We want the diagnostic, and if the load is optimized away, we still want to
trap it. Stop checking non-default address spaces; that doesn't work in
general.
llvm-svn: 167219
If HA can only partially fit into VFP registers, we add padding to make sure
HA will be on stack and later VFP CPRCs will be on stack as well.
llvm-svn: 167058
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
varargs parameter passing.
A strict reading of the ABI indicates that any argument with alignment greater
than 8 may require skipping doublewords in the parameter save area to align
the argument, and hence require skipping GPRs. In practice, this is not done
by GCC. The alignment restriction is used for internal alignment of a
structure, but a structure with 16-byte alignment, for example, is not
itself 16-byte aligned in the parameter save area. Although this is messy,
it has become the de facto standard used in building existing libraries.
My initial varargs support followed the ABI language, but not the de facto
standard. Running the GCC compatibility test suite exposed this issue, and
indeed showed that LLVM didn't pass parameters self-consistently with my
original logic. Removing the additional alignment logic allows the affected
tests to now pass.
I modified the ppc64-varargs-struct.c test case to remove the existing test
for generation of alignment code, which is no longer appropriate.
Built and tested on powerpc64-unknown-linux-gnu with no new regressions.
llvm-svn: 166805
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
will be represented in the IR as a plain "i32" type. This causes the
tests to spuriously fail on platforms where int is not a 32-bit type,
or where the ABI requires attributes like "signext" or "zeroext" to
be used.
This patch adds -triple or -target parameters to force those tests
to use the i386-unknown-unknown target.
llvm-svn: 166551
are no known current users of column info. Robustify and fix up
a few tests in the process. Reduces the size of debug information
by a small amount.
Part of PR14106
llvm-svn: 166236
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
We expand varargs in clang and the call site is handled in the back end, it is
hard to match exactly how illegal vectors are handled in the backend. Therefore,
we legalize the illegal vector types in clang:
if (Size <= 32), legalize to i32.
if (Size == 64), legalize to v2i32.
if (Size == 128), legalize to v4i32.
if (Size > 128), use indirect.
rdar://12439123
llvm-svn: 166043
We create an aligned temporary space and copy the content over from ap.cur to
the temporary space. This is necessary if the natural alignment of the type is
greater than the ABI alignment.
rdar://12439123
llvm-svn: 166040
don't try the normal GetOrCreateLLVM. The latter could drop the weak
atrtibute on the second reference, if there is no explicit declaration
of the aliasee.
llvm-svn: 166032
-The front-end now builds a single assembly string and feeds it to the
AsmParser. The front-end iterates on a per statement basis by calling the
ParseStatement() function. Please note, the calling of ParseStatement() and
and any notion of MCAsmParsedOperands will be sunk into the MC layer in the
near future. I plan to expose more basic APIs such as getClobbers, etc.
-The enumeration of the AsmString expressions have been reworked to use SMLocs
rather than assembly Pieces, which were being parsed in the front-end.
-The test case, t8(), was modified due to r129223. I'll have to find a way to
work around things such as these.
Sorry for the large commit, but breaking this in multiple smaller commits proved
too irritating.
llvm-svn: 165957
For 64-bit PowerPC SVR4, an aggregate containing only one
floating-point field (float, double, or long double) must be passed in
a register as though just that field were present. This patch
addresses the issue during Clang code generation by specifying in the
ABIArgInfo for the argument that the underlying type is passed
directly in a register. The included test case verifies flat and
nested structs for the three data types.
llvm-svn: 165816
now unused static helper function.
The test case needs to be remove temporarily until I can better filter memory
operands that aren't actual variable reference.
llvm-svn: 165751
The issue arises when coercing to/from types of different sizes. We need
to be certain that the allocation on either end has sufficient room for
the coerced type. When it doesn't, we need to make room, copy across,
and then proceed. PR11905 handled the case of storing function arguments
back into allocas in the function prolog, this patch handles the case of
setting up the function arguments in a call expression.
This is actually significantly simpler than the fix for PR11905. It ends
up being a trivial change to create a temporary alloca when the source
is too small and memcpy across. This should preserve the compile-time
fast-isel benefits of doing gep+load sequences and avoiding FCAs.
Reviewed by Benjamin and Evgeniy (who fixed PR11905).
llvm-svn: 165615
clang itself. This dates back to clang's early days and while it looks like
some of it is still used (for kext for example), other parts are probably dead.
Remove the -ccc-clang-archs option and associated code. I don't think there
is any remaining setup where clang doesn't support an architecture but it can
expect an working gcc cross compiler to be available.
A nice side effect is that tests no longer need to differentiate architectures
that are included in production builds of clang and those that are not.
llvm-svn: 165545
GCC has always supported this on PowerPC and 4.8 supports it on all platforms,
so it's a good idea to expose it in clang too. LLVM supports this on all targets.
llvm-svn: 165362
- outside C++, return undef (behavior is not undefined unless the value is used)
- in C++, with -fcatch-undefined-behavior, perform an appropriate trap
- in C++, produce an 'unreachable' (behavior is undefined immediately)
llvm-svn: 165273
64-bit PPC SVR4 ABI.
The test verifies passing of structures, items with 16-byte alignment, and
small items that are passed right-justified in the parameter save area slot.
llvm-svn: 165245
Clang will now honor the FP_CONTRACT pragma and emit LLVM
fmuladd intrinsics for expressions of the form A * B + C (when they occur in a
single statement).
llvm-svn: 164989
more robust way to address a few FIXMEs.
The initial implementation, r163342, built the IR asm string and then tried to
patch things on the fly without enough context. Specifically, it didn't skip
mnemonics nor did it track with assembly instruction an expression was related
to. The new implementation patches the operands and then builds the final
IR string.
llvm-svn: 163756
MCOperands then iterate over all of then when computing clobbers, inputs and
outputs.
On x86 the 1-to-many mapping is a memory operand that includes a BaseReg(reg),
MemScale(imm), MemIndexReg(reg), an Expr(MCExpr or imm) and a MemSegReg(reg).
Invalid register (Op.getReg() == 0) are not considered when computing clobber.
llvm-svn: 163728
the trap BB out of the individual checks and into a common function, to prepare
for making this code call into a runtime library. Rename the existing EmitCheck
to EmitTypeCheck to clarify it and to move it out of the way of the new
EmitCheck.
llvm-svn: 163451
This patch uses a new ABIInfo implementation specific to the le32
target, rather than falling back to DefaultABIInfo. Its behavior is
basically the same, but it also allows the regparm argument attribute.
It also includes basic tests for argument codegen and attributes.
llvm-svn: 163333
by this mode, and also check for signed left shift overflow. The rules for the
latter are a little subtle:
* neither C89 nor C++98 specify the behavior of a signed left shift at all
* in C99 and C11, shifting a 1 bit into the sign bit has undefined behavior
* in C++11, with core issue 1457, shifting a 1 bit *out* of the sign bit has
undefined behavior
As of this change, we use the C99 rules for all C language variants, and the
C++11 rules for all C++ language variants. Once we have individual
-fcatch-undefined-behavior= flags, this should be revisited.
llvm-svn: 162634
* when checking that a pointer or reference refers to appropriate storage for a type, also check the alignment and perform a null check
* check that references are bound to appropriate storage
* check that 'this' has appropriate storage in member accesses and member function calls
llvm-svn: 162523
The conditions described by POSIX can never happen with IEEE-754 floats.
When the function is const we can emit a single sse4.1 instruction for
it, without losing anything :)
llvm-svn: 162379
There were missed optimizations when the system headers didn't have attributes
in place, specifically:
- Add copysign, exp2, log2, nearbyint, rint and trunc to the list.
These are functions that get inlined by LLVM's optimizer, but only when they
have the right attributes.
- Mark copysign, fabs, fmax, fmin and trunc const unconditionally.
Previously these were only const with -fno-math-errno, but they never set
errno per POSIX.
For ceil/floor/nearbyint/round I'm not aware of any implementation that sets
errno, but POSIX says it may signal overflow so I left them alone for now.
llvm-svn: 162375
These require special handling, which we don't currently handle. This is being
put in place to ensure we don't do invalid symbol table lookups or try to parse
invalid assembly. The test cases just makes sure the latter isn't happening.
llvm-svn: 162050
variables, function or label references. The former is a potential clobber.
The latter is either an input or an output. Unfortunately, it's difficult to
test this patch at the moment, but the added test case will eventually do so.
llvm-svn: 162026
statement. For example,
if (x)
__asm out dx, ax __asm out dx, ax
results in a single inline asm statement (i.e., both "out dx, ax" statements are
predicated on if(x)).
llvm-svn: 161986
The backend has to legalize i64 types by splitting them into two 32-bit pieces,
which leads to poor quality code. If we produce code for these intrinsics that
uses one-element vector types, which can live in Neon vector registers without
getting split up, then the generated code is much better. Radar 11998303.
llvm-svn: 161879
The AsmParser expects a single asm instruction, but valid ms-style inline asm
statements may contain multiple instructions.
This happens with asm blocks
__asm {
mov ebx, eax
mov ecx, ebx
}
or when multiple asm statements are adjacent to one another
__asm mov ebx, eax
__asm mov ecx, ebx
and
__asm mov ebx, eax __asm mov ecx, ebx
Currently, asm blocks are not properly handled.
llvm-svn: 161780
attribute. It is a variation of the x86_64 ABI:
* A struct returned indirectly uses the first register argument to pass the
pointer.
* Floats, Doubles and structs containing only one of them are not passed in
registers.
* Other structs are split into registers if they fit on the remaining ones.
Otherwise they are passed in memory.
* When a struct doesn't fit it still consumes the registers.
llvm-svn: 161022
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
there's something going on there. Remove the unconditional line entry
and only add one if we're emitting cleanups (any other statements
would be handled normally).
Fixes rdar://9199234
llvm-svn: 160033
Lots of tests are using an explicit target triple w/o first checking that the
target is actually available. Add a REQUIRES clause to a bunch of them. This should
hopefully unbreak bots which don't configure w/ all targets enabled.
llvm-svn: 159949
This flag sets the 'fp-contract' mode, which controls the formation of fused
floating point operations. Available modes are:
- Fast: Form fused operations anywhere.
- On: Form fused operations where allowed by FP_CONTRACT. This is the default
mode.
- Off: Don't form fused operations (in future this may be relaxed to forming
fused operations where it can be proved that the result won't be
affected).
Currently clang doesn't support the FP_CONTRACT pragma, so the 'On' and 'Off'
modes are equivalent.
llvm-svn: 159794
values:
- Return integer vectors in integer registers.
- Pass vector arguments in integer registers.
- Set an upper bound for argument alignment. The largest alignment is 8-byte
for O32 and 16-byte for N32/64.
llvm-svn: 159676
if we want to ignore a result, the Dest will be null. Otherwise,
we must copy into it. This means we need to ensure a slot when
loading from a volatile l-value.
With all that in place, fix a bug with chained assignments into
__block variables of aggregate type where we were losing insight into
the actual source of the value during the second assignment.
llvm-svn: 159630
Corrected type for index of _mm256_mask_i32gather_pd
from 256-bit to 128-bit
Corrected types for src|dst|mask of _mm256_mask_i64gather_ps
from 256-bit to 128-bit
Support the following intrinsics:
_mm_mask_i32gather_epi64, _mm256_mask_i32gather_epi64,
_mm_mask_i64gather_epi64, _mm256_mask_i64gather_epi64,
_mm_mask_i32gather_epi32, _mm256_mask_i32gather_epi32,
_mm_mask_i64gather_epi32, _mm256_mask_i64gather_epi32
llvm-svn: 159403
Support the following intrinsics:
_mm_mask_i32gather_pd, _mm256_mask_i32gather_pd, _mm_mask_i64gather_pd
_mm256_mask_i64gather_pd, _mm_mask_i32gather_ps, _mm256_mask_i32gather_ps
_mm_mask_i64gather_ps, _mm256_mask_i64gather_ps
llvm-svn: 159222
literal helper functions. All helper functions (global
and locals) use block_invoke as their prefix. Local literal
helper names are prefixed by their enclosing mangled function
names. Blocks in non-local initializers (e.g. a global variable
or a C++11 field) are prefixed by their mangled variable name.
The descriminator number added to end of the name starts off
with blank (for first block) and _<N> (for the N+2-th block).
llvm-svn: 159206
This adds support for the tls_model attribute. This allows the user to
choose a TLS model that is better than what LLVM would select by
default. For example, a variable might be declared as:
__thread int x __attribute__((tls_model("initial-exec")));
if it will not be used in a shared library that is dlopen'ed.
This depends on LLVM r159077.
llvm-svn: 159078
The original r158700 caused crashes in the gcc test suite,
g++.abi/vtable3a.C among others. It also caused failures in the libc++
test suite.
llvm-svn: 158749
pointer, but such folding encounters side-effects, ignore the side-effects
rather than performing them at runtime: CodeGen generates wrong code for
__builtin_object_size in that case.
llvm-svn: 157310
Currently cold functions are marked with the "optsize" attribute in CodeGen
so they are always optimized for size. The hot attribute is just ignored,
LLVM doesn't have a way to express hotness at the moment.
llvm-svn: 156723
A vector should be returned via the hidden pointer argument except if its size
is equal to or smaller than 16-bytes and the target ABI is N32 or N64.
llvm-svn: 156642
It reduces the amount of emitted debug information:
1) DIEs in .debug_info have types DW_TAG_compile_unit, DW_TAG_subprogram,
DW_TAG_inlined_subroutine (for opt builds) and DW_TAG_lexical_block only.
2) .debug_str contains only function names.
3) No debug data for types/namespaces/variables is emitted.
4) The data in .debug_line is enough to produce valid stack traces with
function names and line numbers.
Reviewed by Eric Christopher.
llvm-svn: 156160
goodness because it provides opportunites to cleanup things. For example,
uint64_t t1(__m128i vA)
{
uint64_t Alo;
_mm_storel_epi64((__m128i*)&Alo, vA);
return Alo;
}
was generating
movq %xmm0, -8(%rbp)
movq -8(%rbp), %rax
and now generates
movd %xmm0, %rax
rdar://11282581
llvm-svn: 155924
i32 __builtin_annotation(i32, string);
Applying it to i64 (e.g., long long) generates the following IR.
trunc i64 {{.*}} to i32
call i32 @llvm.annotation.i32
zext i32 {{.*}} to i64
The redundant truncation and extension make the result difficult to use.
This patch makes __builtin_annotation() generic.
type __builtin_annotation(type, string);
For the i64 example, it simplifies the generated IR to:
call i64 @llvm.annotation.i64
Patch by Xi Wang!
llvm-svn: 155764
more aligned than the block header but also contains something
smaller than the block-header alignment but not exactly half
the difference between the large alignment and the header
alignment. Got that?
I'm really not sure what I was thinking with the buggy computation
here, but the fix is pretty obvious.
llvm-svn: 155662
With -fno-math-errno (the default for Darwin) or -ffast-math these library
function can be marked readnone enabling more opportunities for CSE and other
optimizations.
rdar://11251464
llvm-svn: 155498
NAKAMURA Takumi