Summary:
By manipulating a local variable in the loop, when the loop can
be optimized away (due to no non-trivial destructors), this lets
it be fully optimized away and we modify the __end_ separately.
This results in a substantial improvement in the generated code.
Prior to this change, this would be generated (on x86_64):
movq (%rdi), %rdx
movq 8(%rdi), %rcx
cmpq %rdx, %rcx
je LBB2_2
leaq -12(%rcx), %rax
subq %rdx, %rax
movabsq $-6148914691236517205, %rdx ## imm = 0xAAAAAAAAAAAAAAAB
mulq %rdx
shrq $3, %rdx
notq %rdx
leaq (%rdx,%rdx,2), %rax
leaq (%rcx,%rax,4), %rax
movq %rax, 8(%rdi)
And after:
movq (%rdi), %rax
movq %rax, 8(%rdi)
This brings this in line with what other implementations do.
Subscribers: cfe-commits
Differential Revision: https://reviews.llvm.org/D25241
git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@298601 91177308-0d34-0410-b5e6-96231b3b80d8
Summary: This is my attempt to work around the C1XX bug described to me by @BillyONeal.
Reviewers: BillyONeal, STL_MSFT, CaseyCarter
Reviewed By: BillyONeal
Subscribers: cfe-commits, BillyONeal
Differential Revision: https://reviews.llvm.org/D31260
git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@298554 91177308-0d34-0410-b5e6-96231b3b80d8
The tests for libc++ specify -target on the command-line to the
compiler, but this is problematic for a few reasons.
Firstly, the -target option isn't supported on Apple platforms. Parts
of the triple get dropped and ignored. Instead, software should be
compiled with a combination of the -arch and -m<name>-version-min
options.
Secondly, the generic "darwin" target references a kernel version
instead of a platform version. Each platform has its own independent
versions (with different versions of libc++.1.dylib), independent of the
version of the Darwin kernel.
This commit adds support to the LIT infrastructure for testing against
Apple platforms using -arch and -platform options.
If the host is not on OS X, or the compiler type is not clang or apple-clang, then this commit has NFC.
If the host is on OS X and --param=target_triple=... is specified, then a warning is emitted to use arch and platform instead. Besides the warning, there's NFC.
If the host is on OS X and *no* target-triple is specified, then use the new deployment target logic. This uses two new lit parameters, --param=arch=<arch> and --param=platform=<platform>. <platform> has the form <name>[<version>].
By default, arch is auto-detected from clang -dumpmachine, and platform is "macosx".
If the platform doesn't have a version:
For "macosx", the version is auto-detected from the host system using sw_vers. This may give a different version than the SDK, since new SDKs can be installed on older hosts.
Otherwise, the version is auto-detected from the SDK version using xcrun --show-sdk-path.
-arch <arch> -m<name>-version-min=<version> is added to the compiler flags.
The target triple is computed as <arch>-apple-<platform>. It is *not* passed to clang, but it is available for XFAIL and UNSUPPORTED (as is with_system_cxx_lib=<target>).
For convenience, apple-darwin and <arch>-apple-darwin are added to the set of available features.
There were a number of tests marked to XFAIL on x86_64-apple-darwin11
and x86_64-apple-darwin12. I updated these to
x86_64-apple-macosx10.7 and x86_64-apple-macosx10.8.
git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@297798 91177308-0d34-0410-b5e6-96231b3b80d8
Clang doesn't produce gcov compatible coverage files. This
causes lcov to break because it uses gcov by default. This
patch switches lcov to use llvm-cov as the gcov-tool.
Unfortunatly llvm-cov doesn't provide a gcov like interface by
default so it won't work with lcov. However `llvm-cov gcov` does.
For this reason we generate 'llvm-cov-wrapper' script that always
passes the gcov flag.
git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@297553 91177308-0d34-0410-b5e6-96231b3b80d8
basic_string::replace() has the below line
__sz += __n2 - __n1;
which fails overflow checks if __n1 > __n2, as the negative result
from the subtraction then overflows the original __sz when added to
it.
This behavior is valid as unsigned integer overflow is defined to wrap
around the maximum value and that produces the correct final value for
__sz. Therefore, we disable this check on this function.
git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@297355 91177308-0d34-0410-b5e6-96231b3b80d8
Put proper guards around _LIBCPP_METHOD_TEMPLATE_IMPLICIT_INSTANTIATION_VIS.
No functional change on non-Windows. Avoids incorrect macro redefinition
on Windows.
git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@297330 91177308-0d34-0410-b5e6-96231b3b80d8
