diff --git a/build/make/configure.sh b/build/make/configure.sh
index a48fd9f..3324be3 100755
--- a/build/make/configure.sh
+++ b/build/make/configure.sh
@@ -729,19 +729,18 @@
             add_cflags -arch ${tgt_isa}
             add_ldflags -arch_only ${tgt_isa}
 
-            add_cflags  "-isysroot /Developer/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS4.2.sdk"
+            add_cflags  "-isysroot ${SDK_PATH}/SDKs/iPhoneOS4.3.sdk"
 
             # This should be overridable
-            alt_libc=${SDK_PATH}/SDKs/iPhoneOS4.2.sdk
+            alt_libc=${SDK_PATH}/SDKs/iPhoneOS4.3.sdk
 
             # Add the paths for the alternate libc
-#            for d in usr/include usr/include/gcc/darwin/4.0/; do
-            for d in usr/include usr/include/gcc/darwin/4.0/ usr/lib/gcc/arm-apple-darwin10/4.2.1/include/; do
+            for d in usr/include usr/include/gcc/darwin/4.2/ usr/lib/gcc/arm-apple-darwin10/4.2.1/include/; do
                 try_dir="${alt_libc}/${d}"
                 [ -d "${try_dir}" ] && add_cflags -I"${try_dir}"
             done
 
-            for d in lib usr/lib; do
+            for d in lib usr/lib usr/lib/system; do
                 try_dir="${alt_libc}/${d}"
                 [ -d "${try_dir}" ] && add_ldflags -L"${try_dir}"
             done
@@ -885,6 +884,8 @@
                 link_with_cc=gcc
                 tune_cflags="-march="
             setup_gnu_toolchain
+                #for 32 bit x86 builds, -O3 did not turn on this flag
+                enabled optimizations && check_add_cflags -fomit-frame-pointer
                 ;;
         esac
 
diff --git a/build/make/gen_msvs_proj.sh b/build/make/gen_msvs_proj.sh
index c2ef44a..fb948fd 100755
--- a/build/make/gen_msvs_proj.sh
+++ b/build/make/gen_msvs_proj.sh
@@ -447,6 +447,8 @@
                     obj_int_extract)
                         tag Tool \
                             Name="VCCLCompilerTool" \
+                            Optimization="2" \
+                            FavorSizeorSpeed="1" \
                             AdditionalIncludeDirectories="$incs" \
                             PreprocessorDefinitions="WIN32;NDEBUG;_CONSOLE;_CRT_SECURE_NO_WARNINGS;_CRT_SECURE_NO_DEPRECATE" \
                             RuntimeLibrary="$release_runtime" \
@@ -462,6 +464,8 @@
 
                         tag Tool \
                             Name="VCCLCompilerTool" \
+                            Optimization="2" \
+                            FavorSizeorSpeed="1" \
                             AdditionalIncludeDirectories="$incs" \
                             PreprocessorDefinitions="WIN32;NDEBUG;_CRT_SECURE_NO_WARNINGS;_CRT_SECURE_NO_DEPRECATE;$defines" \
                             RuntimeLibrary="$release_runtime" \
@@ -476,6 +480,8 @@
                         tag Tool \
                             Name="VCCLCompilerTool" \
                             AdditionalIncludeDirectories="$incs" \
+                            Optimization="2" \
+                            FavorSizeorSpeed="1" \
                             PreprocessorDefinitions="WIN32;NDEBUG;_CRT_SECURE_NO_WARNINGS;_CRT_SECURE_NO_DEPRECATE;$defines" \
                             RuntimeLibrary="$release_runtime" \
                             UsePrecompiledHeader="0" \
diff --git a/libmkv/EbmlIDs.h b/libmkv/EbmlIDs.h
index 4297470..3418e36 100644
--- a/libmkv/EbmlIDs.h
+++ b/libmkv/EbmlIDs.h
@@ -120,7 +120,7 @@
     //video
     Video = 0xE0,
     FlagInterlaced = 0x9A,
-//  StereoMode = 0x53B8,
+    StereoMode = 0x53B8,
     PixelWidth = 0xB0,
     PixelHeight = 0xBA,
     PixelCropBottom = 0x54AA,
diff --git a/vp8/common/blockd.h b/vp8/common/blockd.h
index fc8e072..aef6927 100644
--- a/vp8/common/blockd.h
+++ b/vp8/common/blockd.h
@@ -169,12 +169,8 @@
 
     unsigned char partitioning;
     unsigned char mb_skip_coeff;                                /* does this mb has coefficients at all, 1=no coefficients, 0=need decode tokens */
-    unsigned char dc_diff;
     unsigned char need_to_clamp_mvs;
-
     unsigned char segment_id;                  /* Which set of segmentation parameters should be used for this MB */
-
-    unsigned char force_no_skip; /* encoder only */
 } MB_MODE_INFO;
 
 
diff --git a/vp8/common/extend.c b/vp8/common/extend.c
index 47207fa..036bafc 100644
--- a/vp8/common/extend.c
+++ b/vp8/common/extend.c
@@ -13,10 +13,12 @@
 #include "vpx_mem/vpx_mem.h"
 
 
-static void extend_plane_borders
+static void copy_and_extend_plane
 (
     unsigned char *s, /* source */
-    int sp,           /* pitch */
+    int sp,           /* source pitch */
+    unsigned char *d, /* destination */
+    int dp,           /* destination pitch */
     int h,            /* height */
     int w,            /* width */
     int et,           /* extend top border */
@@ -25,7 +27,6 @@
     int er            /* extend right border */
 )
 {
-
     int i;
     unsigned char *src_ptr1, *src_ptr2;
     unsigned char *dest_ptr1, *dest_ptr2;
@@ -34,68 +35,73 @@
     /* copy the left and right most columns out */
     src_ptr1 = s;
     src_ptr2 = s + w - 1;
-    dest_ptr1 = s - el;
-    dest_ptr2 = s + w;
+    dest_ptr1 = d - el;
+    dest_ptr2 = d + w;
 
-    for (i = 0; i < h - 0 + 1; i++)
+    for (i = 0; i < h; i++)
     {
-        /* Some linkers will complain if we call vpx_memset with el set to a
-         * constant 0.
-         */
-        if (el)
-            vpx_memset(dest_ptr1, src_ptr1[0], el);
+        vpx_memset(dest_ptr1, src_ptr1[0], el);
+        vpx_memcpy(dest_ptr1 + el, src_ptr1, w);
         vpx_memset(dest_ptr2, src_ptr2[0], er);
         src_ptr1  += sp;
         src_ptr2  += sp;
-        dest_ptr1 += sp;
-        dest_ptr2 += sp;
+        dest_ptr1 += dp;
+        dest_ptr2 += dp;
     }
 
-    /* Now copy the top and bottom source lines into each line of the respective borders */
-    src_ptr1 = s - el;
-    src_ptr2 = s + sp * (h - 1) - el;
-    dest_ptr1 = s + sp * (-et) - el;
-    dest_ptr2 = s + sp * (h) - el;
-    linesize = el + er + w + 1;
+    /* Now copy the top and bottom lines into each line of the respective
+     * borders
+     */
+    src_ptr1 = d - el;
+    src_ptr2 = d + dp * (h - 1) - el;
+    dest_ptr1 = d + dp * (-et) - el;
+    dest_ptr2 = d + dp * (h) - el;
+    linesize = el + er + w;
 
-    for (i = 0; i < (int)et; i++)
+    for (i = 0; i < et; i++)
     {
         vpx_memcpy(dest_ptr1, src_ptr1, linesize);
-        dest_ptr1 += sp;
+        dest_ptr1 += dp;
     }
 
-    for (i = 0; i < (int)eb; i++)
+    for (i = 0; i < eb; i++)
     {
         vpx_memcpy(dest_ptr2, src_ptr2, linesize);
-        dest_ptr2 += sp;
+        dest_ptr2 += dp;
     }
 }
 
 
-void vp8_extend_to_multiple_of16(YV12_BUFFER_CONFIG *ybf, int width, int height)
+void vp8_copy_and_extend_frame(YV12_BUFFER_CONFIG *src,
+                               YV12_BUFFER_CONFIG *dst)
 {
-    int er = 0xf & (16 - (width & 0xf));
-    int eb = 0xf & (16 - (height & 0xf));
+    int et = dst->border;
+    int el = dst->border;
+    int eb = dst->border + dst->y_height - src->y_height;
+    int er = dst->border + dst->y_width - src->y_width;
 
-    /* check for non multiples of 16 */
-    if (er != 0 || eb != 0)
-    {
-        extend_plane_borders(ybf->y_buffer, ybf->y_stride, height, width, 0, 0, eb, er);
+    copy_and_extend_plane(src->y_buffer, src->y_stride,
+                          dst->y_buffer, dst->y_stride,
+                          src->y_height, src->y_width,
+                          et, el, eb, er);
 
-        /* adjust for uv */
-        height = (height + 1) >> 1;
-        width  = (width  + 1) >> 1;
-        er = 0x7 & (8 - (width  & 0x7));
-        eb = 0x7 & (8 - (height & 0x7));
+    et = (et + 1) >> 1;
+    el = (el + 1) >> 1;
+    eb = (eb + 1) >> 1;
+    er = (er + 1) >> 1;
 
-        if (er || eb)
-        {
-            extend_plane_borders(ybf->u_buffer, ybf->uv_stride, height, width, 0, 0, eb, er);
-            extend_plane_borders(ybf->v_buffer, ybf->uv_stride, height, width, 0, 0, eb, er);
-        }
-    }
+    copy_and_extend_plane(src->u_buffer, src->uv_stride,
+                          dst->u_buffer, dst->uv_stride,
+                          src->uv_height, src->uv_width,
+                          et, el, eb, er);
+
+    copy_and_extend_plane(src->v_buffer, src->uv_stride,
+                          dst->v_buffer, dst->uv_stride,
+                          src->uv_height, src->uv_width,
+                          et, el, eb, er);
 }
 
+
 /* note the extension is only for the last row, for intra prediction purpose */
 void vp8_extend_mb_row(YV12_BUFFER_CONFIG *ybf, unsigned char *YPtr, unsigned char *UPtr, unsigned char *VPtr)
 {
diff --git a/vp8/common/extend.h b/vp8/common/extend.h
index fd0a608..9e0be4e 100644
--- a/vp8/common/extend.h
+++ b/vp8/common/extend.h
@@ -14,8 +14,8 @@
 
 #include "vpx_scale/yv12config.h"
 
-void Extend(YV12_BUFFER_CONFIG *ybf);
 void vp8_extend_mb_row(YV12_BUFFER_CONFIG *ybf, unsigned char *YPtr, unsigned char *UPtr, unsigned char *VPtr);
-void vp8_extend_to_multiple_of16(YV12_BUFFER_CONFIG *ybf, int width, int height);
+void vp8_copy_and_extend_frame(YV12_BUFFER_CONFIG *src,
+                               YV12_BUFFER_CONFIG *dst);
 
 #endif
diff --git a/vp8/common/loopfilter.c b/vp8/common/loopfilter.c
index 37c5b77..221998b 100644
--- a/vp8/common/loopfilter.c
+++ b/vp8/common/loopfilter.c
@@ -353,6 +353,9 @@
         for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
         {
             int Segment = (alt_flt_enabled) ? mbd->mode_info_context->mbmi.segment_id : 0;
+            int skip_lf = (mbd->mode_info_context->mbmi.mode != B_PRED &&
+                            mbd->mode_info_context->mbmi.mode != SPLITMV &&
+                            mbd->mode_info_context->mbmi.mb_skip_coeff);
 
             filter_level = baseline_filter_level[Segment];
 
@@ -367,14 +370,14 @@
                 if (mb_col > 0)
                     cm->lf_mbv(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);
 
-                if (mbd->mode_info_context->mbmi.dc_diff > 0)
+                if (!skip_lf)
                     cm->lf_bv(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);
 
                 /* don't apply across umv border */
                 if (mb_row > 0)
                     cm->lf_mbh(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);
 
-                if (mbd->mode_info_context->mbmi.dc_diff > 0)
+                if (!skip_lf)
                     cm->lf_bh(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);
             }
 
@@ -457,6 +460,10 @@
         for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
         {
             int Segment = (alt_flt_enabled) ? mbd->mode_info_context->mbmi.segment_id : 0;
+            int skip_lf = (mbd->mode_info_context->mbmi.mode != B_PRED &&
+                            mbd->mode_info_context->mbmi.mode != SPLITMV &&
+                            mbd->mode_info_context->mbmi.mb_skip_coeff);
+
             filter_level = baseline_filter_level[Segment];
 
             /* Apply any context driven MB level adjustment */
@@ -467,14 +474,14 @@
                 if (mb_col > 0)
                     cm->lf_mbv(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
 
-                if (mbd->mode_info_context->mbmi.dc_diff > 0)
+                if (!skip_lf)
                     cm->lf_bv(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
 
                 /* don't apply across umv border */
                 if (mb_row > 0)
                     cm->lf_mbh(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
 
-                if (mbd->mode_info_context->mbmi.dc_diff > 0)
+                if (!skip_lf)
                     cm->lf_bh(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
             }
 
@@ -565,6 +572,10 @@
         for (mb_col = 0; mb_col < mb_cols; mb_col++)
         {
             int Segment = (alt_flt_enabled) ? mbd->mode_info_context->mbmi.segment_id : 0;
+            int skip_lf = (mbd->mode_info_context->mbmi.mode != B_PRED &&
+                            mbd->mode_info_context->mbmi.mode != SPLITMV &&
+                            mbd->mode_info_context->mbmi.mb_skip_coeff);
+
             filter_level = baseline_filter_level[Segment];
 
             if (filter_level)
@@ -572,12 +583,12 @@
                 if (mb_col > 0)
                     cm->lf_mbv(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
 
-                if (mbd->mode_info_context->mbmi.dc_diff > 0)
+                if (!skip_lf)
                     cm->lf_bv(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
 
                 cm->lf_mbh(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
 
-                if (mbd->mode_info_context->mbmi.dc_diff > 0)
+                if (!skip_lf)
                     cm->lf_bh(y_ptr, 0, 0, post->y_stride, 0, &lfi[filter_level], 0);
             }
 
diff --git a/vp8/common/onyx.h b/vp8/common/onyx.h
index 2e8f804..a059519 100644
--- a/vp8/common/onyx.h
+++ b/vp8/common/onyx.h
@@ -109,6 +109,7 @@
         int noise_sensitivity;   // parameter used for applying pre processing blur: recommendation 0
         int Sharpness;          // parameter used for sharpening output: recommendation 0:
         int cpu_used;
+        unsigned int rc_max_intra_bitrate_pct;
 
         // mode ->
         //(0)=Realtime/Live Encoding. This mode is optimized for realtim encoding (for example, capturing
diff --git a/vp8/common/postproc.c b/vp8/common/postproc.c
index 5bfc7d6..660880b 100644
--- a/vp8/common/postproc.c
+++ b/vp8/common/postproc.c
@@ -804,11 +804,14 @@
             for (j = 0; j < mb_cols; j++)
             {
                 char zz[4];
+                int dc_diff = !(mi[mb_index].mbmi.mode != B_PRED &&
+                              mi[mb_index].mbmi.mode != SPLITMV &&
+                              mi[mb_index].mbmi.mb_skip_coeff));
 
                 if (oci->frame_type == KEY_FRAME)
                     sprintf(zz, "a");
                 else
-                    sprintf(zz, "%c", mi[mb_index].mbmi.dc_diff + '0');
+                    sprintf(zz, "%c", dc_diff + '0');
 
                 vp8_blit_text(zz, y_ptr, post->y_stride);
                 mb_index ++;
diff --git a/vp8/common/reconinter.c b/vp8/common/reconinter.c
index 7cfab41..6862bae 100644
--- a/vp8/common/reconinter.c
+++ b/vp8/common/reconinter.c
@@ -207,12 +207,12 @@
 }
 
 
+/*encoder only*/
 void vp8_build_inter_predictors_mbuv(MACROBLOCKD *x)
 {
     int i;
 
-    if (x->mode_info_context->mbmi.ref_frame != INTRA_FRAME &&
-        x->mode_info_context->mbmi.mode != SPLITMV)
+    if (x->mode_info_context->mbmi.mode != SPLITMV)
     {
         unsigned char *uptr, *vptr;
         unsigned char *upred_ptr = &x->predictor[256];
@@ -257,69 +257,32 @@
 }
 
 /*encoder only*/
-void vp8_build_inter_predictors_mby(MACROBLOCKD *x)
+void vp8_build_inter16x16_predictors_mby(MACROBLOCKD *x)
 {
+    unsigned char *ptr_base;
+    unsigned char *ptr;
+    unsigned char *pred_ptr = x->predictor;
+    int mv_row = x->mode_info_context->mbmi.mv.as_mv.row;
+    int mv_col = x->mode_info_context->mbmi.mv.as_mv.col;
+    int pre_stride = x->block[0].pre_stride;
 
-  if (x->mode_info_context->mbmi.ref_frame != INTRA_FRAME &&
-      x->mode_info_context->mbmi.mode != SPLITMV)
+    ptr_base = x->pre.y_buffer;
+    ptr = ptr_base + (mv_row >> 3) * pre_stride + (mv_col >> 3);
+
+    if ((mv_row | mv_col) & 7)
     {
-        unsigned char *ptr_base;
-        unsigned char *ptr;
-        unsigned char *pred_ptr = x->predictor;
-        int mv_row = x->mode_info_context->mbmi.mv.as_mv.row;
-        int mv_col = x->mode_info_context->mbmi.mv.as_mv.col;
-        int pre_stride = x->block[0].pre_stride;
-
-        ptr_base = x->pre.y_buffer;
-        ptr = ptr_base + (mv_row >> 3) * pre_stride + (mv_col >> 3);
-
-        if ((mv_row | mv_col) & 7)
-        {
-            x->subpixel_predict16x16(ptr, pre_stride, mv_col & 7, mv_row & 7, pred_ptr, 16);
-        }
-        else
-        {
-            RECON_INVOKE(&x->rtcd->recon, copy16x16)(ptr, pre_stride, pred_ptr, 16);
-        }
+        x->subpixel_predict16x16(ptr, pre_stride, mv_col & 7, mv_row & 7, pred_ptr, 16);
     }
     else
     {
-        int i;
-
-        if (x->mode_info_context->mbmi.partitioning < 3)
-        {
-            for (i = 0; i < 4; i++)
-            {
-                BLOCKD *d = &x->block[bbb[i]];
-                build_inter_predictors4b(x, d, 16);
-            }
-
-        }
-        else
-        {
-            for (i = 0; i < 16; i += 2)
-            {
-                BLOCKD *d0 = &x->block[i];
-                BLOCKD *d1 = &x->block[i+1];
-
-                if (d0->bmi.mv.as_int == d1->bmi.mv.as_int)
-                    build_inter_predictors2b(x, d0, 16);
-                else
-                {
-                    vp8_build_inter_predictors_b(d0, 16, x->subpixel_predict);
-                    vp8_build_inter_predictors_b(d1, 16, x->subpixel_predict);
-                }
-
-            }
-        }
+        RECON_INVOKE(&x->rtcd->recon, copy16x16)(ptr, pre_stride, pred_ptr, 16);
     }
 }
 
 void vp8_build_inter_predictors_mb(MACROBLOCKD *x)
 {
 
-    if (x->mode_info_context->mbmi.ref_frame != INTRA_FRAME &&
-        x->mode_info_context->mbmi.mode != SPLITMV)
+    if (x->mode_info_context->mbmi.mode != SPLITMV)
     {
         int offset;
         unsigned char *ptr_base;
@@ -535,58 +498,58 @@
 
 
 
+void vp8_build_inter16x16_predictors_mb_s(MACROBLOCKD *x)
+{
+    unsigned char *dst_ptr = x->dst.y_buffer;
+
+    int offset;
+    unsigned char *ptr_base;
+    unsigned char *ptr;
+    unsigned char *uptr, *vptr;
+    unsigned char *udst_ptr = x->dst.u_buffer;
+    unsigned char *vdst_ptr = x->dst.v_buffer;
+
+    int mv_row = x->mode_info_context->mbmi.mv.as_mv.row;
+    int mv_col = x->mode_info_context->mbmi.mv.as_mv.col;
+    int pre_stride = x->dst.y_stride; /*x->block[0].pre_stride;*/
+
+    ptr_base = x->pre.y_buffer;
+    ptr = ptr_base + (mv_row >> 3) * pre_stride + (mv_col >> 3);
+
+    if ((mv_row | mv_col) & 7)
+    {
+        x->subpixel_predict16x16(ptr, pre_stride, mv_col & 7, mv_row & 7, dst_ptr, x->dst.y_stride); /*x->block[0].dst_stride);*/
+    }
+    else
+    {
+        RECON_INVOKE(&x->rtcd->recon, copy16x16)(ptr, pre_stride, dst_ptr, x->dst.y_stride); /*x->block[0].dst_stride);*/
+    }
+
+    mv_row = x->block[16].bmi.mv.as_mv.row;
+    mv_col = x->block[16].bmi.mv.as_mv.col;
+    pre_stride >>= 1;
+    offset = (mv_row >> 3) * pre_stride + (mv_col >> 3);
+    uptr = x->pre.u_buffer + offset;
+    vptr = x->pre.v_buffer + offset;
+
+    if ((mv_row | mv_col) & 7)
+    {
+        x->subpixel_predict8x8(uptr, pre_stride, mv_col & 7, mv_row & 7, udst_ptr, x->dst.uv_stride);
+        x->subpixel_predict8x8(vptr, pre_stride, mv_col & 7, mv_row & 7, vdst_ptr, x->dst.uv_stride);
+    }
+    else
+    {
+        RECON_INVOKE(&x->rtcd->recon, copy8x8)(uptr, pre_stride, udst_ptr, x->dst.uv_stride);
+        RECON_INVOKE(&x->rtcd->recon, copy8x8)(vptr, pre_stride, vdst_ptr, x->dst.uv_stride);
+    }
+}
 void vp8_build_inter_predictors_mb_s(MACROBLOCKD *x)
 {
-    /*unsigned char *pred_ptr = x->block[0].predictor;
-    unsigned char *dst_ptr = *(x->block[0].base_dst) + x->block[0].dst;*/
-    unsigned char *pred_ptr = x->predictor;
     unsigned char *dst_ptr = x->dst.y_buffer;
 
     if (x->mode_info_context->mbmi.mode != SPLITMV)
     {
-        int offset;
-        unsigned char *ptr_base;
-        unsigned char *ptr;
-        unsigned char *uptr, *vptr;
-        /*unsigned char *pred_ptr = x->predictor;
-        unsigned char *upred_ptr = &x->predictor[256];
-        unsigned char *vpred_ptr = &x->predictor[320];*/
-        unsigned char *udst_ptr = x->dst.u_buffer;
-        unsigned char *vdst_ptr = x->dst.v_buffer;
-
-        int mv_row = x->mode_info_context->mbmi.mv.as_mv.row;
-        int mv_col = x->mode_info_context->mbmi.mv.as_mv.col;
-        int pre_stride = x->dst.y_stride; /*x->block[0].pre_stride;*/
-
-        ptr_base = x->pre.y_buffer;
-        ptr = ptr_base + (mv_row >> 3) * pre_stride + (mv_col >> 3);
-
-        if ((mv_row | mv_col) & 7)
-        {
-            x->subpixel_predict16x16(ptr, pre_stride, mv_col & 7, mv_row & 7, dst_ptr, x->dst.y_stride); /*x->block[0].dst_stride);*/
-        }
-        else
-        {
-            RECON_INVOKE(&x->rtcd->recon, copy16x16)(ptr, pre_stride, dst_ptr, x->dst.y_stride); /*x->block[0].dst_stride);*/
-        }
-
-        mv_row = x->block[16].bmi.mv.as_mv.row;
-        mv_col = x->block[16].bmi.mv.as_mv.col;
-        pre_stride >>= 1;
-        offset = (mv_row >> 3) * pre_stride + (mv_col >> 3);
-        uptr = x->pre.u_buffer + offset;
-        vptr = x->pre.v_buffer + offset;
-
-        if ((mv_row | mv_col) & 7)
-        {
-            x->subpixel_predict8x8(uptr, pre_stride, mv_col & 7, mv_row & 7, udst_ptr, x->dst.uv_stride);
-            x->subpixel_predict8x8(vptr, pre_stride, mv_col & 7, mv_row & 7, vdst_ptr, x->dst.uv_stride);
-        }
-        else
-        {
-            RECON_INVOKE(&x->rtcd->recon, copy8x8)(uptr, pre_stride, udst_ptr, x->dst.uv_stride);
-            RECON_INVOKE(&x->rtcd->recon, copy8x8)(vptr, pre_stride, vdst_ptr, x->dst.uv_stride);
-        }
+        vp8_build_inter16x16_predictors_mb_s(x);
     }
     else
     {
@@ -599,25 +562,20 @@
         {
             for (i = 0; i < 4; i++)
             {
+                unsigned char *ptr_base;
+                unsigned char *ptr;
                 BLOCKD *d = &x->block[bbb[i]];
-                /*build_inter_predictors4b(x, d, 16);*/
 
+                ptr_base = *(d->base_pre);
+                ptr = ptr_base + d->pre + (d->bmi.mv.as_mv.row >> 3) * d->pre_stride + (d->bmi.mv.as_mv.col >> 3);
+
+                if (d->bmi.mv.as_mv.row & 7 || d->bmi.mv.as_mv.col & 7)
                 {
-                    unsigned char *ptr_base;
-                    unsigned char *ptr;
-                    unsigned char *pred_ptr = d->predictor;
-
-                    ptr_base = *(d->base_pre);
-                    ptr = ptr_base + d->pre + (d->bmi.mv.as_mv.row >> 3) * d->pre_stride + (d->bmi.mv.as_mv.col >> 3);
-
-                    if (d->bmi.mv.as_mv.row & 7 || d->bmi.mv.as_mv.col & 7)
-                    {
-                        x->subpixel_predict8x8(ptr, d->pre_stride, d->bmi.mv.as_mv.col & 7, d->bmi.mv.as_mv.row & 7, dst_ptr, x->dst.y_stride); /*x->block[0].dst_stride);*/
-                    }
-                    else
-                    {
-                        RECON_INVOKE(&x->rtcd->recon, copy8x8)(ptr, d->pre_stride, dst_ptr, x->dst.y_stride); /*x->block[0].dst_stride);*/
-                    }
+                    x->subpixel_predict8x8(ptr, d->pre_stride, d->bmi.mv.as_mv.col & 7, d->bmi.mv.as_mv.row & 7, dst_ptr, x->dst.y_stride); /*x->block[0].dst_stride);*/
+                }
+                else
+                {
+                    RECON_INVOKE(&x->rtcd->recon, copy8x8)(ptr, d->pre_stride, dst_ptr, x->dst.y_stride); /*x->block[0].dst_stride);*/
                 }
             }
         }
@@ -633,7 +591,6 @@
                     /*build_inter_predictors2b(x, d0, 16);*/
                     unsigned char *ptr_base;
                     unsigned char *ptr;
-                    unsigned char *pred_ptr = d0->predictor;
 
                     ptr_base = *(d0->base_pre);
                     ptr = ptr_base + d0->pre + (d0->bmi.mv.as_mv.row >> 3) * d0->pre_stride + (d0->bmi.mv.as_mv.col >> 3);
@@ -665,7 +622,6 @@
                 /*build_inter_predictors2b(x, d0, 8);*/
                 unsigned char *ptr_base;
                 unsigned char *ptr;
-                unsigned char *pred_ptr = d0->predictor;
 
                 ptr_base = *(d0->base_pre);
                 ptr = ptr_base + d0->pre + (d0->bmi.mv.as_mv.row >> 3) * d0->pre_stride + (d0->bmi.mv.as_mv.col >> 3);
diff --git a/vp8/common/reconinter.h b/vp8/common/reconinter.h
index 7c1dee4..688bebe 100644
--- a/vp8/common/reconinter.h
+++ b/vp8/common/reconinter.h
@@ -14,8 +14,9 @@
 
 extern void vp8_build_inter_predictors_mb(MACROBLOCKD *x);
 extern void vp8_build_inter_predictors_mb_s(MACROBLOCKD *x);
+extern void vp8_build_inter16x16_predictors_mb_s(MACROBLOCKD *x);
 
-extern void vp8_build_inter_predictors_mby(MACROBLOCKD *x);
+extern void vp8_build_inter16x16_predictors_mby(MACROBLOCKD *x);
 extern void vp8_build_uvmvs(MACROBLOCKD *x, int fullpixel);
 extern void vp8_build_inter_predictors_b(BLOCKD *d, int pitch, vp8_subpix_fn_t sppf);
 extern void vp8_build_inter_predictors_mbuv(MACROBLOCKD *x);
diff --git a/vp8/common/threading.h b/vp8/common/threading.h
index 44eaf08..b7542b3 100644
--- a/vp8/common/threading.h
+++ b/vp8/common/threading.h
@@ -12,8 +12,6 @@
 #ifndef _PTHREAD_EMULATION
 #define _PTHREAD_EMULATION
 
-#define VPXINFINITE 10000       /* 10second. */
-
 #if CONFIG_OS_SUPPORT && CONFIG_MULTITHREAD
 
 /* Thread management macros */
@@ -28,7 +26,7 @@
 #define pthread_t HANDLE
 #define pthread_attr_t DWORD
 #define pthread_create(thhandle,attr,thfunc,tharg) (int)((*thhandle=(HANDLE)_beginthreadex(NULL,0,(unsigned int (__stdcall *)(void *))thfunc,tharg,0,NULL))==NULL)
-#define pthread_join(thread, result) ((WaitForSingleObject((thread),VPXINFINITE)!=WAIT_OBJECT_0) || !CloseHandle(thread))
+#define pthread_join(thread, result) ((WaitForSingleObject((thread),INFINITE)!=WAIT_OBJECT_0) || !CloseHandle(thread))
 #define pthread_detach(thread) if(thread!=NULL)CloseHandle(thread)
 #define thread_sleep(nms) Sleep(nms)
 #define pthread_cancel(thread) terminate_thread(thread,0)
@@ -62,7 +60,7 @@
 #define sem_t HANDLE
 #define pause(voidpara) __asm PAUSE
 #define sem_init(sem, sem_attr1, sem_init_value) (int)((*sem = CreateEvent(NULL,FALSE,FALSE,NULL))==NULL)
-#define sem_wait(sem) (int)(WAIT_OBJECT_0 != WaitForSingleObject(*sem,VPXINFINITE))
+#define sem_wait(sem) (int)(WAIT_OBJECT_0 != WaitForSingleObject(*sem,INFINITE))
 #define sem_post(sem) SetEvent(*sem)
 #define sem_destroy(sem) if(*sem)((int)(CloseHandle(*sem))==TRUE)
 #define thread_sleep(nms) Sleep(nms)
diff --git a/vp8/common/x86/idctllm_sse2.asm b/vp8/common/x86/idctllm_sse2.asm
index edee157..34a7e18 100644
--- a/vp8/common/x86/idctllm_sse2.asm
+++ b/vp8/common/x86/idctllm_sse2.asm
@@ -32,9 +32,6 @@
         mov         rdx,            arg(1) ; dequant
         mov         rax,            arg(0) ; qcoeff
 
-    ; Zero out xmm7, for use unpacking
-        pxor        xmm7,           xmm7
-
         movd        xmm4,           [rax]
         movd        xmm5,           [rdx]
 
@@ -43,9 +40,12 @@
 
         pmullw      xmm4,           xmm5
 
+    ; Zero out xmm5, for use unpacking
+        pxor        xmm5,           xmm5
+
     ; clear coeffs
-        movd        [rax],          xmm7
-        movd        [rax+32],       xmm7
+        movd        [rax],          xmm5
+        movd        [rax+32],       xmm5
 ;pshufb
         pshuflw     xmm4,           xmm4,       00000000b
         pshufhw     xmm4,           xmm4,       00000000b
@@ -62,10 +62,10 @@
         lea         rcx,            [3*rcx]
         movq        xmm3,           [rax+rcx]
 
-        punpcklbw   xmm0,           xmm7
-        punpcklbw   xmm1,           xmm7
-        punpcklbw   xmm2,           xmm7
-        punpcklbw   xmm3,           xmm7
+        punpcklbw   xmm0,           xmm5
+        punpcklbw   xmm1,           xmm5
+        punpcklbw   xmm2,           xmm5
+        punpcklbw   xmm3,           xmm5
 
         mov         rax,            arg(3) ; dst
         movsxd      rdx,            dword ptr arg(4) ; dst_stride
@@ -77,10 +77,10 @@
         paddw       xmm3,           xmm4
 
     ; pack up before storing
-        packuswb    xmm0,           xmm7
-        packuswb    xmm1,           xmm7
-        packuswb    xmm2,           xmm7
-        packuswb    xmm3,           xmm7
+        packuswb    xmm0,           xmm5
+        packuswb    xmm1,           xmm5
+        packuswb    xmm2,           xmm5
+        packuswb    xmm3,           xmm5
 
     ; store blocks back out
         movq        [rax],          xmm0
@@ -102,6 +102,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -347,6 +348,7 @@
     pop         rdi
     pop         rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -377,8 +379,8 @@
         mov         rdi,            arg(3) ; dst
         mov         rdx,            arg(5) ; dc
 
-    ; Zero out xmm7, for use unpacking
-        pxor        xmm7,           xmm7
+    ; Zero out xmm5, for use unpacking
+        pxor        xmm5,           xmm5
 
     ; load up 2 dc words here == 2*16 = doubleword
         movd        xmm4,           [rdx]
@@ -398,10 +400,10 @@
         psraw       xmm4,           3
 
     ; Predict buffer needs to be expanded from bytes to words
-        punpcklbw   xmm0,           xmm7
-        punpcklbw   xmm1,           xmm7
-        punpcklbw   xmm2,           xmm7
-        punpcklbw   xmm3,           xmm7
+        punpcklbw   xmm0,           xmm5
+        punpcklbw   xmm1,           xmm5
+        punpcklbw   xmm2,           xmm5
+        punpcklbw   xmm3,           xmm5
 
     ; Add to predict buffer
         paddw       xmm0,           xmm4
@@ -410,10 +412,10 @@
         paddw       xmm3,           xmm4
 
     ; pack up before storing
-        packuswb    xmm0,           xmm7
-        packuswb    xmm1,           xmm7
-        packuswb    xmm2,           xmm7
-        packuswb    xmm3,           xmm7
+        packuswb    xmm0,           xmm5
+        packuswb    xmm1,           xmm5
+        packuswb    xmm2,           xmm5
+        packuswb    xmm3,           xmm5
 
     ; Load destination stride before writing out,
     ;   doesn't need to persist
@@ -441,6 +443,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -692,6 +695,7 @@
     pop         rdi
     pop         rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
diff --git a/vp8/common/x86/iwalsh_sse2.asm b/vp8/common/x86/iwalsh_sse2.asm
index 83c97df..1da4fd8 100644
--- a/vp8/common/x86/iwalsh_sse2.asm
+++ b/vp8/common/x86/iwalsh_sse2.asm
@@ -17,7 +17,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 2
-    SAVE_XMM
+    SAVE_XMM 6
     push        rsi
     push        rdi
     ; end prolog
@@ -41,7 +41,7 @@
     movdqa    xmm4, xmm0
     punpcklqdq  xmm0, xmm3          ;d1 a1
     punpckhqdq  xmm4, xmm3          ;c1 b1
-    movd    xmm7, eax
+    movd    xmm6, eax
 
     movdqa    xmm1, xmm4          ;c1 b1
     paddw   xmm4, xmm0          ;dl+cl a1+b1 aka op[4] op[0]
@@ -66,7 +66,7 @@
     pshufd    xmm2, xmm1, 4eh       ;ip[8] ip[12]
     movdqa    xmm3, xmm4          ;ip[4] ip[0]
 
-    pshufd    xmm7, xmm7, 0       ;03 03 03 03 03 03 03 03
+    pshufd    xmm6, xmm6, 0       ;03 03 03 03 03 03 03 03
 
     paddw   xmm4, xmm2          ;ip[4]+ip[8] ip[0]+ip[12] aka b1 a1
     psubw   xmm3, xmm2          ;ip[4]-ip[8] ip[0]-ip[12] aka c1 d1
@@ -90,8 +90,8 @@
     punpcklwd xmm5, xmm0          ; 31 21 11 01 30 20 10 00
     punpckhwd xmm1, xmm0          ; 33 23 13 03 32 22 12 02
 ;~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-    paddw   xmm5, xmm7
-    paddw   xmm1, xmm7
+    paddw   xmm5, xmm6
+    paddw   xmm1, xmm6
 
     psraw   xmm5, 3
     psraw   xmm1, 3
diff --git a/vp8/common/x86/loopfilter_sse2.asm b/vp8/common/x86/loopfilter_sse2.asm
index 849133d..c2ce1a1 100644
--- a/vp8/common/x86/loopfilter_sse2.asm
+++ b/vp8/common/x86/loopfilter_sse2.asm
@@ -288,7 +288,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -338,7 +338,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -584,7 +584,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -634,7 +634,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1024,7 +1024,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1091,7 +1091,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1249,7 +1249,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1318,7 +1318,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1386,7 +1386,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1503,7 +1503,7 @@
     push        rbp         ; save old base pointer value.
     mov         rbp, rsp    ; set new base pointer value.
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx         ; save callee-saved reg
     push        rsi
     push        rdi
diff --git a/vp8/common/x86/postproc_sse2.asm b/vp8/common/x86/postproc_sse2.asm
index 30b4bf5..06d51ec 100644
--- a/vp8/common/x86/postproc_sse2.asm
+++ b/vp8/common/x86/postproc_sse2.asm
@@ -26,7 +26,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -256,7 +256,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 5
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -456,7 +456,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 5
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
diff --git a/vp8/common/x86/recon_sse2.asm b/vp8/common/x86/recon_sse2.asm
index 4ad3973..67b6420 100644
--- a/vp8/common/x86/recon_sse2.asm
+++ b/vp8/common/x86/recon_sse2.asm
@@ -67,7 +67,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 4
-    SAVE_XMM
+    SAVE_XMM 7
     push        rsi
     push        rdi
     ; end prolog
diff --git a/vp8/common/x86/subpixel_sse2.asm b/vp8/common/x86/subpixel_sse2.asm
index b87cad2..83e3b14 100644
--- a/vp8/common/x86/subpixel_sse2.asm
+++ b/vp8/common/x86/subpixel_sse2.asm
@@ -37,7 +37,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -157,7 +157,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -333,7 +333,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 8
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -428,7 +428,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 8
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -538,7 +538,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -651,7 +651,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -816,7 +816,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -908,7 +908,6 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 5
-    ;SAVE_XMM                          ;xmm6, xmm7 are not used here.
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -948,7 +947,6 @@
     pop rdi
     pop rsi
     RESTORE_GOT
-    ;RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -969,7 +967,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1238,7 +1236,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
diff --git a/vp8/common/x86/subpixel_ssse3.asm b/vp8/common/x86/subpixel_ssse3.asm
index 7f6fd93..1ddbc54 100644
--- a/vp8/common/x86/subpixel_ssse3.asm
+++ b/vp8/common/x86/subpixel_ssse3.asm
@@ -39,6 +39,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -107,6 +108,7 @@
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -162,6 +164,7 @@
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -179,7 +182,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -194,10 +197,6 @@
 
     mov         rdi, arg(2)                     ;output_ptr
 
-;;
-;;    cmp         esi, DWORD PTR [rax]
-;;    je          vp8_filter_block1d16_h4_ssse3
-
     mov         rsi, arg(0)                     ;src_ptr
 
     movdqa      xmm4, XMMWORD PTR [rax]         ;k0_k5
@@ -271,61 +270,7 @@
     pop rdi
     pop rsi
     RESTORE_GOT
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-vp8_filter_block1d16_h4_ssse3:
-    movdqa      xmm5, XMMWORD PTR [rax+256]     ;k2_k4
-    movdqa      xmm6, XMMWORD PTR [rax+128]     ;k1_k3
-
-    mov         rsi, arg(0)             ;src_ptr
-    movsxd      rax, dword ptr arg(1)   ;src_pixels_per_line
-    movsxd      rcx, dword ptr arg(4)   ;output_height
-    movsxd      rdx, dword ptr arg(3)   ;output_pitch
-
-filter_block1d16_h4_rowloop_ssse3:
-    movdqu      xmm1,   XMMWORD PTR [rsi - 2]
-
-    movdqa      xmm2, xmm1
-    pshufb      xmm1, [GLOBAL(shuf2b)]
-    pshufb      xmm2, [GLOBAL(shuf3b)]
-    pmaddubsw   xmm1, xmm5
-
-    movdqu      xmm3,   XMMWORD PTR [rsi + 6]
-
-    pmaddubsw   xmm2, xmm6
-    movdqa      xmm0, xmm3
-    pshufb      xmm3, [GLOBAL(shuf3b)]
-    pshufb      xmm0, [GLOBAL(shuf2b)]
-
-    paddsw      xmm1, [GLOBAL(rd)]
-    paddsw      xmm1, xmm2
-
-    pmaddubsw   xmm0, xmm5
-    pmaddubsw   xmm3, xmm6
-
-    psraw       xmm1, 7
-    packuswb    xmm1, xmm1
-    lea         rsi,    [rsi + rax]
-    paddsw      xmm3, xmm0
-    paddsw      xmm3, [GLOBAL(rd)]
-    psraw       xmm3, 7
-    packuswb    xmm3, xmm3
-
-    punpcklqdq  xmm1, xmm3
-
-    movdqa      XMMWORD Ptr [rdi], xmm1
-
-    add         rdi, rdx
-    dec         rcx
-    jnz         filter_block1d16_h4_rowloop_ssse3
-
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -344,6 +289,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -451,6 +397,7 @@
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -471,6 +418,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -566,6 +514,7 @@
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -638,6 +587,7 @@
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -656,6 +606,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -728,6 +679,7 @@
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -776,6 +728,7 @@
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -932,7 +885,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -1195,7 +1148,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
diff --git a/vp8/decoder/decodframe.c b/vp8/decoder/decodframe.c
index 82841e8..9598508 100644
--- a/vp8/decoder/decodframe.c
+++ b/vp8/decoder/decodframe.c
@@ -111,16 +111,15 @@
  */
 static void skip_recon_mb(VP8D_COMP *pbi, MACROBLOCKD *xd)
 {
-    if (xd->frame_type == KEY_FRAME  ||  xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME)
+    if (xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME)
     {
-
         vp8_build_intra_predictors_mbuv_s(xd);
         RECON_INVOKE(&pbi->common.rtcd.recon,
                      build_intra_predictors_mby_s)(xd);
     }
     else
     {
-        vp8_build_inter_predictors_mb_s(xd);
+        vp8_build_inter16x16_predictors_mb_s(xd);
     }
 }
 
@@ -195,11 +194,10 @@
         clamp_mvs(xd);
     }
 
-    xd->mode_info_context->mbmi.dc_diff = 1;
-
-    if (xd->mode_info_context->mbmi.mode != B_PRED && xd->mode_info_context->mbmi.mode != SPLITMV && eobtotal == 0)
+    eobtotal |= (xd->mode_info_context->mbmi.mode == B_PRED ||
+                  xd->mode_info_context->mbmi.mode == SPLITMV);
+    if (!eobtotal)
     {
-        xd->mode_info_context->mbmi.dc_diff = 0;
         skip_recon_mb(pbi, xd);
         return;
     }
@@ -208,7 +206,7 @@
         mb_init_dequantizer(pbi, xd);
 
     /* do prediction */
-    if (xd->frame_type == KEY_FRAME  ||  xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME)
+    if (xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME)
     {
         vp8_build_intra_predictors_mbuv(xd);
 
@@ -255,7 +253,7 @@
                          xd->predictor, xd->dst.y_buffer,
                          xd->dst.y_stride, xd->eobs, xd->block[24].diff);
     }
-    else if ((xd->frame_type == KEY_FRAME  ||  xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME) && xd->mode_info_context->mbmi.mode == B_PRED)
+    else if (xd->mode_info_context->mbmi.mode == B_PRED)
     {
         for (i = 0; i < 16; i++)
         {
diff --git a/vp8/decoder/detokenize.c b/vp8/decoder/detokenize.c
old mode 100755
new mode 100644
diff --git a/vp8/decoder/onyxd_if.c b/vp8/decoder/onyxd_if.c
index ef2e00d..1e83ab5 100644
--- a/vp8/decoder/onyxd_if.c
+++ b/vp8/decoder/onyxd_if.c
@@ -76,7 +76,6 @@
     pbi->common.current_video_frame = 0;
     pbi->ready_for_new_data = 1;
 
-    pbi->CPUFreq = 0; /*vp8_get_processor_freq();*/
 #if CONFIG_MULTITHREAD
     pbi->max_threads = oxcf->max_threads;
     vp8_decoder_create_threads(pbi);
@@ -252,7 +251,6 @@
     VP8D_COMP *pbi = (VP8D_COMP *) ptr;
     VP8_COMMON *cm = &pbi->common;
     int retcode = 0;
-    struct vpx_usec_timer timer;
 
     /*if(pbi->ready_for_new_data == 0)
         return -1;*/
@@ -317,8 +315,6 @@
 
     pbi->common.error.setjmp = 1;
 
-    vpx_usec_timer_start(&timer);
-
     /*cm->current_video_frame++;*/
     pbi->Source = source;
     pbi->source_sz = size;
@@ -379,15 +375,9 @@
 
         if(pbi->common.filter_level)
         {
-            struct vpx_usec_timer lpftimer;
-            vpx_usec_timer_start(&lpftimer);
             /* Apply the loop filter if appropriate. */
-
             vp8_loop_filter_frame(cm, &pbi->mb, cm->filter_level);
 
-            vpx_usec_timer_mark(&lpftimer);
-            pbi->time_loop_filtering += vpx_usec_timer_elapsed(&lpftimer);
-
             cm->last_frame_type = cm->frame_type;
             cm->last_filter_type = cm->filter_type;
             cm->last_sharpness_level = cm->sharpness_level;
@@ -398,11 +388,6 @@
 
     vp8_clear_system_state();
 
-    vpx_usec_timer_mark(&timer);
-    pbi->decode_microseconds = vpx_usec_timer_elapsed(&timer);
-
-    pbi->time_decoding += pbi->decode_microseconds;
-
     /*vp8_print_modes_and_motion_vectors( cm->mi, cm->mb_rows,cm->mb_cols, cm->current_video_frame);*/
 
     if (cm->show_frame)
diff --git a/vp8/decoder/onyxd_int.h b/vp8/decoder/onyxd_int.h
index ac1e332..9b91756 100644
--- a/vp8/decoder/onyxd_int.h
+++ b/vp8/decoder/onyxd_int.h
@@ -81,12 +81,6 @@
     const unsigned char *Source;
     unsigned int   source_sz;
 
-
-    unsigned int CPUFreq;
-    unsigned int decode_microseconds;
-    unsigned int time_decoding;
-    unsigned int time_loop_filtering;
-
 #if CONFIG_MULTITHREAD
     /* variable for threading */
 
diff --git a/vp8/decoder/threading.c b/vp8/decoder/threading.c
index 3d9d428..c92de00 100644
--- a/vp8/decoder/threading.c
+++ b/vp8/decoder/threading.c
@@ -108,12 +108,10 @@
         clamp_mvs(xd);
     }
 
-    xd->mode_info_context->mbmi.dc_diff = 1;
-
-    if (xd->mode_info_context->mbmi.mode != B_PRED && xd->mode_info_context->mbmi.mode != SPLITMV && eobtotal == 0)
+    eobtotal |= (xd->mode_info_context->mbmi.mode == B_PRED ||
+                  xd->mode_info_context->mbmi.mode == SPLITMV);
+    if (!eobtotal)
     {
-        xd->mode_info_context->mbmi.dc_diff = 0;
-
         /*mt_skip_recon_mb(pbi, xd, mb_row, mb_col);*/
         if (xd->frame_type == KEY_FRAME  ||  xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME)
         {
@@ -122,7 +120,7 @@
         }
         else
         {
-            vp8_build_inter_predictors_mb_s(xd);
+            vp8_build_inter16x16_predictors_mb_s(xd);
         }
         return;
     }
@@ -322,6 +320,7 @@
 
                         if (pbi->common.filter_level)
                         {
+                            int skip_lf;
                             if( mb_row != pc->mb_rows-1 )
                             {
                                 /* Save decoded MB last row data for next-row decoding */
@@ -349,6 +348,10 @@
 
                             /* update loopfilter info */
                             Segment = (alt_flt_enabled) ? xd->mode_info_context->mbmi.segment_id : 0;
+                            skip_lf = (xd->mode_info_context->mbmi.mode != B_PRED &&
+                                            xd->mode_info_context->mbmi.mode != SPLITMV &&
+                                            xd->mode_info_context->mbmi.mb_skip_coeff);
+
                             filter_level = pbi->mt_baseline_filter_level[Segment];
                             /* Distance of Mb to the various image edges.
                              * These are specified to 8th pel as they are always compared to values that are in 1/8th pel units
@@ -362,14 +365,14 @@
                                 if (mb_col > 0)
                                     pc->lf_mbv(xd->dst.y_buffer, xd->dst.u_buffer, xd->dst.v_buffer, recon_y_stride, recon_uv_stride, &lfi[filter_level], pc->simpler_lpf);
 
-                                if (xd->mode_info_context->mbmi.dc_diff > 0)
+                                if (!skip_lf)
                                     pc->lf_bv(xd->dst.y_buffer, xd->dst.u_buffer, xd->dst.v_buffer, recon_y_stride, recon_uv_stride, &lfi[filter_level], pc->simpler_lpf);
 
                                 /* don't apply across umv border */
                                 if (mb_row > 0)
                                     pc->lf_mbh(xd->dst.y_buffer, xd->dst.u_buffer, xd->dst.v_buffer, recon_y_stride, recon_uv_stride, &lfi[filter_level], pc->simpler_lpf);
 
-                                if (xd->mode_info_context->mbmi.dc_diff > 0)
+                                if (!skip_lf)
                                     pc->lf_bh(xd->dst.y_buffer, xd->dst.u_buffer, xd->dst.v_buffer, recon_y_stride, recon_uv_stride, &lfi[filter_level], pc->simpler_lpf);
                             }
                         }
@@ -810,6 +813,7 @@
 
                 if (pbi->common.filter_level)
                 {
+                    int skip_lf;
                     /* Save decoded MB last row data for next-row decoding */
                     if(mb_row != pc->mb_rows-1)
                     {
@@ -837,6 +841,9 @@
 
                     /* update loopfilter info */
                     Segment = (alt_flt_enabled) ? xd->mode_info_context->mbmi.segment_id : 0;
+                    skip_lf = (xd->mode_info_context->mbmi.mode != B_PRED &&
+                                    xd->mode_info_context->mbmi.mode != SPLITMV &&
+                                    xd->mode_info_context->mbmi.mb_skip_coeff);
                     filter_level = pbi->mt_baseline_filter_level[Segment];
                     /* Distance of Mb to the various image edges.
                      * These are specified to 8th pel as they are always compared to values that are in 1/8th pel units
@@ -850,14 +857,14 @@
                         if (mb_col > 0)
                             pc->lf_mbv(xd->dst.y_buffer, xd->dst.u_buffer, xd->dst.v_buffer, recon_y_stride, recon_uv_stride, &lfi[filter_level], pc->simpler_lpf);
 
-                        if (xd->mode_info_context->mbmi.dc_diff > 0)
+                        if (!skip_lf)
                             pc->lf_bv(xd->dst.y_buffer, xd->dst.u_buffer, xd->dst.v_buffer, recon_y_stride, recon_uv_stride, &lfi[filter_level], pc->simpler_lpf);
 
                         /* don't apply across umv border */
                         if (mb_row > 0)
                             pc->lf_mbh(xd->dst.y_buffer, xd->dst.u_buffer, xd->dst.v_buffer, recon_y_stride, recon_uv_stride, &lfi[filter_level], pc->simpler_lpf);
 
-                        if (xd->mode_info_context->mbmi.dc_diff > 0)
+                        if (!skip_lf)
                             pc->lf_bh(xd->dst.y_buffer, xd->dst.u_buffer, xd->dst.v_buffer, recon_y_stride, recon_uv_stride, &lfi[filter_level], pc->simpler_lpf);
                     }
                 }
diff --git a/vp8/encoder/block.h b/vp8/encoder/block.h
index 5a2568d..6de4c85 100644
--- a/vp8/encoder/block.h
+++ b/vp8/encoder/block.h
@@ -34,7 +34,7 @@
     // 16 Y blocks, 4 U blocks, 4 V blocks each with 16 entries
     short *quant;
     short *quant_fast;
-    short *quant_shift;
+    unsigned char *quant_shift;
     short *zbin;
     short *zrun_zbin_boost;
     short *round;
diff --git a/vp8/encoder/encodeframe.c b/vp8/encoder/encodeframe.c
index ab4071d..f48a1fc 100644
--- a/vp8/encoder/encodeframe.c
+++ b/vp8/encoder/encodeframe.c
@@ -147,7 +147,7 @@
 #define EXACT_QUANT
 #ifdef EXACT_QUANT
 static void vp8cx_invert_quant(int improved_quant, short *quant,
-                               short *shift, short d)
+                               unsigned char *shift, short d)
 {
     if(improved_quant)
     {
@@ -1157,7 +1157,6 @@
 
     x->e_mbd.mode_info_context->mbmi.ref_frame = INTRA_FRAME;
 
-#if !(CONFIG_REALTIME_ONLY)
     if (cpi->sf.RD && cpi->compressor_speed != 2)
     {
         vp8_rd_pick_intra_mbuv_mode(cpi, x, &rateuv, &rateuv_tokenonly, &distuv);
@@ -1170,7 +1169,6 @@
         rate += (Error4x4 < Error16x16) ? rate4x4 : rate16x16;
     }
     else
-#endif
     {
         int rate2, best_distortion;
         MB_PREDICTION_MODE mode, best_mode = DC_PRED;
@@ -1188,7 +1186,7 @@
                 (&x->e_mbd);
             distortion2 = VARIANCE_INVOKE(&cpi->rtcd.variance, get16x16prederror)(x->src.y_buffer, x->src.y_stride, x->e_mbd.predictor, 16, 0x7fffffff);
             rate2  = x->mbmode_cost[x->e_mbd.frame_type][mode];
-            this_rd = RD_ESTIMATE(x->rdmult, x->rddiv, rate2, distortion2);
+            this_rd = RDCOST(x->rdmult, x->rddiv, rate2, distortion2);
 
             if (Error16x16 > this_rd)
             {
@@ -1242,8 +1240,6 @@
     else
         x->encode_breakout = cpi->oxcf.encode_breakout;
 
-#if !(CONFIG_REALTIME_ONLY)
-
     if (cpi->sf.RD)
     {
         int zbin_mode_boost_enabled = cpi->zbin_mode_boost_enabled;
@@ -1270,7 +1266,6 @@
 
     }
     else
-#endif
         vp8_pick_inter_mode(cpi, x, recon_yoffset, recon_uvoffset, &rate, &distortion, &intra_error);
 
     cpi->prediction_error += distortion;
@@ -1386,7 +1381,7 @@
             cpi->MVcount[1][mv_max+((xd->block[0].bmi.mv.as_mv.col - best_ref_mv.col) >> 1)]++;
         }
 
-        if (!x->skip && !x->e_mbd.mode_info_context->mbmi.force_no_skip)
+        if (!x->skip)
         {
             vp8_encode_inter16x16(IF_RTCD(&cpi->rtcd), x);
 
@@ -1396,7 +1391,7 @@
 
         }
         else
-            vp8_stuff_inter16x16(x);
+            vp8_build_inter_predictors_mb_s(xd);
     }
 
     if (!x->skip)
@@ -1405,11 +1400,6 @@
     {
         if (cpi->common.mb_no_coeff_skip)
         {
-            if (xd->mode_info_context->mbmi.mode != B_PRED && xd->mode_info_context->mbmi.mode != SPLITMV)
-                xd->mode_info_context->mbmi.dc_diff = 0;
-            else
-                xd->mode_info_context->mbmi.dc_diff = 1;
-
             xd->mode_info_context->mbmi.mb_skip_coeff = 1;
             cpi->skip_true_count ++;
             vp8_fix_contexts(xd);
diff --git a/vp8/encoder/encodeintra.c b/vp8/encoder/encodeintra.c
index 4400006..7d835a1 100644
--- a/vp8/encoder/encodeintra.c
+++ b/vp8/encoder/encodeintra.c
@@ -75,14 +75,9 @@
 
     vp8_quantize_mby(x);
 
-#if !(CONFIG_REALTIME_ONLY)
-#if 1
     if (x->optimize)
         vp8_optimize_mby(x, rtcd);
 
-#endif
-#endif
-
     vp8_inverse_transform_mby(IF_RTCD(&rtcd->common->idct), &x->e_mbd);
 
     RECON_INVOKE(&rtcd->common->recon, recon_mby)
@@ -126,15 +121,9 @@
 
     vp8_quantize_mbuv(x);
 
-#if !(CONFIG_REALTIME_ONLY)
-#if 1
-
     if (x->optimize==2 ||(x->optimize && x->rddiv > 1))
         vp8_optimize_mbuv(x, rtcd);
 
-#endif
-#endif
-
     vp8_inverse_transform_mbuv(IF_RTCD(&rtcd->common->idct), &x->e_mbd);
 
     vp8_recon_intra_mbuv(IF_RTCD(&rtcd->common->recon), &x->e_mbd);
diff --git a/vp8/encoder/encodemb.c b/vp8/encoder/encodemb.c
index 463dbca..2509e06 100644
--- a/vp8/encoder/encodemb.c
+++ b/vp8/encoder/encodemb.c
@@ -19,6 +19,7 @@
 #include "vp8/common/reconintra.h"
 #include "dct.h"
 #include "vpx_mem/vpx_mem.h"
+#include "rdopt.h"
 
 #if CONFIG_RUNTIME_CPU_DETECT
 #define IF_RTCD(x) (x)
@@ -195,42 +196,7 @@
 }
 
 
-void vp8_stuff_inter16x16(MACROBLOCK *x)
-{
-    vp8_build_inter_predictors_mb_s(&x->e_mbd);
-    /*
-        // recon = copy from predictors to destination
-        {
-            BLOCKD *b = &x->e_mbd.block[0];
-            unsigned char *pred_ptr = b->predictor;
-            unsigned char *dst_ptr = *(b->base_dst) + b->dst;
-            int stride = b->dst_stride;
 
-            int i;
-            for(i=0;i<16;i++)
-                vpx_memcpy(dst_ptr+i*stride,pred_ptr+16*i,16);
-
-            b = &x->e_mbd.block[16];
-            pred_ptr = b->predictor;
-            dst_ptr = *(b->base_dst) + b->dst;
-            stride = b->dst_stride;
-
-            for(i=0;i<8;i++)
-                vpx_memcpy(dst_ptr+i*stride,pred_ptr+8*i,8);
-
-            b = &x->e_mbd.block[20];
-            pred_ptr = b->predictor;
-            dst_ptr = *(b->base_dst) + b->dst;
-            stride = b->dst_stride;
-
-            for(i=0;i<8;i++)
-                vpx_memcpy(dst_ptr+i*stride,pred_ptr+8*i,8);
-        }
-    */
-}
-
-#if !(CONFIG_REALTIME_ONLY)
-#define RDCOST(RM,DM,R,D) ( ((128+(R)*(RM)) >> 8) + (DM)*(D) )
 #define RDTRUNC(RM,DM,R,D) ( (128+(R)*(RM)) & 0xFF )
 
 typedef struct vp8_token_state vp8_token_state;
@@ -608,7 +574,6 @@
             ta + vp8_block2above[b], tl + vp8_block2left[b], rtcd);
     }
 }
-#endif
 
 void vp8_encode_inter16x16(const VP8_ENCODER_RTCD *rtcd, MACROBLOCK *x)
 {
@@ -620,10 +585,8 @@
 
     vp8_quantize_mb(x);
 
-#if !(CONFIG_REALTIME_ONLY)
     if (x->optimize)
         optimize_mb(x, rtcd);
-#endif
 
     vp8_inverse_transform_mb(IF_RTCD(&rtcd->common->idct), &x->e_mbd);
 
@@ -635,7 +598,7 @@
 /* this funciton is used by first pass only */
 void vp8_encode_inter16x16y(const VP8_ENCODER_RTCD *rtcd, MACROBLOCK *x)
 {
-    vp8_build_inter_predictors_mby(&x->e_mbd);
+    vp8_build_inter16x16_predictors_mby(&x->e_mbd);
 
     ENCODEMB_INVOKE(&rtcd->encodemb, submby)(x->src_diff, x->src.y_buffer, x->e_mbd.predictor, x->src.y_stride);
 
diff --git a/vp8/encoder/encodemb.h b/vp8/encoder/encodemb.h
index 8c93aa1..47fc72d 100644
--- a/vp8/encoder/encodemb.h
+++ b/vp8/encoder/encodemb.h
@@ -95,8 +95,6 @@
 struct VP8_ENCODER_RTCD;
 void vp8_encode_inter16x16(const struct VP8_ENCODER_RTCD *rtcd, MACROBLOCK *x);
 
-extern void vp8_stuff_inter16x16(MACROBLOCK *x);
-
 void vp8_build_dcblock(MACROBLOCK *b);
 void vp8_transform_mb(MACROBLOCK *mb);
 void vp8_transform_mbuv(MACROBLOCK *x);
diff --git a/vp8/encoder/firstpass.c b/vp8/encoder/firstpass.c
index 8f24a11..6f33099 100644
--- a/vp8/encoder/firstpass.c
+++ b/vp8/encoder/firstpass.c
@@ -786,7 +786,8 @@
 
         // TODO:  handle the case when duration is set to 0, or something less
         // than the full time between subsequent cpi->source_time_stamp s  .
-        fps.duration = cpi->source_end_time_stamp - cpi->source_time_stamp;
+        fps.duration = cpi->source->ts_end
+                       - cpi->source->ts_start;
 
         // don't want to do output stats with a stack variable!
         memcpy(cpi->this_frame_stats,
diff --git a/vp8/encoder/generic/csystemdependent.c b/vp8/encoder/generic/csystemdependent.c
index 1d672be..d48c95b 100644
--- a/vp8/encoder/generic/csystemdependent.c
+++ b/vp8/encoder/generic/csystemdependent.c
@@ -89,9 +89,7 @@
 
     cpi->rtcd.quantize.quantb                = vp8_regular_quantize_b;
     cpi->rtcd.quantize.fastquantb            = vp8_fast_quantize_b_c;
-#if !(CONFIG_REALTIME_ONLY)
     cpi->rtcd.search.full_search             = vp8_full_search_sad;
-#endif
     cpi->rtcd.search.diamond_search          = vp8_diamond_search_sad;
 #if !(CONFIG_REALTIME_ONLY)
     cpi->rtcd.temporal.apply                 = vp8_temporal_filter_apply_c;
diff --git a/vp8/encoder/lookahead.c b/vp8/encoder/lookahead.c
new file mode 100644
index 0000000..3b86d40
--- /dev/null
+++ b/vp8/encoder/lookahead.c
@@ -0,0 +1,157 @@
+/*
+ *  Copyright (c) 2011 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 <stdlib.h>
+#include "vpx_config.h"
+#include "lookahead.h"
+#include "vp8/common/extend.h"
+
+#define MAX_LAG_BUFFERS (CONFIG_REALTIME_ONLY? 1 : 25)
+
+struct lookahead_ctx
+{
+    unsigned int max_sz;         /* Absolute size of the queue */
+    unsigned int sz;             /* Number of buffers currently in the queue */
+    unsigned int read_idx;       /* Read index */
+    unsigned int write_idx;      /* Write index */
+    struct lookahead_entry *buf; /* Buffer list */
+};
+
+
+/* Return the buffer at the given absolute index and increment the index */
+static struct lookahead_entry *
+pop(struct lookahead_ctx *ctx,
+    unsigned int         *idx)
+{
+    unsigned int            index = *idx;
+    struct lookahead_entry *buf = ctx->buf + index;
+
+    assert(index < ctx->max_sz);
+    if(++index >= ctx->max_sz)
+        index -= ctx->max_sz;
+    *idx = index;
+    return buf;
+}
+
+
+void
+vp8_lookahead_destroy(struct lookahead_ctx *ctx)
+{
+    if(ctx)
+    {
+        if(ctx->buf)
+        {
+            int i;
+
+            for(i = 0; i < ctx->max_sz; i++)
+                vp8_yv12_de_alloc_frame_buffer(&ctx->buf[i].img);
+            free(ctx->buf);
+        }
+        free(ctx);
+    }
+}
+
+
+struct lookahead_ctx*
+vp8_lookahead_init(unsigned int width,
+                   unsigned int height,
+                   unsigned int depth)
+{
+    struct lookahead_ctx *ctx = NULL;
+    int i;
+
+    /* Clamp the lookahead queue depth */
+    if(depth < 1)
+        depth = 1;
+    else if(depth > MAX_LAG_BUFFERS)
+        depth = MAX_LAG_BUFFERS;
+
+    /* Align the buffer dimensions */
+    width = (width + 15) & ~15;
+    height = (height + 15) & ~15;
+
+    /* Allocate the lookahead structures */
+    ctx = calloc(1, sizeof(*ctx));
+    if(ctx)
+    {
+        ctx->max_sz = depth;
+        ctx->buf = calloc(depth, sizeof(*ctx->buf));
+        if(!ctx->buf)
+            goto bail;
+        for(i=0; i<depth; i++)
+            if (vp8_yv12_alloc_frame_buffer(&ctx->buf[i].img, width, height, 16))
+                goto bail;
+    }
+    return ctx;
+bail:
+    vp8_lookahead_destroy(ctx);
+    return NULL;
+}
+
+
+int
+vp8_lookahead_push(struct lookahead_ctx *ctx,
+                   YV12_BUFFER_CONFIG   *src,
+                   int64_t               ts_start,
+                   int64_t               ts_end,
+                   unsigned int          flags)
+{
+    struct lookahead_entry* buf;
+
+    if(ctx->sz + 1 > ctx->max_sz)
+        return 1;
+    ctx->sz++;
+    buf = pop(ctx, &ctx->write_idx);
+    vp8_copy_and_extend_frame(src, &buf->img);
+    buf->ts_start = ts_start;
+    buf->ts_end = ts_end;
+    buf->flags = flags;
+    return 0;
+}
+
+
+struct lookahead_entry*
+vp8_lookahead_pop(struct lookahead_ctx *ctx,
+                  int                   drain)
+{
+    struct lookahead_entry* buf = NULL;
+
+    if(ctx->sz && (drain || ctx->sz == ctx->max_sz))
+    {
+        buf = pop(ctx, &ctx->read_idx);
+        ctx->sz--;
+    }
+    return buf;
+}
+
+
+struct lookahead_entry*
+vp8_lookahead_peek(struct lookahead_ctx *ctx,
+                   int                   index)
+{
+    struct lookahead_entry* buf = NULL;
+
+    assert(index < ctx->max_sz);
+    if(index < ctx->sz)
+    {
+        index += ctx->read_idx;
+        if(index >= ctx->max_sz)
+            index -= ctx->max_sz;
+        buf = ctx->buf + index;
+    }
+    return buf;
+}
+
+
+unsigned int
+vp8_lookahead_depth(struct lookahead_ctx *ctx)
+{
+    return ctx->sz;
+}
diff --git a/vp8/encoder/lookahead.h b/vp8/encoder/lookahead.h
new file mode 100644
index 0000000..a483d7e
--- /dev/null
+++ b/vp8/encoder/lookahead.h
@@ -0,0 +1,101 @@
+/*
+ *  Copyright (c) 2011 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.
+ */
+#ifndef LOOKAHEAD_H
+#define LOOKAHEAD_H
+#include "vpx_scale/yv12config.h"
+#include "vpx/vpx_integer.h"
+
+struct lookahead_entry
+{
+    YV12_BUFFER_CONFIG  img;
+    int64_t             ts_start;
+    int64_t             ts_end;
+    unsigned int        flags;
+};
+
+
+struct lookahead_ctx;
+
+/**\brief Initializes the lookahead stage
+ *
+ * The lookahead stage is a queue of frame buffers on which some analysis
+ * may be done when buffers are enqueued.
+ *
+ *
+ */
+struct lookahead_ctx* vp8_lookahead_init(unsigned int width,
+                                         unsigned int height,
+                                         unsigned int depth
+                                         );
+
+
+/**\brief Destroys the lookahead stage
+ *
+ */
+void vp8_lookahead_destroy(struct lookahead_ctx *ctx);
+
+
+/**\brief Enqueue a source buffer
+ *
+ * This function will copy the source image into a new framebuffer with
+ * the expected stride/border.
+ *
+ * \param[in] ctx       Pointer to the lookahead context
+ * \param[in] src       Pointer to the image to enqueue
+ * \param[in] ts_start  Timestamp for the start of this frame
+ * \param[in] ts_end    Timestamp for the end of this frame
+ * \param[in] flags     Flags set on this frame
+ */
+int
+vp8_lookahead_push(struct lookahead_ctx *ctx,
+                   YV12_BUFFER_CONFIG   *src,
+                   int64_t               ts_start,
+                   int64_t               ts_end,
+                   unsigned int          flags);
+
+
+/**\brief Get the next source buffer to encode
+ *
+ *
+ * \param[in] ctx       Pointer to the lookahead context
+ * \param[in] drain     Flag indicating the buffer should be drained
+ *                      (return a buffer regardless of the current queue depth)
+ *
+ * \retval NULL, if drain set and queue is empty
+ * \retval NULL, if drain not set and queue not of the configured depth
+ *
+ */
+struct lookahead_entry*
+vp8_lookahead_pop(struct lookahead_ctx *ctx,
+                  int                   drain);
+
+
+/**\brief Get a future source buffer to encode
+ *
+ * \param[in] ctx       Pointer to the lookahead context
+ * \param[in] index     Index of the frame to be returned, 0 == next frame
+ *
+ * \retval NULL, if no buffer exists at the specified index
+ *
+ */
+struct lookahead_entry*
+vp8_lookahead_peek(struct lookahead_ctx *ctx,
+                   int                   index);
+
+
+/**\brief Get the number of frames currently in the lookahead queue
+ *
+ * \param[in] ctx       Pointer to the lookahead context
+ */
+unsigned int
+vp8_lookahead_depth(struct lookahead_ctx *ctx);
+
+
+#endif
diff --git a/vp8/encoder/mcomp.c b/vp8/encoder/mcomp.c
index 716f514..651890d 100644
--- a/vp8/encoder/mcomp.c
+++ b/vp8/encoder/mcomp.c
@@ -194,13 +194,13 @@
 #define DIST(r,c) vfp->svf( PRE(r,c), d->pre_stride, SP(c),SP(r), z,b->src_stride,&sse) // returns subpixel variance error function.
 #define IFMVCV(r,c,s,e) if ( c >= minc && c <= maxc && r >= minr && r <= maxr) s else e;
 #define ERR(r,c) (MVC(r,c)+DIST(r,c)) // returns distortion + motion vector cost
-#define CHECK_BETTER(v,r,c) IFMVCV(r,c,{if((v = ERR(r,c)) < besterr) { besterr = v; br=r; bc=c; }}, v=INT_MAX;)// checks if (r,c) has better score than previous best
+#define CHECK_BETTER(v,r,c) IFMVCV(r,c,{thismse = DIST(r,c); if((v = (MVC(r,c)+thismse)) < besterr) { besterr = v; br=r; bc=c; *distortion = thismse; *sse1 = sse; }}, v=INT_MAX;)// checks if (r,c) has better score than previous best
 #define MIN(x,y) (((x)<(y))?(x):(y))
 #define MAX(x,y) (((x)>(y))?(x):(y))
 
 //#define CHECK_BETTER(v,r,c) if((v = ERR(r,c)) < besterr) { besterr = v; br=r; bc=c; }
 
-int vp8_find_best_sub_pixel_step_iteratively(MACROBLOCK *x, BLOCK *b, BLOCKD *d, MV *bestmv, MV *ref_mv, int error_per_bit, const vp8_variance_fn_ptr_t *vfp, int *mvcost[2])
+int vp8_find_best_sub_pixel_step_iteratively(MACROBLOCK *x, BLOCK *b, BLOCKD *d, MV *bestmv, MV *ref_mv, int error_per_bit, const vp8_variance_fn_ptr_t *vfp, int *mvcost[2], int *distortion, unsigned int *sse1)
 {
     unsigned char *y = *(d->base_pre) + d->pre + (bestmv->row) * d->pre_stride + bestmv->col;
     unsigned char *z = (*(b->base_src) + b->src);
@@ -214,6 +214,7 @@
     unsigned int whichdir;
     unsigned int halfiters = 4;
     unsigned int quarteriters = 4;
+    int thismse;
 
     int minc = MAX(x->mv_col_min << 2, (ref_mv->col >> 1) - ((1 << mvlong_width) - 1));
     int maxc = MIN(x->mv_col_max << 2, (ref_mv->col >> 1) + ((1 << mvlong_width) - 1));
@@ -225,7 +226,8 @@
     bestmv->col <<= 3;
 
     // calculate central point error
-    besterr = vfp->vf(y, d->pre_stride, z, b->src_stride, &sse);
+    besterr = vfp->vf(y, d->pre_stride, z, b->src_stride, sse1);
+    *distortion = besterr;
     besterr += mv_err_cost(bestmv, ref_mv, mvcost, error_per_bit);
 
     // TODO: Each subsequent iteration checks at least one point in common with the last iteration could be 2 ( if diag selected)
@@ -314,7 +316,7 @@
 #undef CHECK_BETTER
 #undef MIN
 #undef MAX
-int vp8_find_best_sub_pixel_step(MACROBLOCK *x, BLOCK *b, BLOCKD *d, MV *bestmv, MV *ref_mv, int error_per_bit, const vp8_variance_fn_ptr_t *vfp, int *mvcost[2])
+int vp8_find_best_sub_pixel_step(MACROBLOCK *x, BLOCK *b, BLOCKD *d, MV *bestmv, MV *ref_mv, int error_per_bit, const vp8_variance_fn_ptr_t *vfp, int *mvcost[2], int *distortion, unsigned int *sse1)
 {
     int bestmse = INT_MAX;
     MV startmv;
@@ -325,6 +327,7 @@
     int left, right, up, down, diag;
     unsigned int sse;
     int whichdir ;
+    int thismse;
 
 
     // Trap uncodable vectors
@@ -332,6 +335,7 @@
     {
         bestmv->row <<= 3;
         bestmv->col <<= 3;
+        *distortion = INT_MAX;
         return INT_MAX;
     }
 
@@ -341,51 +345,60 @@
     startmv = *bestmv;
 
     // calculate central point error
-    bestmse = vfp->vf(y, d->pre_stride, z, b->src_stride, &sse);
+    bestmse = vfp->vf(y, d->pre_stride, z, b->src_stride, sse1);
+    *distortion = bestmse;
     bestmse += mv_err_cost(bestmv, ref_mv, mvcost, error_per_bit);
 
     // go left then right and check error
     this_mv.row = startmv.row;
     this_mv.col = ((startmv.col - 8) | 4);
-    left = vfp->svf_halfpix_h(y - 1, d->pre_stride, z, b->src_stride, &sse);
-    left += mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
+    thismse = vfp->svf_halfpix_h(y - 1, d->pre_stride, z, b->src_stride, &sse);
+    left = thismse + mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
 
     if (left < bestmse)
     {
         *bestmv = this_mv;
         bestmse = left;
+        *distortion = thismse;
+        *sse1 = sse;
     }
 
     this_mv.col += 8;
-    right = vfp->svf_halfpix_h(y, d->pre_stride, z, b->src_stride, &sse);
-    right += mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
+    thismse = vfp->svf_halfpix_h(y, d->pre_stride, z, b->src_stride, &sse);
+    right = thismse + mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
 
     if (right < bestmse)
     {
         *bestmv = this_mv;
         bestmse = right;
+        *distortion = thismse;
+        *sse1 = sse;
     }
 
     // go up then down and check error
     this_mv.col = startmv.col;
     this_mv.row = ((startmv.row - 8) | 4);
-    up = vfp->svf_halfpix_v(y - d->pre_stride, d->pre_stride, z, b->src_stride, &sse);
-    up += mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
+    thismse =  vfp->svf_halfpix_v(y - d->pre_stride, d->pre_stride, z, b->src_stride, &sse);
+    up = thismse + mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
 
     if (up < bestmse)
     {
         *bestmv = this_mv;
         bestmse = up;
+        *distortion = thismse;
+        *sse1 = sse;
     }
 
     this_mv.row += 8;
-    down = vfp->svf_halfpix_v(y, d->pre_stride, z, b->src_stride, &sse);
-    down += mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
+    thismse = vfp->svf_halfpix_v(y, d->pre_stride, z, b->src_stride, &sse);
+    down = thismse + mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
 
     if (down < bestmse)
     {
         *bestmv = this_mv;
         bestmse = down;
+        *distortion = thismse;
+        *sse1 = sse;
     }
 
 
@@ -400,32 +413,34 @@
     case 0:
         this_mv.col = (this_mv.col - 8) | 4;
         this_mv.row = (this_mv.row - 8) | 4;
-        diag = vfp->svf_halfpix_hv(y - 1 - d->pre_stride, d->pre_stride, z, b->src_stride, &sse);
+        thismse = vfp->svf_halfpix_hv(y - 1 - d->pre_stride, d->pre_stride, z, b->src_stride, &sse);
         break;
     case 1:
         this_mv.col += 4;
         this_mv.row = (this_mv.row - 8) | 4;
-        diag = vfp->svf_halfpix_hv(y - d->pre_stride, d->pre_stride, z, b->src_stride, &sse);
+        thismse = vfp->svf_halfpix_hv(y - d->pre_stride, d->pre_stride, z, b->src_stride, &sse);
         break;
     case 2:
         this_mv.col = (this_mv.col - 8) | 4;
         this_mv.row += 4;
-        diag = vfp->svf_halfpix_hv(y - 1, d->pre_stride, z, b->src_stride, &sse);
+        thismse = vfp->svf_halfpix_hv(y - 1, d->pre_stride, z, b->src_stride, &sse);
         break;
     case 3:
     default:
         this_mv.col += 4;
         this_mv.row += 4;
-        diag = vfp->svf_halfpix_hv(y, d->pre_stride, z, b->src_stride, &sse);
+        thismse = vfp->svf_halfpix_hv(y, d->pre_stride, z, b->src_stride, &sse);
         break;
     }
 
-    diag += mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
+    diag = thismse + mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
 
     if (diag < bestmse)
     {
         *bestmv = this_mv;
         bestmse = diag;
+        *distortion = thismse;
+        *sse1 = sse;
     }
 
 //  }
@@ -448,30 +463,34 @@
     if (startmv.col & 7)
     {
         this_mv.col = startmv.col - 2;
-        left = vfp->svf(y, d->pre_stride, this_mv.col & 7, this_mv.row & 7, z, b->src_stride, &sse);
+        thismse = vfp->svf(y, d->pre_stride, this_mv.col & 7, this_mv.row & 7, z, b->src_stride, &sse);
     }
     else
     {
         this_mv.col = (startmv.col - 8) | 6;
-        left = vfp->svf(y - 1, d->pre_stride, 6, this_mv.row & 7, z, b->src_stride, &sse);
+        thismse = vfp->svf(y - 1, d->pre_stride, 6, this_mv.row & 7, z, b->src_stride, &sse);
     }
 
-    left += mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
+    left = thismse + mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
 
     if (left < bestmse)
     {
         *bestmv = this_mv;
         bestmse = left;
+        *distortion = thismse;
+        *sse1 = sse;
     }
 
     this_mv.col += 4;
-    right = vfp->svf(y, d->pre_stride, this_mv.col & 7, this_mv.row & 7, z, b->src_stride, &sse);
-    right += mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
+    thismse = vfp->svf(y, d->pre_stride, this_mv.col & 7, this_mv.row & 7, z, b->src_stride, &sse);
+    right = thismse + mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
 
     if (right < bestmse)
     {
         *bestmv = this_mv;
         bestmse = right;
+        *distortion = thismse;
+        *sse1 = sse;
     }
 
     // go up then down and check error
@@ -480,30 +499,34 @@
     if (startmv.row & 7)
     {
         this_mv.row = startmv.row - 2;
-        up = vfp->svf(y, d->pre_stride, this_mv.col & 7, this_mv.row & 7, z, b->src_stride, &sse);
+        thismse = vfp->svf(y, d->pre_stride, this_mv.col & 7, this_mv.row & 7, z, b->src_stride, &sse);
     }
     else
     {
         this_mv.row = (startmv.row - 8) | 6;
-        up = vfp->svf(y - d->pre_stride, d->pre_stride, this_mv.col & 7, 6, z, b->src_stride, &sse);
+        thismse = vfp->svf(y - d->pre_stride, d->pre_stride, this_mv.col & 7, 6, z, b->src_stride, &sse);
     }
 
-    up += mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
+    up = thismse + mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
 
     if (up < bestmse)
     {
         *bestmv = this_mv;
         bestmse = up;
+        *distortion = thismse;
+        *sse1 = sse;
     }
 
     this_mv.row += 4;
-    down = vfp->svf(y, d->pre_stride, this_mv.col & 7, this_mv.row & 7, z, b->src_stride, &sse);
-    down += mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
+    thismse = vfp->svf(y, d->pre_stride, this_mv.col & 7, this_mv.row & 7, z, b->src_stride, &sse);
+    down = thismse + mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
 
     if (down < bestmse)
     {
         *bestmv = this_mv;
         bestmse = down;
+        *distortion = thismse;
+        *sse1 = sse;
     }
 
 
@@ -525,12 +548,12 @@
             if (startmv.col & 7)
             {
                 this_mv.col -= 2;
-                diag = vfp->svf(y, d->pre_stride, this_mv.col & 7, this_mv.row & 7, z, b->src_stride, &sse);
+                thismse = vfp->svf(y, d->pre_stride, this_mv.col & 7, this_mv.row & 7, z, b->src_stride, &sse);
             }
             else
             {
                 this_mv.col = (startmv.col - 8) | 6;
-                diag = vfp->svf(y - 1, d->pre_stride, 6, this_mv.row & 7, z, b->src_stride, &sse);;
+                thismse = vfp->svf(y - 1, d->pre_stride, 6, this_mv.row & 7, z, b->src_stride, &sse);;
             }
         }
         else
@@ -540,12 +563,12 @@
             if (startmv.col & 7)
             {
                 this_mv.col -= 2;
-                diag = vfp->svf(y - d->pre_stride, d->pre_stride, this_mv.col & 7, 6, z, b->src_stride, &sse);
+                thismse = vfp->svf(y - d->pre_stride, d->pre_stride, this_mv.col & 7, 6, z, b->src_stride, &sse);
             }
             else
             {
                 this_mv.col = (startmv.col - 8) | 6;
-                diag = vfp->svf(y - d->pre_stride - 1, d->pre_stride, 6, 6, z, b->src_stride, &sse);
+                thismse = vfp->svf(y - d->pre_stride - 1, d->pre_stride, 6, 6, z, b->src_stride, &sse);
             }
         }
 
@@ -556,12 +579,12 @@
         if (startmv.row & 7)
         {
             this_mv.row -= 2;
-            diag = vfp->svf(y, d->pre_stride, this_mv.col & 7, this_mv.row & 7, z, b->src_stride, &sse);
+            thismse = vfp->svf(y, d->pre_stride, this_mv.col & 7, this_mv.row & 7, z, b->src_stride, &sse);
         }
         else
         {
             this_mv.row = (startmv.row - 8) | 6;
-            diag = vfp->svf(y - d->pre_stride, d->pre_stride, this_mv.col & 7, 6, z, b->src_stride, &sse);
+            thismse = vfp->svf(y - d->pre_stride, d->pre_stride, this_mv.col & 7, 6, z, b->src_stride, &sse);
         }
 
         break;
@@ -571,36 +594,36 @@
         if (startmv.col & 7)
         {
             this_mv.col -= 2;
-            diag = vfp->svf(y, d->pre_stride, this_mv.col & 7, this_mv.row & 7, z, b->src_stride, &sse);
+            thismse = vfp->svf(y, d->pre_stride, this_mv.col & 7, this_mv.row & 7, z, b->src_stride, &sse);
         }
         else
         {
             this_mv.col = (startmv.col - 8) | 6;
-            diag = vfp->svf(y - 1, d->pre_stride, 6, this_mv.row & 7, z, b->src_stride, &sse);;
+            thismse = vfp->svf(y - 1, d->pre_stride, 6, this_mv.row & 7, z, b->src_stride, &sse);;
         }
 
         break;
     case 3:
         this_mv.col += 2;
         this_mv.row += 2;
-        diag = vfp->svf(y, d->pre_stride,  this_mv.col & 7, this_mv.row & 7, z, b->src_stride, &sse);
+        thismse = vfp->svf(y, d->pre_stride,  this_mv.col & 7, this_mv.row & 7, z, b->src_stride, &sse);
         break;
     }
 
-    diag += mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
+    diag = thismse + mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
 
     if (diag < bestmse)
     {
         *bestmv = this_mv;
         bestmse = diag;
+        *distortion = thismse;
+        *sse1 = sse;
     }
 
-//  }
-
     return bestmse;
 }
 
-int vp8_find_best_half_pixel_step(MACROBLOCK *mb, BLOCK *b, BLOCKD *d, MV *bestmv, MV *ref_mv, int error_per_bit, const vp8_variance_fn_ptr_t *vfp, int *mvcost[2])
+int vp8_find_best_half_pixel_step(MACROBLOCK *mb, BLOCK *b, BLOCKD *d, MV *bestmv, MV *ref_mv, int error_per_bit, const vp8_variance_fn_ptr_t *vfp, int *mvcost[2], int *distortion, unsigned int *sse1)
 {
     int bestmse = INT_MAX;
     MV startmv;
@@ -610,12 +633,14 @@
     unsigned char *z = (*(b->base_src) + b->src);
     int left, right, up, down, diag;
     unsigned int sse;
+    int thismse;
 
     // Trap uncodable vectors
     if ((abs((bestmv->col << 3) - ref_mv->col) > MAX_FULL_PEL_VAL) || (abs((bestmv->row << 3) - ref_mv->row) > MAX_FULL_PEL_VAL))
     {
         bestmv->row <<= 3;
         bestmv->col <<= 3;
+        *distortion = INT_MAX;
         return INT_MAX;
     }
 
@@ -625,51 +650,60 @@
     startmv = *bestmv;
 
     // calculate central point error
-    bestmse = vfp->vf(y, d->pre_stride, z, b->src_stride, &sse);
+    bestmse = vfp->vf(y, d->pre_stride, z, b->src_stride, sse1);
+    *distortion = bestmse;
     bestmse += mv_err_cost(bestmv, ref_mv, mvcost, error_per_bit);
 
     // go left then right and check error
     this_mv.row = startmv.row;
     this_mv.col = ((startmv.col - 8) | 4);
-    left = vfp->svf_halfpix_h(y - 1, d->pre_stride, z, b->src_stride, &sse);
-    left += mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
+    thismse = vfp->svf_halfpix_h(y - 1, d->pre_stride, z, b->src_stride, &sse);
+    left = thismse + mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
 
     if (left < bestmse)
     {
         *bestmv = this_mv;
         bestmse = left;
+        *distortion = thismse;
+        *sse1 = sse;
     }
 
     this_mv.col += 8;
-    right = vfp->svf_halfpix_h(y, d->pre_stride, z, b->src_stride, &sse);
-    right += mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
+    thismse = vfp->svf_halfpix_h(y, d->pre_stride, z, b->src_stride, &sse);
+    right = thismse + mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
 
     if (right < bestmse)
     {
         *bestmv = this_mv;
         bestmse = right;
+        *distortion = thismse;
+        *sse1 = sse;
     }
 
     // go up then down and check error
     this_mv.col = startmv.col;
     this_mv.row = ((startmv.row - 8) | 4);
-    up = vfp->svf_halfpix_v(y - d->pre_stride, d->pre_stride, z, b->src_stride, &sse);
-    up += mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
+    thismse = vfp->svf_halfpix_v(y - d->pre_stride, d->pre_stride, z, b->src_stride, &sse);
+    up = thismse + mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
 
     if (up < bestmse)
     {
         *bestmv = this_mv;
         bestmse = up;
+        *distortion = thismse;
+        *sse1 = sse;
     }
 
     this_mv.row += 8;
-    down = vfp->svf_halfpix_v(y, d->pre_stride, z, b->src_stride, &sse);
-    down += mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
+    thismse = vfp->svf_halfpix_v(y, d->pre_stride, z, b->src_stride, &sse);
+    down = thismse + mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
 
     if (down < bestmse)
     {
         *bestmv = this_mv;
         bestmse = down;
+        *distortion = thismse;
+        *sse1 = sse;
     }
 
     // somewhat strangely not doing all the diagonals for half pel is slower than doing them.
@@ -713,44 +747,52 @@
 #else
     this_mv.col = (this_mv.col - 8) | 4;
     this_mv.row = (this_mv.row - 8) | 4;
-    diag = vfp->svf_halfpix_hv(y - 1 - d->pre_stride, d->pre_stride, z, b->src_stride, &sse);
-    diag += mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
+    thismse = vfp->svf_halfpix_hv(y - 1 - d->pre_stride, d->pre_stride, z, b->src_stride, &sse);
+    diag = thismse + mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
 
     if (diag < bestmse)
     {
         *bestmv = this_mv;
         bestmse = diag;
+        *distortion = thismse;
+        *sse1 = sse;
     }
 
     this_mv.col += 8;
-    diag = vfp->svf_halfpix_hv(y - d->pre_stride, d->pre_stride, z, b->src_stride, &sse);
-    diag += mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
+    thismse = vfp->svf_halfpix_hv(y - d->pre_stride, d->pre_stride, z, b->src_stride, &sse);
+    diag = thismse + mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
 
     if (diag < bestmse)
     {
         *bestmv = this_mv;
         bestmse = diag;
+        *distortion = thismse;
+        *sse1 = sse;
     }
 
     this_mv.col = (this_mv.col - 8) | 4;
     this_mv.row = startmv.row + 4;
-    diag = vfp->svf_halfpix_hv(y - 1, d->pre_stride, z, b->src_stride, &sse);
-    diag += mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
+    thismse = vfp->svf_halfpix_hv(y - 1, d->pre_stride, z, b->src_stride, &sse);
+    diag = thismse + mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
 
     if (diag < bestmse)
     {
         *bestmv = this_mv;
         bestmse = diag;
+        *distortion = thismse;
+        *sse1 = sse;
     }
 
     this_mv.col += 8;
-    diag = vfp->svf_halfpix_hv(y, d->pre_stride, z, b->src_stride, &sse);
-    diag += mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
+    thismse = vfp->svf_halfpix_hv(y, d->pre_stride, z, b->src_stride, &sse);
+    diag = thismse + mv_err_cost(&this_mv, ref_mv, mvcost, error_per_bit);
 
     if (diag < bestmse)
     {
         *bestmv = this_mv;
         bestmse = diag;
+        *distortion = thismse;
+        *sse1 = sse;
     }
 
 #endif
@@ -1190,8 +1232,6 @@
     + mv_err_cost(&this_mv, center_mv, mvcost, error_per_bit);
 }
 
-
-#if !(CONFIG_REALTIME_ONLY)
 int vp8_full_search_sad(MACROBLOCK *x, BLOCK *b, BLOCKD *d, MV *ref_mv, int error_per_bit, int distance, vp8_variance_fn_ptr_t *fn_ptr, int *mvcost[2], MV *center_mv)
 {
     unsigned char *what = (*(b->base_src) + b->src);
@@ -1571,7 +1611,6 @@
     else
         return INT_MAX;
 }
-#endif /* !(CONFIG_REALTIME_ONLY) */
 
 #ifdef ENTROPY_STATS
 void print_mode_context(void)
diff --git a/vp8/encoder/mcomp.h b/vp8/encoder/mcomp.h
index 5efcec2..b14cbcb 100644
--- a/vp8/encoder/mcomp.h
+++ b/vp8/encoder/mcomp.h
@@ -49,7 +49,7 @@
 
 typedef int (fractional_mv_step_fp)
     (MACROBLOCK *x, BLOCK *b, BLOCKD *d, MV *bestmv, MV *ref_mv,
-     int error_per_bit, const vp8_variance_fn_ptr_t *vfp, int *mvcost[2]);
+     int error_per_bit, const vp8_variance_fn_ptr_t *vfp, int *mvcost[2], int *distortion, unsigned int *sse);
 extern fractional_mv_step_fp vp8_find_best_sub_pixel_step_iteratively;
 extern fractional_mv_step_fp vp8_find_best_sub_pixel_step;
 extern fractional_mv_step_fp vp8_find_best_half_pixel_step;
diff --git a/vp8/encoder/onyx_if.c b/vp8/encoder/onyx_if.c
index a18447d..4208f1f 100644
--- a/vp8/encoder/onyx_if.c
+++ b/vp8/encoder/onyx_if.c
@@ -71,7 +71,7 @@
 int vp8_estimate_entropy_savings(VP8_COMP *cpi);
 int vp8_calc_ss_err(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest, const vp8_variance_rtcd_vtable_t *rtcd);
 
-extern void vp8_temporal_filter_prepare_c(VP8_COMP *cpi);
+extern void vp8_temporal_filter_prepare_c(VP8_COMP *cpi, int distance);
 
 static void set_default_lf_deltas(VP8_COMP *cpi);
 
@@ -287,16 +287,9 @@
     vp8_yv12_de_alloc_frame_buffer(&cpi->last_frame_uf);
     vp8_yv12_de_alloc_frame_buffer(&cpi->scaled_source);
 #if VP8_TEMPORAL_ALT_REF
-    vp8_yv12_de_alloc_frame_buffer(&cpi->alt_ref_buffer.source_buffer);
+    vp8_yv12_de_alloc_frame_buffer(&cpi->alt_ref_buffer);
 #endif
-    {
-        int i;
-
-        for (i = 0; i < MAX_LAG_BUFFERS; i++)
-            vp8_yv12_de_alloc_frame_buffer(&cpi->src_buffer[i].source_buffer);
-
-        cpi->source_buffer_count = 0;
-    }
+    vp8_lookahead_destroy(cpi->lookahead);
 
     vpx_free(cpi->tok);
     cpi->tok = 0;
@@ -1252,35 +1245,23 @@
 }
 static void alloc_raw_frame_buffers(VP8_COMP *cpi)
 {
-    int i, buffers;
-    /* allocate source_buffer to be multiples of 16 */
     int width = (cpi->oxcf.Width + 15) & ~15;
+    int height = (cpi->oxcf.Height + 15) & ~15;
 
-    buffers = cpi->oxcf.lag_in_frames;
-
-    if (buffers > MAX_LAG_BUFFERS)
-        buffers = MAX_LAG_BUFFERS;
-
-    if (buffers < 1)
-        buffers = 1;
-
-    for (i = 0; i < buffers; i++)
-        if (vp8_yv12_alloc_frame_buffer(&cpi->src_buffer[i].source_buffer,
-                                        width, cpi->oxcf.Height,
-                                        16))
-            vpx_internal_error(&cpi->common.error, VPX_CODEC_MEM_ERROR,
-                               "Failed to allocate lag buffer");
+    cpi->lookahead = vp8_lookahead_init(cpi->oxcf.Width, cpi->oxcf.Height,
+                                        cpi->oxcf.lag_in_frames);
+    if(!cpi->lookahead)
+        vpx_internal_error(&cpi->common.error, VPX_CODEC_MEM_ERROR,
+                           "Failed to allocate lag buffers");
 
 #if VP8_TEMPORAL_ALT_REF
 
-    if (vp8_yv12_alloc_frame_buffer(&cpi->alt_ref_buffer.source_buffer,
-                                    width, cpi->oxcf.Height, 16))
+    if (vp8_yv12_alloc_frame_buffer(&cpi->alt_ref_buffer,
+                                    width, height, 16))
         vpx_internal_error(&cpi->common.error, VPX_CODEC_MEM_ERROR,
                            "Failed to allocate altref buffer");
 
 #endif
-
-    cpi->source_buffer_count = 0;
 }
 
 static int vp8_alloc_partition_data(VP8_COMP *cpi)
@@ -1478,10 +1459,6 @@
     cpi->avg_frame_qindex             = cpi->oxcf.worst_allowed_q;
 
     // Initialise the starting buffer levels
-    cpi->oxcf.starting_buffer_level =
-        rescale(cpi->oxcf.starting_buffer_level,
-                cpi->oxcf.target_bandwidth, 1000);
-
     cpi->buffer_level                 = cpi->oxcf.starting_buffer_level;
     cpi->bits_off_target              = cpi->oxcf.starting_buffer_level;
 
@@ -1542,7 +1519,6 @@
 
         break;
 
-#if !(CONFIG_REALTIME_ONLY)
     case MODE_GOODQUALITY:
         cpi->pass = 0;
         cpi->compressor_speed = 1;
@@ -1583,7 +1559,6 @@
         cpi->pass = 2;
         cpi->compressor_speed = 0;
         break;
-#endif
     }
 
     if (cpi->pass == 0)
@@ -1656,6 +1631,10 @@
     // Convert target bandwidth from Kbit/s to Bit/s
     cpi->oxcf.target_bandwidth       *= 1000;
 
+    cpi->oxcf.starting_buffer_level =
+        rescale(cpi->oxcf.starting_buffer_level,
+                cpi->oxcf.target_bandwidth, 1000);
+
     // Set or reset optimal and maximum buffer levels.
     if (cpi->oxcf.optimal_buffer_level == 0)
         cpi->oxcf.optimal_buffer_level = cpi->oxcf.target_bandwidth / 8;
@@ -1720,9 +1699,6 @@
     cm->horiz_scale  = cpi->horiz_scale;
     cm->vert_scale   = cpi->vert_scale ;
 
-    // As per VP8
-    cpi->intra_frame_target = (4 * (cm->Width + cm->Height) / 15) * 1000;
-
     // VP8 sharpness level mapping 0-7 (vs 0-10 in general VPx dialogs)
     if (cpi->oxcf.Sharpness > 7)
         cpi->oxcf.Sharpness = 7;
@@ -1752,10 +1728,6 @@
         vp8_alloc_compressor_data(cpi);
     }
 
-    // Clamp KF frame size to quarter of data rate
-    if (cpi->intra_frame_target > cpi->target_bandwidth >> 2)
-        cpi->intra_frame_target = cpi->target_bandwidth >> 2;
-
     if (cpi->oxcf.fixed_q >= 0)
     {
         cpi->last_q[0] = cpi->oxcf.fixed_q;
@@ -1774,7 +1746,7 @@
         cpi->oxcf.lag_in_frames = MAX_LAG_BUFFERS;
 
     // YX Temp
-    cpi->last_alt_ref_sei    = -1;
+    cpi->alt_ref_source = NULL;
     cpi->is_src_frame_alt_ref = 0;
     cpi->is_next_src_alt_ref = 0;
 
@@ -1981,7 +1953,6 @@
 
     cpi->frames_till_gf_update_due      = 0;
     cpi->key_frame_count              = 1;
-    cpi->tot_key_frame_bits            = 0;
 
     cpi->ni_av_qi                     = cpi->oxcf.worst_allowed_q;
     cpi->ni_tot_qi                    = 0;
@@ -2007,7 +1978,6 @@
 
     for (i = 0; i < KEY_FRAME_CONTEXT; i++)
     {
-        cpi->prior_key_frame_size[i]     = cpi->intra_frame_target;
         cpi->prior_key_frame_distance[i] = (int)cpi->output_frame_rate;
     }
 
@@ -2117,15 +2087,11 @@
     cpi->fn_ptr[BLOCK_4X4].sdx8f          = VARIANCE_INVOKE(&cpi->rtcd.variance, sad4x4x8);
     cpi->fn_ptr[BLOCK_4X4].sdx4df         = VARIANCE_INVOKE(&cpi->rtcd.variance, sad4x4x4d);
 
-#if !(CONFIG_REALTIME_ONLY)
     cpi->full_search_sad = SEARCH_INVOKE(&cpi->rtcd.search, full_search);
-#endif
     cpi->diamond_search_sad = SEARCH_INVOKE(&cpi->rtcd.search, diamond_search);
 
     cpi->ready_for_new_frame = 1;
 
-    cpi->source_encode_index = 0;
-
     // make sure frame 1 is okay
     cpi->error_bins[0] = cpi->common.MBs;
 
@@ -2173,7 +2139,8 @@
         if (cpi->pass != 1)
         {
             FILE *f = fopen("opsnr.stt", "a");
-            double time_encoded = (cpi->source_end_time_stamp - cpi->first_time_stamp_ever) / 10000000.000;
+            double time_encoded = (cpi->last_end_time_stamp_seen
+                                   - cpi->first_time_stamp_ever) / 10000000.000;
             double total_encode_time = (cpi->time_receive_data + cpi->time_compress_data)   / 1000.000;
             double dr = (double)cpi->bytes * (double) 8 / (double)1000  / time_encoded;
 
@@ -2186,7 +2153,7 @@
                 double total_ssim = 100 * pow(cpi->summed_quality / cpi->summed_weights, 8.0);
 
                 fprintf(f, "Bitrate\tAVGPsnr\tGLBPsnr\tAVPsnrP\tGLPsnrP\tVPXSSIM\t  Time(us)\n");
-                fprintf(f, "%7.3f\t%7.3f\t%7.3f\t%7.3f\t%7.3f\t%7.3f %8.0f\n",
+                fprintf(f, "%7.3f\t%7.3f\t%7.3f\t%7.3f\t%7.3f\t%7.3f\t%8.0f\n",
                         dr, cpi->total / cpi->count, total_psnr, cpi->totalp / cpi->count, total_psnr2, total_ssim,
                         total_encode_time);
             }
@@ -2628,37 +2595,13 @@
         vp8_scale_frame(sd, &cpi->scaled_source, cm->temp_scale_frame.y_buffer,
                         tmp_height, hs, hr, vs, vr, 0);
 
+        vp8_yv12_extend_frame_borders(&cpi->scaled_source);
         cpi->Source = &cpi->scaled_source;
 #endif
     }
-    // we may need to copy to a buffer so we can extend the image...
-    else if (cm->Width != cm->yv12_fb[cm->lst_fb_idx].y_width ||
-             cm->Height != cm->yv12_fb[cm->lst_fb_idx].y_height)
-    {
-        //vp8_yv12_copy_frame_ptr(sd, &cpi->scaled_source);
-#if HAVE_ARMV7
-#if CONFIG_RUNTIME_CPU_DETECT
-        if (cm->rtcd.flags & HAS_NEON)
-#endif
-        {
-            vp8_yv12_copy_src_frame_func_neon(sd, &cpi->scaled_source);
-        }
-#if CONFIG_RUNTIME_CPU_DETECT
-        else
-#endif
-#endif
-#if !HAVE_ARMV7 || CONFIG_RUNTIME_CPU_DETECT
-        {
-            vp8_yv12_copy_frame_ptr(sd, &cpi->scaled_source);
-        }
-#endif
-
-        cpi->Source = &cpi->scaled_source;
-    }
-
-    vp8_extend_to_multiple_of16(cpi->Source, cm->Width, cm->Height);
-
 }
+
+
 static void resize_key_frame(VP8_COMP *cpi)
 {
 #if CONFIG_SPATIAL_RESAMPLING
@@ -2718,16 +2661,17 @@
         if (cpi->pass == 2)
             vp8_calc_auto_iframe_target_size(cpi);
 
-        // 1 Pass there is no information on which to base size so use bandwidth per second * fixed fraction
         else
 #endif
-            cpi->this_frame_target = cpi->oxcf.target_bandwidth / 2;
-
-        // in error resilient mode the first frame is bigger since it likely contains
-        // all the static background
-        if (cpi->oxcf.error_resilient_mode == 1 || (cpi->compressor_speed == 2))
         {
-            cpi->this_frame_target *= 3;      // 5;
+            /* 1 Pass there is no information on which to base size so use
+             * bandwidth per second * fraction of the initial buffer
+             * level
+             */
+            cpi->this_frame_target = cpi->oxcf.starting_buffer_level / 2;
+
+            if(cpi->this_frame_target > cpi->oxcf.target_bandwidth * 3 / 2)
+                cpi->this_frame_target = cpi->oxcf.target_bandwidth * 3 / 2;
         }
 
         // Key frame from VFW/auto-keyframe/first frame
@@ -2761,6 +2705,19 @@
         }
     }
 
+    /* Apply limits on keyframe target.
+     *
+     * TODO: move this after consolidating
+     * vp8_calc_iframe_target_size() and vp8_calc_auto_iframe_target_size()
+     */
+    if (cm->frame_type == KEY_FRAME && cpi->oxcf.rc_max_intra_bitrate_pct)
+    {
+        unsigned int max_rate = cpi->av_per_frame_bandwidth
+                                * cpi->oxcf.rc_max_intra_bitrate_pct / 100;
+
+        if (cpi->this_frame_target > max_rate)
+            cpi->this_frame_target = max_rate;
+    }
     return 1;
 }
 
@@ -3571,7 +3528,6 @@
         if (Adjustment)
         {
             int buff_lvl_step;
-            int tmp_lvl = cpi->buffer_level;
 
             if (cpi->buffer_level < cpi->oxcf.maximum_buffer_size)
             {
@@ -4419,9 +4375,9 @@
         vp8_clear_system_state();  //__asm emms;
 
         if (cpi->total_coded_error_left != 0.0)
-            fprintf(f, "%10d %10d %10d %10d %10d %10d %10d %10d %6ld %6ld"
-                       "%6ld %6ld %6ld %5ld %5ld %5ld %8ld %8.2f %10d %10.3f"
-                       "%10.3f %8ld\n",
+            fprintf(f, "%10d %10d %10d %10d %10d %10d %10d %10d %6d %6d"
+                       "%6d %6d %6d %5d %5d %5d %8d %8.2f %10d %10.3f"
+                       "%10.3f %8d\n",
                        cpi->common.current_video_frame, cpi->this_frame_target,
                        cpi->projected_frame_size,
                        (cpi->projected_frame_size - cpi->this_frame_target),
@@ -4438,9 +4394,9 @@
                        (double)cpi->bits_left / cpi->total_coded_error_left,
                        cpi->tot_recode_hits);
         else
-            fprintf(f, "%10d %10d %10d %10d %10d %10d %10d %10d %6ld %6ld"
-                       "%6ld %6ld %6ld %5ld %5ld %5ld %8ld %8.2f %10d %10.3f"
-                       "%8ld\n",
+            fprintf(f, "%10d %10d %10d %10d %10d %10d %10d %10d %6d %6d"
+                       "%6d %6d %6d %5d %5d %5d %8d %8.2f %10d %10.3f"
+                       "%8d\n",
                        cpi->common.current_video_frame,
                        cpi->this_frame_target, cpi->projected_frame_size,
                        (cpi->projected_frame_size - cpi->this_frame_target),
@@ -4672,17 +4628,17 @@
 extern void vp8_push_neon(INT64 *store);
 extern void vp8_pop_neon(INT64 *store);
 #endif
+
+
 int vp8_receive_raw_frame(VP8_PTR ptr, unsigned int frame_flags, YV12_BUFFER_CONFIG *sd, INT64 time_stamp, INT64 end_time)
 {
 #if HAVE_ARMV7
     INT64 store_reg[8];
 #endif
-    VP8_COMP *cpi = (VP8_COMP *) ptr;
-    VP8_COMMON *cm = &cpi->common;
+    VP8_COMP              *cpi = (VP8_COMP *) ptr;
+    VP8_COMMON            *cm = &cpi->common;
     struct vpx_usec_timer  timer;
-
-    if (!cpi)
-        return -1;
+    int                    res = 0;
 
 #if HAVE_ARMV7
 #if CONFIG_RUNTIME_CPU_DETECT
@@ -4694,75 +4650,10 @@
 #endif
 
     vpx_usec_timer_start(&timer);
-
-    // no more room for frames;
-    if (cpi->source_buffer_count != 0 && cpi->source_buffer_count >= cpi->oxcf.lag_in_frames)
-    {
-#if HAVE_ARMV7
-#if CONFIG_RUNTIME_CPU_DETECT
-        if (cm->rtcd.flags & HAS_NEON)
-#endif
-        {
-            vp8_pop_neon(store_reg);
-        }
-#endif
-        return -1;
-    }
-
-    //printf("in-cpi->source_buffer_count: %d\n", cpi->source_buffer_count);
-
+    if(vp8_lookahead_push(cpi->lookahead, sd, time_stamp, end_time,
+                          frame_flags))
+        res = -1;
     cm->clr_type = sd->clrtype;
-
-    // make a copy of the frame for use later...
-#if !(CONFIG_REALTIME_ONLY)
-
-    if (cpi->oxcf.allow_lag)
-    {
-        int which_buffer =  cpi->source_encode_index - 1;
-        SOURCE_SAMPLE *s;
-
-        if (which_buffer == -1)
-            which_buffer = cpi->oxcf.lag_in_frames - 1;
-
-        if (cpi->source_buffer_count < cpi->oxcf.lag_in_frames - 1)
-            which_buffer = cpi->source_buffer_count;
-
-        s = &cpi->src_buffer[which_buffer];
-
-        s->source_time_stamp = time_stamp;
-        s->source_end_time_stamp = end_time;
-        s->source_frame_flags = frame_flags;
-        vp8_yv12_copy_frame_ptr(sd, &s->source_buffer);
-
-        cpi->source_buffer_count ++;
-    }
-    else
-#endif
-    {
-        SOURCE_SAMPLE *s;
-        s = &cpi->src_buffer[0];
-        s->source_end_time_stamp = end_time;
-        s->source_time_stamp = time_stamp;
-        s->source_frame_flags = frame_flags;
-#if HAVE_ARMV7
-#if CONFIG_RUNTIME_CPU_DETECT
-        if (cm->rtcd.flags & HAS_NEON)
-#endif
-        {
-            vp8_yv12_copy_src_frame_func_neon(sd, &s->source_buffer);
-        }
-#if CONFIG_RUNTIME_CPU_DETECT
-        else
-#endif
-#endif
-#if !HAVE_ARMV7 || CONFIG_RUNTIME_CPU_DETECT
-        {
-            vp8_yv12_copy_frame_ptr(sd, &s->source_buffer);
-        }
-#endif
-        cpi->source_buffer_count = 1;
-    }
-
     vpx_usec_timer_mark(&timer);
     cpi->time_receive_data += vpx_usec_timer_elapsed(&timer);
 
@@ -4775,8 +4666,10 @@
     }
 #endif
 
-    return 0;
+    return res;
 }
+
+
 int vp8_get_compressed_data(VP8_PTR ptr, unsigned int *frame_flags, unsigned long *size, unsigned char *dest, INT64 *time_stamp, INT64 *time_end, int flush)
 {
 #if HAVE_ARMV7
@@ -4787,6 +4680,7 @@
     struct vpx_usec_timer  tsctimer;
     struct vpx_usec_timer  ticktimer;
     struct vpx_usec_timer  cmptimer;
+    YV12_BUFFER_CONFIG    *force_src_buffer = NULL;
 
     if (!cpi)
         return -1;
@@ -4802,95 +4696,24 @@
 
     vpx_usec_timer_start(&cmptimer);
 
-
-    // flush variable tells us that even though we have less than 10 frames
-    // in our buffer we need to start producing compressed frames.
-    // Probably because we are at the end of a file....
-    if ((cpi->source_buffer_count == cpi->oxcf.lag_in_frames && cpi->oxcf.lag_in_frames > 0)
-        || (!cpi->oxcf.allow_lag && cpi->source_buffer_count > 0)
-        || (flush && cpi->source_buffer_count > 0))
-    {
-
-        SOURCE_SAMPLE *s;
-
-        s = &cpi->src_buffer[cpi->source_encode_index];
-        cpi->source_time_stamp = s->source_time_stamp;
-        cpi->source_end_time_stamp = s->source_end_time_stamp;
+    cpi->source = NULL;
 
 #if !(CONFIG_REALTIME_ONLY)
-
-        // Should we code an alternate reference frame
-        if (cpi->oxcf.error_resilient_mode == 0 &&
-            cpi->oxcf.play_alternate &&
-            cpi->source_alt_ref_pending  &&
-            (cpi->frames_till_gf_update_due < cpi->source_buffer_count) &&
-            cpi->oxcf.lag_in_frames != 0)
+    // Should we code an alternate reference frame
+    if (cpi->oxcf.error_resilient_mode == 0 &&
+        cpi->oxcf.play_alternate &&
+        cpi->source_alt_ref_pending)
+    {
+        if ((cpi->source = vp8_lookahead_peek(cpi->lookahead,
+                                              cpi->frames_till_gf_update_due)))
         {
-            cpi->last_alt_ref_sei = (cpi->source_encode_index + cpi->frames_till_gf_update_due) % cpi->oxcf.lag_in_frames;
-
-#if VP8_TEMPORAL_ALT_REF
-
+            cpi->alt_ref_source = cpi->source;
             if (cpi->oxcf.arnr_max_frames > 0)
             {
-#if 0
-                // my attempt at a loop that tests the results of strength filter.
-                int start_frame = cpi->last_alt_ref_sei - 3;
-
-                int i, besti = -1, pastin = cpi->oxcf.arnr_strength;
-
-                int besterr;
-
-                if (start_frame < 0)
-                    start_frame += cpi->oxcf.lag_in_frames;
-
-                besterr = calc_low_ss_err(&cpi->src_buffer[cpi->last_alt_ref_sei].source_buffer,
-                                              &cpi->src_buffer[start_frame].source_buffer, IF_RTCD(&cpi->rtcd.variance));
-
-                for (i = 0; i < 7; i++)
-                {
-                    int thiserr;
-                    cpi->oxcf.arnr_strength = i;
-                    vp8_temporal_filter_prepare_c(cpi);
-
-                    thiserr = calc_low_ss_err(&cpi->alt_ref_buffer.source_buffer,
-                                                  &cpi->src_buffer[start_frame].source_buffer, IF_RTCD(&cpi->rtcd.variance));
-
-                    if (10 * thiserr < besterr * 8)
-                    {
-                        besterr = thiserr;
-                        besti = i;
-                    }
-                }
-
-                if (besti != -1)
-                {
-                    cpi->oxcf.arnr_strength = besti;
-                    vp8_temporal_filter_prepare_c(cpi);
-                    s = &cpi->alt_ref_buffer;
-
-                    // FWG not sure if I need to copy this data for the Alt Ref frame
-                    s->source_time_stamp = cpi->src_buffer[cpi->last_alt_ref_sei].source_time_stamp;
-                    s->source_end_time_stamp = cpi->src_buffer[cpi->last_alt_ref_sei].source_end_time_stamp;
-                    s->source_frame_flags = cpi->src_buffer[cpi->last_alt_ref_sei].source_frame_flags;
-                }
-                else
-                    s = &cpi->src_buffer[cpi->last_alt_ref_sei];
-
-#else
-                vp8_temporal_filter_prepare_c(cpi);
-                s = &cpi->alt_ref_buffer;
-
-                // FWG not sure if I need to copy this data for the Alt Ref frame
-                s->source_time_stamp = cpi->src_buffer[cpi->last_alt_ref_sei].source_time_stamp;
-                s->source_end_time_stamp = cpi->src_buffer[cpi->last_alt_ref_sei].source_end_time_stamp;
-                s->source_frame_flags = cpi->src_buffer[cpi->last_alt_ref_sei].source_frame_flags;
-
-#endif
+                vp8_temporal_filter_prepare_c(cpi,
+                                              cpi->frames_till_gf_update_due);
+                force_src_buffer = &cpi->alt_ref_buffer;
             }
-            else
-#endif
-                s = &cpi->src_buffer[cpi->last_alt_ref_sei];
-
             cm->frames_till_alt_ref_frame = cpi->frames_till_gf_update_due;
             cm->refresh_alt_ref_frame = 1;
             cm->refresh_golden_frame = 0;
@@ -4900,40 +4723,33 @@
             cpi->is_src_frame_alt_ref = 0;
             cpi->is_next_src_alt_ref = 0;
         }
-        else
+    }
 #endif
+
+    if (!cpi->source)
+    {
+        if ((cpi->source = vp8_lookahead_pop(cpi->lookahead, flush)))
         {
             cm->show_frame = 1;
-#if !(CONFIG_REALTIME_ONLY)
 
-            if (cpi->oxcf.allow_lag)
-            {
-                if (cpi->source_encode_index ==  cpi->last_alt_ref_sei)
-                {
-                    cpi->is_src_frame_alt_ref = 1;
-                    cpi->last_alt_ref_sei    = -1;
-                }
-                else
-                    cpi->is_src_frame_alt_ref = 0;
+            cpi->is_src_frame_alt_ref = cpi->alt_ref_source
+                                        && (cpi->source == cpi->alt_ref_source);
 
-                cpi->source_encode_index = (cpi->source_encode_index + 1) % cpi->oxcf.lag_in_frames;
-
-                if(cpi->source_encode_index == cpi->last_alt_ref_sei)
-                    cpi->is_next_src_alt_ref = 1;
-                else
-                    cpi->is_next_src_alt_ref = 0;
-            }
-
-#endif
-            cpi->source_buffer_count--;
+            cpi->is_next_src_alt_ref = cpi->alt_ref_source
+                                        && (vp8_lookahead_peek(cpi->lookahead, 0)
+                                            == cpi->alt_ref_source);
+            if(cpi->is_src_frame_alt_ref)
+                cpi->alt_ref_source = NULL;
         }
+    }
 
-        cpi->un_scaled_source = &s->source_buffer;
-        cpi->Source = &s->source_buffer;
-        cpi->source_frame_flags = s->source_frame_flags;
-
-        *time_stamp = cpi->source_time_stamp;
-        *time_end = cpi->source_end_time_stamp;
+    if (cpi->source)
+    {
+        cpi->un_scaled_source =
+        cpi->Source = force_src_buffer ? force_src_buffer : &cpi->source->img;
+        *time_stamp = cpi->source->ts_start;
+        *time_end = cpi->source->ts_end;
+        *frame_flags = cpi->source->flags;
     }
     else
     {
@@ -4959,26 +4775,24 @@
         return -1;
     }
 
-    *frame_flags = cpi->source_frame_flags;
-
-    if (cpi->source_time_stamp < cpi->first_time_stamp_ever)
+    if (cpi->source->ts_start < cpi->first_time_stamp_ever)
     {
-        cpi->first_time_stamp_ever = cpi->source_time_stamp;
-        cpi->last_end_time_stamp_seen = cpi->source_time_stamp;
+        cpi->first_time_stamp_ever = cpi->source->ts_start;
+        cpi->last_end_time_stamp_seen = cpi->source->ts_start;
     }
 
     // adjust frame rates based on timestamps given
     if (!cm->refresh_alt_ref_frame)
     {
-        if (cpi->source_time_stamp == cpi->first_time_stamp_ever)
+        if (cpi->source->ts_start == cpi->first_time_stamp_ever)
         {
-            double this_fps = 10000000.000 / (cpi->source_end_time_stamp - cpi->source_time_stamp);
+            double this_fps = 10000000.000 / (cpi->source->ts_end - cpi->source->ts_start);
 
             vp8_new_frame_rate(cpi, this_fps);
         }
         else
         {
-            long long nanosecs = cpi->source_end_time_stamp
+            long long nanosecs = cpi->source->ts_end
                 - cpi->last_end_time_stamp_seen;
 
             if (nanosecs > 0)
@@ -4989,8 +4803,8 @@
 
         }
 
-        cpi->last_time_stamp_seen = cpi->source_time_stamp;
-        cpi->last_end_time_stamp_seen = cpi->source_end_time_stamp;
+        cpi->last_time_stamp_seen = cpi->source->ts_start;
+        cpi->last_end_time_stamp_seen = cpi->source->ts_end;
     }
 
     if (cpi->compressor_speed == 2)
@@ -5111,7 +4925,6 @@
 
             if (cpi->b_calculate_psnr)
             {
-                double y, u, v;
                 double ye,ue,ve;
                 double frame_psnr;
                 YV12_BUFFER_CONFIG      *orig = cpi->Source;
@@ -5144,7 +4957,7 @@
                 cpi->total_sq_error += sq_error;
                 cpi->total  += frame_psnr;
                 {
-                    double y2, u2, v2, frame_psnr2, frame_ssim2 = 0;
+                    double frame_psnr2, frame_ssim2 = 0;
                     double weight = 0;
 
                     vp8_deblock(cm->frame_to_show, &cm->post_proc_buffer, cm->filter_level * 10 / 6, 1, 0, IF_RTCD(&cm->rtcd.postproc));
diff --git a/vp8/encoder/onyx_int.h b/vp8/encoder/onyx_int.h
index 0e53f68..fdd3b99 100644
--- a/vp8/encoder/onyx_int.h
+++ b/vp8/encoder/onyx_int.h
@@ -29,6 +29,7 @@
 #include "mcomp.h"
 #include "temporal_filter.h"
 #include "vp8/common/findnearmv.h"
+#include "lookahead.h"
 
 //#define SPEEDSTATS 1
 #define MIN_GF_INTERVAL             4
@@ -217,14 +218,6 @@
     void *ptr1;
 } LPFTHREAD_DATA;
 
-typedef struct
-{
-    INT64  source_time_stamp;
-    INT64  source_end_time_stamp;
-
-    DECLARE_ALIGNED(16, YV12_BUFFER_CONFIG, source_buffer);
-    unsigned int source_frame_flags;
-} SOURCE_SAMPLE;
 
 typedef struct VP8_ENCODER_RTCD
 {
@@ -251,17 +244,17 @@
 {
 
     DECLARE_ALIGNED(16, short, Y1quant[QINDEX_RANGE][16]);
-    DECLARE_ALIGNED(16, short, Y1quant_shift[QINDEX_RANGE][16]);
+    DECLARE_ALIGNED(16, unsigned char, Y1quant_shift[QINDEX_RANGE][16]);
     DECLARE_ALIGNED(16, short, Y1zbin[QINDEX_RANGE][16]);
     DECLARE_ALIGNED(16, short, Y1round[QINDEX_RANGE][16]);
 
     DECLARE_ALIGNED(16, short, Y2quant[QINDEX_RANGE][16]);
-    DECLARE_ALIGNED(16, short, Y2quant_shift[QINDEX_RANGE][16]);
+    DECLARE_ALIGNED(16, unsigned char, Y2quant_shift[QINDEX_RANGE][16]);
     DECLARE_ALIGNED(16, short, Y2zbin[QINDEX_RANGE][16]);
     DECLARE_ALIGNED(16, short, Y2round[QINDEX_RANGE][16]);
 
     DECLARE_ALIGNED(16, short, UVquant[QINDEX_RANGE][16]);
-    DECLARE_ALIGNED(16, short, UVquant_shift[QINDEX_RANGE][16]);
+    DECLARE_ALIGNED(16, unsigned char, UVquant_shift[QINDEX_RANGE][16]);
     DECLARE_ALIGNED(16, short, UVzbin[QINDEX_RANGE][16]);
     DECLARE_ALIGNED(16, short, UVround[QINDEX_RANGE][16]);
 
@@ -280,19 +273,17 @@
 
     VP8_CONFIG oxcf;
 
+    struct lookahead_ctx    *lookahead;
+    struct lookahead_entry  *source;
+    struct lookahead_entry  *alt_ref_source;
+
     YV12_BUFFER_CONFIG *Source;
     YV12_BUFFER_CONFIG *un_scaled_source;
-    INT64 source_time_stamp;
-    INT64 source_end_time_stamp;
-    unsigned int source_frame_flags;
     YV12_BUFFER_CONFIG scaled_source;
 
-    int source_buffer_count;    // number of src_buffers in use for lagged encoding
-    int source_encode_index;    // index of buffer in src_buffer to encode
     int source_alt_ref_pending; // frame in src_buffers has been identified to be encoded as an alt ref
     int source_alt_ref_active;  // an alt ref frame has been encoded and is usable
 
-    int last_alt_ref_sei;       // index into src_buffers of frame used as alt reference
     int is_src_frame_alt_ref;   // source of frame to encode is an exact copy of an alt ref frame
     int is_next_src_alt_ref;    // source of next frame to encode is an exact copy of an alt ref frame
 
@@ -301,8 +292,6 @@
     int gold_is_alt;  // don't do both alt and gold search ( just do gold).
 
     //int refresh_alt_ref_frame;
-    SOURCE_SAMPLE src_buffer[MAX_LAG_BUFFERS];
-
     YV12_BUFFER_CONFIG last_frame_uf;
 
     TOKENEXTRA *tok;
@@ -396,14 +385,11 @@
     int active_arnr_frames;           // <= cpi->oxcf.arnr_max_frames
 
     INT64 key_frame_count;
-    INT64 tot_key_frame_bits;
-    int prior_key_frame_size[KEY_FRAME_CONTEXT];
     int prior_key_frame_distance[KEY_FRAME_CONTEXT];
     int per_frame_bandwidth;          // Current section per frame bandwidth target
     int av_per_frame_bandwidth;        // Average frame size target for clip
     int min_frame_bandwidth;          // Minimum allocation that should be used for any frame
     int last_key_frame_size;
-    int intra_frame_target;
     int inter_frame_target;
     double output_frame_rate;
     long long last_time_stamp_seen;
@@ -638,7 +624,7 @@
     VP8_ENCODER_RTCD            rtcd;
 #endif
 #if VP8_TEMPORAL_ALT_REF
-    SOURCE_SAMPLE alt_ref_buffer;
+    YV12_BUFFER_CONFIG alt_ref_buffer;
     YV12_BUFFER_CONFIG *frames[MAX_LAG_BUFFERS];
     int fixed_divide[512];
 #endif
diff --git a/vp8/encoder/pickinter.c b/vp8/encoder/pickinter.c
index 0759e2d..765ff26 100644
--- a/vp8/encoder/pickinter.c
+++ b/vp8/encoder/pickinter.c
@@ -50,7 +50,7 @@
 extern void vp8_set_mbmode_and_mvs(MACROBLOCK *x, MB_PREDICTION_MODE mb, MV *mv);
 
 
-int vp8_skip_fractional_mv_step(MACROBLOCK *mb, BLOCK *b, BLOCKD *d, MV *bestmv, MV *ref_mv, int error_per_bit, const vp8_variance_fn_ptr_t *vfp, int *mvcost[2])
+int vp8_skip_fractional_mv_step(MACROBLOCK *mb, BLOCK *b, BLOCKD *d, MV *bestmv, MV *ref_mv, int error_per_bit, const vp8_variance_fn_ptr_t *vfp, int *mvcost[2], int *distortion, unsigned int *sse)
 {
     (void) b;
     (void) d;
@@ -58,6 +58,8 @@
     (void) error_per_bit;
     (void) vfp;
     (void) mvcost;
+    (void) distortion;
+    (void) sse;
     bestmv->row <<= 3;
     bestmv->col <<= 3;
     return 0;
@@ -194,7 +196,7 @@
         rate = mode_costs[mode];
         vp8_predict_intra4x4(b, mode, b->predictor);
         distortion = get_prediction_error(be, b, &rtcd->variance);
-        this_rd = RD_ESTIMATE(x->rdmult, x->rddiv, rate, distortion);
+        this_rd = RDCOST(x->rdmult, x->rddiv, rate, distortion);
 
         if (this_rd < best_rd)
         {
@@ -212,7 +214,13 @@
 }
 
 
-int vp8_pick_intra4x4mby_modes(const VP8_ENCODER_RTCD *rtcd, MACROBLOCK *mb, int *Rate, int *best_dist)
+int vp8_pick_intra4x4mby_modes
+(
+    const VP8_ENCODER_RTCD *rtcd,
+    MACROBLOCK *mb,
+    int *Rate,
+    int *best_dist
+)
 {
     MACROBLOCKD *const xd = &mb->e_mbd;
     int i;
@@ -239,20 +247,18 @@
 
         mic->bmi[i].mode = xd->block[i].bmi.mode = best_mode;
 
-        // Break out case where we have already exceeded best so far value that was bassed in
+        // Break out case where we have already exceeded best so far value
+        // that was passed in
         if (distortion > *best_dist)
             break;
     }
 
-    for (i = 0; i < 16; i++)
-        xd->block[i].bmi.mv.as_int = 0;
-
     *Rate = cost;
 
     if (i == 16)
     {
         *best_dist = distortion;
-        error = RD_ESTIMATE(mb->rdmult, mb->rddiv, cost, distortion);
+        error = RDCOST(mb->rdmult, mb->rddiv, cost, distortion);
     }
     else
     {
@@ -260,6 +266,9 @@
         error = INT_MAX;
     }
 
+    for (i = 0; i < 16; i++)
+        xd->block[i].bmi.mv.as_int = 0;
+
     return error;
 }
 
@@ -435,7 +444,7 @@
     int bestsme;
     //int all_rds[MAX_MODES];         // Experimental debug code.
     int best_mode_index = 0;
-    int sse = INT_MAX;
+    unsigned int sse = INT_MAX;
 
     MV mvp;
     int near_sadidx[8] = {0, 1, 2, 3, 4, 5, 6, 7};
@@ -452,6 +461,8 @@
 
     int skip_mode[4] = {0, 0, 0, 0};
 
+    int have_subp_search = cpi->sf.half_pixel_search;  /* In real-time mode, when Speed >= 15, no sub-pixel search. */
+
     vpx_memset(mode_mv, 0, sizeof(mode_mv));
     vpx_memset(nearest_mv, 0, sizeof(nearest_mv));
     vpx_memset(near_mv, 0, sizeof(near_mv));
@@ -632,10 +643,10 @@
         switch (this_mode)
         {
         case B_PRED:
-            distortion2 = *returndistortion;                    // Best so far passed in as breakout value to vp8_pick_intra4x4mby_modes
-            vp8_pick_intra4x4mby_modes(IF_RTCD(&cpi->rtcd), x, &rate, &distortion2);
-            rate2 += rate;
-            distortion2 = VARIANCE_INVOKE(&cpi->rtcd.variance, get16x16prederror)(x->src.y_buffer, x->src.y_stride, x->e_mbd.predictor, 16, 0x7fffffff);
+            // Pass best so far to vp8_pick_intra4x4mby_modes to use as breakout
+            distortion2 = *returndistortion;
+            vp8_pick_intra4x4mby_modes(IF_RTCD(&cpi->rtcd), x,
+                                         &rate, &distortion2);
 
             if (distortion2 == INT_MAX)
             {
@@ -643,7 +654,12 @@
             }
             else
             {
-                this_rd = RD_ESTIMATE(x->rdmult, x->rddiv, rate2, distortion2);
+                rate2 += rate;
+                distortion2 = VARIANCE_INVOKE
+                                (&cpi->rtcd.variance, get16x16prederror)(
+                                    x->src.y_buffer, x->src.y_stride,
+                                    x->e_mbd.predictor, 16, 0x7fffffff);
+                this_rd = RDCOST(x->rdmult, x->rddiv, rate2, distortion2);
 
                 if (this_rd < best_intra_rd)
                 {
@@ -667,7 +683,7 @@
                 (&x->e_mbd);
             distortion2 = VARIANCE_INVOKE(&cpi->rtcd.variance, get16x16prederror)(x->src.y_buffer, x->src.y_stride, x->e_mbd.predictor, 16, 0x7fffffff);
             rate2 += x->mbmode_cost[x->e_mbd.frame_type][x->e_mbd.mode_info_context->mbmi.mode];
-            this_rd = RD_ESTIMATE(x->rdmult, x->rddiv, rate2, distortion2);
+            this_rd = RDCOST(x->rdmult, x->rddiv, rate2, distortion2);
 
             if (this_rd < best_intra_rd)
             {
@@ -781,7 +797,7 @@
             }
 
             if (bestsme < INT_MAX)
-                cpi->find_fractional_mv_step(x, b, d, &d->bmi.mv.as_mv, &best_ref_mv, x->errorperbit, &cpi->fn_ptr[BLOCK_16X16], cpi->mb.mvcost);
+                cpi->find_fractional_mv_step(x, b, d, &d->bmi.mv.as_mv, &best_ref_mv, x->errorperbit, &cpi->fn_ptr[BLOCK_16X16], cpi->mb.mvcost, &distortion2, &sse);
 
             mode_mv[NEWMV].row = d->bmi.mv.as_mv.row;
             mode_mv[NEWMV].col = d->bmi.mv.as_mv.col;
@@ -811,9 +827,10 @@
             x->e_mbd.block[0].bmi.mode = this_mode;
             x->e_mbd.block[0].bmi.mv.as_int = x->e_mbd.mode_info_context->mbmi.mv.as_int;
 
-            distortion2 = get_inter_mbpred_error(x, &cpi->fn_ptr[BLOCK_16X16], (unsigned int *)(&sse));
+            if((this_mode != NEWMV) || !(have_subp_search))
+                distortion2 = get_inter_mbpred_error(x, &cpi->fn_ptr[BLOCK_16X16], &sse);
 
-            this_rd = RD_ESTIMATE(x->rdmult, x->rddiv, rate2, distortion2);
+            this_rd = RDCOST(x->rdmult, x->rddiv, rate2, distortion2);
 
             if (cpi->active_map_enabled && x->active_ptr[0] == 0)
             {
@@ -921,7 +938,6 @@
         best_mbmode.uv_mode = 0;
         best_mbmode.mb_skip_coeff = (cpi->common.mb_no_coeff_skip) ? 1 : 0;
         best_mbmode.partitioning = 0;
-        best_mbmode.dc_diff = 0;
 
         vpx_memcpy(&x->e_mbd.mode_info_context->mbmi, &best_mbmode, sizeof(MB_MODE_INFO));
         vpx_memcpy(x->partition_info, &best_partition, sizeof(PARTITION_INFO));
@@ -932,6 +948,7 @@
         }
 
         x->e_mbd.mode_info_context->mbmi.mv.as_int = 0;
+        return;
     }
 
 
diff --git a/vp8/encoder/pickinter.h b/vp8/encoder/pickinter.h
index 8fea983..f96fc53 100644
--- a/vp8/encoder/pickinter.h
+++ b/vp8/encoder/pickinter.h
@@ -14,7 +14,6 @@
 #include "vpx_ports/config.h"
 #include "vp8/common/onyxc_int.h"
 
-#define RD_ESTIMATE(RM,DM,R,D) ( ((128+(R)*(RM)) >> 8) + (DM)*(D) )
 extern int vp8_pick_intra4x4mby_modes(const VP8_ENCODER_RTCD *, MACROBLOCK *mb, int *Rate, int *Distortion);
 extern void vp8_pick_intra_mbuv_mode(MACROBLOCK *mb);
 extern void vp8_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int recon_uvoffset, int *returnrate, int *returndistortion, int *returnintra);
diff --git a/vp8/encoder/quantize.c b/vp8/encoder/quantize.c
index 803e3a5..86ed267 100644
--- a/vp8/encoder/quantize.c
+++ b/vp8/encoder/quantize.c
@@ -27,7 +27,7 @@
     short *zbin_ptr        = b->zbin;
     short *round_ptr       = b->round;
     short *quant_ptr       = b->quant_fast;
-    short *quant_shift_ptr = b->quant_shift;
+    unsigned char *quant_shift_ptr = b->quant_shift;
     short *qcoeff_ptr      = d->qcoeff;
     short *dqcoeff_ptr     = d->dqcoeff;
     short *dequant_ptr     = d->dequant;
@@ -112,7 +112,7 @@
     short *zbin_ptr        = b->zbin;
     short *round_ptr       = b->round;
     short *quant_ptr       = b->quant;
-    short *quant_shift_ptr = b->quant_shift;
+    unsigned char *quant_shift_ptr = b->quant_shift;
     short *qcoeff_ptr      = d->qcoeff;
     short *dqcoeff_ptr     = d->dqcoeff;
     short *dequant_ptr     = d->dequant;
@@ -166,7 +166,7 @@
     int sz;
     short *coeff_ptr;
     short *quant_ptr;
-    short *quant_shift_ptr;
+    unsigned char *quant_shift_ptr;
     short *qcoeff_ptr;
     short *dqcoeff_ptr;
     short *dequant_ptr;
diff --git a/vp8/encoder/ratectrl.c b/vp8/encoder/ratectrl.c
index 767b271..8e8854a 100644
--- a/vp8/encoder/ratectrl.c
+++ b/vp8/encoder/ratectrl.c
@@ -1451,86 +1451,85 @@
     return (bits_per_mb_at_max_q * cpi->common.MBs) >> BPER_MB_NORMBITS;
 }
 
-void vp8_adjust_key_frame_context(VP8_COMP *cpi)
+
+static int estimate_keyframe_frequency(VP8_COMP *cpi)
 {
     int i;
-    int av_key_frames_per_second;
 
-    // Average key frame frequency and size
-    unsigned int total_weight = 0;
-    unsigned int av_key_frame_frequency = 0;
-    unsigned int av_key_frame_bits = 0;
+    // Average key frame frequency
+    int av_key_frame_frequency = 0;
 
-    unsigned int output_frame_rate = (unsigned int)(100 * cpi->output_frame_rate);
-    unsigned int target_bandwidth = (unsigned int)(100 * cpi->target_bandwidth);
-
-    // Clear down mmx registers to allow floating point in what follows
-    vp8_clear_system_state();  //__asm emms;
-
-    // Update the count of total key frame bits
-    cpi->tot_key_frame_bits += cpi->projected_frame_size;
-
-    // First key frame at start of sequence is a special case. We have no frequency data.
+    /* First key frame at start of sequence is a special case. We have no
+     * frequency data.
+     */
     if (cpi->key_frame_count == 1)
     {
-        av_key_frame_frequency = (int)cpi->output_frame_rate * 2;            // Assume a default of 1 kf every 2 seconds
-        av_key_frame_bits = cpi->projected_frame_size;
-        av_key_frames_per_second  = output_frame_rate / av_key_frame_frequency;  // Note output_frame_rate not cpi->output_frame_rate
+        /* Assume a default of 1 kf every 2 seconds, or the max kf interval,
+         * whichever is smaller.
+         */
+        int key_freq = cpi->oxcf.key_freq>0 ? cpi->oxcf.key_freq : 1;
+        av_key_frame_frequency = (int)cpi->output_frame_rate * 2;
+
+        if (cpi->oxcf.auto_key && av_key_frame_frequency > key_freq)
+            av_key_frame_frequency = cpi->oxcf.key_freq;
+
+        cpi->prior_key_frame_distance[KEY_FRAME_CONTEXT - 1]
+            = av_key_frame_frequency;
     }
     else
     {
+        unsigned int total_weight = 0;
         int last_kf_interval =
                 (cpi->frames_since_key > 0) ? cpi->frames_since_key : 1;
 
-        // reset keyframe context and calculate weighted average of last KEY_FRAME_CONTEXT keyframes
+        /* reset keyframe context and calculate weighted average of last
+         * KEY_FRAME_CONTEXT keyframes
+         */
         for (i = 0; i < KEY_FRAME_CONTEXT; i++)
         {
             if (i < KEY_FRAME_CONTEXT - 1)
-            {
-                cpi->prior_key_frame_size[i]     = cpi->prior_key_frame_size[i+1];
-                cpi->prior_key_frame_distance[i] = cpi->prior_key_frame_distance[i+1];
-            }
+                cpi->prior_key_frame_distance[i]
+                    = cpi->prior_key_frame_distance[i+1];
             else
-            {
-                cpi->prior_key_frame_size[i]     = cpi->projected_frame_size;
                 cpi->prior_key_frame_distance[i] = last_kf_interval;
-            }
 
-            av_key_frame_bits      += prior_key_frame_weight[i] * cpi->prior_key_frame_size[i];
-            av_key_frame_frequency += prior_key_frame_weight[i] * cpi->prior_key_frame_distance[i];
-            total_weight         += prior_key_frame_weight[i];
+            av_key_frame_frequency += prior_key_frame_weight[i]
+                                      * cpi->prior_key_frame_distance[i];
+            total_weight += prior_key_frame_weight[i];
         }
 
-        av_key_frame_bits       /= total_weight;
         av_key_frame_frequency  /= total_weight;
-        av_key_frames_per_second  = output_frame_rate / av_key_frame_frequency;
 
     }
+    return av_key_frame_frequency;
+}
+
+
+void vp8_adjust_key_frame_context(VP8_COMP *cpi)
+{
+    // Clear down mmx registers to allow floating point in what follows
+    vp8_clear_system_state();
 
     // Do we have any key frame overspend to recover?
-    if ((cpi->pass != 2) && (cpi->projected_frame_size > cpi->per_frame_bandwidth))
+    // Two-pass overspend handled elsewhere.
+    if ((cpi->pass != 2)
+         && (cpi->projected_frame_size > cpi->per_frame_bandwidth))
     {
-        // Update the count of key frame overspend to be recovered in subsequent frames
-        // A portion of the KF overspend is treated as gf overspend (and hence recovered more quickly)
-        // as the kf is also a gf. Otherwise the few frames following each kf tend to get more bits
-        // allocated than those following other gfs.
-        cpi->kf_overspend_bits += (cpi->projected_frame_size - cpi->per_frame_bandwidth) * 7 / 8;
-        cpi->gf_overspend_bits += (cpi->projected_frame_size - cpi->per_frame_bandwidth) * 1 / 8;
-        if(!av_key_frame_frequency)
-            av_key_frame_frequency = 60;
+        int overspend;
 
-        // Work out how much to try and recover per frame.
-        // For one pass we estimate the number of frames to spread it over based upon past history.
-        // For two pass we know how many frames there will be till the next kf.
-        if (cpi->pass == 2)
-        {
-            if (cpi->frames_to_key > 16)
-                cpi->kf_bitrate_adjustment = cpi->kf_overspend_bits / (int)cpi->frames_to_key;
-            else
-                cpi->kf_bitrate_adjustment = cpi->kf_overspend_bits / 16;
-        }
-        else
-            cpi->kf_bitrate_adjustment = cpi->kf_overspend_bits / (int)av_key_frame_frequency;
+        /* Update the count of key frame overspend to be recovered in
+         * subsequent frames. A portion of the KF overspend is treated as gf
+         * overspend (and hence recovered more quickly) as the kf is also a
+         * gf. Otherwise the few frames following each kf tend to get more
+         * bits allocated than those following other gfs.
+         */
+        overspend = (cpi->projected_frame_size - cpi->per_frame_bandwidth);
+        cpi->kf_overspend_bits += overspend * 7 / 8;
+        cpi->gf_overspend_bits += overspend * 1 / 8;
+
+        /* Work out how much to try and recover per frame. */
+        cpi->kf_bitrate_adjustment = cpi->kf_overspend_bits
+                                     / estimate_keyframe_frequency(cpi);
     }
 
     cpi->frames_since_key = 0;
@@ -1539,6 +1538,7 @@
 }
 
 void vp8_compute_frame_size_bounds(VP8_COMP *cpi, int *frame_under_shoot_limit, int *frame_over_shoot_limit)
+void vp8_compute_frame_size_bounds(VP8_COMP *cpi, int *frame_under_shoot_limit, int *frame_over_shoot_limit)
 {
     // Set-up bounds on acceptable frame size:
     if (cpi->oxcf.fixed_q >= 0)
diff --git a/vp8/encoder/rdopt.c b/vp8/encoder/rdopt.c
index 863b6d4..1f329d7 100644
--- a/vp8/encoder/rdopt.c
+++ b/vp8/encoder/rdopt.c
@@ -31,7 +31,7 @@
 #include "vp8/common/g_common.h"
 #include "variance.h"
 #include "mcomp.h"
-
+#include "rdopt.h"
 #include "vpx_mem/vpx_mem.h"
 #include "dct.h"
 #include "vp8/common/systemdependent.h"
@@ -46,13 +46,8 @@
 extern void vp8cx_mb_init_quantizer(VP8_COMP *cpi, MACROBLOCK *x);
 extern void vp8_update_zbin_extra(VP8_COMP *cpi, MACROBLOCK *x);
 
-
-#define RDCOST(RM,DM,R,D) ( ((128+(R)*(RM)) >> 8) + (DM)*(D) )
-
 #define MAXF(a,b)            (((a) > (b)) ? (a) : (b))
 
-
-
 static const int auto_speed_thresh[17] =
 {
     1000,
@@ -480,7 +475,6 @@
 
 }
 
-#if !(CONFIG_REALTIME_ONLY)
 static int cost_coeffs(MACROBLOCK *mb, BLOCKD *b, int type, ENTROPY_CONTEXT *a, ENTROPY_CONTEXT *l)
 {
     int c = !type;              /* start at coef 0, unless Y with Y2 */
@@ -847,7 +841,6 @@
 
     x->e_mbd.mode_info_context->mbmi.uv_mode = mode_selected;
 }
-#endif
 
 int vp8_cost_mv_ref(MB_PREDICTION_MODE m, const int near_mv_ref_ct[4])
 {
@@ -875,7 +868,6 @@
     }
 }
 
-#if !(CONFIG_REALTIME_ONLY)
 static int labels2mode(
     MACROBLOCK *x,
     int const *labelings, int which_label,
@@ -1213,12 +1205,15 @@
 
                 if (bestsme < INT_MAX)
                 {
+                    int distortion;
+                    unsigned int sse;
+
                     if (!cpi->common.full_pixel)
                         cpi->find_fractional_mv_step(x, c, e, &mode_mv[NEW4X4],
-                                                     bsi->ref_mv, x->errorperbit / 2, v_fn_ptr, x->mvcost);
+                                                     bsi->ref_mv, x->errorperbit / 2, v_fn_ptr, x->mvcost, &distortion, &sse);
                     else
                         vp8_skip_fractional_mv_step(x, c, e, &mode_mv[NEW4X4],
-                                                    bsi->ref_mv, x->errorperbit, v_fn_ptr, x->mvcost);
+                                                    bsi->ref_mv, x->errorperbit, v_fn_ptr, x->mvcost, &distortion, &sse);
                 }
             } /* NEW4X4 */
 
@@ -1437,7 +1432,6 @@
 
     return bsi.segment_rd;
 }
-#endif
 
 static void swap(int *x,int *y)
 {
@@ -1725,7 +1719,6 @@
     }
 }
 
-#if !(CONFIG_REALTIME_ONLY)
 void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int recon_uvoffset, int *returnrate, int *returndistortion, int *returnintra)
 {
     BLOCK *b = &x->block[0];
@@ -1760,9 +1753,6 @@
     //int intermodecost[MAX_MODES];
 
     MB_PREDICTION_MODE uv_intra_mode;
-
-    int force_no_skip = 0;
-
     MV mvp;
     int near_sadidx[8] = {0, 1, 2, 3, 4, 5, 6, 7};
     int saddone=0;
@@ -1865,8 +1855,6 @@
         int disable_skip = 0;
         int other_cost = 0;
 
-        force_no_skip = 0;
-
         // Experimental debug code.
         // Record of rd values recorded for this MB. -1 indicates not measured
         //all_rds[mode_index] = -1;
@@ -2198,8 +2186,11 @@
                 x->mv_row_max = tmp_row_max;
 
                 if (bestsme < INT_MAX)
-                    // cpi->find_fractional_mv_step(x,b,d,&d->bmi.mv.as_mv,&best_ref_mv,x->errorperbit/2,cpi->fn_ptr.svf,cpi->fn_ptr.vf,x->mvcost);  // normal mvc=11
-                    cpi->find_fractional_mv_step(x, b, d, &d->bmi.mv.as_mv, &best_ref_mv, x->errorperbit / 4, &cpi->fn_ptr[BLOCK_16X16], x->mvcost);
+                {
+                    int dis; /* TODO: use dis in distortion calculation later. */
+                    unsigned int sse;
+                    cpi->find_fractional_mv_step(x, b, d, &d->bmi.mv.as_mv, &best_ref_mv, x->errorperbit / 4, &cpi->fn_ptr[BLOCK_16X16], x->mvcost, &dis, &sse);
+                }
 
                 mode_mv[NEWMV].row = d->bmi.mv.as_mv.row;
                 mode_mv[NEWMV].col = d->bmi.mv.as_mv.col;
@@ -2230,8 +2221,6 @@
 
         case ZEROMV:
 
-        mv_selected:
-
             // Trap vectors that reach beyond the UMV borders
             // Note that ALL New MV, Nearest MV Near MV and Zero MV code drops through to this point
             // because of the lack of break statements in the previous two cases.
@@ -2240,14 +2229,15 @@
                 continue;
 
             vp8_set_mbmode_and_mvs(x, this_mode, &mode_mv[this_mode]);
-            vp8_build_inter_predictors_mby(&x->e_mbd);
+            vp8_build_inter16x16_predictors_mby(&x->e_mbd);
 
             if (cpi->active_map_enabled && x->active_ptr[0] == 0) {
                 x->skip = 1;
             }
             else if (x->encode_breakout)
             {
-                int sum, sse;
+                int sum;
+                unsigned int sse;
                 int threshold = (xd->block[0].dequant[1]
                             * xd->block[0].dequant[1] >>4);
 
@@ -2256,7 +2246,7 @@
 
                 VARIANCE_INVOKE(&cpi->rtcd.variance, get16x16var)
                     (x->src.y_buffer, x->src.y_stride,
-                     x->e_mbd.predictor, 16, (unsigned int *)(&sse), &sum);
+                     x->e_mbd.predictor, 16, &sse, &sum);
 
                 if (sse < threshold)
                 {
@@ -2280,8 +2270,7 @@
                             distortion_uv = sse2;
 
                             disable_skip = 1;
-                            this_rd = RDCOST(x->rdmult, x->rddiv, rate2,
-                                             distortion2);
+                            this_rd = RDCOST(x->rdmult, x->rddiv, rate2, distortion2);
 
                             break;
                         }
@@ -2376,7 +2365,6 @@
         {
             // Note index of best mode so far
             best_mode_index = mode_index;
-            x->e_mbd.mode_info_context->mbmi.force_no_skip = force_no_skip;
 
             if (this_mode <= B_PRED)
             {
@@ -2473,7 +2461,6 @@
         best_mbmode.uv_mode = 0;
         best_mbmode.mb_skip_coeff = (cpi->common.mb_no_coeff_skip) ? 1 : 0;
         best_mbmode.partitioning = 0;
-        best_mbmode.dc_diff = 0;
 
         vpx_memcpy(&x->e_mbd.mode_info_context->mbmi, &best_mbmode, sizeof(MB_MODE_INFO));
         vpx_memcpy(x->partition_info, &best_partition, sizeof(PARTITION_INFO));
@@ -2484,6 +2471,7 @@
         }
 
         x->e_mbd.mode_info_context->mbmi.mv.as_int = 0;
+        return;
     }
 
 
@@ -2507,4 +2495,3 @@
 
     x->e_mbd.mode_info_context->mbmi.mv.as_mv = x->e_mbd.block[15].bmi.mv.as_mv;
 }
-#endif
diff --git a/vp8/encoder/rdopt.h b/vp8/encoder/rdopt.h
index 1d1be11..1d5f9a3 100644
--- a/vp8/encoder/rdopt.h
+++ b/vp8/encoder/rdopt.h
@@ -11,6 +11,9 @@
 
 #ifndef __INC_RDOPT_H
 #define __INC_RDOPT_H
+
+#define RDCOST(RM,DM,R,D) ( ((128+(R)*(RM)) >> 8) + (DM)*(D) )
+
 extern void vp8_initialize_rd_consts(VP8_COMP *cpi, int Qvalue);
 extern int vp8_rd_pick_intra4x4mby_modes(VP8_COMP *cpi, MACROBLOCK *mb, int *rate, int *rate_to, int *distortion, int best_rd);
 extern int vp8_rd_pick_intra16x16mby_mode(VP8_COMP *cpi, MACROBLOCK *x, int *returnrate, int *rate_to, int *returndistortion);
diff --git a/vp8/encoder/ssim.c b/vp8/encoder/ssim.c
index 64d67c6..c78be37 100644
--- a/vp8/encoder/ssim.c
+++ b/vp8/encoder/ssim.c
@@ -290,8 +290,8 @@
      }
 }
 
-const static long long c1 =  426148; // (256^2*(.01*255)^2
-const static long long c2 = 3835331; //(256^2*(.03*255)^2
+const static long long cc1 =  26634; // (64^2*(.01*255)^2
+const static long long cc2 = 239708; // (64^2*(.03*255)^2
 
 static double similarity
 (
@@ -303,10 +303,19 @@
     int count
 )
 {
-    long long ssim_n = (2*sum_s*sum_r+ c1)*(2*count*sum_sxr-2*sum_s*sum_r+c2);
+    long long ssim_n, ssim_d;
+    long long c1, c2;
 
-    long long ssim_d = (sum_s*sum_s +sum_r*sum_r+c1)*
-            (count*sum_sq_s-sum_s*sum_s + count*sum_sq_r-sum_r*sum_r +c2) ;
+    //scale the constants by number of pixels
+    c1 = (cc1*count*count)>>12;
+    c2 = (cc2*count*count)>>12;
+
+    ssim_n = (2*sum_s*sum_r+ c1)*((long long) 2*count*sum_sxr-
+          (long long) 2*sum_s*sum_r+c2);
+
+    ssim_d = (sum_s*sum_s +sum_r*sum_r+c1)*
+        ((long long)count*sum_sq_s-(long long)sum_s*sum_s +
+        (long long)count*sum_sq_r-(long long) sum_r*sum_r +c2) ;
 
     return ssim_n * 1.0 / ssim_d;
 }
@@ -332,18 +341,33 @@
            const vp8_variance_rtcd_vtable_t *rtcd)
 {
     unsigned long sum_s=0,sum_r=0,sum_sq_s=0,sum_sq_r=0,sum_sxr=0;
-    double ssim3;
-    long long ssim_n;
-    long long ssim_d;
+    long long ssim3;
+    long long ssim_n,ssim_n1,ssim_n2;
+    long long ssim_d,ssim_d1,ssim_d2;
+    long long ssim_t1,ssim_t2;
+    long long c1, c2;
+
+    // normalize by 256/64
+    c1 = cc1*16;
+    c2 = cc2*16;
 
     rtcd->ssimpf(s, sp, r, rp, &sum_s, &sum_r, &sum_sq_s, &sum_sq_r, &sum_sxr);
-    ssim_n = (2*sum_s*sum_r+ c1)*(2*256*sum_sxr-2*sum_s*sum_r+c2);
+    ssim_n1 = (2*sum_s*sum_r+ c1);
 
-    ssim_d = (sum_s*sum_s +sum_r*sum_r+c1)*
-            (256*sum_sq_s-sum_s*sum_s + 256*sum_sq_r-sum_r*sum_r +c2) ;
+    ssim_n2 =((long long) 2*256*sum_sxr-(long long) 2*sum_s*sum_r+c2);
 
-    ssim3 = 256 * (ssim_d-ssim_n) / ssim_d;
-    return (long)( 256*ssim3 * ssim3 );
+    ssim_d1 =((long long)sum_s*sum_s +(long long)sum_r*sum_r+c1);
+
+    ssim_d2 = (256 * (long long) sum_sq_s-(long long) sum_s*sum_s +
+                    (long long) 256*sum_sq_r-(long long) sum_r*sum_r +c2) ;
+
+    ssim_t1 = 256 - 256 * ssim_n1 / ssim_d1;
+    ssim_t2 = 256 - 256 * ssim_n2 / ssim_d2;
+
+    ssim3 = 256 *ssim_t1 * ssim_t2;
+    if(ssim3 <0 )
+        ssim3=0;
+    return (long)( ssim3  );
 }
 // TODO: (jbb) this 8x8 window might be too big + we may want to pick pixels
 // such that the window regions overlap block boundaries to penalize blocking
@@ -361,18 +385,20 @@
 )
 {
     int i,j;
-
+    int samples =0;
     double ssim_total=0;
 
-    // we can sample points as frequently as we like start with 1 per 8x8
-    for(i=0; i < height; i+=8, img1 += stride_img1*8, img2 += stride_img2*8)
+    // we can sample points as frequently as we like start with 1 per 4x4
+    for(i=0; i < height-8; i+=4, img1 += stride_img1*4, img2 += stride_img2*4)
     {
-        for(j=0; j < width; j+=8 )
+        for(j=0; j < width-8; j+=4 )
         {
-            ssim_total += ssim_8x8(img1, stride_img1, img2, stride_img2, rtcd);
+            double v = ssim_8x8(img1+j, stride_img1, img2+j, stride_img2, rtcd);
+            ssim_total += v;
+            samples++;
         }
     }
-    ssim_total /= (width/8 * height /8);
+    ssim_total /= samples;
     return ssim_total;
 
 }
@@ -405,4 +431,4 @@
     *weight = 1;
 
     return ssimv;
-}
+}
\ No newline at end of file
diff --git a/vp8/encoder/temporal_filter.c b/vp8/encoder/temporal_filter.c
index cec9518..b771955 100644
--- a/vp8/encoder/temporal_filter.c
+++ b/vp8/encoder/temporal_filter.c
@@ -208,10 +208,12 @@
     // Try sub-pixel MC?
     //if (bestsme > error_thresh && bestsme < INT_MAX)
     {
+        int distortion;
+        unsigned int sse;
         bestsme = cpi->find_fractional_mv_step(x, b, d,
                     &d->bmi.mv.as_mv, &best_ref_mv1,
                     x->errorperbit, &cpi->fn_ptr[BLOCK_16X16],
-                    mvcost);
+                    mvcost, &distortion, &sse);
     }
 #endif
 
@@ -357,8 +359,8 @@
             }
 
             // Normalize filter output to produce AltRef frame
-            dst1 = cpi->alt_ref_buffer.source_buffer.y_buffer;
-            stride = cpi->alt_ref_buffer.source_buffer.y_stride;
+            dst1 = cpi->alt_ref_buffer.y_buffer;
+            stride = cpi->alt_ref_buffer.y_stride;
             byte = mb_y_offset;
             for (i = 0,k = 0; i < 16; i++)
             {
@@ -377,9 +379,9 @@
                 byte += stride - 16;
             }
 
-            dst1 = cpi->alt_ref_buffer.source_buffer.u_buffer;
-            dst2 = cpi->alt_ref_buffer.source_buffer.v_buffer;
-            stride = cpi->alt_ref_buffer.source_buffer.uv_stride;
+            dst1 = cpi->alt_ref_buffer.u_buffer;
+            dst2 = cpi->alt_ref_buffer.v_buffer;
+            stride = cpi->alt_ref_buffer.uv_stride;
             byte = mb_uv_offset;
             for (i = 0,k = 256; i < 8; i++)
             {
@@ -422,7 +424,8 @@
 
 void vp8_temporal_filter_prepare_c
 (
-    VP8_COMP *cpi
+    VP8_COMP *cpi,
+    int distance
 )
 {
     int frame = 0;
@@ -441,12 +444,9 @@
 
     int max_frames = cpi->active_arnr_frames;
 
-    num_frames_backward = cpi->last_alt_ref_sei - cpi->source_encode_index;
-
-    if (num_frames_backward < 0)
-        num_frames_backward += cpi->oxcf.lag_in_frames;
-
-    num_frames_forward = cpi->oxcf.lag_in_frames - (num_frames_backward + 1);
+    num_frames_backward = distance;
+    num_frames_forward = vp8_lookahead_depth(cpi->lookahead)
+                         - (num_frames_backward + 1);
 
     switch (blur_type)
     {
@@ -498,8 +498,7 @@
         break;
     }
 
-    start_frame = (cpi->last_alt_ref_sei
-                    + frames_to_blur_forward) % cpi->oxcf.lag_in_frames;
+    start_frame = distance + frames_to_blur_forward;
 
 #ifdef DEBUGFWG
     // DEBUG FWG
@@ -520,12 +519,9 @@
     for (frame = 0; frame < frames_to_blur; frame++)
     {
         int which_buffer =  start_frame - frame;
-
-        if (which_buffer < 0)
-            which_buffer += cpi->oxcf.lag_in_frames;
-
-        cpi->frames[frames_to_blur-1-frame]
-                = &cpi->src_buffer[which_buffer].source_buffer;
+        struct lookahead_entry* buf = vp8_lookahead_peek(cpi->lookahead,
+                                                         which_buffer);
+        cpi->frames[frames_to_blur-1-frame] = &buf->img;
     }
 
     vp8_temporal_filter_iterate_c (
diff --git a/vp8/encoder/tokenize.c b/vp8/encoder/tokenize.c
index e3f423f..1c59238 100644
--- a/vp8/encoder/tokenize.c
+++ b/vp8/encoder/tokenize.c
@@ -224,28 +224,9 @@
     int plane_type;
     int b;
 
-    TOKENEXTRA *start = *t;
-    TOKENEXTRA *tp = *t;
-
-    x->mode_info_context->mbmi.dc_diff = 1;
-
-#if 0
-
-    if (x->mbmi.force_no_skip)
-    {
-        x->mbmi.mb_skip_coeff = 1;
-        //reset for next_mb.
-        x->mbmi.force_no_skip = 0;
-    }
-
-#endif
-
-#if 1
-
     x->mode_info_context->mbmi.mb_skip_coeff = mb_is_skippable(x);
     if (x->mode_info_context->mbmi.mb_skip_coeff)
     {
-
         cpi->skip_true_count++;
 
         if (!cpi->common.mb_no_coeff_skip)
@@ -255,17 +236,11 @@
             vp8_fix_contexts(x);
         }
 
-        if (x->mode_info_context->mbmi.mode != B_PRED && x->mode_info_context->mbmi.mode != SPLITMV)
-            x->mode_info_context->mbmi.dc_diff = 0;
-        else
-            x->mode_info_context->mbmi.dc_diff = 1;
-
-
         return;
     }
 
     cpi->skip_false_count++;
-#endif
+
 #if 0
     vpx_memcpy(cpi->coef_counts_backup, cpi->coef_counts, sizeof(cpi->coef_counts));
 #endif
@@ -292,42 +267,6 @@
                             A + vp8_block2above[b],
                             L + vp8_block2left[b], cpi);
 
-#if 0
-
-    if (cpi->common.mb_no_coeff_skip)
-    {
-        int skip = 1;
-
-        while ((tp != *t) && skip)
-        {
-            skip = (skip && (tp->Token == DCT_EOB_TOKEN));
-            tp ++;
-        }
-
-        if (skip != x->mbmi.mb_skip_coeff)
-            skip += 0;
-
-        x->mbmi.mb_skip_coeff = skip;
-
-        if (x->mbmi.mb_skip_coeff == 1)
-        {
-            x->mbmi.dc_diff = 0;
-            //redo the coutnts
-            vpx_memcpy(cpi->coef_counts, cpi->coef_counts_backup, sizeof(cpi->coef_counts));
-
-            *t = start;
-            cpi->skip_true_count++;
-            //skip_true_count++;
-        }
-        else
-        {
-
-            cpi->skip_false_count++;
-            //skip_false_count++;
-        }
-    }
-
-#endif
 }
 
 
@@ -510,13 +449,6 @@
                      A + vp8_block2above[24], L + vp8_block2left[24], cpi);
     plane_type = 0;
 
-
-    if (x->mode_info_context->mbmi.mode != B_PRED && x->mode_info_context->mbmi.mode != SPLITMV)
-        x->mode_info_context->mbmi.dc_diff = 0;
-    else
-        x->mode_info_context->mbmi.dc_diff = 1;
-
-
     for (b = 0; b < 16; b++)
         stuff1st_order_b(x->block + b, t, plane_type, x->frame_type,
                          A + vp8_block2above[b],
diff --git a/vp8/encoder/x86/dct_sse2.asm b/vp8/encoder/x86/dct_sse2.asm
index 652dd98..3d52a5d 100644
--- a/vp8/encoder/x86/dct_sse2.asm
+++ b/vp8/encoder/x86/dct_sse2.asm
@@ -33,6 +33,7 @@
     %define     input       rcx
     %define     output      rdx
     %define     pitch       r8
+    SAVE_XMM 7, u
   %else
     %define     input       rdi
     %define     output      rsi
@@ -53,6 +54,7 @@
     pop         rbp
 %else
   %ifidn __OUTPUT_FORMAT__,x64
+    RESTORE_XMM
   %endif
 %endif
     ret
diff --git a/vp8/encoder/x86/encodeopt.asm b/vp8/encoder/x86/encodeopt.asm
index c0f06bb..9946294 100644
--- a/vp8/encoder/x86/encodeopt.asm
+++ b/vp8/encoder/x86/encodeopt.asm
@@ -22,33 +22,33 @@
     ; end prologue
 
         mov         rsi,        arg(0) ;coeff_ptr
-
         mov         rdi,        arg(1) ;dcoef_ptr
-        movdqa      xmm3,       [rsi]
 
-        movdqa      xmm4,       [rdi]
-        movdqa      xmm5,       [rsi+16]
+        movdqa      xmm0,       [rsi]
+        movdqa      xmm1,       [rdi]
 
-        movdqa      xmm6,       [rdi+16]
-        psubw       xmm3,       xmm4
+        movdqa      xmm2,       [rsi+16]
+        movdqa      xmm3,       [rdi+16]
 
-        psubw       xmm5,       xmm6
-        pmaddwd     xmm3,       xmm3
-        pmaddwd     xmm5,       xmm5
+        psubw       xmm0,       xmm1
+        psubw       xmm2,       xmm3
 
-        paddd       xmm3,       xmm5
+        pmaddwd     xmm0,       xmm0
+        pmaddwd     xmm2,       xmm2
 
-        pxor        xmm7,       xmm7
-        movdqa      xmm0,       xmm3
+        paddd       xmm0,       xmm2
 
-        punpckldq   xmm0,       xmm7
-        punpckhdq   xmm3,       xmm7
+        pxor        xmm5,       xmm5
+        movdqa      xmm1,       xmm0
 
-        paddd       xmm0,       xmm3
-        movdqa      xmm3,       xmm0
+        punpckldq   xmm0,       xmm5
+        punpckhdq   xmm1,       xmm5
+
+        paddd       xmm0,       xmm1
+        movdqa      xmm1,       xmm0
 
         psrldq      xmm0,       8
-        paddd       xmm0,       xmm3
+        paddd       xmm0,       xmm1
 
         movq        rax,        xmm0
 
@@ -208,53 +208,54 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 3
+    SAVE_XMM 6
     push rsi
     push rdi
     ; end prolog
 
 
         mov         rsi,        arg(0) ;coeff_ptr
-        pxor        xmm7,       xmm7
+        pxor        xmm6,       xmm6
 
         mov         rdi,        arg(1) ;dcoef_ptr
-        pxor        xmm2,       xmm2
+        pxor        xmm4,       xmm4
 
-        movd        xmm1,       dword ptr arg(2) ;dc
-        por         xmm1,       xmm2
+        movd        xmm5,       dword ptr arg(2) ;dc
+        por         xmm5,       xmm4
 
-        pcmpeqw     xmm1,       xmm7
+        pcmpeqw     xmm5,       xmm6
         mov         rcx,        16
 
 mberror_loop:
-        movdqa      xmm3,       [rsi]
-        movdqa      xmm4,       [rdi]
+        movdqa      xmm0,       [rsi]
+        movdqa      xmm1,       [rdi]
 
-        movdqa      xmm5,       [rsi+16]
-        movdqa      xmm6,       [rdi+16]
+        movdqa      xmm2,       [rsi+16]
+        movdqa      xmm3,       [rdi+16]
 
 
-        psubw       xmm5,       xmm6
-        pmaddwd     xmm5,       xmm5
+        psubw       xmm2,       xmm3
+        pmaddwd     xmm2,       xmm2
 
-        psubw       xmm3,       xmm4
-        pand        xmm3,       xmm1
+        psubw       xmm0,       xmm1
+        pand        xmm0,       xmm5
 
-        pmaddwd     xmm3,       xmm3
+        pmaddwd     xmm0,       xmm0
         add         rsi,        32
 
         add         rdi,        32
 
         sub         rcx,        1
-        paddd       xmm2,       xmm5
+        paddd       xmm4,       xmm2
 
-        paddd       xmm2,       xmm3
+        paddd       xmm4,       xmm0
         jnz         mberror_loop
 
-        movdqa      xmm0,       xmm2
-        punpckldq   xmm0,       xmm7
+        movdqa      xmm0,       xmm4
+        punpckldq   xmm0,       xmm6
 
-        punpckhdq   xmm2,       xmm7
-        paddd       xmm0,       xmm2
+        punpckhdq   xmm4,       xmm6
+        paddd       xmm0,       xmm4
 
         movdqa      xmm1,       xmm0
         psrldq      xmm0,       8
@@ -265,6 +266,7 @@
     pop rdi
     pop rsi
     ; begin epilog
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -342,7 +344,7 @@
         mov             rdi,        arg(1) ;d_ptr
 
         mov             rcx,        16
-        pxor            xmm7,       xmm7
+        pxor            xmm3,       xmm3
 
 mbuverror_loop:
 
@@ -352,7 +354,7 @@
         psubw           xmm1,       xmm2
         pmaddwd         xmm1,       xmm1
 
-        paddd           xmm7,       xmm1
+        paddd           xmm3,       xmm1
 
         add             rsi,        16
         add             rdi,        16
@@ -361,7 +363,7 @@
         jnz             mbuverror_loop
 
         pxor        xmm0,           xmm0
-        movdqa      xmm1,           xmm7
+        movdqa      xmm1,           xmm3
 
         movdqa      xmm2,           xmm1
         punpckldq   xmm1,           xmm0
diff --git a/vp8/encoder/x86/fwalsh_sse2.asm b/vp8/encoder/x86/fwalsh_sse2.asm
index 39439f0..71efd56 100644
--- a/vp8/encoder/x86/fwalsh_sse2.asm
+++ b/vp8/encoder/x86/fwalsh_sse2.asm
@@ -17,7 +17,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 3
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
diff --git a/vp8/encoder/x86/quantize_sse2.asm b/vp8/encoder/x86/quantize_sse2.asm
index 5e40dc7..056b64c 100644
--- a/vp8/encoder/x86/quantize_sse2.asm
+++ b/vp8/encoder/x86/quantize_sse2.asm
@@ -20,7 +20,7 @@
 sym(vp8_regular_quantize_b_sse2):
     push        rbp
     mov         rbp, rsp
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
 
 %if ABI_IS_32BIT
@@ -142,7 +142,7 @@
     movsx       edi, WORD PTR[rsp + temp_qcoeff + %1 * 2]
 
     ; downshift by quant_shift[rc]
-    movsx       ecx, WORD PTR[rax + %1 * 2] ; quant_shift_ptr[rc]
+    movsx       cx, BYTE PTR[rax + %1]      ; quant_shift_ptr[rc]
     sar         edi, cl                     ; also sets Z bit
     je          rq_zigzag_loop_%1           ; !y
     mov         WORD PTR[rsp + qcoeff + %1 * 2], di ;qcoeff_ptr[rc] = temp_qcoeff[rc]
diff --git a/vp8/encoder/x86/quantize_sse4.asm b/vp8/encoder/x86/quantize_sse4.asm
new file mode 100644
index 0000000..258899e
--- /dev/null
+++ b/vp8/encoder/x86/quantize_sse4.asm
@@ -0,0 +1,254 @@
+;
+;  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+;  Use of this source code is governed by a BSD-style license and patent
+;  grant that can be found in the LICENSE file in the root of the source
+;  tree. All contributing project authors may be found in the AUTHORS
+;  file in the root of the source tree.
+;
+
+
+%include "vpx_ports/x86_abi_support.asm"
+%include "asm_enc_offsets.asm"
+
+
+; void vp8_regular_quantize_b_sse4 | arg
+;  (BLOCK  *b,                     |  0
+;   BLOCKD *d)                     |  1
+
+global sym(vp8_regular_quantize_b_sse4)
+sym(vp8_regular_quantize_b_sse4):
+
+%if ABI_IS_32BIT
+    push        rbp
+    mov         rbp, rsp
+    GET_GOT     rbx
+    push        rdi
+    push        rsi
+
+    ALIGN_STACK 16, rax
+    %define qcoeff      0 ; 32
+    %define stack_size 32
+    sub         rsp, stack_size
+%else
+  %ifidn __OUTPUT_FORMAT__,x64
+    SAVE_XMM 8, u
+    push        rdi
+    push        rsi
+  %endif
+%endif
+    ; end prolog
+
+%if ABI_IS_32BIT
+    mov         rdi, arg(0)                 ; BLOCK *b
+    mov         rsi, arg(1)                 ; BLOCKD *d
+%else
+  %ifidn __OUTPUT_FORMAT__,x64
+    mov         rdi, rcx                    ; BLOCK *b
+    mov         rsi, rdx                    ; BLOCKD *d
+  %else
+    ;mov         rdi, rdi                    ; BLOCK *b
+    ;mov         rsi, rsi                    ; BLOCKD *d
+  %endif
+%endif
+
+    mov         rax, [rdi + vp8_block_coeff]
+    mov         rcx, [rdi + vp8_block_zbin]
+    mov         rdx, [rdi + vp8_block_round]
+    movd        xmm7, [rdi + vp8_block_zbin_extra]
+
+    ; z
+    movdqa      xmm0, [rax]
+    movdqa      xmm1, [rax + 16]
+
+    ; duplicate zbin_oq_value
+    pshuflw     xmm7, xmm7, 0
+    punpcklwd   xmm7, xmm7
+
+    movdqa      xmm2, xmm0
+    movdqa      xmm3, xmm1
+
+    ; sz
+    psraw       xmm0, 15
+    psraw       xmm1, 15
+
+    ; (z ^ sz)
+    pxor        xmm2, xmm0
+    pxor        xmm3, xmm1
+
+    ; x = abs(z)
+    psubw       xmm2, xmm0
+    psubw       xmm3, xmm1
+
+    ; zbin
+    movdqa      xmm4, [rcx]
+    movdqa      xmm5, [rcx + 16]
+
+    ; *zbin_ptr + zbin_oq_value
+    paddw       xmm4, xmm7
+    paddw       xmm5, xmm7
+
+    movdqa      xmm6, xmm2
+    movdqa      xmm7, xmm3
+
+    ; x - (*zbin_ptr + zbin_oq_value)
+    psubw       xmm6, xmm4
+    psubw       xmm7, xmm5
+
+    ; round
+    movdqa      xmm4, [rdx]
+    movdqa      xmm5, [rdx + 16]
+
+    mov         rax, [rdi + vp8_block_quant_shift]
+    mov         rcx, [rdi + vp8_block_quant]
+    mov         rdx, [rdi + vp8_block_zrun_zbin_boost]
+
+    ; x + round
+    paddw       xmm2, xmm4
+    paddw       xmm3, xmm5
+
+    ; quant
+    movdqa      xmm4, [rcx]
+    movdqa      xmm5, [rcx + 16]
+
+    ; y = x * quant_ptr >> 16
+    pmulhw      xmm4, xmm2
+    pmulhw      xmm5, xmm3
+
+    ; y += x
+    paddw       xmm2, xmm4
+    paddw       xmm3, xmm5
+
+    pxor        xmm4, xmm4
+%if ABI_IS_32BIT
+    movdqa      [rsp + qcoeff], xmm4
+    movdqa      [rsp + qcoeff + 16], xmm4
+%else
+    pxor        xmm8, xmm8
+%endif
+
+    ; quant_shift
+    movdqa      xmm5, [rax]
+
+    ; zrun_zbin_boost
+    mov         rax, rdx
+
+%macro ZIGZAG_LOOP 5
+    ; x
+    pextrw      ecx, %4, %2
+
+    ; if (x >= zbin)
+    sub         cx, WORD PTR[rdx]           ; x - zbin
+    lea         rdx, [rdx + 2]              ; zbin_boost_ptr++
+    jl          rq_zigzag_loop_%1           ; x < zbin
+
+    pextrw      edi, %3, %2                 ; y
+
+    ; downshift by quant_shift[rc]
+    pextrb      ecx, xmm5, %1               ; quant_shift[rc]
+    sar         edi, cl                     ; also sets Z bit
+    je          rq_zigzag_loop_%1           ; !y
+%if ABI_IS_32BIT
+    mov         WORD PTR[rsp + qcoeff + %1 *2], di
+%else
+    pinsrw      %5, edi, %2                 ; qcoeff[rc]
+%endif
+    mov         rdx, rax                    ; reset to b->zrun_zbin_boost
+rq_zigzag_loop_%1:
+%endmacro
+; in vp8_default_zig_zag1d order: see vp8/common/entropy.c
+ZIGZAG_LOOP  0, 0, xmm2, xmm6, xmm4
+ZIGZAG_LOOP  1, 1, xmm2, xmm6, xmm4
+ZIGZAG_LOOP  4, 4, xmm2, xmm6, xmm4
+ZIGZAG_LOOP  8, 0, xmm3, xmm7, xmm8
+ZIGZAG_LOOP  5, 5, xmm2, xmm6, xmm4
+ZIGZAG_LOOP  2, 2, xmm2, xmm6, xmm4
+ZIGZAG_LOOP  3, 3, xmm2, xmm6, xmm4
+ZIGZAG_LOOP  6, 6, xmm2, xmm6, xmm4
+ZIGZAG_LOOP  9, 1, xmm3, xmm7, xmm8
+ZIGZAG_LOOP 12, 4, xmm3, xmm7, xmm8
+ZIGZAG_LOOP 13, 5, xmm3, xmm7, xmm8
+ZIGZAG_LOOP 10, 2, xmm3, xmm7, xmm8
+ZIGZAG_LOOP  7, 7, xmm2, xmm6, xmm4
+ZIGZAG_LOOP 11, 3, xmm3, xmm7, xmm8
+ZIGZAG_LOOP 14, 6, xmm3, xmm7, xmm8
+ZIGZAG_LOOP 15, 7, xmm3, xmm7, xmm8
+
+    mov         rcx, [rsi + vp8_blockd_dequant]
+    mov         rdi, [rsi + vp8_blockd_dqcoeff]
+
+%if ABI_IS_32BIT
+    movdqa      xmm4, [rsp + qcoeff]
+    movdqa      xmm5, [rsp + qcoeff + 16]
+%else
+    %define     xmm5 xmm8
+%endif
+
+    ; y ^ sz
+    pxor        xmm4, xmm0
+    pxor        xmm5, xmm1
+    ; x = (y ^ sz) - sz
+    psubw       xmm4, xmm0
+    psubw       xmm5, xmm1
+
+    ; dequant
+    movdqa      xmm0, [rcx]
+    movdqa      xmm1, [rcx + 16]
+
+    mov         rcx, [rsi + vp8_blockd_qcoeff]
+
+    pmullw      xmm0, xmm4
+    pmullw      xmm1, xmm5
+
+    ; store qcoeff
+    movdqa      [rcx], xmm4
+    movdqa      [rcx + 16], xmm5
+
+    ; store dqcoeff
+    movdqa      [rdi], xmm0
+    movdqa      [rdi + 16], xmm1
+
+    ; select the last value (in zig_zag order) for EOB
+    pxor        xmm6, xmm6
+    pcmpeqw     xmm4, xmm6
+    pcmpeqw     xmm5, xmm6
+
+    packsswb    xmm4, xmm5
+    pshufb      xmm4, [GLOBAL(zig_zag1d)]
+    pmovmskb    edx, xmm4
+    xor         rdi, rdi
+    mov         eax, -1
+    xor         dx, ax
+    bsr         eax, edx
+    sub         edi, edx
+    sar         edi, 31
+    add         eax, 1
+    and         eax, edi
+
+    mov         [rsi + vp8_blockd_eob], eax
+
+    ; begin epilog
+%if ABI_IS_32BIT
+    add         rsp, stack_size
+    pop         rsp
+
+    pop         rsi
+    pop         rdi
+    RESTORE_GOT
+    pop         rbp
+%else
+  %undef xmm5
+  %ifidn __OUTPUT_FORMAT__,x64
+    pop         rsi
+    pop         rdi
+    RESTORE_XMM
+  %endif
+%endif
+
+    ret
+
+SECTION_RODATA
+align 16
+; vp8/common/entropy.c: vp8_default_zig_zag1d
+zig_zag1d:
+    db 0, 1, 4, 8, 5, 2, 3, 6, 9, 12, 13, 10, 7, 11, 14, 15
diff --git a/vp8/encoder/x86/quantize_x86.h b/vp8/encoder/x86/quantize_x86.h
index f093580..bbe475f 100644
--- a/vp8/encoder/x86/quantize_x86.h
+++ b/vp8/encoder/x86/quantize_x86.h
@@ -51,4 +51,17 @@
 
 #endif /* HAVE_SSSE3 */
 
+
+#if HAVE_SSE4_1
+extern prototype_quantize_block(vp8_regular_quantize_b_sse4);
+
+#if !CONFIG_RUNTIME_CPU_DETECT
+
+#undef vp8_quantize_quantb
+#define vp8_quantize_quantb vp8_regular_quantize_b_sse4
+
+#endif /* !CONFIG_RUNTIME_CPU_DETECT */
+
+#endif /* HAVE_SSE4_1 */
+
 #endif /* QUANTIZE_X86_H */
diff --git a/vp8/encoder/x86/sad_sse2.asm b/vp8/encoder/x86/sad_sse2.asm
index cc6bc3c..04ee72f 100644
--- a/vp8/encoder/x86/sad_sse2.asm
+++ b/vp8/encoder/x86/sad_sse2.asm
@@ -21,6 +21,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 4
+    SAVE_XMM 6
     push        rsi
     push        rdi
     ; end prolog
@@ -34,7 +35,7 @@
         lea             rcx,        [rsi+rax*8]
 
         lea             rcx,        [rcx+rax*8]
-        pxor            xmm7,       xmm7
+        pxor            xmm6,       xmm6
 
 x16x16sad_wmt_loop:
 
@@ -52,32 +53,33 @@
         punpcklbw       xmm1,       xmm3
 
         psadbw          xmm0,       xmm1
-        movq            xmm6,       QWORD PTR [rsi+rax+8]
+        movq            xmm2,       QWORD PTR [rsi+rax+8]
 
         movq            xmm3,       QWORD PTR [rdi+rdx+8]
         lea             rsi,        [rsi+rax*2]
 
         lea             rdi,        [rdi+rdx*2]
-        punpcklbw       xmm4,       xmm6
+        punpcklbw       xmm4,       xmm2
 
         punpcklbw       xmm5,       xmm3
         psadbw          xmm4,       xmm5
 
-        paddw           xmm7,       xmm0
-        paddw           xmm7,       xmm4
+        paddw           xmm6,       xmm0
+        paddw           xmm6,       xmm4
 
         cmp             rsi,        rcx
         jne             x16x16sad_wmt_loop
 
-        movq            xmm0,       xmm7
-        psrldq          xmm7,       8
+        movq            xmm0,       xmm6
+        psrldq          xmm6,       8
 
-        paddw           xmm0,       xmm7
+        paddw           xmm0,       xmm6
         movq            rax,        xmm0
 
     ; begin epilog
     pop rdi
     pop rsi
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
diff --git a/vp8/encoder/x86/sad_sse3.asm b/vp8/encoder/x86/sad_sse3.asm
index f0336ab..2dbcc7d 100644
--- a/vp8/encoder/x86/sad_sse3.asm
+++ b/vp8/encoder/x86/sad_sse3.asm
@@ -33,14 +33,15 @@
     movsxd      rdx,        dword ptr arg(3)    ; ref_stride
 %else
   %ifidn __OUTPUT_FORMAT__,x64
+    SAVE_XMM 7, u
     %define     src_ptr     rcx
     %define     src_stride  rdx
     %define     ref_ptr     r8
     %define     ref_stride  r9
     %define     end_ptr     r10
     %define     ret_var     r11
-    %define     result_ptr  [rsp+8+4*8]
-    %define     max_err     [rsp+8+4*8]
+    %define     result_ptr  [rsp+xmm_stack_space+8+4*8]
+    %define     max_err     [rsp+xmm_stack_space+8+4*8]
   %else
     %define     src_ptr     rdi
     %define     src_stride  rsi
@@ -72,6 +73,7 @@
     pop         rbp
 %else
   %ifidn __OUTPUT_FORMAT__,x64
+    RESTORE_XMM
   %endif
 %endif
     ret
@@ -106,6 +108,7 @@
     xchg        rbx,        rax
 %else
   %ifidn __OUTPUT_FORMAT__,x64
+    SAVE_XMM 7, u
     %define     src_ptr     rcx
     %define     src_stride  rdx
     %define     r0_ptr      rsi
@@ -113,7 +116,7 @@
     %define     r2_ptr      r11
     %define     r3_ptr      r8
     %define     ref_stride  r9
-    %define     result_ptr  [rsp+16+4*8]
+    %define     result_ptr  [rsp+xmm_stack_space+16+4*8]
     push        rsi
 
     LOAD_X4_ADDRESSES r8, r0_ptr, r1_ptr, r2_ptr, r3_ptr
@@ -151,6 +154,7 @@
 %else
   %ifidn __OUTPUT_FORMAT__,x64
     pop         rsi
+    RESTORE_XMM
   %endif
 %endif
     ret
diff --git a/vp8/encoder/x86/sad_ssse3.asm b/vp8/encoder/x86/sad_ssse3.asm
index 69c5eae..6ecf081 100644
--- a/vp8/encoder/x86/sad_ssse3.asm
+++ b/vp8/encoder/x86/sad_ssse3.asm
@@ -157,6 +157,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 5
+    SAVE_XMM 7
     push        rsi
     push        rdi
     push        rcx
@@ -253,6 +254,7 @@
     pop         rcx
     pop         rdi
     pop         rsi
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -268,6 +270,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 5
+    SAVE_XMM 7
     push        rsi
     push        rdi
     push        rcx
@@ -361,6 +364,7 @@
     pop         rcx
     pop         rdi
     pop         rsi
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
diff --git a/vp8/encoder/x86/ssim_opt.asm b/vp8/encoder/x86/ssim_opt.asm
index c267cdb..d5d267a 100644
--- a/vp8/encoder/x86/ssim_opt.asm
+++ b/vp8/encoder/x86/ssim_opt.asm
@@ -16,12 +16,12 @@
         paddusw         xmm14, xmm4  ; sum_r
         movdqa          xmm1, xmm3
         pmaddwd         xmm1, xmm1
-        paddq           xmm13, xmm1 ; sum_sq_s
+        paddd           xmm13, xmm1 ; sum_sq_s
         movdqa          xmm2, xmm4
         pmaddwd         xmm2, xmm2
-        paddq           xmm12, xmm2 ; sum_sq_r
+        paddd           xmm12, xmm2 ; sum_sq_r
         pmaddwd         xmm3, xmm4
-        paddq           xmm11, xmm3  ; sum_sxr
+        paddd           xmm11, xmm3  ; sum_sxr
 %endmacro
 
 ; Sum across the register %1 starting with q words
@@ -66,6 +66,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 9
+    SAVE_XMM 15
     push        rsi
     push        rdi
     ; end prolog
@@ -115,19 +116,20 @@
     SUM_ACROSS_Q    xmm11
 
     mov             rdi,arg(4)
-    movq            [rdi], xmm15;
+    movd            [rdi], xmm15;
     mov             rdi,arg(5)
-    movq            [rdi], xmm14;
+    movd            [rdi], xmm14;
     mov             rdi,arg(6)
-    movq            [rdi], xmm13;
+    movd            [rdi], xmm13;
     mov             rdi,arg(7)
-    movq            [rdi], xmm12;
+    movd            [rdi], xmm12;
     mov             rdi,arg(8)
-    movq            [rdi], xmm11;
+    movd            [rdi], xmm11;
 
     ; begin epilog
     pop         rdi
     pop         rsi
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -154,6 +156,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 9
+    SAVE_XMM 15
     push        rsi
     push        rdi
     ; end prolog
@@ -174,11 +177,8 @@
 NextRow2:
 
     ;grab source and reference pixels
-    movq            xmm5, [rsi]
-    movq            xmm6, [rdi]
-
-    movdqa          xmm3, xmm5
-    movdqa          xmm4, xmm6
+    movq            xmm3, [rsi]
+    movq            xmm4, [rdi]
     punpcklbw       xmm3, xmm0 ; low_s
     punpcklbw       xmm4, xmm0 ; low_r
 
@@ -197,19 +197,20 @@
     SUM_ACROSS_Q    xmm11
 
     mov             rdi,arg(4)
-    movq            [rdi], xmm15;
+    movd            [rdi], xmm15;
     mov             rdi,arg(5)
-    movq            [rdi], xmm14;
+    movd            [rdi], xmm14;
     mov             rdi,arg(6)
-    movq            [rdi], xmm13;
+    movd            [rdi], xmm13;
     mov             rdi,arg(7)
-    movq            [rdi], xmm12;
+    movd            [rdi], xmm12;
     mov             rdi,arg(8)
-    movq            [rdi], xmm11;
+    movd            [rdi], xmm11;
 
     ; begin epilog
     pop         rdi
     pop         rsi
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
diff --git a/vp8/encoder/x86/subtract_sse2.asm b/vp8/encoder/x86/subtract_sse2.asm
index 3fb23d0..95888f6 100644
--- a/vp8/encoder/x86/subtract_sse2.asm
+++ b/vp8/encoder/x86/subtract_sse2.asm
@@ -77,7 +77,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 4
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push rsi
     push rdi
diff --git a/vp8/encoder/x86/temporal_filter_apply_sse2.asm b/vp8/encoder/x86/temporal_filter_apply_sse2.asm
index 0127b01..b777ef5 100644
--- a/vp8/encoder/x86/temporal_filter_apply_sse2.asm
+++ b/vp8/encoder/x86/temporal_filter_apply_sse2.asm
@@ -26,7 +26,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 8
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push        rsi
     push        rdi
@@ -164,10 +164,10 @@
         movdqa      xmm6,           [rdi+32]
         movdqa      xmm7,           [rdi+48]
         ; += modifier
-        paddw       xmm4,           xmm0
-        paddw       xmm5,           xmm2
-        paddw       xmm6,           xmm1
-        paddw       xmm7,           xmm3
+        paddd       xmm4,           xmm0
+        paddd       xmm5,           xmm2
+        paddd       xmm6,           xmm1
+        paddd       xmm7,           xmm3
         ; write back
         movdqa      [rdi],          xmm4
         movdqa      [rdi+16],       xmm5
diff --git a/vp8/encoder/x86/variance_impl_sse2.asm b/vp8/encoder/x86/variance_impl_sse2.asm
index c2c30de..5becc73 100644
--- a/vp8/encoder/x86/variance_impl_sse2.asm
+++ b/vp8/encoder/x86/variance_impl_sse2.asm
@@ -85,6 +85,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
+    SAVE_XMM 7
     push rbx
     push rsi
     push rdi
@@ -206,6 +207,7 @@
     pop rdi
     pop rsi
     pop rbx
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -223,6 +225,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 4
+    SAVE_XMM 7
     GET_GOT     rbx
     push rsi
     push rdi
@@ -321,6 +324,7 @@
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -341,6 +345,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 6
+    SAVE_XMM 7
     GET_GOT     rbx
     push rsi
     push rdi
@@ -506,6 +511,7 @@
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -528,7 +534,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 9
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push rsi
     push rdi
@@ -805,6 +811,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
+    SAVE_XMM 7
     GET_GOT     rbx
     push rsi
     push rdi
@@ -906,6 +913,7 @@
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -925,7 +933,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push rsi
     push rdi
@@ -1041,6 +1049,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
+    SAVE_XMM 7
     GET_GOT     rbx
     push rsi
     push rdi
@@ -1127,6 +1136,7 @@
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -1146,7 +1156,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push rsi
     push rdi
@@ -1254,6 +1264,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
+    SAVE_XMM 7
     GET_GOT     rbx
     push rsi
     push rdi
@@ -1338,6 +1349,7 @@
     pop rdi
     pop rsi
     RESTORE_GOT
+    RESTORE_XMM
     UNSHADOW_ARGS
     pop         rbp
     ret
@@ -1357,7 +1369,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 7
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push rsi
     push rdi
diff --git a/vp8/encoder/x86/variance_impl_ssse3.asm b/vp8/encoder/x86/variance_impl_ssse3.asm
index 3c0fef9..a582f8d 100644
--- a/vp8/encoder/x86/variance_impl_ssse3.asm
+++ b/vp8/encoder/x86/variance_impl_ssse3.asm
@@ -34,7 +34,7 @@
     push        rbp
     mov         rbp, rsp
     SHADOW_ARGS_TO_STACK 9
-    SAVE_XMM
+    SAVE_XMM 7
     GET_GOT     rbx
     push rsi
     push rdi
diff --git a/vp8/encoder/x86/x86_csystemdependent.c b/vp8/encoder/x86/x86_csystemdependent.c
index 8f2774b..b01319f 100644
--- a/vp8/encoder/x86/x86_csystemdependent.c
+++ b/vp8/encoder/x86/x86_csystemdependent.c
@@ -271,9 +271,7 @@
         cpi->rtcd.variance.sad8x16x3             = vp8_sad8x16x3_sse3;
         cpi->rtcd.variance.sad8x8x3              = vp8_sad8x8x3_sse3;
         cpi->rtcd.variance.sad4x4x3              = vp8_sad4x4x3_sse3;
-#if !(CONFIG_REALTIME_ONLY)
         cpi->rtcd.search.full_search             = vp8_full_search_sadx3;
-#endif
         cpi->rtcd.variance.sad16x16x4d           = vp8_sad16x16x4d_sse3;
         cpi->rtcd.variance.sad16x8x4d            = vp8_sad16x8x4d_sse3;
         cpi->rtcd.variance.sad8x16x4d            = vp8_sad8x16x4d_sse3;
@@ -314,9 +312,9 @@
         cpi->rtcd.variance.sad8x16x8             = vp8_sad8x16x8_sse4;
         cpi->rtcd.variance.sad8x8x8              = vp8_sad8x8x8_sse4;
         cpi->rtcd.variance.sad4x4x8              = vp8_sad4x4x8_sse4;
-#if !(CONFIG_REALTIME_ONLY)
         cpi->rtcd.search.full_search             = vp8_full_search_sadx8;
-#endif
+
+        cpi->rtcd.quantize.quantb                = vp8_regular_quantize_b_sse4;
     }
 #endif
 
diff --git a/vp8/vp8_cx_iface.c b/vp8/vp8_cx_iface.c
index 6b13033..db60bfe 100644
--- a/vp8/vp8_cx_iface.c
+++ b/vp8/vp8_cx_iface.c
@@ -137,8 +137,8 @@
                                        const vpx_codec_enc_cfg_t *cfg,
                                        const struct vp8_extracfg *vp8_cfg)
 {
-    RANGE_CHECK(cfg, g_w,                   1, 16384);
-    RANGE_CHECK(cfg, g_h,                   1, 16384);
+    RANGE_CHECK(cfg, g_w,                   1, 16383); /* 14 bits available */
+    RANGE_CHECK(cfg, g_h,                   1, 16383); /* 14 bits available */
     RANGE_CHECK(cfg, g_timebase.den,        1, 1000000000);
     RANGE_CHECK(cfg, g_timebase.num,        1, cfg->g_timebase.den);
     RANGE_CHECK_HI(cfg, g_profile,          3);
@@ -175,16 +175,13 @@
               "or kf_max_dist instead.");
 
     RANGE_CHECK_BOOL(vp8_cfg,               enable_auto_alt_ref);
+    RANGE_CHECK(vp8_cfg, cpu_used,           -16, 16);
+
 #if !(CONFIG_REALTIME_ONLY)
     RANGE_CHECK(vp8_cfg, encoding_mode,      VP8_BEST_QUALITY_ENCODING, VP8_REAL_TIME_ENCODING);
-    RANGE_CHECK(vp8_cfg, cpu_used,           -16, 16);
     RANGE_CHECK_HI(vp8_cfg, noise_sensitivity,  6);
 #else
     RANGE_CHECK(vp8_cfg, encoding_mode,      VP8_REAL_TIME_ENCODING, VP8_REAL_TIME_ENCODING);
-
-    if (!((vp8_cfg->cpu_used >= -16 && vp8_cfg->cpu_used <= -4) || (vp8_cfg->cpu_used >= 4 && vp8_cfg->cpu_used <= 16)))
-        ERROR("cpu_used out of range [-16..-4] or [4..16]");
-
     RANGE_CHECK(vp8_cfg, noise_sensitivity,  0, 0);
 #endif
 
@@ -198,8 +195,6 @@
 #if !(CONFIG_REALTIME_ONLY)
     if (cfg->g_pass == VPX_RC_LAST_PASS)
     {
-        int              mb_r = (cfg->g_h + 15) / 16;
-        int              mb_c = (cfg->g_w + 15) / 16;
         size_t           packet_sz = sizeof(FIRSTPASS_STATS);
         int              n_packets = cfg->rc_twopass_stats_in.sz / packet_sz;
         FIRSTPASS_STATS *stats;
@@ -310,6 +305,7 @@
     }
 
     oxcf->target_bandwidth       = cfg.rc_target_bitrate;
+    oxcf->rc_max_intra_bitrate_pct = cfg.rc_max_intra_bitrate_pct;
 
     oxcf->best_allowed_q          = cfg.rc_min_quantizer;
     oxcf->worst_allowed_q         = cfg.rc_max_quantizer;
@@ -1090,7 +1086,7 @@
         {0},                /* rc_twopass_stats_in */
 #endif
         256,                /* rc_target_bandwidth */
-
+        0,                  /* rc_max_intra_bitrate_pct */
         4,                  /* rc_min_quantizer */
         63,                 /* rc_max_quantizer */
         100,                /* rc_undershoot_pct */
diff --git a/vp8/vp8cx.mk b/vp8/vp8cx.mk
index 8037f9a..c178371 100644
--- a/vp8/vp8cx.mk
+++ b/vp8/vp8cx.mk
@@ -52,6 +52,8 @@
 VP8_CX_SRCS-yes += encoder/encodemb.h
 VP8_CX_SRCS-yes += encoder/encodemv.h
 VP8_CX_SRCS-yes += encoder/firstpass.h
+VP8_CX_SRCS-yes += encoder/lookahead.c
+VP8_CX_SRCS-yes += encoder/lookahead.h
 VP8_CX_SRCS-yes += encoder/mcomp.h
 VP8_CX_SRCS-yes += encoder/modecosts.h
 VP8_CX_SRCS-yes += encoder/onyx_int.h
@@ -115,6 +117,7 @@
 VP8_CX_SRCS-$(HAVE_SSSE3) += encoder/x86/variance_impl_ssse3.asm
 VP8_CX_SRCS-$(HAVE_SSSE3) += encoder/x86/quantize_ssse3.asm
 VP8_CX_SRCS-$(HAVE_SSE4_1) += encoder/x86/sad_sse4.asm
+VP8_CX_SRCS-$(HAVE_SSE4_1) += encoder/x86/quantize_sse4.asm
 VP8_CX_SRCS-$(ARCH_X86)$(ARCH_X86_64) += encoder/x86/quantize_mmx.asm
 VP8_CX_SRCS-$(ARCH_X86)$(ARCH_X86_64) += encoder/x86/encodeopt.asm
 VP8_CX_SRCS-$(ARCH_X86_64) += encoder/x86/ssim_opt.asm
diff --git a/vpx/vp8cx.h b/vpx/vp8cx.h
index 5ff6bdc..6fd161b 100644
--- a/vpx/vp8cx.h
+++ b/vpx/vp8cx.h
@@ -138,7 +138,7 @@
      * onyx_if.c:vp8_set_speed_features().
      * \todo List highlights of the changes at various levels.
      *
-     * \note Valid range: -16..16 or {-16..-4, 4..16} w/CONFIG_REALTIME_ONLY
+     * \note Valid range: -16..16
      */
     VP8E_SET_CPUUSED           = 13,
     VP8E_SET_ENABLEAUTOALTREF,       /**< control function to enable vp8 to automatic set and use altref frame */
diff --git a/vpx/vpx_encoder.h b/vpx/vpx_encoder.h
index 3760241..6dbce0d 100644
--- a/vpx/vpx_encoder.h
+++ b/vpx/vpx_encoder.h
@@ -398,6 +398,21 @@
         unsigned int           rc_target_bitrate;
 
 
+        /*!\brief Max data rate for Intra frames
+         *
+         * This value controls additional clamping on the maximum size of a
+         * keyframe. It is expressed as a percentage of the average
+         * per-frame bitrate, with the special (and default) value 0 meaning
+         * unlimited, or no additional clamping beyond the codec's built-in
+         * algorithm.
+         *
+         * For example, to allocate no more than 4.5 frames worth of bitrate
+         * to a keyframe, set this to 450.
+         *
+         */
+        unsigned int           rc_max_intra_bitrate_pct;
+
+
         /*
          * quantizer settings
          */
diff --git a/vpx_ports/x86.h b/vpx_ports/x86.h
index 190c864..b0130fb 100644
--- a/vpx_ports/x86.h
+++ b/vpx_ports/x86.h
@@ -151,8 +151,8 @@
     __asm__ __volatile__ ("pause \n\t")
 #else
 #if ARCH_X86_64
-/* No pause intrinsic for windows x64 */
-#define x86_pause_hint()
+#define x86_pause_hint()\
+    _mm_pause();
 #else
 #define x86_pause_hint()\
     __asm pause
diff --git a/vpx_ports/x86_abi_support.asm b/vpx_ports/x86_abi_support.asm
index be64cd7..37a3205 100644
--- a/vpx_ports/x86_abi_support.asm
+++ b/vpx_ports/x86_abi_support.asm
@@ -255,21 +255,48 @@
   %define UNSHADOW_ARGS mov rsp, rbp
 %endif
 
-; must keep XMM6:XMM15 (libvpx uses XMM6 and XMM7) on Win64 ABI
-; rsp register has to be aligned
+; Win64 ABI requires that XMM6:XMM15 are callee saved
+; SAVE_XMM n, [u]
+; store registers 6-n on the stack
+; if u is specified, use unaligned movs.
+; Win64 ABI requires 16 byte stack alignment, but then pushes an 8 byte return
+; value. Typically we follow this up with 'push rbp' - re-aligning the stack -
+; but in some cases this is not done and unaligned movs must be used.
 %ifidn __OUTPUT_FORMAT__,x64
-%macro SAVE_XMM 0
-  sub rsp, 32
-  movdqa XMMWORD PTR [rsp], xmm6
-  movdqa XMMWORD PTR [rsp+16], xmm7
+%macro SAVE_XMM 1-2 a
+  %if %1 < 6
+    %error Only xmm registers 6-15 must be preserved
+  %else
+    %assign last_xmm %1
+    %define movxmm movdq %+ %2
+    %assign xmm_stack_space ((last_xmm - 5) * 16)
+    sub rsp, xmm_stack_space
+    %assign i 6
+    %rep (last_xmm - 5)
+      movxmm [rsp + ((i - 6) * 16)], xmm %+ i
+      %assign i i+1
+    %endrep
+  %endif
 %endmacro
 %macro RESTORE_XMM 0
-  movdqa xmm6, XMMWORD PTR [rsp]
-  movdqa xmm7, XMMWORD PTR [rsp+16]
-  add rsp, 32
+  %ifndef last_xmm
+    %error RESTORE_XMM must be paired with SAVE_XMM n
+  %else
+    %assign i last_xmm
+    %rep (last_xmm - 5)
+      movxmm xmm %+ i, [rsp +((i - 6) * 16)]
+      %assign i i-1
+    %endrep
+    add rsp, xmm_stack_space
+    ; there are a couple functions which return from multiple places.
+    ; otherwise, we could uncomment these:
+    ; %undef last_xmm
+    ; %undef xmm_stack_space
+    ; %undef movxmm
+  %endif
 %endmacro
 %else
-%macro SAVE_XMM 0
+%macro SAVE_XMM 1-2
 %endmacro
 %macro RESTORE_XMM 0
 %endmacro
diff --git a/vpxenc.c b/vpxenc.c
old mode 100755
new mode 100644
index 39256b6..6ea83fb
--- a/vpxenc.c
+++ b/vpxenc.c
@@ -256,6 +256,16 @@
     return stats->buf;
 }
 
+/* Stereo 3D packed frame format */
+typedef enum stereo_format
+{
+    STEREO_FORMAT_MONO       = 0,
+    STEREO_FORMAT_LEFT_RIGHT = 1,
+    STEREO_FORMAT_BOTTOM_TOP = 2,
+    STEREO_FORMAT_TOP_BOTTOM = 3,
+    STEREO_FORMAT_RIGHT_LEFT = 11
+} stereo_format_t;
+
 enum video_file_type
 {
     FILE_TYPE_RAW,
@@ -606,7 +616,8 @@
 static void
 write_webm_file_header(EbmlGlobal                *glob,
                        const vpx_codec_enc_cfg_t *cfg,
-                       const struct vpx_rational *fps)
+                       const struct vpx_rational *fps,
+                       stereo_format_t            stereo_fmt)
 {
     {
         EbmlLoc start;
@@ -650,6 +661,7 @@
                     Ebml_StartSubElement(glob, &videoStart, Video);
                     Ebml_SerializeUnsigned(glob, PixelWidth, pixelWidth);
                     Ebml_SerializeUnsigned(glob, PixelHeight, pixelHeight);
+                    Ebml_SerializeUnsigned(glob, StereoMode, stereo_fmt);
                     Ebml_SerializeFloat(glob, FrameRate, frameRate);
                     Ebml_EndSubElement(glob, &videoStart); //Video
                 }
@@ -916,6 +928,16 @@
         "Frame width");
 static const arg_def_t height           = ARG_DEF("h", "height", 1,
         "Frame height");
+static const struct arg_enum_list stereo_mode_enum[] = {
+    {"mono"      , STEREO_FORMAT_MONO},
+    {"left-right", STEREO_FORMAT_LEFT_RIGHT},
+    {"bottom-top", STEREO_FORMAT_BOTTOM_TOP},
+    {"top-bottom", STEREO_FORMAT_TOP_BOTTOM},
+    {"right-left", STEREO_FORMAT_RIGHT_LEFT},
+    {NULL, 0}
+};
+static const arg_def_t stereo_mode      = ARG_DEF_ENUM(NULL, "stereo-mode", 1,
+        "Stereo 3D video format", stereo_mode_enum);
 static const arg_def_t timebase         = ARG_DEF(NULL, "timebase", 1,
         "Stream timebase (frame duration)");
 static const arg_def_t error_resilient  = ARG_DEF(NULL, "error-resilient", 1,
@@ -926,7 +948,7 @@
 static const arg_def_t *global_args[] =
 {
     &use_yv12, &use_i420, &usage, &threads, &profile,
-    &width, &height, &timebase, &framerate, &error_resilient,
+    &width, &height, &stereo_mode, &timebase, &framerate, &error_resilient,
     &lag_in_frames, NULL
 };
 
@@ -962,11 +984,14 @@
         "Client initial buffer size (ms)");
 static const arg_def_t buf_optimal_sz     = ARG_DEF(NULL, "buf-optimal-sz", 1,
         "Client optimal buffer size (ms)");
+static const arg_def_t max_intra_rate_pct = ARG_DEF(NULL, "max-intra-rate", 1,
+        "Max I-frame bitrate (pct)");
 static const arg_def_t *rc_args[] =
 {
     &dropframe_thresh, &resize_allowed, &resize_up_thresh, &resize_down_thresh,
     &end_usage, &target_bitrate, &min_quantizer, &max_quantizer,
     &undershoot_pct, &overshoot_pct, &buf_sz, &buf_initial_sz, &buf_optimal_sz,
+    &max_intra_rate_pct,
     NULL
 };
 
@@ -1084,7 +1109,6 @@
 
 #define ARG_CTRL_CNT_MAX 10
 
-
 int main(int argc, const char **argv_)
 {
     vpx_codec_ctx_t        encoder;
@@ -1120,6 +1144,7 @@
     uint64_t                 psnr_samples_total = 0;
     double                   psnr_totals[4] = {0, 0, 0, 0};
     int                      psnr_count = 0;
+    stereo_format_t          stereo_fmt = STEREO_FORMAT_MONO;
 
     exec_name = argv_[0];
     ebml.last_pts_ms = -1;
@@ -1259,6 +1284,8 @@
             cfg.g_w = arg_parse_uint(&arg);
         else if (arg_match(&arg, &height, argi))
             cfg.g_h = arg_parse_uint(&arg);
+        else if (arg_match(&arg, &stereo_mode, argi))
+            stereo_fmt = arg_parse_enum_or_int(&arg);
         else if (arg_match(&arg, &timebase, argi))
             cfg.g_timebase = arg_parse_rational(&arg);
         else if (arg_match(&arg, &error_resilient, argi))
@@ -1279,6 +1306,8 @@
             cfg.rc_end_usage = arg_parse_enum_or_int(&arg);
         else if (arg_match(&arg, &target_bitrate, argi))
             cfg.rc_target_bitrate = arg_parse_uint(&arg);
+        else if (arg_match(&arg, &max_intra_rate_pct, argi))
+            cfg.rc_max_intra_bitrate_pct = arg_parse_uint(&arg);
         else if (arg_match(&arg, &min_quantizer, argi))
             cfg.rc_min_quantizer = arg_parse_uint(&arg);
         else if (arg_match(&arg, &max_quantizer, argi))
@@ -1557,7 +1586,7 @@
         if(write_webm)
         {
             ebml.stream = outfile;
-            write_webm_file_header(&ebml, &cfg, &arg_framerate);
+            write_webm_file_header(&ebml, &cfg, &arg_framerate, stereo_fmt);
         }
         else
             write_ivf_file_header(outfile, &cfg, codec->fourcc, 0);