The tables in the BSD tolower/toupper are slower for ASCII than just
doing the bit twiddling.
We can't actually remove the tables on LP32, so move them into the
"cruft" we keep around for backwards compatibility (but remove them for
LP64 where they were never exposed).
I noticed that the new bit-twiddling tolower(3) was performing better
on arm64 than toupper(3). The 0xdf constant was requiring an extra MOV,
and there isn't a BIC that takes an immediate value. Since we've already
done the comparison to check that we're in the right range (where the
bit is always set), though, we can EOR 0x20 to get the same result as
the missing BIC 0x20 in just one instruction.
I've applied that same optimization to towupper(3) too.
Before:
BM_ctype_tolower_n 3.30 ns 3.30 ns 212353035
BM_ctype_tolower_y 3.31 ns 3.30 ns 211234204
BM_ctype_toupper_n 3.30 ns 3.29 ns 214161246
BM_ctype_toupper_y 3.29 ns 3.28 ns 207643473
BM_wctype_towupper_ascii_n 3.53 ns 3.53 ns 195944444
BM_wctype_towupper_ascii_y 3.48 ns 3.48 ns 199233248
After:
BM_ctype_tolower_n 2.93 ns 2.92 ns 242373703
BM_ctype_tolower_y 2.88 ns 2.87 ns 245365309
BM_ctype_toupper_n 2.93 ns 2.93 ns 243049353
BM_ctype_toupper_y 2.89 ns 2.89 ns 245072521
BM_wctype_towupper_ascii_n 3.34 ns 3.33 ns 212951912
BM_wctype_towupper_ascii_y 3.29 ns 3.29 ns 214651254
(Why do both the "y" and "n" variants speed up with the EOR
change? Because the compiler transforms the code so that we
unconditionally do the bit twiddling and then use CSEL to decide whether
or not to actually use the result.)
We also save 1028 bytes of data in the LP64 libc.so.
Test: ran the bionic benchmarks and tests
Change-Id: I7829339f8cb89a58efe539c2a01c51807413aa2d
There are places in frameworks and art code that directly included
private bionic header files. Move these files to the new platform
include files.
This change also moves the __get_tls.h header file to tls.h and includes
the tls defines header so that there is a single header that platform
code can use to get __get_tls and the defines.
Also, simplify the visibility rules for platform includes.
Bug: 141560639
Test: Builds and bionic unit tests pass.
Change-Id: I9e5e9c33fe8a85260f69823468bc9d340ab7a1f9
Merged-In: I9e5e9c33fe8a85260f69823468bc9d340ab7a1f9
(cherry picked from commit 44631c919a)
Just to sanity check that toupper/tolower and towupper/towlower are in
the same ballpark for the ASCII range.
Test: ran benchmarks
Change-Id: I1ddc3f4f4478b4075107831f27bf4d4b4a3bc5e8
This reverts commit 3bf27c86d1.
Disable the 32-bit glibc target. We're not interested in benchmarking that
target, and the bench_noop_nostl executable doesn't build when ASAN is
turned on, because the libclang_rt.asan_cxx-i386.a lib contains
ubsan_type_hash_itanium.cc.o, which needs __dynamic_cast and various
STL typeinfo variables. The equivalent libclang_rt.asan_cxx-x86_64.a
lib doesn't have ubsan_type_hash_itanium.cc.o in it.
Bug: http://b/141693636
Test: m bench_noop_nostl bench_noop bench_noop_static ASAN_OPTIONS=detect_leaks=0 SANITIZE_HOST=address
Change-Id: Id6de17e622682f3a166367ad670cba5bfa6aee47
The accompanying soong change causes sanitize attributes to be allowed on
cc_object targets and propagates sanitize attributes into dependencies. This
is problematic for the crt objects in sanitizer builds because everything
depends on them including the sanitizer runtime, so a circular dependency
would otherwise be created. Furthermore, some of the code in these objects
runs before sanitizer initialization so it is unlikely that sanitizing them
would work anyway. Therefore, disable sanitization on these objects.
Change-Id: I25380dfc8eed5db34b034ba127a9d6b5674032fa
Many of our benchmarks are basically just "call one function with a
fixed argument". We don't need to keep repeating all the boilerplate for
that.
This also ensures we don't forget the benchmark::DoNotOptimize, which --
in addition to being a good idea in general -- specifically solves the
problem with gettid benchmark and provides a more accurate result by
removing the indirection through a function pointer.
Test: ran benchmarks
Change-Id: Id67535243678cd0d48f51cf25141e2040da9af03
Specifically, test the wall-clock time for running:
* a C program that does nothing (i.e. no STL)
* a C++ program that does nothing
* a statically-linked C++ program
* toybox true (in /system/bin and /vendor/bin)
* mksh -c true (in /system/bin and /vendor/bin)
Test: bionic-spawn-benchmarks
Change-Id: I961850ec90004cac83088feab5783f4f27768be1
* Fix the path to bionic-benchmarks-glibc
* Add symlinks for the toybox symlink commands. Each symlink bypasses the
intermediate symlink in ${OUT}/system/bin and points to the final
toybox binary. Suppress a bunch of warnings by skipping symlinks for
non-existent files.
The new spawn benchmarks try to run /system/bin/true. (They also try to
run /vendor/bin/true and print an error.)
* Quote "$@"
* Use soong_ui.bash --dumpvars-mode to set a bunch of variables, rather
than get_build_var, which invokes Soong once per variable. This reduces
the "build/run-on-host.sh" runtime from 4s to 1.3s.
* build/run-on-host.sh isn't executable and is only useful when it's
sourced into another shell, so remove its shebang to reduce confusion.
Bug: none
Test: \
. build/envsetup.sh
lunch aosp_x86_64-userdebug
. bionic/build/run-on-host.sh
prepare MODULES-IN-bionic MODULES-IN-external-toybox
/system/bin/true
Change-Id: I59e9a6aca77d35b16bdf51759c5fc7e725bfc67c
A kernel update broke three clock_getres tests, so disable
them while the kernel is fixed.
Bug: 141515847
Test: All tests pass on cuttlefish.
Change-Id: I7db789b2b8ba0bc5f8b13bb06e85711031735925
Use generated linker config by default, but with some back up plan
(sys.linker.use_generated_config property). Linker config generator
still does not support non-treblelized devices and vndk-lite, so these
cases should be handled later.
Bug: 138920271
Test: m -j && atest passed
Test: Tested from cuttlefish
Change-Id: I39e9d089a82f9409eccdcaa4fb26660caf3f5779
