mirror of
https://github.com/FFmpeg/FFmpeg.git
synced 2024-12-02 03:06:28 +02:00
34454c761f
The 32bits targets have been compiled with -mfpmath=sse for proper reference.
sbr_sum_square C /32bits: 82c (unrolled)/102c
C /64bits: 69c (unrolled)/82c
SSE/32bits: 42c
SSE/64bits: 31c
Use of SSE4.1 dpps to perform the final sum is slower.
Not unrolling to perform 8 operations in a loop yields 10 more cycles.
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
80 lines
4.3 KiB
Makefile
80 lines
4.3 KiB
Makefile
OBJS-$(CONFIG_MLP_DECODER) += x86/mlpdsp.o
|
|
OBJS-$(CONFIG_TRUEHD_DECODER) += x86/mlpdsp.o
|
|
|
|
YASM-OBJS-$(CONFIG_DCT) += x86/dct32_sse.o
|
|
|
|
YASM-OBJS-FFT-$(HAVE_AMD3DNOW) += x86/fft_3dn.o
|
|
YASM-OBJS-FFT-$(HAVE_AMD3DNOWEXT) += x86/fft_3dn2.o
|
|
YASM-OBJS-FFT-$(HAVE_SSE) += x86/fft_sse.o
|
|
YASM-OBJS-$(CONFIG_FFT) += x86/fft_mmx.o \
|
|
$(YASM-OBJS-FFT-yes)
|
|
|
|
YASM-OBJS-$(CONFIG_H264CHROMA) += x86/h264_chromamc.o \
|
|
x86/h264_chromamc_10bit.o
|
|
|
|
MMX-OBJS-$(CONFIG_H264DSP) += x86/h264dsp_mmx.o
|
|
YASM-OBJS-$(CONFIG_H264DSP) += x86/h264_deblock.o \
|
|
x86/h264_deblock_10bit.o \
|
|
x86/h264_idct.o \
|
|
x86/h264_idct_10bit.o \
|
|
x86/h264_weight.o \
|
|
x86/h264_weight_10bit.o \
|
|
|
|
YASM-OBJS-$(CONFIG_H264PRED) += x86/h264_intrapred.o \
|
|
x86/h264_intrapred_10bit.o
|
|
MMX-OBJS-$(CONFIG_H264PRED) += x86/h264_intrapred_init.o
|
|
|
|
MMX-OBJS-$(CONFIG_RV30_DECODER) += x86/rv34dsp_init.o
|
|
YASM-OBJS-$(CONFIG_RV30_DECODER) += x86/rv34dsp.o
|
|
MMX-OBJS-$(CONFIG_RV40_DECODER) += x86/rv34dsp_init.o \
|
|
x86/rv40dsp_init.o
|
|
YASM-OBJS-$(CONFIG_RV40_DECODER) += x86/rv34dsp.o \
|
|
x86/rv40dsp.o
|
|
|
|
YASM-OBJS-$(CONFIG_VC1_DECODER) += x86/vc1dsp_yasm.o
|
|
|
|
MMX-OBJS-$(CONFIG_AC3DSP) += x86/ac3dsp_mmx.o
|
|
YASM-OBJS-$(CONFIG_AC3DSP) += x86/ac3dsp.o
|
|
MMX-OBJS-$(CONFIG_CAVS_DECODER) += x86/cavsdsp_mmx.o
|
|
MMX-OBJS-$(CONFIG_DNXHD_ENCODER) += x86/dnxhd_mmx.o
|
|
MMX-OBJS-$(CONFIG_MPEGAUDIODSP) += x86/mpegaudiodec_mmx.o
|
|
YASM-OBJS-$(CONFIG_MPEGAUDIODSP) += x86/imdct36_sse.o
|
|
MMX-OBJS-$(CONFIG_ENCODERS) += x86/dsputilenc_mmx.o
|
|
YASM-OBJS-$(CONFIG_ENCODERS) += x86/dsputilenc_yasm.o
|
|
MMX-OBJS-$(CONFIG_GPL) += x86/idct_mmx.o
|
|
MMX-OBJS-$(CONFIG_LPC) += x86/lpc_mmx.o
|
|
YASM-OBJS-$(CONFIG_PNG_DECODER) += x86/pngdsp.o
|
|
MMX-OBJS-$(CONFIG_PNG_DECODER) += x86/pngdsp-init.o
|
|
YASM-OBJS-$(CONFIG_PRORES_DECODER) += x86/proresdsp.o
|
|
MMX-OBJS-$(CONFIG_PRORES_DECODER) += x86/proresdsp-init.o
|
|
MMX-OBJS-$(CONFIG_AAC_DECODER) += x86/sbrdsp_init.o
|
|
YASM-OBJS-$(CONFIG_AAC_DECODER) += x86/sbrdsp.o
|
|
MMX-OBJS-$(CONFIG_DWT) += x86/snowdsp_mmx.o
|
|
MMX-OBJS-$(CONFIG_VC1_DECODER) += x86/vc1dsp_mmx.o
|
|
YASM-OBJS-$(CONFIG_VP3_DECODER) += x86/vp3dsp.o
|
|
YASM-OBJS-$(CONFIG_VP5_DECODER) += x86/vp3dsp.o
|
|
MMX-OBJS-$(CONFIG_VP5_DECODER) += x86/vp56dsp_init.o
|
|
YASM-OBJS-$(CONFIG_VP6_DECODER) += x86/vp3dsp.o \
|
|
x86/vp56dsp.o
|
|
MMX-OBJS-$(CONFIG_VP6_DECODER) += x86/vp56dsp_init.o
|
|
YASM-OBJS-$(CONFIG_VP8_DECODER) += x86/vp8dsp.o
|
|
MMX-OBJS-$(CONFIG_VP8_DECODER) += x86/vp8dsp-init.o
|
|
MMX-OBJS-$(HAVE_YASM) += x86/dsputil_yasm.o \
|
|
x86/deinterlace.o \
|
|
x86/fmtconvert.o \
|
|
x86/h264_qpel_10bit.o \
|
|
$(YASM-OBJS-yes)
|
|
|
|
MMX-OBJS-$(CONFIG_FFT) += x86/fft.o
|
|
|
|
OBJS-$(HAVE_MMX) += x86/dsputil_mmx.o \
|
|
x86/fdct_mmx.o \
|
|
x86/fmtconvert_mmx.o \
|
|
x86/idct_mmx_xvid.o \
|
|
x86/idct_sse2_xvid.o \
|
|
x86/motion_est_mmx.o \
|
|
x86/mpegvideo_mmx.o \
|
|
x86/simple_idct_mmx.o \
|
|
|
|
OBJS-$(CONFIG_XMM_CLOBBER_TEST) += x86/w64xmmtest.o
|