available, the only thing this affects is that we produce
.set in one case we didn't before, which shouldn't harm
anything. Make EmitSectionOffset call EmitDifference
instead of duplicating it.
llvm-svn: 98005
immediate instructions cannot set the condition codes, so they do not have
the extra cc_out operand. We hit an assertion during tail duplication
because the instruction being duplicated had more operands that expected.
llvm-svn: 98001
is a workaround for <rdar://problem/7672401/> (which I filed).
This let's us build Wine on Darwin, and it gets the Qt build there a little bit
further (so Doug says).
llvm-svn: 97845
CALL ... %RAX<imp-def>
... [not using %RAX]
%EAX = ..., %RAX<imp-use, kill>
RET %EAX<imp-use,kill>
Now we do this:
CALL ... %RAX<imp-def, dead>
... [not using %RAX]
%EAX = ...
RET %EAX<imp-use,kill>
By not artificially keeping %RAX alive, we lower register pressure a bit.
The correct number of instructions for 2008-08-05-SpillerBug.ll is obviously
55, anybody can see that. Sheesh.
llvm-svn: 97838
The MicroBlaze backend was generating stack layouts that did not
conform correctly to the ABI. This update generates stack layouts
which are closer to what GCC does.
Variable arguments support was added as well but the stack layout
for varargs has not been finalized.
llvm-svn: 97807
node which has a flag. That flag in turn was used by an
already-selected adde which turned into an ADC32ri8 which
used a selected load which was chained to the load we
folded. This flag use caused us to form a cycle. Fix
this by not ignoring chains in IsLegalToFold even in
cases where the isel thinks it can.
llvm-svn: 97791
This code:
float floatingPointComparison(float x, float y) {
double product = (double)x * y;
if (product == 0.0)
return product;
return product - 1.0;
}
produces this:
_floatingPointComparison:
0000000000000000 cvtss2sd %xmm1,%xmm1
0000000000000004 cvtss2sd %xmm0,%xmm0
0000000000000008 mulsd %xmm1,%xmm0
000000000000000c pxor %xmm1,%xmm1
0000000000000010 ucomisd %xmm1,%xmm0
0000000000000014 jne 0x00000004
0000000000000016 jp 0x00000002
0000000000000018 jmp 0x00000008
000000000000001a addsd 0x00000006(%rip),%xmm0
0000000000000022 cvtsd2ss %xmm0,%xmm0
0000000000000026 ret
The "jne/jp/jmp" sequence can be reduced to this instead:
_floatingPointComparison:
0000000000000000 cvtss2sd %xmm1,%xmm1
0000000000000004 cvtss2sd %xmm0,%xmm0
0000000000000008 mulsd %xmm1,%xmm0
000000000000000c pxor %xmm1,%xmm1
0000000000000010 ucomisd %xmm1,%xmm0
0000000000000014 jp 0x00000002
0000000000000016 je 0x00000008
0000000000000018 addsd 0x00000006(%rip),%xmm0
0000000000000020 cvtsd2ss %xmm0,%xmm0
0000000000000024 ret
for a savings of 2 bytes.
This xform can happen when we recognize that jne and jp jump to the same "true"
MBB, the unconditional jump would jump to the "false" MBB, and the "true" branch
is the fall-through MBB.
llvm-svn: 97766
an undef value. This is only going to come up for bugpoint-reduced tests --
correct programs will not access memory at undefined addresses -- so it's not
worth the effort of doing anything more aggressive.
llvm-svn: 97745
These instructions technically define AL,AH, but a trick in X86ISelDAGToDAG
reads AX in order to avoid reading AH with a REX instruction.
Fix PR6489.
llvm-svn: 97742
long test(long x) { return (x & 123124) | 3; }
Currently compiles to:
_test:
orl $3, %edi
movq %rdi, %rax
andq $123127, %rax
ret
This is because instruction and DAG combiners canonicalize
(or (and x, C), D) -> (and (or, D), (C | D))
However, this is only profitable if (C & D) != 0. It gets in the way of the
3-addressification because the input bits are known to be zero.
llvm-svn: 97616
CopyToReg/CopyFromReg/INLINEASM. These are annoying because
they have the same opcode before an after isel. Fix this by
setting their NodeID to -1 to indicate that they are selected,
just like what automatically happens when selecting things that
end up being machine nodes.
With that done, give IsLegalToFold a new flag that causes it to
ignore chains. This lets the HandleMergeInputChains routine be
the one place that validates chains after a match is successful,
enabling the new hotness in chain processing. This smarter
chain processing eliminates the need for "PreprocessRMW" in the
X86 and MSP430 backends and enables MSP to start matching it's
multiple mem operand instructions more aggressively.
I currently #if out the dead code in the X86 backend and MSP
backend, I'll remove it for real in a follow-on patch.
The testcase changes are:
test/CodeGen/X86/sse3.ll: we generate better code
test/CodeGen/X86/store_op_load_fold2.ll: PreprocessRMW was
miscompiling this before, we now generate correct code
Convert it to filecheck while I'm at it.
test/CodeGen/MSP430/Inst16mm.ll: Add a testcase for mem/mem
folding to make anton happy. :)
llvm-svn: 97596
was that we weren't properly handling the case when interior
nodes of a matched pattern become dead after updating chain
and flag uses. Now we handle this explicitly in
UpdateChainsAndFlags.
llvm-svn: 97561
stuff now that we don't care about emulating the old broken
behavior of the old isel. This eliminates the
'CheckChainCompatible' check (along with IsChainCompatible) which
did an incorrect and inefficient scan *up* the chain nodes which
happened as the pattern was being formed and does the validation
at the end in HandleMergeInputChains when it forms a structural
pattern. This scans "down" the graph, which means that it is
quickly bounded by nodes already selected. This also handles
token factors that get "trapped" in the dag.
Removing the CheckChainCompatible nodes also shrinks the
generated tables by about 6K for X86 (down to 83K).
There are two pieces remaining before I can nuke PreprocessRMW:
1. I xfailed a test because we're now producing worse code in a
case that has nothing to do with the change: it turns out that
our use of MorphNodeTo will leave dead nodes in the graph
which (depending on how the graph is walked) end up causing
bogus uses of chains and blocking matches. This is really
bad for other reasons, so I'll fix this in a follow-up patch.
2. CheckFoldableChainNode needs to be improved to handle the TF.
llvm-svn: 97539
ordered correctly. Previously it would get in trouble when
two patterns were too similar and give them nondet ordering.
We force this by using the record ID order as a fallback.
The testsuite diff is due to alpha patterns being ordered
slightly differently, the change is a semantic noop afaict:
< lda $0,-100($16)
---
> subq $16,100,$0
llvm-svn: 97509
The PowerPC floating point registers can represent both f32 and f64 via the
two register classes F4RC and F8RC. F8RC is considered a subclass of F4RC to
allow cross-class coalescing. This coalescing only affects whether registers
are spilled as f32 or f64.
Spill slots must be accessed with load/store instructions corresponding to the
class of the spilled register. PPCInstrInfo::foldMemoryOperandImpl was looking
at the instruction opcode which is wrong.
X86 has similar floating point register classes, but doesn't try to fold
memory operands, so there is no problem there.
llvm-svn: 97262
instead of to have a chained series of scope nodes. This makes
the generated table smaller, improves the efficiency of the
interpreter, and make the factoring optimization much more
reasonable to implement.
llvm-svn: 97160
terms of store and load, which means bitcasting between scalar
integer and vector has endian-specific results, which undermines
this whole approach.
llvm-svn: 97137
which branch on undef to branch on a boolean constant for the edge
exiting the loop. This helps ScalarEvolution compute trip counts for
loops.
Teach ScalarEvolution to recognize single-value PHIs, when safe, and
ForgetSymbolicName to forget such single-value PHI nodes as apprpriate
in ForgetSymbolicName.
llvm-svn: 97126
- Function uses all scratch registers AND
- Function does not use any callee saved registers AND
- Stack size is too big to address with immediate offsets.
In this case a register must be scavenged to calculate the address of a stack
object, and the scavenger needs a spare register or emergency spill slot.
llvm-svn: 97071
greater-than-or-equal SELECT_CCs to NEON vmin/vmax instructions. This is
only allowed when UnsafeFPMath is set or when at least one of the operands
is known to be nonzero.
llvm-svn: 97065
the number of value bits, not the number of bits of allocation for in-memory
storage.
Make getTypeStoreSize and getTypeAllocSize work consistently for arrays and
vectors.
Fix several places in CodeGen which compute offsets into in-memory vectors
to use TargetData information.
This fixes PR1784.
llvm-svn: 97064
necessary to swap the operands to handle NaN and negative zero properly.
Also, reintroduce logic for checking for NaN conditions when forming
SSE min and max instructions, fixed to take into consideration NaNs and
negative zeros. This allows forming min and max instructions in more
cases.
llvm-svn: 97025
to adding them in a determinstic order (bottom up from
the root) based on the structure of the graph itself.
This updates tests for some random changes, interesting
bits: CodeGen/Blackfin/promote-logic.ll no longer crashes.
I have no idea why, but that's good right?
CodeGen/X86/2009-07-16-LoadFoldingBug.ll also fails, but
now compiles to have one fewer constant pool entry, making
the expected load that was being folded disappear. Since it
is an unreduced mass of gnast, I just removed it.
This fixes PR6370
llvm-svn: 97023
Previously, LiveIntervalAnalysis would infer phi joins by looking for multiply
defined registers. That doesn't work if the phi join is implicitly defined in
all but one of the predecessors.
llvm-svn: 96994
The MicroBlaze is a highly configurable 32-bit soft-microprocessor for
use on Xilinx FPGAs. For more information see:
http://www.xilinx.com/tools/microblaze.htmhttp://en.wikipedia.org/wiki/MicroBlaze
The current LLVM MicroBlaze backend generates assembly which can be
compiled using the an appropriate binutils assembler.
llvm-svn: 96969
operators.
The test difference is just due to the multiplication operands
being commuted (and thus requiring a more elaborate match). In
optimized code, that expression would be folded.
llvm-svn: 96816
during a tail call. A parameter might overwrite this stack slot during the tail
call.
The sequence during a tail call is:
1.) load return address to temp reg
2.) move parameters (might involve storing to return address stack slot)
3.) store return address to new location from temp reg
If the stack location is marked immutable CodeGen can colocate load (1) with the
store (3).
This fixes bug 6225.
llvm-svn: 96783
SSE min and max instructions. The real thing this code needs to be
concerned about is negative zero.
Update the sse-minmax.ll test accordingly, and add tests for
-enable-unsafe-fp-math mode as well.
llvm-svn: 96775
induction variable value and a loop-variant value, don't force the
insert position to be at the post-increment position, because it may
not be dominated by the loop-variant value. This fixes a
use-before-def problem noticed on PPC.
llvm-svn: 96774
dragonegg self-host build. I reverted 96640 in order to revert
96556 (96640 goes on top of 96556), but it also looks like with
both of them applied the breakage happens even earlier. The
symptom of the 96556 miscompile is the following crash:
llvm[3]: Compiling AlphaISelLowering.cpp for Release build
cc1plus: /home/duncan/tmp/tmp/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp:4982: void llvm::SelectionDAG::ReplaceAllUsesWith(llvm::SDNode*, llvm::SDNode*, llvm::SelectionDAG::DAGUpdateListener*): Assertion `(!From->hasAnyUseOfValue(i) || From->getValueType(i) == To->getValueType(i)) && "Cannot use this version of ReplaceAllUsesWith!"' failed.
Stack dump:
0. Running pass 'X86 DAG->DAG Instruction Selection' on function '@_ZN4llvm19AlphaTargetLowering14LowerOperationENS_7SDValueERNS_12SelectionDAGE'
g++: Internal error: Aborted (program cc1plus)
This occurs when building LLVM using LLVM built by LLVM (via
dragonegg). Probably LLVM has miscompiled itself, though it
may have miscompiled GCC and/or dragonegg itself: at this point
of the self-host build, all of GCC, LLVM and dragonegg were built
using LLVM. Unfortunately this kind of thing is extremely hard
to debug, and while I did rummage around a bit I didn't find any
smoking guns, aka obviously miscompiled code.
Found by bisection.
r96556 | evancheng | 2010-02-18 03:13:50 +0100 (Thu, 18 Feb 2010) | 5 lines
Some dag combiner goodness:
Transform br (xor (x, y)) -> br (x != y)
Transform br (xor (xor (x,y), 1)) -> br (x == y)
Also normalize (and (X, 1) == / != 1 -> (and (X, 1)) != / == 0 to match to "test on x86" and "tst on arm"
r96640 | evancheng | 2010-02-19 01:34:39 +0100 (Fri, 19 Feb 2010) | 16 lines
Transform (xor (setcc), (setcc)) == / != 1 to
(xor (setcc), (setcc)) != / == 1.
e.g. On x86_64
%0 = icmp eq i32 %x, 0
%1 = icmp eq i32 %y, 0
%2 = xor i1 %1, %0
br i1 %2, label %bb, label %return
=>
testl %edi, %edi
sete %al
testl %esi, %esi
sete %cl
cmpb %al, %cl
je LBB1_2
llvm-svn: 96672
strides in foreign loops. This helps locate reuse opportunities
with existing induction variables in foreign loops and reduces
the need for inserting new ones. This fixes rdar://7657764.
llvm-svn: 96629
Moderate the weight given to very small intervals.
The spill weight given to new intervals created when spilling was not
normalized in the same way as the original spill weights calculated by
CalcSpillWeights. That meant that restored registers would tend to hang around
because they had a much higher spill weight that unspilled registers.
This improves the runtime of a few tests by up to 10%, and there are no
significant regressions.
llvm-svn: 96613
checking whether AnalyzeBranch disagrees with the CFG
directly, rather than looking for EH_LABEL instructions.
EH_LABEL instructions aren't always at the end of the
block, due to FP_REG_KILL and other things. This fixes
an infinite loop compiling MultiSource/Benchmarks/Bullet.
llvm-svn: 96611
into a roundss intrinsic, producing a cyclic dag. The root cause
of this is badness handling ComplexPattern nodes in the old dagisel
that I noticed through inspection. Eliminate a copy of the of the
code that handled ComplexPatterns by making EmitChildMatchCode call
into EmitMatchCode.
llvm-svn: 96408
If there exists a use of a build_vector that's the bitwise complement of the mask,
then transform the node to
(and (xor x, (build_vector -1,-1,-1,-1)), (build_vector ~c1,~c2,~c3,~c4)).
Since this transformation is only useful when 1) the given build_vector will
become a load from constpool, and 2) (and (xor x -1), y) matches to a single
instruction, I decided this is appropriate as a x86 specific transformation.
rdar://7323335
llvm-svn: 96389
non-temporal. Fix from r96241 for botched encoding of MOVNTDQ.
Add documentation for !nontemporal metadata.
Add a simpler movnt testcase.
llvm-svn: 96386
branch in ARM v4 code, since it gets clobbered by the return address before
it is used. Instead of adding a new register class containing all the GPRs
except LR, just use the existing tGPR class.
llvm-svn: 96360
A virtual register can be used before it is defined in the same MBB if the MBB
is part of a loop. Teach the implicit-def pass about this case.
llvm-svn: 96279
When coalescing with a physreg, remember to add imp-def and imp-kill when
dealing with sub-registers.
Also fix a related bug in VirtRegRewriter where substitutePhysReg may
reallocate the operand list on an instruction and invalidate the reg_iterator.
This can happen when a register is mentioned twice on the same instruction.
llvm-svn: 96072
phi cycles. Adjust a few tests to keep dead instructions from being optimized
away. This (together with my previous change for phi cycles) fixes Apple
radar 7627077.
llvm-svn: 96057
stack frame, the prolog/epilog code was using the same
register for the copy of CR and the address of the save slot. Oops.
This is fixed here for Darwin, sort of, by reserving R2 for this case.
A better way would be to do the store before the decrement of SP,
which is safe on Darwin due to the red zone.
SVR4 probably has the same problem, but I don't know how to fix it;
there is no red zone and R2 is already used for something else.
I'm going to leave it to someone interested in that target.
Better still would be to rewrite the CR-saving code completely;
spilling each CR subregister individually is horrible code.
llvm-svn: 96015
bug fixes, and with improved heuristics for analyzing foreign-loop
addrecs.
This change also flattens IVUsers, eliminating the stride-oriented
groupings, which makes it easier to work with.
llvm-svn: 95975
reduce down to a single value. InstCombine already does this transformation
but DAG legalization may introduce new opportunities. This has turned out to
be important for ARM where 64-bit values are split up during type legalization:
InstCombine is not able to remove the PHI cycles on the 64-bit values but
the separate 32-bit values can be optimized. I measured the compile time
impact of this (running llc on 176.gcc) and it was not significant.
llvm-svn: 95951
lowering and requires that certain types exist in ValueTypes.h. Modified widening to
check if an op can trap and if so, the widening algorithm will apply only the op on
the defined elements. It is safer to do this in widening because the optimizer can't
guarantee removing unused ops in some cases.
llvm-svn: 95823
legalization even when the IR-level optimizer has removed dead phis, such
as when the high half of an i64 value is unused on a 32-bit target.
I had to adjust a few test cases that had dead phis.
This is a partial fix for Radar 7627077.
llvm-svn: 95816
in global initializers. Instead of aborting, attempt to fold them on the
spot. If folding succeeds, emit the folded expression instead.
This fixes PR6255.
llvm-svn: 95583
only run for x86 with fastisel. I've found it being very effective in
eliminating some obvious dead code as result of formal parameter lowering
especially when tail call optimization eliminated the need for some of the loads
from fixed frame objects. It also shrinks a number of the tests. A couple of
tests no longer make sense and are now eliminated.
llvm-svn: 95493
following it. However, the EmitGlobalConstant method wasn't emitting a body for
the constant. The assembler doesn't like that. Before, we were generating this:
.zerofill __DATA, __common, __cmd, 1, 3
This fix puts us back to that semantic.
llvm-svn: 95336
where callee's arguments are already in the caller's own caller's stack and
they line up perfectly. e.g.
extern int foo(int a, int b, int c);
int bar(int a, int b, int c) {
return foo(a, b, c);
}
llvm-svn: 95053
