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
Fix the width and align of bool type on Darwin to be 32bits
while keeping it 8 everywhere else.
Change the definition of va_list to default to SV4 ABI one
and let darwin subtarget override this.
Both changes submitted by Nathan Whitehorn and reviewed
by Rafael Espindola.
llvm-svn: 122956
a specific language. We are adding such language info. by
extensing Builtins.def and via a language flag added
to LIBBUILTIN/BUILTIN and check for that when deciding
a name is builtin or not. Implements //rdar://8689273.
llvm-svn: 120429
in asm's. PR 8501, 8602988.
I don't like including Type.h where it is; the idea was
to get references to X86_MMXTy out of the common code.
Maybe there's a better way?
llvm-svn: 117736
This adds an option to set the _MSC_VER macro without
recompiling. This is very useful when testing compatibility
with the Windows SDK and c++stdlib headers.
-fmsc-version=<version> (defaults to VS2003 (1300))
llvm-svn: 116999
whether to use objc_msgSend_fpret; the choice is target dependent, not Obj-C ABI
dependent.
- <rdar://problem/8139758> arm objc _objc_msgSend_fpret bug
llvm-svn: 108379
the x86-64 __va_list_tag with this attribute. The attribute causes the
affected type to behave like a fundamental type when considered by ADL.
(x86-64 is the only target we currently provide with a struct-based
__builtin_va_list)
Fixes PR6762.
llvm-svn: 104941
- This fixes the last known ABI issues with ARM/APCS.
- I've run the first 1k ABITests with '--no-unsigned --no-vector --no-complex'
on {armv6, armv7} x {-mno-thumb, -mthumb}, and the first 10k tests for armv7
-mthumb, for both function return types and single argument calls. These all
pass now (they failed horribly before without --no-bitfield).
llvm-svn: 102070
by setting the section of the generated global. This is an
optimization done by the code generator, and the code being
removed didn't handle the case when the string contained an
embedded nul (which the code generator does correctly
handle). This is rdar://7589850
llvm-svn: 95003
- Correctly is in quotes, because we are following what I interpreted as GCC's
intent (which diverges from practice, naturally).
- Also, fix the arch define for arm1136jf-s.
llvm-svn: 91855
- In particular, it can claim features for itself instead of always passing them on to LLVM.
- This allows using the target features as a generic mechanism for passing target specific options to the TargetInfo instance, which may need them for initializing preprocessor defines, etc.
llvm-svn: 91753
1) -fwritable-string does affect the non-utf16 version of cfstrings
just not the utf16 ones.
2) utf16 strings should always be marked constant, as the __TEXT segment
is readonly.
3) The name of the global doesn't matter, remove it from TargetInfo.
4) Trust the asmprinter to drop cstrings into the right section, like llvmgcc does now.
This fixes rdar://7115750
llvm-svn: 84077
- Change TargetData string to match llvm-gcc.
- Some -target-abi support for 'apcs-gnu', most importantly the alignment of double and long long changes.
llvm-svn: 81732
1. Passing something that isn't a string used to cause:
"argument to annotate attribute was not a string literal"
make it say "section attribute" instead.
2. Fix the location of the above message to point to the
bad argument instead of the section token.
3. Implement rdar://4341926, by diagnosing invalid section
specifiers in the frontend rather than letting them slip all
the way to the assembler (a QoI win).
An example of #3 is that we used to produce something like this:
/var/folders/n7/n7Yno9ihEm894640nJdSQU+++TI/-Tmp-//ccFPFGtT.s:2:Expected comma after segment-name
/var/folders/n7/n7Yno9ihEm894640nJdSQU+++TI/-Tmp-//ccFPFGtT.s:2:Rest of line ignored. 1st junk character valued 46 (.).
Daniel improved clang to use llvm_report_error, so now we got:
$ clang t.c -c
fatal error: error in backend: Global variable 'x' has an invalid section specifier 'sadf': mach-o section specifier
requires a segment and section separated by a comma.
with no loc info. Now we get:
$ clang t.c -fsyntax-only
t.c:4:30: error: argument to 'section' attribute is not valid for this target: mach-o section specifier requires a segment
and section separated by a comma
int x __attribute__((section("sadf")));
^
which is nice :)
llvm-svn: 78586
Note that I'm guessing that *BSD and Solaris do the same thing as Linux
here, but it's quite possible I'm wrong; if the following testcase
gives an error on x86-64 with gcc for any of those operating systems, please
tell me:
#include <stdint.h>
int64_t x; long x;
llvm-svn: 74583
This unifies all the targets supported by an OS into a template.
It also cleans up the differences between the darwin targets.
Also __LP64__ wasn't needed for *BSD, since x86-64 target defines it anyway.
llvm-svn: 74532
function attributes. There are predefined macros that are defined when stack
protectors are used: __SSP__=1 with -fstack-protector and __SSP_ALL__=2 with
-fstack-protector-all.
llvm-svn: 74405
- Apologies for the extremely gross code duplication, I want to get
this working and then decide how to get this information out of the
back end.
- This replaces -m[no-]sse4[12] by -m[no-]sse4, it appears gcc
doesn't distinguish them?
- -msse, etc. now properly disable/enable related features.
- Don't always define __SSE3__...
- The main missing functionality bit here is that we don't initialize
the features based on the CPU for all -march options.
llvm-svn: 71117
- Default to yonah on Darwin (to get SSE3).
- Default to Pentium4 (32-bit) and x86-64 (64-bit) on
non-Darwin. Welcome to the 21st century.
llvm-svn: 71069
- This is a WIP...
- This adds -march= handling to the driver, and fixes the defaulting
of -mcpu on Darwin (which was using the wrong test).
Instead of handling -m{sse, ...} in the driver, pass them to clang-cc as
-target-feature [+-]name
In clang-cc, communicate with the (clang) target to discover the legal
features of a target, and the features which are enabled based on
-mcpu. This is currently hardcoded just enough to not be a feature
regression, we need to get this information from the backend's
TableGen information somehow.
This is used to construct the full list of features which are being
used, which is in turn used to initialize the predefines.
llvm-svn: 71061
Let me know if I messed up for some target. Note that for Windows, we
should be able to support it (MSVC supports "__declspec(thread)"), but
I'm pretty sure LLVM doesn't know how to generate the correct code.
llvm-svn: 69552
functions in glibc header files that use FP Stack inline asm which the
backend can't deal with (PR879).
This "fixes" PR3970 for linux. Other affected systems should do similar
things. Maybe this should just go to the general i386/x86-64 sections?
llvm-svn: 69527
- Patch by Shantonu Sen (with a minor tweak to split out
getDarwin{OSX,IPhoneOS}Defines)!
- <rdar://problem/6776277> Need clang-cc/ccc-analyzer support for
-miphoneos-version-min
llvm-svn: 68815
and are even set in C mode. As such, move them to Targets.cpp.
__OBJC_GC__ is also darwin specific, but seems reasonable to always
define it when in objc-gc mode.
This fixes rdar://6761450
llvm-svn: 68494
- Notably, set section on cfstring literal string data (for now, this
is done everywhere because it matches what we were already doing
for the CFString data itself)
- <rdar://problem/6599098> [irgen] linker requires objc string data
to go into cstring
llvm-svn: 68160
and defining target-specific macros based on them (like __SSE3__ and
friends). After extensive discussion with Daniel, this work will need
driver support, which will translate things like -msse3 into a -mattr
feature. Until this work is done, the code in clang.cpp is disabled and
the X86TargetInfo ctor still defaults to SSE2. With these two things
changed, this code will work. PR3634
llvm-svn: 65966
a target.
Make Preprocessor.cpp define a new __INTPTR_TYPE__ macro based on this.
On linux/32, set intptr_t to int, instead of long. This fixes PR3563.
llvm-svn: 64495
specific targets default them to on. Default blocks to on on 10.6 and later.
Add a -fblocks option that allows the user to override the target's default.
Use -fblocks in the various testcases that use blocks.
llvm-svn: 60563
the types for size_t and ptrdiff_t more accurate. I think all of these
are correct, but please compare the defines for __PTRDIFF_TYPE__ and
__SIZE_TYPE__ to gcc to double-check; this particularly applies to
those on BSD variants, since I'm not sure what they do here; I assume
here that they're the same as on Linux.
Fixes wchar_t to be "int", not "unsigned int" (which I think is
correct on everything but Windows).
Fixes ptrdiff_t to be "int" rather than "short" on PIC16; "short" is an
somewhat strange choice because it normally gets promoted, and it's not
consistent with the choice for size_t.
llvm-svn: 58556
etc more generic. For some targets, long may not be equal to pointer size. For
example: PIC16 has int as i16, ptr as i16 but long as i32.
Also fixed a few build warnings in assert() functions in CFRefCount.cpp,
CGDecl.cpp, SemaDeclCXX.cpp and ParseDeclCXX.cpp.
llvm-svn: 58501
target indep code.
Note that this changes functionality on PIC16: it defines __INT_MAX__
correctly for it, and it changes sizeof(long) to 16-bits (to match
the size of pointer).
llvm-svn: 57132
If you're on some other platform, the correct definition for this macro
would be appreciated; to find the correct definition, just run the
following command:
echo | gcc -dM -E - | grep USER_LABEL_PREFIX
llvm-svn: 55869
- Used to autoselect runtime when neither -fnext-runtime nor
-fgnu-runtime is specified.
- Default impl is false, all darwin targets set it to true.
llvm-svn: 55231
difference from generic x86 is the defines. The rest is non-trivial to
implement.
I'm not planning on adding any more targets myself; if there are any
targets anyone is currently using that are missing, feel free to add
them, or ask me to add them.
This concludes the work I'm planning for the TargetInfo
implementations at the moment; all the other issues with TargetInfo require
some API changes, and I haven't really thought it through. Some of the
remaining issues: allowing targets to define size_t and wchar_t properly,
adding some sort of __builtin_type_info intrinsic so we can finish clang's
limits.h and float.h and get rid of a massive number of macro
definitions, allowing target-specific command-line options, allowing
target-specific defaults for certain command-line options like
-fms-extensions, exposing vector alignment outside of the description
string, exposing endianness outside of the description string, allowing
targets to expose special bit-field layout requirements, exposing some
sort of custom hook for call generation in CodeGen, and adding CPU
selection to control defines like __SSE__.
llvm-svn: 55098
This approach allows adding OS-specific targets/defines/etc. without
completely breaking unknown subtargets. No new subtargets yet, although
I plan to add x86-Linux soon. Others can add targets that they use as
needed; adding a new subtarget takes very little code.
Also does some fixups for description strings; a lot of them were
unspecified. I think all the ones I added are correct, but
they're unverified; corrections are welcome.
llvm-svn: 55091
cleaned it up a bit, including fixing the definition of va_list; this
shouldn't break anything, but anyone using Sparc should watch for
regressions.
llvm-svn: 55041
visible effects, but this will significantly reduce the amount of
boilerplate code necessary to add subtargets.
If this looks okay, I'll do the rest of the processors (PPC, Sparc, ARM)
soon.
llvm-svn: 55036
- Kill unnecessary #includes in .cpp files. This is an automatic
sweep so some things removed are actually used, but happen to be
included by a previous header. I tried to get rid of the obvious
examples and this was the easiest way to trim the #includes in one
fell swoop.
- We now return to regularly scheduled development.
llvm-svn: 54632
hardcoded data layout in getTargetDescription. Hopefully fixes a test
failure.
Of course, this should be fixed properly, but that's a bigger fix.
llvm-svn: 51948
lib dir and move all the libraries into it. This follows the main
llvm tree, and allows the libraries to be built in parallel. The
top level now enforces that all the libs are built before Driver,
but we don't care what order the libs are built in. This speeds
up parallel builds, particularly incremental ones.
llvm-svn: 48402
Chandler Carruth