This changes getARMFloatABI to use the ToolChain and Args instead of Driver,
Args, Triple. Although this pushes the Triple calculation/parsing into the
function itself, it enables the use of the function for a future change. The
reason to sink the triple calculation here is to avoid threading the Triple
through multiple layers in a future change.
llvm-svn: 248095
Rather than using re-calculating the effective triple, thread the already
calculated value down into AddARMTargetArgs. This avoids both recreating the
triple, as well as re-parsing the triple as it was already done in the previous
frame.
llvm-svn: 248094
128-bit vector integer sign extensions correctly lower to the pmovsx instructions even for debug builds.
This patch removes the builtins and reimplements the _mm_cvtepi*_epi* intrinsics __using builtin_shufflevector (to extract the bottom most subvector) and __builtin_convertvector (to actually perform the sign extension).
Differential Revision: http://reviews.llvm.org/D12835
llvm-svn: 248092
Now that we have fast vector CTPOP implementations we can use this to speed up vector CTTZ using the pattern (cttz(x) = ctpop((x & -x) - 1))
Additionally, for AVX512CD that provides lzcnt instructions we can use the pattern (cttz_undef(x) = (width - 1) - ctlz(x & -x))
Differential Revision: http://reviews.llvm.org/D12663
llvm-svn: 248091
If an import directive was put into wrong context, the error message was obscure,
complaining on misbalanced braces. To get more descriptive messages, annotation
tokens related to modules are processed where they must not be seen.
Differential Revision: http://reviews.llvm.org/D11844
llvm-svn: 248085
Use an enumeration for the Floating Point ABIs supported on MIPS. This is
replicating the ARM change to avoid string based tracking of the floating point
ABI. NFC.
llvm-svn: 248083
InputFile::parse() can be called in parallel with other calls of
the same function. By doing that, time to self-link improves from
741 ms to 654 ms or 12% faster.
This is probably the last low hanging fruit in terms of parallelism.
Input file parsing and symbol table insertion takes 450 ms in total.
If we want to optimize further, we probably have to parallelize
symbol table insertion using concurrent hashmap or something.
That's doable, but that's not easy, especially if you want to keep
the exact same semantics and linking order. I'm not going to do that
at least soon.
Anyway, compared to r248019 (the change before the first attempt for
parallelism), we achieved 36% performance improvement from 1022 ms
to 654 ms. MSVC linker takes 3.3 seconds to link the same program.
MSVC's ICF feature is very slow for some reason, but even if we
disable the feature, it still takes about 1.2 seconds.
Our number is probably good enough.
llvm-svn: 248078
The test events had worker indexes coming across as strings. I
want them to be ints. worker_index now comes across as an int in
the dicationary.
The optional type can be specified with:
--event-add-entries key=val[:type][,key2=val2[:type2]...]
The type piece may be 'int' at this time. That is all. Otherwise
it will be a string.
llvm-svn: 248066
Binary literals predate C++14, but they are listed as a C++14 extension since
this was the first time they were standardized in the language. Move the
warning into a subgroup so it can be selectively disabled when checking for
other C++14 features.
llvm-svn: 248064
The failure mode when one gets this wrong is quite gnarly to then walk oneself out of, and if you can't actually find the library, trying to build against it is fairly pointless anyway
This + my previous skip_if_library_missing change should make running the libc++ tests on a Linux machine without it much more seamless
llvm-svn: 248057
Rui Ueyama