[FFmpeg-devel] Profiling the library
Thorsten Jordan
tjordan
Thu May 10 15:37:31 CEST 2007
Hello,
i got some more information out of oprofile. The values depend of course
on the input, but maybe it is interesting to the developers as well.
Here is an excerpt. If this is too off-topic or uninteresting, please
tell me - i am new to the list and don't want to be known as spammer.
Discovery HD material 1920x1088, ca. 17.5mbit for Video:
CPU: P4 / Xeon with 2 hyper-threads, speed 3066.9 MHz (estimated)
Counted GLOBAL_POWER_EVENTS events (time during which processor is not
stopped) with a unit mask of 0x01 (mandatory) count 100000
samples % app name symbol name
641419 16.3455 no-vmlinux (no symbols)
480138 12.2355 libc-2.5.so (no symbols)
474999 12.1046 libavcodec.so decode_cabac_residual
241597 6.1567 libavcodec.so decode_mb_cabac
181427 4.6234 libavcodec.so h264_h_loop_filter_luma_mmx2
161359 4.1120 libavcodec.so fill_caches
129169 3.2917 libavcodec.so hl_decode_mb_simple
118638 3.0233 libavcodec.so put_h264_chroma_mc8_mmx
100001 2.5484 libavcodec.so h264_h_loop_filter_chroma_mmx2
94209 2.4008 libavcodec.so h264_v_loop_filter_luma_mmx2
79825 2.0342 libavcodec.so put_h264_qpel8_h_lowpass_l2_mmx2
75998 1.9367 libavcodec.so put_h264_qpel8or16_v_lowpass_mmx2
74051 1.8871 libavcodec.so filter_mb_fast
55606 1.4170 libavcodec.so decode_cabac_mb_mvd
51846 1.3212 libavcodec.so put_h264_qpel16_mc00_mmx2
45264 1.1535 libavcodec.so h264_loop_filter_strength_mmx2
42432 1.0813 libavcodec.so get_cabac_noinline
41434 1.0559 oprofiled (no symbols)
39429 1.0048 libavcodec.so h264_v_loop_filter_chroma_mmx2
36018 0.9179 libavcodec.so ff_h264_idct_add_mmx
35390 0.9019 Xorg (no symbols)
34774 0.8862 libavcodec.so filter_mb_edgev
34472 0.8785 libavcodec.so filter_mb_edgeh
31097 0.7925 libavcodec.so mc_part
28953 0.7378 libavcodec.so clear_blocks_mmx
25934 0.6609 libavcodec.so hl_motion
25625 0.6530 libavcodec.so h264_parse
Anixe HD material 1920x1088, ca. 11.2mbit for Video
CPU: P4 / Xeon with 2 hyper-threads, speed 3066.9 MHz (estimated)
Counted GLOBAL_POWER_EVENTS events (time during which processor is not
stopped) with a unit mask of 0x01 (mandatory) count 100000
samples % app name symbol name
1729288 17.8433 libavcodec.so decode_cabac_residual
1241987 12.8152 libc-2.5.so (no symbols)
990787 10.2232 libavcodec.so decode_mb_cabac
864689 8.9221 no-vmlinux (no symbols)
618225 6.3790 libavcodec.so put_h264_qpel8or16_v_lowpass_mmx2
594189 6.1310 libavcodec.so put_h264_chroma_mc8_mmx
494330 5.1006 libavcodec.so put_h264_qpel8_h_lowpass_l2_mmx2
484531 4.9995 libavcodec.so fill_caches
399129 4.1183 libavcodec.so hl_decode_mb_simple
389900 4.0231 libavcodec.so decode_cabac_mb_mvd
180724 1.8648 libavcodec.so get_cabac_noinline
149591 1.5435 libavcodec.so clear_blocks_mmx
130306 1.3445 libavcodec.so mc_part
113568 1.1718 libavcodec.so hl_motion
102844 1.0612 oprofiled (no symbols)
97346 1.0044 libavcodec.so ff_h264_idct_add_mmx
92063 0.9499 libavcodec.so put_h264_qpel8_h_lowpass_mmx2
82303 0.8492 libavcodec.so ff_h264_idct_dc_add_mmx2
76850 0.7930 libavcodec.so put_h264_qpel16_mc00_mmx2
55097 0.5685 Xorg (no symbols)
51270 0.5290 libavcodec.so prefetch_mmx2
47297 0.4880 libavcodec.so decode_slice
38728 0.3996 libavcodec.so draw_edges_mmx
33169 0.3422 libX11.so.6.2.0 (no symbols)
30026 0.3098 libpthread-2.5.so pthread_mutex_lock
25970 0.2680 libglib-1.2.so.0.0.10 (no symbols)
24621 0.2540 libavcodec.so h264_parse
Interesting: decode_cabac_residual uses more cpu on the stream with
lower bitrate.
--
Mit freundlichen Gr??en, Thorsten
More information about the ffmpeg-devel
mailing list