da6822f584
Add codegen for __builtin_abort. Convert to FileCheck.
...
llvm-svn: 83427
2009-10-06 22:58:45 +00:00
78a2e6356b
remove this test, grepping for define isn't really testing anything.
...
llvm-svn: 83325
2009-10-05 21:16:22 +00:00
80401b96c2
FileCheckize test case.
...
llvm-svn: 83244
2009-10-02 10:32:51 +00:00
8a744ad8a6
Don't update the struct alignment when adding fields to a packed struct. Fixes PR5118.
...
llvm-svn: 83235
2009-10-02 04:52:12 +00:00
39e3eb12ae
When building constant structs, check if the resulting LLVM struct will be bigger than the record layout size and use a packed struct if that's the case. Fixes PR5108.
...
llvm-svn: 83230
2009-10-02 02:15:20 +00:00
3dc959f88b
Tweak CHECK lines to eliminate a failure on i686-apple-darwin10
...
llvm-svn: 83173
2009-09-30 21:39:51 +00:00
24440102aa
Convert from nonportable grep to filecheck, patch by John Thompson
...
llvm-svn: 83158
2009-09-30 19:55:07 +00:00
e33eed5c1e
Set alignment on static function level decls and VLAs. Fixes PR5060.
...
llvm-svn: 82868
2009-09-26 18:16:06 +00:00
b3b1e53d33
Darwin/x86-32: Enumerated types and block pointer types in structures were not
...
handled correctly.
- <rdar://problem/7247671> Function arguments incorrect when function returns a
struct on i386 w/ llvm-g++ and clang
llvm-svn: 82681
2009-09-24 05:12:36 +00:00
1da76c4e58
Add ARM register names and aliases.
...
- Patch by Shantonu Sen!
<rdar://problem/6922650> clang doesn't know about ARM registers for inline asm clobber lists
llvm-svn: 82132
2009-09-17 07:03:19 +00:00
09d3362bf5
ARM/[A]APCS: Ignore empty records passed as arguments.
...
llvm-svn: 81798
2009-09-14 21:54:03 +00:00
ff0553ec61
Fix subtle bug in generating LLVM function declarations for builtin functions.
...
The decl wasn't being passed down, which meant that function attributes were not
being set correctly. This is particularly important for ARM, since it wants to
override the calling convention. Instead we would emit the builtin with the
wrong calling convention, and instcombine would come along and merrily shred all
the calls to it. :)
llvm-svn: 81756
2009-09-14 04:33:21 +00:00
d53bac7fa4
ARM/APCS: Don't treat structs w/ floating point types as "integer like".
...
llvm-svn: 81748
2009-09-14 02:20:34 +00:00
1ce7251a0a
Some minor clang/ARM/AAPCS tweaks.
...
llvm-svn: 81737
2009-09-14 00:56:55 +00:00
b4091a9c6a
Add TargetInfo::getABI(), and base ARM APCS vs AAPCS choice on that.
...
llvm-svn: 81735
2009-09-14 00:35:03 +00:00
ccbabc9645
Fix another byref bug. This should hopefully get QuickLookPlugins building successfully.
...
llvm-svn: 81681
2009-09-13 17:55:13 +00:00
626f1d8c3a
ARM/APCS: Only "integer like" aggregates should be returned in r0 (following
...
gcc's interpretation of APCS' somewhat loose specification).
llvm-svn: 81671
2009-09-13 08:03:58 +00:00
299d76e901
Rework the way we determine whether an externally visible symbol is
...
generated for an inline function definition, taking into account C99
and GNU inline/extern inline semantics. This solution is simpler,
cleaner, and fixes PR4536.
llvm-svn: 81670
2009-09-13 07:46:26 +00:00
0482cfd790
Don't use the PredefinedExpr string as the global variable name, these don't
...
make very nice symbols, just use the function name.
llvm-svn: 81653
2009-09-12 23:06:21 +00:00
f8e94f2008
Add support for __block variables with alignment greater than __alignof(void *).
...
llvm-svn: 81602
2009-09-12 02:44:18 +00:00
020daa9476
Stub out room for ARM APCS ABI implementation (and AAPCS_VFP, although you can't
...
hit this via command line options yet).
llvm-svn: 81595
2009-09-12 01:00:39 +00:00
0533ffa2a6
Update this test to expect the "inbounds" keyword, which LLVM's constant
...
folder is now automatically adding.
llvm-svn: 81491
2009-09-11 00:27:06 +00:00
10f2c10b83
Make the forwarding member of block byref structs be a pointer to the block byref struct itself.
...
llvm-svn: 81423
2009-09-10 01:32:12 +00:00
11289f4280
Remove tabs, and whitespace cleanups.
...
llvm-svn: 81346
2009-09-09 15:08:12 +00:00
6426345b8f
convert this to filecheck, hopefully it will fix PR4888. If nothing
...
else it will make tests run faster and make 4888 easier to diagnose.
llvm-svn: 81238
2009-09-08 18:43:45 +00:00
2fb0824197
Vastly improve PredefinedExpr output, both in Sema and CodeGen. Patch by Sam Weinig!
...
llvm-svn: 81237
2009-09-08 18:24:21 +00:00
0e91275ab4
If the alignment of the chosen field in a union is greater than the alignment of the union, we need to use a packed LLVM struct. Fixes <rdar://problem/7184250>.
...
llvm-svn: 80964
2009-09-03 22:56:02 +00:00
09a3774cd3
Packed unions should be packed. Fixes an assert Daniel reported.
...
llvm-svn: 80808
2009-09-02 17:51:33 +00:00
1c277d0fe8
PR4836, part 2: CodeGen for __builtin_isnan.
...
llvm-svn: 80655
2009-09-01 04:19:44 +00:00
7ef882dca0
convert to filecheck and force a triple so that this passes on big
...
endian hosts. Fixes PR4792
llvm-svn: 80301
2009-08-27 23:22:19 +00:00
95f896380f
Make address-space qualification work correctly for compound literals.
...
Issue reported on cfe-dev.
Also fixed the code to use isConstant to determine whether to generate a
constant global, to be consistent with CodeGenModule. This probably
needs to be refactored to deal with C++, though.
llvm-svn: 80131
2009-08-26 20:01:39 +00:00
a98e0e73de
Update clang for raw_fd_ostream no longer requiring F_Force.
...
llvm-svn: 79991
2009-08-25 15:36:09 +00:00
87db734400
Fix a few tests to be -Asserts agnostic.
...
- Ugh.
llvm-svn: 79860
2009-08-23 19:28:59 +00:00
2a4061929f
Fix some made up triples.
...
llvm-svn: 79316
2009-08-18 05:30:27 +00:00
c0f6d44ef6
Make test a bit more precise.
...
llvm-svn: 79073
2009-08-15 03:35:55 +00:00
04fddf0d1f
Fix for PR4721: adjust CodeGen and ASTContext so that we have a
...
consistent model for handling size expressions for VLAs.
The model is essentially as follows: VLA types own their associated
expression. In some cases, we need to create multiple VLA types to
represent a given VLA (for canonical types, or qualifiers on array types,
or type merging). If we need to create multiple types based off of
the same VLA declaration, we use the new refcounting functionality so they can
all own the expression. The VLASizeMap in CodeGenFunction then uses the size
expression to identify the group of VLA types based off of the same original
declaration.
I'm not particularly attached to the VLA types owning the expression,
but we're stuck with at least until someone comes up with a way
to walk the VLA expressions for a declaration.
I did the parallel fix in ASTContext for DependentSizedArrayType, but I
haven't really looked closely at it, so there might still be issues
there.
I'll clean up the code duplication in ASTContext in a followup commit.
llvm-svn: 79071
2009-08-15 02:50:32 +00:00
e5515289fa
Update test
...
llvm-svn: 78877
2009-08-13 01:27:45 +00:00
d362857f4c
Update test case.
...
llvm-svn: 78699
2009-08-11 18:37:40 +00:00
2375952a47
these tests include section specifiers that aren't valid on the
...
mac, use an explicit triple.
llvm-svn: 78583
2009-08-10 18:56:44 +00:00
e139f868b7
Add tests.
...
llvm-svn: 78491
2009-08-08 19:43:14 +00:00
f49573d1ee
weak globals that are const should get weak_odr linkage.
...
add a fixme about C++ const.
llvm-svn: 78159
2009-08-05 05:20:29 +00:00
c0693bc2ea
rdar://7119244 - globals with an explicit section specified don't get
...
common linkage.
llvm-svn: 78158
2009-08-05 04:56:58 +00:00
5638895b2a
Revert the fix for PR3800, it broke things.
...
llvm-svn: 78084
2009-08-04 18:18:36 +00:00
e9863976b3
Fix test case for Darwin10 (which sets ssp), and move to CodeGen/
...
llvm-svn: 78049
2009-08-04 04:38:31 +00:00
4d93a4f9d4
Make sure UTF-16 strings end with a 16-bit null (as opposed to 8-bit).
...
llvm-svn: 78001
2009-08-03 21:47:08 +00:00
7df966e040
Don't evaluate inout constraints twice. Fixes PR3800.
...
llvm-svn: 77854
2009-08-02 03:40:19 +00:00
62561fe9b0
Fix a minor issue with unions in the new struct building code.
...
llvm-svn: 77829
2009-08-01 23:11:24 +00:00
fe6bbd1145
New test case for rev. 77694.
...
llvm-svn: 77698
2009-07-31 18:24:12 +00:00
72fb384a65
Enable the new struct type builder now that the constant struct builder works. (The old code will still be there until we know that everything works well.
...
llvm-svn: 77190
2009-07-27 15:31:55 +00:00
b7130ed888
Enable the new constant struct builder by default. The old code is still in place but will be removed shortly. The new struct builder works on big endian systems.
...
llvm-svn: 77185
2009-07-27 05:54:15 +00:00
8c5d7996e8
Add noreturn as a type attribute, handle printing for them and handle
...
calls to noreturn function pointers when CFG building.
llvm-svn: 77089
2009-07-25 21:26:53 +00:00
6d8adef9a0
Remove empty tests, left over from an svn revert.
...
llvm-svn: 77068
2009-07-25 12:41:25 +00:00
97ea867690
MultiTestRunner: Validate '&&' at the end of RUN lines.
...
- This is just to normalize, these will go away soon hopefully.
Added all the missing '&&'s that have crept in. :)
llvm-svn: 77062
2009-07-25 11:27:37 +00:00
02dee0a46a
Semantic checking for main().
...
Fix some invalid main() methods in the test suite that were nicely
exposed by the new checks.
llvm-svn: 77047
2009-07-25 04:36:53 +00:00
8e79f99b16
Implement new warning for functions declared 'noreturn' when they fall off the end.
...
llvm-svn: 76932
2009-07-24 02:49:01 +00:00
91ade14197
Output UTF-16 string literals independent of host byte order.
...
- Steve, can you take a look at this? It seems like this code should live
elsewhere, and there is a FIXME about having Sema validates the UTF-8 to
UTF-16 conversion.
llvm-svn: 76915
2009-07-23 23:41:22 +00:00
516e5e7835
Correct a thinko in bitfield layout code. Fixes PR4611.
...
llvm-svn: 76898
2009-07-23 21:16:33 +00:00
d78fc89fcb
We don't need to keep track of the packed alignment, just whether the struct is packed or not. Fixes PR4610.
...
llvm-svn: 76884
2009-07-23 17:24:40 +00:00
718a89a501
Use arrays as union padding. Also, since the resulting struct will always contain a single element and either a single i8 element or an array of i8s, there's no reason to use a packed struct.
...
llvm-svn: 76854
2009-07-23 04:50:01 +00:00
5ac69057c0
Make vectorized floating-point comparisons work without crashing.
...
llvm-svn: 76726
2009-07-22 06:07:16 +00:00
acedf7768f
Preserve address space information through member accesses, e.g.,
...
__attribute__((address_space(1))) struct {int arr[ 3 ]; } *p1;
... = p1->arr[2]; // load from address space 1
llvm-svn: 76717
2009-07-22 03:08:17 +00:00
5c1acd1f33
Revert this, was a bug in my new warning code, not the test case.
...
llvm-svn: 76690
2009-07-21 23:50:15 +00:00
5e7869f63e
Prep for new warning.
...
llvm-svn: 76638
2009-07-21 20:52:43 +00:00
c3ab4c6c98
Detect when the current generation point is unreachable after emitting
...
expressions.
- This generally catches the important case of noreturn functions.
- With the last two changes, we are down to 152 unreachable blocks emitted on
403.gcc, vs the 1805 we started with.
llvm-svn: 76364
2009-07-19 08:23:12 +00:00
b6adc43f6e
Avoid generation of dead code in a few more situations.
...
- Emit variable declarations as "simple", we want to avoid forcing the creation
of a dummy basic block, but still need to make the variable available for
later use.
- With that, we can now skip IRgen for other unreachable statements (which
don't define a label).
- Anders, I added two fixmes on calls to EmitVLASize, can you check them?
llvm-svn: 76361
2009-07-19 06:58:07 +00:00
3afa3e1d91
codegen string literals using private linkage now like llvm-gcc, eliminating
...
some target hooks.
llvm-svn: 75895
2009-07-16 05:03:48 +00:00
2a7deb64c0
reimplement vector comparisons as [fi]cmp+sext instead of using v[if]cmp.
...
Also, enable them in sema so that they are tested, and now that the x86 backend
has stablized.
llvm-svn: 74983
2009-07-08 01:08:03 +00:00
877a48e67a
make these tests pass with the stack canary stuff even on targets where they default to on.
...
llvm-svn: 74412
2009-06-28 19:49:49 +00:00
d63bbadbef
Add stack protector support to clang. This generates the 'ssp' and 'sspreq'
...
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
2009-06-28 07:36:13 +00:00
0d425350b2
Revrt PR4228 fix for now.
...
llvm-svn: 74304
2009-06-26 18:32:22 +00:00
4c8da96ea9
fix PR4423.
...
llvm-svn: 73938
2009-06-23 01:38:41 +00:00
c619d6fab5
Add a target triple to a couple of tests which depend on it. Reported
...
by Mark Cianciosa on cfe-dev.
llvm-svn: 73672
2009-06-18 02:04:19 +00:00
9fc4341eec
It is possible that main input file does not have any symbol with debug info. To handle this edge case, always create main compile unit first.
...
This fixes PR 4228.
llvm-svn: 73520
2009-06-16 18:02:02 +00:00
943658f00b
testcase for r72869, it turns out that it happens in Ruby.
...
llvm-svn: 73514
2009-06-16 17:39:02 +00:00
6282b3cff8
PR4390: Make sure to handle anonymous unions correctly while building
...
static intializers for structs.
llvm-svn: 73349
2009-06-14 21:41:37 +00:00
3192cc8526
Fix the calling convention for structs/unions containing SSE vectors on
...
x86-32. This is slightly messy, but I think it's consistent with gcc.
llvm-svn: 73306
2009-06-13 21:37:10 +00:00
4ca97c3b9e
Fix PR4372, another case where non-prototyped functions can prevent
...
always_inline from working.
llvm-svn: 73273
2009-06-13 00:26:38 +00:00
fd262b4a62
Fix test, which could miss failures, and also avoid leaving temporary .i file in
...
source directory.
llvm-svn: 73094
2009-06-08 22:44:26 +00:00
eed62b7c4b
Add stack alignment to x86_64 target data.
...
- <rdar://problem/6948443> WARNING: Linking two modules of different data
layouts!
llvm-svn: 73093
2009-06-08 22:39:13 +00:00
e9ff191459
Remove a few more vector builtins.
...
llvm-svn: 73022
2009-06-07 09:32:56 +00:00
5a996fc0fc
Now that LLVM CodeGen can handle the generic variations a bit better,
...
get rid of a few more clang vector builtins.
llvm-svn: 73015
2009-06-07 07:12:56 +00:00
07bbeca406
PR4339: make sure to properly extend/trunc the index of a vector element
...
insert/extract; the relevant instructions are defined to take only an
i32.
llvm-svn: 73005
2009-06-06 19:09:26 +00:00
5f75ff84b7
Test changes to account for removed builtins.
...
llvm-svn: 73004
2009-06-06 18:15:42 +00:00
feeeda70e2
weak_import should not make definitions have weak linkage.
...
- <rdar://problem/6948703> clang treats weak_import like weak
llvm-svn: 72967
2009-06-05 22:58:34 +00:00
4be99ff767
ABI handling: Fix nasty thinko where IRgen could generate an out-of-bounds read
...
when generating a coercion for ABI handling purposes.
- This may only manifest itself when building at -O0, but the practical effect
is that other arguments may get clobbered.
- <rdar://problem/6930451> [irgen] ABI coercion clobbers other arguments
llvm-svn: 72932
2009-06-05 07:58:54 +00:00
dd27484888
PR4316: Fix IRGen for cast-to-union extension.
...
llvm-svn: 72803
2009-06-03 20:45:06 +00:00
877b30f83c
Driver: Support -mllvm; this just forwards options to clang-cc.
...
llvm-svn: 72780
2009-06-03 16:16:27 +00:00
4b665ebb01
Add clang-cc support for -disable-llvm-optzns.
...
- Avoids running any LLVM optimizations, even at -O2, etc., while still keeping
any language changes these optimizations imply.
llvm-svn: 72742
2009-06-02 22:07:45 +00:00
cb9d07caeb
Add support for __builtin_unwind_init.
...
Also, committing an #if 0'ed __builtin_setjmp and __builtin_longjmp
implementation I've had sitting in my tree for a while. I haven't
enabled it because the LLVM backend support isn't complete yet.
llvm-svn: 72727
2009-06-02 09:37:50 +00:00
e32c02114f
PR4289: Make sure "&func" has the right LLVM type when "func" is a
...
K&R-style definition.
llvm-svn: 72690
2009-06-01 10:04:20 +00:00
202a68a917
Simplify run line.
...
llvm-svn: 72687
2009-06-01 08:55:08 +00:00
2346cd2a10
Improve __builtin_nanf support; we now can deal with them as constants.
...
llvm-svn: 72607
2009-05-30 03:56:50 +00:00
9444638e4e
Re-add a slightly more general version of the check from r72578; it is
...
actually necessary in some obscure cases.
llvm-svn: 72585
2009-05-29 19:23:46 +00:00
de83126b80
We don't want to validate bad code,
...
llvm-svn: 72574
2009-05-29 16:24:13 +00:00
f9f89a3fee
Note another case that doesn't work yet.
...
llvm-svn: 72573
2009-05-29 16:12:36 +00:00
df0fe27b66
Fixup the rest of the trivial cases of the codegen of volatile. If
...
any body can spot codegen bugs with volatile, or knows of any in the
bug database, let me know.
llvm-svn: 72572
2009-05-29 15:46:01 +00:00
2a69547f38
PR4281: Fix bogus CodeGen assertion. The issue is that
...
getUnqualifiedType() doesn't strip off all qualifiers for non-canonical
types.
llvm-svn: 72552
2009-05-28 23:04:00 +00:00
16f422ec86
Fix a use-of-undefined, IRgen expects the RValue for "void" to be a scalar RValue.
...
llvm-svn: 72504
2009-05-27 23:45:33 +00:00
de79ea2f51
Update comment.
...
llvm-svn: 72444
2009-05-26 23:35:31 +00:00
6adbd62790
And an additional testcase that also works.
...
llvm-svn: 72441
2009-05-26 23:10:55 +00:00
ec3cbfe8c6
Fixup codegen for volatile structs in the trivial cases (a a=a and a=a=a).
...
llvm-svn: 72439
2009-05-26 22:03:21 +00:00
1518b64ddc
When trying to pass an argument on the stack, assume LLVM will do the right
...
thing for non-aggregate types.
- Otherwise we unnecessarily pin values to the stack and currently end up
triggering a backend bug in one case.
- This loose cooperation with LLVM to implement the ABI is pretty ugly.
- <rdar://problem/6918722> [irgen] clang miscompile of many pointer varargs on
x86-64
llvm-svn: 72419
2009-05-26 16:37:37 +00:00
895771aa4b
Handle the edge case of a weak function with incomplete type correctly.
...
Found by code inspection; I haven't seen this in real-world code.
llvm-svn: 72408
2009-05-26 01:22:57 +00:00
c63428b192
Fixup codegen for __block int i; i += rhs();. Should also slightly
...
improve codegen in some cases.
llvm-svn: 72273
2009-05-22 19:07:20 +00:00
499f3f9c5d
x86_64 ABI: Account for sret parameters consuming an integer register.
...
- PR4242.
llvm-svn: 72268
2009-05-22 17:33:44 +00:00
5b34933b90
Set correct calling convention even if there is a bitcast in the way.
...
This attempts to fix PR4239.
llvm-svn: 72251
2009-05-22 07:25:06 +00:00
aed08f9929
Fixup blocks codegen for { __block i; i = rhs(); }, we want the rhs
...
evaluated first. This can also improve codegen just a bit as we might
have another register to play with for the evaluation of the rhs.
llvm-svn: 72226
2009-05-21 21:05:15 +00:00
018958d94d
Update test
...
llvm-svn: 72110
2009-05-19 16:09:59 +00:00
45966b4671
Remove the -arch option from clang-cc: for all practical purposes, it's
...
redundant with -triple.
llvm-svn: 72108
2009-05-19 11:12:40 +00:00
8dd2947696
Remove an unused builtin.
...
llvm-svn: 72033
2009-05-18 19:25:54 +00:00
2081200b8c
Add 'cmp' SSE builtins and get rid of a bunch of other builtins.
...
llvm-svn: 72032
2009-05-18 19:16:46 +00:00
2daacd1f08
Avoid generating temp in source directory
...
llvm-svn: 71776
2009-05-14 17:00:11 +00:00
ffdb8439d7
ABI handling: Fix invalid assertion, it is possible for a valid
...
coercion to be specified which truncates padding bits. It would be
nice to still have the assert, but we don't have any API call for the
unpadding size of a type yet.
llvm-svn: 71695
2009-05-13 18:54:26 +00:00
afde259240
implement __sync_synchronize and __sync_lock_release,
...
rdar://6880573
llvm-svn: 71637
2009-05-13 04:46:13 +00:00
e541ea3a61
implement l-value codegen of comma expr
...
llvm-svn: 71595
2009-05-12 21:28:12 +00:00
bbfd054746
Darwin x86-32 ABI: Now that structure passing is farther along, we
...
don't need special treatment for unions.
llvm-svn: 71559
2009-05-12 17:00:20 +00:00
203e2e8dd8
x86-64 ABI: clang incorrectly passes union { long double, float } in
...
register.
- Merge algorithm was returning MEMORY as it should.
llvm-svn: 71556
2009-05-12 15:22:40 +00:00
097353cbb5
Darwin x86-32: Multi-dimensional arrays were not handled correctly,
...
spotted by Eli!
llvm-svn: 71490
2009-05-11 23:01:34 +00:00
e92449e4cc
Make cleanup-stack.c test stronger, and work in release-asserts mode.
...
llvm-svn: 71484
2009-05-11 22:30:29 +00:00
2ce6b3f91c
Darwin x86_32: Treat records with unnamed bit-fields as "empty".
...
llvm-svn: 71461
2009-05-11 18:58:49 +00:00
da38fda583
force a target triple so that the right greppable output happens.
...
llvm-svn: 71361
2009-05-09 17:36:58 +00:00
b997f3bcc3
x86_64 ABI: Ignore padding bit-fields during classification.
...
- {return-types,single-args}-{32,64} pass the first 1k ABI tests with
bit-fields enabled.
llvm-svn: 71272
2009-05-08 22:26:44 +00:00
4752783057
Darwin x86_32: When coercing a "single element" structure, make sure
...
to use a wide enough type. This might be wider than the "single
element"'s type in the presence of padding bit-fields.
- Darwin x86_32 now passes the first 1k ABI tests with bit-field
generation enabled.
llvm-svn: 71270
2009-05-08 21:30:11 +00:00
fdda3501a0
Darwin x86_32: Ignore padding bit-fields when looking for "single
...
element" structures.
llvm-svn: 71266
2009-05-08 21:04:47 +00:00
4861346c44
Darwin x86_32: Improve bit-field handling for returning records.
...
- This turns out to be a no-op now that most of the handling for
everything else is in place.
llvm-svn: 71261
2009-05-08 20:55:49 +00:00
85f4028f2e
Darwin x86_32: Ignore arrays of empty structures inside records.
...
- This eliminates 5/1000 failures on return-types-32, on the current
ABITest config.
llvm-svn: 71250
2009-05-08 20:21:04 +00:00
5b9241b2a6
Fix the atomics sema code to convert operands to the argument types
...
of the underlying _N builtin, not the the type of the pointee of the
actual type. This ensures that atomics involving pointers end up
using the correct integer type when they are resolved, avoiding
aborts in codegen.
llvm-svn: 71218
2009-05-08 15:36:58 +00:00
dc04654697
reimplement __sync_* builtins to be variadic and to follow the same
...
semantic rules that gcc and icc use. This implements the variadic
and concrete versions as builtins and has sema do the
disambiguation. There are probably a bunch of details to finish up
but this seems like a large monotonic step forward :)
llvm-svn: 71212
2009-05-08 06:58:22 +00:00
7ce29a18e1
Fix crash with constant initialization of bit-fields in unions.
...
llvm-svn: 71194
2009-05-07 23:42:42 +00:00
4dbaaa6f43
Improve handling of (X86) target features.
...
- 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
2009-05-06 03:16:41 +00:00
36797ab251
When defining a function whose type has no prototype, make an effort
...
to go back and clean up existing uses of the bitcasted function. This
is not just an optimization: it is required for correctness to get
always inline functions to work, see testcases in function-attributes.c.
llvm-svn: 70971
2009-05-05 06:16:31 +00:00
2ad7e17096
PR4143: don't crash generating debug info for incomplete enum types.
...
llvm-svn: 70825
2009-05-04 04:39:55 +00:00
5b73b5e197
PR4134: Implement __builtin_extract_return_addr.
...
llvm-svn: 70794
2009-05-03 19:23:23 +00:00
c0042d804c
PR4133: fix always_inline implementation to be consistent with gcc.
...
llvm-svn: 70786
2009-05-03 18:13:43 +00:00
61af27860d
look at the right operand when increasing the size of an asm output,
...
this fixes http://llvm.org/bugs/show_bug.cgi?id=3373#c20
llvm-svn: 70685
2009-05-03 09:05:53 +00:00
cc1cde9c57
allow references to the larger value in a tied constraint
...
from the asm string, but reject references to the smaller one.
llvm-svn: 70679
2009-05-03 08:32:32 +00:00
59c3a9cd54
add support for tying asm operands where the result is smaller than
...
the input. This is part of PR3373.
llvm-svn: 70677
2009-05-03 08:21:20 +00:00
c7121faa71
Remove typo
...
llvm-svn: 70676
2009-05-03 08:00:14 +00:00
10f221f321
implement support for asm outputs targetting non-simple lvalue destinations
...
like bitfields. incidentally llvm-gcc crashes on this sort of thing also. :)
llvm-svn: 70675
2009-05-03 07:53:25 +00:00
b65933eaa2
handle codegen of asms where a small input is tied to a large output.
...
llvm-svn: 70672
2009-05-03 07:27:51 +00:00
8d9c509975
Implement bit-field promotion rules for C99. Fixes PR3500.
...
llvm-svn: 70571
2009-05-01 20:41:21 +00:00
72a57b89aa
Add testcase that illustrates the problem from r69699 regarding tentative definitions of statics
...
llvm-svn: 70543
2009-05-01 15:45:53 +00:00
d898026e8a
Don't assert when we think we need copy/dispose, but don't need them.
...
Radar 6838889
llvm-svn: 70525
2009-05-01 01:31:57 +00:00
b4e463287e
Don't use indirect memory destinations for inline asm. Fixes 6841383.
...
llvm-svn: 70523
2009-05-01 00:16:04 +00:00
902fddd1cf
Fix for PR4108: be a bit looser with the casts that we accept in
...
constant initializers.
llvm-svn: 70483
2009-04-30 07:03:22 +00:00
7d4f5c47ce
only support int128_t on 64-bit and larger targets. 32-bit targets don't
...
have support for __divti3 and friends.
llvm-svn: 70480
2009-04-30 06:18:40 +00:00
f122cef4df
initial support for __[u]int128_t, which should be basically
...
compatible with VC++ and GCC. The codegen/mangling angle hasn't
been fully ironed out yet. Note that we accept int128_t even in
32-bit mode, unlike gcc.
llvm-svn: 70464
2009-04-30 02:43:43 +00:00
76fe50c654
Improve compatibility with GCC regarding inline semantics in GNU89
...
mode and in the presence of __gnu_inline__ attributes. This should fix
both PR3989 and PR4069.
As part of this, we now keep track of all of the attributes attached
to each declaration even after we've performed declaration
merging. This fixes PR3264.
llvm-svn: 70292
2009-04-28 06:37:30 +00:00
2b680b43e9
Simplify the scheme used for keywords, and change the classification
...
scheme to be more useful.
The new scheme introduces a set of categories that should be more
readable, and also reflects what we want to consider as an extension
more accurately. Specifically, it makes the "what is a keyword"
determination accurately reflect whether the keyword is a GNU or
Microsoft extension.
I also introduced separate flags for keyword aliases; this is useful
because the classification of the aliases is mostly unrelated to the
classification of the original keyword.
This patch treats anything that's in the implementation
namespace (prefixed with "__", or "_X" where "X" is any upper-case
letter) as a keyword without marking it as an extension. This is
consistent with the standards in that an implementation is allowed to define
arbitrary extensions in the implementation namespace without violating
the standard. This gets rid of all the nasty "extension used" warnings
for stuff like __attribute__ in -pedantic mode. We still warn for
extensions outside of the the implementation namespace, like typeof.
If someone wants to implement -Wextensions or something like that, we
could add additional information to the keyword table.
This also removes processing for the unused "Boolean" language option;
such an extension isn't supported on any other C implementation, so I
don't see any point to adding it.
The changes to test/CodeGen/inline.c are required because previously, we
weren't actually disabling the "inline" keyword in -std=c89 mode.
I'll remove Boolean and NoExtensions from LangOptions in a follow-up
commit.
llvm-svn: 70281
2009-04-28 03:13:54 +00:00
01910a50c4
x86-32 ABI: Fix crash on return of structure with flexible array
...
member.
Also, spell bitfield more consistently as bit-field.
llvm-svn: 70220
2009-04-27 18:31:32 +00:00
Mike Stump