ELF subtarget.
The existing description string is moved from PPC64TargetInfo to its
DarwinTargetInfo subclass, to avoid any changes to the Darwin ABI.
PPC64TargetInfo now has two possible description strings: one for FreeBSD,
which requires 8-byte alignment, and a default string that requires
16-byte alignment.
I've added a test for PPC64 Linux to verify the 16-byte alignment. If
somebody wants to add a separate test for FreeBSD, that would be great.
Note that there is a companion patch to update the alignment information
in LLVM, which I am committing now as well.
llvm-svn: 166927
This code checks the ASM string to see if the output size is able to fit within
the variable specified as the output. For instance, scalar-to-vector conversions
may not really work. It's on by default, but can be turned off with a flag if
you think you know what you're doing.
This is placed under a flag ('-Wasm-operand-widths') and flag group ('-Wasm').
<rdar://problem/12284092>
llvm-svn: 166737
Because PNaCl bitcode must be target-independent, it uses some
different bitcode representations from other targets (e.g. byval and
sret for structures). This means that without additional type
information, it cannot meet some native ABI requirements for some
targets (e.g. passing structures containing unions by value on
x86-64). To allow generation of code which uses the correct native
ABIs, we also support triples such as x86_64-nacl, which uses
target-dependent IR (as opposed to le32-nacl, which uses byval and
sret).
To allow interoperation between the two types of code, this patch adds
a calling convention attribute to be used in code compiled with the
target-dependent triple, which will generate code using the le32-style
bitcode. This calling convention does not need to be explicitly
supported in the backend because it determines bitcode representation
rather than native conventions (the backend just needs to undersand
how to handle byval and sret for the Native Client OS).
This patch implements __attribute__((pnaclcall)) to generate calls in
bitcode according to the le32 bitcode conventions, an attribute which
is accepted by any Native Client target, but issues a warning
otherwise.
llvm-svn: 166065
AAPCS ABI Section 7.1.4 [1] specifies that va_list
should be defined as struct __va_list { void *__ap;};
And in C++, it is defined in namespace std.
[1] http://infocenter.arm.com/help/topic
/com.arm.doc.ihi0042d/IHI0042D_aapcs.pdf
Patch by Weiming Zhao.
llvm-svn: 165609
Most of the code guarded with ANDROIDEABI are not
ARM-specific, and having no relation with arm-eabi.
Thus, it will be more natural to call this
environment "Android" instead of "ANDROIDEABI".
Note: We are not using ANDROID because several projects
are using "-DANDROID" as the conditional compilation
flag.
llvm-svn: 163088
to define all macros for MIPS targets. Remove redundant virtual function
getArchDefines(). Two virtual functions for this task are really too much.
llvm-svn: 162853
Instead of adding it to each individual subclass in
Targets.cpp, simply check the appropriate target
values.
Where before it was only on x86_64 and ppc64, it's now
also defined on mips64 and nvptx64.
Also add a bunch of negative tests to ensure it is *not*
defined on any other architectures while we're here.
llvm-svn: 161685
AVX). Currently, if no aligned attribute is specified the alignment of a vector is
inferred from its size. Thus, very large vectors will be over-aligned with no
benefit. Target owners should set this target max.
llvm-svn: 160209
This functionality is based on what is done on ARM, and enables selecting PPC CPUs
in a way compatible with gcc's driver. Also, mirroring gcc (and what is done on x86),
-mcpu=native support was added. This uses the host cpu detection from LLVM
(which will also soon be updated by refactoring code currently in backend).
In order for this to work, the target needs a list of valid CPUs -- we now accept all CPUs accepted by LLVM.
A few preprocessor defines for common CPU types have been added.
llvm-svn: 158334
Because in CUDA types do not have associated address spaces,
globals are declared in their "native" address space, and accessed
by bitcasting the pointer to address space 0. This relies on address
space 0 being a unified address space.
llvm-svn: 157167
r155047. See the LLVM log for the primary motivation:
http://llvm.org/viewvc/llvm-project?rev=155047&view=rev
Primary commit r154828:
- Several issues were raised in review, and fixed in subsequent
commits.
- Follow-up commits also reverted, and which should be folded into the
original before reposting:
- r154837: Re-add the 'undef BUILTIN' thing to fix the build.
- r154928: Fix build warnings, re-add (and correct) header and
license
- r154937: Typo fix.
Please resubmit this patch with the relevant LLVM resubmission.
llvm-svn: 155048
uses Neon instructions for single-precision FP.
-mfpmath=neon is analogous to passing llc -mattr=+neonfp.
-mfpmath=[vfp|vfp2|vfp3|vfp4] is analogous to passing llc -mattr=-neonfp.
rdar://11108618
llvm-svn: 154046
ptrdiff_t on PPC32 on Linux, etc. should be int not long.
This does not matter for C, but it does matter for C++ because of
name mangling.
The preprocessor test has been changed accordingly.
llvm-svn: 151935
Unconditionally define __C99FEATURES__ when using C++ on Solaris. This is a
(hopefully temporary) work around for libc++ exposing C99-but-not-C++98
features in C++98 mode.
llvm-svn: 151889
each of the targets. Use this for module requirements, so that we can
pin the availability of certain modules to certain target features,
e.g., provide a module for xmmintrin.h only when SSE support is
available.
Use these feature names to provide a nearly-complete module map for
Clang's built-in headers. Only mm_alloc.h and unwind.h are missing,
and those two are fairly specialized at the moment. Finishes
<rdar://problem/10710060>.
llvm-svn: 149227
like Darwin that don't support it. We should also complain about
invalid -fvisibility=protected, but that information doesn't seem
to exist at the most appropriate time, so I've left a FIXME behind.
llvm-svn: 149186
ARM supports clz and ctz directly and both operations have well-defined
results for zero. There is no disadvantage in performance to using the
defined-at-zero versions of llvm.ctlz/cttz intrinsics. We're running into
ARM-specific code written with the assumption that __builtin_clz(0) == 32,
even though that value is technically undefined. The code is failing now
because of llvm optimizations that are taking advantage of the undef
behavior (specifically svn r147255). There's nothing wrong with that
optimization on x86 where any incorrect assumptions about __builtin_clz(0)
will quickly be exposed. For ARM, though, optimizations based on that undef
behavior are likely to cause subtle bugs. Other targets with defined-at-zero
clz/ctz support may want to override the default behavior as well.
llvm-svn: 149086
Patch from Jyotsna Verma:
I have made the changes to remove assertions in the Hexagon backend
specific clang driver. Instead of asserting on invalid arch name, it has
been modified to use the default value.
I have changed the implementation of the CPU flag validation for the
Hexagon backend. Earlier, the clang driver performed the check and
asserted on invalid inputs. In the new implementation, the driver passes
the last CPU flag (or sets to "v4" if not specified) to the compiler (and
also to the assembler and linker which perform their own check) instead of
asserting on incorrect values. This patch changes the setCPU function for
the Hexagon backend in clang/lib/Basic/Targets.cpp which causes the
compiler to error out on incorrect CPU flag values.
llvm-svn: 148139
- Support gcc-compatible vfpv3 name in addition to vfp3.
- Support vfpv3-d16.
- Disable neon feature for -mfpu=vfp* (yes, we were emitting Neon instructions
for those!).
llvm-svn: 147943
This is equal to alignof(std::max_align_t) on the platform and equal to the
alignment provided by malloc. (Platform owners please double-check your
platform's value.)
llvm-svn: 146762
. move compiler-rt to a separate directory so the -L argument only includes compiler-rt (thanks joerg)
. build all clang subdirs
. switches the Minix platform to ELF
. normalizes toolchain invocation
Patch by Ben Gras.
llvm-svn: 146206
* Enabling sse enables mmx.
* Disabling (-mno-mmx) mmx, doesn't disable sse (we got this right already).
* The order in not important. -msse -mno-mmx is the same as -mno-mmx -msse.
llvm-svn: 145194
Original behaviour of defining wchar_t as signed int has been kept for apcs-gnu as I don't have any spec for this to validate against.
llvm-svn: 145102
This only has an effect with fairly new binutils (2.21.51 or later). Other ELF targets probably want this as well, but on BSDs binutils is usually old so it doesn't matter.
llvm-svn: 142076
predefines based on the output of GCC as well as the CPU predefines.
Invert tests for __AVX__, Clang's AVX feature is hard coded off still.
Switch Atom from 'SSE3' to 'SSSE3'. This matches GCC's behavior, Intel's
documentation, and ICC's documentation (such as I could dig up).
Switch Athlon and Geode to enable 3dnowa rather than just 3dnow and
nothing (resp.).
llvm-svn: 140692
fallthrough now that we're working with a switch. Also remove a dubious
"feature" regarding k6 processors and 3dnow and leave a fixme... Not
that anyone is likely to care about correct tuning for k6 processors
with and w/o 3dnow...
llvm-svn: 140687
selected CPU model to the enumeration. This parses the string
representation once using a StringSwitch on SetCPU. It returns an error
for strings which are not recognized (yay!). Finally it replaces
ridiculous if-chains with switches that cover all enumerators.
The last change required adding several missing entries to the features
function. These were obvious on inspection. Yay for a pattern that gives
warnings when we miss one.
No new test cases yet, as I want to get the 64-bit errors working first.
I'll then start fleshing out the testing more. Currently I'm primarily
testing on Linux, but I'm hoping check whether there are interesting
differences on darwin before long...
llvm-svn: 140685
it an error if a CPU is provided for a target that doesn't implement
logic handling CPU settings, to match the ABI settings. It also removes
the CPU parameter from the getDefaultFeatures method. This parameter was
always filled in with the same value as setCPU was called with, and at
this point every single target implementation that referenced the CPU
within this function has needed to store the CPU via setCPU anyways in
order to implement other interface points.
llvm-svn: 140683
is *very* much a WIP that I'll be refining over the next several
commits, but I need to get this checkpoint in place for sanity.
This also adds a much more comprehensive test for architecture macros,
which is roughly generated by inspecting the behavior of a trunk build
of GCC. It still requires some massaging, but eventually I'll even check
in the script that generates these so that others can use it to append
more tests for more architectures, etc.
Next up is a bunch of simplification of the Targets.cpp code, followed
by a lot more test cases once we can reject invalid architectures.
llvm-svn: 140673
change __builtin_va_list to from a structure to int[4] (same alignment
and size, but with a simpler representation). Patch by David Meyer!
llvm-svn: 140144
language options. Use that .def file to declare the LangOptions class
and initialize all of its members, eliminating a source of annoying
initialization bugs.
AST serialization changes are next up.
llvm-svn: 139605
- wrong alignment for double (it was 4, but 8 is desired),
- added checks for _REENTRANT define,
- fixed the issue that defines were not tested (because the check for inside #ifdef).
llvm-svn: 138775
alignment. This fixes cases where the anonymous bitfield is followed by a
non-bitfield member. E.g.,
struct t4
{
int foo : 1;
long : 0;
char bar;
};
Part of rdar://9859156
llvm-svn: 136858
specified, 128 avx code is used and we're not sure yet if this the behavior
we want (and if it does, some improvements are needed before relying on it).
llvm-svn: 134939
Note that because we don't usually touch the MMX registers anyway, all -mno-mmx needs to do is tweak the x86-32 calling convention a little for vectors that look like MMX vectors, and prevent the definition of __MMX__.
clang doesn't actually stop the user from using MMX inline asm operands or MMX builtins in -mno-mmx mode; as a QOI issue, it would be nice to diagnose, but I doubt it really matters much.
<rdar://problem/9694837>
llvm-svn: 134770
change.
Previously clang was passing the following feature strings to the ARM backend
when CPU is cortex-a8: +neon,-vfp2,-vfp3
This used to work because -vfp2,-vfp3 had no effect after +neon. Now that the
features are controlled by individual bits (with implied hierarchy), the net
effect is all three features will be turned off.
llvm-svn: 134691
Language-design credit goes to a lot of people, but I particularly want
to single out Blaine Garst and Patrick Beard for their contributions.
Compiler implementation credit goes to Argyrios, Doug, Fariborz, and myself,
in no particular order.
llvm-svn: 133103
Sandeep Patel noticed that the alignment was wrong for Neon vector types,
and this change is partly derived from his patch. For the APCS ABI, however,
additional changes were required: the maximum ABI alignment is 32 bits and
the preferred alignment for i64 and f64 types should be 64 bits.
llvm-svn: 128825
which versions of an OS provide a certain facility. For example,
void foo()
__attribute__((availability(macosx,introduced=10.2,deprecated=10.4,obsoleted=10.6)));
says that the function "foo" was introduced in 10.2, deprecated in
10.4, and completely obsoleted in 10.6. This attribute ties in with
the deployment targets (e.g., -mmacosx-version-min=10.1 specifies that
we want to deploy back to Mac OS X 10.1). There are several concrete
behaviors that this attribute enables, as illustrated with the
function foo() above:
- If we choose a deployment target >= Mac OS X 10.4, uses of "foo"
will result in a deprecation warning, as if we had placed
attribute((deprecated)) on it (but with a better diagnostic)
- If we choose a deployment target >= Mac OS X 10.6, uses of "foo"
will result in an "unavailable" warning (in C)/error (in C++), as
if we had placed attribute((unavailable)) on it
- If we choose a deployment target prior to 10.2, foo() is
weak-imported (if it is a kind of entity that can be weak
imported), as if we had placed the weak_import attribute on it.
Naturally, there can be multiple availability attributes on a
declaration, for different platforms; only the current platform
matters when checking availability attributes.
The only platforms this attribute currently works for are "ios" and
"macosx", since we already have -mxxxx-version-min flags for them and we
have experience there with macro tricks translating down to the
deprecated/unavailable/weak_import attributes. The end goal is to open
this up to other platforms, and even extension to other "platforms"
that are really libraries (say, through a #pragma clang
define_system), but that hasn't yet been designed and we may want to
shake out more issues with this narrower problem first.
Addresses <rdar://problem/6690412>.
As a drive-by bug-fix, if an entity is both deprecated and
unavailable, we only emit the "unavailable" diagnostic.
llvm-svn: 128127
TCE target has some too strict alignment rules (that the HW really does not require, but which caused problems elsewhere) for data types and an ABI change was decided.
llvm-svn: 125833
Bill Schmidt