Unfortunately we still need to build the NDK sysroot as non-NEON
since that's still supported, though it's no longer the default.
This reverts commit f4e0601c7e.
Test: treehugger
Test: rebuilt NDK sysroot and imported into the NDK, ran NDK tests
Bug: None
This enables using arch-dependent property values in modules defined in
android/.
Bug: 122332178
Test: Soong tests + TreeHugger
Change-Id: I89869e395fabf0e69f505b77eab8a4221384124e
Support Kryo 385 cpu variant which should be used for
Qualcomm Snapdragon 845.
Motivation:
kryo385 CPU *IS NOT* an Arm Cortex-A75.
Particular problem:
Cortex-A75 CPU must support optional ARMv8.2 dot product
feature; kryo385 CPU (Qualcomm Snapdragon 845) doesn't
support it.
Test: Builds when kryo385 is used as cpu variant.
Test: Pixel 3 boots to GUI with kryo385 as CPU variant
with this patch on AOSP 9.0.0 r16.
Bug: 119564566
Change-Id: I62ffb46b1977b48446c6c1ca1400b1b39f7a8457
Add arch. variant option to support cores that implement armv8-2a ISAs.
Test: set TARGET_2ND_ARCH_VARIANT := armv8-2a, build successful and 32bit app launch ok
BUG: 118414869
Change-Id: I755b8858726bd887068923123bad106aed7b1ec8
Clang is always used now, so we can remove all the GCC checks. Removing
GCC-specific configuration will happen in the next CL.
Test: m
Change-Id: I4835ecf6062159315d0dfb07b098e60bff033a8a
... and rename SanitizerRuntimeLibrary to LibclangRuntimeLibrary since
some of them are not sanitizers.
Bug: 29275768
Test: m
Change-Id: Id638adea42d22b02438fce26120aa155bea14d42
Bug: 78133793
Bug: 78242072
Test: Builds when using cortex-a75 as a target.
Change-Id: Ie3fbe40b15f4d89eeba0d630a82097122fc83b33
Merged-In: Ie3fbe40b15f4d89eeba0d630a82097122fc83b33
(cherry picked from commit 0612597a41)
* USE_CLANG_LLD is unedefined in current builds.
* When USE_CLANG_LLD is defined to 'true' or '1',
use clang's lld instead of ld or ld.gold.
* When lld is enabled:
* ld-only flags are not passed to 'lld'.
* location_packer is disabled.
* Use new lld's --pack-dyn-relocs=android.
* When lld does not work:
* In Android.mk files use LOCAL_USE_CLANG_LLD := false.
* In Android.bp files use use_clang_lld: false.
* Only arm, arm64, x86, and x86_64_devices have LLD flags;
all other hosts and targets do not call lld yet.
Bug: 73768157
Test: make checkbuild and boot
Change-Id: I06b8a1e868a600997a7e70fe05c299d751d23d5f
This reverts commit fa00dfda13.
Reason for revert: New Build Breakages in git_pi-dev-plus-aosp-without-vendor/aosp_walleye-userdebug and git_pi-dev-plus-aosp/aosp_crosshatch-userdebug
Change-Id: Iec9e695cccc3f206df67f0c01cfd4c2c3023dcec
Proposal for adding SHT_RELR sections in the generic-abi is at
https://groups.google.com/forum/#!topic/generic-abi/bX460iggiKg
Comparison with relocation_packer for aosp_marlin-userdebug:
$ du -B1 -s */out/target/product/marlin/system
996958208 base/out/target/product/marlin/system
1008340992 no_pack/out/target/product/marlin/system
997801984 sht_relr/out/target/product/marlin/system
base contains a build with current settings.
no_pack contains a build with relocation_packer disabled.
sht_relr contains a build with SHT_RELR sections enabled.
relocation_packer was saving 11,382,784 bytes (no_pack - base).
SHT_RELR sections are saving 10,539,008 bytes (no_pack - sht_relr).
Bug: None
Test: Built aosp_marlin-userdebug image, boots on device.
Test: Built aosp_x86_64-userdebug image, boots in emulator.
Change-Id: I448dc95baa4976b68ce2659b80498539426775f9
* This is a workaround of llvm LTO bug.
* Keep using current NDK include directory path.
NDKTriple is default ClangTriple but kept as GccTriple for arm device.
Bug: 72619014
Test: make checkbuild
Change-Id: I5dc63c99760325c60bc2da98fd6a3125cef7267d
This architecture only existed for unbundled use, but even the NDK is
removing support in their r17 release, so just remove support for it.
Test: build/soong/build_test.bash -only-soong
Change-Id: I4bd23babf567128d2d242cbdee3311abb198dd7c
Bug: http://b/69481735
While Clang supports -mcpu=kryo, the GNU assembler doesn't. Use
Cortex-a53 instead, which is close to Kryo. The current alternative of
using Krait causes problems when also using armv8, which is not
available in Krait.
Test: Build marlin with internal CL 3248394 cherry-picked.
Change-Id: I571739e6ab4f1b37fafb304ecad2865c9394e04f
This reverts commit 20823f95e6.
Reason for revert: caused a minor performance regression in some benchmarks.
Bug: 69470341
Test: builds
Change-Id: I1a852b5d7a2aa1d08ecb54617898f814cddd7600
-fomit-frame-pointer was only being set for arm and mips. Since
we always use unwind tables and not frame pointers to unwind, and
since ART generated code does not use frame pointers, just turn
off frame pointers everywhere to gain an extra register.
Bug: 68951394
Test: m checkbuild
Change-Id: I9237d486a0c0215cdafd96d66712082df0eba785
These flags were added for gcc, but are always stripped out when
compiling for clang. Since gcc is barely used, removed them.
Bug: 68855788
Bug: 68947919
Test: m checkbuild
Change-Id: Iae2bda9808dd9499848ce145ccdf71c4c490b80e
Move -fdata-sections and -fno-short-enums to global device flags.
-fdata-sections was not previously set on x86[_64], -fno-short-enums
was not previously set on mips[64].
Bug: 68855788
Test: m checkbuild
Change-Id: I68e64888d5414fc022366eb2b6c5cd92c28a5542
It was previously set on arm[64] and mips[64], this will cause it
to be set for x86[_64] too.
Bug: 68855788
Test: m checkbuild
Change-Id: I75af16e7d259963ad633cc664929144332bb435d
When the following lines are specified in the BoardConfig.mk
TARGET_2ND_ARCH_VARIANT := armv8-a
TARGET_2ND_CPU_VARIANT := generic
The resulted binaries won't be built using 32-bit armv8-a integer
division instructions. Instead, division routines are used because
the build system will set a compiler option "-march=armv7-a", and
hence overwiting the previous "-march=armv8-a", when
TARGET_2ND_CPU_VARIANT is set to generic (or unset).
Bug: 67446726
Test: lunch aosp_arm64; make -j; emulator # boot to home screen
Also verify -march=armv7-a is not specified in the compiler
command line and the resulted binary actually uses integer
division instructions.
Change-Id: I430687aa5a658d2ec9f325a66d849b4c8898c7d5
-fomit-frame-pointer was specified twice, once for arm and once for
thumb. Move it to the shared cflags.
Bug: 68855788
Test: m checkbuild
Change-Id: Iab1299c247808f1a2542b468084600b7c32996e8
Move ldflags that are specified for all devices into
deviceGlobalLdflags, and add them to linker.go:
-Wl,-z,noexecstack
-Wl,-z,relro
-Wl,-z,now
-Wl,--build-id=md5
-Wl,--warn-shared-textrel
-Wl,--fatal-warnings
-Wl,--no-undefined-version
Bug: 68855788
Test: m checkbuild
Change-Id: I82561b4189287d7638006f9e298c5151f9930c5e
Move cflags that are set on all devices to deviceGlobalCflags:
-fno-canonical-system-headers
-ffunction-sections
-funwind-tables
-fstack-protector-strong
-Wa,--noexecstack
-D_FORTIFY_SOURCE=2
-Wstrict-aliasing=2
-Werror=format-security
Bug: 68855788
Test: m checkbuild
Change-Id: Iefec689fdd2749013d0cc003b3abec674a85fb74
Move all the flags that are used everywhere (or should be used
everywhere) into cc/config/global.go:
-no-canonical-prefixes
-fno-exceptions
-Wno-multichar
-O2
-g
-fno-strict-aliasing
Also remove flags that are already in noOverrideGlobalCflags:
-Werror=pointer-to-int-cast
-Werror=int-to-pointer-cast
x86 and x86_64 were specifying -fstrict-aliasing, but that was
being overriden later by -fno-strict-aliasing, so remove it.
ARM devices still override -fno-strict-aliasing to -fstrict-aliasing
when using ARM (vs. Thumb) instruction set.
Bug: 68855788
Test: m checkbuild
Change-Id: Ia2b5891bdefb60f974ad92b4b84a8548c2dcc7dc
This reverts commit 80b27e4e20.
Fixed CTS build issue by adding -march=armv7-a in cts/tests/tests/os/jni/Android.mk since the library built here needs/uses instruction (swp) obsolete in ARMv8.
Bug: 64964151
Test: lunch aosp_arm64-userdebug; make -j cts
lunch aosp_x86-userdebug; make -j cts
lunch aosp_sailfish-userdebug; make -j cts
lunch aosp_bullhead-userdebug; make -j cts
Change-Id: I6ccafce043e101c769ad370a81ffa0eb62719b5e
Add armv8-a as a valid/supported 32-bit arm architecture variant. If
some known armv8-a core, like cortex-a53, cortex-a73, etc. is specified
as TARGET(_2ND)_CPU_VARIANT, the associated TARGET(_2ND)_ARCH_VARIANT
will be treated as if armv8-a is specified.
Bug: 62895439
Test: "bionic-unit-tests-static --gtest_filter=*strlen*" on Nexus 4
(armv7-krait), emulator (armv7), and sailfish (armv8-kryo).
The test binary for the first 2 is built with armv7-a as its
TARGET_ARCH_VARIANT; The test binary for the last is built with
armv8-a as its TARGET_2ND_ARCH_VARIANT.
TARGET(_2ND)_CPU_VARIANTs of both binaries are set to "generic".
Change-Id: I53bee2974346cf485b2da54cf7aad010b0425910
Gold recognizes a different set of emulation strings to specify targets
than the bfd linker accepts. Clang only passes bfd emulations to the
linker, so we need to override these with the versions accepted by gold
when using gold.
Gold normally picks up the correct target from the first ELF input file
if it cannot parse the emulation parameter. However when using LTO, all
objects may be bitcode files, which causes gold to rely on the passed
--oformat or -m parameter to determine the proper target. If gold cannot
parse the emulation string passed by clang, it fails to link.
Test: build LTO version of libc
Change-Id: I38e78bb912fd3cc5fb7b4a762284f50ddd4f3998
* Cortex-A15 supports VFPv4 instructions
Test: make otapackage
Test: Built angler with 32 bit variant set to cortex-a15.
Test: Boot this version of the angler, ran bionic unit tests.
Change-Id: I94f00c9486ece201aa2c230308b9f90b0209dc4a
Also, update flags for cortex-a53.
Bug: 37647380
Test: Built target with cortex-a73 for both 32 bit/64 bit.
Test: Built target with cortex-a53.cortex-a57 for both 32 bit/64 bit.
Test: Ran bionic unit tests, ran art target tests.
(cherry picked from commit 1d9aa26d44)
Change-Id: I3325f60add7f424f8cca53c22919fc481ef5e787
Current modules must use armv7_a_neon to specify source files that
compile only with neon. If a future arch variant also supports neon,
all these modules will fall back to non-neon. Support a neon arch
feature that modules can use instead. Similarly, support dspr2 for
mips.
arm_device.go was also mixing armv7-a-neon with armv7_a_neon. Use
armv7-a-neon consistently, and fix the - to _ when creating the
property structs.
Test: m -j checkbuild
Change-Id: I24d3764280ab3bcbb9a73c0934edc9b99fc7f6a0
* GCC and Clang both support this, so we may as well simplify
the logic.
Test: make otapackage
Test: verified that the options are exactly the same, just slightly different order.
Change-Id: I8c2e9f3875cb662db708c799c9ce54f9fdd55981
Clang supports specific CPU targeting and optimization
for Kryo.
This switches us to using the specific Kryo targeting
when Clang is used. For other compilers, we fallback
to cortex-a57 targeting.
Also, move the replaceFirst function to a shared location.
Bug: 36728278
Test: Built and booted sailfish, ran bionic unit tests and art tests.
Test: Disassembled libc.so before this change and after and looked at
Test: the differences. Mostly the results were the order of instructions
Test: changing.
Test: Verified with the clang person (srhines) that our clang has this support
Test: and that it appears to be mostly instruction scheduling changes.
Change-Id: I4ee73d8bcc1e4f5eccb162c18937811fe199b16f
Signed-off-by: Alex Naidis <alex.naidis@linux.com>
* Kryo is closer to Krait than to cortex-a53, so choose Krait defaults.
* Made together with Jake Weinstein (xboxfanj)
Bug: 36728278
Test: Built and boot sailfish set up as kryo.
Change-Id: I04fa0a1dca6f97ae19202d28ee3ce8a59bf169b5
Signed-off-by: Alex Naidis <alex.naidis@linux.com>
and export the library name to make. Refactor the code a bit to avoid repeating the library name
multiple times.
Bug: 22033465
Test: Ran external/clang/build.py for aosp-llvm
Change-Id: I25eb3858eb92e1dd493b09524d559802551b2547
Luca Stefani