ref: b6c2d872f0a48350cd5d4fcdbc7af98e4c46af27
parent: 650e6324004bbabd855180c1314074709aed0531
author: Ronald S. Bultje <[email protected]>
date: Wed May 1 10:45:27 EDT 2013
Fix some crashes in sb8x8 experiment. Change-Id: I390bb1cedc835f439fd5dd6cda6572b29cbb139c
--- a/vp9/common/vp9_rtcd_defs.sh
+++ b/vp9/common/vp9_rtcd_defs.sh
@@ -342,6 +342,12 @@
vp9_variance8x8_sse2=vp9_variance8x8_wmt
vp9_variance8x8_mmx=vp9_variance8x8_mmx
+prototype unsigned int vp9_variance8x4 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
+specialize vp9_variance8x4
+
+prototype unsigned int vp9_variance4x8 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
+specialize vp9_variance4x8
+
prototype unsigned int vp9_variance4x4 "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse"
specialize vp9_variance4x4 mmx sse2
vp9_variance4x4_sse2=vp9_variance4x4_wmt
--- a/vp9/encoder/vp9_onyx_if.c
+++ b/vp9/encoder/vp9_onyx_if.c
@@ -1647,6 +1647,12 @@
BFP(BLOCK_8X8, vp9_sad8x8, vp9_variance8x8, vp9_sub_pixel_variance8x8,
NULL, NULL, NULL, vp9_sad8x8x3, vp9_sad8x8x8, vp9_sad8x8x4d)
+#if CONFIG_SB8X8
+ BFP(BLOCK_4X8, NULL, vp9_variance4x8, NULL,
+ NULL, NULL, NULL, NULL, NULL, NULL)
+ BFP(BLOCK_8X4, NULL, vp9_variance8x4, NULL,
+ NULL, NULL, NULL, NULL, NULL, NULL)
+#endif
BFP(BLOCK_4X4, vp9_sad4x4, vp9_variance4x4, vp9_sub_pixel_variance4x4,
NULL, NULL, NULL, vp9_sad4x4x3, vp9_sad4x4x8, vp9_sad4x4x4d)
--- a/vp9/encoder/vp9_onyx_int.h
+++ b/vp9/encoder/vp9_onyx_int.h
@@ -283,6 +283,8 @@
enum BlockSize {#if CONFIG_SB8X8
BLOCK_4X4,
+ BLOCK_4X8,
+ BLOCK_8X4,
BLOCK_8X8,
BLOCK_8X16,
BLOCK_16X8,
--- a/vp9/encoder/vp9_rdopt.c
+++ b/vp9/encoder/vp9_rdopt.c
@@ -1874,7 +1874,7 @@
raster_block_offset_uint8(&x->e_mbd, BLOCK_SIZE_SB8X8, 0, n,
x->plane[0].src.buf,
x->plane[0].src.stride);
- assert(((intptr_t)x->e_mbd.plane[0].pre[0].buf & 0xf) == 0);
+ assert(((intptr_t)x->e_mbd.plane[0].pre[0].buf & 0x7) == 0);
x->e_mbd.plane[0].pre[0].buf =
raster_block_offset_uint8(&x->e_mbd, BLOCK_SIZE_SB8X8, 0, n,
x->e_mbd.plane[0].pre[0].buf,
@@ -3295,6 +3295,11 @@
case BLOCK_32X16: return BLOCK_16X8;
case BLOCK_16X32: return BLOCK_8X16;
case BLOCK_16X16: return BLOCK_8X8;
+#if CONFIG_SB8X8
+ case BLOCK_16X8: return BLOCK_8X4;
+ case BLOCK_8X16: return BLOCK_4X8;
+ case BLOCK_8X8: return BLOCK_4X4;
+#endif
default:
assert(0);
return -1;
@@ -3310,6 +3315,11 @@
case BLOCK_SIZE_SB32X16: return BLOCK_32X16;
case BLOCK_SIZE_SB16X32: return BLOCK_16X32;
case BLOCK_SIZE_MB16X16: return BLOCK_16X16;
+#if CONFIG_SB8X8
+ case BLOCK_SIZE_SB16X8: return BLOCK_16X8;
+ case BLOCK_SIZE_SB8X16: return BLOCK_8X16;
+ case BLOCK_SIZE_SB8X8: return BLOCK_8X8;
+#endif
default:
assert(0);
return -1;
@@ -4927,7 +4937,9 @@
if (cpi->Speed == 0
|| (cpi->Speed > 0 && (ref_frame_mask & (1 << INTRA_FRAME)))) {mbmi->mode = DC_PRED;
- for (i = 0; i <= ((bsize < BLOCK_SIZE_SB64X64) ? TX_16X16 : TX_32X32);
+ for (i = 0; i <= (bsize < BLOCK_SIZE_MB16X16 ? TX_4X4 :
+ (bsize < BLOCK_SIZE_SB32X32 ? TX_8X8 :
+ (bsize < BLOCK_SIZE_SB64X64 ? TX_16X16 : TX_32X32)));
i++) {mbmi->txfm_size = i;
rd_pick_intra_sbuv_mode(cpi, x, &rate_uv_intra[i], &rate_uv_tokenonly[i],
@@ -5097,6 +5109,8 @@
bsize, txfm_cache);
uv_tx = mbmi->txfm_size;
+ if (bsize < BLOCK_SIZE_MB16X16 && uv_tx == TX_8X8)
+ uv_tx = TX_4X4;
if (bsize < BLOCK_SIZE_SB32X32 && uv_tx == TX_16X16)
uv_tx = TX_8X8;
else if (bsize < BLOCK_SIZE_SB64X64 && uv_tx == TX_32X32)
--- a/vp9/encoder/vp9_variance_c.c
+++ b/vp9/encoder/vp9_variance_c.c
@@ -239,6 +239,32 @@
return (var - (((unsigned int)avg * avg) >> 6));
}
+unsigned int vp9_variance8x4_c(const uint8_t *src_ptr,
+ int source_stride,
+ const uint8_t *ref_ptr,
+ int recon_stride,
+ unsigned int *sse) {+ unsigned int var;
+ int avg;
+
+ variance(src_ptr, source_stride, ref_ptr, recon_stride, 8, 4, &var, &avg);
+ *sse = var;
+ return (var - (((unsigned int)avg * avg) >> 5));
+}
+
+unsigned int vp9_variance4x8_c(const uint8_t *src_ptr,
+ int source_stride,
+ const uint8_t *ref_ptr,
+ int recon_stride,
+ unsigned int *sse) {+ unsigned int var;
+ int avg;
+
+ variance(src_ptr, source_stride, ref_ptr, recon_stride, 4, 8, &var, &avg);
+ *sse = var;
+ return (var - (((unsigned int)avg * avg) >> 5));
+}
+
unsigned int vp9_variance4x4_c(const uint8_t *src_ptr,
int source_stride,
const uint8_t *ref_ptr,