Reviewed-by: Michael Niedermayer <michael@niedermayer.cc> Signed-off-by: Thomas Mundt <tmundt75@gmail.com> Signed-off-by: James Almer <jamrial@gmail.com>tags/n3.4
| @@ -25,9 +25,11 @@ | |||
| #ifndef AVFILTER_INTERLACE_H | |||
| #define AVFILTER_INTERLACE_H | |||
| #include "libavutil/bswap.h" | |||
| #include "libavutil/common.h" | |||
| #include "libavutil/imgutils.h" | |||
| #include "libavutil/opt.h" | |||
| #include "libavutil/pixdesc.h" | |||
| #include "avfilter.h" | |||
| #include "formats.h" | |||
| @@ -55,8 +57,9 @@ typedef struct InterlaceContext { | |||
| enum ScanMode scan; // top or bottom field first scanning | |||
| int lowpass; // enable or disable low pass filtering | |||
| AVFrame *cur, *next; // the two frames from which the new one is obtained | |||
| const AVPixFmtDescriptor *csp; | |||
| void (*lowpass_line)(uint8_t *dstp, ptrdiff_t linesize, const uint8_t *srcp, | |||
| ptrdiff_t mref, ptrdiff_t pref); | |||
| ptrdiff_t mref, ptrdiff_t pref, int clip_max); | |||
| } InterlaceContext; | |||
| void ff_interlace_init_x86(InterlaceContext *interlace); | |||
| @@ -27,7 +27,9 @@ | |||
| #ifndef AVFILTER_TINTERLACE_H | |||
| #define AVFILTER_TINTERLACE_H | |||
| #include "libavutil/bswap.h" | |||
| #include "libavutil/opt.h" | |||
| #include "libavutil/pixdesc.h" | |||
| #include "drawutils.h" | |||
| #include "avfilter.h" | |||
| @@ -60,8 +62,9 @@ typedef struct TInterlaceContext { | |||
| int black_linesize[4]; | |||
| FFDrawContext draw; | |||
| FFDrawColor color; | |||
| const AVPixFmtDescriptor *csp; | |||
| void (*lowpass_line)(uint8_t *dstp, ptrdiff_t width, const uint8_t *srcp, | |||
| ptrdiff_t mref, ptrdiff_t pref); | |||
| ptrdiff_t mref, ptrdiff_t pref, int clip_max); | |||
| } TInterlaceContext; | |||
| void ff_tinterlace_init_x86(TInterlaceContext *interlace); | |||
| @@ -61,8 +61,8 @@ static const AVOption interlace_options[] = { | |||
| AVFILTER_DEFINE_CLASS(interlace); | |||
| static void lowpass_line_c(uint8_t *dstp, ptrdiff_t linesize, | |||
| const uint8_t *srcp, | |||
| ptrdiff_t mref, ptrdiff_t pref) | |||
| const uint8_t *srcp, ptrdiff_t mref, | |||
| ptrdiff_t pref, int clip_max) | |||
| { | |||
| const uint8_t *srcp_above = srcp + mref; | |||
| const uint8_t *srcp_below = srcp + pref; | |||
| @@ -75,9 +75,28 @@ static void lowpass_line_c(uint8_t *dstp, ptrdiff_t linesize, | |||
| } | |||
| } | |||
| static void lowpass_line_c_16(uint8_t *dst8, ptrdiff_t linesize, | |||
| const uint8_t *src8, ptrdiff_t mref, | |||
| ptrdiff_t pref, int clip_max) | |||
| { | |||
| uint16_t *dstp = (uint16_t *)dst8; | |||
| const uint16_t *srcp = (const uint16_t *)src8; | |||
| const uint16_t *srcp_above = srcp + mref / 2; | |||
| const uint16_t *srcp_below = srcp + pref / 2; | |||
| int i, src_x; | |||
| for (i = 0; i < linesize; i++) { | |||
| // this calculation is an integer representation of | |||
| // '0.5 * current + 0.25 * above + 0.25 * below' | |||
| // '1 +' is for rounding. | |||
| src_x = av_le2ne16(srcp[i]) << 1; | |||
| dstp[i] = av_le2ne16((1 + src_x + av_le2ne16(srcp_above[i]) | |||
| + av_le2ne16(srcp_below[i])) >> 2); | |||
| } | |||
| } | |||
| static void lowpass_line_complex_c(uint8_t *dstp, ptrdiff_t linesize, | |||
| const uint8_t *srcp, | |||
| ptrdiff_t mref, ptrdiff_t pref) | |||
| const uint8_t *srcp, ptrdiff_t mref, | |||
| ptrdiff_t pref, int clip_max) | |||
| { | |||
| const uint8_t *srcp_above = srcp + mref; | |||
| const uint8_t *srcp_below = srcp + pref; | |||
| @@ -103,11 +122,51 @@ static void lowpass_line_complex_c(uint8_t *dstp, ptrdiff_t linesize, | |||
| } | |||
| } | |||
| static void lowpass_line_complex_c_16(uint8_t *dst8, ptrdiff_t linesize, | |||
| const uint8_t *src8, ptrdiff_t mref, | |||
| ptrdiff_t pref, int clip_max) | |||
| { | |||
| uint16_t *dstp = (uint16_t *)dst8; | |||
| const uint16_t *srcp = (const uint16_t *)src8; | |||
| const uint16_t *srcp_above = srcp + mref / 2; | |||
| const uint16_t *srcp_below = srcp + pref / 2; | |||
| const uint16_t *srcp_above2 = srcp + mref; | |||
| const uint16_t *srcp_below2 = srcp + pref; | |||
| int i, dst_le, src_le, src_x, src_ab; | |||
| for (i = 0; i < linesize; i++) { | |||
| // this calculation is an integer representation of | |||
| // '0.75 * current + 0.25 * above + 0.25 * below - 0.125 * above2 - 0.125 * below2' | |||
| // '4 +' is for rounding. | |||
| src_le = av_le2ne16(srcp[i]); | |||
| src_x = src_le << 1; | |||
| src_ab = av_le2ne16(srcp_above[i]) + av_le2ne16(srcp_below[i]); | |||
| dst_le = av_clip((4 + ((src_le + src_x + src_ab) << 1) | |||
| - av_le2ne16(srcp_above2[i]) | |||
| - av_le2ne16(srcp_below2[i])) >> 3, 0, clip_max); | |||
| // Prevent over-sharpening: | |||
| // dst must not exceed src when the average of above and below | |||
| // is less than src. And the other way around. | |||
| if (src_ab > src_x) { | |||
| if (dst_le < src_le) | |||
| dstp[i] = av_le2ne16(src_le); | |||
| else | |||
| dstp[i] = av_le2ne16(dst_le); | |||
| } else if (dst_le > src_le) { | |||
| dstp[i] = av_le2ne16(src_le); | |||
| } else | |||
| dstp[i] = av_le2ne16(dst_le); | |||
| } | |||
| } | |||
| static const enum AVPixelFormat formats_supported[] = { | |||
| AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUV444P, | |||
| AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUV410P, AV_PIX_FMT_YUVA420P, | |||
| AV_PIX_FMT_GRAY8, AV_PIX_FMT_YUVJ420P, AV_PIX_FMT_YUVJ422P, | |||
| AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_YUVJ440P, AV_PIX_FMT_NONE | |||
| AV_PIX_FMT_YUV410P, AV_PIX_FMT_YUV411P, | |||
| AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUV444P, | |||
| AV_PIX_FMT_YUV420P10LE, AV_PIX_FMT_YUV422P10LE, AV_PIX_FMT_YUV444P10LE, | |||
| AV_PIX_FMT_YUV420P12LE, AV_PIX_FMT_YUV422P12LE, AV_PIX_FMT_YUV444P12LE, | |||
| AV_PIX_FMT_YUVA420P, AV_PIX_FMT_YUVA422P, AV_PIX_FMT_YUVA444P, | |||
| AV_PIX_FMT_YUVA420P10LE, AV_PIX_FMT_YUVA422P10LE, AV_PIX_FMT_YUVA444P10LE, | |||
| AV_PIX_FMT_GRAY8, AV_PIX_FMT_YUVJ420P, AV_PIX_FMT_YUVJ422P, | |||
| AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_YUVJ440P, AV_PIX_FMT_NONE | |||
| }; | |||
| static int query_formats(AVFilterContext *ctx) | |||
| @@ -150,12 +209,19 @@ static int config_out_props(AVFilterLink *outlink) | |||
| outlink->time_base.num *= 2; | |||
| outlink->frame_rate.den *= 2; | |||
| s->csp = av_pix_fmt_desc_get(outlink->format); | |||
| if (s->lowpass) { | |||
| if (s->lowpass == VLPF_LIN) | |||
| s->lowpass_line = lowpass_line_c; | |||
| else if (s->lowpass == VLPF_CMP) | |||
| s->lowpass_line = lowpass_line_complex_c; | |||
| if (s->lowpass == VLPF_LIN) { | |||
| if (s->csp->comp[0].depth > 8) | |||
| s->lowpass_line = lowpass_line_c_16; | |||
| else | |||
| s->lowpass_line = lowpass_line_c; | |||
| } else if (s->lowpass == VLPF_CMP) { | |||
| if (s->csp->comp[0].depth > 8) | |||
| s->lowpass_line = lowpass_line_complex_c_16; | |||
| else | |||
| s->lowpass_line = lowpass_line_complex_c; | |||
| } | |||
| if (ARCH_X86) | |||
| ff_interlace_init_x86(s); | |||
| } | |||
| @@ -183,6 +249,7 @@ static void copy_picture_field(InterlaceContext *s, | |||
| const uint8_t *srcp = src_frame->data[plane]; | |||
| int srcp_linesize = src_frame->linesize[plane] * 2; | |||
| int dstp_linesize = dst_frame->linesize[plane] * 2; | |||
| int clip_max = (1 << s->csp->comp[plane].depth) - 1; | |||
| av_assert0(cols >= 0 || lines >= 0); | |||
| @@ -202,11 +269,13 @@ static void copy_picture_field(InterlaceContext *s, | |||
| mref = 0; | |||
| else if (j <= (1 + x)) | |||
| pref = 0; | |||
| s->lowpass_line(dstp, cols, srcp, mref, pref); | |||
| s->lowpass_line(dstp, cols, srcp, mref, pref, clip_max); | |||
| dstp += dstp_linesize; | |||
| srcp += srcp_linesize; | |||
| } | |||
| } else { | |||
| if (s->csp->comp[plane].depth > 8) | |||
| cols *= 2; | |||
| av_image_copy_plane(dstp, dstp_linesize, srcp, srcp_linesize, cols, lines); | |||
| } | |||
| } | |||
| @@ -78,7 +78,12 @@ static int query_formats(AVFilterContext *ctx) | |||
| AV_PIX_FMT_YUV410P, AV_PIX_FMT_YUV411P, | |||
| AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV422P, | |||
| AV_PIX_FMT_YUV440P, AV_PIX_FMT_YUV444P, | |||
| AV_PIX_FMT_YUV420P10LE, AV_PIX_FMT_YUV422P10LE, | |||
| AV_PIX_FMT_YUV440P10LE, AV_PIX_FMT_YUV444P10LE, | |||
| AV_PIX_FMT_YUV420P12LE, AV_PIX_FMT_YUV422P12LE, | |||
| AV_PIX_FMT_YUV440P12LE, AV_PIX_FMT_YUV444P12LE, | |||
| AV_PIX_FMT_YUVA420P, AV_PIX_FMT_YUVA422P, AV_PIX_FMT_YUVA444P, | |||
| AV_PIX_FMT_YUVA420P10LE, AV_PIX_FMT_YUVA422P10LE, AV_PIX_FMT_YUVA444P10LE, | |||
| AV_PIX_FMT_GRAY8, FULL_SCALE_YUVJ_FORMATS, | |||
| AV_PIX_FMT_NONE | |||
| }; | |||
| @@ -90,7 +95,7 @@ static int query_formats(AVFilterContext *ctx) | |||
| } | |||
| static void lowpass_line_c(uint8_t *dstp, ptrdiff_t width, const uint8_t *srcp, | |||
| ptrdiff_t mref, ptrdiff_t pref) | |||
| ptrdiff_t mref, ptrdiff_t pref, int clip_max) | |||
| { | |||
| const uint8_t *srcp_above = srcp + mref; | |||
| const uint8_t *srcp_below = srcp + pref; | |||
| @@ -103,8 +108,26 @@ static void lowpass_line_c(uint8_t *dstp, ptrdiff_t width, const uint8_t *srcp, | |||
| } | |||
| } | |||
| static void lowpass_line_c_16(uint8_t *dst8, ptrdiff_t width, const uint8_t *src8, | |||
| ptrdiff_t mref, ptrdiff_t pref, int clip_max) | |||
| { | |||
| uint16_t *dstp = (uint16_t *)dst8; | |||
| const uint16_t *srcp = (const uint16_t *)src8; | |||
| const uint16_t *srcp_above = srcp + mref / 2; | |||
| const uint16_t *srcp_below = srcp + pref / 2; | |||
| int i, src_x; | |||
| for (i = 0; i < width; i++) { | |||
| // this calculation is an integer representation of | |||
| // '0.5 * current + 0.25 * above + 0.25 * below' | |||
| // '1 +' is for rounding. | |||
| src_x = av_le2ne16(srcp[i]) << 1; | |||
| dstp[i] = av_le2ne16((1 + src_x + av_le2ne16(srcp_above[i]) | |||
| + av_le2ne16(srcp_below[i])) >> 2); | |||
| } | |||
| } | |||
| static void lowpass_line_complex_c(uint8_t *dstp, ptrdiff_t width, const uint8_t *srcp, | |||
| ptrdiff_t mref, ptrdiff_t pref) | |||
| ptrdiff_t mref, ptrdiff_t pref, int clip_max) | |||
| { | |||
| const uint8_t *srcp_above = srcp + mref; | |||
| const uint8_t *srcp_below = srcp + pref; | |||
| @@ -130,6 +153,41 @@ static void lowpass_line_complex_c(uint8_t *dstp, ptrdiff_t width, const uint8_t | |||
| } | |||
| } | |||
| static void lowpass_line_complex_c_16(uint8_t *dst8, ptrdiff_t width, const uint8_t *src8, | |||
| ptrdiff_t mref, ptrdiff_t pref, int clip_max) | |||
| { | |||
| uint16_t *dstp = (uint16_t *)dst8; | |||
| const uint16_t *srcp = (const uint16_t *)src8; | |||
| const uint16_t *srcp_above = srcp + mref / 2; | |||
| const uint16_t *srcp_below = srcp + pref / 2; | |||
| const uint16_t *srcp_above2 = srcp + mref; | |||
| const uint16_t *srcp_below2 = srcp + pref; | |||
| int i, dst_le, src_le, src_x, src_ab; | |||
| for (i = 0; i < width; i++) { | |||
| // this calculation is an integer representation of | |||
| // '0.75 * current + 0.25 * above + 0.25 * below - 0.125 * above2 - 0.125 * below2' | |||
| // '4 +' is for rounding. | |||
| src_le = av_le2ne16(srcp[i]); | |||
| src_x = src_le << 1; | |||
| src_ab = av_le2ne16(srcp_above[i]) + av_le2ne16(srcp_below[i]); | |||
| dst_le = av_clip((4 + ((src_le + src_x + src_ab) << 1) | |||
| - av_le2ne16(srcp_above2[i]) | |||
| - av_le2ne16(srcp_below2[i])) >> 3, 0, clip_max); | |||
| // Prevent over-sharpening: | |||
| // dst must not exceed src when the average of above and below | |||
| // is less than src. And the other way around. | |||
| if (src_ab > src_x) { | |||
| if (dst_le < src_le) | |||
| dstp[i] = av_le2ne16(src_le); | |||
| else | |||
| dstp[i] = av_le2ne16(dst_le); | |||
| } else if (dst_le > src_le) { | |||
| dstp[i] = av_le2ne16(src_le); | |||
| } else | |||
| dstp[i] = av_le2ne16(dst_le); | |||
| } | |||
| } | |||
| static av_cold void uninit(AVFilterContext *ctx) | |||
| { | |||
| TInterlaceContext *tinterlace = ctx->priv; | |||
| @@ -198,12 +256,19 @@ static int config_out_props(AVFilterLink *outlink) | |||
| (tinterlace->flags & TINTERLACE_FLAG_EXACT_TB)) | |||
| outlink->time_base = tinterlace->preout_time_base; | |||
| tinterlace->csp = av_pix_fmt_desc_get(outlink->format); | |||
| if (tinterlace->flags & TINTERLACE_FLAG_CVLPF) { | |||
| tinterlace->lowpass_line = lowpass_line_complex_c; | |||
| if (tinterlace->csp->comp[0].depth > 8) | |||
| tinterlace->lowpass_line = lowpass_line_complex_c_16; | |||
| else | |||
| tinterlace->lowpass_line = lowpass_line_complex_c; | |||
| if (ARCH_X86) | |||
| ff_tinterlace_init_x86(tinterlace); | |||
| } else if (tinterlace->flags & TINTERLACE_FLAG_VLPF) { | |||
| tinterlace->lowpass_line = lowpass_line_c; | |||
| if (tinterlace->csp->comp[0].depth > 8) | |||
| tinterlace->lowpass_line = lowpass_line_c_16; | |||
| else | |||
| tinterlace->lowpass_line = lowpass_line_c; | |||
| if (ARCH_X86) | |||
| ff_tinterlace_init_x86(tinterlace); | |||
| } | |||
| @@ -250,6 +315,7 @@ void copy_picture_field(TInterlaceContext *tinterlace, | |||
| const uint8_t *srcp = src[plane]; | |||
| int srcp_linesize = src_linesize[plane] * k; | |||
| int dstp_linesize = dst_linesize[plane] * (interleave ? 2 : 1); | |||
| int clip_max = (1 << tinterlace->csp->comp[plane].depth) - 1; | |||
| lines = (lines + (src_field == FIELD_UPPER)) / k; | |||
| if (src_field == FIELD_LOWER) | |||
| @@ -267,11 +333,13 @@ void copy_picture_field(TInterlaceContext *tinterlace, | |||
| if (h >= (lines - x)) mref = 0; // there is no line above | |||
| else if (h <= (1 + x)) pref = 0; // there is no line below | |||
| tinterlace->lowpass_line(dstp, cols, srcp, mref, pref); | |||
| tinterlace->lowpass_line(dstp, cols, srcp, mref, pref, clip_max); | |||
| dstp += dstp_linesize; | |||
| srcp += srcp_linesize; | |||
| } | |||
| } else { | |||
| if (tinterlace->csp->comp[plane].depth > 8) | |||
| cols *= 2; | |||
| av_image_copy_plane(dstp, dstp_linesize, srcp, srcp_linesize, cols, lines); | |||
| } | |||
| } | |||
| @@ -30,27 +30,26 @@ pw_4: times 8 dw 4 | |||
| SECTION .text | |||
| %macro LOWPASS_LINE 0 | |||
| cglobal lowpass_line, 5, 5, 7, dst, h, src, mref, pref | |||
| %macro LOWPASS 1 | |||
| add dstq, hq | |||
| add srcq, hq | |||
| add mrefq, srcq | |||
| add prefq, srcq | |||
| neg hq | |||
| pcmpeqb m6, m6 | |||
| pcmpeq%1 m6, m6 | |||
| .loop: | |||
| mova m0, [mrefq+hq] | |||
| mova m1, [mrefq+hq+mmsize] | |||
| pavgb m0, [prefq+hq] | |||
| pavgb m1, [prefq+hq+mmsize] | |||
| pavg%1 m0, [prefq+hq] | |||
| pavg%1 m1, [prefq+hq+mmsize] | |||
| pxor m0, m6 | |||
| pxor m1, m6 | |||
| pxor m2, m6, [srcq+hq] | |||
| pxor m3, m6, [srcq+hq+mmsize] | |||
| pavgb m0, m2 | |||
| pavgb m1, m3 | |||
| pavg%1 m0, m2 | |||
| pavg%1 m1, m3 | |||
| pxor m0, m6 | |||
| pxor m1, m6 | |||
| mova [dstq+hq], m0 | |||
| @@ -59,7 +58,15 @@ cglobal lowpass_line, 5, 5, 7, dst, h, src, mref, pref | |||
| add hq, 2*mmsize | |||
| jl .loop | |||
| REP_RET | |||
| %endmacro | |||
| %macro LOWPASS_LINE 0 | |||
| cglobal lowpass_line, 5, 5, 7, dst, h, src, mref, pref | |||
| LOWPASS b | |||
| cglobal lowpass_line_16, 5, 5, 7, dst, h, src, mref, pref | |||
| shl hq, 1 | |||
| LOWPASS w | |||
| %endmacro | |||
| %macro LOWPASS_LINE_COMPLEX 0 | |||
| @@ -124,6 +131,65 @@ cglobal lowpass_line_complex, 5, 5, 8, dst, h, src, mref, pref | |||
| jg .loop | |||
| REP_RET | |||
| cglobal lowpass_line_complex_12, 5, 5, 8, 16, dst, h, src, mref, pref, clip_max | |||
| movd m7, DWORD clip_maxm | |||
| SPLATW m7, m7, 0 | |||
| mova [rsp], m7 | |||
| .loop: | |||
| mova m0, [srcq+mrefq] | |||
| mova m1, [srcq+mrefq+mmsize] | |||
| mova m2, [srcq+prefq] | |||
| mova m3, [srcq+prefq+mmsize] | |||
| paddw m0, m2 | |||
| paddw m1, m3 | |||
| mova m6, m0 | |||
| mova m7, m1 | |||
| mova m2, [srcq] | |||
| mova m3, [srcq+mmsize] | |||
| paddw m0, m2 | |||
| paddw m1, m3 | |||
| psllw m2, 1 | |||
| psllw m3, 1 | |||
| paddw m0, m2 | |||
| paddw m1, m3 | |||
| psllw m0, 1 | |||
| psllw m1, 1 | |||
| pcmpgtw m6, m2 | |||
| pcmpgtw m7, m3 | |||
| mova m2, [srcq+2*mrefq] | |||
| mova m3, [srcq+2*mrefq+mmsize] | |||
| mova m4, [srcq+2*prefq] | |||
| mova m5, [srcq+2*prefq+mmsize] | |||
| paddw m2, m4 | |||
| paddw m3, m5 | |||
| paddw m0, [pw_4] | |||
| paddw m1, [pw_4] | |||
| psubusw m0, m2 | |||
| psubusw m1, m3 | |||
| psrlw m0, 3 | |||
| psrlw m1, 3 | |||
| pminsw m0, [rsp] | |||
| pminsw m1, [rsp] | |||
| mova m2, m0 | |||
| mova m3, m1 | |||
| pmaxsw m0, [srcq] | |||
| pmaxsw m1, [srcq+mmsize] | |||
| pminsw m2, [srcq] | |||
| pminsw m3, [srcq+mmsize] | |||
| pand m0, m6 | |||
| pand m1, m7 | |||
| pandn m6, m2 | |||
| pandn m7, m3 | |||
| por m0, m6 | |||
| por m1, m7 | |||
| mova [dstq], m0 | |||
| mova [dstq+mmsize], m1 | |||
| add dstq, 2*mmsize | |||
| add srcq, 2*mmsize | |||
| sub hd, mmsize | |||
| jg .loop | |||
| REP_RET | |||
| %endmacro | |||
| INIT_XMM sse2 | |||
| @@ -27,27 +27,50 @@ | |||
| #include "libavfilter/interlace.h" | |||
| void ff_lowpass_line_sse2(uint8_t *dstp, ptrdiff_t linesize, | |||
| const uint8_t *srcp, | |||
| ptrdiff_t mref, ptrdiff_t pref); | |||
| const uint8_t *srcp, ptrdiff_t mref, | |||
| ptrdiff_t pref, int clip_max); | |||
| void ff_lowpass_line_avx (uint8_t *dstp, ptrdiff_t linesize, | |||
| const uint8_t *srcp, | |||
| ptrdiff_t mref, ptrdiff_t pref); | |||
| const uint8_t *srcp, ptrdiff_t mref, | |||
| ptrdiff_t pref, int clip_max); | |||
| void ff_lowpass_line_16_sse2(uint8_t *dstp, ptrdiff_t linesize, | |||
| const uint8_t *srcp, ptrdiff_t mref, | |||
| ptrdiff_t pref, int clip_max); | |||
| void ff_lowpass_line_16_avx (uint8_t *dstp, ptrdiff_t linesize, | |||
| const uint8_t *srcp, ptrdiff_t mref, | |||
| ptrdiff_t pref, int clip_max); | |||
| void ff_lowpass_line_complex_sse2(uint8_t *dstp, ptrdiff_t linesize, | |||
| const uint8_t *srcp, | |||
| ptrdiff_t mref, ptrdiff_t pref); | |||
| const uint8_t *srcp, ptrdiff_t mref, | |||
| ptrdiff_t pref, int clip_max); | |||
| void ff_lowpass_line_complex_12_sse2(uint8_t *dstp, ptrdiff_t linesize, | |||
| const uint8_t *srcp, ptrdiff_t mref, | |||
| ptrdiff_t pref, int clip_max); | |||
| av_cold void ff_interlace_init_x86(InterlaceContext *s) | |||
| { | |||
| int cpu_flags = av_get_cpu_flags(); | |||
| if (EXTERNAL_SSE2(cpu_flags)) { | |||
| if (s->lowpass == VLPF_LIN) | |||
| s->lowpass_line = ff_lowpass_line_sse2; | |||
| else if (s->lowpass == VLPF_CMP) | |||
| s->lowpass_line = ff_lowpass_line_complex_sse2; | |||
| if (s->csp->comp[0].depth > 8) { | |||
| if (EXTERNAL_SSE2(cpu_flags)) { | |||
| if (s->lowpass == VLPF_LIN) | |||
| s->lowpass_line = ff_lowpass_line_16_sse2; | |||
| else if (s->lowpass == VLPF_CMP) | |||
| s->lowpass_line = ff_lowpass_line_complex_12_sse2; | |||
| } | |||
| if (EXTERNAL_AVX(cpu_flags)) | |||
| if (s->lowpass == VLPF_LIN) | |||
| s->lowpass_line = ff_lowpass_line_16_avx; | |||
| } else { | |||
| if (EXTERNAL_SSE2(cpu_flags)) { | |||
| if (s->lowpass == VLPF_LIN) | |||
| s->lowpass_line = ff_lowpass_line_sse2; | |||
| else if (s->lowpass == VLPF_CMP) | |||
| s->lowpass_line = ff_lowpass_line_complex_sse2; | |||
| } | |||
| if (EXTERNAL_AVX(cpu_flags)) | |||
| if (s->lowpass == VLPF_LIN) | |||
| s->lowpass_line = ff_lowpass_line_avx; | |||
| } | |||
| if (EXTERNAL_AVX(cpu_flags)) | |||
| if (s->lowpass == VLPF_LIN) | |||
| s->lowpass_line = ff_lowpass_line_avx; | |||
| } | |||
| @@ -28,27 +28,50 @@ | |||
| #include "libavfilter/tinterlace.h" | |||
| void ff_lowpass_line_sse2(uint8_t *dstp, ptrdiff_t linesize, | |||
| const uint8_t *srcp, | |||
| ptrdiff_t mref, ptrdiff_t pref); | |||
| const uint8_t *srcp, ptrdiff_t mref, | |||
| ptrdiff_t pref, int clip_max); | |||
| void ff_lowpass_line_avx (uint8_t *dstp, ptrdiff_t linesize, | |||
| const uint8_t *srcp, | |||
| ptrdiff_t mref, ptrdiff_t pref); | |||
| const uint8_t *srcp, ptrdiff_t mref, | |||
| ptrdiff_t pref, int clip_max); | |||
| void ff_lowpass_line_16_sse2(uint8_t *dstp, ptrdiff_t linesize, | |||
| const uint8_t *srcp, ptrdiff_t mref, | |||
| ptrdiff_t pref, int clip_max); | |||
| void ff_lowpass_line_16_avx (uint8_t *dstp, ptrdiff_t linesize, | |||
| const uint8_t *srcp, ptrdiff_t mref, | |||
| ptrdiff_t pref, int clip_max); | |||
| void ff_lowpass_line_complex_sse2(uint8_t *dstp, ptrdiff_t linesize, | |||
| const uint8_t *srcp, | |||
| ptrdiff_t mref, ptrdiff_t pref); | |||
| const uint8_t *srcp, ptrdiff_t mref, | |||
| ptrdiff_t pref, int clip_max); | |||
| void ff_lowpass_line_complex_12_sse2(uint8_t *dstp, ptrdiff_t linesize, | |||
| const uint8_t *srcp, ptrdiff_t mref, | |||
| ptrdiff_t pref, int clip_max); | |||
| av_cold void ff_tinterlace_init_x86(TInterlaceContext *s) | |||
| { | |||
| int cpu_flags = av_get_cpu_flags(); | |||
| if (EXTERNAL_SSE2(cpu_flags)) { | |||
| if (!(s->flags & TINTERLACE_FLAG_CVLPF)) | |||
| s->lowpass_line = ff_lowpass_line_sse2; | |||
| else | |||
| s->lowpass_line = ff_lowpass_line_complex_sse2; | |||
| if (s->csp->comp[0].depth > 8) { | |||
| if (EXTERNAL_SSE2(cpu_flags)) { | |||
| if (!(s->flags & TINTERLACE_FLAG_CVLPF)) | |||
| s->lowpass_line = ff_lowpass_line_16_sse2; | |||
| else | |||
| s->lowpass_line = ff_lowpass_line_complex_12_sse2; | |||
| } | |||
| if (EXTERNAL_AVX(cpu_flags)) | |||
| if (!(s->flags & TINTERLACE_FLAG_CVLPF)) | |||
| s->lowpass_line = ff_lowpass_line_16_avx; | |||
| } else { | |||
| if (EXTERNAL_SSE2(cpu_flags)) { | |||
| if (!(s->flags & TINTERLACE_FLAG_CVLPF)) | |||
| s->lowpass_line = ff_lowpass_line_sse2; | |||
| else | |||
| s->lowpass_line = ff_lowpass_line_complex_sse2; | |||
| } | |||
| if (EXTERNAL_AVX(cpu_flags)) | |||
| if (!(s->flags & TINTERLACE_FLAG_CVLPF)) | |||
| s->lowpass_line = ff_lowpass_line_avx; | |||
| } | |||
| if (EXTERNAL_AVX(cpu_flags)) | |||
| if (!(s->flags & TINTERLACE_FLAG_CVLPF)) | |||
| s->lowpass_line = ff_lowpass_line_avx; | |||
| } | |||
| @@ -2,12 +2,23 @@ gray 9849d71519ae9c584ae8abfa8adb2f8e | |||
| yuv410p 44ee4b74b95c82d6f79ddf53b5e3aa9d | |||
| yuv411p 5fa9d1fba7adfd6f7fa04464332b631a | |||
| yuv420p ee9591ea3ab06c73be902c4b8868c69e | |||
| yuv420p10le 19b2dcf5e82725b2c9e366d5d3a4b67b | |||
| yuv420p12le e5b76eb58cab2dfa42738f665d8b8059 | |||
| yuv422p b1be7b55567bde86d655adf80fac1257 | |||
| yuv422p10le 66f7433c7423d6289ee20c19c4e8dd90 | |||
| yuv422p12le b4a4b5ed81341b36f685b14b2bb8a798 | |||
| yuv440p ddf6ee697f4ff4f90d501e6869392309 | |||
| yuv440p10le 7e44a92d9e47e15f0728b125725ccd41 | |||
| yuv440p12le c2cac82661e10be4917bd9dacb419fce | |||
| yuv444p 7cb5d0c0997c8c2545a16bfc4cb9fd6d | |||
| yuv444p10le 08fb13c69b40229014a6ccf0bdd5498f | |||
| yuv444p12le 796c5559b9f426470eb2997c21be5d14 | |||
| yuva420p ee0761e2f76ec441c545feede77103e4 | |||
| yuva420p10le 4bb6b3a1cbf2fd1fe5aefcc11ce8029f | |||
| yuva422p a8da2806e21a88449079faa7f4303ffa | |||
| yuva422p10le d2965b5b5a43a7165badaff0718a17d8 | |||
| yuva444p a3f57734d6f72bdf37f8f612ea7cce63 | |||
| yuva444p10le e020512901fd9ac7088898a4e3a8c7c1 | |||
| yuvj420p 9f358e311b694bcd01e1a07d1120ade5 | |||
| yuvj422p 9a7628a9f1630d35c7176951ddc1b2f6 | |||
| yuvj440p 112fe35292c687746ec0c622a42c611b | |||
| @@ -2,12 +2,23 @@ gray fab3a7abc4f076cf926205aeacadbe51 | |||
| yuv410p d4506e49eeb64c7ce714c07597e7dd69 | |||
| yuv411p 2e8bb385cb4a53a0f3771815020f7213 | |||
| yuv420p c967d3d5f3200d1b2417b0f2356f12fc | |||
| yuv420p10le 404d90eca9b93f7c0949be23187804c7 | |||
| yuv420p12le eb2ef81b82f1342618e4c9fe8e086592 | |||
| yuv422p febaa84ea2e3246af742a7ed37c030f6 | |||
| yuv422p10le 0dc5f3833c0f11a0fcd0422d76508c10 | |||
| yuv422p12le 190eb7e76af5fc481588ded472901b58 | |||
| yuv440p 13a934b42df65f11e153314ebb4f311e | |||
| yuv440p10le 7fabea51d6e52b718ef3ab919b24f8d2 | |||
| yuv440p12le c7b138374a18f258bdd1a2d21b4421b0 | |||
| yuv444p 45d4466f5689942a4effb0fd23e44949 | |||
| yuv444p10le fde95627ba0e66be7ec863386b3e5ca6 | |||
| yuv444p12le c1e15c934ce3bc562a869dd78c75db17 | |||
| yuva420p dc1173a07c3f993b277ea0c94d513e1f | |||
| yuva420p10le d85972762437105a932759c4bb9759c9 | |||
| yuva422p ca200be80e5bfdb159e1aea57129ed3a | |||
| yuva422p10le 06d4f79ee2ddf31d9fe15af8ca573f46 | |||
| yuva444p 9f39c35d6899dcb8b9a9b07c339ca365 | |||
| yuva444p10le b0c54fc3efad73f252d86127407aa1fd | |||
| yuvj420p 844359293bb6ff81549f3fc0090cc587 | |||
| yuvj422p 526af049d43974822baa7b48aa1e1098 | |||
| yuvj440p af9285194da8efbc40d93bf8109f9dc5 | |||
| @@ -2,12 +2,23 @@ gray 7ef396fecd8d1c9fe32173e4415ba671 | |||
| yuv410p 35bc11d0d32efc9e9a969be7d720f4e6 | |||
| yuv411p 17ef3cd22a74f7368b5e02f68779f294 | |||
| yuv420p 93d5b6a4c44d67e4d4447e8dd0bf3d33 | |||
| yuv420p10le 14e754c6e9d41cb048ce3c93512d7d35 | |||
| yuv420p12le ce54a2e38d121a7575dff30542facaad | |||
| yuv422p 3ee40b0b6533b9183764b85c853ec3f9 | |||
| yuv422p10le d4b61a84b93e74b07b7020ceed40e39e | |||
| yuv422p12le 8fd90be12a97307645ecfcd09d576643 | |||
| yuv440p 1d3c1258a51d09e778cd8368b1a4126f | |||
| yuv440p10le 29d116cb550a05920e5619ab58284d30 | |||
| yuv440p12le d7518f941a3b2c137f944afe9da816a1 | |||
| yuv444p 1093568ad8f479ec20e738d018dd3f8f | |||
| yuv444p10le c65233e1d8b01f3369c20738f7386801 | |||
| yuv444p12le 70dc51a857bfb215b3a81fceb114b74c | |||
| yuva420p 4588aef20c0010e514550c9391219724 | |||
| yuva420p10le 3181e84fd7aaed606bb86eecd2e13f20 | |||
| yuva422p 3426ed1ac9429202d8c29fa62a04d4c3 | |||
| yuva422p10le c00acd7c437d41755dff09c5ca3642cf | |||
| yuva444p 1b9fc791c7d774b4ba8c9dc836f78cf5 | |||
| yuva444p10le 616b42a232c83b8f9e5c5168ec4b5da5 | |||
| yuvj420p 9a872e0c1b3c0b6fe856415696b758bd | |||
| yuvj422p da3c9ef25528a2ee96746ce44e6969f3 | |||
| yuvj440p a9a5495c6b0e2bf6e561998ea1c356a7 | |||
| @@ -2,12 +2,23 @@ gray b79791449947c25cd5b36d9d3b9d1831 | |||
| yuv410p 5bc03f4cf6b441b421f0fdaeeff1e9ed | |||
| yuv411p 19046df1876c46ed1ef0458680270bd3 | |||
| yuv420p 69c743b84996be9430b051a55cfbcb29 | |||
| yuv420p10le 85948ad609abded6b50882d459f5a2f8 | |||
| yuv420p12le 7cebe45f51bdadc766f66c68db8d347d | |||
| yuv422p d710ccd1941f6f389c97a09bc977e709 | |||
| yuv422p10le c54873f77dac1d710fb2aa1b0ce2669c | |||
| yuv422p12le 94a527bb787b9d121ffbbcb3a6c545d8 | |||
| yuv440p 1a482a23fe5a9b7d02388c299fd0a423 | |||
| yuv440p10le 506efa287ecce9c951da2039fa1de2ae | |||
| yuv440p12le 631bcf190f409ccbc5c27b9f0f6ba5e2 | |||
| yuv444p c968a92f4b7ab6706ee9b425eb5345b5 | |||
| yuv444p10le 0af437e635d49feccf7dfae201e6dfc5 | |||
| yuv444p12le 2e9e9f7caae1fae3b026810246fc6ac1 | |||
| yuva420p 3f89a166f309c0cda8b91a9e8a0ce937 | |||
| yuva420p10le 79de1cc549c03d4893cf6f1aca86e057 | |||
| yuva422p ef8fdbe910d68e88e98227b0e99fb5a6 | |||
| yuva422p10le 257a4aec41f9b5412179272d8a7fb6f7 | |||
| yuva444p 3662eadd5f61a6edbc9d715ea8591415 | |||
| yuva444p10le 0905cf5b7f42c11be3f0486a66533c71 | |||
| yuvj420p 14c4390b319c5d679184503309060ac3 | |||
| yuvj422p bbe00a26526931b72a024febe1cd6b90 | |||
| yuvj440p f654cf28b7879c6a6c950c3cb9612580 | |||