The GCC construct __attribute__((aligned)) is defined to set alignment
to "the default alignment for the target architecture" according to
the GCC documentation:
The default alignment is sufficient for all scalar types, but may not be
enough for all vector types on a target that supports vector operations.
The default alignment is fixed for a particular target ABI.
clang currently hard-coded an alignment of 16 bytes for that construct,
which is correct on some platforms (including X86), but wrong on others
(including SystemZ). Since this value is ABI-relevant, it is important
to get correct for compatibility purposes.
This patch adds a new TargetInfo member "DefaultAlignForAttributeAligned"
that targets can set to the appropriate default __attribute__((aligned))
value.
Note that I'm deliberately *not* using the existing "SuitableAlign"
value, which is used to set the pre-defined macro __BIGGEST_ALIGNMENT__,
since those two values may not be the same on all platforms. In fact,
on X86, __attribute__((aligned)) always uses 16-byte alignment, while
__BIGGEST_ALIGNMENT__ may be larger if AVX-2 or AVX-512 are supported.
(This is actually not yet correctly implemented in clang either.)
The patch provides a value for DefaultAlignForAttributeAligned only for
SystemZ, and leaves the default for all other targets at 16, which means
no visible change in behavior on all other targets. (The value is still
wrong for some other targets, but I'd prefer to leave it to the target
maintainers for those platforms to fix.)
llvm-svn: 235397
This is a tool for checking consistency of code generation with different
compiler options (such as -g or outputting to .s). This tool has found a number
of code generation issues. The script acts as a wrapper to clang or clang++
performing 2 (or more) compiles then comparing the object files. Instructions
for use are in check_cfc.py including how to use with LNT.
Differential Revision: http://reviews.llvm.org/D8723
llvm-svn: 233919
We know all subclasses in tblgen so just generate a giant switch for
the few virtual methods or turn them into a member variable using spare
bits. The giant jump tables aren't pretty but still much smaller than
a vtable for every attribute, shrinking Release+Asserts clang by ~400k.
Also halves the size of the Attr base class. No functional change
intended.
llvm-svn: 232726
Now that SmallString is a first-class citizen, most SmallString::str()
calls are not required. This patch removes a whole bunch of them, yet
there are lots more.
There are two use cases where str() is really needed:
1) To use one of StringRef member functions which is not available in
SmallString.
2) To convert to std::string, as StringRef implicitly converts while
SmallString do not. We may wish to change this, but it may introduce
ambiguity.
llvm-svn: 232622
We do not implicitly create an OpenCLImageAccessAttr, so this change only affects out of tree users. There is no way to test this behavior specifically that I can see, since this only affects implicit creation of attributes.
Fixes PR22403.
llvm-svn: 231803
This attribute serves as a hint to improve warnings about the ranges of
enumerators used as flag types. It currently has no working C++ implementation
due to different semantics for enums in C++. For more explanation, see the docs
and testcases.
Reviewed by Aaron Ballman.
llvm-svn: 222906
Instead of manually maintaining a flag indicating whether we're about to print
out the last child of the parent node (to determine whether we print "`" or
"|"), capture a callable to print that child and defer printing it until we
either see a next child or finish the parent.
No functionality change intended.
llvm-svn: 220930
Previously loop hints such as #pragma loop vectorize_width(#) required a constant. This patch allows a constant expression to be used as well. Such as a non-type template parameter or an expression (2 * c + 1).
Reviewed by Richard Smith
llvm-svn: 219589
When generating records/unions, the same enum type may be generated more
than once (with different names). In these cases, the name of the enum
values are not sufficiently unique to prevent multiple declarations. E.g:
typedef enum T3 { enum0val0 } T3;
typedef T3 T2[3];
typedef enum T4 { enum0val0 } T4;
typedef union T1 { T2 field0; T4 field1; char field2; } T1;
Added a unique suffix to enum values so that multiple identical enum types do
not use the same enum value names.
One example of this bug is produced by:
ABITestGen.py --no-unsigned --no-vector --no-complex --no-bool \
--max-args 0 --max-record-depth 1 -o inputs/test.9921.a.c \
-T inputs/test.9921.b.c -D inputs/test.9921.driver.c \
--min=9921 --count=1
llvm-svn: 216166
This function might be a bit easier if it were split in two with a lot
of early returns - and that setOptional bit in the outer function, but
anyway.
llvm-svn: 215263
Updating the diagnostics in the launch_bounds test since they have been improved in that case. Adding a test for nonnull since it has little test coverage, but has truly variadic arguments.
llvm-svn: 214407
The NEON intrinsics in arm_neon.h are designed to work on vectors
"as-if" loaded by (V)LDR. We load vectors "as-if" (V)LD1, so the
intrinsics are currently incorrect.
This patch adds big-endian versions of the intrinsics that does the
"obvious but dumb" thing of reversing all vector inputs and all
vector outputs. This will produce extra REVs, but we trust the
optimizer to remove them.
llvm-svn: 211893
There comes a time in the life of any amateur code generator when dumb string
concatenation just won't cut it any more. For NeonEmitter.cpp, that time has
come.
There were a bunch of magic type codes which meant different things depending on
the context. There were a bunch of special cases that really had no reason to be
there but the whole thing was so creaky that removing them would cause something
weird to fall over. There was a 1000 line switch statement for code generation
involving string concatenation, which actually did lexical scoping to an extent
(!!) with a bunch of semi-repeated cases.
I tried to refactor this three times in three different ways without
success. The only way forward was to rewrite the entire thing. Luckily the
testing coverage on this stuff is absolutely massive, both with regression tests
and the "emperor" random test case generator.
The main change is that previously, in arm_neon.td a bunch of "Operation"s were
defined with special names. NeonEmitter.cpp knew about these Operations and
would emit code based on a huge switch. Actually this doesn't make much sense -
the type information was held as strings, so type checking was impossible. Also
TableGen's DAG type actually suits this sort of code generation very well
(surprising that...)
So now every operation is defined in terms of TableGen DAGs. There are a bunch
of operators to use, including "op" (a generic unary or binary operator), "call"
(to call other intrinsics) and "shuffle" (take a guess...). One of the main
advantages of this apart from making it more obvious what is going on, is that
we have proper type inference. This has two obvious advantages:
1) TableGen can error on bad intrinsic definitions easier, instead of just
generating wrong code.
2) Calls to other intrinsics are typechecked too. So
we no longer need to work out whether the thing we call needs to be the Q-lane
version or the D-lane version - TableGen knows that itself!
Here's an example: before:
case OpAbdl: {
std::string abd = MangleName("vabd", typestr, ClassS) + "(__a, __b)";
if (typestr[0] != 'U') {
// vabd results are always unsigned and must be zero-extended.
std::string utype = "U" + typestr.str();
s += "(" + TypeString(proto[0], typestr) + ")";
abd = "(" + TypeString('d', utype) + ")" + abd;
s += Extend(utype, abd) + ";";
} else {
s += Extend(typestr, abd) + ";";
}
break;
}
after:
def OP_ABDL : Op<(cast "R", (call "vmovl", (cast $p0, "U",
(call "vabd", $p0, $p1))))>;
As an example of what happens if you do something wrong now, here's what happens
if you make $p0 unsigned before the call to "vabd" - that is, $p0 -> (cast "U",
$p0):
arm_neon.td:574:1: error: No compatible intrinsic found - looking up intrinsic 'vabd(uint8x8_t, int8x8_t)'
Available overloads:
- float64x2_t vabdq_v(float64x2_t, float64x2_t)
- float64x1_t vabd_v(float64x1_t, float64x1_t)
- float64_t vabdd_f64(float64_t, float64_t)
- float32_t vabds_f32(float32_t, float32_t)
... snip ...
This makes it seriously easy to work out what you've done wrong in fairly nasty
intrinsics.
As part of this I've massively beefed up the documentation in arm_neon.td too.
Things still to do / on the radar:
- Testcase generation. This was implemented in the previous version and not in
the new one, because
- Autogenerated tests are not being run. The testcase in test/ differs from
the autogenerated version.
- There were a whole slew of special cases in the testcase generation that just
felt (and looked) like hacks.
If someone really feels strongly about this, I can try and reimplement it too.
- Big endian. That's coming soon and should be a very small diff on top of this one.
llvm-svn: 211101
By describing system header suppressions directly in tablegen we eliminate
special cases in getDiagnosticSeverity().
Dropping the reliance on builtin diagnostic classes when mapping also gets us
closer to the goal of reusing the diagnostic machinery for custom diagnostics.
No change in functionality.
llvm-svn: 211023
hint attributes. Includes tests for pragma printing and for attribute order
which is incorrectly reversed by ParsedAttributes.
Reviewed by Aaron Ballman
llvm-svn: 210925
This begins to address cognitive dissonance caused by treating the Note
diagnostic level as a severity in the diagnostic engine.
No change in functionality.
llvm-svn: 210758
will never be true in a well-defined context. The checking for null pointers
has been moved into the caller logic so it does not rely on undefined behavior.
llvm-svn: 210498
I was bitten by this when working with the dll attributes: when a dll
attribute was cloned from a class template declaration to its
specialization, the Inherited flag didn't get cloned.
Differential Revision: http://reviews.llvm.org/D3972
llvm-svn: 209950
The attribute emitter was using FunctionTemplate to map the diagnostic to "functions or methods", but that isn't a particularly clear diagnostic in these cases anyway (since they do not apply to ObjC methods). Updated the attribute emitter to remove custom logic for FunctionTemplateDecl, and updated the test cases for the change in diagnostic wording.
llvm-svn: 209209
Craig Topper