ffmpeg/libavcodec/x86/diracdsp.asm

348 lines
8.3 KiB
NASM

;******************************************************************************
;* Copyright (c) 2010 David Conrad
;*
;* This file is part of FFmpeg.
;*
;* FFmpeg is free software; you can redistribute it and/or
;* modify it under the terms of the GNU Lesser General Public
;* License as published by the Free Software Foundation; either
;* version 2.1 of the License, or (at your option) any later version.
;*
;* FFmpeg is distributed in the hope that it will be useful,
;* but WITHOUT ANY WARRANTY; without even the implied warranty of
;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
;* Lesser General Public License for more details.
;*
;* You should have received a copy of the GNU Lesser General Public
;* License along with FFmpeg; if not, write to the Free Software
;* 51, Inc., Foundation Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
;******************************************************************************
%include "libavutil/x86/x86util.asm"
SECTION_RODATA
pw_7: times 8 dw 7
convert_to_unsigned_10bit: times 4 dd 0x200
clip_10bit: times 8 dw 0x3ff
cextern pw_3
cextern pw_16
cextern pw_32
cextern pb_80
section .text
%macro UNPACK_ADD 6
mov%5 %1, %3
mov%6 m5, %4
mova m4, %1
mova %2, m5
punpcklbw %1, m7
punpcklbw m5, m7
punpckhbw m4, m7
punpckhbw %2, m7
paddw %1, m5
paddw %2, m4
%endmacro
%macro HPEL_FILTER 1
; dirac_hpel_filter_v_sse2(uint8_t *dst, uint8_t *src, int stride, int width);
cglobal dirac_hpel_filter_v_%1, 4,6,8, dst, src, stride, width, src0, stridex3
mov src0q, srcq
lea stridex3q, [3*strideq]
sub src0q, stridex3q
pxor m7, m7
.loop:
; 7*(src[0] + src[1])
UNPACK_ADD m0, m1, [srcq], [srcq + strideq], a,a
pmullw m0, [pw_7]
pmullw m1, [pw_7]
; 3*( ... + src[-2] + src[3])
UNPACK_ADD m2, m3, [src0q + strideq], [srcq + stridex3q], a,a
paddw m0, m2
paddw m1, m3
pmullw m0, [pw_3]
pmullw m1, [pw_3]
; ... - 7*(src[-1] + src[2])
UNPACK_ADD m2, m3, [src0q + strideq*2], [srcq + strideq*2], a,a
pmullw m2, [pw_7]
pmullw m3, [pw_7]
psubw m0, m2
psubw m1, m3
; ... - (src[-3] + src[4])
UNPACK_ADD m2, m3, [src0q], [srcq + strideq*4], a,a
psubw m0, m2
psubw m1, m3
paddw m0, [pw_16]
paddw m1, [pw_16]
psraw m0, 5
psraw m1, 5
packuswb m0, m1
mova [dstq], m0
add dstq, mmsize
add srcq, mmsize
add src0q, mmsize
sub widthd, mmsize
jg .loop
RET
; dirac_hpel_filter_h_sse2(uint8_t *dst, uint8_t *src, int width);
cglobal dirac_hpel_filter_h_%1, 3,3,8, dst, src, width
dec widthd
pxor m7, m7
and widthd, ~(mmsize-1)
.loop:
; 7*(src[0] + src[1])
UNPACK_ADD m0, m1, [srcq + widthq], [srcq + widthq + 1], u,u
pmullw m0, [pw_7]
pmullw m1, [pw_7]
; 3*( ... + src[-2] + src[3])
UNPACK_ADD m2, m3, [srcq + widthq - 2], [srcq + widthq + 3], u,u
paddw m0, m2
paddw m1, m3
pmullw m0, [pw_3]
pmullw m1, [pw_3]
; ... - 7*(src[-1] + src[2])
UNPACK_ADD m2, m3, [srcq + widthq - 1], [srcq + widthq + 2], u,u
pmullw m2, [pw_7]
pmullw m3, [pw_7]
psubw m0, m2
psubw m1, m3
; ... - (src[-3] + src[4])
UNPACK_ADD m2, m3, [srcq + widthq - 3], [srcq + widthq + 4], u,u
psubw m0, m2
psubw m1, m3
paddw m0, [pw_16]
paddw m1, [pw_16]
psraw m0, 5
psraw m1, 5
packuswb m0, m1
mova [dstq + widthq], m0
sub widthd, mmsize
jge .loop
RET
%endmacro
%macro PUT_RECT 1
; void put_rect_clamped(uint8_t *dst, int dst_stride, int16_t *src, int src_stride, int width, int height)
cglobal put_signed_rect_clamped_%1, 5,9,3, dst, dst_stride, src, src_stride, w, dst2, src2
mova m0, [pb_80]
add wd, (mmsize-1)
and wd, ~(mmsize-1)
%if ARCH_X86_64
movsxd dst_strideq, dst_strided
movsxd src_strideq, src_strided
mov r7d, r5m
mov r8d, wd
%define wspill r8d
%define hd r7d
%else
mov r4m, wd
%define wspill r4m
%define hd r5mp
%endif
.loopy:
lea src2q, [srcq+src_strideq]
lea dst2q, [dstq+dst_strideq]
.loopx:
sub wd, mmsize
mova m1, [srcq +2*wq]
mova m2, [src2q+2*wq]
packsswb m1, [srcq +2*wq+mmsize]
packsswb m2, [src2q+2*wq+mmsize]
paddb m1, m0
paddb m2, m0
mova [dstq +wq], m1
mova [dst2q+wq], m2
jg .loopx
lea srcq, [srcq+src_strideq*2]
lea dstq, [dstq+dst_strideq*2]
sub hd, 2
mov wd, wspill
jg .loopy
RET
%endm
%macro ADD_RECT 1
; void add_rect_clamped(uint8_t *dst, uint16_t *src, int stride, int16_t *idwt, int idwt_stride, int width, int height)
cglobal add_rect_clamped_%1, 7,9,3, dst, src, stride, idwt, idwt_stride, w, h
mova m0, [pw_32]
add wd, (mmsize-1)
and wd, ~(mmsize-1)
%if ARCH_X86_64
movsxd strideq, strided
movsxd idwt_strideq, idwt_strided
mov r8d, wd
%define wspill r8d
%else
mov r5m, wd
%define wspill r5m
%endif
.loop:
sub wd, mmsize
movu m1, [srcq +2*wq] ; FIXME: ensure alignment
paddw m1, m0
psraw m1, 6
movu m2, [srcq +2*wq+mmsize] ; FIXME: ensure alignment
paddw m2, m0
psraw m2, 6
paddw m1, [idwtq+2*wq]
paddw m2, [idwtq+2*wq+mmsize]
packuswb m1, m2
mova [dstq +wq], m1
jg .loop
lea srcq, [srcq + 2*strideq]
add dstq, strideq
lea idwtq, [idwtq+ 2*idwt_strideq]
sub hd, 1
mov wd, wspill
jg .loop
RET
%endm
%macro ADD_OBMC 2
; void add_obmc(uint16_t *dst, uint8_t *src, int stride, uint8_t *obmc_weight, int yblen)
cglobal add_dirac_obmc%1_%2, 6,6,5, dst, src, stride, obmc, yblen
pxor m4, m4
.loop:
%assign i 0
%rep %1 / mmsize
mova m0, [srcq+i]
mova m1, m0
punpcklbw m0, m4
punpckhbw m1, m4
mova m2, [obmcq+i]
mova m3, m2
punpcklbw m2, m4
punpckhbw m3, m4
pmullw m0, m2
pmullw m1, m3
movu m2, [dstq+2*i]
movu m3, [dstq+2*i+mmsize]
paddw m0, m2
paddw m1, m3
movu [dstq+2*i], m0
movu [dstq+2*i+mmsize], m1
%assign i i+mmsize
%endrep
lea srcq, [srcq+strideq]
lea dstq, [dstq+2*strideq]
add obmcq, 32
sub yblend, 1
jg .loop
RET
%endm
INIT_MMX
%if ARCH_X86_64 == 0
PUT_RECT mmx
ADD_RECT mmx
HPEL_FILTER mmx
ADD_OBMC 32, mmx
ADD_OBMC 16, mmx
%endif
ADD_OBMC 8, mmx
INIT_XMM
PUT_RECT sse2
ADD_RECT sse2
HPEL_FILTER sse2
ADD_OBMC 32, sse2
ADD_OBMC 16, sse2
INIT_XMM sse4
; void dequant_subband_32(uint8_t *src, uint8_t *dst, ptrdiff_t stride, const int qf, const int qs, int tot_v, int tot_h)
cglobal dequant_subband_32, 7, 7, 4, src, dst, stride, qf, qs, tot_v, tot_h
movd m2, qfd
movd m3, qsd
SPLATD m2
SPLATD m3
mov r4, tot_hq
mov r3, dstq
.loop_v:
mov tot_hq, r4
mov dstq, r3
.loop_h:
movu m0, [srcq]
pabsd m1, m0
pmulld m1, m2
paddd m1, m3
psrld m1, 2
psignd m1, m0
movu [dstq], m1
add srcq, mmsize
add dstq, mmsize
sub tot_hd, 4
jg .loop_h
add r3, strideq
dec tot_vd
jg .loop_v
RET
INIT_XMM sse4
; void put_signed_rect_clamped_10(uint8_t *dst, int dst_stride, const uint8_t *src, int src_stride, int width, int height)
%if ARCH_X86_64
cglobal put_signed_rect_clamped_10, 6, 8, 5, dst, dst_stride, src, src_stride, w, h, t1, t2
%else
cglobal put_signed_rect_clamped_10, 5, 7, 5, dst, dst_stride, src, src_stride, w, t1, t2
%define hd r5mp
%endif
shl wd, 2
add srcq, wq
neg wq
mov t2q, dstq
mov t1q, wq
pxor m2, m2
mova m3, [clip_10bit]
mova m4, [convert_to_unsigned_10bit]
.loop_h:
mov dstq, t2q
mov wq, t1q
.loop_w:
movu m0, [srcq+wq+0*mmsize]
movu m1, [srcq+wq+1*mmsize]
paddd m0, m4
paddd m1, m4
packusdw m0, m0, m1
CLIPW m0, m2, m3 ; packusdw saturates so it's fine
movu [dstq], m0
add dstq, 1*mmsize
add wq, 2*mmsize
jl .loop_w
add srcq, src_strideq
add t2q, dst_strideq
sub hd, 1
jg .loop_h
RET