Move GetConstantStringInfo to lib/Analysis. Remove
string output routine from Constant. Update all
callers. Change debug intrinsic api slightly to
accomodate move of routine, these now return values
instead of strings.
This unbreaks llvm-gcc bootstrap.
llvm-svn: 52884
string output routine from Constant. Update all
callers. Change debug intrinsic api slightly to
accomodate move of routine, these now return values
instead of strings.
llvm-svn: 52748
in the presence of out-of-loop users of in-loop values and the trip
count is not a known multiple of the unroll count, and to be a bit
simpler overall. This fixes PR2253.
llvm-svn: 52645
structures. Its default threshold is to promote things that are
smaller than 128 bytes, which is sane. However, it is not sane
to do this for things that turn into 128 *registers*. Add a cap
on the number of registers introduced, defaulting to 128/4=32.
llvm-svn: 52611
DeadArgumentElimination and assert that the function type does not change if
nothing was changed. This should catch subtle changes in function type that are
not intended.
llvm-svn: 52536
This is a fixed version that no longer uses multimap::equal_range, which
resulted in a pointer invalidation problem.
Also, DAE::InspectedFunctions was not really necessary, so it got removed.
Lastly, this version no longer applies the extra arg hack on functions who did
not have any arguments to start with.
llvm-svn: 52532
dependencies between return values and/or arguments. Also make the handling of
arguments and return values the same.
The pass now looks properly inside returned structs, but only at the first
level (ie, not inside nested structs).
Also add a testcase for testing various variations of (multiple) dead rerturn
values.
llvm-svn: 52459
speaking these are not constant values. However, when a function always returns
one of its arguments, then from the point of view of each caller the return
value is constant (or at least a known value) and can be replaced.
llvm-svn: 52397
individually.
Also learn IPConstProp how returning first class aggregates work, in addition
to old style multiple return instructions.
Modify the return-constants testscase to confirm this behaviour.
llvm-svn: 52396
when changing the stride of a comparison so that it's slightly
more precise, by having it scan the instruction list to determine
if there is a use of the condition after the point where the
condition will be inserted.
llvm-svn: 52371
I'm at it, rename it to FindInsertedValue.
The only functional change is that newly created instructions are no longer
added to instcombine's worklist, but that is not really necessary anyway (and
I'll commit some improvements next that will completely remove the need).
llvm-svn: 52315
of apint codegen failure is the DAG combiner doing
the wrong thing because it was comparing MVT's using
< rather than comparing the number of bits. Removing
the < method makes this mistake impossible to commit.
Instead, add helper methods for comparing bits and use
them.
llvm-svn: 52098
and better control the abstraction. Rename the type
to MVT. To update out-of-tree patches, the main
thing to do is to rename MVT::ValueType to MVT, and
rewrite expressions like MVT::getSizeInBits(VT) in
the form VT.getSizeInBits(). Use VT.getSimpleVT()
to extract a MVT::SimpleValueType for use in switch
statements (you will get an assert failure if VT is
an extended value type - these shouldn't exist after
type legalization).
This results in a small speedup of codegen and no
new testsuite failures (x86-64 linux).
llvm-svn: 52044
work and how to replace them into individual values. Also, when trying to
replace an aggregrate that is used by load or store with a single (large)
integer, don't crash (but don't replace the aggregrate either).
Also adds a testcase for both structs and arrays.
llvm-svn: 51997
are the same as in unpacked structs, only field
positions differ. This only matters for structs
containing x86 long double or an apint; it may
cause backwards compatibility problems if someone
has bitcode containing a packed struct with a
field of one of those types.
The issue is that only 10 bytes are needed to
hold an x86 long double: the store size is 10
bytes, but the ABI size is 12 or 16 bytes (linux/
darwin) which comes from rounding the store size
up by the alignment. Because it seemed silly not
to pack an x86 long double into 10 bytes in a
packed struct, this is what was done. I now
think this was a mistake. Reserving the ABI size
for an x86 long double field even in a packed
struct makes things more uniform: the ABI size is
now always used when reserving space for a type.
This means that developers are less likely to
make mistakes. It also makes life easier for the
CBE which otherwise could not represent all LLVM
packed structs (PR2402).
Front-end people might need to adjust the way
they create LLVM structs - see following change
to llvm-gcc.
llvm-svn: 51928
out of instcombine into a new file in libanalysis. This also teaches
ComputeNumSignBits about the number of sign bits in a constantint.
llvm-svn: 51863
the conditions for performing the transform when only the
function declaration is available: no longer allow turning
i32 into i64 for example. Only allow changing between
pointer types, and between pointer types and integers of
the same size. For return values ptr -> intptr was already
allowed; I added ptr -> ptr and intptr -> ptr while there.
As shown by a recent objc testcase, changing the way
parameters/return values are passed can be fatal when calling
code written in assembler that directly manipulates call
arguments and return values unless the transform has no
impact on the way they are passed at the codegen level.
While it is possible to imagine an ABI that treats integers
of pointer size differently to pointers, I don't think LLVM
supports any so the transform should now be safe while still
being useful.
llvm-svn: 51834
the one case that ADCE catches that normal DCE doesn't: non-induction variable
loop computations.
This implementation handles this problem without using postdominators.
llvm-svn: 51668
the section or the visibility from one global
value to another: copyAttributesFrom. This is
particularly useful for duplicating functions:
previously this was done by explicitly copying
each attribute in turn at each place where a
new function was created out of an old one, with
the result that obscure attributes were regularly
forgotten (like the collector or the section).
Hopefully now everything is uniform and nothing
is forgotten.
llvm-svn: 51567
Analysis/ConstantFolding to fold ConstantExpr's, then make instcombine use it
to try to use targetdata to fold constant expressions on void instructions.
Also extend the icmp(inttoptr, inttoptr) folding to handle the case where
int size != ptr size.
llvm-svn: 51559
The SimplifyCFG pass looks at basic blocks that contain only phi nodes,
followed by an unconditional branch. In a lot of cases, such a block (BB) can
be merged into their successor (Succ).
This merging is performed by TryToSimplifyUncondBranchFromEmptyBlock. It does
this by taking all phi nodes in the succesor block Succ and expanding them to
include the predecessors of BB. Furthermore, any phi nodes in BB are moved to
Succ and expanded to include the predecessors of Succ as well.
Before attempting this merge, CanPropagatePredecessorsForPHIs checks to see if
all phi nodes can be properly merged. All functional changes are made to
this function, only comments were updated in
TryToSimplifyUncondBranchFromEmptyBlock.
In the original code, CanPropagatePredecessorsForPHIs looks quite convoluted
and more like stack of checks added to handle different kinds of situations
than a comprehensive check. In particular the first check in the function did
some value checking for the case that BB and Succ have a common predecessor,
while the last check in the function simply rejected all cases where BB and
Succ have a common predecessor. The first check was still useful in the case
that BB did not contain any phi nodes at all, though, so it was not completely
useless.
Now, CanPropagatePredecessorsForPHIs is restructured to to look a lot more
similar to the code that actually performs the merge. Both functions now look
at the same phi nodes in about the same order. Any conflicts (phi nodes with
different values for the same source) that could arise from merging or moving
phi nodes are detected. If no conflicts are found, the merge can happen.
Apart from only restructuring the checks, two main changes in functionality
happened.
Firstly, the old code rejected blocks with common predecessors in most cases.
The new code performs some extra checks so common predecessors can be handled
in a lot of cases. Wherever common predecessors still pose problems, the
blocks are left untouched.
Secondly, the old code rejected the merge when values (phi nodes) from BB were
used in any other place than Succ. However, it does not seem that there is any
situation that would require this check. Even more, this can be proven.
Consider that BB is a block containing of a single phi node "%a" and a branch
to Succ. Now, since the definition of %a will dominate all of its uses, BB
will dominate all blocks that use %a. Furthermore, since the branch from BB to
Succ is unconditional, Succ will also dominate all uses of %a.
Now, assume that one predecessor of Succ is not dominated by BB (and thus not
dominated by Succ). Since at least one use of %a (but in reality all of them)
is reachable from Succ, you could end up at a use of %a without passing
through it's definition in BB (by coming from X through Succ). This is a
contradiction, meaning that our original assumption is wrong. Thus, all
predecessors of Succ must also be dominated by BB (and thus also by Succ).
This means that moving the phi node %a from BB to Succ does not pose any
problems when the two blocks are merged, and any use checks are not needed.
llvm-svn: 51478
ScalarEvolution::deleteValueFromRecords on it before doing the
replaceAllUsesWith, because ScalarEvolution looks at the instruction's
users to find SCEV references to the instruction's SCEV object in its
internal maps.
Move all of LSR's loop-related state clearing after processing the loop
and before cleaning up dead PHI nodes. This eliminates all of LSR's SCEV
references just before the calls to ScalarEvolution::deleteValueFromRecords
so that when ScalarEvolution drops its own SCEV references, the reference
counts will reach zero and the SCEVs will be deleted immediately.
These changes fix some compiler aborts involving ScalarEvolution holding
onto and reusing SCEV objects for instructions that have been deleted.
No regression test unfortunately; because the symptoms were due to
dangling pointers, reduced testcases ended up being fairly arbitrary.
llvm-svn: 51359
replaced is a PHI. This prevents it from inserting uses before defs
in the case that it isn't a PHI and it depends on other instructions
later in the block. This fixes the 447.dealII regression on x86-64.
llvm-svn: 51292
type and the other operand is a constant into integer comparisons.
This happens surprisingly frequently (e.g. 10 times in 471.omnetpp),
which are things like this:
%tmp8283 = sitofp i32 %tmp82 to double
%tmp1013 = fcmp ult double %tmp8283, 0.0
Clearly comparing tmp82 against i32 0 is cheaper here.
this also triggers 8 times in gobmk, including this one:
%tmp375376 = sitofp i32 %tmp375 to double
%tmp377 = fcmp ogt double %tmp375376, 8.150000e+01
which is comparing an integer against 81.5 :).
llvm-svn: 51268
intersecting bits. This triggers all over the place, for example in lencode,
with adds of stuff like:
%tmp580 = mul i32 %tmp579, 2
%tmp582 = and i32 %b8, 1
and
%tmp28 = shl i32 %abs.i, 1
%sign.0 = select i1 %tmp23, i32 1, i32 0
and
%tmp344 = shl i32 %tmp343, 2
%tmp346 = and i32 %tmp96, 3
etc.
llvm-svn: 51263
replaced at linktime with a body that throws, even
if the body in this file does not. Make PruneEH
be more conservative in this case.
g++.dg/eh/weak1.C
llvm-svn: 51207
use-before-def. The problem comes up in code with multiple PHIs where
one PHI is being rewritten in terms of the other, but the other needs
to be casted first. LLVM rules requre the cast instruction to be
inserted after any PHI instructions, but when instructions were
inserted to replace the second PHI value with a function of the first,
they were ended up going before the cast instruction. Avoid this
problem by remembering the location of the cast instruction, when one
is needed, and inserting the expansion of the new value after it.
This fixes a bug that surfaced in 255.vortex on x86-64 when
instcombine was removed from the middle of the loop optimization
passes.
llvm-svn: 51169
is bitcast to return a floating point value. The result of the instruction may
not be used by the program afterwards, and LLVM will happily remove all
instructions except the call. But, on some platforms, if a value is returned as
a floating point, it may need to be removed from the stack (like x87). Thus, we
can't get rid of the bitcast even if there isn't a use of the value.
llvm-svn: 51134
bug as well as a missed optimization. We weren't properly checking for local
dependencies before moving on to non-local ones when doing non-local read-only
call CSE.
llvm-svn: 51082
address of the PassInfo directly instead of calling getPassInfo.
This eliminates a bunch of dynamic initializations of static data.
Also, fold RegisterPassBase into PassInfo, make a bunch of its
data members const, and rearrange some code to initialize data
members in constructors instead of using setter member functions.
llvm-svn: 51022
method. DOUT statements are disabled when assertions are off, but the
side effects of getName() are still evaluated. Just call getNameSTart,
which is close enough and doesn't cause heap traffic.
llvm-svn: 50958
a FunctionPass. This makes it simpler, fixes dozens of bugs, adds
a couple of minor features, and shrinks is considerably: from
2214 to 1437 lines.
llvm-svn: 50520
we were checking for it in the wrong order. This caused a miscompilation because the
return slot optimization assumes that the call it is dealing with is NOT a memcpy.
llvm-svn: 50444
generalizes the previous code to handle the case when the string is not
an immediate to the strlen call (for example, crazy stuff like
strlen(c ? "foo" : "bart"+1) -> 3). This implements
gcc.c-torture/execute/builtins/strlen-2.c. I will generalize other
cases in simplifylibcalls to use the same routine later.
llvm-svn: 50408
ComputeMaskedBits knows about cttz, ctlz, and ctpop. Teach
SelectionDAG's ComputeMaskedBits what InstCombine's knows
about SRem. And teach them both some things about high bits
in Mul, UDiv, URem, and Sub. This allows instcombine and
dagcombine to eliminate sign-extension operations in
several new cases.
llvm-svn: 50358
When choosing between constraints with multiple options,
like "ir", test to see if we can use the 'i' constraint and
go with that if possible. This produces more optimal ASM in
all cases (sparing a register and an instruction to load it),
and fixes inline asm like this:
void test () {
asm volatile (" %c0 %1 " : : "imr" (42), "imr"(14));
}
Previously we would dump "42" into a memory location (which
is ok for the 'm' constraint) which would cause a problem
because the 'c' modifier is not valid on memory operands.
Isn't it great how inline asm turns 'missed optimization'
into 'compile failed'??
Incidentally, this was the todo in
PowerPC/2007-04-24-InlineAsm-I-Modifier.ll
Please do NOT pull this into Tak.
llvm-svn: 50315
to the block that defines their operands. This doesn't work in the
case that the operand is an invoke, because invoke is a terminator
and must be the last instruction in a block.
Replace it with support in SelectionDAGISel for copying struct values
into sequences of virtual registers.
llvm-svn: 50279
getelementptr-seteq.ll into:
define i1 @test(i64 %X, %S* %P) {
%C = icmp eq i64 %X, -1 ; <i1> [#uses=1]
ret i1 %C
}
instead of:
define i1 @test(i64 %X, %S* %P) {
%A.idx.mask = and i64 %X, 4611686018427387903 ; <i64> [#uses=1]
%C = icmp eq i64 %A.idx.mask, 4611686018427387903 ; <i1> [#uses=1]
ret i1 %C
}
And fixes the second half of PR2235. This speeds up the insertion sort
case by 45%, from 1.12s to 0.77s. In practice, this will significantly
speed up for loops structured like:
for (double *P = Base + N; P != Base; --P)
...
Which happens frequently for C++ iterators.
llvm-svn: 50079
as a global helper function. At the same type, switch it from taking
a vector of predecessors to an arbitrary sequential input. This allows
us to switch LoopSimplify to use a SmallVector for various temporary
vectors that it passed into SplitBlockPredecessors.
llvm-svn: 50020
in addition to integer expressions. Rewrite GetOrEnforceKnownAlignment
as a ComputeMaskedBits problem, moving all of its special alignment
knowledge to ComputeMaskedBits as low-zero-bits knowledge.
Also, teach ComputeMaskedBits a few basic things about Mul and PHI
instructions.
This improves ComputeMaskedBits-based simplifications in a few cases,
but more noticeably it significantly improves instcombine's alignment
detection for loads, stores, and memory intrinsics.
llvm-svn: 49492
needs to be fixed here - a previous commit made sure
that intrinsics always get the right attributes.
So remove no-longer needed code, and while there use
Intrinsic::getDeclaration rather than getOrInsertFunction.
llvm-svn: 49337
2. Do not use # of basic blocks as part of the cost computation since it doesn't really figure into function size.
3. More aggressively inline function with vector code.
llvm-svn: 49061
not marked nounwind, or for all functions when -enable-eh
is set, provided the target supports Dwarf EH.
llvm-gcc generates nounwind in the right places; other FEs
will need to do so also. Given such a FE, -enable-eh should
no longer be needed.
llvm-svn: 49006
when something changes, instead of moving forward. This allows us to
simplify memset lowering, inserting the memset at the end of the range of
stuff we're touching instead of at the start.
This, in turn, allows us to make use of the addressing instructions already
used in the function instead of inserting our own. For example, we now
codegen:
%tmp41 = getelementptr [8 x i8]* %ref_idx, i32 0, i32 0 ; <i8*> [#uses=2]
call void @llvm.memset.i64( i8* %tmp41, i8 -1, i64 8, i32 1 )
instead of:
%tmp20 = getelementptr [8 x i8]* %ref_idx, i32 0, i32 7 ; <i8*> [#uses=1]
%ptroffset = getelementptr i8* %tmp20, i64 -7 ; <i8*> [#uses=1]
call void @llvm.memset.i64( i8* %ptroffset, i8 -1, i64 8, i32 1 )
llvm-svn: 48940
memsets that initialize "structs of arrays" and other store sequences
that are not sequential. This is still only enabled if you pass
-form-memset-from-stores. The flag is not heavily tested and I haven't
analyzed the perf regressions when -form-memset-from-stores is passed
either, but this causes no make check regressions.
llvm-svn: 48909
Furthermore, double the limit when more than 10% of the callee instructions are vector instructions. Multimedia kernels tend to love inlining.
llvm-svn: 48725
This fires dozens of times across spec and multisource, but I don't know
if it actually speeds stuff up. Hopefully the testers will show something
nice :)
llvm-svn: 48680
