| commit | 65a6211f4d5aa9cf19e6f1d6a43aadb60583eb45 | [log] [tgz] |
|---|---|---|
| author | Adhemerval Zanella <adhemerval.zanella@linaro.org> | Fri Jun 22 15:30:03 2018 -0300 |
| committer | Haibo Huang <hhb@google.com> | Tue Aug 21 19:50:09 2018 +0000 |
| tree | c9805f240955fa1784c5ffab7928717931e4bef8 | |
| parent | 2019f962471d5789b23d037ede06f4c76e46cde6 [diff] |
[AArch64] Improve strncmp for mutually misaligned inputs
This patch was originally written by Siddhesh Poyarekar and pushed on
cortex-strings [1]. The mutually misaligned inputs on aarch64 are
compared with a simple byte copy, which is not very efficient.
This patch enhances the comparison similar to strcmp by loading a
double-word at a time.
Comparison on the default bionic and proposed optimized routines
shows the following performance improvements on A54 (using the
new proposed memcmp input data from test_strncmp.xml):
- No noticeable change on aligned inputs or with same alignment.
- Large improvements on unaligned inputs from sizes larger than
16 bytes.
Benchmark Time CPU Time Old Time New CPU Old CPU New
--------------------------------------------------------------------------------------------------------------------
BM_string_strncmp/1/0/0 -0.0954 -0.0954 19 17 19 17
BM_string_strncmp/2/0/0 -0.0344 -0.0344 19 18 19 18
BM_string_strncmp/3/0/0 +0.1768 +0.1768 15 18 15 18
BM_string_strncmp/4/0/0 -0.0344 -0.0344 19 18 19 18
BM_string_strncmp/5/0/0 -0.0344 -0.0344 19 18 19 18
BM_string_strncmp/6/0/0 +0.1589 +0.1589 15 18 15 18
BM_string_strncmp/7/0/0 -0.0344 -0.0344 19 18 19 18
BM_string_strncmp/8/0/0 -0.0998 -0.0998 19 17 19 17
BM_string_strncmp/9/0/0 -0.0277 -0.0277 23 22 23 22
BM_string_strncmp/10/0/0 -0.0270 -0.0270 23 22 23 22
BM_string_strncmp/11/0/0 -0.0331 -0.0331 23 22 23 22
BM_string_strncmp/12/0/0 -0.0270 -0.0270 23 22 23 22
BM_string_strncmp/13/0/0 -0.0284 -0.0284 23 22 23 22
BM_string_strncmp/14/0/0 +0.1042 +0.1042 20 22 20 22
BM_string_strncmp/15/0/0 -0.0277 -0.0277 23 22 23 22
BM_string_strncmp/16/0/0 +0.0214 +0.0215 22 22 22 22
BM_string_strncmp/24/0/0 -0.1291 -0.1291 24 21 24 21
BM_string_strncmp/32/0/0 -0.0470 -0.0470 27 26 27 26
BM_string_strncmp/40/0/0 -0.0433 -0.0433 29 28 29 28
BM_string_strncmp/48/0/0 -0.0301 -0.0301 31 30 31 30
BM_string_strncmp/56/0/0 -0.0800 -0.0800 33 31 33 31
BM_string_strncmp/64/0/0 +0.0188 +0.0188 34 34 34 34
BM_string_strncmp/72/0/0 -0.0334 -0.0334 38 37 38 37
BM_string_strncmp/80/0/0 -0.0000 -0.0000 40 40 40 40
BM_string_strncmp/88/0/0 +0.0413 +0.0413 61 64 61 64
BM_string_strncmp/96/0/0 -0.0215 -0.0216 69 67 69 67
BM_string_strncmp/104/0/0 -0.0208 -0.0208 72 70 72 70
BM_string_strncmp/112/0/0 -0.0173 -0.0173 75 74 75 74
BM_string_strncmp/120/0/0 -0.0166 -0.0166 78 77 78 77
BM_string_strncmp/128/0/0 -0.0158 -0.0158 81 80 81 80
BM_string_strncmp/136/0/0 -0.0149 -0.0149 84 83 84 83
BM_string_strncmp/144/0/0 -0.0201 -0.0201 88 86 88 86
BM_string_strncmp/160/0/0 -0.0136 -0.0136 94 93 94 93
BM_string_strncmp/176/0/0 +0.0224 +0.0224 96 98 96 98
BM_string_strncmp/192/0/0 +0.0289 +0.0289 102 105 102 105
BM_string_strncmp/208/0/0 +0.0101 +0.0101 111 112 111 112
BM_string_strncmp/224/0/0 -0.0107 -0.0107 119 118 119 118
BM_string_strncmp/240/0/0 -0.0088 -0.0088 126 125 126 125
BM_string_strncmp/256/0/0 -0.0101 -0.0101 132 131 132 131
BM_string_strncmp/512/0/0 -0.0056 -0.0056 235 233 235 233
BM_string_strncmp/1024/0/0 -0.0030 -0.0030 439 437 439 437
BM_string_strncmp/8192/0/0 -0.0431 -0.0431 3799 3635 3799 3635
BM_string_strncmp/16384/0/0 -0.0069 -0.0069 6778 6732 6779 6732
BM_string_strncmp/32768/0/0 -0.0001 -0.0002 13405 13403 13405 13403
BM_string_strncmp/65536/0/0 +0.0005 +0.0005 26968 26981 26968 26981
BM_string_strncmp/131072/0/0 -0.0057 -0.0057 53959 53650 53958 53650
BM_string_strncmp/1/4/0 -0.1352 -0.1352 12 10 12 10
BM_string_strncmp/2/4/0 +0.0020 +0.0020 15 15 15 15
BM_string_strncmp/3/4/0 -0.1560 -0.1560 20 17 20 17
BM_string_strncmp/4/4/0 +0.0296 +0.0296 22 22 22 22
BM_string_strncmp/5/4/0 +0.0573 +0.0573 22 23 22 23
BM_string_strncmp/6/4/0 -0.0340 -0.0340 25 24 25 24
BM_string_strncmp/7/4/0 +0.0185 +0.0185 26 26 26 26
BM_string_strncmp/8/4/0 -0.0050 -0.0050 27 27 27 27
BM_string_strncmp/9/4/0 -0.1294 -0.1294 28 24 28 24
BM_string_strncmp/10/4/0 +0.0109 +0.0109 29 29 29 29
BM_string_strncmp/11/4/0 -0.0000 -0.0001 30 30 30 30
BM_string_strncmp/12/4/0 +0.0055 +0.0055 50 50 50 50
BM_string_strncmp/13/4/0 -0.0249 -0.0249 51 50 51 50
BM_string_strncmp/14/4/0 -0.0289 -0.0289 53 52 53 52
BM_string_strncmp/15/4/0 -0.0205 -0.0205 55 54 55 54
BM_string_strncmp/16/4/0 -0.4616 -0.4616 57 31 57 31
BM_string_strncmp/24/4/0 -0.4871 -0.4871 72 37 72 37
BM_string_strncmp/32/4/0 -0.5549 -0.5549 87 39 87 39
BM_string_strncmp/40/4/0 -0.5964 -0.5964 103 42 103 42
BM_string_strncmp/48/4/0 -0.6647 -0.6647 118 40 118 40
BM_string_strncmp/56/4/0 -0.6551 -0.6551 134 46 134 46
BM_string_strncmp/64/4/0 -0.6609 -0.6609 145 49 145 49
BM_string_strncmp/72/4/0 -0.5709 -0.5710 164 70 164 70
BM_string_strncmp/80/4/0 -0.5929 -0.5929 180 73 180 73
BM_string_strncmp/88/4/0 -0.6051 -0.6051 195 77 195 77
BM_string_strncmp/96/4/0 -0.6160 -0.6160 210 81 210 81
BM_string_strncmp/104/4/0 -0.6199 -0.6199 223 85 223 85
BM_string_strncmp/112/4/0 -0.6293 -0.6293 240 89 240 89
BM_string_strncmp/120/4/0 -0.6439 -0.6439 255 91 255 91
BM_string_strncmp/128/4/0 -0.6493 -0.6493 271 95 271 95
BM_string_strncmp/136/4/0 -0.6704 -0.6704 287 95 287 95
BM_string_strncmp/144/4/0 -0.6744 -0.6744 302 98 302 98
BM_string_strncmp/160/4/0 -0.6700 -0.6700 333 110 333 110
BM_string_strncmp/176/4/0 -0.6821 -0.6821 364 116 364 116
BM_string_strncmp/192/4/0 -0.6887 -0.6887 394 123 394 123
BM_string_strncmp/208/4/0 -0.6949 -0.6949 425 130 425 130
BM_string_strncmp/224/4/0 -0.7069 -0.7069 456 134 456 134
BM_string_strncmp/240/4/0 -0.7042 -0.7042 486 144 486 144
BM_string_strncmp/256/4/0 -0.7043 -0.7043 514 152 514 152
BM_string_strncmp/1/0/4 +0.0227 +0.0227 14 14 14 14
BM_string_strncmp/2/0/4 +0.0442 +0.0442 15 16 15 16
BM_string_strncmp/3/0/4 +0.5829 +0.5829 17 27 17 27
BM_string_strncmp/4/0/4 -0.1593 -0.1593 22 19 22 19
BM_string_strncmp/5/0/4 -0.0516 -0.0516 23 22 23 22
BM_string_strncmp/6/0/4 -0.1684 -0.1684 25 20 25 20
BM_string_strncmp/7/0/4 +0.0170 +0.0170 26 26 26 26
BM_string_strncmp/8/0/4 +0.0006 +0.0006 27 27 27 27
BM_string_strncmp/9/0/4 +0.1272 +0.1272 25 28 25 28
BM_string_strncmp/10/0/4 +0.0108 +0.0108 29 29 29 29
BM_string_strncmp/11/0/4 -0.0001 -0.0001 30 30 30 30
BM_string_strncmp/12/0/4 -0.3557 -0.3557 50 32 50 32
BM_string_strncmp/13/0/4 -0.3370 -0.3370 51 34 51 34
BM_string_strncmp/14/0/4 -0.3444 -0.3444 53 35 53 35
BM_string_strncmp/15/0/4 +0.0946 +0.0946 51 56 51 56
BM_string_strncmp/16/0/4 -0.5203 -0.5203 53 25 53 25
BM_string_strncmp/24/0/4 -0.6109 -0.6109 72 28 72 28
BM_string_strncmp/32/0/4 -0.6934 -0.6934 88 27 88 27
BM_string_strncmp/40/0/4 -0.6833 -0.6833 103 33 103 33
BM_string_strncmp/48/0/4 -0.6973 -0.6973 118 36 118 36
BM_string_strncmp/56/0/4 -0.7116 -0.7116 134 39 134 39
BM_string_strncmp/64/0/4 -0.6017 -0.6018 149 59 149 59
BM_string_strncmp/72/0/4 -0.6268 -0.6268 164 61 164 61
BM_string_strncmp/80/0/4 -0.6409 -0.6409 179 64 179 64
BM_string_strncmp/88/0/4 -0.6465 -0.6465 195 69 195 69
BM_string_strncmp/96/0/4 -0.6551 -0.6551 210 72 210 72
BM_string_strncmp/104/0/4 -0.6662 -0.6662 227 76 227 76
BM_string_strncmp/112/0/4 -0.6700 -0.6700 240 79 240 79
BM_string_strncmp/120/0/4 -0.6740 -0.6740 256 83 256 83
BM_string_strncmp/128/0/4 -0.6862 -0.6862 271 85 271 85
BM_string_strncmp/136/0/4 -0.6883 -0.6883 287 89 287 89
BM_string_strncmp/144/0/4 -0.7031 -0.7031 297 88 297 88
BM_string_strncmp/160/0/4 -0.6985 -0.6985 333 100 333 100
BM_string_strncmp/176/0/4 -0.7082 -0.7082 364 106 364 106
BM_string_strncmp/192/0/4 -0.7223 -0.7223 396 110 396 110
BM_string_strncmp/208/0/4 -0.7135 -0.7135 421 121 421 121
BM_string_strncmp/224/0/4 -0.7194 -0.7194 455 128 455 128
BM_string_strncmp/240/0/4 -0.7233 -0.7233 487 135 487 135
BM_string_strncmp/256/0/4 -0.7239 -0.7239 516 143 516 143
BM_string_strncmp/1/4/4 +0.0224 +0.0225 21 22 21 22
BM_string_strncmp/2/4/4 -0.0001 -0.0001 22 22 22 22
BM_string_strncmp/3/4/4 -0.0001 -0.0001 22 22 22 22
BM_string_strncmp/4/4/4 -0.0435 -0.0435 22 21 22 21
BM_string_strncmp/5/4/4 -0.0118 -0.0118 27 27 27 27
BM_string_strncmp/6/4/4 -0.0118 -0.0118 27 27 27 27
BM_string_strncmp/7/4/4 -0.0117 -0.0117 27 27 27 27
BM_string_strncmp/8/4/4 -0.0118 -0.0118 27 27 27 27
BM_string_strncmp/9/4/4 -0.0117 -0.0117 27 27 27 27
BM_string_strncmp/10/4/4 +0.1447 +0.1447 23 27 23 27
BM_string_strncmp/11/4/4 -0.0062 -0.0062 27 27 27 27
BM_string_strncmp/12/4/4 -0.0454 -0.0454 28 27 28 27
BM_string_strncmp/13/4/4 -0.1507 -0.1507 29 24 29 24
BM_string_strncmp/14/4/4 -0.0003 -0.0003 29 29 29 29
BM_string_strncmp/15/4/4 -0.0002 -0.0003 29 29 29 29
BM_string_strncmp/16/4/4 +0.0047 +0.0047 29 29 29 29
BM_string_strncmp/24/4/4 -0.0104 -0.0104 31 30 31 30
BM_string_strncmp/32/4/4 -0.0290 -0.0290 33 32 33 32
BM_string_strncmp/40/4/4 -0.0189 -0.0189 34 33 34 33
BM_string_strncmp/48/4/4 -0.0059 -0.0059 36 36 36 36
BM_string_strncmp/56/4/4 +0.0000 +0.0000 39 39 39 39
BM_string_strncmp/64/4/4 +0.0000 +0.0000 42 42 42 42
BM_string_strncmp/72/4/4 +0.0000 +0.0000 45 45 45 45
BM_string_strncmp/80/4/4 +0.0391 +0.0392 65 68 65 68
BM_string_strncmp/88/4/4 -0.0090 -0.0090 71 70 71 70
BM_string_strncmp/96/4/4 -0.0034 -0.0034 74 74 74 74
BM_string_strncmp/104/4/4 -0.0482 -0.0482 77 73 77 73
BM_string_strncmp/112/4/4 +0.0387 +0.0387 77 80 77 80
BM_string_strncmp/120/4/4 -0.0072 -0.0073 84 83 84 83
BM_string_strncmp/128/4/4 -0.0071 -0.0071 87 86 87 86
BM_string_strncmp/136/4/4 +0.0366 +0.0366 86 89 86 89
BM_string_strncmp/144/4/4 -0.0068 -0.0068 93 93 93 93
BM_string_strncmp/160/4/4 -0.0064 -0.0064 100 99 100 99
BM_string_strncmp/176/4/4 -0.0063 -0.0063 106 105 106 105
BM_string_strncmp/192/4/4 -0.0012 -0.0012 112 112 112 112
BM_string_strncmp/208/4/4 -0.0098 -0.0098 119 118 119 118
BM_string_strncmp/224/4/4 -0.0050 -0.0050 125 125 125 125
BM_string_strncmp/240/4/4 -0.0060 -0.0060 132 131 132 131
BM_string_strncmp/256/4/4 -0.0046 -0.0046 138 137 138 137
[1] Commit id: 26cc4faec37a55529e5d0a39949f7b6ec81008f9
Test: bionic tests and benchmarks on aarch64.
Change-Id: Ied579d2044b4092fc95fad486af6541d1eb71dc3
See the additional documentation.
The C library. Stuff like fopen(3) and kill(2).
The math library. Traditionally Unix systems kept stuff like sin(3) and cos(3) in a separate library to save space in the days before shared libraries.
The dynamic linker interface library. This is actually just a bunch of stubs that the dynamic linker replaces with pointers to its own implementation at runtime. This is where stuff like dlopen(3) lives.
The C++ ABI support functions. The C++ compiler doesn't know how to implement thread-safe static initialization and the like, so it just calls functions that are supplied by the system. Stuff like __cxa_guard_acquire and __cxa_pure_virtual live here.
The dynamic linker. When you run a dynamically-linked executable, its ELF file has a DT_INTERP entry that says "use the following program to start me". On Android, that's either linker or linker64 (depending on whether it's a 32-bit or 64-bit executable). It's responsible for loading the ELF executable into memory and resolving references to symbols (so that when your code tries to jump to fopen(3), say, it lands in the right place).
The tests/ directory contains unit tests. Roughly arranged as one file per publicly-exported header file.
The benchmarks/ directory contains benchmarks, with its own documentation.
The first question you should ask is "should I add a libc wrapper for this system call?". The answer is usually "no".
The answer is "yes" if the system call is part of the POSIX standard.
The answer is probably "yes" if the system call has a wrapper in at least one other C library.
The answer may be "yes" if the system call has three/four distinct users in different projects, and there isn't a more specific library that would make more sense as the place to add the wrapper.
In all other cases, you should use syscall(3) instead.
Adding a system call usually involves:
__INTRODUCED_IN()../libc/tools/genversion-scripts.py.As mentioned above, this is currently a two-step process:
Note that if you're actually just trying to expose device-specific headers to build your device drivers, you shouldn't modify bionic. Instead use TARGET_DEVICE_KERNEL_HEADERS and friends described in config.mk.
This is fully automated (and these days handled by the libcore team, because they own icu, and that needs to be updated in sync with bionic):
If you make a change that is likely to have a wide effect on the tree (such as a libc header change), you should run make checkbuild. A regular make will not build the entire tree; just the minimum number of projects that are required for the device. Tests, additional developer tools, and various other modules will not be built. Note that make checkbuild will not be complete either, as make tests covers a few additional modules, but generally speaking make checkbuild is enough.
The tests are all built from the tests/ directory.
$ mma # In $ANDROID_ROOT/bionic.
$ adb root && adb remount && adb sync
$ adb shell /data/nativetest/bionic-unit-tests/bionic-unit-tests
$ adb shell \
/data/nativetest/bionic-unit-tests-static/bionic-unit-tests-static
# Only for 64-bit targets
$ adb shell /data/nativetest64/bionic-unit-tests/bionic-unit-tests
$ adb shell \
/data/nativetest64/bionic-unit-tests-static/bionic-unit-tests-static
Note that we use our own custom gtest runner that offers a superset of the options documented at https://github.com/google/googletest/blob/master/googletest/docs/AdvancedGuide.md#running-test-programs-advanced-options, in particular for test isolation and parallelism (both on by default).
Most of the unit tests are executed by CTS. By default, CTS runs as a non-root user, so the unit tests must also pass when not run as root. Some tests cannot do any useful work unless run as root. In this case, the test should check getuid() == 0 and do nothing otherwise (typically we log in this case to prevent accidents!). Obviously, if the test can be rewritten to not require root, that's an even better solution.
Currently, the list of bionic CTS tests is generated at build time by running a host version of the test executable and dumping the list of all tests. In order for this to continue to work, all architectures must have the same number of tests, and the host version of the executable must also have the same number of tests.
Running the gtests directly is orders of magnitude faster than using CTS, but in cases where you really have to run CTS:
$ make cts # In $ANDROID_ROOT.
$ adb unroot # Because real CTS doesn't run as root.
# This will sync any *test* changes, but not *code* changes:
$ cts-tradefed \
run singleCommand cts --skip-preconditions -m CtsBionicTestCases
The host tests require that you have lunched either an x86 or x86_64 target. Note that due to ABI limitations (specifically, the size of pthread_mutex_t), 32-bit bionic requires PIDs less than 65536. To enforce this, set /proc/sys/kernel/pid_max to 65536.
$ ./tests/run-on-host.sh 32 $ ./tests/run-on-host.sh 64 # For x86_64-bit *targets* only.
You can supply gtest flags as extra arguments to this script.
As a way to check that our tests do in fact test the correct behavior (and not just the behavior we think is correct), it is possible to run the tests against the host's glibc.
$ ./tests/run-on-host.sh glibc
For either host or target coverage, you must first:
$ export NATIVE_COVERAGE=truebionic_coverage=true in libc/Android.mk and libm/Android.mk.$ mma
$ adb sync
$ adb shell \
GCOV_PREFIX=/data/local/tmp/gcov \
GCOV_PREFIX_STRIP=`echo $ANDROID_BUILD_TOP | grep -o / | wc -l` \
/data/nativetest/bionic-unit-tests/bionic-unit-tests
$ acov
acov will pull all coverage information from the device, push it to the right directories, run lcov, and open the coverage report in your browser.
First, build and run the host tests as usual (see above).
$ croot $ lcov -c -d $ANDROID_PRODUCT_OUT -o coverage.info $ genhtml -o covreport coverage.info # or lcov --list coverage.info
The coverage report is now available at covreport/index.html.
Bionic's test runner will run each test in its own process by default to prevent tests failures from impacting other tests. This also has the added benefit of running them in parallel, so they are much faster.
However, this also makes it difficult to run the tests under GDB. To prevent each test from being forked, run the tests with the flag --no-isolate.
See 32-bit ABI bugs.