Merge "Adding search_site_config struct."
diff --git a/vp8/common/arm/neon/buildintrapredictorsmby_neon.asm b/vp8/common/arm/neon/buildintrapredictorsmby_neon.asm
index e3ea91f..a8730aa 100644
--- a/vp8/common/arm/neon/buildintrapredictorsmby_neon.asm
+++ b/vp8/common/arm/neon/buildintrapredictorsmby_neon.asm
@@ -26,6 +26,7 @@
|vp8_build_intra_predictors_mby_neon_func| PROC
push {r4-r8, lr}
+ vpush {d8-d15}
cmp r3, #0
beq case_dc_pred
@@ -37,8 +38,8 @@
beq case_tm_pred
case_dc_pred
- ldr r4, [sp, #24] ; Up
- ldr r5, [sp, #28] ; Left
+ ldr r4, [sp, #88] ; Up
+ ldr r5, [sp, #92] ; Left
; Default the DC average to 128
mov r12, #128
@@ -143,6 +144,7 @@
vst1.u8 {q0}, [r1]!
vst1.u8 {q0}, [r1]!
+ vpop {d8-d15}
pop {r4-r8,pc}
case_v_pred
; Copy down above row
@@ -165,6 +167,7 @@
vst1.u8 {q0}, [r1]!
vst1.u8 {q0}, [r1]!
vst1.u8 {q0}, [r1]!
+ vpop {d8-d15}
pop {r4-r8,pc}
case_h_pred
@@ -224,6 +227,7 @@
vst1.u8 {q2}, [r1]!
vst1.u8 {q3}, [r1]!
+ vpop {d8-d15}
pop {r4-r8,pc}
case_tm_pred
@@ -293,6 +297,7 @@
subs r12, r12, #1
bne case_tm_pred_loop
+ vpop {d8-d15}
pop {r4-r8,pc}
ENDP
@@ -307,6 +312,7 @@
|vp8_build_intra_predictors_mby_s_neon_func| PROC
push {r4-r8, lr}
+ vpush {d8-d15}
mov r1, r0 ; unsigned char *ypred_ptr = x->dst.y_buffer; //x->Predictor;
@@ -320,8 +326,8 @@
beq case_tm_pred_s
case_dc_pred_s
- ldr r4, [sp, #24] ; Up
- ldr r5, [sp, #28] ; Left
+ ldr r4, [sp, #88] ; Up
+ ldr r5, [sp, #92] ; Left
; Default the DC average to 128
mov r12, #128
@@ -426,6 +432,7 @@
vst1.u8 {q0}, [r1], r2
vst1.u8 {q0}, [r1], r2
+ vpop {d8-d15}
pop {r4-r8,pc}
case_v_pred_s
; Copy down above row
@@ -448,6 +455,8 @@
vst1.u8 {q0}, [r1], r2
vst1.u8 {q0}, [r1], r2
vst1.u8 {q0}, [r1], r2
+
+ vpop {d8-d15}
pop {r4-r8,pc}
case_h_pred_s
@@ -507,6 +516,7 @@
vst1.u8 {q2}, [r1], r2
vst1.u8 {q3}, [r1], r2
+ vpop {d8-d15}
pop {r4-r8,pc}
case_tm_pred_s
@@ -576,6 +586,7 @@
subs r12, r12, #1
bne case_tm_pred_loop_s
+ vpop {d8-d15}
pop {r4-r8,pc}
ENDP
diff --git a/vp8/common/arm/neon/idct_dequant_0_2x_neon.asm b/vp8/common/arm/neon/idct_dequant_0_2x_neon.asm
index 6c29c55..3a39210 100644
--- a/vp8/common/arm/neon/idct_dequant_0_2x_neon.asm
+++ b/vp8/common/arm/neon/idct_dequant_0_2x_neon.asm
@@ -22,6 +22,7 @@
; r3 stride
|idct_dequant_0_2x_neon| PROC
push {r4, r5}
+ vpush {d8-d15}
add r12, r2, #4
vld1.32 {d2[0]}, [r2], r3
@@ -72,6 +73,7 @@
vst1.32 {d4[1]}, [r2]
vst1.32 {d10[1]}, [r0]
+ vpop {d8-d15}
pop {r4, r5}
bx lr
diff --git a/vp8/common/arm/neon/idct_dequant_full_2x_neon.asm b/vp8/common/arm/neon/idct_dequant_full_2x_neon.asm
index d5dce63..8da0fa0 100644
--- a/vp8/common/arm/neon/idct_dequant_full_2x_neon.asm
+++ b/vp8/common/arm/neon/idct_dequant_full_2x_neon.asm
@@ -22,6 +22,8 @@
; r2 *dst
; r3 stride
|idct_dequant_full_2x_neon| PROC
+ vpush {d8-d15}
+
vld1.16 {q0, q1}, [r1] ; dq (same l/r)
vld1.16 {q2, q3}, [r0] ; l q
add r0, r0, #32
@@ -184,6 +186,7 @@
vst1.32 {d3[0]}, [r2]
vst1.32 {d3[1]}, [r1]
+ vpop {d8-d15}
bx lr
ENDP ; |idct_dequant_full_2x_neon|
diff --git a/vp8/common/arm/neon/loopfilter_neon.asm b/vp8/common/arm/neon/loopfilter_neon.asm
index e44be0a..c4f09c7 100644
--- a/vp8/common/arm/neon/loopfilter_neon.asm
+++ b/vp8/common/arm/neon/loopfilter_neon.asm
@@ -24,10 +24,12 @@
; sp unsigned char thresh,
|vp8_loop_filter_horizontal_edge_y_neon| PROC
push {lr}
+ vpush {d8-d15}
+
vdup.u8 q0, r2 ; duplicate blimit
vdup.u8 q1, r3 ; duplicate limit
sub r2, r0, r1, lsl #2 ; move src pointer down by 4 lines
- ldr r3, [sp, #4] ; load thresh
+ ldr r3, [sp, #68] ; load thresh
add r12, r2, r1
add r1, r1, r1
@@ -52,6 +54,7 @@
vst1.u8 {q7}, [r2@128], r1 ; store oq0
vst1.u8 {q8}, [r12@128], r1 ; store oq1
+ vpop {d8-d15}
pop {pc}
ENDP ; |vp8_loop_filter_horizontal_edge_y_neon|
@@ -64,10 +67,12 @@
; sp+4 unsigned char *v
|vp8_loop_filter_horizontal_edge_uv_neon| PROC
push {lr}
+ vpush {d8-d15}
+
vdup.u8 q0, r2 ; duplicate blimit
vdup.u8 q1, r3 ; duplicate limit
- ldr r12, [sp, #4] ; load thresh
- ldr r2, [sp, #8] ; load v ptr
+ ldr r12, [sp, #68] ; load thresh
+ ldr r2, [sp, #72] ; load v ptr
vdup.u8 q2, r12 ; duplicate thresh
sub r3, r0, r1, lsl #2 ; move u pointer down by 4 lines
@@ -104,6 +109,7 @@
vst1.u8 {d16}, [r0@64] ; store u oq1
vst1.u8 {d17}, [r2@64] ; store v oq1
+ vpop {d8-d15}
pop {pc}
ENDP ; |vp8_loop_filter_horizontal_edge_uv_neon|
@@ -120,11 +126,13 @@
|vp8_loop_filter_vertical_edge_y_neon| PROC
push {lr}
+ vpush {d8-d15}
+
vdup.u8 q0, r2 ; duplicate blimit
vdup.u8 q1, r3 ; duplicate limit
sub r2, r0, #4 ; src ptr down by 4 columns
add r1, r1, r1
- ldr r3, [sp, #4] ; load thresh
+ ldr r3, [sp, #68] ; load thresh
add r12, r2, r1, asr #1
vld1.u8 {d6}, [r2], r1
@@ -194,6 +202,7 @@
vst4.8 {d14[6], d15[6], d16[6], d17[6]}, [r0]
vst4.8 {d14[7], d15[7], d16[7], d17[7]}, [r12]
+ vpop {d8-d15}
pop {pc}
ENDP ; |vp8_loop_filter_vertical_edge_y_neon|
@@ -210,9 +219,11 @@
; sp+4 unsigned char *v
|vp8_loop_filter_vertical_edge_uv_neon| PROC
push {lr}
+ vpush {d8-d15}
+
vdup.u8 q0, r2 ; duplicate blimit
sub r12, r0, #4 ; move u pointer down by 4 columns
- ldr r2, [sp, #8] ; load v ptr
+ ldr r2, [sp, #72] ; load v ptr
vdup.u8 q1, r3 ; duplicate limit
sub r3, r2, #4 ; move v pointer down by 4 columns
@@ -233,7 +244,7 @@
vld1.u8 {d20}, [r12]
vld1.u8 {d21}, [r3]
- ldr r12, [sp, #4] ; load thresh
+ ldr r12, [sp, #68] ; load thresh
;transpose to 8x16 matrix
vtrn.32 q3, q7
@@ -281,6 +292,7 @@
vst4.8 {d10[7], d11[7], d12[7], d13[7]}, [r0]
vst4.8 {d14[7], d15[7], d16[7], d17[7]}, [r2]
+ vpop {d8-d15}
pop {pc}
ENDP ; |vp8_loop_filter_vertical_edge_uv_neon|
diff --git a/vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.asm b/vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.asm
index adf848b..6eb0651 100644
--- a/vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.asm
+++ b/vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.asm
@@ -9,7 +9,6 @@
;
- ;EXPORT |vp8_loop_filter_simple_horizontal_edge_neon|
EXPORT |vp8_loop_filter_bhs_neon|
EXPORT |vp8_loop_filter_mbhs_neon|
ARM
@@ -22,7 +21,7 @@
; q1 limit, PRESERVE
|vp8_loop_filter_simple_horizontal_edge_neon| PROC
-
+ vpush {d8-d15}
sub r3, r0, r1, lsl #1 ; move src pointer down by 2 lines
vld1.u8 {q7}, [r0@128], r1 ; q0
@@ -82,6 +81,7 @@
vst1.u8 {q6}, [r3@128] ; store op0
vst1.u8 {q7}, [r0@128] ; store oq0
+ vpop {d8-d15}
bx lr
ENDP ; |vp8_loop_filter_simple_horizontal_edge_neon|
diff --git a/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.asm b/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.asm
index e690df2..78d13c8 100644
--- a/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.asm
+++ b/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.asm
@@ -9,7 +9,6 @@
;
- ;EXPORT |vp8_loop_filter_simple_vertical_edge_neon|
EXPORT |vp8_loop_filter_bvs_neon|
EXPORT |vp8_loop_filter_mbvs_neon|
ARM
@@ -22,6 +21,8 @@
; q1 limit, PRESERVE
|vp8_loop_filter_simple_vertical_edge_neon| PROC
+ vpush {d8-d15}
+
sub r0, r0, #2 ; move src pointer down by 2 columns
add r12, r1, r1
add r3, r0, r1
@@ -120,6 +121,7 @@
vst2.8 {d14[6], d15[6]}, [r0], r12
vst2.8 {d14[7], d15[7]}, [r3]
+ vpop {d8-d15}
bx lr
ENDP ; |vp8_loop_filter_simple_vertical_edge_neon|
diff --git a/vp8/common/arm/neon/mbloopfilter_neon.asm b/vp8/common/arm/neon/mbloopfilter_neon.asm
index f41c156..d200c30 100644
--- a/vp8/common/arm/neon/mbloopfilter_neon.asm
+++ b/vp8/common/arm/neon/mbloopfilter_neon.asm
@@ -28,8 +28,10 @@
; sp unsigned char thresh,
|vp8_mbloop_filter_horizontal_edge_y_neon| PROC
push {lr}
+ vpush {d8-d15}
+
add r1, r1, r1 ; double stride
- ldr r12, [sp, #4] ; load thresh
+ ldr r12, [sp, #68] ; load thresh
sub r0, r0, r1, lsl #1 ; move src pointer down by 4 lines
vdup.u8 q2, r12 ; thresh
add r12, r0, r1, lsr #1 ; move src pointer up by 1 line
@@ -55,6 +57,7 @@
vst1.u8 {q8}, [r12@128] ; store oq1
vst1.u8 {q9}, [r0@128] ; store oq2
+ vpop {d8-d15}
pop {pc}
ENDP ; |vp8_mbloop_filter_horizontal_edge_y_neon|
@@ -72,10 +75,12 @@
|vp8_mbloop_filter_horizontal_edge_uv_neon| PROC
push {lr}
- ldr r12, [sp, #4] ; load thresh
+ vpush {d8-d15}
+
+ ldr r12, [sp, #68] ; load thresh
sub r0, r0, r1, lsl #2 ; move u pointer down by 4 lines
vdup.u8 q2, r12 ; thresh
- ldr r12, [sp, #8] ; load v ptr
+ ldr r12, [sp, #72] ; load v ptr
sub r12, r12, r1, lsl #2 ; move v pointer down by 4 lines
vld1.u8 {d6}, [r0@64], r1 ; p3
@@ -116,6 +121,7 @@
vst1.u8 {d18}, [r0@64], r1 ; store u oq2
vst1.u8 {d19}, [r12@64], r1 ; store v oq2
+ vpop {d8-d15}
pop {pc}
ENDP ; |vp8_mbloop_filter_horizontal_edge_uv_neon|
@@ -130,7 +136,9 @@
; sp unsigned char thresh,
|vp8_mbloop_filter_vertical_edge_y_neon| PROC
push {lr}
- ldr r12, [sp, #4] ; load thresh
+ vpush {d8-d15}
+
+ ldr r12, [sp, #68] ; load thresh
sub r0, r0, #4 ; move src pointer down by 4 columns
vdup.s8 q2, r12 ; thresh
add r12, r0, r1, lsl #3 ; move src pointer down by 8 lines
@@ -208,6 +216,7 @@
vst1.8 {d20}, [r0]
vst1.8 {d21}, [r12]
+ vpop {d8-d15}
pop {pc}
ENDP ; |vp8_mbloop_filter_vertical_edge_y_neon|
@@ -224,10 +233,12 @@
; sp+4 unsigned char *v
|vp8_mbloop_filter_vertical_edge_uv_neon| PROC
push {lr}
- ldr r12, [sp, #4] ; load thresh
+ vpush {d8-d15}
+
+ ldr r12, [sp, #68] ; load thresh
sub r0, r0, #4 ; move u pointer down by 4 columns
vdup.u8 q2, r12 ; thresh
- ldr r12, [sp, #8] ; load v ptr
+ ldr r12, [sp, #72] ; load v ptr
sub r12, r12, #4 ; move v pointer down by 4 columns
vld1.u8 {d6}, [r0], r1 ;load u data
@@ -303,6 +314,7 @@
vst1.8 {d20}, [r0]
vst1.8 {d21}, [r12]
+ vpop {d8-d15}
pop {pc}
ENDP ; |vp8_mbloop_filter_vertical_edge_uv_neon|
diff --git a/vp8/common/arm/neon/sad16_neon.asm b/vp8/common/arm/neon/sad16_neon.asm
index d7c590e..7197e56 100644
--- a/vp8/common/arm/neon/sad16_neon.asm
+++ b/vp8/common/arm/neon/sad16_neon.asm
@@ -24,6 +24,7 @@
; r3 int ref_stride
|vp8_sad16x16_neon| PROC
;;
+ vpush {d8-d15}
vld1.8 {q0}, [r0], r1
vld1.8 {q4}, [r2], r3
@@ -132,6 +133,7 @@
vmov.32 r0, d0[0]
+ vpop {d8-d15}
bx lr
ENDP
@@ -143,6 +145,8 @@
; unsigned char *ref_ptr,
; int ref_stride)
|vp8_sad16x8_neon| PROC
+ vpush {d8-d15}
+
vld1.8 {q0}, [r0], r1
vld1.8 {q4}, [r2], r3
@@ -200,6 +204,7 @@
vmov.32 r0, d0[0]
+ vpop {d8-d15}
bx lr
ENDP
diff --git a/vp8/common/arm/neon/sad8_neon.asm b/vp8/common/arm/neon/sad8_neon.asm
index 23ba6df..6b849d9 100644
--- a/vp8/common/arm/neon/sad8_neon.asm
+++ b/vp8/common/arm/neon/sad8_neon.asm
@@ -25,6 +25,7 @@
; int ref_stride)
|vp8_sad8x8_neon| PROC
+ vpush {d8-d15}
vld1.8 {d0}, [r0], r1
vld1.8 {d8}, [r2], r3
@@ -70,6 +71,7 @@
vmov.32 r0, d0[0]
+ vpop {d8-d15}
bx lr
ENDP
@@ -82,6 +84,7 @@
; int ref_stride)
|vp8_sad8x16_neon| PROC
+ vpush {d8-d15}
vld1.8 {d0}, [r0], r1
vld1.8 {d8}, [r2], r3
@@ -167,6 +170,7 @@
vmov.32 r0, d0[0]
+ vpop {d8-d15}
bx lr
ENDP
@@ -179,6 +183,7 @@
; int ref_stride)
|vp8_sad4x4_neon| PROC
+ vpush {d8-d15}
vld1.8 {d0}, [r0], r1
vld1.8 {d8}, [r2], r3
@@ -202,6 +207,7 @@
vpaddl.u32 d0, d1
vmov.32 r0, d0[0]
+ vpop {d8-d15}
bx lr
ENDP
diff --git a/vp8/common/arm/neon/shortidct4x4llm_neon.asm b/vp8/common/arm/neon/shortidct4x4llm_neon.asm
index 67d2ab0..87ca887 100644
--- a/vp8/common/arm/neon/shortidct4x4llm_neon.asm
+++ b/vp8/common/arm/neon/shortidct4x4llm_neon.asm
@@ -37,12 +37,14 @@
; result of the multiplication that is needed in IDCT.
|vp8_short_idct4x4llm_neon| PROC
+ vpush {d8-d15}
+
adr r12, idct_coeff
vld1.16 {q1, q2}, [r0]
vld1.16 {d0}, [r12]
vswp d3, d4 ;q2(vp[4] vp[12])
- ldr r0, [sp] ; stride
+ ldr r0, [sp, #64] ; stride
vqdmulh.s16 q3, q2, d0[2]
vqdmulh.s16 q4, q2, d0[0]
@@ -125,6 +127,7 @@
vst1.32 d2[0], [r3], r0
vst1.32 d2[1], [r3], r0
+ vpop {d8-d15}
bx lr
ENDP
diff --git a/vp8/common/arm/neon/sixtappredict16x16_neon.asm b/vp8/common/arm/neon/sixtappredict16x16_neon.asm
index 9fdafd3..dd27719 100644
--- a/vp8/common/arm/neon/sixtappredict16x16_neon.asm
+++ b/vp8/common/arm/neon/sixtappredict16x16_neon.asm
@@ -43,10 +43,11 @@
|vp8_sixtap_predict16x16_neon| PROC
push {r4-r5, lr}
+ vpush {d8-d15}
adr r12, filter16_coeff
- ldr r4, [sp, #12] ;load parameters from stack
- ldr r5, [sp, #16] ;load parameters from stack
+ ldr r4, [sp, #76] ;load parameters from stack
+ ldr r5, [sp, #80] ;load parameters from stack
cmp r2, #0 ;skip first_pass filter if xoffset=0
beq secondpass_filter16x16_only
@@ -291,6 +292,8 @@
bne filt_blk2d_sp16x16_outloop_neon
add sp, sp, #336
+
+ vpop {d8-d15}
pop {r4-r5,pc}
;--------------------
@@ -384,6 +387,7 @@
bne filt_blk2d_fpo16x16_loop_neon
+ vpop {d8-d15}
pop {r4-r5,pc}
;--------------------
@@ -482,6 +486,7 @@
bne filt_blk2d_spo16x16_outloop_neon
+ vpop {d8-d15}
pop {r4-r5,pc}
ENDP
diff --git a/vp8/common/arm/neon/sixtappredict4x4_neon.asm b/vp8/common/arm/neon/sixtappredict4x4_neon.asm
index a4222bc..e32e713 100644
--- a/vp8/common/arm/neon/sixtappredict4x4_neon.asm
+++ b/vp8/common/arm/neon/sixtappredict4x4_neon.asm
@@ -35,10 +35,11 @@
|vp8_sixtap_predict4x4_neon| PROC
push {r4, lr}
+ vpush {d8-d15}
adr r12, filter4_coeff
- ldr r4, [sp, #8] ;load parameters from stack
- ldr lr, [sp, #12] ;load parameters from stack
+ ldr r4, [sp, #72] ;load parameters from stack
+ ldr lr, [sp, #76] ;load parameters from stack
cmp r2, #0 ;skip first_pass filter if xoffset=0
beq secondpass_filter4x4_only
@@ -261,6 +262,7 @@
vst1.32 {d4[0]}, [r1]
vst1.32 {d4[1]}, [r2]
+ vpop {d8-d15}
pop {r4, pc}
@@ -348,6 +350,7 @@
vst1.32 {d28[0]}, [r1]
vst1.32 {d28[1]}, [r2]
+ vpop {d8-d15}
pop {r4, pc}
@@ -413,6 +416,7 @@
vst1.32 {d4[0]}, [r1]
vst1.32 {d4[1]}, [r2]
+ vpop {d8-d15}
pop {r4, pc}
ENDP
diff --git a/vp8/common/arm/neon/sixtappredict8x4_neon.asm b/vp8/common/arm/neon/sixtappredict8x4_neon.asm
index a57ec01..d19bf89 100644
--- a/vp8/common/arm/neon/sixtappredict8x4_neon.asm
+++ b/vp8/common/arm/neon/sixtappredict8x4_neon.asm
@@ -35,10 +35,11 @@
|vp8_sixtap_predict8x4_neon| PROC
push {r4-r5, lr}
+ vpush {d8-d15}
adr r12, filter8_coeff
- ldr r4, [sp, #12] ;load parameters from stack
- ldr r5, [sp, #16] ;load parameters from stack
+ ldr r4, [sp, #76] ;load parameters from stack
+ ldr r5, [sp, #80] ;load parameters from stack
cmp r2, #0 ;skip first_pass filter if xoffset=0
beq secondpass_filter8x4_only
@@ -297,6 +298,8 @@
vst1.u8 {d9}, [r4], r5
add sp, sp, #32
+
+ vpop {d8-d15}
pop {r4-r5,pc}
;--------------------
@@ -392,6 +395,7 @@
vst1.u8 {d24}, [r4], r5
vst1.u8 {d25}, [r4], r5
+ vpop {d8-d15}
pop {r4-r5,pc}
;---------------------
@@ -464,6 +468,7 @@
vst1.u8 {d8}, [r4], r5
vst1.u8 {d9}, [r4], r5
+ vpop {d8-d15}
pop {r4-r5,pc}
ENDP
diff --git a/vp8/common/arm/neon/sixtappredict8x8_neon.asm b/vp8/common/arm/neon/sixtappredict8x8_neon.asm
index 00ed5ae..4b04925 100644
--- a/vp8/common/arm/neon/sixtappredict8x8_neon.asm
+++ b/vp8/common/arm/neon/sixtappredict8x8_neon.asm
@@ -35,11 +35,11 @@
|vp8_sixtap_predict8x8_neon| PROC
push {r4-r5, lr}
-
+ vpush {d8-d15}
adr r12, filter8_coeff
- ldr r4, [sp, #12] ;load parameters from stack
- ldr r5, [sp, #16] ;load parameters from stack
+ ldr r4, [sp, #76] ;load parameters from stack
+ ldr r5, [sp, #80] ;load parameters from stack
cmp r2, #0 ;skip first_pass filter if xoffset=0
beq secondpass_filter8x8_only
@@ -324,6 +324,8 @@
bne filt_blk2d_sp8x8_loop_neon
add sp, sp, #64
+
+ vpop {d8-d15}
pop {r4-r5,pc}
;---------------------
@@ -428,6 +430,7 @@
bne filt_blk2d_fpo8x8_loop_neon
+ vpop {d8-d15}
pop {r4-r5,pc}
;---------------------
@@ -515,6 +518,7 @@
bne filt_blk2d_spo8x8_loop_neon
+ vpop {d8-d15}
pop {r4-r5,pc}
ENDP
diff --git a/vp8/common/arm/neon/variance_neon.asm b/vp8/common/arm/neon/variance_neon.asm
index e3b4832..8ecad72 100644
--- a/vp8/common/arm/neon/variance_neon.asm
+++ b/vp8/common/arm/neon/variance_neon.asm
@@ -26,6 +26,7 @@
; r3 int recon_stride
; stack unsigned int *sse
|vp8_variance16x16_neon| PROC
+ vpush {q5}
vmov.i8 q8, #0 ;q8 - sum
vmov.i8 q9, #0 ;q9, q10 - sse
vmov.i8 q10, #0
@@ -67,7 +68,7 @@
vadd.u32 q10, q9, q10 ;accumulate sse
vpaddl.s32 q0, q8 ;accumulate sum
- ldr r12, [sp] ;load *sse from stack
+ ldr r12, [sp, #16] ;load *sse from stack
vpaddl.u32 q1, q10
vadd.s64 d0, d0, d1
@@ -87,6 +88,8 @@
vsub.u32 d0, d1, d10
vmov.32 r0, d0[0] ;return
+
+ vpop {q5}
bx lr
ENDP
@@ -99,6 +102,8 @@
; int recon_stride,
; unsigned int *sse)
|vp8_variance16x8_neon| PROC
+ vpush {q5}
+
vmov.i8 q8, #0 ;q8 - sum
vmov.i8 q9, #0 ;q9, q10 - sse
vmov.i8 q10, #0
@@ -137,7 +142,7 @@
vadd.u32 q10, q9, q10 ;accumulate sse
vpaddl.s32 q0, q8 ;accumulate sum
- ldr r12, [sp] ;load *sse from stack
+ ldr r12, [sp, #16] ;load *sse from stack
vpaddl.u32 q1, q10
vadd.s64 d0, d0, d1
@@ -149,6 +154,8 @@
vsub.u32 d0, d1, d10
vmov.32 r0, d0[0] ;return
+
+ vpop {q5}
bx lr
ENDP
@@ -162,6 +169,8 @@
; unsigned int *sse)
|vp8_variance8x16_neon| PROC
+ vpush {q5}
+
vmov.i8 q8, #0 ;q8 - sum
vmov.i8 q9, #0 ;q9, q10 - sse
vmov.i8 q10, #0
@@ -192,7 +201,7 @@
vadd.u32 q10, q9, q10 ;accumulate sse
vpaddl.s32 q0, q8 ;accumulate sum
- ldr r12, [sp] ;load *sse from stack
+ ldr r12, [sp, #16] ;load *sse from stack
vpaddl.u32 q1, q10
vadd.s64 d0, d0, d1
@@ -204,6 +213,8 @@
vsub.u32 d0, d1, d10
vmov.32 r0, d0[0] ;return
+
+ vpop {q5}
bx lr
ENDP
@@ -215,6 +226,8 @@
; r3 int recon_stride
; stack unsigned int *sse
|vp8_variance8x8_neon| PROC
+ vpush {q5}
+
vmov.i8 q8, #0 ;q8 - sum
vmov.i8 q9, #0 ;q9, q10 - sse
vmov.i8 q10, #0
@@ -257,7 +270,7 @@
vadd.u32 q10, q9, q10 ;accumulate sse
vpaddl.s32 q0, q8 ;accumulate sum
- ldr r12, [sp] ;load *sse from stack
+ ldr r12, [sp, #16] ;load *sse from stack
vpaddl.u32 q1, q10
vadd.s64 d0, d0, d1
@@ -269,6 +282,8 @@
vsub.u32 d0, d1, d10
vmov.32 r0, d0[0] ;return
+
+ vpop {q5}
bx lr
ENDP
diff --git a/vp8/common/arm/neon/vp8_subpixelvariance16x16_neon.asm b/vp8/common/arm/neon/vp8_subpixelvariance16x16_neon.asm
index 9d22c52..adc5b7e 100644
--- a/vp8/common/arm/neon/vp8_subpixelvariance16x16_neon.asm
+++ b/vp8/common/arm/neon/vp8_subpixelvariance16x16_neon.asm
@@ -31,11 +31,12 @@
|vp8_sub_pixel_variance16x16_neon_func| PROC
push {r4-r6, lr}
+ vpush {d8-d15}
adr r12, bilinear_taps_coeff
- ldr r4, [sp, #16] ;load *dst_ptr from stack
- ldr r5, [sp, #20] ;load dst_pixels_per_line from stack
- ldr r6, [sp, #24] ;load *sse from stack
+ ldr r4, [sp, #80] ;load *dst_ptr from stack
+ ldr r5, [sp, #84] ;load dst_pixels_per_line from stack
+ ldr r6, [sp, #88] ;load *sse from stack
cmp r2, #0 ;skip first_pass filter if xoffset=0
beq secondpass_bfilter16x16_only
@@ -416,6 +417,7 @@
add sp, sp, #528
vmov.32 r0, d0[0] ;return
+ vpop {d8-d15}
pop {r4-r6,pc}
ENDP
diff --git a/vp8/common/arm/neon/vp8_subpixelvariance16x16s_neon.asm b/vp8/common/arm/neon/vp8_subpixelvariance16x16s_neon.asm
index 155be4f..b0829af 100644
--- a/vp8/common/arm/neon/vp8_subpixelvariance16x16s_neon.asm
+++ b/vp8/common/arm/neon/vp8_subpixelvariance16x16s_neon.asm
@@ -31,9 +31,10 @@
;================================================
|vp8_variance_halfpixvar16x16_h_neon| PROC
push {lr}
+ vpush {d8-d15}
mov r12, #4 ;loop counter
- ldr lr, [sp, #4] ;load *sse from stack
+ ldr lr, [sp, #68] ;load *sse from stack
vmov.i8 q8, #0 ;q8 - sum
vmov.i8 q9, #0 ;q9, q10 - sse
vmov.i8 q10, #0
@@ -116,6 +117,8 @@
vsub.u32 d0, d1, d10
vmov.32 r0, d0[0] ;return
+
+ vpop {d8-d15}
pop {pc}
ENDP
@@ -131,11 +134,12 @@
;================================================
|vp8_variance_halfpixvar16x16_v_neon| PROC
push {lr}
+ vpush {d8-d15}
mov r12, #4 ;loop counter
vld1.u8 {q0}, [r0], r1 ;load src data
- ldr lr, [sp, #4] ;load *sse from stack
+ ldr lr, [sp, #68] ;load *sse from stack
vmov.i8 q8, #0 ;q8 - sum
vmov.i8 q9, #0 ;q9, q10 - sse
@@ -212,6 +216,8 @@
vsub.u32 d0, d1, d10
vmov.32 r0, d0[0] ;return
+
+ vpop {d8-d15}
pop {pc}
ENDP
@@ -227,10 +233,11 @@
;================================================
|vp8_variance_halfpixvar16x16_hv_neon| PROC
push {lr}
+ vpush {d8-d15}
vld1.u8 {d0, d1, d2, d3}, [r0], r1 ;load src data
- ldr lr, [sp, #4] ;load *sse from stack
+ ldr lr, [sp, #68] ;load *sse from stack
vmov.i8 q13, #0 ;q8 - sum
vext.8 q1, q0, q1, #1 ;construct src_ptr[1]
@@ -331,6 +338,8 @@
vsub.u32 d0, d1, d10
vmov.32 r0, d0[0] ;return
+
+ vpop {d8-d15}
pop {pc}
ENDP
@@ -349,10 +358,11 @@
|vp8_sub_pixel_variance16x16s_neon| PROC
push {r4, lr}
+ vpush {d8-d15}
- ldr r4, [sp, #8] ;load *dst_ptr from stack
- ldr r12, [sp, #12] ;load dst_pixels_per_line from stack
- ldr lr, [sp, #16] ;load *sse from stack
+ ldr r4, [sp, #72] ;load *dst_ptr from stack
+ ldr r12, [sp, #76] ;load dst_pixels_per_line from stack
+ ldr lr, [sp, #80] ;load *sse from stack
cmp r2, #0 ;skip first_pass filter if xoffset=0
beq secondpass_bfilter16x16s_only
@@ -566,6 +576,7 @@
add sp, sp, #256
vmov.32 r0, d0[0] ;return
+ vpop {d8-d15}
pop {r4, pc}
ENDP
diff --git a/vp8/common/arm/neon/vp8_subpixelvariance8x8_neon.asm b/vp8/common/arm/neon/vp8_subpixelvariance8x8_neon.asm
index f6b6847..9d9f9e0 100644
--- a/vp8/common/arm/neon/vp8_subpixelvariance8x8_neon.asm
+++ b/vp8/common/arm/neon/vp8_subpixelvariance8x8_neon.asm
@@ -26,11 +26,12 @@
|vp8_sub_pixel_variance8x8_neon| PROC
push {r4-r5, lr}
+ vpush {d8-d15}
adr r12, bilinear_taps_coeff
- ldr r4, [sp, #12] ;load *dst_ptr from stack
- ldr r5, [sp, #16] ;load dst_pixels_per_line from stack
- ldr lr, [sp, #20] ;load *sse from stack
+ ldr r4, [sp, #76] ;load *dst_ptr from stack
+ ldr r5, [sp, #80] ;load dst_pixels_per_line from stack
+ ldr lr, [sp, #84] ;load *sse from stack
cmp r2, #0 ;skip first_pass filter if xoffset=0
beq skip_firstpass_filter
@@ -210,6 +211,8 @@
vsub.u32 d0, d1, d10
vmov.32 r0, d0[0] ;return
+
+ vpop {d8-d15}
pop {r4-r5, pc}
ENDP
diff --git a/vp8/encoder/arm/neon/subtract_neon.asm b/vp8/encoder/arm/neon/subtract_neon.asm
index 5bda786..840cb33 100644
--- a/vp8/encoder/arm/neon/subtract_neon.asm
+++ b/vp8/encoder/arm/neon/subtract_neon.asm
@@ -65,8 +65,10 @@
; unsigned char *pred, int pred_stride)
|vp8_subtract_mby_neon| PROC
push {r4-r7}
+ vpush {d8-d15}
+
mov r12, #4
- ldr r4, [sp, #16] ; pred_stride
+ ldr r4, [sp, #80] ; pred_stride
mov r6, #32 ; "diff" stride x2
add r5, r0, #16 ; second diff pointer
@@ -101,6 +103,7 @@
subs r12, r12, #1
bne subtract_mby_loop
+ vpop {d8-d15}
pop {r4-r7}
bx lr
ENDP
@@ -112,9 +115,11 @@
|vp8_subtract_mbuv_neon| PROC
push {r4-r7}
- ldr r4, [sp, #16] ; upred
- ldr r5, [sp, #20] ; vpred
- ldr r6, [sp, #24] ; pred_stride
+ vpush {d8-d15}
+
+ ldr r4, [sp, #80] ; upred
+ ldr r5, [sp, #84] ; vpred
+ ldr r6, [sp, #88] ; pred_stride
add r0, r0, #512 ; short *udiff = diff + 256;
mov r12, #32 ; "diff" stride x2
add r7, r0, #16 ; second diff pointer
@@ -191,6 +196,7 @@
vst1.16 {q14}, [r0], r12
vst1.16 {q15}, [r7], r12
+ vpop {d8-d15}
pop {r4-r7}
bx lr
diff --git a/vp8/encoder/arm/neon/vp8_memcpy_neon.asm b/vp8/encoder/arm/neon/vp8_memcpy_neon.asm
index 5b9f11e..d219e2d 100644
--- a/vp8/encoder/arm/neon/vp8_memcpy_neon.asm
+++ b/vp8/encoder/arm/neon/vp8_memcpy_neon.asm
@@ -21,6 +21,7 @@
;void vp8_memcpy_partial_neon(unsigned char *dst_ptr, unsigned char *src_ptr,
; int sz);
|vp8_memcpy_partial_neon| PROC
+ vpush {d8-d15}
;pld [r1] ;preload pred data
;pld [r1, #128]
;pld [r1, #256]
@@ -64,6 +65,7 @@
bne extra_copy_neon_loop
done_copy_neon_loop
+ vpop {d8-d15}
bx lr
ENDP
diff --git a/vp8/encoder/arm/neon/vp8_mse16x16_neon.asm b/vp8/encoder/arm/neon/vp8_mse16x16_neon.asm
index 55edbf5..f82af3e 100644
--- a/vp8/encoder/arm/neon/vp8_mse16x16_neon.asm
+++ b/vp8/encoder/arm/neon/vp8_mse16x16_neon.asm
@@ -27,6 +27,8 @@
;from vp8_variance().
|vp8_mse16x16_neon| PROC
+ vpush {q7}
+
vmov.i8 q7, #0 ;q7, q8, q9, q10 - sse
vmov.i8 q8, #0
vmov.i8 q9, #0
@@ -62,7 +64,7 @@
vadd.u32 q7, q7, q8
vadd.u32 q9, q9, q10
- ldr r12, [sp] ;load *sse from stack
+ ldr r12, [sp, #16] ;load *sse from stack
vadd.u32 q10, q7, q9
vpaddl.u32 q1, q10
@@ -71,6 +73,7 @@
vst1.32 {d0[0]}, [r12]
vmov.32 r0, d0[0]
+ vpop {q7}
bx lr
ENDP
@@ -82,6 +85,8 @@
; r2 unsigned char *ref_ptr,
; r3 int recon_stride
|vp8_get4x4sse_cs_neon| PROC
+ vpush {q7}
+
vld1.8 {d0}, [r0], r1 ;Load up source and reference
vld1.8 {d4}, [r2], r3
vld1.8 {d1}, [r0], r1
@@ -109,6 +114,8 @@
vadd.u64 d0, d2, d3
vmov.32 r0, d0[0]
+
+ vpop {q7}
bx lr
ENDP
diff --git a/vp9/common/vp9_rtcd_defs.pl b/vp9/common/vp9_rtcd_defs.pl
index a0666f4..99fd6ca 100644
--- a/vp9/common/vp9_rtcd_defs.pl
+++ b/vp9/common/vp9_rtcd_defs.pl
@@ -563,33 +563,6 @@
add_proto qw/unsigned int vp9_sad4x4_avg/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, const uint8_t *second_pred, unsigned int max_sad";
specialize qw/vp9_sad4x4_avg/, "$sse_x86inc";
-add_proto qw/unsigned int vp9_variance_halfpixvar16x16_h/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
-specialize qw/vp9_variance_halfpixvar16x16_h/, "$sse2_x86inc";
-
-add_proto qw/unsigned int vp9_variance_halfpixvar16x16_v/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
-specialize qw/vp9_variance_halfpixvar16x16_v/, "$sse2_x86inc";
-
-add_proto qw/unsigned int vp9_variance_halfpixvar16x16_hv/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
-specialize qw/vp9_variance_halfpixvar16x16_hv/, "$sse2_x86inc";
-
-add_proto qw/unsigned int vp9_variance_halfpixvar64x64_h/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
-specialize qw/vp9_variance_halfpixvar64x64_h/;
-
-add_proto qw/unsigned int vp9_variance_halfpixvar64x64_v/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
-specialize qw/vp9_variance_halfpixvar64x64_v/;
-
-add_proto qw/unsigned int vp9_variance_halfpixvar64x64_hv/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
-specialize qw/vp9_variance_halfpixvar64x64_hv/;
-
-add_proto qw/unsigned int vp9_variance_halfpixvar32x32_h/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
-specialize qw/vp9_variance_halfpixvar32x32_h/;
-
-add_proto qw/unsigned int vp9_variance_halfpixvar32x32_v/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
-specialize qw/vp9_variance_halfpixvar32x32_v/;
-
-add_proto qw/unsigned int vp9_variance_halfpixvar32x32_hv/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sse";
-specialize qw/vp9_variance_halfpixvar32x32_hv/;
-
add_proto qw/void vp9_sad64x64x3/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int ref_stride, unsigned int *sad_array";
specialize qw/vp9_sad64x64x3/;
diff --git a/vp9/common/vp9_tapify.py b/vp9/common/vp9_tapify.py
deleted file mode 100644
index 99529cf..0000000
--- a/vp9/common/vp9_tapify.py
+++ /dev/null
@@ -1,106 +0,0 @@
-"""
- * Copyright (c) 2012 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.
-"""
-#!/usr/bin/env python
-import sys,string,os,re,math,numpy
-scale = 2**16
-def dist(p1,p2):
- x1,y1 = p1
- x2,y2 = p2
- if x1==x2 and y1==y2 :
- return 1.0
- return 1/ math.sqrt((x1-x2)*(x1-x2)+(y1-y2)*(y1-y2))
-
-def gettaps(p):
- def l(b):
- return int(math.floor(b))
- def h(b):
- return int(math.ceil(b))
- def t(b,p,s):
- return int((scale*dist(b,p)+s/2)/s)
- r,c = p
- ul=[l(r),l(c)]
- ur=[l(r),h(c)]
- ll=[h(r),l(c)]
- lr=[h(r),h(c)]
- sum = dist(ul,p)+dist(ur,p)+dist(ll,p)+dist(lr,p)
- t4 = scale - t(ul,p,sum) - t(ur,p,sum) - t(ll,p,sum);
- return [[ul,t(ul,p,sum)],[ur,t(ur,p,sum)],
- [ll,t(ll,p,sum)],[lr,t4]]
-
-def print_mb_taps(angle,blocksize):
- theta = angle / 57.2957795;
- affine = [[math.cos(theta),-math.sin(theta)],
- [math.sin(theta),math.cos(theta)]]
- radius = (float(blocksize)-1)/2
- print " // angle of",angle,"degrees"
- for y in range(blocksize) :
- for x in range(blocksize) :
- r,c = numpy.dot(affine,[y-radius, x-radius])
- tps = gettaps([r+radius,c+radius])
- for t in tps :
- p,t = t
- tr,tc = p
- print " %2d, %2d, %5d, " % (tr,tc,t,),
- print " // %2d,%2d " % (y,x)
-
-i=float(sys.argv[1])
-while i <= float(sys.argv[2]) :
- print_mb_taps(i,float(sys.argv[4]))
- i=i+float(sys.argv[3])
-"""
-
-taps = []
-pt=dict()
-ptr=dict()
-for y in range(16) :
- for x in range(16) :
- r,c = numpy.dot(affine,[y-7.5, x-7.5])
- tps = gettaps([r+7.5,c+7.5])
- j=0
- for tp in tps :
- p,i = tp
- r,c = p
- pt[y,x,j]= [p,i]
- try:
- ptr[r,j,c].append([y,x])
- except:
- ptr[r,j,c]=[[y,x]]
- j = j+1
-
-for key in sorted(pt.keys()) :
- print key,pt[key]
-
-lr = -99
-lj = -99
-lc = 0
-
-shuf=""
-mask=""
-for r,j,c in sorted(ptr.keys()) :
- for y,x in ptr[r,j,c] :
- if lr != r or lj != j :
- print "shuf_"+str(lr)+"_"+str(lj)+"_"+shuf.ljust(16,"0"), lc
- shuf=""
- lc = 0
- for i in range(lc,c-1) :
- shuf = shuf +"0"
- shuf = shuf + hex(x)[2]
- lc =c
- break
- lr = r
- lj = j
-# print r,j,c,ptr[r,j,c]
-# print
-
-for r,j,c in sorted(ptr.keys()) :
- for y,x in ptr[r,j,c] :
- print r,j,c,y,x
- break
-"""
diff --git a/vp9/decoder/vp9_decoder.c b/vp9/decoder/vp9_decoder.c
index faf710c..385b2eb 100644
--- a/vp9/decoder/vp9_decoder.c
+++ b/vp9/decoder/vp9_decoder.c
@@ -32,74 +32,6 @@
#include "vp9/decoder/vp9_detokenize.h"
#include "vp9/decoder/vp9_dthread.h"
-#define WRITE_RECON_BUFFER 0
-#if WRITE_RECON_BUFFER == 1
-static void recon_write_yuv_frame(const char *name,
- const YV12_BUFFER_CONFIG *s,
- int w, int _h) {
- FILE *yuv_file = fopen(name, "ab");
- const uint8_t *src = s->y_buffer;
- int h = _h;
-
- do {
- fwrite(src, w, 1, yuv_file);
- src += s->y_stride;
- } while (--h);
-
- src = s->u_buffer;
- h = (_h + 1) >> 1;
- w = (w + 1) >> 1;
-
- do {
- fwrite(src, w, 1, yuv_file);
- src += s->uv_stride;
- } while (--h);
-
- src = s->v_buffer;
- h = (_h + 1) >> 1;
-
- do {
- fwrite(src, w, 1, yuv_file);
- src += s->uv_stride;
- } while (--h);
-
- fclose(yuv_file);
-}
-#endif
-#if WRITE_RECON_BUFFER == 2
-void write_dx_frame_to_file(YV12_BUFFER_CONFIG *frame, int this_frame) {
- // write the frame
- FILE *yframe;
- int i;
- char filename[255];
-
- snprintf(filename, sizeof(filename)-1, "dx\\y%04d.raw", this_frame);
- yframe = fopen(filename, "wb");
-
- for (i = 0; i < frame->y_height; i++)
- fwrite(frame->y_buffer + i * frame->y_stride,
- frame->y_width, 1, yframe);
-
- fclose(yframe);
- snprintf(filename, sizeof(filename)-1, "dx\\u%04d.raw", this_frame);
- yframe = fopen(filename, "wb");
-
- for (i = 0; i < frame->uv_height; i++)
- fwrite(frame->u_buffer + i * frame->uv_stride,
- frame->uv_width, 1, yframe);
-
- fclose(yframe);
- snprintf(filename, sizeof(filename)-1, "dx\\v%04d.raw", this_frame);
- yframe = fopen(filename, "wb");
-
- for (i = 0; i < frame->uv_height; i++)
- fwrite(frame->v_buffer + i * frame->uv_stride,
- frame->uv_width, 1, yframe);
-
- fclose(yframe);
-}
-#endif
-
void vp9_initialize_dec() {
static int init_done = 0;
@@ -348,15 +280,6 @@
swap_frame_buffers(pbi);
-#if WRITE_RECON_BUFFER == 2
- if (cm->show_frame)
- write_dx_frame_to_file(cm->frame_to_show,
- cm->current_video_frame);
- else
- write_dx_frame_to_file(cm->frame_to_show,
- cm->current_video_frame + 1000);
-#endif
-
if (!pbi->do_loopfilter_inline) {
// If multiple threads are used to decode tiles, then we use those threads
// to do parallel loopfiltering.
@@ -367,21 +290,6 @@
}
}
-#if WRITE_RECON_BUFFER == 2
- if (cm->show_frame)
- write_dx_frame_to_file(cm->frame_to_show,
- cm->current_video_frame + 2000);
- else
- write_dx_frame_to_file(cm->frame_to_show,
- cm->current_video_frame + 3000);
-#endif
-
-#if WRITE_RECON_BUFFER == 1
- if (cm->show_frame)
- recon_write_yuv_frame("recon.yuv", cm->frame_to_show,
- cm->width, cm->height);
-#endif
-
vp9_clear_system_state();
cm->last_width = cm->width;
diff --git a/vp9/encoder/vp9_encoder.c b/vp9/encoder/vp9_encoder.c
index f04b515..395d26a 100644
--- a/vp9/encoder/vp9_encoder.c
+++ b/vp9/encoder/vp9_encoder.c
@@ -970,95 +970,73 @@
cpi->rd.thresh_freq_fact[i][j] = 32;
}
-#define BFP(BT, SDF, SDAF, VF, SVF, SVAF, SVFHH, SVFHV, SVFHHV, \
- SDX3F, SDX8F, SDX4DF)\
+#define BFP(BT, SDF, SDAF, VF, SVF, SVAF, SDX3F, SDX8F, SDX4DF)\
cpi->fn_ptr[BT].sdf = SDF; \
cpi->fn_ptr[BT].sdaf = SDAF; \
cpi->fn_ptr[BT].vf = VF; \
cpi->fn_ptr[BT].svf = SVF; \
cpi->fn_ptr[BT].svaf = SVAF; \
- cpi->fn_ptr[BT].svf_halfpix_h = SVFHH; \
- cpi->fn_ptr[BT].svf_halfpix_v = SVFHV; \
- cpi->fn_ptr[BT].svf_halfpix_hv = SVFHHV; \
cpi->fn_ptr[BT].sdx3f = SDX3F; \
cpi->fn_ptr[BT].sdx8f = SDX8F; \
cpi->fn_ptr[BT].sdx4df = SDX4DF;
BFP(BLOCK_32X16, vp9_sad32x16, vp9_sad32x16_avg,
vp9_variance32x16, vp9_sub_pixel_variance32x16,
- vp9_sub_pixel_avg_variance32x16, NULL, NULL,
- NULL, NULL, NULL,
- vp9_sad32x16x4d)
+ vp9_sub_pixel_avg_variance32x16, NULL, NULL, vp9_sad32x16x4d)
BFP(BLOCK_16X32, vp9_sad16x32, vp9_sad16x32_avg,
vp9_variance16x32, vp9_sub_pixel_variance16x32,
- vp9_sub_pixel_avg_variance16x32, NULL, NULL,
- NULL, NULL, NULL,
- vp9_sad16x32x4d)
+ vp9_sub_pixel_avg_variance16x32, NULL, NULL, vp9_sad16x32x4d)
BFP(BLOCK_64X32, vp9_sad64x32, vp9_sad64x32_avg,
vp9_variance64x32, vp9_sub_pixel_variance64x32,
- vp9_sub_pixel_avg_variance64x32, NULL, NULL,
- NULL, NULL, NULL,
- vp9_sad64x32x4d)
+ vp9_sub_pixel_avg_variance64x32, NULL, NULL, vp9_sad64x32x4d)
BFP(BLOCK_32X64, vp9_sad32x64, vp9_sad32x64_avg,
vp9_variance32x64, vp9_sub_pixel_variance32x64,
- vp9_sub_pixel_avg_variance32x64, NULL, NULL,
- NULL, NULL, NULL,
- vp9_sad32x64x4d)
+ vp9_sub_pixel_avg_variance32x64, NULL, NULL, vp9_sad32x64x4d)
BFP(BLOCK_32X32, vp9_sad32x32, vp9_sad32x32_avg,
vp9_variance32x32, vp9_sub_pixel_variance32x32,
- vp9_sub_pixel_avg_variance32x32, vp9_variance_halfpixvar32x32_h,
- vp9_variance_halfpixvar32x32_v,
- vp9_variance_halfpixvar32x32_hv, vp9_sad32x32x3, vp9_sad32x32x8,
+ vp9_sub_pixel_avg_variance32x32, vp9_sad32x32x3, vp9_sad32x32x8,
vp9_sad32x32x4d)
BFP(BLOCK_64X64, vp9_sad64x64, vp9_sad64x64_avg,
vp9_variance64x64, vp9_sub_pixel_variance64x64,
- vp9_sub_pixel_avg_variance64x64, vp9_variance_halfpixvar64x64_h,
- vp9_variance_halfpixvar64x64_v,
- vp9_variance_halfpixvar64x64_hv, vp9_sad64x64x3, vp9_sad64x64x8,
+ vp9_sub_pixel_avg_variance64x64, vp9_sad64x64x3, vp9_sad64x64x8,
vp9_sad64x64x4d)
BFP(BLOCK_16X16, vp9_sad16x16, vp9_sad16x16_avg,
vp9_variance16x16, vp9_sub_pixel_variance16x16,
- vp9_sub_pixel_avg_variance16x16, vp9_variance_halfpixvar16x16_h,
- vp9_variance_halfpixvar16x16_v,
- vp9_variance_halfpixvar16x16_hv, vp9_sad16x16x3, vp9_sad16x16x8,
+ vp9_sub_pixel_avg_variance16x16, vp9_sad16x16x3, vp9_sad16x16x8,
vp9_sad16x16x4d)
BFP(BLOCK_16X8, vp9_sad16x8, vp9_sad16x8_avg,
vp9_variance16x8, vp9_sub_pixel_variance16x8,
- vp9_sub_pixel_avg_variance16x8, NULL, NULL, NULL,
+ vp9_sub_pixel_avg_variance16x8,
vp9_sad16x8x3, vp9_sad16x8x8, vp9_sad16x8x4d)
BFP(BLOCK_8X16, vp9_sad8x16, vp9_sad8x16_avg,
vp9_variance8x16, vp9_sub_pixel_variance8x16,
- vp9_sub_pixel_avg_variance8x16, NULL, NULL, NULL,
+ vp9_sub_pixel_avg_variance8x16,
vp9_sad8x16x3, vp9_sad8x16x8, vp9_sad8x16x4d)
BFP(BLOCK_8X8, vp9_sad8x8, vp9_sad8x8_avg,
vp9_variance8x8, vp9_sub_pixel_variance8x8,
- vp9_sub_pixel_avg_variance8x8, NULL, NULL, NULL,
+ vp9_sub_pixel_avg_variance8x8,
vp9_sad8x8x3, vp9_sad8x8x8, vp9_sad8x8x4d)
BFP(BLOCK_8X4, vp9_sad8x4, vp9_sad8x4_avg,
vp9_variance8x4, vp9_sub_pixel_variance8x4,
- vp9_sub_pixel_avg_variance8x4, NULL, NULL,
- NULL, NULL, vp9_sad8x4x8,
- vp9_sad8x4x4d)
+ vp9_sub_pixel_avg_variance8x4, NULL, vp9_sad8x4x8, vp9_sad8x4x4d)
BFP(BLOCK_4X8, vp9_sad4x8, vp9_sad4x8_avg,
vp9_variance4x8, vp9_sub_pixel_variance4x8,
- vp9_sub_pixel_avg_variance4x8, NULL, NULL,
- NULL, NULL, vp9_sad4x8x8,
- vp9_sad4x8x4d)
+ vp9_sub_pixel_avg_variance4x8, NULL, vp9_sad4x8x8, vp9_sad4x8x4d)
BFP(BLOCK_4X4, vp9_sad4x4, vp9_sad4x4_avg,
vp9_variance4x4, vp9_sub_pixel_variance4x4,
- vp9_sub_pixel_avg_variance4x4, NULL, NULL, NULL,
+ vp9_sub_pixel_avg_variance4x4,
vp9_sad4x4x3, vp9_sad4x4x8, vp9_sad4x4x4d)
cpi->full_search_sad = vp9_full_search_sad;
@@ -1440,77 +1418,67 @@
}
#endif
-static void scale_and_extend_frame_nonnormative(YV12_BUFFER_CONFIG *src_fb,
- YV12_BUFFER_CONFIG *dst_fb) {
- const int in_w = src_fb->y_crop_width;
- const int in_h = src_fb->y_crop_height;
- const int out_w = dst_fb->y_crop_width;
- const int out_h = dst_fb->y_crop_height;
- const int in_w_uv = src_fb->uv_crop_width;
- const int in_h_uv = src_fb->uv_crop_height;
- const int out_w_uv = dst_fb->uv_crop_width;
- const int out_h_uv = dst_fb->uv_crop_height;
+static void scale_and_extend_frame_nonnormative(const YV12_BUFFER_CONFIG *src,
+ YV12_BUFFER_CONFIG *dst) {
+ // TODO(dkovalev): replace YV12_BUFFER_CONFIG with vpx_image_t
int i;
+ const uint8_t *const srcs[4] = {src->y_buffer, src->u_buffer, src->v_buffer,
+ src->alpha_buffer};
+ const int src_strides[4] = {src->y_stride, src->uv_stride, src->uv_stride,
+ src->alpha_stride};
+ const int src_widths[4] = {src->y_crop_width, src->uv_crop_width,
+ src->uv_crop_width, src->y_crop_width};
+ const int src_heights[4] = {src->y_crop_height, src->uv_crop_height,
+ src->uv_crop_height, src->y_crop_height};
+ uint8_t *const dsts[4] = {dst->y_buffer, dst->u_buffer, dst->v_buffer,
+ dst->alpha_buffer};
+ const int dst_strides[4] = {dst->y_stride, dst->uv_stride, dst->uv_stride,
+ dst->alpha_stride};
+ const int dst_widths[4] = {dst->y_crop_width, dst->uv_crop_width,
+ dst->uv_crop_width, dst->y_crop_width};
+ const int dst_heights[4] = {dst->y_crop_height, dst->uv_crop_height,
+ dst->uv_crop_height, dst->y_crop_height};
- uint8_t *srcs[4] = {src_fb->y_buffer, src_fb->u_buffer, src_fb->v_buffer,
- src_fb->alpha_buffer};
- int src_strides[4] = {src_fb->y_stride, src_fb->uv_stride, src_fb->uv_stride,
- src_fb->alpha_stride};
+ for (i = 0; i < MAX_MB_PLANE; ++i)
+ vp9_resize_plane(srcs[i], src_heights[i], src_widths[i], src_strides[i],
+ dsts[i], dst_heights[i], dst_widths[i], dst_strides[i]);
- uint8_t *dsts[4] = {dst_fb->y_buffer, dst_fb->u_buffer, dst_fb->v_buffer,
- dst_fb->alpha_buffer};
- int dst_strides[4] = {dst_fb->y_stride, dst_fb->uv_stride, dst_fb->uv_stride,
- dst_fb->alpha_stride};
-
- for (i = 0; i < MAX_MB_PLANE; ++i) {
- if (i == 0 || i == 3) {
- // Y and alpha planes
- vp9_resize_plane(srcs[i], in_h, in_w, src_strides[i],
- dsts[i], out_h, out_w, dst_strides[i]);
- } else {
- // Chroma planes
- vp9_resize_plane(srcs[i], in_h_uv, in_w_uv, src_strides[i],
- dsts[i], out_h_uv, out_w_uv, dst_strides[i]);
- }
- }
// TODO(hkuang): Call C version explicitly
// as neon version only expand border size 32.
- vp8_yv12_extend_frame_borders_c(dst_fb);
+ vp8_yv12_extend_frame_borders_c(dst);
}
-static void scale_and_extend_frame(YV12_BUFFER_CONFIG *src_fb,
- YV12_BUFFER_CONFIG *dst_fb) {
- const int in_w = src_fb->y_crop_width;
- const int in_h = src_fb->y_crop_height;
- const int out_w = dst_fb->y_crop_width;
- const int out_h = dst_fb->y_crop_height;
+static void scale_and_extend_frame(const YV12_BUFFER_CONFIG *src,
+ YV12_BUFFER_CONFIG *dst) {
+ const int src_w = src->y_crop_width;
+ const int src_h = src->y_crop_height;
+ const int dst_w = dst->y_crop_width;
+ const int dst_h = dst->y_crop_height;
+ const uint8_t *const srcs[4] = {src->y_buffer, src->u_buffer, src->v_buffer,
+ src->alpha_buffer};
+ const int src_strides[4] = {src->y_stride, src->uv_stride, src->uv_stride,
+ src->alpha_stride};
+ uint8_t *const dsts[4] = {dst->y_buffer, dst->u_buffer, dst->v_buffer,
+ dst->alpha_buffer};
+ const int dst_strides[4] = {dst->y_stride, dst->uv_stride, dst->uv_stride,
+ dst->alpha_stride};
int x, y, i;
- uint8_t *srcs[4] = {src_fb->y_buffer, src_fb->u_buffer, src_fb->v_buffer,
- src_fb->alpha_buffer};
- int src_strides[4] = {src_fb->y_stride, src_fb->uv_stride, src_fb->uv_stride,
- src_fb->alpha_stride};
-
- uint8_t *dsts[4] = {dst_fb->y_buffer, dst_fb->u_buffer, dst_fb->v_buffer,
- dst_fb->alpha_buffer};
- int dst_strides[4] = {dst_fb->y_stride, dst_fb->uv_stride, dst_fb->uv_stride,
- dst_fb->alpha_stride};
-
- for (y = 0; y < out_h; y += 16) {
- for (x = 0; x < out_w; x += 16) {
+ for (y = 0; y < dst_h; y += 16) {
+ for (x = 0; x < dst_w; x += 16) {
for (i = 0; i < MAX_MB_PLANE; ++i) {
const int factor = (i == 0 || i == 3 ? 1 : 2);
- const int x_q4 = x * (16 / factor) * in_w / out_w;
- const int y_q4 = y * (16 / factor) * in_h / out_h;
+ const int x_q4 = x * (16 / factor) * src_w / dst_w;
+ const int y_q4 = y * (16 / factor) * src_h / dst_h;
const int src_stride = src_strides[i];
const int dst_stride = dst_strides[i];
- uint8_t *src = srcs[i] + y / factor * in_h / out_h * src_stride +
- x / factor * in_w / out_w;
- uint8_t *dst = dsts[i] + y / factor * dst_stride + x / factor;
+ const uint8_t *src_ptr = srcs[i] + (y / factor) * src_h / dst_h *
+ src_stride + (x / factor) * src_w / dst_w;
+ uint8_t *dst_ptr = dsts[i] + (y / factor) * dst_stride + (x / factor);
- vp9_convolve8(src, src_stride, dst, dst_stride,
- vp9_sub_pel_filters_8[x_q4 & 0xf], 16 * in_w / out_w,
- vp9_sub_pel_filters_8[y_q4 & 0xf], 16 * in_h / out_h,
+ vp9_convolve8(src_ptr, src_stride, dst_ptr, dst_stride,
+ vp9_sub_pel_filters_8[x_q4 & 0xf], 16 * src_w / dst_w,
+ vp9_sub_pel_filters_8[y_q4 & 0xf], 16 * src_h / dst_h,
16 / factor, 16 / factor);
}
}
@@ -1518,7 +1486,7 @@
// TODO(hkuang): Call C version explicitly
// as neon version only expand border size 32.
- vp8_yv12_extend_frame_borders_c(dst_fb);
+ vp8_yv12_extend_frame_borders_c(dst);
}
static int find_fp_qindex() {
@@ -1697,7 +1665,7 @@
for (ref_frame = LAST_FRAME; ref_frame <= ALTREF_FRAME; ++ref_frame) {
const int idx = cm->ref_frame_map[get_ref_frame_idx(cpi, ref_frame)];
- YV12_BUFFER_CONFIG *const ref = &cm->frame_bufs[idx].buf;
+ const YV12_BUFFER_CONFIG *const ref = &cm->frame_bufs[idx].buf;
if (ref->y_crop_width != cm->width ||
ref->y_crop_height != cm->height) {
diff --git a/vp9/encoder/vp9_mcomp.c b/vp9/encoder/vp9_mcomp.c
index 7496cc0..bbec4da 100644
--- a/vp9/encoder/vp9_mcomp.c
+++ b/vp9/encoder/vp9_mcomp.c
@@ -1359,108 +1359,89 @@
int sad_per_bit, int distance,
const vp9_variance_fn_ptr_t *fn_ptr,
const MV *center_mv, MV *best_mv) {
+ int r;
const MACROBLOCKD *const xd = &x->e_mbd;
- const uint8_t *const what = x->plane[0].src.buf;
- const int what_stride = x->plane[0].src.stride;
- const uint8_t *const in_what = xd->plane[0].pre[0].buf;
- const int in_what_stride = xd->plane[0].pre[0].stride;
- MV this_mv;
- unsigned int bestsad = INT_MAX;
- int r, c;
- int ref_row = ref_mv->row;
- int ref_col = ref_mv->col;
-
- // Apply further limits to prevent us looking using vectors that stretch
- // beyond the UMV border
- const int row_min = MAX(ref_row - distance, x->mv_row_min);
- const int row_max = MIN(ref_row + distance, x->mv_row_max);
- const int col_min = MAX(ref_col - distance, x->mv_col_min);
- const int col_max = MIN(ref_col + distance, x->mv_col_max);
- DECLARE_ALIGNED_ARRAY(16, uint32_t, sad_array8, 8);
- unsigned int sad_array[3];
+ const struct buf_2d *const what = &x->plane[0].src;
+ const struct buf_2d *const in_what = &xd->plane[0].pre[0];
+ const int row_min = MAX(ref_mv->row - distance, x->mv_row_min);
+ const int row_max = MIN(ref_mv->row + distance, x->mv_row_max);
+ const int col_min = MAX(ref_mv->col - distance, x->mv_col_min);
+ const int col_max = MIN(ref_mv->col + distance, x->mv_col_max);
const MV fcenter_mv = {center_mv->row >> 3, center_mv->col >> 3};
+ unsigned int best_sad = fn_ptr->sdf(what->buf, what->stride,
+ get_buf_from_mv(in_what, ref_mv), in_what->stride, 0x7fffffff) +
+ mvsad_err_cost(x, ref_mv, &fcenter_mv, sad_per_bit);
+ *best_mv = *ref_mv;
- // Work out the mid point for the search
- const uint8_t *bestaddress = &in_what[ref_row * in_what_stride + ref_col];
+ for (r = row_min; r < row_max; ++r) {
+ int c = col_min;
+ const uint8_t *check_here = &in_what->buf[r * in_what->stride + c];
- best_mv->row = ref_row;
- best_mv->col = ref_col;
+ if (fn_ptr->sdx8f != NULL) {
+ while ((c + 7) < col_max) {
+ int i;
+ unsigned int sads[8];
- // Baseline value at the center
- bestsad = fn_ptr->sdf(what, what_stride,
- bestaddress, in_what_stride, 0x7fffffff)
- + mvsad_err_cost(x, best_mv, &fcenter_mv, sad_per_bit);
+ fn_ptr->sdx8f(what->buf, what->stride, check_here, in_what->stride,
+ sads);
- for (r = row_min; r < row_max; r++) {
- const uint8_t *check_here = &in_what[r * in_what_stride + col_min];
- this_mv.row = r;
- c = col_min;
-
- while ((c + 7) < col_max) {
- int i;
-
- fn_ptr->sdx8f(what, what_stride, check_here, in_what_stride, sad_array8);
-
- for (i = 0; i < 8; i++) {
- unsigned int thissad = (unsigned int)sad_array8[i];
-
- if (thissad < bestsad) {
- this_mv.col = c;
- thissad += mvsad_err_cost(x, &this_mv, &fcenter_mv, sad_per_bit);
- if (thissad < bestsad) {
- bestsad = thissad;
- best_mv->row = r;
- best_mv->col = c;
+ for (i = 0; i < 8; ++i) {
+ unsigned int sad = sads[i];
+ if (sad < best_sad) {
+ const MV mv = {r, c};
+ sad += mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit);
+ if (sad < best_sad) {
+ best_sad = sad;
+ *best_mv = mv;
+ }
}
+ ++check_here;
+ ++c;
}
-
- check_here++;
- c++;
}
}
- while ((c + 2) < col_max && fn_ptr->sdx3f != NULL) {
- int i;
+ if (fn_ptr->sdx3f != NULL) {
+ while ((c + 2) < col_max) {
+ int i;
+ unsigned int sads[3];
- fn_ptr->sdx3f(what, what_stride, check_here, in_what_stride, sad_array);
+ fn_ptr->sdx3f(what->buf, what->stride, check_here, in_what->stride,
+ sads);
- for (i = 0; i < 3; i++) {
- unsigned int thissad = sad_array[i];
-
- if (thissad < bestsad) {
- this_mv.col = c;
- thissad += mvsad_err_cost(x, &this_mv, &fcenter_mv, sad_per_bit);
- if (thissad < bestsad) {
- bestsad = thissad;
- best_mv->row = r;
- best_mv->col = c;
+ for (i = 0; i < 3; ++i) {
+ unsigned int sad = sads[i];
+ if (sad < best_sad) {
+ const MV mv = {r, c};
+ sad += mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit);
+ if (sad < best_sad) {
+ best_sad = sad;
+ *best_mv = mv;
+ }
}
+ ++check_here;
+ ++c;
}
-
- check_here++;
- c++;
}
}
while (c < col_max) {
- unsigned int thissad = fn_ptr->sdf(what, what_stride,
- check_here, in_what_stride, bestsad);
-
- if (thissad < bestsad) {
- this_mv.col = c;
- thissad += mvsad_err_cost(x, &this_mv, &fcenter_mv, sad_per_bit);
- if (thissad < bestsad) {
- bestsad = thissad;
- best_mv->row = r;
- best_mv->col = c;
+ unsigned int sad = fn_ptr->sdf(what->buf, what->stride,
+ check_here, in_what->stride, best_sad);
+ if (sad < best_sad) {
+ const MV mv = {r, c};
+ sad += mvsad_err_cost(x, &mv, &fcenter_mv, sad_per_bit);
+ if (sad < best_sad) {
+ best_sad = sad;
+ *best_mv = mv;
}
}
-
- check_here++;
- c++;
+ ++check_here;
+ ++c;
}
}
- return bestsad;
+
+ return best_sad;
}
int vp9_refining_search_sad_c(const MACROBLOCK *x,
diff --git a/vp9/encoder/vp9_variance.c b/vp9/encoder/vp9_variance.c
index 520ee44..ae3c86a 100644
--- a/vp9/encoder/vp9_variance.c
+++ b/vp9/encoder/vp9_variance.c
@@ -276,87 +276,6 @@
SUBPIX_VAR(64, 64)
SUBPIX_AVG_VAR(64, 64)
-unsigned int vp9_variance_halfpixvar16x16_h_c(const uint8_t *src_ptr,
- int source_stride,
- const uint8_t *ref_ptr,
- int recon_stride,
- unsigned int *sse) {
- return vp9_sub_pixel_variance16x16_c(src_ptr, source_stride, 8, 0,
- ref_ptr, recon_stride, sse);
-}
-
-unsigned int vp9_variance_halfpixvar32x32_h_c(const uint8_t *src_ptr,
- int source_stride,
- const uint8_t *ref_ptr,
- int recon_stride,
- unsigned int *sse) {
- return vp9_sub_pixel_variance32x32_c(src_ptr, source_stride, 8, 0,
- ref_ptr, recon_stride, sse);
-}
-
-unsigned int vp9_variance_halfpixvar64x64_h_c(const uint8_t *src_ptr,
- int source_stride,
- const uint8_t *ref_ptr,
- int recon_stride,
- unsigned int *sse) {
- return vp9_sub_pixel_variance64x64_c(src_ptr, source_stride, 8, 0,
- ref_ptr, recon_stride, sse);
-}
-
-unsigned int vp9_variance_halfpixvar16x16_v_c(const uint8_t *src_ptr,
- int source_stride,
- const uint8_t *ref_ptr,
- int recon_stride,
- unsigned int *sse) {
- return vp9_sub_pixel_variance16x16_c(src_ptr, source_stride, 0, 8,
- ref_ptr, recon_stride, sse);
-}
-
-unsigned int vp9_variance_halfpixvar32x32_v_c(const uint8_t *src_ptr,
- int source_stride,
- const uint8_t *ref_ptr,
- int recon_stride,
- unsigned int *sse) {
- return vp9_sub_pixel_variance32x32_c(src_ptr, source_stride, 0, 8,
- ref_ptr, recon_stride, sse);
-}
-
-unsigned int vp9_variance_halfpixvar64x64_v_c(const uint8_t *src_ptr,
- int source_stride,
- const uint8_t *ref_ptr,
- int recon_stride,
- unsigned int *sse) {
- return vp9_sub_pixel_variance64x64_c(src_ptr, source_stride, 0, 8,
- ref_ptr, recon_stride, sse);
-}
-
-unsigned int vp9_variance_halfpixvar16x16_hv_c(const uint8_t *src_ptr,
- int source_stride,
- const uint8_t *ref_ptr,
- int recon_stride,
- unsigned int *sse) {
- return vp9_sub_pixel_variance16x16_c(src_ptr, source_stride, 8, 8,
- ref_ptr, recon_stride, sse);
-}
-
-unsigned int vp9_variance_halfpixvar32x32_hv_c(const uint8_t *src_ptr,
- int source_stride,
- const uint8_t *ref_ptr,
- int recon_stride,
- unsigned int *sse) {
- return vp9_sub_pixel_variance32x32_c(src_ptr, source_stride, 8, 8,
- ref_ptr, recon_stride, sse);
-}
-
-unsigned int vp9_variance_halfpixvar64x64_hv_c(const uint8_t *src_ptr,
- int source_stride,
- const uint8_t *ref_ptr,
- int recon_stride,
- unsigned int *sse) {
- return vp9_sub_pixel_variance64x64_c(src_ptr, source_stride, 8, 8,
- ref_ptr, recon_stride, sse);
-}
-
void vp9_comp_avg_pred(uint8_t *comp_pred, const uint8_t *pred, int width,
int height, const uint8_t *ref, int ref_stride) {
int i, j;
diff --git a/vp9/encoder/vp9_variance.h b/vp9/encoder/vp9_variance.h
index 4c8be71..152c3d9 100644
--- a/vp9/encoder/vp9_variance.h
+++ b/vp9/encoder/vp9_variance.h
@@ -82,9 +82,6 @@
vp9_variance_fn_t vf;
vp9_subpixvariance_fn_t svf;
vp9_subp_avg_variance_fn_t svaf;
- vp9_variance_fn_t svf_halfpix_h;
- vp9_variance_fn_t svf_halfpix_v;
- vp9_variance_fn_t svf_halfpix_hv;
vp9_sad_multi_fn_t sdx3f;
vp9_sad_multi_fn_t sdx8f;
vp9_sad_multi_d_fn_t sdx4df;
diff --git a/vp9/encoder/x86/vp9_variance_sse2.c b/vp9/encoder/x86/vp9_variance_sse2.c
index 9e65694..25d5946 100644
--- a/vp9/encoder/x86/vp9_variance_sse2.c
+++ b/vp9/encoder/x86/vp9_variance_sse2.c
@@ -494,58 +494,3 @@
#undef FNS
#undef FN
-
-unsigned int vp9_variance_halfpixvar16x16_h_sse2(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse) {
- int xsum0;
- unsigned int xxsum0;
-
- vp9_half_horiz_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum0, &xxsum0);
-
- *sse = xxsum0;
- return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 8));
-}
-
-
-unsigned int vp9_variance_halfpixvar16x16_v_sse2(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse) {
- int xsum0;
- unsigned int xxsum0;
- vp9_half_vert_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum0, &xxsum0);
-
- *sse = xxsum0;
- return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 8));
-}
-
-
-unsigned int vp9_variance_halfpixvar16x16_hv_sse2(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse) {
- int xsum0;
- unsigned int xxsum0;
-
- vp9_half_horiz_vert_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum0, &xxsum0);
-
- *sse = xxsum0;
- return (xxsum0 - (((unsigned int)xsum0 * xsum0) >> 8));
-}
