This change attempts to simplify (X^Y) -> X or Y in the user's context if we know that
only bits from X or Y are demanded.
A minimized case is provided bellow. This change will simplify "t>>16" into "var1 >>16".
=============================================================
unsigned foo (unsigned val1, unsigned val2) {
unsigned t = val1 ^ 1234;
return (t >> 16) | t; // NOTE: t is used more than once.
}
=============================================================
Note that if the "t" were used only once, the expression would be finally optimized as well.
However, with with this change, the optimization will take place earlier.
Reviewed by Nadav, Thanks a lot!
llvm-svn: 169317
The count attribute is more accurate with regards to the size of an array. It
also obviates the upper bound attribute in the subrange. We can also better
handle an unbound array by setting the count to -1 instead of the lower bound to
1 and upper bound to 0.
llvm-svn: 169312
This reapplies the fix for PR13303 now with more justification. Based on my
execution of the GDB 7.5 test suite this results in:
expected passes: 16101 -> 20890 (+30%)
unexpected failures: 4826 -> 637 (-77%)
There are 23 checks that used to pass and now fail. They are all in
gdb.reverse. Investigating a few looks like they were accidentally passing
due to extra breakpoints being set by this bug. They're generally due to the
difference in end location between gcc and clang, the test suite is trying to
set breakpoints on the closing '}' that clang doesn't associate with any
instructions.
llvm-svn: 169304
textually as NativeClient. Also added a link to the native client project for
readers unfamiliar with it.
A Clang patch will follow shortly.
llvm-svn: 169291
on 64-bit PowerPC ELF.
The patch includes code to handle external assembly and MC output with the
integrated assembler. It intentionally does not support the "old" JIT.
For the initial-exec TLS model, the ABI requires the following to calculate
the address of external thread-local variable x:
Code sequence Relocation Symbol
ld 9,x@got@tprel(2) R_PPC64_GOT_TPREL16_DS x
add 9,9,x@tls R_PPC64_TLS x
The register 9 is arbitrary here. The linker will replace x@got@tprel
with the offset relative to the thread pointer to the generated GOT
entry for symbol x. It will replace x@tls with the thread-pointer
register (13).
The two test cases verify correct assembly output and relocation output
as just described.
PowerPC-specific selection node variants are added for the two
instructions above: LD_GOT_TPREL and ADD_TLS. These are inserted
when an initial-exec global variable is encountered by
PPCTargetLowering::LowerGlobalTLSAddress(), and later lowered to
machine instructions LDgotTPREL and ADD8TLS. LDgotTPREL is a pseudo
that uses the same LDrs support added for medium code model's LDtocL,
with a different relocation type.
The rest of the processing is straightforward.
llvm-svn: 169281
Again, tools are trickier to pick the main module header for than
library source files. I've started to follow the pattern of using
LLVMContext.h when it is included as a stub for program source files.
llvm-svn: 169252
missed in the first pass because the script didn't yet handle include
guards.
Note that the script is now able to handle all of these headers without
manual edits. =]
llvm-svn: 169224
1) Teach it to handle files with #include on the first line -- these do
actually exist in LLVM.
2) Support llvm-c and clang-c include projects.
3) Nuke some stail imports.
4) Switch to using os.path to split the file extension off.
5) Remove debugging leftovers.
6) Add docstring (a really puny one) for the sort function.
I'm continuing te avoid stripping the whitespace on the RHS to preserve
whatever newline characters happen to be in the original file.
llvm-svn: 169222
The count field is necessary because there isn't a difference between the 'lo'
and 'hi' attributes for a one-element array and a zero-element array. When the
count is '0', we know that this is a zero-element array. When it's >=1, then
it's a normal constant sized array. When it's -1, then the array is unbounded.
llvm-svn: 169218
Added the code that actually performs the if-conversion during vectorization.
We can now vectorize this code:
for (int i=0; i<n; ++i) {
unsigned k = 0;
if (a[i] > b[i]) <------ IF inside the loop.
k = k * 5 + 3;
a[i] = k; <---- K is a phi node that becomes vector-select.
}
llvm-svn: 169217
the alignment is clamped to TargetFrameLowering.getStackAlignment if the target
does not support stack realignment or the option "realign-stack" is off.
This will cause miscompile if the address is treated as aligned and add is
replaced with or in DAGCombine.
Added a bool StackRealignable to TargetFrameLowering to check whether stack
realignment is implemented for the target. Also added a bool RealignOption
to MachineFrameInfo to check whether the option "realign-stack" is on.
rdar://12713765
llvm-svn: 169197
Now that there can be multiple hint registers from targets, it doesn't
make sense to have a function that returns 'the' preferred register.
llvm-svn: 169190
Targets can provide multiple hints now, so getRegAllocPref() doesn't
make sense any longer because it only returns one preferred register.
Replace it with getSimpleHint() in the remaining heuristics. This
function only
llvm-svn: 169188
This change tries to simmplify E1 = " X >> C1 << C2" into :
- E2 = "X << (C2 - C1)" if C2 > C1, or
- E2 = "X >> (C1 - C2)" if C1 > C2, or
- E2 = X if C1 == C2.
Reviewed by Nadav. Thanks!
llvm-svn: 169182
Virtual registers with a known preferred register are prioritized by
RAGreedy. This function makes the condition explicit without depending
on getRegAllocPref().
llvm-svn: 169179
This small change adds support for that. It will make all MCJIT tests pass
in make-check on BigEndian platforms.
Patch by Petar Jovanovic.
llvm-svn: 169178
This change adds endian-awareness to MipsJITInfo and emitWordLE in
MipsCodeEmitter has become emitWord now to support both endianness.
Patch by Petar Jovanovic.
llvm-svn: 169177
This provides the same functionality as getRawAllocationOrder() for the
even/odd hints, but without the many constant register arrays.
llvm-svn: 169169
"Windows.h" includes <Windows.h> which defines a bunch of stuff it shouldn't
(even with all the restriction macros). We have no control over this file, so
make it's scope as small as possible.
llvm-svn: 169165
The TargetRegisterInfo::getRegAllocationHints() function is going to
replace the existing mechanisms for providing target-dependent hints to
the register allocator: ResolveRegAllocHint() and
getRawAllocationOrder().
The new hook is more flexible because it allows the target to provide
multiple preferred candidate registers for each virtual register, and it
is easier to use because targets are not required to return a reference
to a constant array like getRawAllocationOrder().
An optional VirtRegMap argument can be used to provide target-dependent
hints that depend on the provisional assignments of other virtual
registers.
llvm-svn: 169154
which is the legality of the if-conversion transformation. The next step is to
implement the cost-model for the if-converted code as well as the
vectorization itself.
llvm-svn: 169152
is not yet good enough for more sophistication. The important goal of this
test is to make sure llc doesn't crash on this IR like it used to.
llvm-svn: 169146
AKA: Recompile *ALL* the source code!
This one went much better. No manual edits here. I spot-checked for
silliness and grep-checked for really broken edits and everything seemed
good. It all still compiles. Yell if you see something that looks goofy.
llvm-svn: 169133
Sooooo many of these had incorrect or strange main module includes.
I have manually inspected all of these, and fixed the main module
include to be the nearest plausible thing I could find. If you own or
care about any of these source files, I encourage you to take some time
and check that these edits were sensible. I can't have broken anything
(I strictly added headers, and reordered them, never removed), but they
may not be the headers you'd really like to identify as containing the
API being implemented.
Many forward declarations and missing includes were added to a header
files to allow them to parse cleanly when included first. The main
module rule does in fact have its merits. =]
llvm-svn: 169131
standards.
I am a terrible Python programmer. Patches more the welcome. Please tell
me how this should look if it should look differently. It's just a tiny
little script so it didn't make sense to go through pre-commit review,
especially as someone who actually knows python may want to just rip it
apart and do it The Right Way.
I will be preparing a commit shortly that uses this script to
canonicalize *all* of the #include lines in LLVM. Really, all of them.
llvm-svn: 169125
The partitioning logic attempted to handle uses of an alloca with an
offset starting before the alloca so long as the use had some overlap
with the alloca itself. However, there was a bug where we tested
'(uint64_t)Offset >= AllocSize' without first checking whether 'Offset'
was positive. As a consequence, essentially every negative offset (that
is, starting *before* the alloca does) would be thrown out, even if it
was overlapping. The subsequent code to throw out negative offsets which
were actually non-overlapping was essentially dead. The code to *handle*
overlapping negative offsets was actually dead!
I've just removed all of this, and taught SROA to discard any uses which
start prior to the alloca from the beginning. It has the lovely property
of simplifying the code. =] All the tests still pass, and in fact no new
tests are needed as this is already covered by our testsuite. Fixing the
code so that negative offsets work the way the comments indicate they
were supposed to work causes regressions. That's how I found this.
Anyways, this is all progress in the correct direction -- tightening up
SROA to be maximally aggressive. Some day, I really hope to turn
out-of-bounds accesses to an alloca into 'unreachable'.
llvm-svn: 169120
; CHECK: [[VAR:[a-z]]]
The problem was that to find the end of the regex var definition, it was
simplistically looking for the next ]] and finding the incorrect one. A
better approach is to count nesting of brackets (taking escaping into
account). This way the brackets that are part of the regex can be discovered
and skipped properly, and the ]] ending is detected in the right place.
llvm-svn: 169109
Also check in a case to repeat the issue, on which 'opt -globalopt' consumes 1.6GB memory.
The big memory footprint cause is that current GlobalOpt one by one hoists and stores the leaf element constant into the global array, in each iteration, it recreates the global array initializer constant and leave the old initializer alone. This may result in many obsolete constants left.
For example: we have global array @rom = global [16 x i32] zeroinitializer
After the first element value is hoisted and installed: @rom = global [16 x i32] [ 1, 0, 0, ... ]
After the second element value is installed: @rom = global [16 x 32] [ 1, 2, 0, 0, ... ] // here the previous initializer is obsolete
...
When the transform is done, we have 15 obsolete initializers left useless.
llvm-svn: 169079
- Each macro instantiation introduces a new buffer, and FindBufferForLoc() is
linear, so previously macro instantiation could be N^2 for some pathological
inputs.
llvm-svn: 169073
The TwoAddressInstructionPass takes the machine code out of SSA form by
expanding REG_SEQUENCE instructions into copies. It is no longer
necessary to rewrite the registers used by a REG_SEQUENCE instruction
because the new coalescer algorithm can do it now.
REG_SEQUENCE is just converted to a sequence of sub-register copies now.
llvm-svn: 169067
MachineCopyPropagation doesn't understand super-register liveness well
enough to be able to remove implicit defs of super-registers.
This fixes a problem in ARM/2012-01-26-CopyPropKills.ll that is exposed
by an future TwoAddressInstructionPass change. The KILL instructions are
removed before the machine code is emitted.
llvm-svn: 169060
uses. APFloat::convert() takes the pointer to the fltSemantics
variable, which is later accessed it in ~APFloat() desctructor.
That is, semantics must still be alive at the moment we delete
APFloat.
Found by experimental AddressSanitizer use-after-scope checker.
llvm-svn: 169047
The original patch removed a bunch of code that the SjLjEHPrepare pass placed
into the entry block if all of the landing pads were removed during the
CodeGenPrepare class. The more natural way of doing things is to run the CGP
*before* we run the SjLjEHPrepare pass.
Make it so!
llvm-svn: 169044
This causes llc to repeat the module compilation N times, making it
possible to get more accurate information from -time-passes when
compiling small modules.
llvm-svn: 169040
Codegen was failing with an assertion because of unexpected vector
operands when legalizing the selection DAG for a MUL instruction.
The asserting code was legalizing multiplies for vectors of size 128
bits. It uses a custom lowering to try and detect cases where it can
use a VMULL instruction instead of a VMOVL + VMUL. The code was
looking for input operands to the MUL that had been sign or zero
extended. If it found the extended operands it would drop the
sign/zero extension and use the original vector size as input to a
VMULL instruction.
The code assumed that the original input vector was 64 bits so that
after dropping the extension it would fit directly into a D register
and could be used as an operand of a VMULL instruction. The input
code that trigger the failure used a vector of <4 x i8> that was
sign extended to <4 x i32>. It was not safe to drop the sign
extension in this case because the original vector is only 32 bits
wide. The fix is to insert a sign extension for the vector to reach
the required 64 bit size. In this particular example, the vector would
need to be sign extented to a <4 x i16>.
llvm-svn: 169024
instruction (vmaddfp) to conform with IEEE to ensure the sign of a zero
result when resulting product is -0.0.
The -0.0 vector addend to vmaddfp is generated by a creating a vector
with full bits sets and then shifting each elements by 31-bits to the
left, resulting in a vector of 0x80000000 (or -0.0 as float).
The 'buildvec_canonicalize.ll' was adjusted to reflect this change and
the 'vec_mul.ll' was complemented with the float vector multiplication
test.
llvm-svn: 168998
Rationale:
1) This was the name in the comment block. ;]
2) It matches Clang's __has_feature naming convention.
3) It matches other compiler-feature-test conventions.
Sorry for the noise. =]
I've also switch the comment block to use a \brief tag and not duplicate
the name.
llvm-svn: 168996
references from whether it supports an R-value reference *this. No
version of GCC today supports the latter, which breaks GCC C++11
compiles of LLVM and Clang now.
Also add doxygen comments clarifying what's going on here, and update
the usage in Optional. I'll update the usages in Clang next.
llvm-svn: 168993
For example, don't allow empty strings to be passed to getInt.
Move asserts inside parseSpecifier. (One day we may want to pass parse
error messages to the user - from LLParser - instead of using asserts,
but keep the code simple until then. There have been an attempt to do
this. See r142288, which got reverted, and r142605.)
llvm-svn: 168991
We're iterating over a non-deterministically ordered container looking
for two saturating flags. To do this correctly, we have to saturate
both, and only stop looping if both saturate to their final value.
Otherwise, which flag we see first changes the result.
This is also a micro-optimization of the previous version as now we
don't go into the (possibly expensive) test logic once the first
violation of either constraint is detected.
llvm-svn: 168989
functionality changed.
Evan's commit r168970 moved the code that the primary comment in this
function referred to to the other end of the function without moving the
comment, and there has been a steady creep of "boolean" logic in it that
is simpler if handled via early exit. That way each special case can
have its own comments. I've also made the variable name a bit more
explanatory than "AllFit". This is in preparation to fix the
non-deterministic output of this function.
llvm-svn: 168988
The simplify-libcalls pass maintained a statistic to count the number
of library calls that have been simplified. Now that library call
simplification is being carried out in instcombine the statistic should
be moved to there.
llvm-svn: 168975
depends on the IR infrastructure, there is no sense in it being off in
Support land.
This is in preparation to start working to expand InstVisitor into more
special-purpose visitors that are still generic and can be re-used
across different passes. The expansion will go into the Analylis tree
though as nothing in VMCore needs it.
llvm-svn: 168972
This expands to '&', and is intended to be used when an /optional/ rvalue
override is available.
Before:
void foo() const { ... }
After:
void foo() const LLVM_LVALUE_FUNCTION { ... }
void foo() && { ... }
This is used to allow moving the contents of an Optional.
llvm-svn: 168963
This revision attempts to recognize following population-count pattern:
while(a) { c++; ... ; a &= a - 1; ... },
where <c> and <a>could be used multiple times in the loop body.
TODO: On X8664 and ARM, __buildin_ctpop() are not expanded to a efficent
instruction sequence, which need to be improved in the following commits.
Reviewed by Nadav, really appreciate!
llvm-svn: 168931
the last invoke instruction in the function. This also removes the last landing
pad in an function. This is fine, but with SjLj EH code, we've already placed a
bunch of code in the 'entry' block, which expects the landing pad to stick
around.
When we get to the situation where CGP has removed the last landing pad, go
ahead and nuke the SjLj instructions from the 'entry' block.
<rdar://problem/12721258>
llvm-svn: 168930
This patch migrates the puts optimizations from the simplify-libcalls
pass into the instcombine library call simplifier.
All the simplifiers from simplify-libcalls have now been migrated to
instcombine. Yay! Just a few other bits to migrate (prototype attribute
inference and a few statistics) and simplify-libcalls can finally be put
to rest.
llvm-svn: 168925
MachOObjectFile owns a MachOObj, but never frees it. Both MachOObjectFile
and MachOObj want to own the MemoryBuffer, though, so we have to be careful
and give them each one of their own.
Thanks to Greg Clayton, Eric Christopher and Michael Spencer for helping
figure out what's going wrong here.
rdar://12561773
llvm-svn: 168923
The old version failed on a 3-arg instruction with (-1, 0, 0) shadows (it would
pick the 3rd operand origin irrespective of its shadow).
The new version always picks the origin of the rightmost poisoned operand.
llvm-svn: 168887
This was found by MSVC10's STL debug mode on a test from the test suite. Sadly
std::is_heap isn't standard so there is no way to assert this without writing
our own heap verify, which looks like overkill to me.
llvm-svn: 168885
If we need to split the operand of a VSELECT, it must be the mask operand. We
split the entire VSELECT operand with EXTRACT_SUBVECTOR.
llvm-svn: 168883
Rewrite getOriginPtr in a way that lets subsequent optimizations factor out
the common part of Shadow and Origin address calculation. Improves perf by
up to 5%.
llvm-svn: 168879
This was already done for memmove, where it is required for correctness.
This change improves performance by avoiding copyingthe same memory twice.
Also, the library functions are given __msan_ prefix to prevent instcombine
pass from converting them back to intrinsics.
llvm-svn: 168876
Also a couple not-user-visible changes; using empty() instead of size(), and
make inSection() not insert NULL Regex*'s into StringMap when doing a lookup.
llvm-svn: 168833
No functional change, just moved header files.
Targets can inject custom passes between register allocation and
rewriting. This makes it possible to tweak the register allocation
before rewriting, using the full global interference checking available
from LiveRegMatrix.
llvm-svn: 168806
appropriate unit tests. This change in itself is not expected to
affect any functionality at this point, but it will serve as a
stepping stone to improve FileCheck's variable matching capabilities.
Luckily, our regex implementation already supports backreferences,
although a bit of hacking is required to enable it. It supports both
Basic Regular Expressions (BREs) and Extended Regular Expressions
(EREs), without supporting backrefs for EREs, following POSIX strictly
in this respect. And EREs is what we actually use (rightly). This is
contrary to many implementations (including the default on Linux) of
POSIX regexes, that do allow backrefs in EREs.
Adding backref support to our EREs is a very simple change in the
regcomp parsing code. I fail to think of significant cases where it
would clash with existing things, and can bring more versatility to
the regexes we write. There's always the danger of a backref in a
specially crafted regex causing exponential matching times, but since
we mainly use them for testing purposes I don't think it's a big
problem. [it can also be placed behind a flag specific to FileCheck,
if needed].
For more details, see:
* http://lists.cs.uiuc.edu/pipermail/llvmdev/2012-November/055840.html
* http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20121126/156878.html
llvm-svn: 168802
The createPPCMCAsmInfo routine used PPC::R1 as the initial frame
pointer register, but on PPC64 the 32-bit R1 register does not
have a corresponding DWARF number, causing invalid CIE initial
frame state to be emitted. Fix by using PPC::X1 instead.
llvm-svn: 168799
Accordingly, update a testcase with a broken datalayout string.
Also, we never parse negative numbers, because '-' is used as a
separator. Therefore, use unsigned as result type.
llvm-svn: 168785
This is a simple, cheap infrastructure for analyzing the shape of a
DAG. It recognizes uniform DAGs that take the shape of bottom-up
subtrees, such as the included matrix multiplication example. This is
useful for heuristics that balance register pressure with ILP. Two
canonical expressions of the heuristic are implemented in scheduling
modes: -misched-ilpmin and -misched-ilpmax.
llvm-svn: 168773
This fixes a hole in the "cheap" alias analysis logic implemented within
the DAG builder itself, regardless of whether proper alias analysis is
enabled. It now handles this pattern produced by LSR+CodeGenPrepare.
%sunkaddr1 = ptrtoint * %obj to i64
%sunkaddr2 = add i64 %sunkaddr1, %lsr.iv
%sunkaddr3 = inttoptr i64 %sunkaddr2 to i32*
store i32 %v, i32* %sunkaddr3
llvm-svn: 168768
The Target library is not allowed to depend on the large CodeGen
library, but the TRI and TII classes provide abstract interfaces that
require both caller and callee to link to CodeGen.
The implementation files for these classes provide default
implementations of some of the hooks. These methods may need to
reference CodeGen, so they belong in that library.
We already have a number of methods implemented in the
TargetInstrInfoImpl sub-class because of that. I will merge that class
into the parent next.
llvm-svn: 168758
When the CodeGenInfo is to be created for the PPC64 target machine,
a default code-model selection is converted to CodeModel::Medium
provided we are not targeting the Darwin OS. Defaults for Darwin
are unaffected.
llvm-svn: 168747
classes. The vast majority of the remaining issues are due to uses of
invalid registers, which are defined by getRegForValue(). Those will be
a little more challenging to cleanup.
rdar://12719844
llvm-svn: 168735
The functionality of SectionMemoryManager is equivalent to the LLIMCJITMemoryManager being replaced except that it allocates memory as RW and later changes it to RX or R as needed. The page permissions are set in the call to MCJIT::finalizeObject.
llvm-svn: 168722
The SectionMemoryManager now supports (and requires) applying section-specific page permissions. Clients using this memory manager must call either MCJIT::finalizeObject() or SectionMemoryManager::applyPermissions() before executing JITed code.
See r168718 for changes from the previous implementation.
llvm-svn: 168721
This commit is primarily here for the revision history. I'm about to move the SectionMemoryManager into the RuntimeDyld library, but I wanted to check the changes in here so people could see the differences in the updated implementation.
llvm-svn: 168718
boundaries.
Given the following case:
BB0
%vreg1<def> = SUBrr %vreg0, %vreg7
%vreg2<def> = COPY %vreg7
BB1
%vreg10<def> = SUBrr %vreg0, %vreg2
We should be able to CSE between SUBrr in BB0 and SUBrr in BB1.
rdar://12462006
llvm-svn: 168717
My commit to migrate the printf simplifiers from the simplify-libcalls
in r168604 introduced a regression reported by Duncan [1]. The problem
is that in some cases the library call simplifier can return a new value
that has no uses and the new value's type is different than the old value's
type (which is fine because there are no uses). The specific case that
triggered the bug looked something like:
declare void @printf(i8*, ...)
...
call void (i8*, ...)* @printf(i8* %fmt)
Which we want to optimized into:
call i32 @putchar(i32 104)
However, the code was attempting to replace all uses of the printf with
the putchar and the types differ, hence a crash. This is fixed by *just*
deleting the original instruction when there are no uses. The old
simplify-libcalls pass is already doing something similar.
[1] http://lists.cs.uiuc.edu/pipermail/llvmdev/2012-November/056338.html
llvm-svn: 168716
when the destination register is wider than the memory load.
These load instructions load from m32 or m64 and set the upper bits to zero,
while the folded instructions may accept m128.
rdar://12721174
llvm-svn: 168710
The default for 64-bit PowerPC is small code model, in which TOC entries
must be addressable using a 16-bit offset from the TOC pointer. Additionally,
only TOC entries are addressed via the TOC pointer.
With medium code model, TOC entries and data sections can all be addressed
via the TOC pointer using a 32-bit offset. Cooperation with the linker
allows 16-bit offsets to be used when these are sufficient, reducing the
number of extra instructions that need to be executed. Medium code model
also does not generate explicit TOC entries in ".section toc" for variables
that are wholly internal to the compilation unit.
Consider a load of an external 4-byte integer. With small code model, the
compiler generates:
ld 3, .LC1@toc(2)
lwz 4, 0(3)
.section .toc,"aw",@progbits
.LC1:
.tc ei[TC],ei
With medium model, it instead generates:
addis 3, 2, .LC1@toc@ha
ld 3, .LC1@toc@l(3)
lwz 4, 0(3)
.section .toc,"aw",@progbits
.LC1:
.tc ei[TC],ei
Here .LC1@toc@ha is a relocation requesting the upper 16 bits of the
32-bit offset of ei's TOC entry from the TOC base pointer. Similarly,
.LC1@toc@l is a relocation requesting the lower 16 bits. Note that if
the linker determines that ei's TOC entry is within a 16-bit offset of
the TOC base pointer, it will replace the "addis" with a "nop", and
replace the "ld" with the identical "ld" instruction from the small
code model example.
Consider next a load of a function-scope static integer. For small code
model, the compiler generates:
ld 3, .LC1@toc(2)
lwz 4, 0(3)
.section .toc,"aw",@progbits
.LC1:
.tc test_fn_static.si[TC],test_fn_static.si
.type test_fn_static.si,@object
.local test_fn_static.si
.comm test_fn_static.si,4,4
For medium code model, the compiler generates:
addis 3, 2, test_fn_static.si@toc@ha
addi 3, 3, test_fn_static.si@toc@l
lwz 4, 0(3)
.type test_fn_static.si,@object
.local test_fn_static.si
.comm test_fn_static.si,4,4
Again, the linker may replace the "addis" with a "nop", calculating only
a 16-bit offset when this is sufficient.
Note that it would be more efficient for the compiler to generate:
addis 3, 2, test_fn_static.si@toc@ha
lwz 4, test_fn_static.si@toc@l(3)
The current patch does not perform this optimization yet. This will be
addressed as a peephole optimization in a later patch.
For the moment, the default code model for 64-bit PowerPC will remain the
small code model. We plan to eventually change the default to medium code
model, which matches current upstream GCC behavior. Note that the different
code models are ABI-compatible, so code compiled with different models will
be linked and execute correctly.
I've tested the regression suite and the application/benchmark test suite in
two ways: Once with the patch as submitted here, and once with additional
logic to force medium code model as the default. The tests all compile
cleanly, with one exception. The mandel-2 application test fails due to an
unrelated ABI compatibility with passing complex numbers. It just so happens
that small code model was incredibly lucky, in that temporary values in
floating-point registers held the expected values needed by the external
library routine that was called incorrectly. My current thought is to correct
the ABI problems with _Complex before making medium code model the default,
to avoid introducing this "regression."
Here are a few comments on how the patch works, since the selection code
can be difficult to follow:
The existing logic for small code model defines three pseudo-instructions:
LDtoc for most uses, LDtocJTI for jump table addresses, and LDtocCPT for
constant pool addresses. These are expanded by SelectCodeCommon(). The
pseudo-instruction approach doesn't work for medium code model, because
we need to generate two instructions when we match the same pattern.
Instead, new logic in PPCDAGToDAGISel::Select() intercepts the TOC_ENTRY
node for medium code model, and generates an ADDIStocHA followed by either
a LDtocL or an ADDItocL. These new node types correspond naturally to
the sequences described above.
The addis/ld sequence is generated for the following cases:
* Jump table addresses
* Function addresses
* External global variables
* Tentative definitions of global variables (common linkage)
The addis/addi sequence is generated for the following cases:
* Constant pool entries
* File-scope static global variables
* Function-scope static variables
Expanding to the two-instruction sequences at select time exposes the
instructions to subsequent optimization, particularly scheduling.
The rest of the processing occurs at assembly time, in
PPCAsmPrinter::EmitInstruction. Each of the instructions is converted to
a "real" PowerPC instruction. When a TOC entry needs to be created, this
is done here in the same manner as for the existing LDtoc, LDtocJTI, and
LDtocCPT pseudo-instructions (I factored out a new routine to handle this).
I had originally thought that if a TOC entry was needed for LDtocL or
ADDItocL, it would already have been generated for the previous ADDIStocHA.
However, at higher optimization levels, the ADDIStocHA may appear in a
different block, which may be assembled textually following the block
containing the LDtocL or ADDItocL. So it is necessary to include the
possibility of creating a new TOC entry for those two instructions.
Note that for LDtocL, we generate a new form of LD called LDrs. This
allows specifying the @toc@l relocation for the offset field of the LD
instruction (i.e., the offset is replaced by a SymbolLo relocation).
When the peephole optimization described above is added, we will need
to do similar things for all immediate-form load and store operations.
The seven "mcm-n.ll" test cases are kept separate because otherwise the
intermingling of various TOC entries and so forth makes the tests fragile
and hard to understand.
The above assumes use of an external assembler. For use of the
integrated assembler, new relocations are added and used by
PPCELFObjectWriter. Testing is done with "mcm-obj.ll", which tests for
proper generation of the various relocations for the same sequences
tested with the external assembler.
llvm-svn: 168708
argument. Instead, use a pair of .local and .comm directives.
This avoids spurious differences between binaries built by the
integrated assembler vs. those built by the external assembler,
since the external assembler may impose alignment requirements
on .lcomm symbols where the integrated assembler does not.
llvm-svn: 168704
If the Src and Dst are the same instruction,
no loop-independent dependence is possible,
so we force the PossiblyLoopIndependent flag to false.
The test case results are updated appropriately.
llvm-svn: 168678
Added in first optimization using fast-math flags to serve as an example for following optimizations. SimplifyInstruction will now try to optimize an fmul observing its FastMathFlags to see if it can fold multiply by zero when 'nnan' and 'nsz' flags are set.
llvm-svn: 168648
Added in bitcode enum for the serializing of fast-math flags. Added in the reading/writing of fast-math flags from the OptimizationFlags record for BinaryOps.
llvm-svn: 168646
Added in the ability to read LLVM IR text that contains fast-math flags as a sequence of capital letters separated by spaces in any order. Added in the printing of the fast-math flags in a canonical order, and don't print the other flags when 'fast' is specified, as 'fast' implies all the others.
llvm-svn: 168645
Add in getter/setter methods for Instructions, allowing them to be the interface to FPMathOperator similarly to now NUS/NSW is handled.
llvm-svn: 168642
Created FastMathFlags convenience struct for the getting and setting of fast-math flags en masse. Added SubclassOptionalData bitfields and corresponding getters/setters to FPMathOperator for the various fast-math flags.
llvm-svn: 168641
It currently assumes register numbering and any harmless change in the X86
register naming makes it fail. It's enough to match the register names.
llvm-svn: 168632
r168627), we no longer need to call the freezeReservedRegs() function a second
time. Previously, this pass was conservatively adding the FP to the set of
reserved registers, requiring the second update to the reserved registers.
rdar://12719844
llvm-svn: 168631
r168627), we no longer need to call the freezeReservedRegs() function a second
time. Previously, this pass was conservatively adding the FP to the set of
reserved registers, requiring the second update to the reserved registers.
rdar://12719844
llvm-svn: 168630
InstCombineLoadStoreAlloca.cpp, which had many issues.
(At least two bugs were noted on llvm-commits, and it was overly conservative.)
Instead, use getOrEnforceKnownAlignment.
llvm-svn: 168629
This pass was conservative in that it always reserved the FP to enable dynamic
stack realignment, which allowed the RA to use aligned spills for vector
registers. This happens even when spills were not necessary. The RA has
since been improved to use unaligned spills when necessary.
The new behavior is to realign the stack if the frame pointer was already
reserved for some other reason, but don't reserve the frame pointer just
because a function contains vector virtual registers.
Part of rdar://12719844
llvm-svn: 168627
Enhancement to InstCombine. Try to catch this opportunity:
---------------------------------------------------------------
((X^C1) >> C2) ^ C3 => (X>>C2) ^ ((C1>>C2)^C3)
where the subexpression "X ^ C1" has more than one uses, and
"(X^C1) >> C2" has single use.
----------------------------------------------------------------
Reviewed by Nadav (with minor change per his request).
llvm-svn: 168615
In preparation for the FileCheck functionality change which will allow using
a variable later on the same line.
No functionality change.
llvm-svn: 168588
The bug can be triggered when we require LoopInfo analysis ahead of DominatorTree construction in a Module Pass. The cause is that the LoopInfo analysis itself also invokes DominatorTree construction, therefore, when PassManager schedules LoopInfo, it will add DominatorTree first. Then after that, when the PassManger turns to schedule DominatorTree invoked by the above ModulePass, it finds there is already a DominatorTree, so it delete the redundant one. However, somehow it still try to access that pass pointer after free as code pasted below, which results in segment fault.
llvm-svn: 168581
Shuxin Yang