depth first order, so it wouldn't process unreachable blocks.
When compiling at -O0, late dead block elimination isn't done
and the bad instructions got to isel.
llvm-svn: 81187
from floating-point to integer first, and bitcast the result
back to floating-point. Previously, this test was passing by
falling back to SelectionDAG lowering. The resulting code isn't
as nice, but it's correct and CodeGen now stays on the fast path.
llvm-svn: 81171
linear scan reg alloc. This fixes a problem I ran into where extracting
a function from a larger file caused the generated code to change (masking
the problem I was trying to debug) because the allocator behaved differently.
This changes the results for two X86 regression checks. stack-color-with-reg
is improved, with one less instruction, but pr3495 is worse, with one more
copy. As far as I can tell, these tests were just getting lucky or unlucky,
so I've changed the expected results.
llvm-svn: 81060
tied to different source registers, the TwoAddressInstructionPass needs to
be smarter. Change it to check before replacing a source register whether
that source register is tied to a different destination register, and if so,
defer handling it until a subsequent iteration.
llvm-svn: 80654
makes an eggregious hack somewhat more palatable. Bringing the LSDA forward
and making it a GV available for reference would be even better, but is
beyond the scope of what I'm looking to solve at this point.
Objective C++ code could generate function names that broke the previous
scheme. This fixes that.
llvm-svn: 80649
moves. This avoids the need to promote the operands (or implicitly
extend them, a partial register update condition), and can reduce
i8 register pressure. This substantially speeds up code such as
write_hex in lib/Support/raw_ostream.cpp.
subclass-coalesce.ll is too trivial and no longer tests what it was
originally intended to test.
llvm-svn: 80184
leads to partial-register definitions. To help avoid redundant
zero-extensions, also teach the h-register matching patterns that
use movzbl to match anyext as well as zext.
llvm-svn: 80099
more and is much nicer to the OS.
- Dan, please check. If there are parts of the test you think I should strip
out so it doesn't cause random failures let me know (there are still some PIC
label numbers in it, for example).
llvm-svn: 80019
code, according to Anton (I'm not totally convinced, but we can always
resurrect patches if we need to do so.)
- Start moving CellSPU's tests to prefer FileCheck.
llvm-svn: 79958
all Darwin targets; could be split into separate tests for
the chip subdirectories, but from Chris' last mail on testing
I assume he'd rather have only one test. Generic seems to be
the best available, maybe there should be a Darwin subdirectory?
llvm-svn: 79877
When undoing a reuse in ReuseInfo::GetRegForReload, check if it was only a
sub-register being used. The MachineOperand::getSubReg() method is only valid
for virtual registers, so we have to recover the sub-register index manually.
llvm-svn: 79855
over absolute addressing even in non-PIC mode (unless the address
has an index or something else incompatible), because it has a
smaller encoding.
llvm-svn: 79553
This is derived from a patch by Anton Korzh. I modified it to recognize
the VEXT shuffles during legalization and lower them to a target-specific
DAG node.
llvm-svn: 79428
remove RemoveDuplicateSuccessor, as it is no longer necessary, and because
it breaks assumptions made in
MachineBasicBlock::isOnlyReachableByFallthrough.
Convert test/CodeGen/X86/omit-label.ll to FileCheck and add a testcase
for PR4732.
test/CodeGen/Thumb2/thumb2-ifcvt2.ll sees a diff with this commit due to
it being bugpoint-reduced to the point where it doesn't matter what the
condition for the branch is.
Add some more interesting code to
test/CodeGen/X86/2009-08-06-branchfolder-crash.ll, which is the testcase
that originally motivated the RemoveDuplicateSuccessor code, to help
verify that the original problem isn't being re-broken.
llvm-svn: 79338
for a single "m" constraint; this is wrong because the
opcode of a load or store would have to change in parallel.
This patch makes it always compute addresses into a register,
which is correct but not as efficient as possible. 7144566.
llvm-svn: 79292
support unaligned mem access only for certain types. (Should it be size
instead?)
ARM v7 supports unaligned access for i16 and i32, some v6 variants support it
as well.
llvm-svn: 79127
It is legal for an inline asm operand to use an earlyclobber register if the
use operand is tied to the earlyclobber operand. The issue is discussed here:
http://gcc.gnu.org/ml/gcc/1999-04n/msg00431.html
We should perhaps let only the machine code verifier worry about these finer
details. EarlyClobber operands are not really interesting to the scavenger.
This fixes PR4528 for the third time.
llvm-svn: 79122
In a naked function, the flag is never set and getPristineRegs() returns an
empty list. That means naked functions are able to clobber callee saved
registers, but that is the whole point of naked functions.
This fixes PR4716.
llvm-svn: 79096
In the included test case, a stack load was not included in DistanceMap. That
caused TransferDeadness to ignore the instruction, leading to a scavenger
assert.
llvm-svn: 79090
support for globals going into the appropriate sections with the flags.
This hopefully finishes unbreaking the previous behavior that I broke before.
llvm-svn: 79079
x86_64-apple-darwin10.
--- Reverse-merging r78895 into '.':
U test/CodeGen/PowerPC/2008-12-12-EH.ll
U lib/Target/DarwinTargetAsmInfo.cpp
--- Reverse-merging r78892 into '.':
U include/llvm/Target/DarwinTargetAsmInfo.h
U lib/Target/X86/X86TargetAsmInfo.cpp
U lib/Target/X86/X86TargetAsmInfo.h
U lib/Target/ARM/ARMTargetAsmInfo.h
U lib/Target/ARM/ARMTargetMachine.cpp
U lib/Target/ARM/ARMTargetAsmInfo.cpp
U lib/Target/PowerPC/PPCTargetAsmInfo.cpp
U lib/Target/PowerPC/PPCTargetAsmInfo.h
U lib/Target/PowerPC/PPCTargetMachine.cpp
G lib/Target/DarwinTargetAsmInfo.cpp
llvm-svn: 78919
The register scavenger maintains a DistanceMap that maps MI pointers to their
distance from the top of the current MBB. The DistanceMap is built
incrementally in forward() and in bulk in findFirstUse(). It is used by
scavengeRegister() to determine which candidate register has the longest
unused interval.
Unfortunately the DistanceMap contents can become outdated. The first time
scavengeRegister() is called, the DistanceMap is filled to cover the MBB. If
then instructions are inserted in the MBB (as they always are following
scavengeRegister()), the recorded distances are too short. This causes bad
behaviour in the included test case where a register use /after/ the current
position is ignored because findFirstUse() thinks is is /before/ the current
position. A "using an undefined register" assertion follows promptly.
The fix is to build a fresh DistanceMap at the top of scavengeRegister(), and
discard it after use. This means that DistanceMap is no longer needed as a
RegScavenger member variable, and forward() doesn't need to update it.
The fix then discloses issue number two in the same test case: The candidate
search in scavengeRegister() finds a CSR that has been saved in the prologue,
but is currently unused. It would be both inefficient and wrong to spill such
a register in the emergency spill slot. In the present case, the emergency
slot restore is placed immediately before the normal epilogue restore, leading
to a "Redefining a live register" assertion.
Fix number two: When scavengerRegister() stumbles upon an unused register that
is overwritten later in the MBB, return that register early. It is important
to verify that the register is defined later in the MBB, otherwise it might be
an unspilled CSR.
llvm-svn: 78650
the overloaded vector types allowed floating-point or integer vector elements.
Most of these operations actually depend on the element type, so bitcasting
was not an option.
If you include the vpadd intrinsics that I updated earlier, this gets rid
of 20 intrinsics.
llvm-svn: 78646
MERGE_VALUES nodes. Replacing the result values with the
operands in one MERGE_VALUES node may cause another
MERGE_VALUES node be CSE'd with the first one, and bring
its uses along, so that the first one isn't dead, as this
code expects. Fix this by iterating until the node is
really dead. This fixes PR4699.
llvm-svn: 78619
instead of syntactically as a string. This means that it keeps track of the
segment, section, flags, etc directly and asmprints them in the right format.
This also includes parsing and validation support for llvm-mc and
"attribute(section)", so we should now start getting errors about invalid
section attributes from the compiler instead of the assembler on darwin.
Still todo:
1) Uniquing of darwin mcsections
2) Move all the Darwin stuff out to MCSectionMachO.[cpp|h]
3) there are a few FIXMEs, for example what is the syntax to get the
S_GB_ZEROFILL segment type?
llvm-svn: 78547
Blackfin supports and/or/xor on i32 but not on i16. Teach
DAGCombiner::SimplifyBinOpWithSameOpcodeHands to not produce illegal nodes
after legalize ops.
llvm-svn: 78497
Verify that early clobber registers and their aliases are not used.
All changes to RegsAvailable are now done as a transaction so the order of
operands makes no difference.
The included test case is from PR4686. It has behaviour that was dependent on the order of operands.
llvm-svn: 78465
This patch takes pain to ensure all the PEI lowering code does the right thing when lowering frame indices, insert code to manipulate stack pointers, etc. It's also custom lowering dynamic stack alloc into pseudo instructions so we can insert the right instructions at scheduling time.
This fixes PR4659 and PR4682.
llvm-svn: 78361
by aggressive chain operand optimization. UpdateNodeOperands
does not modify the node in place if it would result in
a node identical to an existing node.
llvm-svn: 78297
and high-bits values in ways that weren't correct for integer
types wider than 64 bits. This fixes a miscompile in
PPMacroExpansion.cpp in clang on x86-64.
llvm-svn: 78295
Instead of awkwardly encoding calling-convention information with ISD::CALL,
ISD::FORMAL_ARGUMENTS, ISD::RET, and ISD::ARG_FLAGS nodes, TargetLowering
provides three virtual functions for targets to override:
LowerFormalArguments, LowerCall, and LowerRet, which replace the custom
lowering done on the special nodes. They provide the same information, but
in a more immediately usable format.
This also reworks much of the target-independent tail call logic. The
decision of whether or not to perform a tail call is now cleanly split
between target-independent portions, and the target dependent portion
in IsEligibleForTailCallOptimization.
This also synchronizes all in-tree targets, to help enable future
refactoring and feature work.
llvm-svn: 78142
When LowerExtract eliminates an EXTRACT_SUBREG with a kill flag, it moves the
kill flag to the place where the sub-register is killed. This can accidentally
overlap with the use of a sibling sub-register, and we have trouble.
In the test case we have this code:
Live Ins: %R0 %R1 %R2
%R2L<def> = EXTRACT_SUBREG %R2<kill>, 1
%R2H<def> = LOAD16fi <fi#-1>, 0, Mem:LD(2,4) [FixedStack-1 + 0]
%R1L<def> = EXTRACT_SUBREG %R1<kill>, 1
%R0L<def> = EXTRACT_SUBREG %R0<kill>, 1
%R0H<def> = ADD16 %R2H<kill>, %R2L<kill>, %AZ<imp-def>, %AN<imp-def>, %AC0<imp-def>, %V<imp-def>, %VS<imp-def>
subreg: CONVERTING: %R2L<def> = EXTRACT_SUBREG %R2<kill>, 1
subreg: eliminated!
subreg: killed here: %R0H<def> = ADD16 %R2H, %R2L, %R2<imp-use,kill>, %AZ<imp-def>, %AN<imp-def>, %AC0<imp-def>, %V<imp-def>, %VS<imp-def>
The kill flag on %R2 is moved to the last instruction, and the live range overlaps with the definition of %R2H:
*** Bad machine code: Redefining a live physical register ***
- function: f
- basic block: 0x18358c0 (#0)
- instruction: %R2H<def> = LOAD16fi <fi#-1>, 0, Mem:LD(2,4) [FixedStack-1 + 0]
Register R2H was defined but already live.
The fix is to replace EXTRACT_SUBREG with IMPLICIT_DEF instead of eliminating
it completely:
subreg: CONVERTING: %R2L<def> = EXTRACT_SUBREG %R2<kill>, 1
subreg: replace by: %R2L<def> = IMPLICIT_DEF %R2<kill>
Note that these IMPLICIT_DEF instructions survive to the asm output. It is
necessary to fix the stack-color-with-reg test case because of that.
llvm-svn: 78093
killed by another operand.
There is probably a better fix. Either 1) scavenger can look at other operands, or
2) livevariables can be smarter about kill markers. Patches welcome.
llvm-svn: 78072
When LowerSubregsInstructionPass::LowerInsert eliminates an INSERT_SUBREG
instriction because it is an identity copy, make sure that the same registers
are alive before and after the elimination.
When the super-register is marked <undef> this requires inserting an
IMPLICIT_DEF instruction to make sure the super register is live.
Fix a related bug where a kill flag on the inserted sub-register was not transferred properly.
Finally, clear the undef flag in MachineInstr::addRegisterKilled. Undef implies dead and kill implies live, so they cant both be valid.
llvm-svn: 77989
This is not just a matter of passing in the target triple from the module;
currently backends are making decisions based on the build and host
architecture. The goal is to migrate to making these decisions based off of the
triple (in conjunction with the feature string). Thus most clients pass in the
target triple, or the host triple if that is empty.
This has one important change in the way behavior of the JIT and llc.
For the JIT, it was previously selecting the Target based on the host
(naturally), but it was setting the target machine features based on the triple
from the module. Now it is setting the target machine features based on the
triple of the host.
For LLC, -march was previously only used to select the target, the target
machine features were initialized from the module's triple (which may have been
empty). Now the target triple is taken from the module, or the host's triple is
used if that is empty. Then the triple is adjusted to match -march.
The take away is that -march for llc is now used in conjunction with the host
triple to initialize the subtarget. If users want more deterministic behavior
from llc, they should use -mtriple, or set the triple in the input module.
llvm-svn: 77946
__builtin_bfin_ones does the same as ctpop, so it can be implemented in the front-end.
__builtin_bfin_loadbytes loads from an unaligned pointer with the disalignexcpt instruction. It does the same as loading from a pointer with the low bits masked. It is better if the front-end creates a masked load. We can always instruction select the masked to disalignexcpt+load.
We keep csync/ssync/idle. These intrinsics represent instructions that need workarounds for some silicon revisions. We may even want to convert inline assembler to intrinsics to enable the workarounds.
llvm-svn: 77917
Allow imp-def and imp-use of anything in the scavenger asserts, just like the machine code verifier.
Allow redefinition of a sub-register of a live register.
llvm-svn: 77904
to:
.quad X
even on a 32-bit system, where X is not 64-bits. There isn't much that
we can do here, so we just print:
.quad ((X) & 4294967295)
instead.
llvm-svn: 77818
myself because I'm getting tired of seeing the red buildbots, which have
been red since 5:30PM PDT last night.
Proposed supplement to developer policy: committers should make sure to
be around to watch for buildbot failures after committing.
llvm-svn: 77785
instructions for calls since BL and BLX are always 32-bit long and BX is always
16-bit long.
Also, we should be using BLX to call external function stubs.
llvm-svn: 77756
padding is disabled, tabs get replaced by spaces except in the case of
the first operand, where the tab is output to line up the operands after
the mnemonics.
Add some better comments and eliminate redundant code.
Fix some testcases to not assume tabs.
llvm-svn: 77740
into the mergable section if it is one of our special cases. This could
obviously be improved, but this is the minimal fix and restores us to the
previous behavior.
llvm-svn: 77679
When the return value is not used (i.e. only care about the value in the memory), x86 does not have to use add to implement these. Instead, it can use add, sub, inc, dec instructions with the "lock" prefix.
This is currently implemented using a bit of instruction selection trick. The issue is the target independent pattern produces one output and a chain and we want to map it into one that just output a chain. The current trick is to select it into a merge_values with the first definition being an implicit_def. The proper solution is to add new ISD opcodes for the no-output variant. DAG combiner can then transform the node before it gets to target node selection.
Problem #2 is we are adding a whole bunch of x86 atomic instructions when in fact these instructions are identical to the non-lock versions. We need a way to add target specific information to target nodes and have this information carried over to machine instructions. Asm printer (or JIT) can use this information to add the "lock" prefix.
llvm-svn: 77582
wide vectors. Likewise, change VSTn intrinsics to take separate arguments
for each vector in a multi-vector struct. Adjust tests accordingly.
llvm-svn: 77468
- This change also makes it possible to switch between ARM / Thumb on a
per-function basis.
- Fixed thumb2 routine which expand reg + arbitrary immediate. It was using
using ARM so_imm logic.
- Use movw and movt to do reg + imm when profitable.
- Other code clean ups and minor optimizations.
llvm-svn: 77300
and make it more aggressive, we now put:
const int G2 __attribute__((weak)) = 42;
into the text (readonly) segment like gcc, previously we put
it into the data (readwrite) segment.
llvm-svn: 77104
Before:
adr r12, #LJTI3_0_0
ldr pc, [r12, +r0, lsl #2]
LJTI3_0_0:
.long LBB3_24
.long LBB3_30
.long LBB3_31
.long LBB3_32
After:
adr r12, #LJTI3_0_0
add pc, r12, +r0, lsl #2
LJTI3_0_0:
b.w LBB3_24
b.w LBB3_30
b.w LBB3_31
b.w LBB3_32
This has several advantages.
1. This will make it easier to optimize this to a TBB / TBH instruction +
(smaller) table.
2. This eliminate the need for ugly asm printer hack to force the address
into thumb addresses (bit 0 is one).
3. Same codegen for pic and non-pic.
4. This eliminate the need to align the table so constantpool island pass
won't have to over-estimate the size.
Based on my calculation, the later is probably slightly faster as well since
ldr pc with shifter address is very slow. That is, it should be a win as long
as the HW implementation can do a reasonable job of branch predict the second
branch.
llvm-svn: 77024
dumping ground of various SSE4.1 tests, since filecheck can reasonably
handle them all in one file. Generalize it to check x86-64 stuff as
well since it has a different ABI (a convenient way to test both the
reg and mem forms of these instructions).
llvm-svn: 76848
Getelementptrs that are defined to wrap are virtually useless to
optimization, and getelementptrs that are undefined on any kind
of overflow are too restrictive -- it's difficult to ensure that
all intermediate addresses are within bounds. I'm going to take
a different approach.
Remove a few optimizations that depended on this flag.
llvm-svn: 76437
Inline asm instructions may have additional <imp-def,kill> register operands.
These operands are not marked with a flag like the normal asm operands, so we
must not assert that there is a flag.
llvm-svn: 76373
stack alignment right when it is. This is not
ideal but conservatively correct. Adjust a test
to compensate for changed stack offset value.
gcc.apple/asm-block-57.c
llvm-svn: 76120
The inline asm operands must be parsed from the first flag, you cannot assume
that an immediate operand preceeding a register use operand is the flag.
PowerPC "m" operands are represented as (flag, imm, reg) triples.
isRegTiedToDefOperand() would incorrectly interpret the imm as the flag.
llvm-svn: 76101
additional bug fixes:
1. The bug that everyone hit was a problem in the asmprinter where it
would remove $stub but keep the L prefix on a name when emitting the
indirect symbol. This is easy to fix by keeping the name of the stub
and the name of the symbol in a StringMap instead of just keeping a
StringSet and trying to reconstruct it late.
2. There was a problem printing the personality function. The current
logic to print out the personality function from the DWARF information
is a bit of a cesspool right now that duplicates a bunch of other
logic in the asm printer. The short version of it is that it depends
on emitting both the L and _ prefix for symbols (at least on darwin)
and until I can untangle it, it is best to switch the mangler back to
emitting both prefixes.
llvm-svn: 75646
unbreaking llvm-gcc (on Darwin).
--- Reverse-merging r75620 into '.':
U include/llvm/Support/Mangler.h
--- Reverse-merging r75610 into '.':
U test/CodeGen/X86/loop-hoist.ll
G include/llvm/Support/Mangler.h
U lib/Target/X86/AsmPrinter/X86ATTAsmPrinter.cpp
U lib/VMCore/Mangler.cpp
llvm-svn: 75636
to symbols instead of doing it with "printSuffixedName". This gets us to the point
where there is a real separation between computing a symbol name and printing it,
something I need for MC printer stuff.
This patch also fixes a corner case bug where unnamed private globals wouldn't get
the private label prefix.
Next up, rename all uses of getValueName -> getMangledName for better greppability,
and then tackle the ppc/arm backends to eliminate "printSuffixedName".
llvm-svn: 75610
indicates whether the label is private or not, instead of taking
prefix stuff. One effect of this is that symbols will be generated
with *just* the private prefix, instead of both the private prefix
*and* the user-label-prefix, but this doesn't matter as long as it
is consistent. For example we'll now get "Lfoo" instead of "L_foo".
These are just assembler temporary labels anyway, so they never even
make it into the .o file.
llvm-svn: 75607
1) unique globals with the existing "Count" local in Mangler, not with
atomic nonsense. Using atomics will give us nondeterminstic output
from the compiler when using multiple threads, which is bad.
2) Do not mangle an unknown global name with a type suffix. We don't
need this anymore now that llvm ir doesn't have type planes.
llvm-svn: 75541
of lea. It is better for code size (and presumably efficiency) to use:
movl $foo, %eax
rather than:
leal foo, eax
Both give a nice zero extending "move immediate" instruction, the former is just
smaller. Note that global addresses should be handled different by the x86
backend, but I chose to follow the style already in place and add more fixme's.
llvm-svn: 75403
Basically, using:
lea symbol(%rip), %rax
is not valid in -static mode, because the current RIP may not be
within 32-bits of "symbol" when an app is built partially pic and
partially static. The fix for this is to compile it to:
lea symbol, %rax
It would be better to codegen this as:
movq $symbol, %rax
but that will come next.
The hard part of fixing this bug was fixing abi-isel, which was actively
testing for the wrong behavior. Also, the RUN lines are completely impossible
to understand what they are testing. To help with this, convert the -static
x86-64 codegen tests to use filecheck. This is much more stable and makes it
more clear what the codegen is expected to be.
llvm-svn: 75382
value. Adjust other code to deal with that correctly. Make
DAGTypeLegalizer::PromoteIntRes_EXTRACT_VECTOR_ELT take advantage of
this new flexibility to simplify the code and make it deal with unusual
vectors (like <4 x i1>) correctly. Fixes PR3037.
llvm-svn: 75176
registers based on dynamic conditions. For example, X86 EBP/RBP, when used as
frame register has to be spilled in the first fixed object. It should inform
PEI this so it doesn't get allocated another stack object. Also, it should not
be spilled as other callee-saved registers but rather its spilling and restoring
are being handled by emitPrologue and emitEpilogue. Avoid spilling it twice.
llvm-svn: 75116
as an (index,bool) pair. The bool flag records whether the kill is a
PHI kill or not. This code will be used to enable splitting of live
intervals containing PHI-kills.
A slight change to live interval weights introduced an extra spill
into lsr-code-insertion (outside the critical sections). The test
condition has been updated to reflect this.
llvm-svn: 75097
* remove some old code that was needed when we'd put ESP in the scale instead of
the base of some instructions.
* Fix a bug with the P modifier in inline asm that caused us to drop it.
llvm-svn: 75077
VSETCC must define all bits, which is different than it was documented
to before. Since all targets that implement VSETCC already have this
behavior, and we don't optimize based on this, just change the
documentation. We now get nice code for vec_compare.ll
llvm-svn: 74978
Note, isUndef marker must be placed even on implicit_def def operand or else the scavenger will not ignore it. This is necessary because -O0 path does not use liveintervalanalysis, it treats implicit_def just like any other def.
llvm-svn: 74601
Avoid unnecessary duplication of operand 0 of X86::FpSET_ST0_80. This duplication would
cause one register to remain on the stack at the function return.
llvm-svn: 74534
The register allocator, when it allocates a register to a virtual register defined by an implicit_def, can allocate any physical register without worrying about overlapping live ranges. It should mark all of operands of the said virtual register so later passes will do the right thing.
This is not the best solution. But it should be a lot less fragile to having the scavenger try to track what is defined by implicit_def.
llvm-svn: 74518
After much back and forth, I decided to deviate from ARM design and split LDR into 4 instructions (r + imm12, r + imm8, r + r << imm12, constantpool). The advantage of this is 1) it follows the latest ARM technical manual, and 2) makes it easier to reduce the width of the instruction later. The down side is this creates more inconsistency between the two sub-targets. We should split ARM LDR instruction in a similar fashion later. I've added a README entry for this.
llvm-svn: 74420
implementation primarily differs from the former in that the asmprinter
doesn't make a zillion decisions about whether or not something will be
RIP relative or not. Instead, those decisions are made by isel lowering
and propagated through to the asm printer. To achieve this, we:
1. Represent RIP relative addresses by setting the base of the X86 addr
mode to X86::RIP.
2. When ISel Lowering decides that it is safe to use RIP, it lowers to
X86ISD::WrapperRIP. When it is unsafe to use RIP, it lowers to
X86ISD::Wrapper as before.
3. This removes isRIPRel from X86ISelAddressMode, representing it with
a basereg of RIP instead.
4. The addressing mode matching logic in isel is greatly simplified.
5. The asmprinter is greatly simplified, notably the "NotRIPRel" predicate
passed through various printoperand routines is gone now.
6. The various symbol printing routines in asmprinter now no longer infer
when to emit (%rip), they just print the symbol.
I think this is a big improvement over the previous situation. It does have
two small caveats though: 1. I implemented a horrible "no-rip" modifier for
the inline asm "P" constraint modifier. This is a short term hack, there is
a much better, but more involved, solution. 2. I had to xfail an
-aggressive-remat testcase because it isn't handling the use of RIP in the
constant-pool reading instruction. This specific test is easy to fix without
-aggressive-remat, which I intend to do next.
llvm-svn: 74372
trip counts in more cases.
Generalize ScalarEvolution's isLoopGuardedByCond code to recognize
And and Or conditions, splitting the code out into an
isNecessaryCond helper function so that it can evaluate Ands and Ors
recursively, and make SCEVExpander be much more aggressive about
hoisting instructions out of loops.
test/CodeGen/X86/pr3495.ll has an additional instruction now, but
it appears to be due to an arbitrary register allocation difference.
llvm-svn: 74048
a global with that gets printed with the :mem modifier. All operands to lea's
should be handled with the lea32mem operand kind, and this allows the TLS stuff
to do this. There are several better ways to do this, but I went for the minimal
change since I can't really test this (beyond make check).
This also makes the use of EBX explicit in the operand list in the 32-bit,
instead of implicit in the instruction.
llvm-svn: 73834
while experimenting. I'm reasonably sure this is correct, but please
tell me if these instructions have some strange property which makes this
change unsafe.
llvm-svn: 73746
casted induction variables in cases where the cast
isn't foldable. It ended up being a pessimization in
many cases. This could be fixed, but it would require
a bunch of complicated code in IVUsers' clients. The
advantages of this approach aren't visible enough to
justify it at this time.
llvm-svn: 73706
TurnCopyIntoImpDef turns a copy into implicit_def and remove the val# defined by it. This causes an scavenger assertion later if the def reaches other blocks. Disable the transformation if the value live interval extends beyond its def block.
llvm-svn: 73478
support for x86, and UMULO/SMULO for many architectures, including PPC
(PR4201), ARM, and Cell. The resulting expansion isn't perfect, but it's
not bad.
llvm-svn: 73477
incomming chain of the RETURN node. The incomming chain must
be the outgoing chain of the CALL node. This causes the
backend to identify tail calls that are not tail calls. This
patch fixes this.
llvm-svn: 73387
- Change register allocation hint to a pair of unsigned integers. The hint type is zero (which means prefer the register specified as second part of the pair) or entirely target dependent.
- Allow targets to specify alternative register allocation orders based on allocation hint.
Part 2.
- Use the register allocation hint system to implement more aggressive load / store multiple formation.
- Aggressively form LDRD / STRD. These are formed *before* register allocation. It has to be done this way to shorten live interval of base and offset registers. e.g.
v1025 = LDR v1024, 0
v1026 = LDR v1024, 0
=>
v1025,v1026 = LDRD v1024, 0
If this transformation isn't done before allocation, v1024 will overlap v1025 which means it more difficult to allocate a register pair.
- Even with the register allocation hint, it may not be possible to get the desired allocation. In that case, the post-allocation load / store multiple pass must fix the ldrd / strd instructions. They can either become ldm / stm instructions or back to a pair of ldr / str instructions.
This is work in progress, not yet enabled.
llvm-svn: 73381
consecutive addresses togther. This makes it easier for the post-allocation pass
to form ldm / stm.
This is step 1. We are still missing a lot of ldm / stm opportunities because
of register allocation are not done in the desired order. More enhancements
coming.
llvm-svn: 73291
out of sync with regular cc.
The only difference between the tail call cc and the normal
cc was that one parameter register - R9 - was reserved for
calling functions through a function pointer. After time the
tail call cc has gotten out of sync with the regular cc.
We can use R11 which is also caller saved but not used as
parameter register for potential function pointers and
remove the special tail call cc on x86-64.
llvm-svn: 73233
on x86 to handle more cases. Fix a bug in said code that would cause it
to read past the end of an object. Rewrite the code in
SelectionDAGLegalize::ExpandBUILD_VECTOR to be a bit more general.
Remove PerformBuildVectorCombine, which is no longer necessary with
these changes. In addition to simplifying the code, with this change,
we can now catch a few more cases of consecutive loads.
llvm-svn: 73012
nodes for vectors with an i16 element type. Add an optimization for
building a vector which is all zeros/undef except for the bottom
element, where the bottom element is an i8 or i16.
llvm-svn: 72988
build vectors with i64 elements will only appear on 32b x86 before legalize.
Since vector widening occurs during legalize, and produces i64 build_vector
elements, the dag combiner is never run on these before legalize splits them
into 32b elements.
Teach the build_vector dag combine in x86 back end to recognize consecutive
loads producing the low part of the vector.
Convert the two uses of TLI's consecutive load recognizer to pass LoadSDNodes
since that was required implicitly.
Add a testcase for the transform.
Old:
subl $28, %esp
movl 32(%esp), %eax
movl 4(%eax), %ecx
movl %ecx, 4(%esp)
movl (%eax), %eax
movl %eax, (%esp)
movaps (%esp), %xmm0
pmovzxwd %xmm0, %xmm0
movl 36(%esp), %eax
movaps %xmm0, (%eax)
addl $28, %esp
ret
New:
movl 4(%esp), %eax
pmovzxwd (%eax), %xmm0
movl 8(%esp), %eax
movaps %xmm0, (%eax)
ret
llvm-svn: 72957
integer and floating-point opcodes, introducing
FAdd, FSub, and FMul.
For now, the AsmParser, BitcodeReader, and IRBuilder all preserve
backwards compatability, and the Core LLVM APIs preserve backwards
compatibility for IR producers. Most front-ends won't need to change
immediately.
This implements the first step of the plan outlined here:
http://nondot.org/sabre/LLVMNotes/IntegerOverflow.txt
llvm-svn: 72897
Update code generator to use this attribute and remove DisableRedZone target option.
Update llc to set this attribute when -disable-red-zone command line option is used.
llvm-svn: 72894
e.g.
orl $65536, 8(%rax)
=>
orb $1, 10(%rax)
Since narrowing is not always a win, e.g. i32 -> i16 is a loss on x86, dag combiner consults with the target before performing the optimization.
llvm-svn: 72507
The DAGCombiner created a negative shiftamount, stored in an
unsigned variable. Later the optimizer eliminated the shift entirely as being
undefined.
Example: (srl (shl X, 56) 48). ShiftAmt is 4294967288.
Fix it by checking that the shiftamount is positive, and storing in a signed
variable.
llvm-svn: 72331
and it wasn't generating calls through @PLT for these functions.
hasLocalLinkage() is now false for available_externally,
I attempted to fix the inliner and dce to handle available_externally properly.
It passed make check.
llvm-svn: 72328
code in preparation for code generation. The main thing it does
is handle the case when eh.exception calls (and, in a future
patch, eh.selector calls) are far away from landing pads. Right
now in practice you only find eh.exception calls close to landing
pads: either in a landing pad (the common case) or in a landing
pad successor, due to loop passes shifting them about. However
future exception handling improvements will result in calls far
from landing pads:
(1) Inlining of rewinds. Consider the following case:
In function @f:
...
invoke @g to label %normal unwind label %unwinds
...
unwinds:
%ex = call i8* @llvm.eh.exception()
...
In function @g:
...
invoke @something to label %continue unwind label %handler
...
handler:
%ex = call i8* @llvm.eh.exception()
... perform cleanups ...
"rethrow exception"
Now inline @g into @f. Currently this is turned into:
In function @f:
...
invoke @something to label %continue unwind label %handler
...
handler:
%ex = call i8* @llvm.eh.exception()
... perform cleanups ...
invoke "rethrow exception" to label %normal unwind label %unwinds
unwinds:
%ex = call i8* @llvm.eh.exception()
...
However we would like to simplify invoke of "rethrow exception" into
a branch to the %unwinds label. Then %unwinds is no longer a landing
pad, and the eh.exception call there is then far away from any landing
pads.
(2) Using the unwind instruction for cleanups.
It would be nice to have codegen handle the following case:
invoke @something to label %continue unwind label %run_cleanups
...
handler:
... perform cleanups ...
unwind
This requires turning "unwind" into a library call, which
necessarily takes a pointer to the exception as an argument
(this patch also does this unwind lowering). But that means
you are using eh.exception again far from a landing pad.
(3) Bugpoint simplifications. When bugpoint is simplifying
exception handling code it often generates eh.exception calls
far from a landing pad, which then causes codegen to assert.
Bugpoint then latches on to this assertion and loses sight
of the original problem.
Note that it is currently rare for this pass to actually do
anything. And in fact it normally shouldn't do anything at
all given the code coming out of llvm-gcc! But it does fire
a few times in the testsuite. As far as I can see this is
almost always due to the LoopStrengthReduce codegen pass
introducing pointless loop preheader blocks which are landing
pads and only contain a branch to another block. This other
block contains an eh.exception call. So probably by tweaking
LoopStrengthReduce a bit this can be avoided.
llvm-svn: 72276
build an integer and cast that to a float. This fixes a crash
caused by trying to split an f32 into two f16's.
This changes the behavior in test/CodeGen/XCore/fneg.ll because that
testcase now triggers a DAGCombine which converts the fneg into an integer
operation. If someone is interested, it's probably possible to tweak
the test to generate an actual fneg.
llvm-svn: 72162
When a test fails with more than a pipeful of output on stdout AND stderr, one
of the DejaGnu programs blocks. The problem can be avoided by redirecting
stdout to a file.
llvm-svn: 71919
Anton Korobeynikov