"halfpack" (yuv planar 4:2:0 -> packed 4:2:2, half height) video filter

(useful for downsampling luma for low-res output devices without
losing chroma samples, when hardware downscaling is poor quality or
unavailable)


git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@7156 b3059339-0415-0410-9bf9-f77b7e298cf2

3 files changed, 203 insertions, 1 deletions

diff --git a/libmpcodecs/Makefile b/libmpcodecs/Makefile
index 8b68c81edf..45e13e2ec0 100644
--- a/libmpcodecs/Makefile
+++ b/libmpcodecs/Makefile
@@ -6,7 +6,7 @@ LIBNAME2 = libmpencoders.a
 
 AUDIO_SRCS=dec_audio.c ad.c ad_a52.c ad_acm.c ad_alaw.c ad_dk3adpcm.c ad_dshow.c ad_dvdpcm.c ad_ffmpeg.c ad_hwac3.c ad_imaadpcm.c ad_mp3.c ad_msadpcm.c ad_pcm.c ad_roqaudio.c ad_msgsm.c ad_faad.c ad_vorbis.c ad_libmad.c ad_real.c ad_libdv.c
 VIDEO_SRCS=dec_video.c vd.c vd_null.c vd_real.c vd_cinepak.c vd_qtrpza.c vd_ffmpeg.c vd_dshow.c vd_vfw.c vd_odivx.c vd_divx4.c vd_raw.c vd_xanim.c vd_msvidc.c vd_fli.c vd_qtrle.c vd_qtsmc.c vd_roqvideo.c vd_cyuv.c vd_nuv.c vd_libmpeg2.c vd_msrle.c vd_huffyuv.c vd_zlib.c vd_mpegpes.c vd_svq1.c vd_xvid.c vd_libdv.c
-VFILTER_SRCS=vf.c vf_vo.c vf_crop.c vf_expand.c vf_pp.c vf_scale.c vf_format.c vf_yuy2.c vf_flip.c vf_rgb2bgr.c vf_rotate.c vf_mirror.c vf_palette.c vf_lavc.c vf_dvbscale.c vf_cropdetect.c vf_test.c vf_noise.c vf_yvu9.c vf_rectangle.c vf_lavcdeint.c vf_eq.c
+VFILTER_SRCS=vf.c vf_vo.c vf_crop.c vf_expand.c vf_pp.c vf_scale.c vf_format.c vf_yuy2.c vf_flip.c vf_rgb2bgr.c vf_rotate.c vf_mirror.c vf_palette.c vf_lavc.c vf_dvbscale.c vf_cropdetect.c vf_test.c vf_noise.c vf_yvu9.c vf_rectangle.c vf_lavcdeint.c vf_eq.c vf_halfpack.c
 ENCODER_SRCS=ve.c ve_divx4.c ve_lavc.c ve_vfw.c ve_rawrgb.c ve_libdv.c
 NATIVE_SRCS=native/RTjpegN.c native/cinepak.c native/cyuv.c native/fli.c native/minilzo.c native/msvidc.c native/nuppelvideo.c native/qtrle.c native/qtrpza.c native/qtsmc.c native/roqav.c native/xa_gsm.c native/svq1.c
 
diff --git a/libmpcodecs/vf.c b/libmpcodecs/vf.c
index 9638c16236..dcbd2e5ec4 100644
--- a/libmpcodecs/vf.c
+++ b/libmpcodecs/vf.c
@@ -39,6 +39,7 @@ extern vf_info_t vf_info_noise;
 extern vf_info_t vf_info_yvu9;
 extern vf_info_t vf_info_lavcdeint;
 extern vf_info_t vf_info_eq;
+extern vf_info_t vf_info_halfpack;
 
 char** vo_plugin_args=(char**) NULL;
 
@@ -71,6 +72,7 @@ static vf_info_t* filter_list[]={
     &vf_info_noise,
     &vf_info_yvu9,
     &vf_info_eq,
+    &vf_info_halfpack,
     NULL
 };
 
diff --git a/libmpcodecs/vf_halfpack.c b/libmpcodecs/vf_halfpack.c
new file mode 100644
index 0000000000..6f60045017
--- /dev/null
+++ b/libmpcodecs/vf_halfpack.c
@@ -0,0 +1,200 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <inttypes.h>
+
+#include "../config.h"
+#include "../mp_msg.h"
+#include "../cpudetect.h"
+
+#include "img_format.h"
+#include "mp_image.h"
+#include "vf.h"
+
+#include "../libvo/fastmemcpy.h"
+#include "../postproc/rgb2rgb.h"
+
+
+#ifdef HAVE_MMX
+static void halfpack_MMX(unsigned char *dst, unsigned char *src[3],
+		     unsigned int dststride, unsigned int srcstride[3],
+		     int w, int h)
+{
+	int j;
+	unsigned char *y1, *y2, *u, *v;
+	unsigned int dstinc, yinc, uinc, vinc;
+
+	y1 = src[0];
+	y2 = src[0] + srcstride[0];
+	u = src[1];
+	v = src[2];
+
+	dstinc = dststride - 2*w;
+	yinc = 2*srcstride[0] - w;
+	uinc = srcstride[1] - w/2;
+	vinc = srcstride[2] - w/2;
+
+	for (h/=2; h; h--) {
+		asm (
+			"pxor %%mm0, %%mm0 \n\t"
+			".align 16 \n\t"
+			"1: \n\t"
+			"movq (%0), %%mm1 \n\t"
+			"movq (%0), %%mm2 \n\t"
+			"movq (%1), %%mm3 \n\t"
+			"movq (%1), %%mm4 \n\t"
+			"punpcklbw %%mm0, %%mm1 \n\t"
+			"punpckhbw %%mm0, %%mm2 \n\t"
+			"punpcklbw %%mm0, %%mm3 \n\t"
+			"punpckhbw %%mm0, %%mm4 \n\t"
+			"paddw %%mm3, %%mm1 \n\t"
+			"paddw %%mm4, %%mm2 \n\t"
+			"psrlw $1, %%mm1 \n\t"
+			"psrlw $1, %%mm2 \n\t"
+			
+			"movq (%2), %%mm3 \n\t"
+			"movq (%3), %%mm5 \n\t"
+			"punpcklbw %%mm0, %%mm3 \n\t"
+			"punpcklbw %%mm0, %%mm5 \n\t"
+			"movq %%mm3, %%mm4 \n\t"
+			"movq %%mm5, %%mm6 \n\t"
+			"punpcklwd %%mm0, %%mm3 \n\t"
+			"punpckhwd %%mm0, %%mm4 \n\t"
+			"punpcklwd %%mm0, %%mm5 \n\t"
+			"punpckhwd %%mm0, %%mm6 \n\t"
+			"pslld $8, %%mm3 \n\t"
+			"pslld $8, %%mm4 \n\t"
+			"pslld $24, %%mm5 \n\t"
+			"pslld $24, %%mm6 \n\t"
+
+			"por %%mm3, %%mm1 \n\t"
+			"por %%mm4, %%mm2 \n\t"
+			"por %%mm5, %%mm1 \n\t"
+			"por %%mm6, %%mm2 \n\t"
+
+			"addl $8, %0 \n\t"
+			"addl $8, %1 \n\t"
+			"addl $4, %2 \n\t"
+			"addl $4, %3 \n\t"
+			"movq %%mm1, (%8) \n\t"
+			"movq %%mm2, 8(%8) \n\t"
+			"addl $16, %8 \n\t"
+			"decl %9 \n\t"
+			"jnz 1b \n\t"
+			: "=r" (y1), "=r" (y2), "=r" (u), "=r" (v)
+			: "0" (y1), "1" (y2), "2" (u), "3" (v), "r" (dst), "r" (w/8)
+			: "memory"
+		);
+		for (j = (w&7)/2; j; j--) {
+			*dst++ = (*y1++ + *y2++)/2;
+			*dst++ = *u++;
+			*dst++ = (*y1++ + *y2++)/2;
+			*dst++ = *v++;
+		}
+		y1 += yinc;
+		y2 += yinc;
+		u += uinc;
+		v += vinc;
+		dst += dstinc;
+	}
+	asm volatile ( "emms \n\t" ::: "memory" );
+}
+#endif
+
+
+
+static void halfpack_C(unsigned char *dst, unsigned char *src[3],
+		     unsigned int dststride, unsigned int srcstride[3],
+		     int w, int h)
+{
+	int i, j;
+	unsigned char *y1, *y2, *u, *v;
+	unsigned int dstinc, yinc, uinc, vinc;
+
+	y1 = src[0];
+	y2 = src[0] + srcstride[0];
+	u = src[1];
+	v = src[2];
+
+	dstinc = dststride - 2*w;
+	yinc = 2*srcstride[0] - w;
+	uinc = srcstride[1] - w/2;
+	vinc = srcstride[2] - w/2;
+
+	for (i = h/2; i; i--) {
+		for (j = w/2; j; j--) {
+			*dst++ = (*y1++ + *y2++)>>1;
+			*dst++ = *u++;
+			*dst++ = (*y1++ + *y2++)>>1;
+			*dst++ = *v++;
+		}
+		y1 += yinc;
+		y2 += yinc;
+		u += uinc;
+		v += vinc;
+		dst += dstinc;
+	}
+}
+
+static void (*halfpack)(unsigned char *dst, unsigned char *src[3],
+	unsigned int dststride, unsigned int srcstride[3], int w, int h);
+
+
+static void put_image(struct vf_instance_s* vf, mp_image_t *mpi)
+{
+	mp_image_t *dmpi;
+
+	// hope we'll get DR buffer:
+	dmpi=vf_get_image(vf->next, IMGFMT_YUY2,
+			  MP_IMGTYPE_TEMP, MP_IMGFLAG_ACCEPT_STRIDE,
+			  mpi->w, mpi->h/2);
+
+	halfpack(dmpi->planes[0], mpi->planes,
+		 dmpi->stride[0], mpi->stride,
+		 mpi->w, mpi->h);
+
+	vf_next_put_image(vf,dmpi);
+}
+
+static int config(struct vf_instance_s* vf,
+		  int width, int height, int d_width, int d_height,
+		  unsigned int flags, unsigned int outfmt)
+{
+	/* FIXME - also support UYVY output? */
+	return vf_next_config(vf, width, height/2, d_width, d_height, flags, IMGFMT_YUY2);
+}
+
+
+static int query_format(struct vf_instance_s* vf, unsigned int fmt)
+{
+	/* FIXME - really any YUV 4:2:0 input format should work */
+	switch (fmt) {
+	case IMGFMT_YV12:
+	case IMGFMT_IYUV:
+	case IMGFMT_I420:
+		return vf_next_query_format(vf,IMGFMT_YUY2);
+	}
+	return 0;
+}
+
+
+static int open(vf_instance_t *vf, char* args)
+{
+	vf->config=config;
+	vf->query_format=query_format;
+	vf->put_image=put_image;
+	halfpack = halfpack_C;
+#ifdef HAVE_MMX
+	if(gCpuCaps.hasMMX) halfpack = halfpack_MMX;
+#endif
+	return 1;
+}
+
+vf_info_t vf_info_halfpack = {
+	"yuv planar 4:2:0 -> packed 4:2:2, half height",
+	"halfpack",
+	"Richard Felker",
+	"",
+	open
+};
+