the key. This will cause it to create a new std::string, which isn't
wanted. Instead, pass back the "const char*". Modify the EmitString() method to
take a "const char*".
llvm-svn: 68741
register destinations that are tied to source operands. The
TargetInstrDescr::findTiedToSrcOperand method silently fails for inline
assembly. The existing MachineInstr::isRegReDefinedByTwoAddr was very
close to doing what is needed, so this revision makes a few changes to
that method and also renames it to isRegTiedToUseOperand (for consistency
with the very similar isRegTiedToDefOperand and because it handles both
two-address instructions and inline assembly with tied registers).
llvm-svn: 68714
with SUBREG_TO_REG, teach SimpleRegisterCoalescing to coalesce
SUBREG_TO_REG instructions (which are similar to INSERT_SUBREG
instructions), and teach the DAGCombiner to take advantage of this on
targets which support it. This eliminates many redundant
zero-extension operations on x86-64.
This adds a new TargetLowering hook, isZExtFree. It's similar to
isTruncateFree, except it only applies to actual definitions, and not
no-op truncates which may not zero the high bits.
Also, this adds a new optimization to SimplifyDemandedBits: transform
operations like x+y into (zext (add (trunc x), (trunc y))) on targets
where all the casts are no-ops. In contexts where the high part of the
add is explicitly masked off, this allows the mask operation to be
eliminated. Fix the DAGCombiner to avoid undoing these transformations
to eliminate casts on targets where the casts are no-ops.
Also, this adds a new two-address lowering heuristic. Since
two-address lowering runs before coalescing, it helps to be able to
look through copies when deciding whether commuting and/or
three-address conversion are profitable.
Also, fix a bug in LiveInterval::MergeInClobberRanges. It didn't handle
the case that a clobber range extended both before and beyond an
existing live range. In that case, multiple live ranges need to be
added. This was exposed by the new subreg coalescing code.
Remove 2008-05-06-SpillerBug.ll. It was bugpoint-reduced, and the
spiller behavior it was looking for no longer occurrs with the new
instruction selection.
llvm-svn: 68576
When compiling in Thumb mode, only the low (R0-R7) registers are available
for most instructions. Breaking the low registers into a new register class
handles this. Uses of R12, SP, etc, are handled explicitly where needed
with copies inserted to move results into low registers where the rest of
the code generator can deal with them.
llvm-svn: 68545
elements in a form that is efficient for the reader to just get a
pointer in memory and start reading. APIs to do efficient reading
and writing are still todo.
llvm-svn: 68465
Constant, MDString and MDNode which can only be used by globals with a name
that starts with "llvm." or as arguments to a function with the same naming
restriction.
llvm-svn: 68420
- Particularly nice for small constant strings, which get optimized
down nicely. On a synthetic benchmark writing out "hello" in a
loop, this is about 2x faster with gcc and 3x faster with
llvm-gcc. llc on insn-attrtab.bc from 403.gcc is about .5% faster.
- I tried for a fancier solution which wouldn't increase code size as
much (by trying to match constant arrays), but can't quite make it
fly.
llvm-svn: 68396
"The code was doing "if (End+NumInputs > Capacity) ...". If End is
close to 0xFFFFFFFF and NumInputs is large, it'll overflow, the
condition will come out false, and the vector won't grow to
accommodate the new elements, and the program will crash in memmove."
Patch by Jeffrey Yasskin!
llvm-svn: 68277
- The code is silly, I'm just amusing myself. Rewrite to be efficient
if you like. :)
Also, if you wish to debate the proper names of the triple components
I'm all ears.
llvm-svn: 68252
which are effectively smart pointers to Value*'s. They are both very light
weight and simple, and react to values being destroyed or being RAUW'd.
WeakVN does a best effort to follow a value around, including through RAUW
operations and will get nulled out of the value is destroyed. This is useful
for the eventual "metadata that references a value" work, because it is a
reference to a value that does not show up on its use_* list.
AssertingVH is a pointer that compiles down to a dumb raw pointer when
assertions are disabled. When enabled, it emits an assertion if the
pointed-to value is destroyed while it is still being referenced. This
is very useful for Maps and other things, and should have caught the recent
bugs in CallGraph and Reassociate, for example.
llvm-svn: 68149
entered via fall-through. Don't miss fallthroughs from blocks
terminated by conditional branches. Also, move
isOnlyReachableByFallthrough out of line.
llvm-svn: 68129
llvm::sys::getOS{Name,Version}.
Right now the implementation just derives from LLVM_HOSTTRIPLE (which
is wrong, but it doesn't look like we have a define for the target
triple). Ideally this routine would actually be able to compute the
triple for targets we care about.
llvm-svn: 68118
only reachable via fall-through edges. This dramatically reduces the
number of labels printed, and thus also the number of labels the
assembler must parse and remember.
llvm-svn: 68073
you to do things like:
/// PointerUnion<int*, float*> P;
/// P = (int*)0;
/// printf("%d %d", P.is<int*>(), P.is<float*>()); // prints "1 0"
/// X = P.get<int*>(); // ok.
/// Y = P.get<float*>(); // runtime assertion failure.
/// Z = P.get<double*>(); // does not compile.
/// P = (float*)0;
/// Y = P.get<float*>(); // ok.
/// X = P.get<int*>(); // runtime assertion failure.
llvm-svn: 67987
function with a new NumLowBitsAvailable enum, which makes the
value available as an integer constant expression.
Add PointerLikeTypeTraits specializations for Instruction* and
Use** since they are only guaranteed 4-byte aligned.
Enhance PointerIntPair to know about (and enforce) the alignment
specified by PointerLikeTypeTraits. This should allow things
like PointerIntPair<PointerIntPair<void*, 1,bool>, 1, bool>
because the inner one knows that 2 low bits are free.
llvm-svn: 67979
x * 40
=>
shlq $3, %rdi
leaq (%rdi,%rdi,4), %rax
This has the added benefit of allowing more multiply to be folded into addressing mode. e.g.
a * 24 + b
=>
leaq (%rdi,%rdi,2), %rax
leaq (%rsi,%rax,8), %rax
llvm-svn: 67917
causing a bootstrap failure. Bootstraps here on
x86-32-linux and x86-64-linux. Requested by the
author Gabor Greif who says that a bug that might
have been causing the failure has since been fixed.
llvm-svn: 67844
- Make type declarations match the struct/class keyword of the definition.
- Move AddSignalHandler into the namespace where it belongs.
- Correctly call functions from template base.
- Some other small changes.
With this patch, LLVM and Clang should build properly and with far less noise under VS2008.
llvm-svn: 67347
the inliner; prevents nondeterministic behavior
when the same address is reallocated.
Don't build call graph nodes for debug intrinsic calls;
they're useless, and there were typically a lot of them.
llvm-svn: 67311
the set of blocks in which values are used, the set in which
values are live-through, and the set in which values are
killed. For the live-through and killed sets, conservative
approximations are used.
llvm-svn: 67309
a single character requires only one branch to follow slow path.
- Never use a buffer when writing on an unbuffered stream.
- Move default buffer size to header.
llvm-svn: 67066
write as arguments.
- Add raw_ostream::GetNumBytesInBuffer.
- Privatize buffer pointers.
- Get rid of slow and unnecessary code for writing out large strings.
llvm-svn: 67060
- Flush a known non-empty buffers; enforces the interface to
flush_impl and kills off HandleFlush (which I saw no reason to be
an inline method, Chris?).
- Clarify invariant that flush_impl is only called with OutBufCur >
OutBufStart.
- This also cleary collects all places where we have to deal with the
buffer possibly not existing.
- A few more comments and fixing the unbuffered behavior remain in
this commit sequence.
llvm-svn: 67057
by inserting explicit zero extensions where necessary. Included
is a testcase where SelectionDAG produces a virtual register
holding an i1 value which FastISel previously mistakenly assumed
to be zero-extended.
llvm-svn: 66941
changes.
For InvokeInst now all arguments begin at op_begin().
The Callee, Cont and Fail are now faster to get by
access relative to op_end().
This patch introduces some temporary uglyness in CallSite.
Next I'll bring CallInst up to a similar scheme and then
the uglyness will magically vanish.
This patch also exposes all the reliance of the libraries
on InvokeInst's operand ordering. I am thinking of taking
care of that too.
llvm-svn: 66920
1. ConstantPoolSDNode alignment field is log2 value of the alignment requirement. This is not consistent with other SDNode variants.
2. MachineConstantPool alignment field is also a log2 value.
3. However, some places are creating ConstantPoolSDNode with alignment value rather than log2 values. This creates entries with artificially large alignments, e.g. 256 for SSE vector values.
4. Constant pool entry offsets are computed when they are created. However, asm printer group them by sections. That means the offsets are no longer valid. However, asm printer uses them to determine size of padding between entries.
5. Asm printer uses expensive data structure multimap to track constant pool entries by sections.
6. Asm printer iterate over SmallPtrSet when it's emitting constant pool entries. This is non-deterministic.
Solutions:
1. ConstantPoolSDNode alignment field is changed to keep non-log2 value.
2. MachineConstantPool alignment field is also changed to keep non-log2 value.
3. Functions that create ConstantPool nodes are passing in non-log2 alignments.
4. MachineConstantPoolEntry no longer keeps an offset field. It's replaced with an alignment field. Offsets are not computed when constant pool entries are created. They are computed on the fly in asm printer and JIT.
5. Asm printer uses cheaper data structure to group constant pool entries.
6. Asm printer compute entry offsets after grouping is done.
7. Change JIT code to compute entry offsets on the fly.
llvm-svn: 66875
access each with a fixed negative index from op_end().
This has two important implications:
- getUser() will work faster, because there are less iterations
for the waymarking algorithm to perform. This is important
when running various analyses that want to determine callers
of basic blocks.
- getSuccessor() now runs faster, because the indirection via OperandList
is not necessary: Uses corresponding to the successors are at fixed
offset to "this".
The price we pay is the slightly more complicated logic in the operator
User::delete, as it has to pick up the information whether it has to free
the memory of an original unconditional BranchInst or a BranchInst that
was originally conditional, but has been shortened to unconditional.
I was not able to come up with a nicer solution to this problem. (And
rest assured, I tried *a lot*).
Similar reorderings will follow for InvokeInst and CallInst. After that
some optimizations to pred_iterator and CallSite will fall out naturally.
llvm-svn: 66815
1. Use the same value# to represent unknown values being merged into sub-registers.
2. When coalescer commute an instruction and the destination is a physical register, update its sub-registers by merging in the extended ranges.
llvm-svn: 66610
to obtain debug info about them.
Introduce helpers to access debug info for global variables. Also introduce a
helper that works for both local and global variables.
llvm-svn: 66541
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://developer.apple.com/bugreporter> for instructions.
/Volumes/Sandbox/Buildbot/llvm/full-llvm/build/llvmgcc42.roots/llvmgcc42~obj/src/gcc/libgcc2.c: In function '__muldi3':
/Volumes/Sandbox/Buildbot/llvm/full-llvm/build/llvmgcc42.roots/llvmgcc42~obj/src/gcc/libgcc2.c:567: internal compiler error: Bus error
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://developer.apple.com/bugreporter> for instructions.
/Volumes/Sandbox/Buildbot/llvm/full-llvm/build/llvmgcc42.roots/llvmgcc42~obj/src/gcc/libgcc2.c: In function '__lshrdi3':
/Volumes/Sandbox/Buildbot/llvm/full-llvm/build/llvmgcc42.roots/llvmgcc42~obj/src/gcc/libgcc2.c:421: internal compiler error: Bus error
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://developer.apple.com/bugreporter> for instructions.
make[5]: *** [libgcc/./_lshrdi3.o] Error 1
make[5]: *** Waiting for unfinished jobs....
make[5]: *** [libgcc/./_muldi3.o] Error 1
make[5]: *** [libgcc/./_negdi2.o] Error 1
--- Reverse-merging (from foreign repository) r66415 into '.':
U include/llvm/BasicBlock.h
U include/llvm/ADT/ilist_node.h
U include/llvm/CodeGen/SelectionDAG.h
U include/llvm/CodeGen/MachineFunction.h
U include/llvm/CodeGen/MachineBasicBlock.h
U include/llvm/Function.h
llvm-svn: 66426
from 66280. I was unable to verify this with gcc-3.4.6, but with gcc-3.3 it
avoids the "base class with only non-default constructor in class without
a constructor" warning. Apparently that warning was promoted to an error
in gcc-3.4.
llvm-svn: 66424
and extern_weak_odr. These are the same as the non-odr versions,
except that they indicate that the global will only be overridden
by an *equivalent* global. In C, a function with weak linkage can
be overridden by a function which behaves completely differently.
This means that IP passes have to skip weak functions, since any
deductions made from the function definition might be wrong, since
the definition could be replaced by something completely different
at link time. This is not allowed in C++, thanks to the ODR
(One-Definition-Rule): if a function is replaced by another at
link-time, then the new function must be the same as the original
function. If a language knows that a function or other global can
only be overridden by an equivalent global, it can give it the
weak_odr linkage type, and the optimizers will understand that it
is alright to make deductions based on the function body. The
code generators on the other hand map weak and weak_odr linkage
to the same thing.
llvm-svn: 66339
get nice and happy stack traces when we crash in an optimizer or codegen. For
example, an abort put in UnswitchLoops now looks like this:
Stack dump:
0. Program arguments: clang pr3399.c -S -O3
1. <eof> parser at end of file
2. per-module optimization passes
3. Running pass 'CallGraph Pass Manager' on module 'pr3399.c'.
4. Running pass 'Loop Pass Manager' on function '@foo'
5. Running pass 'Unswitch loops' on basic block '%for.inc'
Abort
llvm-svn: 66260
arbitrary functions to be run when a crash happens. Delete
RemoveDirectoryOnSignal as it is dead and has never had clients.
Change PrintStackTraceOnErrorSignal to be implemented in terms of
AddSignalHandler.
I updated the Win32 versions of these APIs, but can't test them.
If there are any problems, I'd be happy to fix them as well.
llvm-svn: 66072
because less bytes are allocated and subobject construction is gone.
For reference how it works, see BasicBlock.h.
Btw. it is very assuring to see that somebody has invented
this ilist-embedded sentinel technique before me :-)
llvm-svn: 66026
While the patch is clearly correct in itself, it's become
apparent other places are assuming debug intrinsics are
marked as touching memory...this needs more testing.
llvm-svn: 65992
arbitrary vector sizes. Add an optional MinSplatBits parameter to specify
a minimum for the splat element size. Update the PPC target to use the
revised interface.
llvm-svn: 65899
User drivers based on llvmc must all share the initialization code.
Putting main() into libCompilerDriver is not a very good idea IMO (and ld gave
me some strange EH-related error anyway).
llvm-svn: 65825
Move the code from 'llvmc/driver' into a new CompilerDriver library, and change
the build system accordingly. Makes it easier for projects using LLVM to build
their own llvmc-based drivers.
Tested with objdir != srcdir.
llvm-svn: 65821
its sentinel. This is quite a win when a function really has a basic block.
When the function is just a declaration (and stays so) the old way did not
allocate a sentinel. So this change is most beneficial when the ratio of
function definition to declaration is high. I.e. linkers etc. Incidentally
these are the most resource demanding applications, so I expect that the
reduced malloc traffic, locality and space savings outweigh the cost of
addition of two pointers to Function.
llvm-svn: 65776
This looks dangerous, but isn't because the sentinel is accessed in special way only,
namely the Next and Prev fields of it, and these are guaranteed to exist.
llvm-svn: 65626
to more accurately describe what it does. Expand its doxygen comment
to describe what the backedge-taken count is and how it differs
from the actual iteration count of the loop. Adjust names and
comments in associated code accordingly.
llvm-svn: 65382
them are generic changes.
- Use the "fast" flag that's already being passed into the asm printers instead
of shoving it into the DwarfWriter.
- Instead of calling "MI->getParent()->getParent()" for every MI, set the
machine function when calling "runOnMachineFunction" in the asm printers.
llvm-svn: 65379
a DBG_LABEL or not. We want to fall back to the original way of emitting debug
info when we're in -O0/-fast mode.
- Add plumbing in to pass the "Fast" flag to places that need it.
- XFAIL DebugInfo/deaddebuglabel.ll. This is finding 11 labels instead of 8. I
need to investigate still.
llvm-svn: 65367
instruction. The class also consolidates the code for detecting constant
splats that's shared across PowerPC and the CellSPU backends (and might be
useful for other backends.) Also introduces SelectionDAG::getBUID_VECTOR() for
generating new BUILD_VECTOR nodes.
llvm-svn: 65296
trip count value when the original loop iteration condition is
signed and the canonical induction variable won't undergo signed
overflow. This isn't required for correctness; it just preserves
more information about original loop iteration values.
Add a getTruncateOrSignExtend method to ScalarEvolution,
following getTruncateOrZeroExtend.
llvm-svn: 64918
that has not been JIT'd yet, the callee is put on a list of pending functions
to JIT. The call is directed through a stub, which is updated with the address
of the function after it has been JIT'd. A new interface for allocating and
updating empty stubs is provided.
Add support for removing the ModuleProvider the JIT was created with, which
would otherwise invalidate the JIT's PassManager, which is initialized with the
ModuleProvider's Module.
Add support under a new ExecutionEngine flag for emitting the infomration
necessary to update Function and GlobalVariable stubs after JITing them, by
recording the address of the stub and the name of the GlobalValue. This allows
code to be copied from one address space to another, where libraries may live
at different virtual addresses, and have the stubs updated with their new
correct target addresses.
llvm-svn: 64906
(Note: Eventually, commits like this will be handled via a pre-commit hook that
does this automagically, as well as expand tabs to spaces and look for 80-col
violations.)
llvm-svn: 64827
Bill Wendling