summaryrefslogtreecommitdiff
path: root/vp8/decoder/arm
diff options
context:
space:
mode:
Diffstat (limited to 'vp8/decoder/arm')
-rw-r--r--vp8/decoder/arm/armv6/dequant_dc_idct_v6.asm218
-rw-r--r--vp8/decoder/arm/armv6/dequant_idct_v6.asm196
-rw-r--r--vp8/decoder/arm/armv6/dequantdcidct_v6.asm203
-rw-r--r--vp8/decoder/arm/armv6/dequantidct_v6.asm184
-rw-r--r--vp8/decoder/arm/dequantize_arm.h18
-rw-r--r--vp8/decoder/arm/neon/dequant_dc_idct_neon.asm (renamed from vp8/decoder/arm/neon/dequantdcidct_neon.asm)72
-rw-r--r--vp8/decoder/arm/neon/dequant_idct_neon.asm (renamed from vp8/decoder/arm/neon/dequantidct_neon.asm)67
7 files changed, 503 insertions, 455 deletions
diff --git a/vp8/decoder/arm/armv6/dequant_dc_idct_v6.asm b/vp8/decoder/arm/armv6/dequant_dc_idct_v6.asm
new file mode 100644
index 000000000..886873c70
--- /dev/null
+++ b/vp8/decoder/arm/armv6/dequant_dc_idct_v6.asm
@@ -0,0 +1,218 @@
+;
+; Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license and patent
+; grant that can be found in the LICENSE file in the root of the source
+; tree. All contributing project authors may be found in the AUTHORS
+; file in the root of the source tree.
+;
+
+
+ EXPORT |vp8_dequant_dc_idct_add_v6|
+
+ AREA |.text|, CODE, READONLY
+
+;void vp8_dequant_dc_idct_v6(short *input, short *dq, unsigned char *pred,
+; unsigned char *dest, int pitch, int stride, int Dc)
+; r0 = input
+; r1 = dq
+; r2 = pred
+; r3 = dest
+; sp + 36 = pitch ; +4 = 40
+; sp + 40 = stride ; +4 = 44
+; sp + 44 = Dc ; +4 = 48
+
+
+|vp8_dequant_dc_idct_add_v6| PROC
+ stmdb sp!, {r4-r11, lr}
+
+ ldr r6, [sp, #44]
+
+ ldr r4, [r0] ;input
+ ldr r5, [r1], #4 ;dq
+
+ sub sp, sp, #4
+ str r3, [sp]
+
+ smultt r7, r4, r5
+
+ ldr r4, [r0, #4] ;input
+ ldr r5, [r1], #4 ;dq
+
+ strh r6, [r0], #2
+ strh r7, [r0], #2
+
+ smulbb r6, r4, r5
+ smultt r7, r4, r5
+
+ ldr r4, [r0, #4] ;input
+ ldr r5, [r1], #4 ;dq
+
+ strh r6, [r0], #2
+ strh r7, [r0], #2
+
+ mov r12, #3
+
+vp8_dequant_dc_add_loop
+ smulbb r6, r4, r5
+ smultt r7, r4, r5
+
+ ldr r4, [r0, #4] ;input
+ ldr r5, [r1], #4 ;dq
+
+ strh r6, [r0], #2
+ strh r7, [r0], #2
+
+ smulbb r6, r4, r5
+ smultt r7, r4, r5
+
+ subs r12, r12, #1
+
+ ldrne r4, [r0, #4]
+ ldrne r5, [r1], #4
+
+ strh r6, [r0], #2
+ strh r7, [r0], #2
+
+ bne vp8_dequant_dc_add_loop
+
+ sub r0, r0, #32
+ mov r1, r0
+
+; short_idct4x4llm_v6_dual
+ ldr r3, cospi8sqrt2minus1
+ ldr r4, sinpi8sqrt2
+ ldr r6, [r0, #8]
+ mov r5, #2
+vp8_dequant_dc_idct_loop1_v6
+ ldr r12, [r0, #24]
+ ldr r14, [r0, #16]
+ smulwt r9, r3, r6
+ smulwb r7, r3, r6
+ smulwt r10, r4, r6
+ smulwb r8, r4, r6
+ pkhbt r7, r7, r9, lsl #16
+ smulwt r11, r3, r12
+ pkhbt r8, r8, r10, lsl #16
+ uadd16 r6, r6, r7
+ smulwt r7, r4, r12
+ smulwb r9, r3, r12
+ smulwb r10, r4, r12
+ subs r5, r5, #1
+ pkhbt r9, r9, r11, lsl #16
+ ldr r11, [r0], #4
+ pkhbt r10, r10, r7, lsl #16
+ uadd16 r7, r12, r9
+ usub16 r7, r8, r7
+ uadd16 r6, r6, r10
+ uadd16 r10, r11, r14
+ usub16 r8, r11, r14
+ uadd16 r9, r10, r6
+ usub16 r10, r10, r6
+ uadd16 r6, r8, r7
+ usub16 r7, r8, r7
+ str r6, [r1, #8]
+ ldrne r6, [r0, #8]
+ str r7, [r1, #16]
+ str r10, [r1, #24]
+ str r9, [r1], #4
+ bne vp8_dequant_dc_idct_loop1_v6
+
+ mov r5, #2
+ sub r0, r1, #8
+vp8_dequant_dc_idct_loop2_v6
+ ldr r6, [r0], #4
+ ldr r7, [r0], #4
+ ldr r8, [r0], #4
+ ldr r9, [r0], #4
+ smulwt r1, r3, r6
+ smulwt r12, r4, r6
+ smulwt lr, r3, r8
+ smulwt r10, r4, r8
+ pkhbt r11, r8, r6, lsl #16
+ pkhbt r1, lr, r1, lsl #16
+ pkhbt r12, r10, r12, lsl #16
+ pkhtb r6, r6, r8, asr #16
+ uadd16 r6, r1, r6
+ pkhbt lr, r9, r7, lsl #16
+ uadd16 r10, r11, lr
+ usub16 lr, r11, lr
+ pkhtb r8, r7, r9, asr #16
+ subs r5, r5, #1
+ smulwt r1, r3, r8
+ smulwb r7, r3, r8
+ smulwt r11, r4, r8
+ smulwb r9, r4, r8
+ pkhbt r1, r7, r1, lsl #16
+ uadd16 r8, r1, r8
+ pkhbt r11, r9, r11, lsl #16
+ usub16 r1, r12, r8
+ uadd16 r8, r11, r6
+ ldr r9, c0x00040004
+ ldr r12, [sp, #40]
+ uadd16 r6, r10, r8
+ usub16 r7, r10, r8
+ uadd16 r7, r7, r9
+ uadd16 r6, r6, r9
+ uadd16 r10, r14, r1
+ usub16 r1, r14, r1
+ uadd16 r10, r10, r9
+ uadd16 r1, r1, r9
+ ldr r11, [r2], r12
+ mov r8, r7, asr #3
+ pkhtb r9, r8, r10, asr #19
+ mov r8, r1, asr #3
+ pkhtb r8, r8, r6, asr #19
+ uxtb16 lr, r11, ror #8
+ qadd16 r9, r9, lr
+ uxtb16 lr, r11
+ qadd16 r8, r8, lr
+ usat16 r9, #8, r9
+ usat16 r8, #8, r8
+ orr r9, r8, r9, lsl #8
+ ldr r11, [r2], r12
+ ldr lr, [sp]
+ ldr r12, [sp, #44]
+ mov r7, r7, lsl #16
+ mov r1, r1, lsl #16
+ mov r10, r10, lsl #16
+ mov r6, r6, lsl #16
+ mov r7, r7, asr #3
+ pkhtb r7, r7, r10, asr #19
+ mov r1, r1, asr #3
+ pkhtb r1, r1, r6, asr #19
+ uxtb16 r8, r11, ror #8
+ qadd16 r7, r7, r8
+ uxtb16 r8, r11
+ qadd16 r1, r1, r8
+ usat16 r7, #8, r7
+ usat16 r1, #8, r1
+ orr r1, r1, r7, lsl #8
+ str r9, [lr], r12
+ str r1, [lr], r12
+ str lr, [sp]
+ bne vp8_dequant_dc_idct_loop2_v6
+
+; vpx_memset
+ sub r0, r0, #32
+ add sp, sp, #4
+
+ mov r12, #0
+ str r12, [r0]
+ str r12, [r0, #4]
+ str r12, [r0, #8]
+ str r12, [r0, #12]
+ str r12, [r0, #16]
+ str r12, [r0, #20]
+ str r12, [r0, #24]
+ str r12, [r0, #28]
+
+ ldmia sp!, {r4 - r11, pc}
+ ENDP ; |vp8_dequant_dc_idct_add_v6|
+
+; Constant Pool
+cospi8sqrt2minus1 DCD 0x00004E7B
+sinpi8sqrt2 DCD 0x00008A8C
+c0x00040004 DCD 0x00040004
+
+ END
diff --git a/vp8/decoder/arm/armv6/dequant_idct_v6.asm b/vp8/decoder/arm/armv6/dequant_idct_v6.asm
new file mode 100644
index 000000000..c13b51299
--- /dev/null
+++ b/vp8/decoder/arm/armv6/dequant_idct_v6.asm
@@ -0,0 +1,196 @@
+;
+; Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license and patent
+; grant that can be found in the LICENSE file in the root of the source
+; tree. All contributing project authors may be found in the AUTHORS
+; file in the root of the source tree.
+;
+
+ EXPORT |vp8_dequant_idct_add_v6|
+
+ AREA |.text|, CODE, READONLY
+;void vp8_dequant_idct_v6(short *input, short *dq, unsigned char *pred,
+; unsigned char *dest, int pitch, int stride)
+; r0 = input
+; r1 = dq
+; r2 = pred
+; r3 = dest
+; sp + 36 = pitch ; +4 = 40
+; sp + 40 = stride ; +4 = 44
+
+
+|vp8_dequant_idct_add_v6| PROC
+ stmdb sp!, {r4-r11, lr}
+
+ ldr r4, [r0] ;input
+ ldr r5, [r1], #4 ;dq
+
+ sub sp, sp, #4
+ str r3, [sp]
+
+ mov r12, #4
+
+vp8_dequant_add_loop
+ smulbb r6, r4, r5
+ smultt r7, r4, r5
+
+ ldr r4, [r0, #4] ;input
+ ldr r5, [r1], #4 ;dq
+
+ strh r6, [r0], #2
+ strh r7, [r0], #2
+
+ smulbb r6, r4, r5
+ smultt r7, r4, r5
+
+ subs r12, r12, #1
+
+ ldrne r4, [r0, #4]
+ ldrne r5, [r1], #4
+
+ strh r6, [r0], #2
+ strh r7, [r0], #2
+
+ bne vp8_dequant_add_loop
+
+ sub r0, r0, #32
+ mov r1, r0
+
+; short_idct4x4llm_v6_dual
+ ldr r3, cospi8sqrt2minus1
+ ldr r4, sinpi8sqrt2
+ ldr r6, [r0, #8]
+ mov r5, #2
+vp8_dequant_idct_loop1_v6
+ ldr r12, [r0, #24]
+ ldr r14, [r0, #16]
+ smulwt r9, r3, r6
+ smulwb r7, r3, r6
+ smulwt r10, r4, r6
+ smulwb r8, r4, r6
+ pkhbt r7, r7, r9, lsl #16
+ smulwt r11, r3, r12
+ pkhbt r8, r8, r10, lsl #16
+ uadd16 r6, r6, r7
+ smulwt r7, r4, r12
+ smulwb r9, r3, r12
+ smulwb r10, r4, r12
+ subs r5, r5, #1
+ pkhbt r9, r9, r11, lsl #16
+ ldr r11, [r0], #4
+ pkhbt r10, r10, r7, lsl #16
+ uadd16 r7, r12, r9
+ usub16 r7, r8, r7
+ uadd16 r6, r6, r10
+ uadd16 r10, r11, r14
+ usub16 r8, r11, r14
+ uadd16 r9, r10, r6
+ usub16 r10, r10, r6
+ uadd16 r6, r8, r7
+ usub16 r7, r8, r7
+ str r6, [r1, #8]
+ ldrne r6, [r0, #8]
+ str r7, [r1, #16]
+ str r10, [r1, #24]
+ str r9, [r1], #4
+ bne vp8_dequant_idct_loop1_v6
+
+ mov r5, #2
+ sub r0, r1, #8
+vp8_dequant_idct_loop2_v6
+ ldr r6, [r0], #4
+ ldr r7, [r0], #4
+ ldr r8, [r0], #4
+ ldr r9, [r0], #4
+ smulwt r1, r3, r6
+ smulwt r12, r4, r6
+ smulwt lr, r3, r8
+ smulwt r10, r4, r8
+ pkhbt r11, r8, r6, lsl #16
+ pkhbt r1, lr, r1, lsl #16
+ pkhbt r12, r10, r12, lsl #16
+ pkhtb r6, r6, r8, asr #16
+ uadd16 r6, r1, r6
+ pkhbt lr, r9, r7, lsl #16
+ uadd16 r10, r11, lr
+ usub16 lr, r11, lr
+ pkhtb r8, r7, r9, asr #16
+ subs r5, r5, #1
+ smulwt r1, r3, r8
+ smulwb r7, r3, r8
+ smulwt r11, r4, r8
+ smulwb r9, r4, r8
+ pkhbt r1, r7, r1, lsl #16
+ uadd16 r8, r1, r8
+ pkhbt r11, r9, r11, lsl #16
+ usub16 r1, r12, r8
+ uadd16 r8, r11, r6
+ ldr r9, c0x00040004
+ ldr r12, [sp, #40]
+ uadd16 r6, r10, r8
+ usub16 r7, r10, r8
+ uadd16 r7, r7, r9
+ uadd16 r6, r6, r9
+ uadd16 r10, r14, r1
+ usub16 r1, r14, r1
+ uadd16 r10, r10, r9
+ uadd16 r1, r1, r9
+ ldr r11, [r2], r12
+ mov r8, r7, asr #3
+ pkhtb r9, r8, r10, asr #19
+ mov r8, r1, asr #3
+ pkhtb r8, r8, r6, asr #19
+ uxtb16 lr, r11, ror #8
+ qadd16 r9, r9, lr
+ uxtb16 lr, r11
+ qadd16 r8, r8, lr
+ usat16 r9, #8, r9
+ usat16 r8, #8, r8
+ orr r9, r8, r9, lsl #8
+ ldr r11, [r2], r12
+ ldr lr, [sp]
+ ldr r12, [sp, #44]
+ mov r7, r7, lsl #16
+ mov r1, r1, lsl #16
+ mov r10, r10, lsl #16
+ mov r6, r6, lsl #16
+ mov r7, r7, asr #3
+ pkhtb r7, r7, r10, asr #19
+ mov r1, r1, asr #3
+ pkhtb r1, r1, r6, asr #19
+ uxtb16 r8, r11, ror #8
+ qadd16 r7, r7, r8
+ uxtb16 r8, r11
+ qadd16 r1, r1, r8
+ usat16 r7, #8, r7
+ usat16 r1, #8, r1
+ orr r1, r1, r7, lsl #8
+ str r9, [lr], r12
+ str r1, [lr], r12
+ str lr, [sp]
+ bne vp8_dequant_idct_loop2_v6
+
+; vpx_memset
+ sub r0, r0, #32
+ add sp, sp, #4
+
+ mov r12, #0
+ str r12, [r0]
+ str r12, [r0, #4]
+ str r12, [r0, #8]
+ str r12, [r0, #12]
+ str r12, [r0, #16]
+ str r12, [r0, #20]
+ str r12, [r0, #24]
+ str r12, [r0, #28]
+
+ ldmia sp!, {r4 - r11, pc}
+ ENDP ; |vp8_dequant_idct_add_v6|
+
+; Constant Pool
+cospi8sqrt2minus1 DCD 0x00004E7B
+sinpi8sqrt2 DCD 0x00008A8C
+c0x00040004 DCD 0x00040004
+
+ END
diff --git a/vp8/decoder/arm/armv6/dequantdcidct_v6.asm b/vp8/decoder/arm/armv6/dequantdcidct_v6.asm
deleted file mode 100644
index 025287223..000000000
--- a/vp8/decoder/arm/armv6/dequantdcidct_v6.asm
+++ /dev/null
@@ -1,203 +0,0 @@
-;
-; Copyright (c) 2010 The VP8 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_dequant_dc_idct_v6|
- ; ARM
- ; REQUIRE8
- ; PRESERVE8
-
- AREA |.text|, CODE, READONLY ; name this block of code
-;void vp8_dequant_dc_idct_v6(short *input, short *dq, short *output, int pitch,int Dc)
-|vp8_dequant_dc_idct_v6| PROC
- stmdb sp!, {r4-r11, lr}
-
- ldr r6, [sp, #36] ;load Dc
-
- ldr r4, [r0] ;input
- ldr r5, [r1], #4 ;dq
-
- sub sp, sp, #4
- str r0, [sp]
-
- smultt r7, r4, r5
-
- ldr r4, [r0, #4] ;input
- ldr r5, [r1], #4 ;dq
-
- strh r6, [r0], #2
- strh r7, [r0], #2
-
- smulbb r6, r4, r5
- smultt r7, r4, r5
-
- ldr r4, [r0, #4] ;input
- ldr r5, [r1], #4 ;dq
-
- strh r6, [r0], #2
- strh r7, [r0], #2
-
- mov r12, #3
-
-dequant_dc_idct_loop
- smulbb r6, r4, r5
- smultt r7, r4, r5
-
- ldr r4, [r0, #4] ;input
- ldr r5, [r1], #4 ;dq
-
- strh r6, [r0], #2
- strh r7, [r0], #2
-
- smulbb r6, r4, r5
- smultt r7, r4, r5
-
- subs r12, r12, #1
-
- ldrne r4, [r0, #4]
- ldrne r5, [r1], #4
-
- strh r6, [r0], #2
- strh r7, [r0], #2
-
- bne dequant_dc_idct_loop
-
- sub r0, r0, #32
- mov r1, r2
- mov r2, r3
-
-; short_idct4x4llm_v6_dual
-
- mov r3, #0x00004E00 ; cos
- orr r3, r3, #0x0000007B ; cospi8sqrt2minus1
- mov r4, #0x00008A00 ; sin
- orr r4, r4, #0x0000008C ; sinpi8sqrt2
- mov r5, #0x2 ; i=2 i
-loop1_dual_11
- ldr r6, [r0, #(4*2)] ; i5 | i4 5|4
- ldr r12, [r0, #(12*2)] ; i13 | i12 13|12
- ldr r14, [r0, #(8*2)] ; i9 | i8 9|8
-
- smulwt r9, r3, r6 ; (ip[5] * cospi8sqrt2minus1) >> 16 5c
- smulwb r7, r3, r6 ; (ip[4] * cospi8sqrt2minus1) >> 16 4c
- smulwt r10, r4, r6 ; (ip[5] * sinpi8sqrt2) >> 16 5s
- smulwb r8, r4, r6 ; (ip[4] * sinpi8sqrt2) >> 16 4s
- pkhbt r7, r7, r9, lsl #16 ; 5c | 4c
- smulwt r11, r3, r12 ; (ip[13] * cospi8sqrt2minus1) >> 16 13c
- pkhbt r8, r8, r10, lsl #16 ; 5s | 4s
- uadd16 r6, r6, r7 ; 5c+5 | 4c+4
- smulwt r7, r4, r12 ; (ip[13] * sinpi8sqrt2) >> 16 13s
- smulwb r9, r3, r12 ; (ip[12] * cospi8sqrt2minus1) >> 16 12c
- smulwb r10, r4, r12 ; (ip[12] * sinpi8sqrt2) >> 16 12s
- subs r5, r5, #0x1 ; i-- --
- pkhbt r9, r9, r11, lsl #16 ; 13c | 12c
- ldr r11, [r0], #0x4 ; i1 | i0 ++ 1|0
- pkhbt r10, r10, r7, lsl #16 ; 13s | 12s
- uadd16 r7, r12, r9 ; 13c+13 | 12c+12
- usub16 r7, r8, r7 ; c c
- uadd16 r6, r6, r10 ; d d
- uadd16 r10, r11, r14 ; a a
- usub16 r8, r11, r14 ; b b
- uadd16 r9, r10, r6 ; a+d a+d
- usub16 r10, r10, r6 ; a-d a-d
- uadd16 r6, r8, r7 ; b+c b+c
- usub16 r7, r8, r7 ; b-c b-c
- str r6, [r1, r2] ; o5 | o4
- add r6, r2, r2 ; pitch * 2 p2
- str r7, [r1, r6] ; o9 | o8
- add r6, r6, r2 ; pitch * 3 p3
- str r10, [r1, r6] ; o13 | o12
- str r9, [r1], #0x4 ; o1 | o0 ++
- bne loop1_dual_11 ;
- mov r5, #0x2 ; i=2 i
- sub r0, r1, #8 ; reset input/output i/o
-loop2_dual_22
- ldr r6, [r0, r2] ; i5 | i4 5|4
- ldr r1, [r0] ; i1 | i0 1|0
- ldr r12, [r0, #0x4] ; i3 | i2 3|2
- add r14, r2, #0x4 ; pitch + 2 p+2
- ldr r14, [r0, r14] ; i7 | i6 7|6
- smulwt r9, r3, r6 ; (ip[5] * cospi8sqrt2minus1) >> 16 5c
- smulwt r7, r3, r1 ; (ip[1] * cospi8sqrt2minus1) >> 16 1c
- smulwt r10, r4, r6 ; (ip[5] * sinpi8sqrt2) >> 16 5s
- smulwt r8, r4, r1 ; (ip[1] * sinpi8sqrt2) >> 16 1s
- pkhbt r11, r6, r1, lsl #16 ; i0 | i4 0|4
- pkhbt r7, r9, r7, lsl #16 ; 1c | 5c
- pkhbt r8, r10, r8, lsl #16 ; 1s | 5s = temp1 © tc1
- pkhtb r1, r1, r6, asr #16 ; i1 | i5 1|5
- uadd16 r1, r7, r1 ; 1c+1 | 5c+5 = temp2 (d) td2
- pkhbt r9, r14, r12, lsl #16 ; i2 | i6 2|6
- uadd16 r10, r11, r9 ; a a
- usub16 r9, r11, r9 ; b b
- pkhtb r6, r12, r14, asr #16 ; i3 | i7 3|7
- subs r5, r5, #0x1 ; i-- --
- smulwt r7, r3, r6 ; (ip[3] * cospi8sqrt2minus1) >> 16 3c
- smulwt r11, r4, r6 ; (ip[3] * sinpi8sqrt2) >> 16 3s
- smulwb r12, r3, r6 ; (ip[7] * cospi8sqrt2minus1) >> 16 7c
- smulwb r14, r4, r6 ; (ip[7] * sinpi8sqrt2) >> 16 7s
-
- pkhbt r7, r12, r7, lsl #16 ; 3c | 7c
- pkhbt r11, r14, r11, lsl #16 ; 3s | 7s = temp1 (d) td1
- uadd16 r6, r7, r6 ; 3c+3 | 7c+7 = temp2 (c) tc2
- usub16 r12, r8, r6 ; c (o1 | o5) c
- uadd16 r6, r11, r1 ; d (o3 | o7) d
- uadd16 r7, r10, r6 ; a+d a+d
- mov r8, #0x4 ; set up 4's 4
- orr r8, r8, #0x40000 ; 4|4
- usub16 r6, r10, r6 ; a-d a-d
- uadd16 r6, r6, r8 ; a-d+4 3|7
- uadd16 r7, r7, r8 ; a+d+4 0|4
- uadd16 r10, r9, r12 ; b+c b+c
- usub16 r1, r9, r12 ; b-c b-c
- uadd16 r10, r10, r8 ; b+c+4 1|5
- uadd16 r1, r1, r8 ; b-c+4 2|6
- mov r8, r10, asr #19 ; o1 >> 3
- strh r8, [r0, #2] ; o1
- mov r8, r1, asr #19 ; o2 >> 3
- strh r8, [r0, #4] ; o2
- mov r8, r6, asr #19 ; o3 >> 3
- strh r8, [r0, #6] ; o3
- mov r8, r7, asr #19 ; o0 >> 3
- strh r8, [r0], r2 ; o0 +p
- sxth r10, r10 ;
- mov r8, r10, asr #3 ; o5 >> 3
- strh r8, [r0, #2] ; o5
- sxth r1, r1 ;
- mov r8, r1, asr #3 ; o6 >> 3
- strh r8, [r0, #4] ; o6
- sxth r6, r6 ;
- mov r8, r6, asr #3 ; o7 >> 3
- strh r8, [r0, #6] ; o7
- sxth r7, r7 ;
- mov r8, r7, asr #3 ; o4 >> 3
- strh r8, [r0], r2 ; o4 +p
-;;;;; subs r5, r5, #0x1 ; i-- --
- bne loop2_dual_22 ;
-
-
-;vpx_memset
- ldr r0, [sp]
- add sp, sp, #4
-
- mov r12, #0
- str r12, [r0]
- str r12, [r0, #4]
- str r12, [r0, #8]
- str r12, [r0, #12]
- str r12, [r0, #16]
- str r12, [r0, #20]
- str r12, [r0, #24]
- str r12, [r0, #28]
-
- ldmia sp!, {r4 - r11, pc} ; replace vars, return restore
-
- ENDP ;|vp8_dequant_dc_idct_v68|
-
- END
diff --git a/vp8/decoder/arm/armv6/dequantidct_v6.asm b/vp8/decoder/arm/armv6/dequantidct_v6.asm
deleted file mode 100644
index 15e4c6814..000000000
--- a/vp8/decoder/arm/armv6/dequantidct_v6.asm
+++ /dev/null
@@ -1,184 +0,0 @@
-;
-; Copyright (c) 2010 The VP8 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_dequant_idct_v6|
- ; ARM
- ; REQUIRE8
- ; PRESERVE8
-
- AREA |.text|, CODE, READONLY ; name this block of code
-;void vp8_dequant_idct_v6(short *input, short *dq, short *output, int pitch)
-|vp8_dequant_idct_v6| PROC
- stmdb sp!, {r4-r11, lr}
-
- ldr r4, [r0] ;input
- ldr r5, [r1], #4 ;dq
-
- sub sp, sp, #4
- str r0, [sp]
-
- mov r12, #4
-
-dequant_idct_loop
- smulbb r6, r4, r5
- smultt r7, r4, r5
-
- ldr r4, [r0, #4] ;input
- ldr r5, [r1], #4 ;dq
-
- strh r6, [r0], #2
- strh r7, [r0], #2
-
- smulbb r6, r4, r5
- smultt r7, r4, r5
-
- subs r12, r12, #1
-
- ldrne r4, [r0, #4]
- ldrne r5, [r1], #4
-
- strh r6, [r0], #2
- strh r7, [r0], #2
-
- bne dequant_idct_loop
-
- sub r0, r0, #32
- mov r1, r2
- mov r2, r3
-
-; short_idct4x4llm_v6_dual
-
- mov r3, #0x00004E00 ; cos
- orr r3, r3, #0x0000007B ; cospi8sqrt2minus1
- mov r4, #0x00008A00 ; sin
- orr r4, r4, #0x0000008C ; sinpi8sqrt2
- mov r5, #0x2 ; i=2 i
-loop1_dual_1
- ldr r6, [r0, #(4*2)] ; i5 | i4 5|4
- ldr r12, [r0, #(12*2)] ; i13 | i12 13|12
- ldr r14, [r0, #(8*2)] ; i9 | i8 9|8
-
- smulwt r9, r3, r6 ; (ip[5] * cospi8sqrt2minus1) >> 16 5c
- smulwb r7, r3, r6 ; (ip[4] * cospi8sqrt2minus1) >> 16 4c
- smulwt r10, r4, r6 ; (ip[5] * sinpi8sqrt2) >> 16 5s
- smulwb r8, r4, r6 ; (ip[4] * sinpi8sqrt2) >> 16 4s
- pkhbt r7, r7, r9, lsl #16 ; 5c | 4c
- smulwt r11, r3, r12 ; (ip[13] * cospi8sqrt2minus1) >> 16 13c
- pkhbt r8, r8, r10, lsl #16 ; 5s | 4s
- uadd16 r6, r6, r7 ; 5c+5 | 4c+4
- smulwt r7, r4, r12 ; (ip[13] * sinpi8sqrt2) >> 16 13s
- smulwb r9, r3, r12 ; (ip[12] * cospi8sqrt2minus1) >> 16 12c
- smulwb r10, r4, r12 ; (ip[12] * sinpi8sqrt2) >> 16 12s
- subs r5, r5, #0x1 ; i-- --
- pkhbt r9, r9, r11, lsl #16 ; 13c | 12c
- ldr r11, [r0], #0x4 ; i1 | i0 ++ 1|0
- pkhbt r10, r10, r7, lsl #16 ; 13s | 12s
- uadd16 r7, r12, r9 ; 13c+13 | 12c+12
- usub16 r7, r8, r7 ; c c
- uadd16 r6, r6, r10 ; d d
- uadd16 r10, r11, r14 ; a a
- usub16 r8, r11, r14 ; b b
- uadd16 r9, r10, r6 ; a+d a+d
- usub16 r10, r10, r6 ; a-d a-d
- uadd16 r6, r8, r7 ; b+c b+c
- usub16 r7, r8, r7 ; b-c b-c
- str r6, [r1, r2] ; o5 | o4
- add r6, r2, r2 ; pitch * 2 p2
- str r7, [r1, r6] ; o9 | o8
- add r6, r6, r2 ; pitch * 3 p3
- str r10, [r1, r6] ; o13 | o12
- str r9, [r1], #0x4 ; o1 | o0 ++
- bne loop1_dual_1 ;
- mov r5, #0x2 ; i=2 i
- sub r0, r1, #8 ; reset input/output i/o
-loop2_dual_2
- ldr r6, [r0, r2] ; i5 | i4 5|4
- ldr r1, [r0] ; i1 | i0 1|0
- ldr r12, [r0, #0x4] ; i3 | i2 3|2
- add r14, r2, #0x4 ; pitch + 2 p+2
- ldr r14, [r0, r14] ; i7 | i6 7|6
- smulwt r9, r3, r6 ; (ip[5] * cospi8sqrt2minus1) >> 16 5c
- smulwt r7, r3, r1 ; (ip[1] * cospi8sqrt2minus1) >> 16 1c
- smulwt r10, r4, r6 ; (ip[5] * sinpi8sqrt2) >> 16 5s
- smulwt r8, r4, r1 ; (ip[1] * sinpi8sqrt2) >> 16 1s
- pkhbt r11, r6, r1, lsl #16 ; i0 | i4 0|4
- pkhbt r7, r9, r7, lsl #16 ; 1c | 5c
- pkhbt r8, r10, r8, lsl #16 ; 1s | 5s = temp1 © tc1
- pkhtb r1, r1, r6, asr #16 ; i1 | i5 1|5
- uadd16 r1, r7, r1 ; 1c+1 | 5c+5 = temp2 (d) td2
- pkhbt r9, r14, r12, lsl #16 ; i2 | i6 2|6
- uadd16 r10, r11, r9 ; a a
- usub16 r9, r11, r9 ; b b
- pkhtb r6, r12, r14, asr #16 ; i3 | i7 3|7
- subs r5, r5, #0x1 ; i-- --
- smulwt r7, r3, r6 ; (ip[3] * cospi8sqrt2minus1) >> 16 3c
- smulwt r11, r4, r6 ; (ip[3] * sinpi8sqrt2) >> 16 3s
- smulwb r12, r3, r6 ; (ip[7] * cospi8sqrt2minus1) >> 16 7c
- smulwb r14, r4, r6 ; (ip[7] * sinpi8sqrt2) >> 16 7s
-
- pkhbt r7, r12, r7, lsl #16 ; 3c | 7c
- pkhbt r11, r14, r11, lsl #16 ; 3s | 7s = temp1 (d) td1
- uadd16 r6, r7, r6 ; 3c+3 | 7c+7 = temp2 (c) tc2
- usub16 r12, r8, r6 ; c (o1 | o5) c
- uadd16 r6, r11, r1 ; d (o3 | o7) d
- uadd16 r7, r10, r6 ; a+d a+d
- mov r8, #0x4 ; set up 4's 4
- orr r8, r8, #0x40000 ; 4|4
- usub16 r6, r10, r6 ; a-d a-d
- uadd16 r6, r6, r8 ; a-d+4 3|7
- uadd16 r7, r7, r8 ; a+d+4 0|4
- uadd16 r10, r9, r12 ; b+c b+c
- usub16 r1, r9, r12 ; b-c b-c
- uadd16 r10, r10, r8 ; b+c+4 1|5
- uadd16 r1, r1, r8 ; b-c+4 2|6
- mov r8, r10, asr #19 ; o1 >> 3
- strh r8, [r0, #2] ; o1
- mov r8, r1, asr #19 ; o2 >> 3
- strh r8, [r0, #4] ; o2
- mov r8, r6, asr #19 ; o3 >> 3
- strh r8, [r0, #6] ; o3
- mov r8, r7, asr #19 ; o0 >> 3
- strh r8, [r0], r2 ; o0 +p
- sxth r10, r10 ;
- mov r8, r10, asr #3 ; o5 >> 3
- strh r8, [r0, #2] ; o5
- sxth r1, r1 ;
- mov r8, r1, asr #3 ; o6 >> 3
- strh r8, [r0, #4] ; o6
- sxth r6, r6 ;
- mov r8, r6, asr #3 ; o7 >> 3
- strh r8, [r0, #6] ; o7
- sxth r7, r7 ;
- mov r8, r7, asr #3 ; o4 >> 3
- strh r8, [r0], r2 ; o4 +p
-;;;;; subs r5, r5, #0x1 ; i-- --
- bne loop2_dual_2 ;
- ;
-
-;vpx_memset
- ldr r0, [sp]
- add sp, sp, #4
-
- mov r12, #0
- str r12, [r0]
- str r12, [r0, #4]
- str r12, [r0, #8]
- str r12, [r0, #12]
- str r12, [r0, #16]
- str r12, [r0, #20]
- str r12, [r0, #24]
- str r12, [r0, #28]
-
- ldmia sp!, {r4 - r11, pc} ; replace vars, return restore
-
- ENDP ;|vp8_dequant_idct_v6|
-
- END
diff --git a/vp8/decoder/arm/dequantize_arm.h b/vp8/decoder/arm/dequantize_arm.h
index 78af0195d..3a044f847 100644
--- a/vp8/decoder/arm/dequantize_arm.h
+++ b/vp8/decoder/arm/dequantize_arm.h
@@ -14,14 +14,32 @@
#if HAVE_ARMV6
extern prototype_dequant_block(vp8_dequantize_b_v6);
+extern prototype_dequant_idct_add(vp8_dequant_idct_add_v6);
+extern prototype_dequant_dc_idct_add(vp8_dequant_dc_idct_add_v6);
#undef vp8_dequant_block
#define vp8_dequant_block vp8_dequantize_b_v6
+#undef vp8_dequant_idct_add
+#define vp8_dequant_idct_add vp8_dequant_idct_add_v6
+
+#undef vp8_dequant_dc_idct_add
+#define vp8_dequant_dc_idct_add vp8_dequant_dc_idct_add_v6
+#endif
+
#if HAVE_ARMV7
extern prototype_dequant_block(vp8_dequantize_b_neon);
+extern prototype_dequant_idct_add(vp8_dequant_idct_add_neon);
+extern prototype_dequant_dc_idct_add(vp8_dequant_dc_idct_add_neon);
#undef vp8_dequant_block
#define vp8_dequant_block vp8_dequantize_b_neon
+#undef vp8_dequant_idct_add
+#define vp8_dequant_idct_add vp8_dequant_idct_add_neon
+
+#undef vp8_dequant_dc_idct_add
+#define vp8_dequant_dc_idct_add vp8_dequant_dc_idct_add_neon
+#endif
+
#endif
diff --git a/vp8/decoder/arm/neon/dequantdcidct_neon.asm b/vp8/decoder/arm/neon/dequant_dc_idct_neon.asm
index ae126f23f..ddb324068 100644
--- a/vp8/decoder/arm/neon/dequantdcidct_neon.asm
+++ b/vp8/decoder/arm/neon/dequant_dc_idct_neon.asm
@@ -9,31 +9,43 @@
;
- EXPORT |vp8_dequant_dc_idct_neon|
+ EXPORT |vp8_dequant_dc_idct_add_neon|
ARM
REQUIRE8
PRESERVE8
AREA ||.text||, CODE, READONLY, ALIGN=2
-;void vp8_dequant_dc_idct_c(short *input, short *dq, short *output, int pitch, int Dc);
+;void vp8_dequant_dc_idct_add_neon(short *input, short *dq, unsigned char *pred,
+; unsigned char *dest, int pitch, int stride,
+; int Dc);
; r0 short *input,
; r1 short *dq,
-; r2 short *output,
-; r3 int pitch,
-; (stack) int Dc
-|vp8_dequant_dc_idct_neon| PROC
+; r2 unsigned char *pred
+; r3 unsigned char *dest
+; sp int pitch
+; sp+4 int stride
+; sp+8 int Dc
+|vp8_dequant_dc_idct_add_neon| PROC
vld1.16 {q3, q4}, [r0]
vld1.16 {q5, q6}, [r1]
- ldr r1, [sp] ;load Dc from stack
+ ldr r1, [sp, #8] ;load Dc from stack
- ldr r12, _dcidct_coeff_
+ ldr r12, _CONSTANTS_
vmul.i16 q1, q3, q5 ;input for short_idct4x4llm_neon
vmul.i16 q2, q4, q6
vmov.16 d2[0], r1
+ ldr r1, [sp] ; pitch
+ vld1.32 {d14[0]}, [r2], r1
+ vld1.32 {d14[1]}, [r2], r1
+ vld1.32 {d15[0]}, [r2], r1
+ vld1.32 {d15[1]}, [r2]
+
+ ldr r1, [sp, #4] ; stride
+
;|short_idct4x4llm_neon| PROC
vld1.16 {d0}, [r12]
vswp d3, d4 ;q2(vp[4] vp[12])
@@ -47,14 +59,9 @@
vshr.s16 q3, q3, #1
vshr.s16 q4, q4, #1
- vqadd.s16 q3, q3, q2 ;modify since sinpi8sqrt2 > 65536/2 (negtive number)
+ vqadd.s16 q3, q3, q2
vqadd.s16 q4, q4, q2
- ;d6 - c1:temp1
- ;d7 - d1:temp2
- ;d8 - d1:temp1
- ;d9 - c1:temp2
-
vqsub.s16 d10, d6, d9 ;c1
vqadd.s16 d11, d7, d8 ;d1
@@ -83,7 +90,7 @@
vshr.s16 q3, q3, #1
vshr.s16 q4, q4, #1
- vqadd.s16 q3, q3, q2 ;modify since sinpi8sqrt2 > 65536/2 (negtive number)
+ vqadd.s16 q3, q3, q2
vqadd.s16 q4, q4, q2
vqsub.s16 d10, d6, d9 ;c1
@@ -101,34 +108,29 @@
vrshr.s16 d4, d4, #3
vrshr.s16 d5, d5, #3
- add r1, r2, r3
- add r12, r1, r3
- add r0, r12, r3
-
vtrn.32 d2, d4
vtrn.32 d3, d5
vtrn.16 d2, d3
vtrn.16 d4, d5
- vst1.16 {d2}, [r2]
- vst1.16 {d3}, [r1]
- vst1.16 {d4}, [r12]
- vst1.16 {d5}, [r0]
+ vaddw.u8 q1, q1, d14
+ vaddw.u8 q2, q2, d15
- bx lr
+ vqmovun.s16 d0, q1
+ vqmovun.s16 d1, q2
+
+ vst1.32 {d0[0]}, [r3], r1
+ vst1.32 {d0[1]}, [r3], r1
+ vst1.32 {d1[0]}, [r3], r1
+ vst1.32 {d1[1]}, [r3]
- ENDP
+ bx lr
-;-----------------
- AREA dcidct4x4_dat, DATA, READWRITE ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
-_dcidct_coeff_
- DCD dcidct_coeff
-dcidct_coeff
- DCD 0x4e7b4e7b, 0x8a8c8a8c
+ ENDP ; |vp8_dequant_dc_idct_add_neon|
-;20091, 20091, 35468, 35468
+; Constant Pool
+_CONSTANTS_ DCD cospi8sqrt2minus1
+cospi8sqrt2minus1 DCD 0x4e7b4e7b
+sinpi8sqrt2 DCD 0x8a8c8a8c
END
diff --git a/vp8/decoder/arm/neon/dequantidct_neon.asm b/vp8/decoder/arm/neon/dequant_idct_neon.asm
index e0888ed35..5c60dd690 100644
--- a/vp8/decoder/arm/neon/dequantidct_neon.asm
+++ b/vp8/decoder/arm/neon/dequant_idct_neon.asm
@@ -9,22 +9,33 @@
;
- EXPORT |vp8_dequant_idct_neon|
+ EXPORT |vp8_dequant_idct_add_neon|
ARM
REQUIRE8
PRESERVE8
AREA ||.text||, CODE, READONLY, ALIGN=2
-;void vp8_dequant_idct_c(short *input, short *dq, short *output, int pitch);
+;void vp8_dequant_idct_neon(short *input, short *dq, unsigned char *pred,
+; unsigned char *dest, int pitch, int stride)
; r0 short *input,
; r1 short *dq,
-; r2 short *output,
-; r3 int pitch,
-|vp8_dequant_idct_neon| PROC
+; r2 unsigned char *pred
+; r3 unsigned char *dest
+; sp int pitch
+; sp+4 int stride
+
+|vp8_dequant_idct_add_neon| PROC
vld1.16 {q3, q4}, [r0]
vld1.16 {q5, q6}, [r1]
+ ldr r1, [sp] ; pitch
+ vld1.32 {d14[0]}, [r2], r1
+ vld1.32 {d14[1]}, [r2], r1
+ vld1.32 {d15[0]}, [r2], r1
+ vld1.32 {d15[1]}, [r2]
+
+ ldr r1, [sp, #4] ; stride
- ldr r12, _didct_coeff_
+ ldr r12, _CONSTANTS_
vmul.i16 q1, q3, q5 ;input for short_idct4x4llm_neon
vmul.i16 q2, q4, q6
@@ -42,14 +53,9 @@
vshr.s16 q3, q3, #1
vshr.s16 q4, q4, #1
- vqadd.s16 q3, q3, q2 ;modify since sinpi8sqrt2 > 65536/2 (negtive number)
+ vqadd.s16 q3, q3, q2
vqadd.s16 q4, q4, q2
- ;d6 - c1:temp1
- ;d7 - d1:temp2
- ;d8 - d1:temp1
- ;d9 - c1:temp2
-
vqsub.s16 d10, d6, d9 ;c1
vqadd.s16 d11, d7, d8 ;d1
@@ -78,7 +84,7 @@
vshr.s16 q3, q3, #1
vshr.s16 q4, q4, #1
- vqadd.s16 q3, q3, q2 ;modify since sinpi8sqrt2 > 65536/2 (negtive number)
+ vqadd.s16 q3, q3, q2
vqadd.s16 q4, q4, q2
vqsub.s16 d10, d6, d9 ;c1
@@ -96,34 +102,29 @@
vrshr.s16 d4, d4, #3
vrshr.s16 d5, d5, #3
- add r1, r2, r3
- add r12, r1, r3
- add r0, r12, r3
-
vtrn.32 d2, d4
vtrn.32 d3, d5
vtrn.16 d2, d3
vtrn.16 d4, d5
- vst1.16 {d2}, [r2]
- vst1.16 {d3}, [r1]
- vst1.16 {d4}, [r12]
- vst1.16 {d5}, [r0]
+ vaddw.u8 q1, q1, d14
+ vaddw.u8 q2, q2, d15
- bx lr
+ vqmovun.s16 d0, q1
+ vqmovun.s16 d1, q2
+
+ vst1.32 {d0[0]}, [r3], r1
+ vst1.32 {d0[1]}, [r3], r1
+ vst1.32 {d1[0]}, [r3], r1
+ vst1.32 {d1[1]}, [r3]
- ENDP
+ bx lr
-;-----------------
- AREA didct4x4_dat, DATA, READWRITE ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
-_didct_coeff_
- DCD didct_coeff
-didct_coeff
- DCD 0x4e7b4e7b, 0x8a8c8a8c
+ ENDP ; |vp8_dequant_idct_add_neon|
-;20091, 20091, 35468, 35468
+; Constant Pool
+_CONSTANTS_ DCD cospi8sqrt2minus1
+cospi8sqrt2minus1 DCD 0x4e7b4e7b
+sinpi8sqrt2 DCD 0x8a8c8a8c
END
generated by cgit v1.2.3 (git 2.25.1) at 2024-10-12 05:28:52 +0000