ARMv6 optimized Intra4x4 prediction
Added ARM optimized intra 4x4 prediction
- 2x faster on Profiler compared to C-code compiled with -O3
- Function interface changed a little to improve BLOCKD structure
access
Change-Id: I9bc2b723155943fe0cf03dd9ca5f1760f7a81f54
diff --git a/vp8/common/arm/arm_systemdependent.c b/vp8/common/arm/arm_systemdependent.c
index 2928851..b5f194d 100644
--- a/vp8/common/arm/arm_systemdependent.c
+++ b/vp8/common/arm/arm_systemdependent.c
@@ -63,6 +63,7 @@
rtcd->recon.copy16x16 = vp8_copy_mem16x16_v6;
rtcd->recon.copy8x8 = vp8_copy_mem8x8_v6;
rtcd->recon.copy8x4 = vp8_copy_mem8x4_v6;
+ rtcd->recon.intra4x4_predict = vp8_intra4x4_predict_armv6;
}
#endif
diff --git a/vp8/common/arm/armv6/intra4x4_predict_v6.asm b/vp8/common/arm/armv6/intra4x4_predict_v6.asm
new file mode 100644
index 0000000..a974cd1
--- /dev/null
+++ b/vp8/common/arm/armv6/intra4x4_predict_v6.asm
@@ -0,0 +1,606 @@
+;
+; 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.
+;
+
+
+ EXPORT |vp8_intra4x4_predict_armv6|
+
+ ARM
+ REQUIRE8
+ PRESERVE8
+
+ AREA ||.text||, CODE, READONLY, ALIGN=2
+
+
+;void vp8_intra4x4_predict(unsigned char *src, int src_stride, int b_mode,
+; unsigned char *dst, int dst_stride)
+
+|vp8_intra4x4_predict_armv6| PROC
+ push {r4-r12, lr}
+
+
+ cmp r2, #10
+ addlt pc, pc, r2, lsl #2 ; position independent switch
+ pop {r4-r12, pc} ; default
+ b b_dc_pred
+ b b_tm_pred
+ b b_ve_pred
+ b b_he_pred
+ b b_ld_pred
+ b b_rd_pred
+ b b_vr_pred
+ b b_vl_pred
+ b b_hd_pred
+ b b_hu_pred
+
+b_dc_pred
+ ; load values
+ ldr r8, [r0, -r1] ; Above
+ ldrb r4, [r0, #-1]! ; Left[0]
+ mov r9, #0
+ ldrb r5, [r0, r1] ; Left[1]
+ ldrb r6, [r0, r1, lsl #1]! ; Left[2]
+ usad8 r12, r8, r9
+ ldrb r7, [r0, r1] ; Left[3]
+
+ ; calculate dc
+ add r4, r4, r5
+ add r4, r4, r6
+ add r4, r4, r7
+ add r4, r4, r12
+ add r4, r4, #4
+ ldr r0, [sp, #40] ; load stride
+ mov r12, r4, asr #3 ; (expected_dc + 4) >> 3
+
+ add r12, r12, r12, lsl #8
+ add r3, r3, r0
+ add r12, r12, r12, lsl #16
+
+ ; store values
+ str r12, [r3, -r0]
+ str r12, [r3]
+ str r12, [r3, r0]
+ str r12, [r3, r0, lsl #1]
+
+ pop {r4-r12, pc}
+
+b_tm_pred
+ sub r10, r0, #1 ; Left
+ ldr r8, [r0, -r1] ; Above
+ ldrb r9, [r10, -r1] ; top_left
+ ldrb r4, [r0, #-1]! ; Left[0]
+ ldrb r5, [r10, r1]! ; Left[1]
+ ldrb r6, [r0, r1, lsl #1] ; Left[2]
+ ldrb r7, [r10, r1, lsl #1] ; Left[3]
+ ldr r0, [sp, #40] ; load stride
+
+
+ add r9, r9, r9, lsl #16 ; [tl|tl]
+ uxtb16 r10, r8 ; a[2|0]
+ uxtb16 r11, r8, ror #8 ; a[3|1]
+ ssub16 r10, r10, r9 ; a[2|0] - [tl|tl]
+ ssub16 r11, r11, r9 ; a[3|1] - [tl|tl]
+
+ add r4, r4, r4, lsl #16 ; l[0|0]
+ add r5, r5, r5, lsl #16 ; l[1|1]
+ add r6, r6, r6, lsl #16 ; l[2|2]
+ add r7, r7, r7, lsl #16 ; l[3|3]
+
+ sadd16 r1, r4, r10 ; l[0|0] + a[2|0] - [tl|tl]
+ sadd16 r2, r4, r11 ; l[0|0] + a[3|1] - [tl|tl]
+ usat16 r1, #8, r1
+ usat16 r2, #8, r2
+
+ sadd16 r4, r5, r10 ; l[1|1] + a[2|0] - [tl|tl]
+ sadd16 r5, r5, r11 ; l[1|1] + a[3|1] - [tl|tl]
+
+ add r12, r1, r2, lsl #8 ; [3|2|1|0]
+ str r12, [r3], r0
+
+ usat16 r4, #8, r4
+ usat16 r5, #8, r5
+
+ sadd16 r1, r6, r10 ; l[2|2] + a[2|0] - [tl|tl]
+ sadd16 r2, r6, r11 ; l[2|2] + a[3|1] - [tl|tl]
+
+ add r12, r4, r5, lsl #8 ; [3|2|1|0]
+ str r12, [r3], r0
+
+ usat16 r1, #8, r1
+ usat16 r2, #8, r2
+
+ sadd16 r4, r7, r10 ; l[3|3] + a[2|0] - [tl|tl]
+ sadd16 r5, r7, r11 ; l[3|3] + a[3|1] - [tl|tl]
+
+ add r12, r1, r2, lsl #8 ; [3|2|1|0]
+
+ usat16 r4, #8, r4
+ usat16 r5, #8, r5
+
+ str r12, [r3], r0
+
+ add r12, r4, r5, lsl #8 ; [3|2|1|0]
+ str r12, [r3], r0
+
+ pop {r4-r12, pc}
+
+b_ve_pred
+ ldr r8, [r0, -r1]! ; a[3|2|1|0]
+ ldr r11, c00FF00FF
+ ldrb r9, [r0, #-1] ; top_left
+ ldrb r10, [r0, #4] ; a[4]
+
+ ldr r0, c00020002
+
+ uxtb16 r4, r8 ; a[2|0]
+ uxtb16 r5, r8, ror #8 ; a[3|1]
+ ldr r2, [sp, #40] ; stride
+ pkhbt r9, r9, r5, lsl #16 ; a[1|-1]
+
+ add r9, r9, r4, lsl #1 ;[a[1]+2*a[2] | tl+2*a[0] ]
+ uxtab16 r9, r9, r5 ;[a[1]+2*a[2]+a[3] | tl+2*a[0]+a[1] ]
+ uxtab16 r9, r9, r0 ;[a[1]+2*a[2]+a[3]+2| tl+2*a[0]+a[1]+2]
+
+ add r0, r0, r10, lsl #16 ;[a[4]+2 | 2]
+ add r0, r0, r4, asr #16 ;[a[4]+2 | a[2]+2]
+ add r0, r0, r5, lsl #1 ;[a[4]+2*a[3]+2 | a[2]+2*a[1]+2]
+ uadd16 r4, r4, r0 ;[a[4]+2*a[3]+a[2]+2|a[2]+2*a[1]+a[0]+2]
+
+ and r9, r11, r9, asr #2
+ and r4, r11, r4, asr #2
+ add r3, r3, r2 ; dst + dst_stride
+ add r9, r9, r4, lsl #8
+
+ ; store values
+ str r9, [r3, -r2]
+ str r9, [r3]
+ str r9, [r3, r2]
+ str r9, [r3, r2, lsl #1]
+
+ pop {r4-r12, pc}
+
+
+b_he_pred
+ sub r10, r0, #1 ; Left
+ ldrb r4, [r0, #-1]! ; Left[0]
+ ldrb r8, [r10, -r1] ; top_left
+ ldrb r5, [r10, r1]! ; Left[1]
+ ldrb r6, [r0, r1, lsl #1] ; Left[2]
+ ldrb r7, [r10, r1, lsl #1] ; Left[3]
+
+ add r8, r8, r4 ; tl + l[0]
+ add r9, r4, r5 ; l[0] + l[1]
+ add r10, r5, r6 ; l[1] + l[2]
+ add r11, r6, r7 ; l[2] + l[3]
+
+ mov r0, #2<<14
+
+ add r8, r8, r9 ; tl + 2*l[0] + l[1]
+ add r4, r9, r10 ; l[0] + 2*l[1] + l[2]
+ add r5, r10, r11 ; l[1] + 2*l[2] + l[3]
+ add r6, r11, r7, lsl #1 ; l[2] + 2*l[3] + l[3]
+
+
+ add r8, r0, r8, lsl #14 ; (tl + 2*l[0] + l[1])>>2 in top half
+ add r9, r0, r4, lsl #14 ; (l[0] + 2*l[1] + l[2])>>2 in top half
+ add r10,r0, r5, lsl #14 ; (l[1] + 2*l[2] + l[3])>>2 in top half
+ add r11,r0, r6, lsl #14 ; (l[2] + 2*l[3] + l[3])>>2 in top half
+
+ pkhtb r8, r8, r8, asr #16 ; l[-|0|-|0]
+ pkhtb r9, r9, r9, asr #16 ; l[-|1|-|1]
+ pkhtb r10, r10, r10, asr #16 ; l[-|2|-|2]
+ pkhtb r11, r11, r11, asr #16 ; l[-|3|-|3]
+
+ ldr r0, [sp, #40] ; stride
+
+ add r8, r8, r8, lsl #8 ; l[0|0|0|0]
+ add r9, r9, r9, lsl #8 ; l[1|1|1|1]
+ add r10, r10, r10, lsl #8 ; l[2|2|2|2]
+ add r11, r11, r11, lsl #8 ; l[3|3|3|3]
+
+ ; store values
+ str r8, [r3], r0
+ str r9, [r3]
+ str r10, [r3, r0]
+ str r11, [r3, r0, lsl #1]
+
+ pop {r4-r12, pc}
+
+b_ld_pred
+ ldr r4, [r0, -r1]! ; Above
+ ldr r12, c00020002
+ ldr r5, [r0, #4]
+ ldr lr, c00FF00FF
+
+ uxtb16 r6, r4 ; a[2|0]
+ uxtb16 r7, r4, ror #8 ; a[3|1]
+ uxtb16 r8, r5 ; a[6|4]
+ uxtb16 r9, r5, ror #8 ; a[7|5]
+ pkhtb r10, r6, r8 ; a[2|4]
+ pkhtb r11, r7, r9 ; a[3|5]
+
+
+ add r4, r6, r7, lsl #1 ; [a2+2*a3 | a0+2*a1]
+ add r4, r4, r10, ror #16 ; [a2+2*a3+a4 | a0+2*a1+a2]
+ uxtab16 r4, r4, r12 ; [a2+2*a3+a4+2 | a0+2*a1+a2+2]
+
+ add r5, r7, r10, ror #15 ; [a3+2*a4 | a1+2*a2]
+ add r5, r5, r11, ror #16 ; [a3+2*a4+a5 | a1+2*a2+a3]
+ uxtab16 r5, r5, r12 ; [a3+2*a4+a5+2 | a1+2*a2+a3+2]
+
+ pkhtb r7, r9, r8, asr #16
+ add r6, r8, r9, lsl #1 ; [a6+2*a7 | a4+2*a5]
+ uadd16 r6, r6, r7 ; [a6+2*a7+a7 | a4+2*a5+a6]
+ uxtab16 r6, r6, r12 ; [a6+2*a7+a7+2 | a4+2*a5+a6+2]
+
+ uxth r7, r9 ; [ a5]
+ add r7, r7, r8, asr #15 ; [ a5+2*a6]
+ add r7, r7, r9, asr #16 ; [ a5+2*a6+a7]
+ uxtah r7, r7, r12 ; [ a5+2*a6+a7+2]
+
+ ldr r0, [sp, #40] ; stride
+
+ ; scale down
+ and r4, lr, r4, asr #2
+ and r5, lr, r5, asr #2
+ and r6, lr, r6, asr #2
+ mov r7, r7, asr #2
+
+ add r8, r4, r5, lsl #8 ; [3|2|1|0]
+ str r8, [r3], r0
+
+ mov r9, r8, lsr #8
+ add r9, r9, r6, lsl #24 ; [4|3|2|1]
+ str r9, [r3], r0
+
+ mov r10, r9, lsr #8
+ add r10, r10, r7, lsl #24 ; [5|4|3|2]
+ str r10, [r3], r0
+
+ mov r6, r6, lsr #16
+ mov r11, r10, lsr #8
+ add r11, r11, r6, lsl #24 ; [6|5|4|3]
+ str r11, [r3], r0
+
+ pop {r4-r12, pc}
+
+b_rd_pred
+ sub r12, r0, r1 ; Above = src - src_stride
+ ldrb r7, [r0, #-1]! ; l[0] = pp[3]
+ ldr lr, [r12] ; Above = pp[8|7|6|5]
+ ldrb r8, [r12, #-1]! ; tl = pp[4]
+ ldrb r6, [r12, r1, lsl #1] ; l[1] = pp[2]
+ ldrb r5, [r0, r1, lsl #1] ; l[2] = pp[1]
+ ldrb r4, [r12, r1, lsl #2] ; l[3] = pp[0]
+
+
+ uxtb16 r9, lr ; p[7|5]
+ uxtb16 r10, lr, ror #8 ; p[8|6]
+ add r4, r4, r6, lsl #16 ; p[2|0]
+ add r5, r5, r7, lsl #16 ; p[3|1]
+ add r6, r6, r8, lsl #16 ; p[4|2]
+ pkhbt r7, r7, r9, lsl #16 ; p[5|3]
+ pkhbt r8, r8, r10, lsl #16 ; p[6|4]
+
+ ldr r12, c00020002
+ ldr lr, c00FF00FF
+
+ add r4, r4, r5, lsl #1 ; [p2+2*p3 | p0+2*p1]
+ add r4, r4, r6 ; [p2+2*p3+p4 | p0+2*p1+p2]
+ uxtab16 r4, r4, r12 ; [p2+2*p3+p4+2 | p0+2*p1+p2+2]
+
+ add r5, r5, r6, lsl #1 ; [p3+2*p4 | p1+2*p2]
+ add r5, r5, r7 ; [p3+2*p4+p5 | p1+2*p2+p3]
+ uxtab16 r5, r5, r12 ; [p3+2*p4+p5+2 | p1+2*p2+p3+2]
+
+ add r6, r7, r8, lsl #1 ; [p5+2*p6 | p3+2*p4]
+ add r6, r6, r9 ; [p5+2*p6+p7 | p3+2*p4+p5]
+ uxtab16 r6, r6, r12 ; [p5+2*p6+p7+2 | p3+2*p4+p5+2]
+
+ add r7, r8, r9, lsl #1 ; [p6+2*p7 | p4+2*p5]
+ add r7, r7, r10 ; [p6+2*p7+p8 | p4+2*p5+p6]
+ uxtab16 r7, r7, r12 ; [p6+2*p7+p8+2 | p4+2*p5+p6+2]
+
+ ldr r0, [sp, #40] ; stride
+
+ ; scale down
+ and r7, lr, r7, asr #2
+ and r6, lr, r6, asr #2
+ and r5, lr, r5, asr #2
+ and r4, lr, r4, asr #2
+
+ add r8, r6, r7, lsl #8 ; [6|5|4|3]
+ str r8, [r3], r0
+
+ mov r9, r8, lsl #8 ; [5|4|3|-]
+ uxtab r9, r9, r4, ror #16 ; [5|4|3|2]
+ str r9, [r3], r0
+
+ mov r10, r9, lsl #8 ; [4|3|2|-]
+ uxtab r10, r10, r5 ; [4|3|2|1]
+ str r10, [r3], r0
+
+ mov r11, r10, lsl #8 ; [3|2|1|-]
+ uxtab r11, r11, r4 ; [3|2|1|0]
+ str r11, [r3], r0
+
+ pop {r4-r12, pc}
+
+b_vr_pred
+ sub r12, r0, r1 ; Above = src - src_stride
+ ldrb r7, [r0, #-1]! ; l[0] = pp[3]
+ ldr lr, [r12] ; Above = pp[8|7|6|5]
+ ldrb r8, [r12, #-1]! ; tl = pp[4]
+ ldrb r6, [r12, r1, lsl #1] ; l[1] = pp[2]
+ ldrb r5, [r0, r1, lsl #1] ; l[2] = pp[1]
+ ldrb r4, [r12, r1, lsl #2] ; l[3] = pp[0]
+
+ add r5, r5, r7, lsl #16 ; p[3|1]
+ add r6, r6, r8, lsl #16 ; p[4|2]
+ uxtb16 r9, lr ; p[7|5]
+ uxtb16 r10, lr, ror #8 ; p[8|6]
+ pkhbt r7, r7, r9, lsl #16 ; p[5|3]
+ pkhbt r8, r8, r10, lsl #16 ; p[6|4]
+
+ ldr r4, c00010001
+ ldr r12, c00020002
+ ldr lr, c00FF00FF
+
+ add r5, r5, r6, lsl #1 ; [p3+2*p4 | p1+2*p2]
+ add r5, r5, r7 ; [p3+2*p4+p5 | p1+2*p2+p3]
+ uxtab16 r5, r5, r12 ; [p3+2*p4+p5+2 | p1+2*p2+p3+2]
+
+ add r6, r6, r7, lsl #1 ; [p4+2*p5 | p2+2*p3]
+ add r6, r6, r8 ; [p4+2*p5+p6 | p2+2*p3+p4]
+ uxtab16 r6, r6, r12 ; [p4+2*p5+p6+2 | p2+2*p3+p4+2]
+
+ uadd16 r11, r8, r9 ; [p6+p7 | p4+p5]
+ uhadd16 r11, r11, r4 ; [(p6+p7+1)>>1 | (p4+p5+1)>>1]
+ ; [F|E]
+
+ add r7, r7, r8, lsl #1 ; [p5+2*p6 | p3+2*p4]
+ add r7, r7, r9 ; [p5+2*p6+p7 | p3+2*p4+p5]
+ uxtab16 r7, r7, r12 ; [p5+2*p6+p7+2 | p3+2*p4+p5+2]
+
+ uadd16 r2, r9, r10 ; [p7+p8 | p5+p6]
+ uhadd16 r2, r2, r4 ; [(p7+p8+1)>>1 | (p5+p6+1)>>1]
+ ; [J|I]
+
+ add r8, r8, r9, lsl #1 ; [p6+2*p7 | p4+2*p5]
+ add r8, r8, r10 ; [p6+2*p7+p8 | p4+2*p5+p6]
+ uxtab16 r8, r8, r12 ; [p6+2*p7+p8+2 | p4+2*p5+p6+2]
+
+ ldr r0, [sp, #40] ; stride
+
+ ; scale down
+ and r5, lr, r5, asr #2 ; [B|A]
+ and r6, lr, r6, asr #2 ; [D|C]
+ and r7, lr, r7, asr #2 ; [H|G]
+ and r8, lr, r8, asr #2 ; [L|K]
+
+ add r12, r11, r2, lsl #8 ; [J|F|I|E]
+ str r12, [r3], r0
+
+ add r12, r7, r8, lsl #8 ; [L|H|K|G]
+ str r12, [r3], r0
+
+ pkhbt r2, r6, r2, lsl #16 ; [-|I|-|C]
+ add r2, r2, r11, lsl #8 ; [F|I|E|C]
+
+ pkhtb r12, r6, r5 ; [-|D|-|A]
+ pkhtb r10, r7, r5, asr #16 ; [-|H|-|B]
+ str r2, [r3], r0
+ add r12, r12, r10, lsl #8 ; [H|D|B|A]
+ str r12, [r3], r0
+
+ pop {r4-r12, pc}
+
+b_vl_pred
+ ldr r4, [r0, -r1]! ; [3|2|1|0]
+ ldr r12, c00020002
+ ldr r5, [r0, #4] ; [7|6|5|4]
+ ldr lr, c00FF00FF
+ ldr r2, c00010001
+
+ mov r0, r4, lsr #16 ; [-|-|3|2]
+ add r0, r0, r5, lsl #16 ; [5|4|3|2]
+ uxtb16 r6, r4 ; [2|0]
+ uxtb16 r7, r4, ror #8 ; [3|1]
+ uxtb16 r8, r0 ; [4|2]
+ uxtb16 r9, r0, ror #8 ; [5|3]
+ uxtb16 r10, r5 ; [6|4]
+ uxtb16 r11, r5, ror #8 ; [7|5]
+
+ uadd16 r4, r6, r7 ; [p2+p3 | p0+p1]
+ uhadd16 r4, r4, r2 ; [(p2+p3+1)>>1 | (p0+p1+1)>>1]
+ ; [B|A]
+
+ add r5, r6, r7, lsl #1 ; [p2+2*p3 | p0+2*p1]
+ add r5, r5, r8 ; [p2+2*p3+p4 | p0+2*p1+p2]
+ uxtab16 r5, r5, r12 ; [p2+2*p3+p4+2 | p0+2*p1+p2+2]
+
+ uadd16 r6, r7, r8 ; [p3+p4 | p1+p2]
+ uhadd16 r6, r6, r2 ; [(p3+p4+1)>>1 | (p1+p2+1)>>1]
+ ; [F|E]
+
+ add r7, r7, r8, lsl #1 ; [p3+2*p4 | p1+2*p2]
+ add r7, r7, r9 ; [p3+2*p4+p5 | p1+2*p2+p3]
+ uxtab16 r7, r7, r12 ; [p3+2*p4+p5+2 | p1+2*p2+p3+2]
+
+ add r8, r8, r9, lsl #1 ; [p4+2*p5 | p2+2*p3]
+ add r8, r8, r10 ; [p4+2*p5+p6 | p2+2*p3+p4]
+ uxtab16 r8, r8, r12 ; [p4+2*p5+p6+2 | p2+2*p3+p4+2]
+
+ add r9, r9, r10, lsl #1 ; [p5+2*p6 | p3+2*p4]
+ add r9, r9, r11 ; [p5+2*p6+p7 | p3+2*p4+p5]
+ uxtab16 r9, r9, r12 ; [p5+2*p6+p7+2 | p3+2*p4+p5+2]
+
+ ldr r0, [sp, #40] ; stride
+
+ ; scale down
+ and r5, lr, r5, asr #2 ; [D|C]
+ and r7, lr, r7, asr #2 ; [H|G]
+ and r8, lr, r8, asr #2 ; [I|D]
+ and r9, lr, r9, asr #2 ; [J|H]
+
+
+ add r10, r4, r6, lsl #8 ; [F|B|E|A]
+ str r10, [r3], r0
+
+ add r5, r5, r7, lsl #8 ; [H|C|G|D]
+ str r5, [r3], r0
+
+ pkhtb r12, r8, r4, asr #16 ; [-|I|-|B]
+ pkhtb r10, r9, r8 ; [-|J|-|D]
+
+ add r12, r6, r12, lsl #8 ; [I|F|B|E]
+ str r12, [r3], r0
+
+ add r10, r7, r10, lsl #8 ; [J|H|D|G]
+ str r10, [r3], r0
+
+ pop {r4-r12, pc}
+
+b_hd_pred
+ sub r12, r0, r1 ; Above = src - src_stride
+ ldrb r7, [r0, #-1]! ; l[0] = pp[3]
+ ldr lr, [r12] ; Above = pp[8|7|6|5]
+ ldrb r8, [r12, #-1]! ; tl = pp[4]
+ ldrb r6, [r0, r1] ; l[1] = pp[2]
+ ldrb r5, [r0, r1, lsl #1] ; l[2] = pp[1]
+ ldrb r4, [r12, r1, lsl #2] ; l[3] = pp[0]
+
+ uxtb16 r9, lr ; p[7|5]
+ uxtb16 r10, lr, ror #8 ; p[8|6]
+
+ add r4, r4, r5, lsl #16 ; p[1|0]
+ add r5, r5, r6, lsl #16 ; p[2|1]
+ add r6, r6, r7, lsl #16 ; p[3|2]
+ add r7, r7, r8, lsl #16 ; p[4|3]
+
+ ldr r12, c00020002
+ ldr lr, c00FF00FF
+ ldr r2, c00010001
+
+ pkhtb r8, r7, r9 ; p[4|5]
+ pkhtb r1, r9, r10 ; p[7|6]
+ pkhbt r10, r8, r10, lsl #16 ; p[6|5]
+
+
+ uadd16 r11, r4, r5 ; [p1+p2 | p0+p1]
+ uhadd16 r11, r11, r2 ; [(p1+p2+1)>>1 | (p0+p1+1)>>1]
+ ; [B|A]
+
+ add r4, r4, r5, lsl #1 ; [p1+2*p2 | p0+2*p1]
+ add r4, r4, r6 ; [p1+2*p2+p3 | p0+2*p1+p2]
+ uxtab16 r4, r4, r12 ; [p1+2*p2+p3+2 | p0+2*p1+p2+2]
+
+ uadd16 r0, r6, r7 ; [p3+p4 | p2+p3]
+ uhadd16 r0, r0, r2 ; [(p3+p4+1)>>1 | (p2+p3+1)>>1]
+ ; [F|E]
+
+ add r5, r6, r7, lsl #1 ; [p3+2*p4 | p2+2*p3]
+ add r5, r5, r8, ror #16 ; [p3+2*p4+p5 | p2+2*p3+p4]
+ uxtab16 r5, r5, r12 ; [p3+2*p4+p5+2 | p2+2*p3+p4+2]
+
+ add r6, r12, r8, ror #16 ; [p5+2 | p4+2]
+ add r6, r6, r10, lsl #1 ; [p5+2+2*p6 | p4+2+2*p5]
+ uxtab16 r6, r6, r1 ; [p5+2+2*p6+p7 | p4+2+2*p5+p6]
+
+ ; scale down
+ and r4, lr, r4, asr #2 ; [D|C]
+ and r5, lr, r5, asr #2 ; [H|G]
+ and r6, lr, r6, asr #2 ; [J|I]
+
+ ldr lr, [sp, #40] ; stride
+
+ pkhtb r2, r0, r6 ; [-|F|-|I]
+ pkhtb r12, r6, r5, asr #16 ; [-|J|-|H]
+ add r12, r12, r2, lsl #8 ; [F|J|I|H]
+ add r2, r0, r5, lsl #8 ; [H|F|G|E]
+ mov r12, r12, ror #24 ; [J|I|H|F]
+ str r12, [r3], lr
+
+
+ mov r7, r11, asr #16 ; [-|-|-|B]
+ str r2, [r3], lr
+ add r7, r7, r0, lsl #16 ; [-|E|-|B]
+ add r7, r7, r4, asr #8 ; [-|E|D|B]
+ add r7, r7, r5, lsl #24 ; [G|E|D|B]
+ str r7, [r3], lr
+
+ add r5, r11, r4, lsl #8 ; [D|B|C|A]
+ str r5, [r3], lr
+
+ pop {r4-r12, pc}
+
+
+
+b_hu_pred
+ ldrb r4, [r0, #-1]! ; Left[0]
+ ldr r12, c00020002
+ ldrb r5, [r0, r1]! ; Left[1]
+ ldr lr, c00FF00FF
+ ldrb r6, [r0, r1]! ; Left[2]
+ ldr r2, c00010001
+ ldrb r7, [r0, r1] ; Left[3]
+
+
+ add r4, r4, r5, lsl #16 ; [1|0]
+ add r5, r5, r6, lsl #16 ; [2|1]
+ add r9, r6, r7, lsl #16 ; [3|2]
+
+ uadd16 r8, r4, r5 ; [p1+p2 | p0+p1]
+ uhadd16 r8, r8, r2 ; [(p1+p2+1)>>1 | (p0+p1+1)>>1]
+ ; [B|A]
+
+ add r4, r4, r5, lsl #1 ; [p1+2*p2 | p0+2*p1]
+ add r4, r4, r9 ; [p1+2*p2+p3 | p0+2*p1+p2]
+ uxtab16 r4, r4, r12 ; [p1+2*p2+p3+2 | p0+2*p1+p2+2]
+ ldr r2, [sp, #40] ; stride
+ and r4, lr, r4, asr #2 ; [D|C]
+
+ add r10, r6, r7 ; [p2+p3]
+ add r11, r10, r7, lsl #1 ; [p2+3*p3]
+ add r10, r10, #1
+ add r11, r11, #2
+ mov r10, r10, asr #1 ; [E]
+ mov r11, r11, asr #2 ; [F]
+
+ add r9, r7, r9, asr #8 ; [-|-|G|G]
+ add r0, r8, r4, lsl #8 ; [D|B|C|A]
+ add r7, r9, r9, lsl #16 ; [G|G|G|G]
+
+ str r0, [r3], r2
+
+ mov r1, r8, asr #16 ; [-|-|-|B]
+ add r1, r1, r4, asr #8 ; [-|-|D|B]
+ add r1, r1, r10, lsl #16 ; [-|E|D|B]
+ add r1, r1, r11, lsl #24 ; [F|E|D|B]
+ str r1, [r3], r2
+
+ add r10, r11, lsl #8 ; [-|-|F|E]
+ add r10, r10, r9, lsl #16 ; [G|G|F|E]
+ str r10, [r3]
+
+ str r7, [r3, r2]
+
+ pop {r4-r12, pc}
+
+ ENDP
+
+; constants
+c00010001
+ DCD 0x00010001
+c00020002
+ DCD 0x00020002
+c00FF00FF
+ DCD 0x00FF00FF
+
+ END
diff --git a/vp8/common/arm/recon_arm.h b/vp8/common/arm/recon_arm.h
index dec7fc4..f8ef56a 100644
--- a/vp8/common/arm/recon_arm.h
+++ b/vp8/common/arm/recon_arm.h
@@ -17,6 +17,7 @@
extern prototype_copy_block(vp8_copy_mem8x8_v6);
extern prototype_copy_block(vp8_copy_mem8x4_v6);
extern prototype_copy_block(vp8_copy_mem16x16_v6);
+extern prototype_intra4x4_predict(vp8_intra4x4_predict_armv6);
#if !CONFIG_RUNTIME_CPU_DETECT
#undef vp8_recon_copy8x8
@@ -27,6 +28,9 @@
#undef vp8_recon_copy16x16
#define vp8_recon_copy16x16 vp8_copy_mem16x16_v6
+
+#undef vp8_recon_intra4x4_predict
+#define vp8_recon_intra4x4_predict vp8_intra4x4_predict_armv6
#endif
#endif
diff --git a/vp8/common/asm_com_offsets.c b/vp8/common/asm_com_offsets.c
index e135a4d..704a3f0 100644
--- a/vp8/common/asm_com_offsets.c
+++ b/vp8/common/asm_com_offsets.c
@@ -13,6 +13,7 @@
#include "vpx/vpx_codec.h"
#include "vpx_ports/asm_offsets.h"
#include "vpx_scale/yv12config.h"
+#include "vp8/common/blockd.h"
BEGIN
@@ -34,6 +35,20 @@
/* add asserts for any offset that is not supported by assembly code */
/* add asserts for any size that is not supported by assembly code */
+#if HAVE_ARMV6
+/* switch case in vp8_intra4x4_predict_armv6 is based on these enumerated values */
+ct_assert(B_DC_PRED, B_DC_PRED == 0);
+ct_assert(B_TM_PRED, B_TM_PRED == 1);
+ct_assert(B_VE_PRED, B_VE_PRED == 2);
+ct_assert(B_HE_PRED, B_HE_PRED == 3);
+ct_assert(B_LD_PRED, B_LD_PRED == 4);
+ct_assert(B_RD_PRED, B_RD_PRED == 5);
+ct_assert(B_VR_PRED, B_VR_PRED == 6);
+ct_assert(B_VL_PRED, B_VL_PRED == 7);
+ct_assert(B_HD_PRED, B_HD_PRED == 8);
+ct_assert(B_HU_PRED, B_HU_PRED == 9);
+#endif
+
#if HAVE_ARMV7
/* vp8_yv12_extend_frame_borders_neon makes several assumptions based on this */
ct_assert(VP8BORDERINPIXELS_VAL, VP8BORDERINPIXELS == 32)
diff --git a/vp8/common/generic/systemdependent.c b/vp8/common/generic/systemdependent.c
index 5c4fbb1..9641d8c 100644
--- a/vp8/common/generic/systemdependent.c
+++ b/vp8/common/generic/systemdependent.c
@@ -88,7 +88,7 @@
rtcd->recon.build_intra_predictors_mbuv_s =
vp8_build_intra_predictors_mbuv_s;
rtcd->recon.intra4x4_predict =
- vp8_intra4x4_predict;
+ vp8_intra4x4_predict_c;
rtcd->subpix.sixtap16x16 = vp8_sixtap_predict16x16_c;
rtcd->subpix.sixtap8x8 = vp8_sixtap_predict8x8_c;
diff --git a/vp8/common/recon.h b/vp8/common/recon.h
index 62bd71a..98c5f63 100644
--- a/vp8/common/recon.h
+++ b/vp8/common/recon.h
@@ -27,7 +27,8 @@
void sym(MACROBLOCKD *x)
#define prototype_intra4x4_predict(sym) \
- void sym(BLOCKD *x, int b_mode, unsigned char *predictor, int stride)
+ void sym(unsigned char *src, int src_stride, int b_mode, \
+ unsigned char *dst, int dst_stride)
struct vp8_recon_rtcd_vtable;
@@ -79,7 +80,7 @@
(vp8_recon_build_intra_predictors_mbuv_s);
#ifndef vp8_recon_intra4x4_predict
-#define vp8_recon_intra4x4_predict vp8_intra4x4_predict
+#define vp8_recon_intra4x4_predict vp8_intra4x4_predict_c
#endif
extern prototype_intra4x4_predict\
(vp8_recon_intra4x4_predict);
diff --git a/vp8/common/reconintra4x4.c b/vp8/common/reconintra4x4.c
index 12430da..64dac5d 100644
--- a/vp8/common/reconintra4x4.c
+++ b/vp8/common/reconintra4x4.c
@@ -9,25 +9,22 @@
*/
-#include "vpx_config.h"
#include "recon.h"
-#include "vpx_mem/vpx_mem.h"
-#include "reconintra.h"
-void vp8_intra4x4_predict(BLOCKD *x,
- int b_mode,
- unsigned char *predictor, int stride)
+void vp8_intra4x4_predict_c(unsigned char *src, int src_stride,
+ int b_mode,
+ unsigned char *dst, int dst_stride)
{
int i, r, c;
- unsigned char *Above = *(x->base_dst) + x->dst - x->dst_stride;
+ unsigned char *Above = src - src_stride;
unsigned char Left[4];
unsigned char top_left = Above[-1];
- Left[0] = (*(x->base_dst))[x->dst - 1];
- Left[1] = (*(x->base_dst))[x->dst - 1 + x->dst_stride];
- Left[2] = (*(x->base_dst))[x->dst - 1 + 2 * x->dst_stride];
- Left[3] = (*(x->base_dst))[x->dst - 1 + 3 * x->dst_stride];
+ Left[0] = src[-1];
+ Left[1] = src[-1 + src_stride];
+ Left[2] = src[-1 + 2 * src_stride];
+ Left[3] = src[-1 + 3 * src_stride];
switch (b_mode)
{
@@ -47,10 +44,10 @@
{
for (c = 0; c < 4; c++)
{
- predictor[c] = expected_dc;
+ dst[c] = expected_dc;
}
- predictor += stride;
+ dst += dst_stride;
}
}
break;
@@ -69,10 +66,10 @@
if (pred > 255)
pred = 255;
- predictor[c] = pred;
+ dst[c] = pred;
}
- predictor += stride;
+ dst += dst_stride;
}
}
break;
@@ -91,10 +88,10 @@
for (c = 0; c < 4; c++)
{
- predictor[c] = ap[c];
+ dst[c] = ap[c];
}
- predictor += stride;
+ dst += dst_stride;
}
}
@@ -114,32 +111,32 @@
{
for (c = 0; c < 4; c++)
{
- predictor[c] = lp[r];
+ dst[c] = lp[r];
}
- predictor += stride;
+ dst += dst_stride;
}
}
break;
case B_LD_PRED:
{
unsigned char *ptr = Above;
- predictor[0 * stride + 0] = (ptr[0] + ptr[1] * 2 + ptr[2] + 2) >> 2;
- predictor[0 * stride + 1] =
- predictor[1 * stride + 0] = (ptr[1] + ptr[2] * 2 + ptr[3] + 2) >> 2;
- predictor[0 * stride + 2] =
- predictor[1 * stride + 1] =
- predictor[2 * stride + 0] = (ptr[2] + ptr[3] * 2 + ptr[4] + 2) >> 2;
- predictor[0 * stride + 3] =
- predictor[1 * stride + 2] =
- predictor[2 * stride + 1] =
- predictor[3 * stride + 0] = (ptr[3] + ptr[4] * 2 + ptr[5] + 2) >> 2;
- predictor[1 * stride + 3] =
- predictor[2 * stride + 2] =
- predictor[3 * stride + 1] = (ptr[4] + ptr[5] * 2 + ptr[6] + 2) >> 2;
- predictor[2 * stride + 3] =
- predictor[3 * stride + 2] = (ptr[5] + ptr[6] * 2 + ptr[7] + 2) >> 2;
- predictor[3 * stride + 3] = (ptr[6] + ptr[7] * 2 + ptr[7] + 2) >> 2;
+ dst[0 * dst_stride + 0] = (ptr[0] + ptr[1] * 2 + ptr[2] + 2) >> 2;
+ dst[0 * dst_stride + 1] =
+ dst[1 * dst_stride + 0] = (ptr[1] + ptr[2] * 2 + ptr[3] + 2) >> 2;
+ dst[0 * dst_stride + 2] =
+ dst[1 * dst_stride + 1] =
+ dst[2 * dst_stride + 0] = (ptr[2] + ptr[3] * 2 + ptr[4] + 2) >> 2;
+ dst[0 * dst_stride + 3] =
+ dst[1 * dst_stride + 2] =
+ dst[2 * dst_stride + 1] =
+ dst[3 * dst_stride + 0] = (ptr[3] + ptr[4] * 2 + ptr[5] + 2) >> 2;
+ dst[1 * dst_stride + 3] =
+ dst[2 * dst_stride + 2] =
+ dst[3 * dst_stride + 1] = (ptr[4] + ptr[5] * 2 + ptr[6] + 2) >> 2;
+ dst[2 * dst_stride + 3] =
+ dst[3 * dst_stride + 2] = (ptr[5] + ptr[6] * 2 + ptr[7] + 2) >> 2;
+ dst[3 * dst_stride + 3] = (ptr[6] + ptr[7] * 2 + ptr[7] + 2) >> 2;
}
break;
@@ -158,22 +155,22 @@
pp[7] = Above[2];
pp[8] = Above[3];
- predictor[3 * stride + 0] = (pp[0] + pp[1] * 2 + pp[2] + 2) >> 2;
- predictor[3 * stride + 1] =
- predictor[2 * stride + 0] = (pp[1] + pp[2] * 2 + pp[3] + 2) >> 2;
- predictor[3 * stride + 2] =
- predictor[2 * stride + 1] =
- predictor[1 * stride + 0] = (pp[2] + pp[3] * 2 + pp[4] + 2) >> 2;
- predictor[3 * stride + 3] =
- predictor[2 * stride + 2] =
- predictor[1 * stride + 1] =
- predictor[0 * stride + 0] = (pp[3] + pp[4] * 2 + pp[5] + 2) >> 2;
- predictor[2 * stride + 3] =
- predictor[1 * stride + 2] =
- predictor[0 * stride + 1] = (pp[4] + pp[5] * 2 + pp[6] + 2) >> 2;
- predictor[1 * stride + 3] =
- predictor[0 * stride + 2] = (pp[5] + pp[6] * 2 + pp[7] + 2) >> 2;
- predictor[0 * stride + 3] = (pp[6] + pp[7] * 2 + pp[8] + 2) >> 2;
+ dst[3 * dst_stride + 0] = (pp[0] + pp[1] * 2 + pp[2] + 2) >> 2;
+ dst[3 * dst_stride + 1] =
+ dst[2 * dst_stride + 0] = (pp[1] + pp[2] * 2 + pp[3] + 2) >> 2;
+ dst[3 * dst_stride + 2] =
+ dst[2 * dst_stride + 1] =
+ dst[1 * dst_stride + 0] = (pp[2] + pp[3] * 2 + pp[4] + 2) >> 2;
+ dst[3 * dst_stride + 3] =
+ dst[2 * dst_stride + 2] =
+ dst[1 * dst_stride + 1] =
+ dst[0 * dst_stride + 0] = (pp[3] + pp[4] * 2 + pp[5] + 2) >> 2;
+ dst[2 * dst_stride + 3] =
+ dst[1 * dst_stride + 2] =
+ dst[0 * dst_stride + 1] = (pp[4] + pp[5] * 2 + pp[6] + 2) >> 2;
+ dst[1 * dst_stride + 3] =
+ dst[0 * dst_stride + 2] = (pp[5] + pp[6] * 2 + pp[7] + 2) >> 2;
+ dst[0 * dst_stride + 3] = (pp[6] + pp[7] * 2 + pp[8] + 2) >> 2;
}
break;
@@ -193,22 +190,22 @@
pp[8] = Above[3];
- predictor[3 * stride + 0] = (pp[1] + pp[2] * 2 + pp[3] + 2) >> 2;
- predictor[2 * stride + 0] = (pp[2] + pp[3] * 2 + pp[4] + 2) >> 2;
- predictor[3 * stride + 1] =
- predictor[1 * stride + 0] = (pp[3] + pp[4] * 2 + pp[5] + 2) >> 2;
- predictor[2 * stride + 1] =
- predictor[0 * stride + 0] = (pp[4] + pp[5] + 1) >> 1;
- predictor[3 * stride + 2] =
- predictor[1 * stride + 1] = (pp[4] + pp[5] * 2 + pp[6] + 2) >> 2;
- predictor[2 * stride + 2] =
- predictor[0 * stride + 1] = (pp[5] + pp[6] + 1) >> 1;
- predictor[3 * stride + 3] =
- predictor[1 * stride + 2] = (pp[5] + pp[6] * 2 + pp[7] + 2) >> 2;
- predictor[2 * stride + 3] =
- predictor[0 * stride + 2] = (pp[6] + pp[7] + 1) >> 1;
- predictor[1 * stride + 3] = (pp[6] + pp[7] * 2 + pp[8] + 2) >> 2;
- predictor[0 * stride + 3] = (pp[7] + pp[8] + 1) >> 1;
+ dst[3 * dst_stride + 0] = (pp[1] + pp[2] * 2 + pp[3] + 2) >> 2;
+ dst[2 * dst_stride + 0] = (pp[2] + pp[3] * 2 + pp[4] + 2) >> 2;
+ dst[3 * dst_stride + 1] =
+ dst[1 * dst_stride + 0] = (pp[3] + pp[4] * 2 + pp[5] + 2) >> 2;
+ dst[2 * dst_stride + 1] =
+ dst[0 * dst_stride + 0] = (pp[4] + pp[5] + 1) >> 1;
+ dst[3 * dst_stride + 2] =
+ dst[1 * dst_stride + 1] = (pp[4] + pp[5] * 2 + pp[6] + 2) >> 2;
+ dst[2 * dst_stride + 2] =
+ dst[0 * dst_stride + 1] = (pp[5] + pp[6] + 1) >> 1;
+ dst[3 * dst_stride + 3] =
+ dst[1 * dst_stride + 2] = (pp[5] + pp[6] * 2 + pp[7] + 2) >> 2;
+ dst[2 * dst_stride + 3] =
+ dst[0 * dst_stride + 2] = (pp[6] + pp[7] + 1) >> 1;
+ dst[1 * dst_stride + 3] = (pp[6] + pp[7] * 2 + pp[8] + 2) >> 2;
+ dst[0 * dst_stride + 3] = (pp[7] + pp[8] + 1) >> 1;
}
break;
@@ -217,22 +214,22 @@
unsigned char *pp = Above;
- predictor[0 * stride + 0] = (pp[0] + pp[1] + 1) >> 1;
- predictor[1 * stride + 0] = (pp[0] + pp[1] * 2 + pp[2] + 2) >> 2;
- predictor[2 * stride + 0] =
- predictor[0 * stride + 1] = (pp[1] + pp[2] + 1) >> 1;
- predictor[1 * stride + 1] =
- predictor[3 * stride + 0] = (pp[1] + pp[2] * 2 + pp[3] + 2) >> 2;
- predictor[2 * stride + 1] =
- predictor[0 * stride + 2] = (pp[2] + pp[3] + 1) >> 1;
- predictor[3 * stride + 1] =
- predictor[1 * stride + 2] = (pp[2] + pp[3] * 2 + pp[4] + 2) >> 2;
- predictor[0 * stride + 3] =
- predictor[2 * stride + 2] = (pp[3] + pp[4] + 1) >> 1;
- predictor[1 * stride + 3] =
- predictor[3 * stride + 2] = (pp[3] + pp[4] * 2 + pp[5] + 2) >> 2;
- predictor[2 * stride + 3] = (pp[4] + pp[5] * 2 + pp[6] + 2) >> 2;
- predictor[3 * stride + 3] = (pp[5] + pp[6] * 2 + pp[7] + 2) >> 2;
+ dst[0 * dst_stride + 0] = (pp[0] + pp[1] + 1) >> 1;
+ dst[1 * dst_stride + 0] = (pp[0] + pp[1] * 2 + pp[2] + 2) >> 2;
+ dst[2 * dst_stride + 0] =
+ dst[0 * dst_stride + 1] = (pp[1] + pp[2] + 1) >> 1;
+ dst[1 * dst_stride + 1] =
+ dst[3 * dst_stride + 0] = (pp[1] + pp[2] * 2 + pp[3] + 2) >> 2;
+ dst[2 * dst_stride + 1] =
+ dst[0 * dst_stride + 2] = (pp[2] + pp[3] + 1) >> 1;
+ dst[3 * dst_stride + 1] =
+ dst[1 * dst_stride + 2] = (pp[2] + pp[3] * 2 + pp[4] + 2) >> 2;
+ dst[0 * dst_stride + 3] =
+ dst[2 * dst_stride + 2] = (pp[3] + pp[4] + 1) >> 1;
+ dst[1 * dst_stride + 3] =
+ dst[3 * dst_stride + 2] = (pp[3] + pp[4] * 2 + pp[5] + 2) >> 2;
+ dst[2 * dst_stride + 3] = (pp[4] + pp[5] * 2 + pp[6] + 2) >> 2;
+ dst[3 * dst_stride + 3] = (pp[5] + pp[6] * 2 + pp[7] + 2) >> 2;
}
break;
@@ -250,22 +247,22 @@
pp[8] = Above[3];
- predictor[3 * stride + 0] = (pp[0] + pp[1] + 1) >> 1;
- predictor[3 * stride + 1] = (pp[0] + pp[1] * 2 + pp[2] + 2) >> 2;
- predictor[2 * stride + 0] =
- predictor[3 * stride + 2] = (pp[1] + pp[2] + 1) >> 1;
- predictor[2 * stride + 1] =
- predictor[3 * stride + 3] = (pp[1] + pp[2] * 2 + pp[3] + 2) >> 2;
- predictor[2 * stride + 2] =
- predictor[1 * stride + 0] = (pp[2] + pp[3] + 1) >> 1;
- predictor[2 * stride + 3] =
- predictor[1 * stride + 1] = (pp[2] + pp[3] * 2 + pp[4] + 2) >> 2;
- predictor[1 * stride + 2] =
- predictor[0 * stride + 0] = (pp[3] + pp[4] + 1) >> 1;
- predictor[1 * stride + 3] =
- predictor[0 * stride + 1] = (pp[3] + pp[4] * 2 + pp[5] + 2) >> 2;
- predictor[0 * stride + 2] = (pp[4] + pp[5] * 2 + pp[6] + 2) >> 2;
- predictor[0 * stride + 3] = (pp[5] + pp[6] * 2 + pp[7] + 2) >> 2;
+ dst[3 * dst_stride + 0] = (pp[0] + pp[1] + 1) >> 1;
+ dst[3 * dst_stride + 1] = (pp[0] + pp[1] * 2 + pp[2] + 2) >> 2;
+ dst[2 * dst_stride + 0] =
+ dst[3 * dst_stride + 2] = (pp[1] + pp[2] + 1) >> 1;
+ dst[2 * dst_stride + 1] =
+ dst[3 * dst_stride + 3] = (pp[1] + pp[2] * 2 + pp[3] + 2) >> 2;
+ dst[2 * dst_stride + 2] =
+ dst[1 * dst_stride + 0] = (pp[2] + pp[3] + 1) >> 1;
+ dst[2 * dst_stride + 3] =
+ dst[1 * dst_stride + 1] = (pp[2] + pp[3] * 2 + pp[4] + 2) >> 2;
+ dst[1 * dst_stride + 2] =
+ dst[0 * dst_stride + 0] = (pp[3] + pp[4] + 1) >> 1;
+ dst[1 * dst_stride + 3] =
+ dst[0 * dst_stride + 1] = (pp[3] + pp[4] * 2 + pp[5] + 2) >> 2;
+ dst[0 * dst_stride + 2] = (pp[4] + pp[5] * 2 + pp[6] + 2) >> 2;
+ dst[0 * dst_stride + 3] = (pp[5] + pp[6] * 2 + pp[7] + 2) >> 2;
}
break;
@@ -273,22 +270,22 @@
case B_HU_PRED:
{
unsigned char *pp = Left;
- predictor[0 * stride + 0] = (pp[0] + pp[1] + 1) >> 1;
- predictor[0 * stride + 1] = (pp[0] + pp[1] * 2 + pp[2] + 2) >> 2;
- predictor[0 * stride + 2] =
- predictor[1 * stride + 0] = (pp[1] + pp[2] + 1) >> 1;
- predictor[0 * stride + 3] =
- predictor[1 * stride + 1] = (pp[1] + pp[2] * 2 + pp[3] + 2) >> 2;
- predictor[1 * stride + 2] =
- predictor[2 * stride + 0] = (pp[2] + pp[3] + 1) >> 1;
- predictor[1 * stride + 3] =
- predictor[2 * stride + 1] = (pp[2] + pp[3] * 2 + pp[3] + 2) >> 2;
- predictor[2 * stride + 2] =
- predictor[2 * stride + 3] =
- predictor[3 * stride + 0] =
- predictor[3 * stride + 1] =
- predictor[3 * stride + 2] =
- predictor[3 * stride + 3] = pp[3];
+ dst[0 * dst_stride + 0] = (pp[0] + pp[1] + 1) >> 1;
+ dst[0 * dst_stride + 1] = (pp[0] + pp[1] * 2 + pp[2] + 2) >> 2;
+ dst[0 * dst_stride + 2] =
+ dst[1 * dst_stride + 0] = (pp[1] + pp[2] + 1) >> 1;
+ dst[0 * dst_stride + 3] =
+ dst[1 * dst_stride + 1] = (pp[1] + pp[2] * 2 + pp[3] + 2) >> 2;
+ dst[1 * dst_stride + 2] =
+ dst[2 * dst_stride + 0] = (pp[2] + pp[3] + 1) >> 1;
+ dst[1 * dst_stride + 3] =
+ dst[2 * dst_stride + 1] = (pp[2] + pp[3] * 2 + pp[3] + 2) >> 2;
+ dst[2 * dst_stride + 2] =
+ dst[2 * dst_stride + 3] =
+ dst[3 * dst_stride + 0] =
+ dst[3 * dst_stride + 1] =
+ dst[3 * dst_stride + 2] =
+ dst[3 * dst_stride + 3] = pp[3];
}
break;
@@ -316,5 +313,3 @@
*dst_ptr1 = *src_ptr;
*dst_ptr2 = *src_ptr;
}
-
-
diff --git a/vp8/decoder/decodframe.c b/vp8/decoder/decodframe.c
index f44c7d3..e501b9e 100644
--- a/vp8/decoder/decodframe.c
+++ b/vp8/decoder/decodframe.c
@@ -211,7 +211,8 @@
int b_mode = xd->mode_info_context->bmi[i].as_mode;
RECON_INVOKE(RTCD_VTABLE(recon), intra4x4_predict)
- (b, b_mode, *(b->base_dst) + b->dst, b->dst_stride);
+ ( *(b->base_dst) + b->dst, b->dst_stride, b_mode,
+ *(b->base_dst) + b->dst, b->dst_stride );
if (xd->eobs[i] )
{
diff --git a/vp8/encoder/encodeintra.c b/vp8/encoder/encodeintra.c
index 1c07cbd..46bf650 100644
--- a/vp8/encoder/encodeintra.c
+++ b/vp8/encoder/encodeintra.c
@@ -64,7 +64,8 @@
BLOCK *be = &x->block[ib];
RECON_INVOKE(&rtcd->common->recon, intra4x4_predict)
- (b, b->bmi.as_mode, b->predictor, 16);
+ (*(b->base_dst) + b->dst, b->dst_stride,
+ b->bmi.as_mode, b->predictor, 16);
ENCODEMB_INVOKE(&rtcd->encodemb, subb)(be, b, 16);
diff --git a/vp8/encoder/pickinter.c b/vp8/encoder/pickinter.c
index 62e644d..4d87341 100644
--- a/vp8/encoder/pickinter.c
+++ b/vp8/encoder/pickinter.c
@@ -157,7 +157,8 @@
rate = mode_costs[mode];
RECON_INVOKE(&rtcd->common->recon, intra4x4_predict)
- (b, mode, b->predictor, 16);
+ (*(b->base_dst) + b->dst, b->dst_stride,
+ mode, b->predictor, 16);
distortion = get_prediction_error(be, b, &rtcd->variance);
this_rd = RDCOST(x->rdmult, x->rddiv, rate, distortion);
diff --git a/vp8/encoder/rdopt.c b/vp8/encoder/rdopt.c
index 7950960..f966b06 100644
--- a/vp8/encoder/rdopt.c
+++ b/vp8/encoder/rdopt.c
@@ -631,7 +631,8 @@
rate = bmode_costs[mode];
RECON_INVOKE(&cpi->rtcd.common->recon, intra4x4_predict)
- (b, mode, b->predictor, 16);
+ (*(b->base_dst) + b->dst, b->dst_stride,
+ mode, b->predictor, 16);
ENCODEMB_INVOKE(IF_RTCD(&cpi->rtcd.encodemb), subb)(be, b, 16);
x->vp8_short_fdct4x4(be->src_diff, be->coeff, 32);
x->quantize_b(be, b);
diff --git a/vp8/vp8_common.mk b/vp8/vp8_common.mk
index 85b8113..5c15a3e 100644
--- a/vp8/vp8_common.mk
+++ b/vp8/vp8_common.mk
@@ -126,6 +126,7 @@
VP8_COMMON_SRCS-$(HAVE_ARMV6) += common/arm/armv6/loopfilter_v6$(ASM)
VP8_COMMON_SRCS-$(HAVE_ARMV6) += common/arm/armv6/simpleloopfilter_v6$(ASM)
VP8_COMMON_SRCS-$(HAVE_ARMV6) += common/arm/armv6/sixtappredict8x4_v6$(ASM)
+VP8_COMMON_SRCS-$(HAVE_ARMV6) += common/arm/armv6/intra4x4_predict_v6$(ASM)
# common (neon)
VP8_COMMON_SRCS-$(HAVE_ARMV7) += common/arm/neon/bilinearpredict4x4_neon$(ASM)
