    unsigned char *above,
+;    unsigned char *left,
+;    int left_stride
 ;    )
 global sym(vp8_intra_pred_y_ve_sse2)
 sym(vp8_intra_pred_y_ve_sse2):
     push        rbp
     mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 4
+    SHADOW_ARGS_TO_STACK 5
     push        rsi
     ; end prolog
 
+    ;arg(3), arg(4) not used
+
+    mov         rax,        arg(2) ;above;
+    mov         rsi,        2
+    movsxd      rdx,        dword ptr arg(1) ;dst_stride
+
     ; read from top
-    mov         rax,        arg(2) ;src;
-    movsxd      rdx,        dword ptr arg(3) ;src_stride;
-    sub         rax,        rdx
     movdqa      xmm1,       [rax]
 
     ; write out
-    mov         rsi,        2
     mov         rax,        arg(0) ;dst;
-    movsxd      rdx,        dword ptr arg(1) ;dst_stride
     lea         rcx,        [rdx*3]
 
 .label
@@ -1004,25 +1016,27 @@ sym(vp8_intra_pred_y_ve_sse2):
 ;void vp8_intra_pred_y_ho_sse2(
 ;    unsigned char *dst,
 ;    int dst_stride
-;    unsigned char *src,
-;    int src_stride,
+;    unsigned char *above,
+;    unsigned char *left,
+;    int left_stride,
 ;    )
 global sym(vp8_intra_pred_y_ho_sse2)
 sym(vp8_intra_pred_y_ho_sse2):
     push        rbp
     mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 4
+    SHADOW_ARGS_TO_STACK 5
     push        rsi
     push        rdi
     ; end prolog
 
+    ;arg(2) not used
+
     ; read from left and write out
     mov         edx,        8
-    mov         rsi,        arg(2) ;src;
-    movsxd      rax,        dword ptr arg(3) ;src_stride;
+    mov         rsi,        arg(3) ;left;
+    movsxd      rax,        dword ptr arg(4) ;left_stride;
     mov         rdi,        arg(0) ;dst;
     movsxd      rcx,        dword ptr arg(1) ;dst_stride
-    dec         rsi
 
 vp8_intra_pred_y_ho_sse2_loop:
     movd        xmm0,       [rsi]
diff --git a/vp8/common/x86/recon_wrapper_sse2.c b/vp8/common/x86/recon_wrapper_sse2.c
index 949b2fb0e..b482faa3f 100644
--- a/vp8/common/x86/recon_wrapper_sse2.c
+++ b/vp8/common/x86/recon_wrapper_sse2.c
@@ -110,23 +110,32 @@ void vp8_build_intra_predictors_mbuv_s_ssse3(MACROBLOCKD *x,
                                         vp8_intra_pred_uv_ho_ssse3);
 }
 
-extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_y_dc_sse2);
-extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_y_dctop_sse2);
-extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_y_dcleft_sse2);
-extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_y_dc128_sse2);
-extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_y_ho_sse2);
-extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_y_ve_sse2);
-extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_y_tm_sse2);
-extern build_intra_predictors_mbuv_prototype(vp8_intra_pred_y_tm_ssse3);
+#define build_intra_predictors_mby_prototype(sym) \
+    void sym(unsigned char *dst, int dst_stride, \
+             const unsigned char *above, \
+             const unsigned char *left, int left_stride)
+typedef build_intra_predictors_mby_prototype((*build_intra_predictors_mby_fn_t));
+
+extern build_intra_predictors_mby_prototype(vp8_intra_pred_y_dc_sse2);
+extern build_intra_predictors_mby_prototype(vp8_intra_pred_y_dctop_sse2);
+extern build_intra_predictors_mby_prototype(vp8_intra_pred_y_dcleft_sse2);
+extern build_intra_predictors_mby_prototype(vp8_intra_pred_y_dc128_sse2);
+extern build_intra_predictors_mby_prototype(vp8_intra_pred_y_ho_sse2);
+extern build_intra_predictors_mby_prototype(vp8_intra_pred_y_ve_sse2);
+extern build_intra_predictors_mby_prototype(vp8_intra_pred_y_tm_sse2);
+extern build_intra_predictors_mby_prototype(vp8_intra_pred_y_tm_ssse3);
 
 static void vp8_build_intra_predictors_mby_x86(MACROBLOCKD *x,
+                                               unsigned char * yabove_row,
                                                unsigned char *dst_y,
                                                int dst_stride,
-                                               build_intra_predictors_mbuv_fn_t tm_func)
+                                               unsigned char * yleft,
+                                               int left_stride,
+                                               build_intra_predictors_mby_fn_t tm_func)
 {
     int mode = x->mode_info_context->mbmi.mode;
     build_intra_predictors_mbuv_fn_t fn;
-    int src_stride = x->dst.y_stride;
+
     switch (mode) {
         case  V_PRED: fn = vp8_intra_pred_y_ve_sse2; break;
         case  H_PRED: fn = vp8_intra_pred_y_ho_sse2; break;
@@ -147,19 +156,31 @@ static void vp8_build_intra_predictors_mby_x86(MACROBLOCKD *x,
         default: return;
     }
 
-//    fn(dst_y, dst_stride, x->dst.y_buffer, src_stride);
+    fn(dst_y, dst_stride, yabove_row, yleft, left_stride);
     return;
 }
 
-void vp8_build_intra_predictors_mby_s_sse2(MACROBLOCKD *x)
+void vp8_build_intra_predictors_mby_s_sse2(MACROBLOCKD *x,
+                                           unsigned char * yabove_row,
+                                           unsigned char * yleft,
+                                           int left_stride,
+                                           unsigned char * ypred_ptr,
+                                           int y_stride)
 {
-    vp8_build_intra_predictors_mby_x86(x, x->dst.y_buffer, x->dst.y_stride,
+    vp8_build_intra_predictors_mby_x86(x, yabove_row, ypred_ptr,
+                                       y_stride, yleft, left_stride,
                                        vp8_intra_pred_y_tm_sse2);
 }
 
-void vp8_build_intra_predictors_mby_s_ssse3(MACROBLOCKD *x)
+void vp8_build_intra_predictors_mby_s_ssse3(MACROBLOCKD *x,
+                                            unsigned char * yabove_row,
+                                            unsigned char * yleft,
+                                            int left_stride,
+                                            unsigned char * ypred_ptr,
+                                            int y_stride)
 {
-    vp8_build_intra_predictors_mby_x86(x, x->dst.y_buffer, x->dst.y_stride,
+    vp8_build_intra_predictors_mby_x86(x, yabove_row, ypred_ptr,
+                                     y_stride, yleft, left_stride,
                                        vp8_intra_pred_y_tm_ssse3);
 
 }
diff --git a/vp8/decoder/threading.c b/vp8/decoder/threading.c
index 20c564b32..845228bb5 100644
--- a/vp8/decoder/threading.c
+++ b/vp8/decoder/threading.c
@@ -815,15 +815,15 @@ void vp8mt_alloc_temp_buffers(VP8D_COMP *pbi, int width, int prev_mb_rows)
         /* Allocate memory for above_row buffers. */
         CHECK_MEM_ERROR(pbi->mt_yabove_row, vpx_malloc(sizeof(unsigned char *) * pc->mb_rows));
         for (i=0; i< pc->mb_rows; i++)
-            CHECK_MEM_ERROR(pbi->mt_yabove_row[i], vpx_calloc(sizeof(unsigned char) * (width + (VP8BORDERINPIXELS<<1)), 1));
+            CHECK_MEM_ERROR(pbi->mt_yabove_row[i], vpx_memalign(16,sizeof(unsigned char) * (width + (VP8BORDERINPIXELS<<1))));
 
         CHECK_MEM_ERROR(pbi->mt_uabove_row, vpx_malloc(sizeof(unsigned char *) * pc->mb_rows));
         for (i=0; i< pc->mb_rows; i++)
-            CHECK_MEM_ERROR(pbi->mt_uabove_row[i], vpx_calloc(sizeof(unsigned char) * (uv_width + VP8BORDERINPIXELS), 1));
+            CHECK_MEM_ERROR(pbi->mt_uabove_row[i], vpx_memalign(16,sizeof(unsigned char) * (uv_width + VP8BORDERINPIXELS)));
 
         CHECK_MEM_ERROR(pbi->mt_vabove_row, vpx_malloc(sizeof(unsigned char *) * pc->mb_rows));
         for (i=0; i< pc->mb_rows; i++)
-            CHECK_MEM_ERROR(pbi->mt_vabove_row[i], vpx_calloc(sizeof(unsigned char) * (uv_width + VP8BORDERINPIXELS), 1));
+            CHECK_MEM_ERROR(pbi->mt_vabove_row[i], vpx_memalign(16,sizeof(unsigned char) * (uv_width + VP8BORDERINPIXELS)));
 
         /* Allocate memory for left_col buffers. */
         CHECK_MEM_ERROR(pbi->mt_yleft_col, vpx_malloc(sizeof(unsigned char *) * pc->mb_rows));