Skip to content

GitLab

Commit 52490354 authored by Johann's avatar Johann Committed by Code Review
Browse files

Merge "neon fast quantizer updated"

parents eeb81173 33fa7c4e
Hide whitespace changes
Inline Side-by-side
build/make/ads2gas.pl
View file @ 52490354
......@@ -23,6 +23,9 @@ print "\t.equ DO1STROUNDING, 0\n";
while (<STDIN>)
{
# Load and store alignment
s/@/,:/g;
# Comment character
s/;/@/g;
......@@ -114,8 +117,8 @@ while (<STDIN>)
# put the colon at the end of the line in the macro
s/^([a-zA-Z_0-9\$]+)/$1:/ if !/EQU/;
# Strip ALIGN
s/\sALIGN/@ ALIGN/g;
# ALIGN directive
s/ALIGN/.balign/g;
# Strip ARM
s/\sARM/@ ARM/g;
......
build/make/ads2gas_apple.pl
View file @ 52490354
......@@ -41,6 +41,9 @@ sub trim($)
while (<STDIN>)
{
# Load and store alignment
s/@/,:/g;
# Comment character
s/;/@/g;
......@@ -137,8 +140,8 @@ while (<STDIN>)
# put the colon at the end of the line in the macro
s/^([a-zA-Z_0-9\$]+)/$1:/ if !/EQU/;
# Strip ALIGN
s/\sALIGN/@ ALIGN/g;
# ALIGN directive
s/ALIGN/.balign/g;
# Strip ARM
s/\sARM/@ ARM/g;
......
vp8/encoder/arm/arm_csystemdependent.c
View file @ 52490354
......@@ -121,12 +121,8 @@ void vp8_arch_arm_encoder_init(VP8_COMP *cpi)
cpi->rtcd.encodemb.submby = vp8_subtract_mby_neon;
cpi->rtcd.encodemb.submbuv = vp8_subtract_mbuv_neon;
/*cpi->rtcd.quantize.quantb = vp8_regular_quantize_b;
cpi->rtcd.quantize.fastquantb = vp8_fast_quantize_b_c;*/
/* The neon quantizer has not been updated to match the new exact
* quantizer introduced in commit e04e2935
*/
/*cpi->rtcd.quantize.fastquantb = vp8_fast_quantize_b_neon;*/
/*cpi->rtcd.quantize.quantb = vp8_regular_quantize_b;*/
cpi->rtcd.quantize.fastquantb = vp8_fast_quantize_b_neon;
}
#endif
......
vp8/encoder/arm/neon/fastquantizeb_neon.asm
View file @ 52490354
;
; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
; 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
......@@ -9,110 +9,122 @@
;
EXPORT |vp8_fast_quantize_b_neon_func|
EXPORT |vp8_fast_quantize_b_neon|
INCLUDE asm_enc_offsets.asm
ARM
REQUIRE8
PRESERVE8
AREA ||.text||, CODE, READONLY, ALIGN=2
AREA ||.text||, CODE, READONLY, ALIGN=4
; r0 short *coeff_ptr
; r1 short *zbin_ptr
; r2 short *qcoeff_ptr
; r3 short *dqcoeff_ptr
; stack short *dequant_ptr
; stack short *scan_mask
; stack short *round_ptr
; stack short *quant_ptr
; return int * eob
|vp8_fast_quantize_b_neon_func| PROC
vld1.16 {q0, q1}, [r0] ;load z
vld1.16 {q10, q11}, [r1] ;load zbin
;void vp8_fast_quantize_b_c(BLOCK *b, BLOCKD *d)
|vp8_fast_quantize_b_neon| PROC
vabs.s16 q4, q0 ;calculate x = abs(z)
vabs.s16 q5, q1
stmfd sp!, {r4-r7}
vcge.s16 q10, q4, q10 ;x>=zbin
vcge.s16 q11, q5, q11
ldr r3, [r0, #vp8_block_coeff]
ldr r4, [r0, #vp8_block_quant_fast]
ldr r5, [r0, #vp8_block_round]
;if x<zbin (q10 & q11 are all 0), go to zero_output
vorr.s16 q6, q10, q11
vorr.s16 d12, d12, d13
vmov r0, r1, d12
orr r0, r0, r1
cmp r0, #0
beq zero_output
vld1.16 {q0, q1}, [r3@128] ; load z
vorr.s16 q14, q0, q1 ; check if all zero (step 1)
ldr r6, [r1, #vp8_blockd_qcoeff]
ldr r7, [r1, #vp8_blockd_dqcoeff]
vorr.s16 d28, d28, d29 ; check if all zero (step 2)
ldr r0, [sp, #8] ;load round_ptr
ldr r12, [sp, #12] ;load quant_ptr
vabs.s16 q12, q0 ; calculate x = abs(z)
vabs.s16 q13, q1
;right shift 15 to get sign, all 0 if it is positive, all 1 if it is negative
vshr.s16 q2, q0, #15 ; sz
vshr.s16 q2, q0, #15 ; sz
vmov r2, r3, d28 ; check if all zero (step 3)
vshr.s16 q3, q1, #15
vld1.s16 {q6, q7}, [r0] ;load round_ptr [0-15]
vld1.s16 {q8, q9}, [r12] ;load quant_ptr [0-15]
vld1.s16 {q14, q15}, [r5@128]; load round_ptr [0-15]
vld1.s16 {q8, q9}, [r4@128] ; load quant_ptr [0-15]
vadd.s16 q4, q6 ;x + Round
vadd.s16 q5, q7
vadd.s16 q12, q14 ; x + Round
vadd.s16 q13, q15
ldr r0, [sp, #4] ;load rvsplus1_scan_order ptr
ldr r0, _inv_zig_zag_ ; load ptr of inverse zigzag table
vqdmulh.s16 q4, q8 ;y = ((Round + abs(z)) * Quant) >> 16
vqdmulh.s16 q5, q9
vqdmulh.s16 q12, q8 ; y = ((Round+abs(z)) * Quant) >> 16
vqdmulh.s16 q13, q9
vld1.16 {q0, q1}, [r0] ;load rvsplus1_scan_order
vceq.s16 q8, q8 ;set q8 to all 1
vld1.16 {q10, q11}, [r0@128]; load inverse scan order
vshr.s16 q4, #1 ;right shift 1 after vqdmulh
vshr.s16 q5, #1
vceq.s16 q8, q8 ; set q8 to all 1
;modify data to have its original sign
veor.s16 q4, q2 ; y^sz
veor.s16 q5, q3
ldr r4, [r1, #vp8_blockd_dequant]
vshr.s16 q12, #1 ; right shift 1 after vqdmulh
vshr.s16 q13, #1
ldr r12, [sp] ;load dequant_ptr
orr r2, r2, r3 ; check if all zero (step 4)
cmp r2, #0 ; check if all zero (step 5)
beq zero_output ; check if all zero (step 6)
;modify data to have its original sign
veor.s16 q12, q2 ; y^sz
veor.s16 q13, q3
vsub.s16 q4, q2 ; x1 = (y^sz) - sz = (y^sz) - (-1) (two's complement)
vsub.s16 q5, q3
vsub.s16 q12, q2 ; x1=(y^sz)-sz = (y^sz)-(-1) (2's complement)
vsub.s16 q13, q3
vand.s16 q4, q10 ;mask off x1 elements
vand.s16 q5, q11
vld1.s16 {q2, q3}, [r4@128] ; load dequant_ptr[i]
vld1.s16 {q6, q7}, [r12] ;load dequant_ptr[i]
vtst.16 q14, q12, q8 ; now find eob
vtst.16 q15, q13, q8 ; non-zero element is set to all 1
vtst.16 q14, q4, q8 ;now find eob
vtst.16 q15, q5, q8 ;non-zero element is set to all 1 in q4, q5
vst1.s16 {q12, q13}, [r6@128]; store: qcoeff = x1
vst1.s16 {q4, q5}, [r2] ;store: qcoeff = x1
vand q10, q10, q14 ; get all valid numbers from scan array
vand q11, q11, q15
vand q0, q0, q14 ;get all valid number from rvsplus1_scan_order array
vand q1, q1, q15
vmax.u16 q0, q0, q1 ;find maximum value in q0, q1
vmax.u16 q0, q10, q11 ; find maximum value in q0, q1
vmax.u16 d0, d0, d1
vmovl.u16 q0, d0
vmul.s16 q6, q4 ;x * Dequant
vmul.s16 q7, q5
vmul.s16 q2, q12 ; x * Dequant
vmul.s16 q3, q13
vmax.u32 d0, d0, d1
vpmax.u32 d0, d0, d0
vst1.s16 {q6, q7}, [r3] ;store dqcoeff = x * Dequant
vst1.s16 {q2, q3}, [r7@128] ; store dqcoeff = x * Dequant
vmov.32 r0, d0[0]
vmov.32 r0, d0[0] ; this instruction takes 1+13 cycles
; if we have vfp, we could use
; vstr s0, [r1, #vp8_blockd_eob]
str r0, [r1, #vp8_blockd_eob]
ldmfd sp!, {r4-r7}
bx lr
zero_output
vst1.s16 {q10, q11}, [r2] ; qcoeff = 0
vst1.s16 {q10, q11}, [r3] ; dqcoeff = 0
mov r0, #0
str r2, [r1, #vp8_blockd_eob]
vst1.s16 {q0, q1}, [r6@128] ; qcoeff = 0
vst1.s16 {q0, q1}, [r7@128] ; dqcoeff = 0
ldmfd sp!, {r4-r7}
bx lr
ENDP
; default inverse zigzag table is defined in vp8/common/entropy.c
_inv_zig_zag_
DCD inv_zig_zag
ALIGN 16 ; enable use of @128 bit aligned loads
inv_zig_zag
DCW 0x0001, 0x0002, 0x0006, 0x0007
DCW 0x0003, 0x0005, 0x0008, 0x000d
DCW 0x0004, 0x0009, 0x000c, 0x000e
DCW 0x000a, 0x000b, 0x000f, 0x0010
END
vp8/encoder/arm/quantize_arm.c deleted 100644 → 0
View file @ eeb81173
/*
* Copyright (c) 2010 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.
*/
#include <math.h>
#include "vpx_mem/vpx_mem.h"
#include "vp8/encoder/quantize.h"
#include "vp8/common/entropy.h"
DECLARE_ALIGNED(16, const short, vp8_rvsplus1_default_zig_zag1d[16]) =
{
1, 2, 6, 7,
3, 5, 8, 13,
4, 9, 12, 14,
10, 11, 15, 16,
};
extern int vp8_fast_quantize_b_neon_func(short *coeff_ptr, short *zbin_ptr, short *qcoeff_ptr, short *dqcoeff_ptr, short *dequant_ptr, const short *scan_mask, short *round_ptr, short *quant_ptr);
void vp8_fast_quantize_b_neon(BLOCK *b, BLOCKD *d)
{
d->eob = vp8_fast_quantize_b_neon_func(b->coeff, b->zbin, d->qcoeff, d->dqcoeff, d->dequant, vp8_rvsplus1_default_zig_zag1d, b->round, b->quant_fast);
}
/*
//neon code is written according to the following rewritten c code
void vp8_fast_quantize_b_neon(BLOCK *b,BLOCKD *d)
{
int i, rc, eob;
int zbin;
int x, x1, y, z, sz;
short *coeff_ptr = &b->Coeff[0];
short *zbin_ptr = &b->Zbin[0][0];
short *round_ptr = &b->Round[0][0];
short *quant_ptr = &b->Quant[0][0];
short *qcoeff_ptr = d->qcoeff;
short *dqcoeff_ptr= d->dqcoeff;
short *dequant_ptr= &d->Dequant[0][0];
eob = 0;
for(i=0;i<16;i++)
{
z = coeff_ptr[i];
zbin = zbin_ptr[i] ;
x = abs(z); // x = abs(z)
if(x>=zbin)
{
sz = (z>>31); // sign of z
y = ((x+round_ptr[i])*quant_ptr[i])>>16; // quantize (x)
x1 = (y^sz) - sz; // get the sign back
qcoeff_ptr[i] = x1; // write to destination
dqcoeff_ptr[i] = x1 * dequant_ptr[i]; // dequantized value
if(y)
{
if(eob<vp8_rvsplus1_default_zig_zag1d[i])
eob=(int)vp8_rvsplus1_default_zig_zag1d[i]; // last nonzero coeffs
}
}else
{
qcoeff_ptr[i] = 0; // write to destination
dqcoeff_ptr[i] = 0; // dequantized value
}
}
d->eob = eob;
}
*/
vp8/encoder/arm/quantize_arm.h
View file @ 52490354
......@@ -23,14 +23,13 @@ extern prototype_quantize_block(vp8_fast_quantize_b_armv6);
#if HAVE_ARMV7
extern prototype_quantize_block(vp8_fast_quantize_b_neon);
/* The neon quantizer has not been updated to match the new exact
* quantizer introduced in commit e04e2935
*/
//#undef vp8_quantize_fastquantb
//#define vp8_quantize_fastquantb vp8_fast_quantize_b_neon
#undef vp8_quantize_fastquantb
#define vp8_quantize_fastquantb vp8_fast_quantize_b_neon
#endif
#endif /* HAVE_ARMV7 */
#endif
vp8/encoder/quantize.h
View file @ 52490354
......@@ -17,6 +17,7 @@
#define prototype_quantize_block(sym) \
void (sym)(BLOCK *b,BLOCKD *d)
#if ARCH_X86 || ARCH_X86_64
#include "x86/quantize_x86.h"
#endif
......@@ -41,6 +42,7 @@ typedef struct
prototype_quantize_block(*fastquantb);
} vp8_quantize_rtcd_vtable_t;
#if CONFIG_RUNTIME_CPU_DETECT
#define QUANTIZE_INVOKE(ctx,fn) (ctx)->fn
#else
......
vp8/vp8cx_arm.mk
View file @ 52490354
......@@ -15,7 +15,6 @@
# encoder
VP8_CX_SRCS-$(ARCH_ARM) += encoder/arm/arm_csystemdependent.c
VP8_CX_SRCS-$(HAVE_ARMV7) += encoder/arm/quantize_arm.c
VP8_CX_SRCS-$(HAVE_ARMV7) += encoder/arm/picklpf_arm.c
VP8_CX_SRCS-$(HAVE_ARMV6) += encoder/arm/dct_arm.c
VP8_CX_SRCS-$(HAVE_ARMV6) += encoder/arm/variance_arm.c
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment