Mercurial > libavcodec.hg
view arm/Makefile @ 12334:435319d67bd8 libavcodec
Use word-writing instead of dword-writing (with two cached but otherwise
unchanged bytes) in the horizontal simple loopfilter. This makes the filter
quite a bit faster in itself (~30 cycles less on Core1), probably mostly
because we don't need a complex 4x4 transpose, but only a simple byte
interleave. Also allows using pextrw on SSE4, which speeds up even more
(e.g. 25% faster on Core i7).
author | rbultje |
---|---|
date | Sat, 31 Jul 2010 23:13:15 +0000 |
parents | 3581c6b92f70 |
children | 2ea2fd28da07 |
line wrap: on
line source
OBJS-$(CONFIG_DCA_DECODER) += arm/dcadsp_init_arm.o \ OBJS-$(CONFIG_VP5_DECODER) += arm/vp56dsp_init_arm.o OBJS-$(CONFIG_VP6_DECODER) += arm/vp56dsp_init_arm.o OBJS-$(CONFIG_H264DSP) += arm/h264dsp_init_arm.o \ arm/h264pred_init_arm.o \ OBJS += arm/dsputil_init_arm.o \ arm/dsputil_arm.o \ arm/fft_init_arm.o \ arm/jrevdct_arm.o \ arm/mpegvideo_arm.o \ arm/simple_idct_arm.o \ OBJS-$(HAVE_ARMV5TE) += arm/dsputil_init_armv5te.o \ arm/mpegvideo_armv5te.o \ arm/mpegvideo_armv5te_s.o \ arm/simple_idct_armv5te.o \ OBJS-$(HAVE_ARMV6) += arm/dsputil_init_armv6.o \ arm/dsputil_armv6.o \ arm/simple_idct_armv6.o \ OBJS-$(HAVE_ARMVFP) += arm/dsputil_vfp.o \ arm/dsputil_init_vfp.o \ OBJS-$(HAVE_IWMMXT) += arm/dsputil_iwmmxt.o \ arm/mpegvideo_iwmmxt.o \ NEON-OBJS-$(CONFIG_FFT) += arm/fft_neon.o \ NEON-OBJS-$(CONFIG_MDCT) += arm/mdct_neon.o \ NEON-OBJS-$(CONFIG_RDFT) += arm/rdft_neon.o \ NEON-OBJS-$(CONFIG_H264DSP) += arm/h264dsp_neon.o \ arm/h264idct_neon.o \ arm/h264pred_neon.o \ NEON-OBJS-$(CONFIG_DCA_DECODER) += arm/dcadsp_neon.o \ arm/synth_filter_neon.o \ NEON-OBJS-$(CONFIG_VP3_DECODER) += arm/vp3dsp_neon.o NEON-OBJS-$(CONFIG_VP5_DECODER) += arm/vp56dsp_neon.o \ arm/vp3dsp_neon.o \ NEON-OBJS-$(CONFIG_VP6_DECODER) += arm/vp56dsp_neon.o \ arm/vp3dsp_neon.o \ OBJS-$(HAVE_NEON) += arm/dsputil_init_neon.o \ arm/dsputil_neon.o \ arm/int_neon.o \ arm/mpegvideo_neon.o \ arm/simple_idct_neon.o \ $(NEON-OBJS-yes)