it down. we effectively were compile the testcase into:
void test14(int x) {
switch (x) {
case 11: break;
case 42: test14(97); // fallthrough
default: test14(42); break;
which is not the same thing at all. This fixes a miscompilation of
MallocBench/gs seen on the clang-x86_64-linux-fnt buildbot.
llvm-svn: 129679
live case of a switch statement when switching on a constant. This is terribly
limited, but enough to handle the trivial example included. Before we would
emit:
define void @test1(i32 %i) nounwind {
entry:
%i.addr = alloca i32, align 4
store i32 %i, i32* %i.addr, align 4
switch i32 1, label %sw.epilog [
i32 1, label %sw.bb
]
sw.bb: ; preds = %entry
%tmp = load i32* %i.addr, align 4
%inc = add nsw i32 %tmp, 1
store i32 %inc, i32* %i.addr, align 4
br label %sw.epilog
sw.epilog: ; preds = %sw.bb, %entry
switch i32 0, label %sw.epilog3 [
i32 1, label %sw.bb1
]
sw.bb1: ; preds = %sw.epilog
%tmp2 = load i32* %i.addr, align 4
%add = add nsw i32 %tmp2, 2
store i32 %add, i32* %i.addr, align 4
br label %sw.epilog3
sw.epilog3: ; preds = %sw.bb1, %sw.epilog
ret void
}
now we emit:
define void @test1(i32 %i) nounwind {
entry:
%i.addr = alloca i32, align 4
store i32 %i, i32* %i.addr, align 4
%tmp = load i32* %i.addr, align 4
%inc = add nsw i32 %tmp, 1
store i32 %inc, i32* %i.addr, align 4
ret void
}
This improves -O0 compile time (less IR to generate and shove through the code
generator) and the clever linux kernel people found a way to fail to build if we
don't do this optimization. This step isn't enough to handle the kernel case
though.
llvm-svn: 126597
LabelDecl and LabelStmt. There is a 1-1 correspondence between the
two, but this simplifies a bunch of code by itself. This is because
labels are the only place where we previously had references to random
other statements, causing grief for AST serialization and other stuff.
This does cause one regression (attr(unused) doesn't silence unused
label warnings) which I'll address next.
This does fix some minor bugs:
1. "The only valid attribute " diagnostic was capitalized.
2. Various diagnostics printed as ''labelname'' instead of 'labelname'
3. This reduces duplication of label checking between functions and blocks.
Review appreciated, particularly for the cindex and template bits.
llvm-svn: 125733
there were only three virtual methods of any significance.
The primary way to grab child iterators now is with
Stmt::child_range children();
Stmt::const_child_range children() const;
where a child_range is just a std::pair of iterators suitable for
being llvm::tie'd to some locals. I've left the old child_begin()
and child_end() accessors in place, but it's probably a substantial
penalty to grab the iterators individually now, since the
switch-based dispatch is kindof inherently slower than vtable
dispatch. Grabbing them together is probably a slight win over the
status quo, although of course we could've achieved that with vtables, too.
I also reclassified SwitchCase (correctly) as an abstract Stmt
class, which (as the first such class that wasn't an Expr subclass)
required some fiddling in a few places.
There are somewhat gross metaprogramming hooks in place to ensure
that new statements/expressions continue to implement
getSourceRange() and children(). I had to work around a recent clang
bug; dgregor actually fixed it already, but I didn't want to
introduce a selfhosting dependency on ToT.
llvm-svn: 125183
delete the block we began emitting into if it had no predecessors. We never
want to do this, because there are several valid cases during statement
emission where an existing block has no known predecessors but will acquire
some later. The case in my test case doesn't inherently fall into this
category, because we could safely emit the case-range code before the statement
body, but there are examples with labels that can't be fallen into
that would also demonstrate this bug.
rdar://problem/8837067
llvm-svn: 123303
in asm statements:
register int foo asm("rdi");
asm("..." : ... "r" (foo) ...
We also only accept these variables if the constraint in the asm statement is "r".
This fixes most of PR3933.
llvm-svn: 122643
Fix a bug in the emission of complex compound assignment l-values.
Introduce a method to emit an expression whose value isn't relevant.
Make that method evaluate its operand as an l-value if it is one.
Fixes our volatile compliance in C++.
llvm-svn: 120931
of all the lines of the inline asm. With the refactoring and enhancement
of the backend, we can now reports errors on the correct source line when
an asm contains multiple lines of text. For something like this:
void foo() {
asm("push %rax\n"
".code32\n");
}
we used to get this: (note that the line 4 in t.c isn't helpful)
t.c:4:7: error: warning: ignoring directive for now
asm("push %rax\n"
^
<inline asm>:2:1: note: instantiated into assembly here
.code32
^
now we get:
t.c:5:8: error: warning: ignoring directive for now
".code32\n"
^
<inline asm>:2:1: note: instantiated into assembly here
.code32
^
Note that we're pointing to line 5 properly now. This implements
rdar://7839391 - inline asm errors should point to the right line in the asm
and makes the error message in PR8595 much less confusing.
llvm-svn: 119489
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
in the scope checker. With that done, turn an indirect goto into a
protected scope into a hard error; otherwise IR generation has to start
worrying about declarations not dominating their scopes, as exemplified
in PR8473.
If this really affects anyone, I can probably adjust this to only hard-error
on possible indirect gotos into VLA scopes rather than arbitrary scopes.
But we'll see how people cope with the aggressive change on the marginal
feature.
llvm-svn: 117539
slot. The easiest way to do that was to bundle up the information
we care about for aggregate slots into a new structure which demands
that its creators at least consider the question.
I could probably be convinced that the ObjC 'needs GC' bit should
be rolled into this structure.
Implement generalized copy elision. The main obstacle here is that
IR-generation must be much more careful about making sure that exactly
llvm-svn: 113962
self-host. Hopefully these results hold up on different platforms.
I tried to keep the GNU ObjC runtime happy, but it's hard for me to test.
Reimplement how clang generates IR for exceptions. Instead of creating new
invoke destinations which sequentially chain to the previous destination,
push a more semantic representation of *why* we need the cleanup/catch/filter
behavior, then collect that information into a single landing pad upon request.
Also reorganizes how normal cleanups (i.e. cleanups triggered by non-exceptional
control flow) are generated, since it's actually fairly closely tied in with
the former. Remove the need to track which cleanup scope a block is associated
with.
Document a lot of previously poorly-understood (by me, at least) behavior.
The new framework implements the Horrible Hack (tm), which requires every
landing pad to have a catch-all so that inlining will work. Clang no longer
requires the Horrible Hack just to make exceptions flow correctly within
a function, however. The HH is an unfortunate requirement of LLVM's EH IR.
llvm-svn: 107631
have CGF create and make accessible standard int32,int64 and
intptr types. This fixes a ton of 80 column violations
introduced by LLVMContextification and cleans up stuff a lot.
llvm-svn: 106977
This works around a crash where malloc reused the memory of an erased BB for a
new BB leaving old cleanup information pointing at the new block.
llvm-svn: 104472
return statements. We perform NRVO only when all of the return
statements in the function return the same variable. Fixes some link
failures in Boost.Interprocess (which is relying on NRVO), and
probably improves performance for some C++ applications.
llvm-svn: 103867
input and output types when the smaller value isn't mentioned in the
asm string. Extend this support from integers to also allowing
fp values to be mismatched (if not mentioned in the asm string).
llvm-svn: 102188
(if there's a current block). The chief advantage of doing this is that it
lets us pick blocks (e.g. EH blocks) to push to the end of the function so
that fallthrough happens consistently --- i.e. it gives us the flexibility
of ordering blocks as we please without having to change the order in which
we generate code. There are standard (?) optimization passes which can do some
of that for us, but better to generate reasonable code to begin with.
llvm-svn: 101997
have the code generate slap a srcloc metadata on inline asm nodes.
This allows us to diagnose invalid inline asms with such nice
diagnostics as:
<inline asm>:1:2: error: unrecognized instruction
abc incl %eax
^
asm.c:2:12: note: generated from here
__asm__ ("abc incl %0" : "+r" (X));
^
2 diagnostics generated.
llvm-svn: 100608
EmitReferenceBindingToExpr() rather than assuming we have an
lvalue. This is just the lowest hanging fruit for PR6024, which still
requires a bit of work.
llvm-svn: 99447
need to deal with aggregates specially; this is consistent with the rest of IRgen.
Also, simplify EmitParmDecl and don't worry about using Decl::getNameAsString.
llvm-svn: 95393
All statements that involve conditions can now hold on to a separate
condition declaration (a VarDecl), and will use a DeclRefExpr
referring to that VarDecl for the condition expression. ForStmts now
have such a VarDecl (I'd missed those in previous commits).
Also, since this change reworks the Action interface for
if/while/switch/for, use FullExprArg for the full expressions in those
expressions, to ensure that we're emitting
Note that we are (still) not generating the right cleanups for
condition variables in for statements. That will be a follow-on
commit.
llvm-svn: 89817
cleanups for while loops:
1) Make sure that we destroy the condition variable of a while statement each time through the loop for, e.g.,
while (shared_ptr<WorkInt> p = getWorkItem()) {
// ...
}
2) Make sure that we always enter a new cleanup scope for the body of the while loop, even when there is no compound expression, e.g.,
while (blah)
RAIIObject raii(blah+1);
llvm-svn: 89800
- Outside the "if", to ensure that we destroy the condition variable
at the end of the "if" statement rather than at the end of the
block containing the "if" statement.
- Inside the "then" and "else" branches, so that we emit then- or
else-local cleanups at the end of the corresponding block when the
block is not a compound statement.
To make adding these new cleanup scopes easier (and since
switch/do/while will all need the same treatment), added the
CleanupScope RAII object to introduce a new cleanup scope and make
sure it gets cleaned up.
llvm-svn: 89773
rather than burying it in a CXXConditionDeclExpr (that occassionally
hides behind implicit conversions). Similar changes for
switch, while, and do-while will follow, then the removal of
CXXConditionDeclExpr. This commit is the canary.
llvm-svn: 89717
using the new LLVM support for this. This is temporarily hiding
behind horrible and ugly #ifdefs until the time when the optimizer
is stable (hopefully a week or so). Until then, lets make it "opt in" :)
llvm-svn: 85446
1. CGF now has fewer bytes of state (one pointer instead of a vector).
2. The generated code is determinstic, instead of getting labels in
'map order' based on pointer addresses.
3. Clang now emits one 'indirect goto switch' for each function, instead
of one for each indirect goto. This fixes an M*N = N^2 IR size issue
when there are lots of address-taken labels and lots of indirect gotos.
4. This also makes the default cause do something useful, reducing the
size of the jump table needed (by one).
llvm-svn: 83952
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
- 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
function calls. For a program like this:
#include <stdio.h>
static __inline__ __attribute__((always_inline))
int bar(int x) { return 4; }
int main() {
int X = bar(4);
printf("%d\n", X);
}
clang was not outputing any debug info for the body of main(). This is
because the backend is getting confused by the region_start/end that clang
is emitting for block scopes. For now, just disable these (matching llvm-gcc),
this stuff is in progress of rework anyway.
llvm-svn: 70889
- <rdar://problem/6732143> Crash when generating @synchronize for
zero-cost exception
- Thanks to Anders for helping track down the problem.
llvm-svn: 68186
really horrible extensions that are disabled by default but that can
be accepted by -fheinous-gnu-extensions (but which always emit a
warning when enabled).
As our first instance of this, implement PR3788/PR3794, which allows
non-lvalues in inline asms in contexts where lvalues are required. bleh.
llvm-svn: 66910
done in sema, and is reflected by the existing PR3258. In the meantime,
fix PR3682 by disabling a bogus assertion (which doesn't account for +
operands).
llvm-svn: 66533
asm. This allows us to properly handle the case when an optimizer duplicates
the asm, such as here:
void bar() {
int i;
for (i = 0; i < 3; ++i)
asm("foo %=" : : "r"(0));
}
we now produce:
_bar:
xorl %eax, %eax
## InlineAsm Start
foo 0
## InlineAsm End
## InlineAsm Start
foo 1
## InlineAsm End
## InlineAsm Start
foo 2
## InlineAsm End
ret
instead of:
_bar:
xorl %eax, %eax
## InlineAsm Start
foo 1
## InlineAsm End
## InlineAsm Start
foo 1
## InlineAsm End
## InlineAsm Start
foo 1
## InlineAsm End
ret
This also fixes a fixme by eliminating a static.
llvm-svn: 66528
it in the stack trace, giving us stuff like:
Stack dump:
0. Program arguments: clang t.c -emit-llvm
1. <eof> parser at end of file
2. t.c:1:5: LLVM IR generation of declaration 'a'
3. t.c:1:9: LLVM IR generation of compound statement ('{}')
4. t.c:2:3: LLVM IR generation of compound statement ('{}')
Abort
llvm-svn: 66154
If people could beat on it and let me know if there are any new
semantics required by newer language standards or DRs or any little
details I goofed on, I'd be happy to fix any issues found.
llvm-svn: 64079
string.
That said, we should probably try and track down the correct clobber
lists for the targets that don't have them (PPC, ARM, and Sparc),
so that we can generate correct code.
llvm-svn: 61298
landing pads.
- Primarily a cleanliness issue instead of a performance issue (this
eliminates all blocks w/o predecessors on 176.gcc/expr.c), but this
also allows subsequent code to recognize it is unreachable and
potentially avoid IRgen.
llvm-svn: 59211
- Use dotted notation for blocks related to a particular statement
type.
- Use .end for landing pads.
No functionality change in NDEBUG mode. :)
llvm-svn: 59210
- Indicates that caller is done with the block and it can be dropped
if it has no predecessors. Useful for callers who need to make
landing pads but which may not be reached.
No functionality change.
llvm-svn: 59207
- Split out "simple" statements which can easily handle IR generation
when there is no insert point. These are generally statements which
start by emitting a new block or are only containers for other
statements.
- This fixes a regression in emitting dummy blocks, notably for case
statements.
- This also fixes spurious emission of a number of debug stoppoint
intrinsic instructions.
Remove unneeded sw.body block, just clear the insertion point.
Lift out CodeGenFunction::EmitStopPoint which calls into the
CGDebugInfo class when generating debug info.
Normalize definitions of Emit{Break,Continue}Stmt and usage of
ErrorUnsupported.
llvm-svn: 59118
CodeGenFunction.cpp. Change VisitConditionalOperator to use
constant fold instead of codegen'ing a constant conditional.
Change ForStmt to use EmitBranchOnBoolExpr, this shrinks
expr.c very slightly to 40239 lines.
llvm-svn: 59113
- EmitStmt is no longer required to finish with a current insertion
point defined (i.e. it does not need to make dummy
blocks). Instead, it can clear the insertion point in the builder
which indicates that the current insertion point is unreachable.
- CodeGenFunction provides HaveInsertPoint and EnsureInsertPoint
which respectively test if there is an insert point and ensure an
insertion point exists (by making a dummy block).
- Clearly mark functions in CodeGenFunction which can be called with
no insertion point defined. Currently this is a limited set, and
EmitStmt simply EnsureInsertPoint()s before emitting subsequent IR.
Remove EmitDummyBlock, which is no longer needed. Clients who haven't
already cleared the insertion point (typically via EmitBranch) can do
so by hand.
Remove isDummyBlock, which has effectively been renamed to
HaveInsertPoint.
The main thrust of this change is that we no longer have create dummy
blocks just to destroy them a short time later in EmitBlock in the
common case that there is no unreachable code following something like
a goto.
Additionally, this means that we are not using the hokey condition in
isDummyBlock that a block without a name is a dummy block. Guess how
well that works when we never emit block names!
llvm-svn: 59089
the size of the -O0 output on some cases. For example, on expr.c from
176.gcc, it shrinks the .ll file from 43164 to 42835 lines, and removed
references to two external symbols.
llvm-svn: 59034
- Provides a basic primitive to jump to an arbitrary basic block,
through the finally code.
- Only used for return statements and rethrow currently. Still need
to handle break, continue and goto.
- Code still needs to be shuffled around to live elsewhere.
llvm-svn: 56827
- Add CodeGenFunction::{EmitReturnOfRValue, EmitIvarOffset}
- Factor EmitLValueForIvar out of EmitObjCIvarRefLValue.
No non-error functionality change (some unsupported errors are
degraded to asserts for the time being).
llvm-svn: 56543
Implement Obj-C lvalue message sends (aggregate returns).
Update several places to emit more precise ErrorUnsupported warnings
for currently unimplemented Obj-C features (main missing chunks are
property references, Obj-C exception handling, and the for ... in
syntax).
llvm-svn: 55234
- 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
- Follows emission scheme used by llvm-gcc, i.e. invent an id for
each label whose address is taken and replace each indirect goto by
a switch to each possible target.
- Currently we emit a switch for each indirect goto instead of
merging them as llvm-gcc does.
llvm-svn: 54318
- Fix multiple issues with the way case ranges were emitted, see test
cases details about the specific issues.
- The root issue was not being careful about how basic blocks were
emitted which led to them being chained together incorrectly,
resulting in improper control flow.
- Fixes <rdar://problem/6098585>
llvm-svn: 54006
- Also cleaned up emission slightly
- Inspection of the code revealed several other bugs, however. Case
ranges are not properly wired and can result in switch cases being
dropped or even infinite loops. See: <rdar://problem/6098585>
Completes: <rdar://problem/6094119>
llvm-svn: 53975
qualifier in the lvalue, and changes lvalue loads/stores to honor
the volatile flag. Places which need some further attention are marked
with FIXMEs.
Patch by Cédric Venet.
llvm-svn: 52264
llvm::Type::isSingleValueType. Currently these two functions have
the same behavior, but soon isFirstClassType will return true for
struct and array types.
Clang may some day want to use of isFirstClassType for some of
these some day as an optimization, but it'll require some
consideration.
llvm-svn: 51446
actually work (instead of crashing llc), and there's enough info emitted
to get line number information in gdb. This should hopefully be helpful
for debugging non-working programs.
I got rid of the begin/endregion calls because the implementation wasn't
working; someone who knows the debugging info a bit better might try to
add it. I really have no clue how a compiler is supposed to emit them.
This commit shouldn't have any effect without -g.
llvm-svn: 51404
Chris Lattner