209 Commits

Author SHA1 Message Date
John McCall
a729c62b81 Whether an argument is required (in contrast with being an
optional argument passed through the variadic ellipsis)
potentially affects how we need to lower it.  Propagate
this information down to the various getFunctionInfo(...)
overloads on CodeGenTypes.  Furthermore, rename those
overloads to clarify their distinct purposes, and make
sure we're calling the right one in the right place.
This has a nice side-effect of making it easier to construct
a function type, since the 'variadic' bit is no longer
separable.

This shouldn't really change anything for our existing
platforms, with one minor exception --- we should now call
variadic ObjC methods with the ... in the "right place"
(see the test case), which I guess matters for anyone
running GNUStep on MIPS.  Mostly it's just a substantial
clean-up.

llvm-svn: 150788
2012-02-17 03:33:10 +00:00
Akira Hatanaka
4984f5dbbe Class objects passed by value follow the same rules as structure objects.
Double fields of by-value class objects should be passed in floating point
registers.

llvm-svn: 150200
2012-02-09 19:54:16 +00:00
Akira Hatanaka
b6f7443c3c Fix bugs in function MipsABIInfo::returnAggregateInRegs. Functions returning
class objects follow the same rules as those returning struct objects.

llvm-svn: 150196
2012-02-09 18:49:26 +00:00
Akira Hatanaka
c07c4655f1 Do not return records with non trivial destructors or copy constructors in
registers.

llvm-svn: 150035
2012-02-08 01:31:22 +00:00
Chris Lattner
ece0409a1a simplify a bunch of code to use the well-known LLVM IR types computed by CodeGenModule.
llvm-svn: 149943
2012-02-07 00:39:47 +00:00
Eli Friedman
a98d1f8ca6 Add some ABI tweaks for i386-pc-win32 triple so that we return structs in an MSVC-compatible way. Patch by Joe Groff.
llvm-svn: 148992
2012-01-25 22:46:34 +00:00
Akira Hatanaka
377152870b Make sure the integer type used to align the vaarg address is the same as the
type of pointers. 

llvm-svn: 148753
2012-01-23 23:59:52 +00:00
Akira Hatanaka
60f5fe6857 Ignore return type if its size is zero.
llvm-svn: 148744
2012-01-23 23:18:57 +00:00
David Blaikie
e4d798f078 More dead code removal (using -Wunreachable-code)
llvm-svn: 148577
2012-01-20 21:50:17 +00:00
Akira Hatanaka
32604a9708 Take into account the pointer to an aggregate that is passed as a hidden
argument when Offset is initialized.

llvm-svn: 147986
2012-01-12 01:10:09 +00:00
Akira Hatanaka
02e13e5005 Call CodeGenTypes::ConvertType to get LLVM::Type*.
Remove function MipsABIInfo::GetFloatingPointTy.

llvm-svn: 147985
2012-01-12 00:52:17 +00:00
Akira Hatanaka
ee65b744e8 Remove switch/case statements and call GetFloatingPointTy() instead to get the
floating point type.

llvm-svn: 147894
2012-01-10 23:30:49 +00:00
Akira Hatanaka
79f0461c38 Flatten float complex arguments. N32/64 requires float complex arguments be
passed in floating point registers.

llvm-svn: 147892
2012-01-10 23:12:19 +00:00
Akira Hatanaka
21ee88cc1c Remove space.
llvm-svn: 147889
2012-01-10 22:44:52 +00:00
Chandler Carruth
c89aa9d964 Add support for the androideabi environment to our triple support, and
for the arm-linux-androideabi triple in particular.

Also use this to do a better job of selecting soft FP settings.

Patch by Evgeniy Stepanov.

llvm-svn: 147872
2012-01-10 19:47:42 +00:00
Joerg Sonnenberger
11aaa7102a Contrary to the other BSDs, NetBSD followed the ELF ABI for structure
passing on i386 and does not use registers.

llvm-svn: 147856
2012-01-10 16:53:17 +00:00
Akira Hatanaka
1632af603d Insert padding before unaligned long double arguments.
llvm-svn: 147791
2012-01-09 19:31:25 +00:00
Akira Hatanaka
f64e1adcbb Add field PaddingType to ABIArgInfo which specifies the type of padding that
is inserted before the real argument. Padding is needed to ensure the backend
reads from or writes to the correct argument slots when the original alignment
of a byval structure is unavailable due to flattening.

llvm-svn: 147699
2012-01-07 00:25:33 +00:00
Akira Hatanaka
f093f5be3e Have functions return structures smaller than 128-bit in registers if ABI
is either N32 or N64. 

llvm-svn: 147520
2012-01-04 03:34:42 +00:00
Tony Linthicum
76329bf83f Hexagon backend support
llvm-svn: 146413
2011-12-12 21:14:55 +00:00
Eli Friedman
b8e45b28da Minor comment update.
llvm-svn: 145905
2011-12-06 03:08:26 +00:00
Eli Friedman
bfd5addf4c When we're passing a vector with an illegal type through memory on x86-64, use byval so we're sure the backend does the right thing. Fixes va_arg with illegal vectors and an obscure ABI mismatch with __m64 vectors.
llvm-svn: 145652
2011-12-02 00:11:43 +00:00
Eli Friedman
f37bd2f2f1 Don't use a varargs convention for calls unprototyped functions where one of the arguments is an AVX vector.
llvm-svn: 145574
2011-12-01 04:53:19 +00:00
Tanya Lattner
71f1b2dcd4 Correct the code generation for function arguments of vec3 types on x86_64 when they are greater than 128 bits. This was incorrectly coercing things like long3 into a double2.
Add test case.

llvm-svn: 145312
2011-11-28 23:18:11 +00:00
Eli Friedman
f22fa9eaef Finish r144971, which was an incomplete commit.
llvm-svn: 144972
2011-11-18 04:01:36 +00:00
Eli Friedman
0b3f201b61 Fix the meaning of an "empty" record for the case of a zero-length array. Use isEmptyRecord for arguments on x86-32; there are structs of size 0 which don't count as empty.
llvm-svn: 144971
2011-11-18 03:47:20 +00:00
Eli Friedman
a1748564b4 Make va_arg on x86-64 compute alignment the same way as argument passing.
Fixes <rdar://problem/10463281>.

llvm-svn: 144966
2011-11-18 02:44:19 +00:00
Eli Friedman
1d7dd3b682 A bunch of fixes to argument passing and va_arg on Darwin x86-32 for structures containing an SSE vector.
llvm-svn: 144963
2011-11-18 02:12:09 +00:00
Eli Friedman
e5c85622c9 Don't try to expand struct arguments containing holes on x86-32. From gcc struct layout tests.
llvm-svn: 144961
2011-11-18 01:32:26 +00:00
Eli Friedman
ee94534662 Simplify code for returning a struct for Darwin x86-32 ABI. Use a better type for a function returning a struct containing only a pointer. Handle the edge case of a struct containing only a float or double plus some dead padding instead of asserting.
llvm-svn: 144960
2011-11-18 01:25:50 +00:00
Eli Friedman
9f061a37bb Ignore empty unions in argument lowering on x86-32. From gcc struct layout tests.
llvm-svn: 144944
2011-11-18 00:28:11 +00:00
Akira Hatanaka
f3879ee666 Use a single integer type for a sub-doubleword part of a byval structure.
llvm-svn: 143666
2011-11-03 23:31:00 +00:00
Jan Wen Voung
180319f72d Do not add "byval" attribute to records with non-trivial copy constructors
and destructors in the DefaultABIInfo.

llvm-svn: 143601
2011-11-03 00:59:44 +00:00
Akira Hatanaka
756ce7f9ab Set MinABIStackAlignInBytes to 8 if ABI is N32/64.
llvm-svn: 143597
2011-11-03 00:05:50 +00:00
Akira Hatanaka
101f70da7c Convert the type of a structure passed by value if it has double precision
fields in order to ease handling of such structures in backend. 

llvm-svn: 143596
2011-11-02 23:54:49 +00:00
Akira Hatanaka
1437852c38 Return function results whose size is smaller than 128-bits in registers if ABI
is N32/64.

llvm-svn: 143589
2011-11-02 23:14:57 +00:00
Bob Wilson
1d9269a8d6 Check for homogeneous aggregate return values with ARM's AAPCS-VFP ABI.
llvm-svn: 143530
2011-11-02 04:51:36 +00:00
Dan Gohman
53bb0ff685 Remove the SystemZ backend.
llvm-svn: 142879
2011-10-24 23:48:52 +00:00
Peter Collingbourne
adcf7c9203 TCE target: conversion of OpenCL reqd_work_group_size attr to metadata
Patch by Pekka Jääskeläinen!

llvm-svn: 141865
2011-10-13 16:24:41 +00:00
Richard Smith
caf3390d44 Constant expression evaluation refactoring:
- Remodel Expr::EvaluateAsInt to behave like the other EvaluateAs* functions,
   and add Expr::EvaluateKnownConstInt to capture the current fold-or-assert
   behaviour.
 - Factor out evaluation of bitfield bit widths.
 - Fix a few places which would evaluate an expression twice: once to determine
   whether it is a constant expression, then again to get the value.

llvm-svn: 141561
2011-10-10 18:28:20 +00:00
Ted Kremenek
c168e50722 Free 'TheTargetCodeGenInfo' in the class CodeGenModule. This looks like an obvious memory leak that was reported from LLDB devs. The comment indicates the leak is deliberate, but I have no idea why this needs to be so. Please comment/revert if you know otherwise.
llvm-svn: 141479
2011-10-08 05:28:26 +00:00
Peter Collingbourne
5bad4afa2f CUDA: set proper calling conventions for PTX
llvm-svn: 141296
2011-10-06 16:49:54 +00:00
Justin Holewinski
38031978b5 PTX: Set proper calling conventions for PTX in OpenCL mode.
llvm-svn: 141193
2011-10-05 17:58:44 +00:00
David Blaikie
aa347f9392 Removing a bunch of dead returns/breaks after llvm_unreachables.
llvm-svn: 140407
2011-09-23 20:26:49 +00:00
David Blaikie
83d382b1ca Switch assert(0/false) llvm_unreachable.
llvm-svn: 140367
2011-09-23 05:06:16 +00:00
John McCall
cbc038a6c3 ANSI C requires that a call to an unprototyped function type succeed
if the definition has a non-variadic prototype with compatible
parameters.  Therefore, the default rule for such calls must be to
use a non-variadic convention.  Achieve this by casting the callee to
the function type with which it is required to be compatible, unless
the target specifically opts out and insists that unprototyped calls
should use the variadic rules.  The only case of that I'm aware of is
the x86-64 convention, which passes arguments the same way in both
cases but also sets a small amount of extra information;  here we seek
to maintain compatibility with GCC, which does set this when calling
an unprototyped function.

Addresses PR10810 and PR10713.

llvm-svn: 140241
2011-09-21 08:08:30 +00:00
Akira Hatanaka
ec11b4fa5a Create a MipsTargetCodeGenInfo object for mips64/mips64el. The size of the
UnwindException structure is 32 for mips64. 

llvm-svn: 140165
2011-09-20 18:30:57 +00:00
Akira Hatanaka
0486db08bd Add a parameter to MipsTargetCodeGenInfo's constructor.
llvm-svn: 140161
2011-09-20 18:23:28 +00:00
Douglas Gregor
e8bbc12152 Extend the ASTContext constructor to delay the initialization of
builtin types (When requested). This is another step toward making
ASTUnit build the ASTContext as needed when loading an AST file,
rather than doing so after the fact. No actual functionality change (yet).

llvm-svn: 138985
2011-09-02 00:18:52 +00:00
John McCall
3480ef24d1 The size of struct UnwindException varies by platform with no
apparent general rule.  Just special-case it as appropriate.
PR10789.

llvm-svn: 138792
2011-08-30 01:42:09 +00:00