Statistics
| Branch: | Revision:

ffmpeg / doc / ffmpeg_powerpc_performance_evaluation_howto.txt @ c08b9856

History | View | Annotate | Download (5.63 KB)

1
FFmpeg & evaluating performance on the PowerPC Architecture HOWTO
2

    
3
(c) 2003-2004 Romain Dolbeau <romain@dolbeau.org>
4

    
5

    
6

    
7
I - Introduction
8

    
9
The PowerPC architecture and its SIMD extension AltiVec offer some
10
interesting tools to evaluate performance and improve the code.
11
This document tries to explain how to use those tools with FFmpeg.
12

    
13
The architecture itself offers two ways to evaluate the performance of
14
a given piece of code:
15

    
16
1) The Time Base Registers (TBL)
17
2) The Performance Monitor Counter Registers (PMC)
18

    
19
The first ones are always available, always active, but they're not very
20
accurate: the registers increment by one every four *bus* cycles. On
21
my 667 Mhz tiBook (ppc7450), this means once every twenty *processor*
22
cycles. So we won't use that.
23

    
24
The PMC are much more useful: not only can they report cycle-accurate
25
timing, but they can also be used to monitor many other parameters,
26
such as the number of AltiVec stalls for every kind of instruction,
27
or instruction cache misses. The downside is that not all processors
28
support the PMC (all G3, all G4 and the 970 do support them), and
29
they're inactive by default - you need to activate them with a
30
dedicated tool. Also, the number of available PMC depends on the
31
procesor: the various 604 have 2, the various 75x (aka. G3) have 4,
32
and the various 74xx (aka G4) have 6.
33

    
34
*WARNING*: The PowerPC 970 is not very well documented, and its PMC
35
registers are 64 bits wide. To properly notify the code, you *must*
36
tune for the 970 (using --tune=970), or the code will assume 32 bit
37
registers.
38

    
39

    
40
II - Enabling FFmpeg PowerPC performance support
41

    
42
This needs to be done by hand. First, you need to configure FFmpeg as
43
usual, but add the "--powerpc-perf-enable" option. For instance:
44

    
45
#####
46
./configure --prefix=/usr/local/ffmpeg-svn --cc=gcc-3.3 --tune=7450 --powerpc-perf-enable
47
#####
48

    
49
This will configure FFmpeg to install inside /usr/local/ffmpeg-svn,
50
compiling with gcc-3.3 (you should try to use this one or a newer
51
gcc), and tuning for the PowerPC 7450 (i.e. the newer G4; as a rule of
52
thumb, those at 550Mhz and more). It will also enable the PMC.
53

    
54
You may also edit the file "config.h" to enable the following line:
55

    
56
#####
57
// #define ALTIVEC_USE_REFERENCE_C_CODE 1
58
#####
59

    
60
If you enable this line, then the code will not make use of AltiVec,
61
but will use the reference C code instead. This is useful to compare
62
performance between two versions of the code.
63

    
64
Also, the number of enabled PMC is defined in "libavcodec/ppc/dsputil_ppc.h":
65

    
66
#####
67
#define POWERPC_NUM_PMC_ENABLED 4
68
#####
69

    
70
If you have a G4 CPU, you can enable all 6 PMC. DO NOT enable more
71
PMC than available on your CPU!
72

    
73
Then, simply compile FFmpeg as usual (make && make install).
74

    
75

    
76

    
77
III - Using FFmpeg PowerPC performance support
78

    
79
This FFmeg can be used exactly as usual. But before exiting, FFmpeg
80
will dump a per-function report that looks like this:
81

    
82
#####
83
PowerPC performance report
84
 Values are from the PMC registers, and represent whatever the
85
 registers are set to record.
86
 Function "gmc1_altivec" (pmc1):
87
        min: 231
88
        max: 1339867
89
        avg: 558.25 (255302)
90
 Function "gmc1_altivec" (pmc2):
91
        min: 93
92
        max: 2164
93
        avg: 267.31 (255302)
94
 Function "gmc1_altivec" (pmc3):
95
        min: 72
96
        max: 1987
97
        avg: 276.20 (255302)
98
(...)
99
#####
100

    
101
In this example, PMC1 was set to record CPU cycles, PMC2 was set to
102
record AltiVec Permute Stall Cycles, and PMC3 was set to record AltiVec
103
Issue Stalls.
104

    
105
The function "gmc1_altivec" was monitored 255302 times, and the
106
minimum execution time was 231 processor cycles. The max and average
107
aren't much use, as it's very likely the OS interrupted execution for
108
reasons of its own :-(
109

    
110
With the exact same settings and source file, but using the reference C
111
code we get:
112

    
113
#####
114
PowerPC performance report
115
 Values are from the PMC registers, and represent whatever the
116
 registers are set to record.
117
 Function "gmc1_altivec" (pmc1):
118
        min: 592
119
        max: 2532235
120
        avg: 962.88 (255302)
121
 Function "gmc1_altivec" (pmc2):
122
        min: 0
123
        max: 33
124
        avg: 0.00 (255302)
125
 Function "gmc1_altivec" (pmc3):
126
        min: 0
127
        max: 350
128
        avg: 0.03 (255302)
129
(...)
130
#####
131

    
132
592 cycles, so the fastest AltiVec execution is about 2.5x faster than
133
the fastest C execution in this example. It's not perfect but it's not
134
bad (well I wrote this function so I can't say otherwise :-).
135

    
136
Once you have that kind of report, you can try to improve things by
137
finding what goes wrong and fixing it; in the example above, one
138
should try to diminish the number of AltiVec stalls, as this *may*
139
improve performance.
140

    
141

    
142

    
143
IV) Enabling the PMC in Mac OS X
144

    
145
This is easy. Use "Monster" and "monster". Those tools come from
146
Apple's CHUD package, and can be found hidden in the developer web
147
site & FTP site. "MONster" is the graphical application, use it to
148
generate a config file specifying what each register should
149
monitor. Then use the command-line application "monster" to use that
150
config file, and enjoy the results.
151

    
152
Note that "MONster" can be used for many other things, but it's
153
documented by Apple, it's not my subject.
154

    
155
If you are using CHUD 4.4.2 or later, you'll notice that MONster is
156
no longer available. It's been superseeded by Shark, where
157
configuration of PMCs is available as a plugin.
158

    
159

    
160

    
161
V) Enabling the PMC on Linux
162

    
163
On linux you may use oprofile from http://oprofile.sf.net, depending on the
164
version and the cpu you may need to apply a patch[1] to access a set of the
165
possibile counters from the userspace application. You can always define them
166
using the kernel interface /dev/oprofile/* .
167

    
168
[1] http://dev.gentoo.org/~lu_zero/development/oprofile-g4-20060423.patch
169

    
170
--
171
Romain Dolbeau <romain@dolbeau.org>
172
Luca Barbato <lu_zero@gentoo.org>