• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1Validing libc Assembler Routines
2================================
3This document describes how to verify incoming assembler libc routines.
4
5## Quick Start
6* First, benchmark the previous version of the routine.
7* Update the routine, run the bionic unit tests to verify the routine doesn't
8have any bugs. See the [Testing](#Testing) section for details about how to
9verify that the routine is being properly tested.
10* Rerun the benchmarks using the updated image that uses the code for
11the new routine. See the [Performance](#Performance) section for details about
12benchmarking.
13* Verify that unwind information for new routine looks sane. See the [Unwind Info](#unwind-info) section for details about how to verify this.
14
15When benchmarking, it's best to verify on the latest Pixel device supported.
16Make sure that you benchmark both the big and little cores to verify that
17there is no major difference in performance on each.
18
19Benchmark 64 bit memcmp:
20
21    /data/benchmarktest64/bionic-benchmarks/bionic-benchmarks --bionic_xml=string.xml --benchmark_filter=memcmp
22
23Benchmark 32 bit memcmp:
24
25    /data/benchmarktest/bionic-benchmarks/bionic-benchmarks --bionic_xml=string.xml --benchmark_filter=memcmp
26
27Locking to a specific cpu:
28
29    /data/benchmarktest/bionic-benchmarks/bionic-benchmarks --bionic_cpu=2 --bionic_xml=string.xml --benchmark_filter=memcmp
30
31## Performance
32The bionic benchmarks are used to verify the performance of changes to
33routines. For most routines, there should already be benchmarks available.
34
35Building
36--------
37The bionic benchmarks are not built by default, they must be built separately
38and pushed on to the device. The commands below show how to do this.
39
40    mmma -j bionic/benchmarks
41    adb sync data
42
43Running
44-------
45There are two bionic benchmarks executables:
46
47    /data/benchmarktest64/bionic-benchmarks/bionic-benchmarks
48
49This is for 64 bit libc routines.
50
51    /data/benchmarktest/bionic-benchmarks/bionic-benchmarks
52
53This is for 32 bit libc routines.
54
55Here is an example of how the benchmark should be executed. For this
56command to work, you need to change directory to one of the above
57directories.
58
59    bionic-benchmarks --bionic_xml=string.xml --benchmark_filter=memcmp
60
61The last argument is the name of the one function that you want to
62benchmark.
63
64Almost all routines are already defined in the **string.xml** file in
65**bionic/benchmarks/suites**. Look at the examples in that file to see
66how to add a benchmark for a function that doesn't already exist.
67
68It can take a long time to run these tests since it attempts to test a
69large number of sizes and alignments.
70
71Results
72-------
73Bionic benchmarks is based on the [Google Benchmarks](https://github.com/google/benchmark)
74library. An example of the output looks like this:
75
76    Run on (8 X 1844 MHz CPU s)
77    CPU Caches:
78      L1 Data 32K (x8)
79      L1 Instruction 32K (x8)
80      L2 Unified 512K (x2)
81    ***WARNING*** CPU scaling is enabled, the benchmark real time measurements may be noisy and will incur extra overhead.
82    -------------------------------------------------------------------------------------------
83    Benchmark                                                    Time           CPU Iterations
84    -------------------------------------------------------------------------------------------
85    BM_string_memcmp/1/0/0                                       6 ns          6 ns  120776418   164.641MB/s
86    BM_string_memcmp/1/1/1                                       6 ns          6 ns  120856788   164.651MB/s
87
88The smaller the time, the better the performance.
89
90Caveats
91-------
92When running the benchmarks, CPU scaling is not normally enabled. This means
93that if the device does not get up to the maximum cpu frequency, the results
94can vary wildly. It's possible to lock the cpu to the maximum frequency, but
95is beyond the scope of this document. However, most of the benchmarks max
96out the cpu very quickly on Pixel devices, and don't affect the results.
97
98Another potential issue is that the device can overheat when running the
99benchmarks. To avoid this, you can run the device in a cool environment,
100or choose a device that is less likely to overheat. To detect these kind
101of issues, you can run a subset of the tests again. At the very least, it's
102always a good idea to rerun the suite a couple of times to verify that
103there isn't a high variation in the numbers.
104
105If you want to verify a single benchmark result, you can run a single test
106using a command like this:
107
108    bionic-benchmarks --bionic_xml=string.xml --benchmark_filter=BM_string_memcmp/1/1/0
109
110Where the argument to the filter argument is the name of the benchmark from
111the output. Sometimes this filter can still match multiple benchmarks, to
112guarantee that you only run the single benchmark, you can execute the benchmark
113like so:
114
115    bionic-benchmarks --bionic_xml=string.xml --benchmark_filter=BM_string_memcmp/1/1/0$
116
117NOTE: It is assumed that these commands are executed in adb as the shell user
118on device. If you are trying to run this using adb directly from a host
119machine, you might need to escape the special shell characters such as **$**.
120
121## Testing
122
123Run the bionic tests to verify that the new routines are valid. However,
124you should verify that there is coverage of the new routines. This is
125especially important if this is the first time a routine is assembler.
126
127Caveats
128-------
129When verifying an assembler routine that operates on buffer data (such as
130memcpy/strcpy), it's important to verify these corner cases:
131
132* Verify the routine does not read past the end of the buffers. Many
133assembler routines optimize by reading multipe bytes at a time and can
134read past the end. This kind of bug results in an infrequent and difficult to
135diagnosis crash.
136* Verify the routine handles unaligned buffers properly. Usually, a failure
137can result in an unaligned exception.
138* Verify the routine handles different sized buffers.
139
140If there are not sufficient tests for a new routine, there are a set of helper
141functions that can be used to verify the above corner cases. See the
142header **bionic/tests/buffer\_tests.h** for these routines and look at
143**bionic/tests/string\_test.cpp** for examples of how to use it.
144
145## Unwind Info
146It is also important to verify that the unwind information for these
147routines are properly set up. Here is a quick checklist of what to check:
148
149* Verify that all labels are of the format .LXXX, where XXX is any valid string
150for a label. If any other label is used, entries in the symbol table
151will be generated that include these labels. In that case, you will get
152an unwind with incorrect function information.
153* Verify that all places where pop/pushes or instructions that modify the
154sp in any way have corresponding cfi information. Along with this item,
155verify that when registers are pushed on the stack that there is cfi
156information indicating how to get the register.
157* Verify that only cfi directives are being used. This only matters for
158arm32, where it's possible to use ARM specific unwind directives.
159
160This list is not meant to be exhaustive, but a minimal set of items to verify
161before submitting a new libc assembler routine. There are difficult
162to verify unwind cases, such as around branches, where unwind information
163can be drastically different for the target of the branch and for the
164code after a branch instruction.
165