Merge remote branch 'origin/master' into experimental

Change-Id: Iaf6e9e14d0cfe5cef3895cfb68524d51139a6d23

11 files changed, 88 insertions, 61 deletions

diff --git a/vp8/encoder/block.h b/vp8/encoder/block.h
index 82a93cdab..0d14b545c 100644
--- a/vp8/encoder/block.h
+++ b/vp8/encoder/block.h
@@ -65,7 +65,9 @@ typedef struct
 {
     DECLARE_ALIGNED(16, short, src_diff[400]);       // 16x16 Y 8x8 U 8x8 V 4x4 2nd Y
     DECLARE_ALIGNED(16, short, coeff[400]);     // 16x16 Y 8x8 U 8x8 V 4x4 2nd Y
+    DECLARE_ALIGNED(16, unsigned char, thismb[256]);
 
+    unsigned char *thismb_ptr;
     // 16 Y blocks, 4 U blocks, 4 V blocks, 1 DC 2nd order block each with 16 entries
     BLOCK block[25];
 
diff --git a/vp8/encoder/encodeframe.c b/vp8/encoder/encodeframe.c
index 7a6650696..2c0048daa 100644
--- a/vp8/encoder/encodeframe.c
+++ b/vp8/encoder/encodeframe.c
@@ -292,6 +292,9 @@ static void build_activity_map( VP8_COMP *cpi )
             xd->left_available = (mb_col != 0);
             recon_yoffset += 16;
 #endif
+            //Copy current mb to a buffer
+            RECON_INVOKE(&xd->rtcd->recon, copy16x16)(x->src.y_buffer, x->src.y_stride, x->thismb, 16);
+
             // measure activity
             mb_activity = mb_activity_measure( cpi, x, mb_row, mb_col );
 
@@ -432,6 +435,9 @@ void encode_mb_row(VP8_COMP *cpi,
         x->rddiv = cpi->RDDIV;
         x->rdmult = cpi->RDMULT;
 
+        //Copy current mb to a buffer
+        RECON_INVOKE(&xd->rtcd->recon, copy16x16)(x->src.y_buffer, x->src.y_stride, x->thismb, 16);
+
 #if CONFIG_MULTITHREAD
         if ((cpi->b_multi_threaded != 0) && (mb_row != 0))
         {
@@ -1013,14 +1019,18 @@ void vp8_build_block_offsets(MACROBLOCK *x)
     vp8_build_block_doffsets(&x->e_mbd);
 
     // y blocks
+    x->thismb_ptr = &x->thismb[0];
     for (br = 0; br < 4; br++)
     {
         for (bc = 0; bc < 4; bc++)
         {
             BLOCK *this_block = &x->block[block];
-            this_block->base_src = &x->src.y_buffer;
-            this_block->src_stride = x->src.y_stride;
-            this_block->src = 4 * br * this_block->src_stride + 4 * bc;
+            //this_block->base_src = &x->src.y_buffer;
+            //this_block->src_stride = x->src.y_stride;
+            //this_block->src = 4 * br * this_block->src_stride + 4 * bc;
+            this_block->base_src = &x->thismb_ptr;
+            this_block->src_stride = 16;
+            this_block->src = 4 * br * 16 + 4 * bc;
             ++block;
         }
     }
diff --git a/vp8/encoder/encodeintra.c b/vp8/encoder/encodeintra.c
index f076bbbb3..59db0253b 100644
--- a/vp8/encoder/encodeintra.c
+++ b/vp8/encoder/encodeintra.c
@@ -91,11 +91,11 @@ void vp8_encode_intra4x4mby(const VP8_ENCODER_RTCD *rtcd, MACROBLOCK *mb)
 
 void vp8_encode_intra16x16mby(const VP8_ENCODER_RTCD *rtcd, MACROBLOCK *x)
 {
-    int b;
+    BLOCK *b = &x->block[0];
 
     RECON_INVOKE(&rtcd->common->recon, build_intra_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);
+    ENCODEMB_INVOKE(&rtcd->encodemb, submby)(x->src_diff, *(b->base_src), x->e_mbd.predictor, b->src_stride);
 
     vp8_transform_intra_mby(x);
 
diff --git a/vp8/encoder/encodemb.c b/vp8/encoder/encodemb.c
index 9c223924d..eb89bba0a 100644
--- a/vp8/encoder/encodemb.c
+++ b/vp8/encoder/encodemb.c
@@ -101,7 +101,9 @@ void vp8_subtract_mby_c(short *diff, unsigned char *src, unsigned char *pred, in
 
 static void vp8_subtract_mb(const VP8_ENCODER_RTCD *rtcd, MACROBLOCK *x)
 {
-    ENCODEMB_INVOKE(&rtcd->encodemb, submby)(x->src_diff, x->src.y_buffer, x->e_mbd.predictor, x->src.y_stride);
+    BLOCK *b = &x->block[0];
+
+    ENCODEMB_INVOKE(&rtcd->encodemb, submby)(x->src_diff, *(b->base_src), x->e_mbd.predictor, b->src_stride);
     ENCODEMB_INVOKE(&rtcd->encodemb, submbuv)(x->src_diff, x->src.u_buffer, x->src.v_buffer, x->e_mbd.predictor, x->src.uv_stride);
 }
 
@@ -602,9 +604,11 @@ void vp8_encode_inter16x16(const VP8_ENCODER_RTCD *rtcd, MACROBLOCK *x)
 /* this funciton is used by first pass only */
 void vp8_encode_inter16x16y(const VP8_ENCODER_RTCD *rtcd, MACROBLOCK *x)
 {
+    BLOCK *b = &x->block[0];
+
     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);
+    ENCODEMB_INVOKE(&rtcd->encodemb, submby)(x->src_diff, *(b->base_src), x->e_mbd.predictor, b->src_stride);
 
     transform_mby(x);
 
diff --git a/vp8/encoder/ethreading.c b/vp8/encoder/ethreading.c
index 420ed8eff..1a37f03b9 100644
--- a/vp8/encoder/ethreading.c
+++ b/vp8/encoder/ethreading.c
@@ -148,6 +148,9 @@ THREAD_FUNCTION thread_encoding_proc(void *p_data)
                     x->rddiv = cpi->RDDIV;
                     x->rdmult = cpi->RDMULT;
 
+                    //Copy current mb to a buffer
+                    RECON_INVOKE(&xd->rtcd->recon, copy16x16)(x->src.y_buffer, x->src.y_stride, x->thismb, 16);
+
                     if (cpi->oxcf.tuning == VP8_TUNE_SSIM)
                         vp8_activity_masking(cpi, x);
 
diff --git a/vp8/encoder/firstpass.c b/vp8/encoder/firstpass.c
index 0b1b48dfb..d3a874cfd 100644
--- a/vp8/encoder/firstpass.c
+++ b/vp8/encoder/firstpass.c
@@ -579,6 +579,9 @@ void vp8_first_pass(VP8_COMP *cpi)
             xd->dst.v_buffer = new_yv12->v_buffer + recon_uvoffset;
             xd->left_available = (mb_col != 0);
 
+            //Copy current mb to a buffer
+            RECON_INVOKE(&xd->rtcd->recon, copy16x16)(x->src.y_buffer, x->src.y_stride, x->thismb, 16);
+
             // do intra 16x16 prediction
             this_error = vp8_encode_intra(cpi, x, use_dc_pred);
 
diff --git a/vp8/encoder/mcomp.c b/vp8/encoder/mcomp.c
index 416948870..6f314a386 100644
--- a/vp8/encoder/mcomp.c
+++ b/vp8/encoder/mcomp.c
@@ -286,8 +286,8 @@ int vp8_find_best_sub_pixel_step_iteratively(MACROBLOCK *x, BLOCK *b, BLOCKD *d,
     bestmv->as_mv.row = br << 1;
     bestmv->as_mv.col = bc << 1;
 
-    if ((abs(bestmv->as_mv.col - ref_mv->as_mv.col) > MAX_FULL_PEL_VAL) ||
-        (abs(bestmv->as_mv.row - ref_mv->as_mv.row) > MAX_FULL_PEL_VAL))
+    if ((abs(bestmv->as_mv.col - ref_mv->as_mv.col) > (MAX_FULL_PEL_VAL<<3)) ||
+        (abs(bestmv->as_mv.row - ref_mv->as_mv.row) > (MAX_FULL_PEL_VAL<<3)))
         return INT_MAX;
 
     return besterr;
diff --git a/vp8/encoder/mcomp.h b/vp8/encoder/mcomp.h
index 44ed055db..416c4d5eb 100644
--- a/vp8/encoder/mcomp.h
+++ b/vp8/encoder/mcomp.h
@@ -22,7 +22,7 @@ extern void accum_mv_refs(MB_PREDICTION_MODE, const int near_mv_ref_cts[4]);
 
 
 #define MAX_MVSEARCH_STEPS 8                                    // The maximum number of steps in a step search given the largest allowed initial step
-#define MAX_FULL_PEL_VAL ((1 << (MAX_MVSEARCH_STEPS+3)) - 8)    // Max full pel mv specified in 1/8 pel units
+#define MAX_FULL_PEL_VAL ((1 << (MAX_MVSEARCH_STEPS)) - 1)      // Max full pel mv specified in 1 pel units
 #define MAX_FIRST_STEP (1 << (MAX_MVSEARCH_STEPS-1))            // Maximum size of the first step in full pel units
 
 extern void print_mode_context(void);
diff --git a/vp8/encoder/pickinter.c b/vp8/encoder/pickinter.c
index b60d2419f..fac068f7a 100644
--- a/vp8/encoder/pickinter.c
+++ b/vp8/encoder/pickinter.c
@@ -588,10 +588,10 @@ void vp8_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset,
 
             /* adjust mvp to make sure it is within MV range */
             vp8_clamp_mv(&mvp,
-                         best_ref_mv.as_mv.col - MAX_FULL_PEL_VAL,
-                         best_ref_mv.as_mv.col + MAX_FULL_PEL_VAL,
-                         best_ref_mv.as_mv.row - MAX_FULL_PEL_VAL,
-                         best_ref_mv.as_mv.row + MAX_FULL_PEL_VAL);
+                         best_ref_mv.as_mv.col - (MAX_FULL_PEL_VAL<<3),
+                         best_ref_mv.as_mv.col + (MAX_FULL_PEL_VAL<<3),
+                         best_ref_mv.as_mv.row - (MAX_FULL_PEL_VAL<<3),
+                         best_ref_mv.as_mv.row + (MAX_FULL_PEL_VAL<<3));
         }
 
         switch (this_mode)
@@ -610,7 +610,7 @@ void vp8_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset,
                 rate2 += rate;
                 distortion2 = VARIANCE_INVOKE
                                 (&cpi->rtcd.variance, var16x16)(
-                                    x->src.y_buffer, x->src.y_stride,
+                                    *(b->base_src), b->src_stride,
                                     x->e_mbd.predictor, 16, &sse);
                 this_rd = RDCOST(x->rdmult, x->rddiv, rate2, distortion2);
 
@@ -635,7 +635,7 @@ void vp8_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset,
             RECON_INVOKE(&cpi->common.rtcd.recon, build_intra_predictors_mby)
                 (&x->e_mbd);
             distortion2 = VARIANCE_INVOKE(&cpi->rtcd.variance, var16x16)
-                                          (x->src.y_buffer, x->src.y_stride,
+                                          (*(b->base_src), b->src_stride,
                                           x->e_mbd.predictor, 16, &sse);
             rate2 += x->mbmode_cost[x->e_mbd.frame_type][x->e_mbd.mode_info_context->mbmi.mode];
             this_rd = RDCOST(x->rdmult, x->rddiv, rate2, distortion2);
@@ -681,10 +681,14 @@ void vp8_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset,
                 mvp.as_int = best_ref_mv.as_int;
             }
 
-            col_min = (best_ref_mv.as_mv.col - MAX_FULL_PEL_VAL) >>3;
-            col_max = (best_ref_mv.as_mv.col + MAX_FULL_PEL_VAL) >>3;
-            row_min = (best_ref_mv.as_mv.row - MAX_FULL_PEL_VAL) >>3;
-            row_max = (best_ref_mv.as_mv.row + MAX_FULL_PEL_VAL) >>3;
+            col_min = (best_ref_mv.as_mv.col < 0)?(-((abs(best_ref_mv.as_mv.col))>>3) - MAX_FULL_PEL_VAL)
+                                                 :((best_ref_mv.as_mv.col>>3) - MAX_FULL_PEL_VAL);
+            col_max = (best_ref_mv.as_mv.col < 0)?(-((abs(best_ref_mv.as_mv.col))>>3) + MAX_FULL_PEL_VAL)
+                                                 :((best_ref_mv.as_mv.col>>3) + MAX_FULL_PEL_VAL);
+            row_min = (best_ref_mv.as_mv.row < 0)?(-((abs(best_ref_mv.as_mv.row))>>3) - MAX_FULL_PEL_VAL)
+                                                 :((best_ref_mv.as_mv.row>>3) - MAX_FULL_PEL_VAL);
+            row_max = (best_ref_mv.as_mv.row < 0)?(-((abs(best_ref_mv.as_mv.row))>>3) + MAX_FULL_PEL_VAL)
+                                                 :((best_ref_mv.as_mv.row>>3) + MAX_FULL_PEL_VAL);
 
             // Get intersection of UMV window and valid MV window to reduce # of checks in diamond search.
             if (x->mv_col_min < col_min )
@@ -904,6 +908,7 @@ void vp8_pick_intra_mode(VP8_COMP *cpi, MACROBLOCK *x, int *rate_)
     MB_PREDICTION_MODE mode, best_mode = DC_PRED;
     int this_rd;
     unsigned int sse;
+    BLOCK *b = &x->block[0];
 
     x->e_mbd.mode_info_context->mbmi.ref_frame = INTRA_FRAME;
 
@@ -915,7 +920,7 @@ void vp8_pick_intra_mode(VP8_COMP *cpi, MACROBLOCK *x, int *rate_)
         RECON_INVOKE(&cpi->common.rtcd.recon, build_intra_predictors_mby)
             (&x->e_mbd);
         distortion = VARIANCE_INVOKE(&cpi->rtcd.variance, var16x16)
-            (x->src.y_buffer, x->src.y_stride, x->e_mbd.predictor, 16, &sse);
+            (*(b->base_src), b->src_stride, x->e_mbd.predictor, 16, &sse);
         rate = x->mbmode_cost[x->e_mbd.frame_type][mode];
         this_rd = RDCOST(x->rdmult, x->rddiv, rate, distortion);
 
diff --git a/vp8/encoder/picklpf.c b/vp8/encoder/picklpf.c
index 4f9d4126e..3d47412f7 100644
--- a/vp8/encoder/picklpf.c
+++ b/vp8/encoder/picklpf.c
@@ -141,13 +141,11 @@ static int get_max_filter_level(VP8_COMP *cpi, int base_qindex)
     // jbb chg: 20100118 - not so any more with this overquant stuff allow high values
     // with lots of intra coming in.
     int max_filter_level = MAX_LOOP_FILTER ;//* 3 / 4;
+    (void)base_qindex;
 
     if (cpi->twopass.section_intra_rating > 8)
         max_filter_level = MAX_LOOP_FILTER * 3 / 4;
 
-    (void) cpi;
-    (void) base_qindex;
-
     return max_filter_level;
 }
 
@@ -157,8 +155,8 @@ void vp8cx_pick_filter_level_fast(YV12_BUFFER_CONFIG *sd, VP8_COMP *cpi)
 
     int best_err = 0;
     int filt_err = 0;
-    int min_filter_level = 0;
-    int max_filter_level = MAX_LOOP_FILTER * 3 / 4;   // PGW August 2006: Highest filter values almost always a bad idea
+    int min_filter_level = get_min_filter_level(cpi, cm->base_qindex);
+    int max_filter_level = get_max_filter_level(cpi, cm->base_qindex);
     int filt_val;
     int best_filt_val = cm->filter_level;
 
@@ -171,10 +169,6 @@ void vp8cx_pick_filter_level_fast(YV12_BUFFER_CONFIG *sd, VP8_COMP *cpi)
     else
         cm->sharpness_level = cpi->oxcf.Sharpness;
 
-    // Enforce a minimum filter level based upon Q
-    min_filter_level = get_min_filter_level(cpi, cm->base_qindex);
-    max_filter_level = get_max_filter_level(cpi, cm->base_qindex);
-
     // Start the search at the previous frame filter level unless it is now out of range.
     if (cm->filter_level < min_filter_level)
         cm->filter_level = min_filter_level;
@@ -294,8 +288,8 @@ void vp8cx_pick_filter_level(YV12_BUFFER_CONFIG *sd, VP8_COMP *cpi)
 
     int best_err = 0;
     int filt_err = 0;
-    int min_filter_level;
-    int max_filter_level;
+    int min_filter_level = get_min_filter_level(cpi, cm->base_qindex);
+    int max_filter_level = get_max_filter_level(cpi, cm->base_qindex);
 
     int filter_step;
     int filt_high = 0;
@@ -329,10 +323,6 @@ void vp8cx_pick_filter_level(YV12_BUFFER_CONFIG *sd, VP8_COMP *cpi)
     else
         cm->sharpness_level = cpi->oxcf.Sharpness;
 
-    // Enforce a minimum filter level based upon Q
-    min_filter_level = get_min_filter_level(cpi, cm->base_qindex);
-    max_filter_level = get_max_filter_level(cpi, cm->base_qindex);
-
     // Start the search at the previous frame filter level unless it is now out of range.
     filt_mid = cm->filter_level;
 
diff --git a/vp8/encoder/rdopt.c b/vp8/encoder/rdopt.c
index 7c05e33a1..ec8957f80 100644
--- a/vp8/encoder/rdopt.c
+++ b/vp8/encoder/rdopt.c
@@ -605,8 +605,8 @@ static void macro_block_yrd( MACROBLOCK *mb,
     BLOCK *beptr;
     int d;
 
-    ENCODEMB_INVOKE(rtcd, submby)( mb->src_diff, mb->src.y_buffer,
-                                   mb->e_mbd.predictor, mb->src.y_stride );
+    ENCODEMB_INVOKE(rtcd, submby)( mb->src_diff, *(mb->block[0].base_src),
+                                   mb->e_mbd.predictor, mb->block[0].src_stride );
 
     // Fdct and building the 2nd order block
     for (beptr = mb->block; beptr < mb->block + 16; beptr += 2)
@@ -1398,10 +1398,14 @@ static int vp8_rd_pick_best_mbsegmentation(VP8_COMP *cpi, MACROBLOCK *x,
 
         if (bsi.segment_rd < best_rd)
         {
-            int col_min = (best_ref_mv->as_mv.col - MAX_FULL_PEL_VAL) >>3;
-            int col_max = (best_ref_mv->as_mv.col + MAX_FULL_PEL_VAL) >>3;
-            int row_min = (best_ref_mv->as_mv.row - MAX_FULL_PEL_VAL) >>3;
-            int row_max = (best_ref_mv->as_mv.row + MAX_FULL_PEL_VAL) >>3;
+            int col_min = (best_ref_mv->as_mv.col < 0)?(-((abs(best_ref_mv->as_mv.col))>>3) - MAX_FULL_PEL_VAL)
+                                                      :((best_ref_mv->as_mv.col>>3) - MAX_FULL_PEL_VAL);
+            int col_max = (best_ref_mv->as_mv.col < 0)?(-((abs(best_ref_mv->as_mv.col))>>3) + MAX_FULL_PEL_VAL)
+                                                      :((best_ref_mv->as_mv.col>>3) + MAX_FULL_PEL_VAL);
+            int row_min = (best_ref_mv->as_mv.row < 0)?(-((abs(best_ref_mv->as_mv.row))>>3) - MAX_FULL_PEL_VAL)
+                                                      :((best_ref_mv->as_mv.row>>3) - MAX_FULL_PEL_VAL);
+            int row_max = (best_ref_mv->as_mv.row < 0)?(-((abs(best_ref_mv->as_mv.row))>>3) + MAX_FULL_PEL_VAL)
+                                                      :((best_ref_mv->as_mv.row>>3) + MAX_FULL_PEL_VAL);
 
             int tmp_col_min = x->mv_col_min;
             int tmp_col_max = x->mv_col_max;
@@ -1701,6 +1705,8 @@ void vp8_cal_sad(VP8_COMP *cpi, MACROBLOCKD *xd, MACROBLOCK *x, int recon_yoffse
 {
 
     int near_sad[8] = {0}; // 0-cf above, 1-cf left, 2-cf aboveleft, 3-lf current, 4-lf above, 5-lf left, 6-lf right, 7-lf below
+    BLOCK *b = &x->block[0];
+    unsigned char *src_y_ptr = *(b->base_src);
 
     //calculate sad for current frame 3 nearby MBs.
     if( xd->mb_to_top_edge==0 && xd->mb_to_left_edge ==0)
@@ -1709,16 +1715,16 @@ void vp8_cal_sad(VP8_COMP *cpi, MACROBLOCKD *xd, MACROBLOCK *x, int recon_yoffse
     }else if(xd->mb_to_top_edge==0)
     {   //only has left MB for sad calculation.
         near_sad[0] = near_sad[2] = INT_MAX;
-        near_sad[1] = cpi->fn_ptr[BLOCK_16X16].sdf(x->src.y_buffer, x->src.y_stride, xd->dst.y_buffer - 16,xd->dst.y_stride, 0x7fffffff);
+        near_sad[1] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, xd->dst.y_buffer - 16,xd->dst.y_stride, 0x7fffffff);
     }else if(xd->mb_to_left_edge ==0)
     {   //only has left MB for sad calculation.
         near_sad[1] = near_sad[2] = INT_MAX;
-        near_sad[0] = cpi->fn_ptr[BLOCK_16X16].sdf(x->src.y_buffer, x->src.y_stride, xd->dst.y_buffer - xd->dst.y_stride *16,xd->dst.y_stride, 0x7fffffff);
+        near_sad[0] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, xd->dst.y_buffer - xd->dst.y_stride *16,xd->dst.y_stride, 0x7fffffff);
     }else
     {
-        near_sad[0] = cpi->fn_ptr[BLOCK_16X16].sdf(x->src.y_buffer, x->src.y_stride, xd->dst.y_buffer - xd->dst.y_stride *16,xd->dst.y_stride, 0x7fffffff);
-        near_sad[1] = cpi->fn_ptr[BLOCK_16X16].sdf(x->src.y_buffer, x->src.y_stride, xd->dst.y_buffer - 16,xd->dst.y_stride, 0x7fffffff);
-        near_sad[2] = cpi->fn_ptr[BLOCK_16X16].sdf(x->src.y_buffer, x->src.y_stride, xd->dst.y_buffer - xd->dst.y_stride *16 -16,xd->dst.y_stride, 0x7fffffff);
+        near_sad[0] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, xd->dst.y_buffer - xd->dst.y_stride *16,xd->dst.y_stride, 0x7fffffff);
+        near_sad[1] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, xd->dst.y_buffer - 16,xd->dst.y_stride, 0x7fffffff);
+        near_sad[2] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, xd->dst.y_buffer - xd->dst.y_stride *16 -16,xd->dst.y_stride, 0x7fffffff);
     }
 
     if(cpi->common.last_frame_type != KEY_FRAME)
@@ -1733,14 +1739,14 @@ void vp8_cal_sad(VP8_COMP *cpi, MACROBLOCKD *xd, MACROBLOCK *x, int recon_yoffse
         if(xd->mb_to_bottom_edge==0) near_sad[7] = INT_MAX;
 
         if(near_sad[4] != INT_MAX)
-            near_sad[4] = cpi->fn_ptr[BLOCK_16X16].sdf(x->src.y_buffer, x->src.y_stride, pre_y_buffer - pre_y_stride *16, pre_y_stride, 0x7fffffff);
+            near_sad[4] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, pre_y_buffer - pre_y_stride *16, pre_y_stride, 0x7fffffff);
         if(near_sad[5] != INT_MAX)
-            near_sad[5] = cpi->fn_ptr[BLOCK_16X16].sdf(x->src.y_buffer, x->src.y_stride, pre_y_buffer - 16, pre_y_stride, 0x7fffffff);
-        near_sad[3] = cpi->fn_ptr[BLOCK_16X16].sdf(x->src.y_buffer, x->src.y_stride, pre_y_buffer, pre_y_stride, 0x7fffffff);
+            near_sad[5] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, pre_y_buffer - 16, pre_y_stride, 0x7fffffff);
+        near_sad[3] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, pre_y_buffer, pre_y_stride, 0x7fffffff);
         if(near_sad[6] != INT_MAX)
-            near_sad[6] = cpi->fn_ptr[BLOCK_16X16].sdf(x->src.y_buffer, x->src.y_stride, pre_y_buffer + 16, pre_y_stride, 0x7fffffff);
+            near_sad[6] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, pre_y_buffer + 16, pre_y_stride, 0x7fffffff);
         if(near_sad[7] != INT_MAX)
-            near_sad[7] = cpi->fn_ptr[BLOCK_16X16].sdf(x->src.y_buffer, x->src.y_stride, pre_y_buffer + pre_y_stride *16, pre_y_stride, 0x7fffffff);
+            near_sad[7] = cpi->fn_ptr[BLOCK_16X16].sdf(src_y_ptr, b->src_stride, pre_y_buffer + pre_y_stride *16, pre_y_stride, 0x7fffffff);
     }
 
     if(cpi->common.last_frame_type != KEY_FRAME)
@@ -1944,10 +1950,10 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int
 
             /* adjust mvp to make sure it is within MV range */
             vp8_clamp_mv(&mvp,
-                         best_ref_mv.as_mv.col - MAX_FULL_PEL_VAL,
-                         best_ref_mv.as_mv.col + MAX_FULL_PEL_VAL,
-                         best_ref_mv.as_mv.row - MAX_FULL_PEL_VAL,
-                         best_ref_mv.as_mv.row + MAX_FULL_PEL_VAL);
+                         best_ref_mv.as_mv.col - (MAX_FULL_PEL_VAL<<3),
+                         best_ref_mv.as_mv.col + (MAX_FULL_PEL_VAL<<3),
+                         best_ref_mv.as_mv.row - (MAX_FULL_PEL_VAL<<3),
+                         best_ref_mv.as_mv.row + (MAX_FULL_PEL_VAL<<3));
         }
 
         // Check to see if the testing frequency for this mode is at its max
@@ -2079,10 +2085,14 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int
 
             int sadpb = x->sadperbit16;
 
-            int col_min = (best_ref_mv.as_mv.col - MAX_FULL_PEL_VAL) >>3;
-            int col_max = (best_ref_mv.as_mv.col + MAX_FULL_PEL_VAL) >>3;
-            int row_min = (best_ref_mv.as_mv.row - MAX_FULL_PEL_VAL) >>3;
-            int row_max = (best_ref_mv.as_mv.row + MAX_FULL_PEL_VAL) >>3;
+            int col_min = (best_ref_mv.as_mv.col < 0)?(-((abs(best_ref_mv.as_mv.col))>>3) - MAX_FULL_PEL_VAL)
+                                                     :((best_ref_mv.as_mv.col>>3) - MAX_FULL_PEL_VAL);
+            int col_max = (best_ref_mv.as_mv.col < 0)?(-((abs(best_ref_mv.as_mv.col))>>3) + MAX_FULL_PEL_VAL)
+                                                     :((best_ref_mv.as_mv.col>>3) + MAX_FULL_PEL_VAL);
+            int row_min = (best_ref_mv.as_mv.row < 0)?(-((abs(best_ref_mv.as_mv.row))>>3) - MAX_FULL_PEL_VAL)
+                                                     :((best_ref_mv.as_mv.row>>3) - MAX_FULL_PEL_VAL);
+            int row_max = (best_ref_mv.as_mv.row < 0)?(-((abs(best_ref_mv.as_mv.row))>>3) + MAX_FULL_PEL_VAL)
+                                                     :((best_ref_mv.as_mv.row>>3) + MAX_FULL_PEL_VAL);
 
             int tmp_col_min = x->mv_col_min;
             int tmp_col_max = x->mv_col_max;
@@ -2233,7 +2243,7 @@ void vp8_rd_pick_inter_mode(VP8_COMP *cpi, MACROBLOCK *x, int recon_yoffset, int
                     threshold = x->encode_breakout;
 
                 var = VARIANCE_INVOKE(&cpi->rtcd.variance, var16x16)
-                        (x->src.y_buffer, x->src.y_stride,
+                        (*(b->base_src), b->src_stride,
                         x->e_mbd.predictor, 16, &sse);
 
                 if (sse < threshold)