The ELF format is used on Windows by the MCJIT engine. Thus, on Windows, the
ELFObjectWriter can encounter symbols mangled using the MS Visual Studio C++
name mangling. Symbols mangled using the MSVC C++ name mangling can legally
have "@@@" as a substring. The EFLObjectWriter should not interpret the "@@@"
substring as specifying GNU-style symbol versioning. The ELFObjectWriter
therefore check for the MSVC C++ name mangling prefix which is either "?", "@?",
"imp_?" or "imp_?@".
llvm-svn: 226830
This solves PR22276.
Splats of constants would sometimes produce redundant shuffles, sometimes ridiculously so (see the PR for details). Fold these shuffles into BUILD_VECTORs early on instead.
Differential Revision: http://reviews.llvm.org/D7093
Fixed recommit of r226811.
llvm-svn: 226816
This solves PR22276.
Splats of constants would sometimes produce redundant shuffles, sometimes ridiculously so (see the PR for details). Fold these shuffles into BUILD_VECTORs early on instead.
Differential Revision: http://reviews.llvm.org/D7093
llvm-svn: 226811
The problem occurs when after vectorization we have type
<2 x i32>. This type is promoted to <2 x i64> and then requires
additional efforts for expanding loads and truncating stores.
I added EXPAND / TRUNCATE attributes to the masked load/store
SDNodes. The code now contains additional shuffles.
I've prepared changes in the cost estimation for masked memory
operations, it will be submitted separately.
llvm-svn: 226808
Type MVT::i1 became legal in KNL, but store operation can't be narrowed to this type,
since the size of VT (1 bit) is not equal to its actual store size(8 bits).
Added a test provided by David (dag@cray.com)
llvm-svn: 226805
There are places where the inductive range check elimination pass
depends on two llvm::Values or llvm::SCEVs to be of the same
llvm::Type when they do not need to be. This patch relaxes those
restrictions (by bailing out of the optimization if the types
mismatch), and adds test cases to trigger those paths.
These issues were found by bootstrapping clang with IRCE running in
the -O3 pass ordering.
Differential Revision: http://reviews.llvm.org/D7082
llvm-svn: 226793
ever stored to always use a legal integer type if one is available.
Regardless of whether this particular type is good or bad, it ensures we
don't get weird differences in generated code (and resulting
performance) from "equivalent" patterns that happen to end up using
a slightly different type.
After some discussion on llvmdev it seems everyone generally likes this
canonicalization. However, there may be some parts of LLVM that handle
it poorly and need to be fixed. I have at least verified that this
doesn't impede GVN and instcombine's store-to-load forwarding powers in
any obvious cases. Subtle cases are exactly what we need te flush out if
they remain.
Also note that this IR pattern should already be hitting LLVM from Clang
at least because it is exactly the IR which would be produced if you
used memcpy to copy a pointer or floating point between memory instead
of a variable.
llvm-svn: 226781
Windows supports a restricted set of relocations (compared to ARM ELF). In some
cases, we may end up generating an unsupported relocation. This can occur with
bad input to the assembler in particular (the frontend should never generate
code that cannot be compiled). Generate an error rather than just aborting.
The change in the API is driven by the desire to provide a slightly more helpful
message for debugging purposes.
llvm-svn: 226779
ScalarEvolution currently lowers a subtraction recurrence to an add
recurrence with the same no-wrap flags as the subtraction. This is
incorrect because `sub nsw X, Y` is not the same as `add nsw X, -Y`
and `sub nuw X, Y` is not the same as `add nuw X, -Y`. This patch
fixes the issue, and adds two test cases demonstrating the bug.
Differential Revision: http://reviews.llvm.org/D7081
llvm-svn: 226755
Added most of the missing integer vector folding patterns for SSE (to SSE42) and AVX1.
The most useful of these are probably the i32/i64 extraction, i8/i16/i32/i64 insertions, zero/sign extension, unsigned saturation subtractions, i64 subtractions and the variable mask blends (pblendvb) - others include CLMUL, SSE42 string comparisons and bit tests.
Differential Revision: http://reviews.llvm.org/D7094
llvm-svn: 226745
When two calls from the same MDLocation are inlined they currently get
treated as one inlined function call (creating difficulty debugging,
duplicate variables, etc).
Clang worked around this by including column information on inline calls
which doesn't address LTO inlining or calls to the same function from
the same line and column (such as through a macro). It also didn't
address ctor and member function calls.
By making the inlinedAt locations distinct, every call site has an
explicitly distinct location that cannot be coalesced with any other
call.
This can produce linearly (2x in the worst case where every call is
inlined and the call instruction has a non-call instruction at the same
location) more debug locations. Any increase beyond that are in cases
where the Clang workaround was insufficient and the new scheme is
creating necessary distinct nodes that were being erroneously coalesced
previously.
After this change to LLVM the incomplete workarounds in Clang. That
should reduce the number of debug locations (in a build without column
info, the default on Darwin, not the default on Linux) by not creating
pseudo-distinct locations for every call to an inline function.
(oh, and I made the inlined-at chain rebuilding iterative instead of
recursive because I was having trouble wrapping my head around it the
way it was - open to discussion on the right design for that function
(including going back to a recursive solution))
llvm-svn: 226736
This patch adds shuffle matching for the SSE3 MOVDDUP, MOVSLDUP and MOVSHDUP instructions. The big use of these being that they avoid many single source shuffles from needing to use (pre-AVX) dual source instructions such as SHUFPD/SHUFPS: causing extra moves and preventing load folds.
Adding these instructions uncovered an issue in XFormVExtractWithShuffleIntoLoad which crashed on single operand shuffle instructions (now fixed). It also involved fixing getTargetShuffleMask to correctly identify theses instructions as unary shuffles.
Also adds a missing tablegen pattern for MOVDDUP.
Differential Revision: http://reviews.llvm.org/D7042
llvm-svn: 226716
Thumbv4t does not have lo->lo copies other than MOVS,
and that can't be predicated. So emit MOVS when needed
and bail if there's a predicate.
http://reviews.llvm.org/D6592
llvm-svn: 226711
This fixes it for SI. It also removes the pattern
used previously for Evergreen for f32. I'm not sure
if the the new R600 output is better or not, but it uses
1 fewer instructions if BFI is available.
llvm-svn: 226682
Now that we can fully specify extload legality, we can declare them
legal for the PMOVSX/PMOVZX instructions. This for instance enables
a DAGCombine to fire on code such as
(and (<zextload-equivalent> ...), <redundant mask>)
to turn it into:
(zextload ...)
as seen in the testcase changes.
There is one regression, in widen_load-2.ll: we're no longer able
to do store-to-load forwarding with illegal extload memory types.
This will be addressed separately.
Differential Revision: http://reviews.llvm.org/D6533
llvm-svn: 226676
AAPCS64 says that it's up to the platform to specify whether x18 is
reserved, and a first step on that way is to add a flag controlling
it.
From: Andrew Turner <andrew@fubar.geek.nz>
llvm-svn: 226664
Previously we always stored 4 bytes of origin at the destination address
even for 8-byte (and longer) stores.
This should fix rare missing, or incorrect, origin stacks in MSan reports.
llvm-svn: 226658
Implement microMIPS 16-bit unconditional branch instruction B.
Implemented 16-bit microMIPS unconditional instruction has real name B16, and
B is an alias which expands to either B16 or BEQ according to the rules:
b 256 --> b16 256 # R_MICROMIPS_PC10_S1
b 12256 --> beq $zero, $zero, 12256 # R_MICROMIPS_PC16_S1
b label --> beq $zero, $zero, label # R_MICROMIPS_PC16_S1
Differential Revision: http://reviews.llvm.org/D3514
llvm-svn: 226657
Changed the AVX1 tests register spill tail call to return a xmm like the SSE42 version - makes doing diffs between them a lot easier without affecting the spills themselves.
llvm-svn: 226623
The SSE42 version of the AVX1 float stack folding tests will be added shortly, this renames the AVX1 file so that the files will be near each other in a directory listing to help ensure they are kept in sync.
llvm-svn: 226620
This addresses part of llvm.org/PR22262. Specifically, it prevents
considering the densities of sub-ranges that have fewer than
TLI.getMinimumJumpTableEntries() elements. Those densities won't help
jump tables.
This is not a complete solution but works around the most pressing
issue.
Review: http://reviews.llvm.org/D7070
llvm-svn: 226600
With the appropriate Verifier changes, exactracting the result out of a
statepoint wrapping a vararg function crashes. However, a void vararg
function works fine: commit this first step.
Differential Revision: http://reviews.llvm.org/D7071
llvm-svn: 226599
This reapplies r225379.
ChangeLog:
- The assertion that this commit previously ran into about the inability
to handle indirect variables has since been removed and the backend
can handle this now.
- Testcases were upgrade to the new MDLocation format.
- Instead of keeping a DebugDeclares map, we now use
llvm::FindAllocaDbgDeclare().
Original commit message follows.
Debug info: Teach SROA how to update debug info for fragmented variables.
This allows us to generate debug info for extremely advanced code such as
typedef struct { long int a; int b;} S;
int foo(S s) {
return s.b;
}
which at -O1 on x86_64 is codegen'd into
define i32 @foo(i64 %s.coerce0, i32 %s.coerce1) #0 {
ret i32 %s.coerce1, !dbg !24
}
with this patch we emit the following debug info for this
TAG_formal_parameter [3]
AT_location( 0x00000000
0x0000000000000000 - 0x0000000000000006: rdi, piece 0x00000008, rsi, piece 0x00000004
0x0000000000000006 - 0x0000000000000008: rdi, piece 0x00000008, rax, piece 0x00000004 )
AT_name( "s" )
AT_decl_file( "/Volumes/Data/llvm/_build.ninja.release/test.c" )
Thanks to chandlerc, dblaikie, and echristo for their feedback on all
previous iterations of this patch!
llvm-svn: 226598
We don't have a good way of legalizing this if the frame index offset
is more than the 12-bits, which is size of MUBUF's offset field, so
now we store the frame index in the vaddr field.
llvm-svn: 226584
Implement microMIPS 16-bit unconditional branch instruction B.
Implemented 16-bit microMIPS unconditional instruction has real name B16, and
B is an alias which expands to either B16 or BEQ according to the rules:
b 256 --> b16 256 # R_MICROMIPS_PC10_S1
b 12256 --> beq $zero, $zero, 12256 # R_MICROMIPS_PC16_S1
b label --> beq $zero, $zero, label # R_MICROMIPS_PC16_S1
Differential Revision: http://reviews.llvm.org/D3514
llvm-svn: 226577
This commits adds the octeon branch instructions bbit0/bbit032/bbit1/bbit132.
It also includes patterns for instruction selection and test cases.
Reviewed by D. Sanders
llvm-svn: 226573
The new code does not create new basic blocks in the case when shadow is a
compile-time constant; it generates either an unconditional __msan_warning
call or nothing instead.
llvm-svn: 226569
pass and a LoopPrinterPass with the expected associated wiring.
I've added a RUN line to the only test case (!!!) we have that actually
prints loops. Everything seems to be working.
This is somewhat exciting as this is the first analysis using another
analysis to go in for the new pass manager. =D I also believe it is the
last analysis necessary for porting instcombine, but of course I may yet
discover more.
llvm-svn: 226560
This patch fixes 2 issues in reorderInputsAccordingToOpcode
1) AllSameOpcodeLeft and AllSameOpcodeRight was being calculated incorrectly resulting in code not being vectorized in few cases.
2) Adds logic to reorder operands if we get longer chain of consecutive loads enabling vectorization. Handled the same for cases were we have AltOpcode.
Thanks Michael for inputs and review.
Review: http://reviews.llvm.org/D6677
llvm-svn: 226547
Now that we can create much more exhaustive X86 memory folding tests, this patch adds the missing AVX1/F16C floating point instruction stack foldings we can easily test for including the scalar intrinsics (add, div, max, min, mul, sub), conversions float/int to double, half precision conversions, rounding, dot product and bit test. The patch also adds a couple of obviously missing SSE instructions (more to follow once we have full SSE testing).
Now that scalar folding is working it broke a very old test (2006-10-07-ScalarSSEMiscompile.ll) - this test appears to make no sense as its trying to ensure that a scalar subtraction isn't folded as it 'would zero the top elts of the loaded vector' - this test just appears to be wrong to me.
Differential Revision: http://reviews.llvm.org/D7055
llvm-svn: 226513
The fixes are to note that AArch64 has additional restrictions on when local
relocations can be used. In particular, ld64 requires that relocations to
cstring/cfstrings use linker visible symbols.
Original message:
In an assembly expression like
bar:
.long L0 + 1
the intended semantics is that bar will contain a pointer one byte past L0.
In sections that are merged by content (strings, 4 byte constants, etc), a
single position in the section doesn't give the linker enough information.
For example, it would not be able to tell a relocation must point to the
end of a string, since that would look just like the start of the next.
The solution used in ELF to use relocation with symbols if there is a non-zero
addend.
In MachO before this patch we would just keep all symbols in some sections.
This would miss some cases (only cstrings on x86_64 were implemented) and was
inefficient since most relocations have an addend of 0 and can be represented
without the symbol.
This patch implements the non-zero addend logic for MachO too.
llvm-svn: 226503
frontends to use a DIExpression with a DW_OP_deref instead.
This is not only a much more natural place for this informationl; there
is also a technical reason: The FlagIndirectVariable is used to mark a
variable that is turned into a reference by virtue of the calling
convention; this happens for example to aggregate return values.
The inliner, for example, may actually need to undo this indirection to
correctly represent the value in its new context. This is impossible to
implement because the DIVariable can't be safely modified. We can however
safely construct a new DIExpression on the fly.
llvm-svn: 226476
Original patch by Luke Iannini. Minor improvements and test added by
Erik de Castro Lopo.
Differential Revision: http://reviews.llvm.org/D6877
From: Erik de Castro Lopo <erikd@mega-nerd.com>
llvm-svn: 226473
An assignment will produce a symbol with a given section and offset. There is
no way to represent something like "1 byte after a common symbol".
This matches the behavior of GNU as.
Part of PR22217.
llvm-svn: 226470
No change in this commit, but clang was changed to also produce trivial comdats when
needed.
Original message:
Don't create new comdats in CodeGen.
This patch stops the implicit creation of comdats during codegen.
Clang now sets the comdat explicitly when it is required. With this patch clang and gcc
now produce the same result in pr19848.
llvm-svn: 226467
Our PPC64 ELF V2 call lowering logic added r2 as an operand to all direct call
instructions in order to represent the dependency on the TOC base pointer
value. Restricting this to ELF V2, however, does not seem to make sense: calls
under ELF V1 have the same dependence, and indirect calls have an r2 dependence
just as direct ones. Make sure the dependence is noted for all calls under both
ELF V1 and ELF V2.
llvm-svn: 226432
The tests for the ISA's should now be approximately diffable. That is, the
output of 'diff valid-mips1.txt valid-mips2.txt' should be emit the lines
for instructions that were added/removed to/from MIPS-I by MIPS-II. This
doesn't work perfectly at the moment due to ordering differences but it
should be close.
llvm-svn: 226408
Begun adding more exhaustive tests - all floating point instructions should now be either tested or have placeholders. We do seem to have a number of missing instructions, I will add a patch for review once the remaining working instructions are added.
I'll then move on to SSE tests and then the integer instructions.
llvm-svn: 226400
The default calling convention specified by the PPC64 ELF (V1 and V2) ABI is
designed to work with both prototyped and non-prototyped/varargs functions. As
a result, GPRs and stack space are allocated for every argument, even those
that are passed in floating-point or vector registers.
GlobalOpt::OptimizeFunctions will transform local non-varargs functions (that
do not have their address taken) to use the 'fast' calling convention.
When functions are using the 'fast' calling convention, don't allocate GPRs for
arguments passed in other types of registers, and don't allocate stack space for
arguments passed in registers. Other changes for the fast calling convention
may be added in the future.
llvm-svn: 226399
R11's status is the same under both the PPC64 ELF V1 and V2 ABIs: it is
reserved for use as an "environment pointer" for compilation models that
require such a thing. We don't, we also don't need a second scratch register,
and because we support only "local" patchpoint call targets, we might as well
let R11 be used for anyregcc patchpoints.
llvm-svn: 226369
Loading 2 2x32-bit float vectors into the bottom half of a 256-bit vector
produced suboptimal code in AVX2 mode with certain IR combinations.
In particular, the IR optimizer folded 2f32 + 2f32 -> 4f32, 4f32 + 4f32
(undef) -> 8f32 into a 2f32 + 2f32 -> 8f32, which seems more canonical,
but then mysteriously generated rather bad code; the movq/movhpd combination
didn't match.
The problem lay in the BUILD_VECTOR optimization path. The 2f32 inputs
would get promoted to 4f32 by the type legalizer, eventually resulting
in a BUILD_VECTOR on two 4f32 into an 8f32. The BUILD_VECTOR then, recognizing
these were both half the output size, concatted them and then produced
a shuffle. However, the resulting concat + shuffle was more complex than
it should be; in the case where the upper half of the output is undef, we
probably want to generate shuffle + concat instead.
This enhancement causes the vector_shuffle combine step to recognize this
suboptimal pattern and correct it. I included it there instead of in BUILD_VECTOR
in case the same suboptimal pattern occurs for other reasons.
This results in the optimizer correctly producing the optimal movq + movhpd
sequence for all three variations on this IR, even with AVX2.
I've included a test case.
Radar link: rdar://problem/19287012
Fix for PR 21943.
From: Fiona Glaser <fglaser@apple.com>
llvm-svn: 226360
This patch disables target specific combine on X86ISD::INSERTPS dag nodes
if optlevel is CodeGenOpt::None.
The backend currently implements a target specific combine rule that converts
a vector load used by an INSERTPS dag node into a scalar load plus a
scalar_to_vector. This allows ISel to select a single INSERTPSrm instead of
two instructions (i.e. a vector load plus INSERTPSrr).
However, the existing target combine rule on INSERTPS nodes only works under
the assumption that ISel will always be able to match an INSERTPSrm. This is
not true in general at -O0, since the backend only allows folding a load into
the memory operand of an instruction if the optimization level is not
CodeGenOpt::None.
In the example below:
//
__m128 test(__m128 a, __m128 *b) {
__m128 c = _mm_insert_ps(a, *b, 1 << 6);
return c;
}
//
Before this patch, at -O0, the backend would have canonicalized the load to 'b'
into a scalar load plus scalar_to_vector. Later on, ISel would have selected an
INSERTPSrr leaving the insertps mask in an inconsistent state:
movss 4(%rdi), %xmm1
insertps $64, %xmm1, %xmm0 # xmm0 = xmm1[1],xmm0[1,2,3].
With this patch, the backend avoids folding the vector load into the operand of
the INSERTPS. The new codegen at -O0 is:
movaps (%rdi), %xmm1
insertps $64, %xmm1, %xmm0 # %xmm1[1],xmm0[1,2,3].
llvm-svn: 226277
The current 'big vectors' stack folded reload testing pattern is very bulky and makes it difficult to test all instructions as big vectors will tend to use only the ymm instruction implementations.
This patch changes the tests to use a nop call that lists explicit xmm registers as sideeffects, with this we can force a partial register spill of the relevant registers and then check that the reload is correctly folded. The asm generated only adds the forced spill, a nop instruction and a couple of extra labels (a fraction of the current approach).
More exhaustive tests will follow shortly, I've added some extra tests (the xmm versions of some of the existing folding tests) as a starting point.
Differential Revision: http://reviews.llvm.org/D6932
llvm-svn: 226264
Bill Schmidt pointed out that some adjustments would be needed to properly
support powerpc64le (using the ELF V2 ABI). For one thing, R11 is not available
as a scratch register, so we need to use R12. R12 is also available under ELF
V1, so to maintain consistency, I flipped the order to make R12 the first
scratch register in the array under both ABIs.
llvm-svn: 226247
This reverts commit r226173, adding r226038 back.
No change in this commit, but clang was changed to also produce trivial comdats for
costructors, destructors and vtables when needed.
Original message:
Don't create new comdats in CodeGen.
This patch stops the implicit creation of comdats during codegen.
Clang now sets the comdat explicitly when it is required. With this patch clang and gcc
now produce the same result in pr19848.
llvm-svn: 226242
IRCE eliminates range checks of the form
0 <= A * I + B < Length
by splitting a loop's iteration space into three segments in a way
that the check is completely redundant in the middle segment. As an
example, IRCE will convert
len = < known positive >
for (i = 0; i < n; i++) {
if (0 <= i && i < len) {
do_something();
} else {
throw_out_of_bounds();
}
}
to
len = < known positive >
limit = smin(n, len)
// no first segment
for (i = 0; i < limit; i++) {
if (0 <= i && i < len) { // this check is fully redundant
do_something();
} else {
throw_out_of_bounds();
}
}
for (i = limit; i < n; i++) {
if (0 <= i && i < len) {
do_something();
} else {
throw_out_of_bounds();
}
}
IRCE can deal with multiple range checks in the same loop (it takes
the intersection of the ranges that will make each of them redundant
individually).
Currently IRCE does not do any profitability analysis. That is a
TODO.
Please note that the status of this pass is *experimental*, and it is
not part of any default pass pipeline. Having said that, I will love
to get feedback and general input from people interested in trying
this out.
This pass was originally r226201. It was reverted because it used C++
features not supported by MSVC 2012.
Differential Revision: http://reviews.llvm.org/D6693
llvm-svn: 226238
Summary:
Shift an older “invalid file” test to get a consistent naming for these tests.
Bugs found by afl-fuzz
Reviewers: rafael
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6945
llvm-svn: 226219
The change used C++11 features not supported by MSVC 2012. I will fix
the change to use things supported MSVC 2012 and recommit shortly.
llvm-svn: 226216
Function pointers under PPC64 ELFv1 (which is used on PPC64/Linux on the
POWER7, A2 and earlier cores) are really pointers to a function descriptor, a
structure with three pointers: the actual pointer to the code to which to jump,
the pointer to the TOC needed by the callee, and an environment pointer. We
used to chain these loads, and make them opaque to the rest of the optimizer,
so that they'd always occur directly before the call. This is not necessary,
and in fact, highly suboptimal on embedded cores. Once the function pointer is
known, the loads can be performed ahead of time; in fact, they can be hoisted
out of loops.
Now these function descriptors are almost always generated by the linker, and
thus the contents of the descriptors are invariant. As a result, by default,
we'll mark the associated loads as invariant (allowing them to be hoisted out
of loops). I've added a target feature to turn this off, however, just in case
someone needs that option (constructing an on-stack descriptor, casting it to a
function pointer, and then calling it cannot be well-defined C/C++ code, but I
can imagine some JIT-compilation system doing so).
Consider this simple test:
$ cat call.c
typedef void (*fp)();
void bar(fp x) {
for (int i = 0; i < 1600000000; ++i)
x();
}
$ cat main.c
typedef void (*fp)();
void bar(fp x);
void foo() {}
int main() {
bar(foo);
}
On the PPC A2 (the BG/Q supercomputer), marking the function-descriptor loads
as invariant brings the execution time down to ~8 seconds from ~32 seconds with
the loads in the loop.
The difference on the POWER7 is smaller. Compiling with:
gcc -std=c99 -O3 -mcpu=native call.c main.c : ~6 seconds [this is 4.8.2]
clang -O3 -mcpu=native call.c main.c : ~5.3 seconds
clang -O3 -mcpu=native call.c main.c -mno-invariant-function-descriptors : ~4 seconds
(looks like we'd benefit from additional loop unrolling here, as a first
guess, because this is faster with the extra loads)
The -mno-invariant-function-descriptors will be added to Clang shortly.
llvm-svn: 226207
IRCE eliminates range checks of the form
0 <= A * I + B < Length
by splitting a loop's iteration space into three segments in a way
that the check is completely redundant in the middle segment. As an
example, IRCE will convert
len = < known positive >
for (i = 0; i < n; i++) {
if (0 <= i && i < len) {
do_something();
} else {
throw_out_of_bounds();
}
}
to
len = < known positive >
limit = smin(n, len)
// no first segment
for (i = 0; i < limit; i++) {
if (0 <= i && i < len) { // this check is fully redundant
do_something();
} else {
throw_out_of_bounds();
}
}
for (i = limit; i < n; i++) {
if (0 <= i && i < len) {
do_something();
} else {
throw_out_of_bounds();
}
}
IRCE can deal with multiple range checks in the same loop (it takes
the intersection of the ranges that will make each of them redundant
individually).
Currently IRCE does not do any profitability analysis. That is a
TODO.
Please note that the status of this pass is *experimental*, and it is
not part of any default pass pipeline. Having said that, I will love
to get feedback and general input from people interested in trying
this out.
Differential Revision: http://reviews.llvm.org/D6693
llvm-svn: 226201
Reapply r226071 with fixes. Two fixes:
1. We need to manually remove the old and create the new 'deaf defs'
associated with physical register definitions when we move the definition of
the physical register from the copy point to the point of the original vreg def.
This problem was picked up by the machinstr verifier, and could trigger a
verification failure on test/CodeGen/X86/2009-02-12-DebugInfoVLA.ll, so I've
turned on the verifier in the tests.
2. When moving the def point of the phys reg up, we need to make sure that it
is neither defined nor read in between the two instructions. We don't, however,
extend the live ranges of phys reg defs to cover uses, so just checking for
live-range overlap between the pair interval and the phys reg aliases won't
pick up reads. As a result, we manually iterate over the range and check for
reads.
A test soon to be committed to the PowerPC backend will test this change.
Original commit message:
[RegisterCoalescer] Remove copies to reserved registers
This allows the RegisterCoalescer to join "non-flipped" range pairs with a
physical destination register -- which allows the RegisterCoalescer to remove
copies like this:
<vreg> = something (maybe a load, for example)
... (things that don't use PHYSREG)
PHYSREG = COPY <vreg>
(with all of the restrictions normally applied by the RegisterCoalescer: having
compatible register classes, etc. )
Previously, the RegisterCoalescer handled only the opposite case (copying
*from* a physical register). I don't handle the problem fully here, but try to
get the common case where there is only one use of <vreg> (the COPY).
An upcoming commit to the PowerPC backend will make this pattern much more
common on PPC64/ELF systems.
llvm-svn: 226200
TargetLibraryAnalysis pass.
There are actually no direct tests of this already in the tree. I've
added the most basic test that the pass manager bits themselves work,
and the TLI object produced will be tested by an upcoming patches as
they port passes which rely on TLI.
This is starting to point out the awkwardness of the invalidate API --
it seems poorly fitting on the *result* object. I suspect I will change
it to live on the analysis instead, but that's not for this change, and
I'd rather have a few more passes ported in order to have more
experience with how this plays out.
I believe there is only one more analysis required in order to start
porting instcombine. =]
llvm-svn: 226160
Reverting this while I investigate some bad behavior this is causing. As a
possibly-related issue, adding -verify-machineinstrs to one of the test cases
now fails because of this change:
llc test/CodeGen/X86/2009-02-12-DebugInfoVLA.ll -march=x86-64 -o - -verify-machineinstrs
*** Bad machine code: No instruction at def index ***
- function: foo
- basic block: BB#0 return (0x10007e21f10) [0B;736B)
- liverange: [128r,128d:9)[160r,160d:8)[176r,176d:7)[336r,336d:6)[464r,464d:5)[480r,480d:4)[624r,624d:3)[752r,752d:2)[768r,768d:1)[78
4r,784d:0) 0@784r 1@768r 2@752r 3@624r 4@480r 5@464r 6@336r 7@176r 8@160r 9@128r
- register: %DS
Valno #3 is defined at 624r
*** Bad machine code: Live segment doesn't end at a valid instruction ***
- function: foo
- basic block: BB#0 return (0x10007e21f10) [0B;736B)
- liverange: [128r,128d:9)[160r,160d:8)[176r,176d:7)[336r,336d:6)[464r,464d:5)[480r,480d:4)[624r,624d:3)[752r,752d:2)[768r,768d:1)[78
4r,784d:0) 0@784r 1@768r 2@752r 3@624r 4@480r 5@464r 6@336r 7@176r 8@160r 9@128r
- register: %DS
[624r,624d:3)
LLVM ERROR: Found 2 machine code errors.
where 624r corresponds exactly to the interval combining change:
624B %RSP<def> = COPY %vreg16; GR64:%vreg16
Considering merging %vreg16 with %RSP
RHS = %vreg16 [608r,624r:0) 0@608r
updated: 608B %RSP<def> = MOV64rm <fi#3>, 1, %noreg, 0, %noreg; mem:LD8[%saved_stack.1]
Success: %vreg16 -> %RSP
Result = %RSP
llvm-svn: 226086
The bug was introduced in r225282. r225282 assumed that sub X, Y is
the same as add X, -Y. This is not correct if we are going to upgrade
the sub to sub nuw. This change fixes the issue by making the
optimization ignore sub instructions.
Differential Revision: http://reviews.llvm.org/D6979
llvm-svn: 226075
This allows the RegisterCoalescer to join "non-flipped" range pairs with a
physical destination register -- which allows the RegisterCoalescer to remove
copies like this:
<vreg> = something (maybe a load, for example)
... (things that don't use PHYSREG)
PHYSREG = COPY <vreg>
(with all of the restrictions normally applied by the RegisterCoalescer: having
compatible register classes, etc. )
Previously, the RegisterCoalescer handled only the opposite case (copying
*from* a physical register). I don't handle the problem fully here, but try to
get the common case where there is only one use of <vreg> (the COPY).
An upcoming commit to the PowerPC backend will make this pattern much more
common on PPC64/ELF systems.
llvm-svn: 226071
Fill out our support for the floating-point status and control register
instructions (mcrfs and friends). As it turns out, these are necessary for
compiling src/test/harness_fp.h in TBB for PowerPC.
Thanks to Raf Schietekat for reporting the issue!
llvm-svn: 226070
it's defined in the current module. Clang generates this situation for the
C++14 sized deallocation functions, because it generates a weak definition in
case one isn't provided by the C++ runtime library.
llvm-svn: 226069
The transform is somewhat involved, but the basic idea is simple: find
derived pointers that have been offset from the base pointer using gep
and replace the relocate of the derived pointer with a gep to the
relocated base pointer (with the same offset).
llvm-svn: 226060
"Write a set of tests that show how name mangling is done for overloaded intrinsics." These happen to use gc.relocates to exercise the codepath in question, but is not a GC specific test.
Patch by: artagnon@gmail.com
Differential Revision: http://reviews.llvm.org/D6915
llvm-svn: 226056
This commit moves `MDLocation`, finishing off PR21433. There's an
accompanying clang commit for frontend testcases. I'll attach the
testcase upgrade script I used to PR21433 to help out-of-tree
frontends/backends.
This changes the schema for `DebugLoc` and `DILocation` from:
!{i32 3, i32 7, !7, !8}
to:
!MDLocation(line: 3, column: 7, scope: !7, inlinedAt: !8)
Note that empty fields (line/column: 0 and inlinedAt: null) don't get
printed by the assembly writer.
llvm-svn: 226048
This patch stops the implicit creation of comdats during codegen.
Clang now sets the comdat explicitly when it is required. With this patch clang and gcc
now produce the same result in pr19848.
llvm-svn: 226038
Some benchmarks have shown that this could lead to a potential
performance benefit, and so adding some flags to try to help measure the
difference.
A possible explanation. In diamond-shaped CFGs (A followed by either
B or C both followed by D), putting B and C both in between A and
D leads to the code being less dense than it could be. Always either
B or C have to be skipped increasing the chance of cache misses etc.
Moving either B or C to after D might be beneficial on average.
In the long run, but we should probably do a better job of analyzing the
basic block and branch probabilities to move the correct one of B or
C to after D. But even if we don't use this in the long run, it is
a good baseline for benchmarking.
Original patch authored by Daniel Jasper with test tweaks and a second
flag added by me.
Differential Revision: http://reviews.llvm.org/D6969
llvm-svn: 226034
Patch by Kit Barton.
Support for the ICBT instruction is currently present, but limited to
embedded processors. This change adds a new FeatureICBT that can be used
to identify whether the ICBT instruction is available on a specific processor.
Two new tests are added:
* Positive test to ensure the icbt instruction is present when using
-mcpu=pwr8
* Negative test to ensure the icbt instruction is not generated when
using -mcpu=pwr7
Both test cases use the Prefetch opcode in LLVM. They are based on the
ppc64-prefetch.ll test case.
llvm-svn: 226033
This can happen if:
* It is present in a comdat in one file.
* It is not present in the comdat of the file that is kept.
* Is is not used.
This should fix the LTO boostrap.
Thanks to Takumi NAKAMURA for setting up the bot!
llvm-svn: 225983
This adds the domtree analysis to the new pass manager. The analysis
returns the same DominatorTree result entity used by the old pass
manager and essentially all of the code is shared. We just have
different boilerplate for running and printing the analysis.
I've converted one test to run in both modes just to make sure this is
exercised while both are live in the tree.
llvm-svn: 225969
This commit refines the pattern for the octeon seq/seqi/sne/snei instructions.
The target register is set to 0 or 1 according to the result of the comparison.
In C, this is something like
rd = (unsigned long)(rs == rt)
This commit adds a zext to bring the result to i64. With this change the
instruction is selected for this type of code. (gcc produces the same code for
the above C code.)
llvm-svn: 225968
The int instruction takes as an operand an 8-bit immediate value. Validate that
the input is valid rather than silently truncating the value.
llvm-svn: 225941
Correct, we have *zero* basic testing of the dominator tree in the
regression test suite. There is a single test that even prints it out,
and that test only checks a single line of the output. There are
a handful of tests that check post dominators, but all of those are
looking for bugs rather than just exercising the basic machinery.
This test is super boring and unexciting. But hey, it's something.
I needed there to be something so I could switch the basic test to run
with both the old and new pass manager.
llvm-svn: 225936
The form of nops used is CPU-specific (some CPUs, such as the POWER7, have
special group-terminating nops). We probably want a different callback for this
kind of nop insertion (something more like MCAsmBackend::writeNopData), or for
PPC to use a different mechanism for scheduling nops, but this will stop the
test from failing for now.
llvm-svn: 225928
Don't do the v4i8 -> v4f32 combine if the load will need to
be expanded due to alignment. This stops adding instructions
to repack into a single register that the v_cvt_ubyteN_f32
instructions read.
llvm-svn: 225926
Now that the source and destination types can be specified,
allow doing an expansion that doesn't use an EXTLOAD of the
result type. Try to do a legal extload to an intermediate type
and extend that if possible.
This generalizes the special case custom lowering of extloads
R600 has been using to work around this problem.
This also happens to fix a bug that would incorrectly use more
aligned loads than should be used.
llvm-svn: 225925
This re-applies r225808, fixed to avoid problems with SDAG dependencies along
with the preceding fix to ScheduleDAGSDNodes::RegDefIter::InitNodeNumDefs.
These problems caused the original regression tests to assert/segfault on many
(but not all) systems.
Original commit message:
This commit does two things:
1. Refactors PPCFastISel to use more of the common infrastructure for call
lowering (this lets us take advantage of this common code for lowering some
common intrinsics, stackmap/patchpoint among them).
2. Adds support for stackmap/patchpoint lowering. For the most part, this is
very similar to the support in the AArch64 target, with the obvious differences
(different registers, NOP instructions, etc.). The test cases are adapted
from the AArch64 test cases.
One difference of note is that the patchpoint call sequence takes 24 bytes, so
you can't use less than that (on AArch64 you can go down to 16). Also, as noted
in the docs, we take the patchpoint address to be the actual code address
(assuming the call is local in the TOC-sharing sense), which should yield
higher performance than generating the full cross-DSO indirect-call sequence
and is likely just as useful for JITed code (if not, we'll change it).
StackMaps and Patchpoints are still marked as experimental, and so this support
is doubly experimental. So go ahead and experiment!
llvm-svn: 225909
A pass that adds random noops to X86 binaries to introduce diversity with the goal of increasing security against most return-oriented programming attacks.
Command line options:
-noop-insertion // Enable noop insertion.
-noop-insertion-percentage=X // X% of assembly instructions will have a noop prepended (default: 50%, requires -noop-insertion)
-max-noops-per-instruction=X // Randomly generate X noops per instruction. ie. roll the dice X times with probability set above (default: 1). This doesn't guarantee X noop instructions.
In addition, the following 'quick switch' in clang enables basic diversity using default settings (currently: noop insertion and schedule randomization; it is intended to be extended in the future).
-fdiversify
This is the llvm part of the patch.
clang part: D3393
http://reviews.llvm.org/D3392
Patch by Stephen Crane (@rinon)
llvm-svn: 225908
This adds handling for ExceptionHandling::MSVC, used by the
x86_64-pc-windows-msvc triple. It assumes that filter functions have
already been outlined in either the frontend or the backend. Filter
functions are used in place of the landingpad catch clause type info
operands. In catch clause order, the first filter to return true will
catch the exception.
The C specific handler table expects the landing pad to be split into
one block per handler, but LLVM IR uses a single landing pad for all
possible unwind actions. This patch papers over the mismatch by
synthesizing single instruction BBs for every catch clause to fill in
the EH selector that the landing pad block expects.
Missing functionality:
- Accessing data in the parent frame from outlined filters
- Cleanups (from __finally) are unsupported, as they will require
outlining and parent frame access
- Filter clauses are unsupported, as there's no clear analogue in SEH
In other words, this is the minimal set of changes needed to write IR to
catch arbitrary exceptions and resume normal execution.
Reviewers: majnemer
Differential Revision: http://reviews.llvm.org/D6300
llvm-svn: 225904
It turns out, all callsites of the simplifier are guarded by a check for
CallInst::getCalledFunction (i.e., to make sure the callee is direct).
This check wasn't done when trying to further optimize a simplified fortified
libcall, introduced by a refactoring in r225640.
Fix that, add a testcase, and document the requirement.
llvm-svn: 225895
emitDebugLocValue() into DwarfExpression.
Ought to be NFC, but it actually uncovered a bug in the debug-loc-asan.ll
testcase. The testcase checks that the address of variable "y" is stored
at [RSP+16], which also lines up with the comment.
It also check(ed) that the *value* of "y" is stored in RDI before that,
but that is actually incorrect, since RDI is the very value that is
stored in [RSP+16]. Here's the assembler output:
movb 2147450880(%rcx), %r8b
#DEBUG_VALUE: bar:y <- RDI
cmpb $0, %r8b
movq %rax, 32(%rsp) # 8-byte Spill
movq %rsi, 24(%rsp) # 8-byte Spill
movq %rdi, 16(%rsp) # 8-byte Spill
.Ltmp3:
#DEBUG_VALUE: bar:y <- [RSP+16]
Fixed the comment to spell out the correct register and the check to
expect an address rather than a value.
Note that the range that is emitted for the RDI location was and is still
wrong, it claims to begin at the function prologue, but really it should
start where RDI is first assigned.
llvm-svn: 225851
Even before I sunk the debug flag into the opt tool this had been made
obsolete by factoring the pass and analysis managers into a single set
of templates that all used the core flag. No functionality changed here.
llvm-svn: 225842
This now handles both 32 and 64-bit element sizes.
In this version, the test are in vector-shuffle-512-v8.ll, canonicalized by
Chandler's update_llc_test_checks.py.
Part of <rdar://problem/17688758>
llvm-svn: 225838
This adds assembly and bitcode support for `MDLocation`. The assembly
side is rather big, since this is the first `MDNode` subclass (that
isn't `MDTuple`). Part of PR21433.
(If you're wondering where the mountains of testcase updates are, we
don't need them until I update `DILocation` and `DebugLoc` to actually
use this class.)
llvm-svn: 225830
Only do for f32 since I'm unclear on both what this is expecting
for the refinement steps in terms of accuracy, and what
f64 instruction actually provides.
llvm-svn: 225827
Speculating things is generally good. SI+ has instructions for these
for 32-bit values. This is still probably better even with the expansion
for 64-bit values, although it is odd that this callback doesn't have
the size as a parameter.
llvm-svn: 225822
This was already done in clang, this commit now uses the integrated
assembler as default when using LLVM tools directly.
A number of test cases deliberately using an invalid instruction in
inline asm now have to use -no-integrated-as.
llvm-svn: 225820
This was already done in clang, this commit now uses the integrated
assembler as default when using LLVM tools directly.
A number of test cases using inline asm had to be adapted, either by
updating the expected output, or by using -no-integrated-as (for such
tests that deliberately use an invalid instruction in inline asm).
llvm-svn: 225819
The ppc64le platform will emit a .localentry directive. This is triggering
a false-positive against a CHECK-NOT: .loc in multiline.ll.
Add a space "{{ }}" to the check-not line to allow for arguments, and
prevent .localentry from matching.
Differential Revision: http://reviews.llvm.org/D6935
llvm-svn: 225810
This commit does two things:
1. Refactors PPCFastISel to use more of the common infrastructure for call
lowering (this lets us take advantage of this common code for lowering some
common intrinsics, stackmap/patchpoint among them).
2. Adds support for stackmap/patchpoint lowering. For the most part, this is
very similar to the support in the AArch64 target, with the obvious differences
(different registers, NOP instructions, etc.). The test cases are adapted
from the AArch64 test cases.
One difference of note is that the patchpoint call sequence takes 24 bytes, so
you can't use less than that (on AArch64 you can go down to 16). Also, as noted
in the docs, we take the patchpoint address to be the actual code address
(assuming the call is local in the TOC-sharing sense), which should yield
higher performance than generating the full cross-DSO indirect-call sequence
and is likely just as useful for JITed code (if not, we'll change it).
StackMaps and Patchpoints are still marked as experimental, and so this support
is doubly experimental. So go ahead and experiment!
llvm-svn: 225808
16 bit instructions are not allowed in jr delay slot. Same stands for
PseudoIndirectBranch and PseudoReturn.
Differential Revision: http://reviews.llvm.org/D6815
llvm-svn: 225798
the generic functionality of the pass managers themselves.
In the new infrastructure, the pass "manager" isn't actually interesting
at all. It just pipelines a single chunk of IR through N passes. We
don't need to know anything about the IR or the passes to do this really
and we can replace the 3 implementations of the exact same functionality
with a single generic PassManager template, complementing the single
generic AnalysisManager template.
I've left typedefs in place to give convenient names to the various
obvious instantiations of the template.
With this, I think I've nuked almost all of the redundant logic in the
managers, and I think the overall design is actually simpler for having
single templates that clearly indicate there is no special logic here.
The logging is made somewhat more annoying by this change, but I don't
think the difference is worth having heavy-weight traits to help log
things.
llvm-svn: 225783
template.
This consolidates three copies of nearly the same core logic. It adds
"complexity" to the ModuleAnalysisManager in that it makes it possible
to share a ModuleAnalysisManager across multiple modules... But it does
so by deleting *all of the code*, so I'm OK with that. This will
naturally make fixing bugs in this code much simpler, etc.
The only down side here is that we have to use 'typename' and 'this->'
in various places, and the implementation is lifted into the header.
I'll take that for the code size reduction.
The convenient names are still typedef-ed and used throughout so that
users can largely ignore this aspect of the implementation.
The follow-up change to this will do the exact same refactoring for the
PassManagers. =D
It turns out that the interesting different code is almost entirely in
the adaptors. At the end, that should be essentially all that is left.
llvm-svn: 225757
This name is less descriptive, but it sort of puts things in the
'llvm.frame...' namespace, relating it to frameallocate and
frameaddress. It also avoids using "allocate" and "allocation" together.
llvm-svn: 225752
These intrinsics allow multiple functions to share a single stack
allocation from one function's call frame. The function with the
allocation may only perform one allocation, and it must be in the entry
block.
Functions accessing the allocation call llvm.recoverframeallocation with
the function whose frame they are accessing and a frame pointer from an
active call frame of that function.
These intrinsics are very difficult to inline correctly, so the
intention is that they be introduced rarely, or at least very late
during EH preparation.
Reviewers: echristo, andrew.w.kaylor
Differential Revision: http://reviews.llvm.org/D6493
llvm-svn: 225746
Otherwise we'll attempt to forward ECX, EDX, and EAX for cdecl and
stdcall thunks, leaving us with no scratch registers for indirect call
targets.
Fixes PR22052.
llvm-svn: 225729
into a new class DwarfExpression that can be shared between AsmPrinter
and DwarfUnit.
This is the first step towards unifying the two entirely redundant
implementations of dwarf expression emission in DwarfUnit and AsmPrinter.
Almost no functional change — Testcases were updated because asm comments
that used to be on two lines now appear on the same line, which is
actually preferable.
llvm-svn: 225706
This happens in the HINT benchmark, where the SLP-vectorizer created
v2f32 fcmp/select code. The "correct" solution would have been to
teach the vectorizer cost model that v2f32 isn't legal (because really,
it isn't), but if we can vectorize we might as well do so.
We legalize these v2f32 FMIN/FMAX nodes by widening to v4f32 later on.
v3f32 were already widened to v4f32 by the generic unroll-and-build-vector
legalization.
rdar://15763436
Differential Revision: http://reviews.llvm.org/D6557
llvm-svn: 225691
There are some operands which can take either immediates or registers
and we were previously using different register class to distinguish
between operands that could take immediates and those that could not.
This patch switches to using RegisterOperands which should simplify the
backend by reducing the number of register classes and also make it
easier to implement the assembler.
llvm-svn: 225662
One is that AArch64 has additional restrictions on when local relocations can
be used. We have to take those into consideration when deciding to put a L
symbol in the symbol table or not.
The other is that ld64 requires the relocations to cstring to use linker
visible symbols on AArch64.
Thanks to Michael Zolotukhin for testing this!
Remove doesSectionRequireSymbols.
In an assembly expression like
bar:
.long L0 + 1
the intended semantics is that bar will contain a pointer one byte past L0.
In sections that are merged by content (strings, 4 byte constants, etc), a
single position in the section doesn't give the linker enough information.
For example, it would not be able to tell a relocation must point to the
end of a string, since that would look just like the start of the next.
The solution used in ELF to use relocation with symbols if there is a non-zero
addend.
In MachO before this patch we would just keep all symbols in some sections.
This would miss some cases (only cstrings on x86_64 were implemented) and was
inefficient since most relocations have an addend of 0 and can be represented
without the symbol.
This patch implements the non-zero addend logic for MachO too.
llvm-svn: 225644
Looking at r225438 inspired me to see how the PowerPC backend handled the
situation (calling a bitcasted TLS global), and it turns out we also produced
an error (cannot select ...). What it means to "call" something that is not a
function is implementation and platform specific, but in the name of doing
something (besides crashing), this makes sure we do what GCC does (treat all
such calls as calls through a function pointer -- meaning that the pointer is
assumed, as is the convention on PPC, to point to a function descriptor
structure holding the actual code address along with the function's TOC pointer
and environment pointer). As GCC does, we now do the same for calling regular
(non-TLS) non-function globals too.
I'm not sure whether this is the most useful way to define the behavior, but at
least we won't be alone.
llvm-svn: 225617
It's possible for the constant pool entry for the shuffle mask to come
from a completely different operation. This occurs when Constants have
the same bit pattern but have different types.
Make DecodePSHUFBMask tolerant of types which, after a bitcast, are
appropriately sized vector types.
This fixes PR22188.
llvm-svn: 225597
Teach the ISelLowering for X86 about the L,M,O target specific constraints.
Although, for the moment, clang performs constraint validation and prevents
passing along inline asm which may have immediate constant constraints violated,
the backend should be able to cope with the invalid inline asm a bit better.
llvm-svn: 225596
This adds support for parsing and emitting the SBREL relocation variant for the
ARM target. Handling this relocation variant is necessary for supporting the
full ARM ELF specification. Addresses PR22128.
llvm-svn: 225595
In the current code we only attempt to match against insertps if we have exactly one element from the second input vector, irrespective of how much of the shuffle result is zeroable.
This patch checks to see if there is a single non-zeroable element from either input that requires insertion. It also supports matching of cases where only one of the inputs need to be referenced.
We also split insertps shuffle matching off into a new lowerVectorShuffleAsInsertPS function.
Differential Revision: http://reviews.llvm.org/D6879
llvm-svn: 225589
This initial implementation of PPCTargetLowering::isZExtFree marks as free
zexts of small scalar loads (that are not sign-extending). This callback is
used by SelectionDAGBuilder's RegsForValue::getCopyToRegs, and thus to
determine whether a zext or an anyext is used to lower illegally-typed PHIs.
Because later truncates of zero-extended values are nops, this allows for the
elimination of later unnecessary truncations.
Fixes the initial complaint associated with PR22120.
llvm-svn: 225584
There is a fair number of relocations that are part of the AAELF specification.
Simply merge the tests into a single test file, otherwise, we will end up with
far too many test files to test each relocation type. NFC.
llvm-svn: 225576
These tests are checking the relocation generation. Use the readobj output as
it is much easier to follow when glancing over the tests.
llvm-svn: 225575
Summary:
In the previous commit, the register was saved, but space was not allocated.
This resulted in the parameter save area potentially clobbering r30, leading to
nasty results.
Test Plan: Tests updated
Reviewers: hfinkel
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6906
llvm-svn: 225573
Now that the way that the partial unrolling threshold for small loops is used
to compute the unrolling factor as been corrected, a slightly smaller threshold
is preferable. This is expected; other targets may need to re-tune as well.
llvm-svn: 225566
When we compute the size of a loop, we include the branch on the backedge and
the comparison feeding the conditional branch. Under normal circumstances,
these don't get replicated with the rest of the loop body when we unroll. This
led to the somewhat surprising behavior that really small loops would not get
unrolled enough -- they could be unrolled more and the resulting loop would be
below the threshold, because we were assuming they'd take
(LoopSize * UnrollingFactor) instructions after unrolling, instead of
(((LoopSize-2) * UnrollingFactor)+2) instructions. This fixes that computation.
llvm-svn: 225565
The bitcode reading interface used std::error_code to report an error to the
callers and it is the callers job to print diagnostics.
This is not ideal for error handling or diagnostic reporting:
* For error handling, all that the callers care about is 3 possibilities:
* It worked
* The bitcode file is corrupted/invalid.
* The file is not bitcode at all.
* For diagnostic, it is user friendly to include far more information
about the invalid case so the user can find out what is wrong with the
bitcode file. This comes up, for example, when a developer introduces a
bug while extending the format.
The compromise we had was to have a lot of error codes.
With this patch we use the DiagnosticHandler to communicate with the
human and std::error_code to communicate with the caller.
This allows us to have far fewer error codes and adds the infrastructure to
print better diagnostics. This is so because the diagnostics are printed when
he issue is found. The code that detected the problem in alive in the stack and
can pass down as much context as needed. As an example the patch updates
test/Bitcode/invalid.ll.
Using a DiagnosticHandler also moves the fatal/non-fatal error decision to the
caller. A simple one like llvm-dis can just use fatal errors. The gold plugin
needs a bit more complex treatment because of being passed non-bitcode files. An
hypothetical interactive tool would make all bitcode errors non-fatal.
llvm-svn: 225562
The previous code assumed that such instructions could not have any uses
outside CaseDest, with the motivation that the instruction could not
dominate CommonDest because CommonDest has phi nodes in it. That simply
isn't true; e.g., CommonDest could have an edge back to itself.
llvm-svn: 225552
pshufb can shuffle in zero bytes as well as bytes from a source vector - we can use this to avoid having to shuffle 2 vectors and ORing the result when the used inputs from a vector are all zeroable.
Differential Revision: http://reviews.llvm.org/D6878
llvm-svn: 225551
doing Load PRE"
It's not really expected to stick around, last time it provoked a weird LTO
build failure that I can't reproduce now, and the bot logs are long gone. I'll
re-revert it if the failures recur.
Original description: Perform Scalar PRE on gep indices that feed loads before
doing Load PRE.
llvm-svn: 225536
Summary:
Mips Linux uses $gp to hold a pointer to thread info structure and accesses it
with a named register. This makes this work for LLVM.
The N32 ABI doesn't quite work yet since the frontend generates incorrect IR
for this case. It neglects to truncate the 64-bit GPR to a 32-bit value before
converting to a pointer. Given correct IR (as in the testcase in this patch),
it works correctly.
Reviewers: sstankovic, vmedic, atanasyan
Reviewed By: atanasyan
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6893
llvm-svn: 225529
The P7 benefits from not have really-small loops so that we either have
multiple dispatch groups in the loop and/or the ability to form more-full
dispatch groups during scheduling. Setting the partial unrolling threshold to
44 seems good, empirically, for the P7. Compared to using no late partial
unrolling, this yields the following test-suite speedups:
SingleSource/Benchmarks/Adobe-C++/simple_types_constant_folding
-66.3253% +/- 24.1975%
SingleSource/Benchmarks/Misc-C++/oopack_v1p8
-44.0169% +/- 29.4881%
SingleSource/Benchmarks/Misc/pi
-27.8351% +/- 12.2712%
SingleSource/Benchmarks/Stanford/Bubblesort
-30.9898% +/- 22.4647%
I've speculatively added a similar setting for the P8. Also, I've noticed that
the unroller does not quite calculate the unrolling factor correctly for really
tiny loops because it neglects to account for the fact that not every loop body
replicant contains an ending branch and counter increment. I'll fix that later.
llvm-svn: 225522
The code that eliminated additional coalescable copies in
removeCopyByCommutingDef() used MergeValueNumberInto() which internally
may merge A into B or B into A. In this case A and B had different Def
points, so we have to reset ValNo.Def to the intended one after merging.
llvm-svn: 225503
On modern cores with lfiw[az]x, we can fold a sign or zero extension from i32
to i64 into the load necessary for an i64 -> fp conversion.
llvm-svn: 225493
Create new copies of distinct `MDNode`s instead of following the
uniquing `MDNode` logic.
Just like self-references (or other cycles), `MapMetadata()` creates a
new node. In practice most calls use `RF_NoModuleLevelChanges`, in
which case nothing is duplicated anyway.
Part of PR22111.
llvm-svn: 225476
Propagate whether `MDNode`s are 'distinct' through the other types of IR
(assembly and bitcode). This adds the `distinct` keyword to assembly.
Currently, no one actually calls `MDNode::getDistinct()`, so these nodes
only get created for:
- self-references, which are never uniqued, and
- nodes whose operands are replaced that hit a uniquing collision.
The concept of distinct nodes is still not quite first-class, since
distinct-ness doesn't yet survive across `MapMetadata()`.
Part of PR22111.
llvm-svn: 225474
MachineLICM uses a callback named hasLowDefLatency to determine if an
instruction def operand has a 'low' latency. If all relevant operands have a
'low' latency, the instruction is considered too cheap to hoist out of loops
even in low-register-pressure situations. On PowerPC cores, both the embedded
cores and the others, there is no reason to believe that this is a good choice:
all instructions have a cost inside a loop, and hoisting them when not limited
by register pressure is a reasonable default.
llvm-svn: 225471
The assert was being triggered when the distance between a constant pool entry
and its user exceeded the maximally allowed distance after thumb2 branch
shortening. A padding was inserted after a thumb2 branch instruction was shrunk,
which caused the user to be out of range. This is wrong as the padding should
have been inserted by the layout algorithm so that the distance between two
instructions doesn't grow later during thumb2 instruction optimization.
This commit fixes the code in ARMConstantIslands::createNewWater to call
computeBlockSize and set BasicBlock::Unalign when a branch instruction is
inserted to create new water after a basic block. A non-zero Unalign causes
the worst-case padding to be inserted when adjustBBOffsetsAfter is called to
recompute the basic block offsets.
rdar://problem/19130476
llvm-svn: 225467
Summary: The PIC additions didn't update the prologue and epilogue code to save and restore r30 (PIC base register). This does that.
Test Plan: Tests updated.
Reviewers: hfinkel
Reviewed By: hfinkel
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6876
llvm-svn: 225450
This partially fixes PR13007 (ARM CodeGen fails with large stack
alignment): for ARM and Thumb2 targets, but not for Thumb1, as it
seems stack alignment for Thumb1 targets hasn't been supported at
all.
Producing an aligned stack pointer is done by zero-ing out the lower
bits of the stack pointer. The BIC instruction was used for this.
However, the immediate field of the BIC instruction only allows to
encode an immediate that can zero out up to a maximum of the 8 lower
bits. When a larger alignment is requested, a BIC instruction cannot
be used; llvm was silently producing incorrect code in this case.
This commit fixes code generation for large stack aligments by
using the BFC instruction instead, when the BFC instruction is
available. When not, it uses 2 instructions: a right shift,
followed by a left shift to zero out the lower bits.
The lowering of ARM::Int_eh_sjlj_dispatchsetup still has code
that unconditionally uses BIC to realign the stack pointer, so it
very likely has the same problem. However, I wasn't able to
produce a test case for that. This commit adds an assert so that
the compiler will fail the assert instead of silently generating
wrong code if this is ever reached.
llvm-svn: 225446
Its functionality has been replaced by calling
SIInstrInfo::legalizeOperands() from
SIISelLowering::AdjstInstrPostInstrSelection() and running the
SIFoldOperands and SIShrinkInstructions passes.
llvm-svn: 225445
The call lowering assumes that if the callee is a global, we want to emit a direct call.
This is correct for regular globals, but not for TLS ones.
Differential Revision: http://reviews.llvm.org/D6862
llvm-svn: 225438
A broken hint is a copy where both ends are assigned different colors. When a
variable gets evicted in the neighborhood of such copies, it is likely we can
reconcile some of them.
** Context **
Copies are inserted during the register allocation via splitting. These split
points are required to relax the constraints on the allocation problem. When
such a point is inserted, both ends of the copy would not share the same color
with respect to the current allocation problem. When variables get evicted,
the allocation problem becomes different and some split point may not be
required anymore. However, the related variables may already have been colored.
This usually shows up in the assembly with pattern like this:
def A
...
save A to B
def A
use A
restore A from B
...
use B
Whereas we could simply have done:
def B
...
def A
use A
...
use B
** Proposed Solution **
A variable having a broken hint is marked for late recoloring if and only if
selecting a register for it evict another variable. Indeed, if no eviction
happens this is pointless to look for recoloring opportunities as it means the
situation was the same as the initial allocation problem where we had to break
the hint.
Finally, when everything has been allocated, we look for recoloring
opportunities for all the identified candidates.
The recoloring is performed very late to rely on accurate copy cost (all
involved variables are allocated).
The recoloring is simple unlike the last change recoloring. It propagates the
color of the broken hint to all its copy-related variables. If the color is
available for them, the recoloring uses it, otherwise it gives up on that hint
even if a more complex coloring would have worked.
The recoloring happens only if it is profitable. The profitability is evaluated
using the expected frequency of the copies of the currently recolored variable
with a) its current color and b) with the target color. If a) is greater or
equal than b), then it is profitable and the recoloring happen.
** Example **
Consider the following example:
BB1:
a =
b =
BB2:
...
= b
= a
Let us assume b gets split:
BB1:
a =
b =
BB2:
c = b
...
d = c
= d
= a
Because of how the allocation work, b, c, and d may be assigned different
colors. Now, if a gets evicted to make room for c, assuming b and d were
assigned to something different than a.
We end up with:
BB1:
a =
st a, SpillSlot
b =
BB2:
c = b
...
d = c
= d
e = ld SpillSlot
= e
This is likely that we can assign the same register for b, c, and d,
getting rid of 2 copies.
** Performances **
Both ARM64 and x86_64 show performance improvements of up to 3% for the
llvm-testsuite + externals with Os and O3. There are a few regressions too that
comes from the (in)accuracy of the block frequency estimate.
<rdar://problem/18312047>
llvm-svn: 225422
I got confused and assumed SrcIdx/DstIdx of the CoalescerPair is a
subregister index in SrcReg/DstReg, but they are actually subregister
indices of the coalesced register that get you back to SrcReg/DstReg
when applied.
Fixed the bug, improved comments and simplified code accordingly.
Testcase by Tom Stellard!
llvm-svn: 225415
Patch by: Ramkumar Ramachandra <artagnon@gmail.com>
"This patch started out as an exploration of gc.relocate, and an attempt
to write a simple test in call-lowering. I then noticed that the
arguments of gc.relocate were not checked fully, so I went in and fixed
a few things. Finally, the most important outcome of this patch is that
my new error handling code caught a bug in a callsite in
stackmap-format."
Differential Revision: http://reviews.llvm.org/D6824
llvm-svn: 225412
`MDNode::replaceOperandWith()` changes all instances of metadata. Stop
using it when linking module flags, since (due to uniquing) the flag
values could be used by other metadata.
Instead, use new API `NamedMDNode::setOperand()` to update the reference
directly.
llvm-svn: 225397
options other than just -disassemble so that universal files can be used with other
options combined with -arch options.
No functional change to existing options and use. One test case added for the
additional functionality with a universal file an a -arch option.
llvm-svn: 225383
This change includes the most basic possible GCStrategy for a GC which is using the statepoint lowering code. At the moment, this GCStrategy doesn't really do much - aside from actually generate correct stackmaps that is - but I went ahead and added a few extra correctness checks as proof of concept. It's mostly here to provide documentation on how to do one, and to provide a point for various optimization legality hooks I'd like to add going forward. (For context, see the TODOs in InstCombine around gc.relocate.)
Most of the validation logic added here as proof of concept will soon move in to the Verifier. That move is dependent on http://reviews.llvm.org/D6811
There was discussion in the review thread about addrspace(1) being reserved for something. I'm going to follow up on a seperate llvmdev thread. If needed, I'll update all the code at once.
Note that I am deliberately not making a GCStrategy required to use gc.statepoints with this change. I want to give folks out of tree - including myself - a chance to migrate. In a week or two, I'll make having a GCStrategy be required for gc.statepoints. To this end, I added the gc tag to one of the test cases but not others.
Differential Revision: http://reviews.llvm.org/D6808
llvm-svn: 225365
LLVM emits stack probes on Windows targets to ensure that the stack is
correctly accessed. However, the amount of stack allocated before
emitting such a probe is hardcoded to 4096.
It is desirable to have this be configurable so that a function might
opt-out of stack probes. Our level of granularity is at the function
level instead of, say, the module level to permit proper generation of
code after LTO.
Patch by Andrew H!
N.B. The inliner needs to be updated to properly consider what happens
after inlining a function with a specific stack-probe-size into another
function with a different stack-probe-size.
llvm-svn: 225360
requiring and invalidating specific analyses. Also make their printed
names match their class names. Writing these out as prose really doesn't
make sense to me any more.
llvm-svn: 225346
Even thouh gcc produces simialr instructions as Owen pointed out the two patterns aren’t equivalent in the case
where the original subtraction could have caused an overflow.
Reverting the same.
llvm-svn: 225341
passes too many time.
I think this is actually the issue that someone raised with me at the
developer's meeting and in an email, but that we never really got to the
bottom of. Having all the testing utilities made it much easier to dig
down and uncover the core issue.
When a pass manager is running many passes over a single function, we
need it to invalidate the analyses between each run so that they can be
re-computed as needed. We also need to track the intersection of
preserved higher-level analyses across all the passes that we run (for
example, if there is one module analysis which all the function analyses
preserve, we want to track that and propagate it). Unfortunately, this
interacted poorly with any enclosing pass adaptor between two IR units.
It would see the intersection of preserved analyses, and need to
invalidate any other analyses, but some of the un-preserved analyses
might have already been invalidated *and recomputed*! We would fail to
propagate the fact that the analysis had already been invalidated.
The solution to this struck me as really strange at first, but the more
I thought about it, the more natural it seemed. After a nice discussion
with Duncan about it on IRC, it seemed even nicer. The idea is that
invalidating an analysis *causes* it to be preserved! Preserving the
lack of result is trivial. If it is recomputed, great. Until something
*else* invalidates it again, we're good.
The consequence of this is that the invalidate methods on the analysis
manager which operate over many passes now consume their
PreservedAnalyses object, update it to "preserve" every analysis pass to
which it delivers an invalidation (regardless of whether the pass
chooses to be removed, or handles the invalidation itself by updating
itself). Then we return this augmented set from the invalidate routine,
letting the pass manager take the result and use the intersection of
*that* across each pass run to compute the final preserved set. This
accounts for all the places where the early invalidation of an analysis
has already "preserved" it for a future run.
I've beefed up the testing and adjusted the assertions to show that we
no longer repeatedly invalidate or compute the analyses across nested
pass managers.
llvm-svn: 225333
In order to make comdats always explicit in the IR, we decided to make
the syntax a bit more compact for the case of a GlobalObject in a
comdat with the same name.
Just dropping the $name causes problems for
@foo = globabl i32 0, comdat
$bar = comdat ...
and
declare void @foo() comdat
$bar = comdat ...
So the syntax is changed to
@g1 = globabl i32 0, comdat($c1)
@g2 = globabl i32 0, comdat
and
declare void @foo() comdat($c1)
declare void @foo() comdat
llvm-svn: 225302
int->fp conversions on PPC must be done through memory loads and stores. On a
modern core, this process begins by storing the int value to memory, then
loading it using a (sometimes special) FP load instruction. Unfortunately, we
would do this even when the value to be converted was itself a load, and we can
just use that same memory location instead of copying it to another first.
There is a slight complication when handling int_to_fp(fp_to_int(x)) pairs,
because the fp_to_int operand has not been lowered when the int_to_fp is being
lowered. We handle this specially by invoking fp_to_int's lowering logic
(partially) and getting the necessary memory location (some trivial refactoring
was done to make this possible).
This is all somewhat ugly, and it would be nice if some later CodeGen stage
could just clean this stuff up, but because doing so would involve modifying
target-specific nodes (or instructions), it is not immediately clear how that
would work.
Also, remove a related entry from the README.txt for which we now generate
reasonable code.
llvm-svn: 225301
This is equivalent to the AMDGPUTargetMachine now, but it is the
starting point for separating R600 and GCN functionality into separate
targets.
It is recommened that users start using the gcn triple for GCN-based
GPUs, because using the r600 triple for these GPUs will be deprecated in
the future.
llvm-svn: 225277
This patch improves the logic added at revision 224899 (see review D6728) that
teaches the backend when it is profitable to speculate calls to cttz/ctlz.
The original algorithm conservatively avoided speculating more than one
instruction from a basic block in a control flow grap modelling an if-statement.
In particular, the only allowed instruction (excluding the terminator) was a
call to cttz/ctlz. However, there are cases where we could be less conservative
and still be able to speculate a call to cttz/ctlz.
With this patch, CodeGenPrepare now tries to speculate a cttz/ctlz if the
result is zero extended/truncated in the same basic block, and the zext/trunc
instruction is "free" for the target.
Added new test cases to CodeGen/X86/cttz-ctlz.ll
Differential Revision: http://reviews.llvm.org/D6853
llvm-svn: 225274
This also rolls in the changes discussed in http://reviews.llvm.org/D6766.
Defers migrating the debug info for new allocas until after all partitions
are created.
Thanks to Chandler for reviewing!
llvm-svn: 225272
In r225251, I removed an old entry from the README.txt file. While there are
several contributing factors (including pieces in Clang's ABI code), upon
further reflection, the backend part deserves a regression test.
llvm-svn: 225268
This is already handled in general when it is known the
conversion can't lose bits with smaller integer types
casted into wider floating point types.
This pattern happens somewhat often in GPU programs that cast
workitem intrinsics to float, which are often compared with 0.
Specifically handle the special case of compares with zero which
should also be known to not lose information. I had a more general
version of this which allows equality compares if the casted float is
exactly representable in the integer, but I'm not 100% confident that
is always correct.
Also fold cases that aren't integers to true / false.
llvm-svn: 225265
Use this to test that path of invalidation. This test actually shows
redundant invalidation here that is really bad. I'm going to work on
fixing that next, but wanted to commit the test harness now that its all
working.
llvm-svn: 225257
Requires new AsmParserOperand types that detect 16-bit and 32/64-bit mode so that we choose the right instruction based on default sizing without predicates. This is necessary since predicates mess up the disassembler table building.
llvm-svn: 225256
Try harder to get rid of bitcast'd calls by ptrtoint/inttoptr'ing
arguments and return values when DataLayout says it is safe to do so.
llvm-svn: 225254
remove an extra, redundant pass manager wrapping every run.
I had kept seeing these when manually testing, but it was getting really
annoying and was going to cause problems with overly eager invalidation.
The root cause was an overly complex and unnecessary pile of code for
parsing the outer layer of the pass pipeline. We can instead delegate
most of this to the recursive pipeline parsing.
I've added some somewhat more basic and precise tests to catch this.
llvm-svn: 225253
"ELF Handling for Thread-Local Storage" specifies that R_X86_64_GOTTPOFF
relocation target a movq or addq instruction.
Prohibit the truncation of such loads to movl or addl.
This fixes PR22083.
Differential Revision: http://reviews.llvm.org/D6839
llvm-svn: 225250
The old target DAG combine that allowed for performing int_to_fp(fp_to_int(x))
without a load/store pair is updated here with support for unsigned integers,
and to support single-precision values without a third rounding step, on newer
cores with the appropriate instructions.
llvm-svn: 225248
a specific analysis result.
This is quite handy to test things, and will also likely be very useful
for debugging issues. You could narrow down pass validation failures by
walking these invalidate pass runs up and down the pass pipeline, etc.
I've added support to the pass pipeline parsing to be able to create one
of these for any analysis pass desired.
Just adding this class uncovered one latent bug where the
AnalysisManager CRTP base class had a hard-coded Module type rather than
using IRUnitT.
I've also added tests for invalidation and caching of analyses in
a basic way across all the pass managers. These in turn uncovered two
more bugs where we failed to correctly invalidate an analysis -- its
results were invalidated but the key for re-running the pass was never
cleared and so it was never re-run. Quite nasty. I'm very glad to debug
this here rather than with a full system.
Also, yes, the naming here is horrid. I'm going to update some of the
names to be slightly less awful shortly. But really, I've no "good"
ideas for naming. I'll be satisfied if I can get it to "not bad".
llvm-svn: 225246
manager tests to use them and be significantly more comprehensive.
This, naturally, uncovered a bug where the CGSCC pass manager wasn't
printing analyses when they were run.
The only remaining core manipulator is I think an invalidate pass
similar to the require pass. That'll be next. =]
llvm-svn: 225240
is a no-op other than requiring some analysis results be available.
This can be used in real pass pipelines to force the usually lazy
analysis running to eagerly compute something at a specific point, and
it can be used to test the pass manager infrastructure (my primary use
at the moment).
I've also added bit of pipeline parsing magic to support generating
these directly from the opt command so that you can directly use these
when debugging your analysis. The syntax is:
require<analysis-name>
This can be used at any level of the pass manager. For example:
cgscc(function(require<my-analysis>,no-op-function))
This would produce a no-op function pass requiring my-analysis, followed
by a fully no-op function pass, both of these in a function pass manager
which is nested inside of a bottom-up CGSCC pass manager which is in the
top-level (implicit) module pass manager.
I have zero attachment to the particular syntax I'm using here. Consider
it a straw man for use while I'm testing and fleshing things out.
Suggestions for better syntax welcome, and I'll update everything based
on any consensus that develops.
I've used this new functionality to more directly test the analysis
printing rather than relying on the cgscc pass manager running an
analysis for me. This is still minimally tested because I need to have
analyses to run first! ;] That patch is next, but wanted to keep this
one separate for easier review and discussion.
llvm-svn: 225236
We now produce the desired code as noted in the README.txt file (no spurious
or). Remove the README entry and improve the regression test.
llvm-svn: 225214
This object is meant to own the ObjectFiles and their underlying
MemoryBuffer. It is basically the equivalent of an OwningBinary
except that it efficiently handles Archives. It is optimized for
efficiently providing mappings of members of the same archive when
they are opened successively (which is standard in Darwin debug
maps, objects from the same archive will be contiguous).
Of course, the BinaryHolder will also be used by the DWARF linker
once it is commited, but for now only the debug map parser uses it.
With this change, you can run llvm-dsymutil on your Darwin debug build
of clang and get a complete debug map for it.
Differential Revision: http://reviews.llvm.org/D6690
llvm-svn: 225207
Consider this function from our README.txt file:
int foo(int a, int b) { return (a < b) << 4; }
We now explicitly track CR bits by default, so the comment in the README.txt
about not really having a SETCC is no longer accurate, but we did generate this
somewhat silly code:
cmpw 0, 3, 4
li 3, 0
li 12, 1
isel 3, 12, 3, 0
sldi 3, 3, 4
blr
which generates the zext as a select between 0 and 1, and then shifts the
result by a constant amount. Here we preprocess the DAG in order to fold the
results of operations on an extension of an i1 value into the SELECT_I[48]
pseudo instruction when the resulting constant can be materialized using one
instruction (just like the 0 and 1). This was not implemented as a DAGCombine
because the resulting code would have been anti-canonical and depends on
replacing chained user nodes, which does not fit well into the lowering
paradigm. Now we generate:
cmpw 0, 3, 4
li 3, 0
li 12, 16
isel 3, 12, 3, 0
blr
which is less silly.
llvm-svn: 225203
The 64-bit semantics of cntlzw are not special, the 32-bit population count is
stored as a 64-bit value in the range [0,32]. As a result, it is always zero
extended, and it can be added to the PPCISelDAGToDAG peephole optimization as a
frontier instruction for the removal of unnecessary zero extensions.
llvm-svn: 225192
lhbrx and lwbrx not only load their data with byte swapping, but also clear the
upper 32 bits (at least). As a result, they can be added to the PPCISelDAGToDAG
peephole optimization as frontier instructions for the removal of unnecessary
zero extensions.
llvm-svn: 225189
We used to generate code similar to:
umov.b w8, v0[2]
strb w8, [x0, x1]
because the STR*ro* patterns were preferred to ST1*.
Instead, we can avoid going through GPRs, and generate:
add x8, x0, x1
st1.b { v0 }[2], [x8]
This patch increases the ST1* AddedComplexity to achieve that.
rdar://16372710
Differential Revision: http://reviews.llvm.org/D6202
llvm-svn: 225183
For 0-lane stores, we used to generate code similar to:
fmov w8, s0
str w8, [x0, x1, lsl #2]
instead of:
str s0, [x0, x1, lsl #2]
To correct that: for store lane 0 patterns, directly match to STR <subreg>0.
Byte-sized instructions don't have the special case for a 0 index,
because FPR8s are defined to have untyped content.
rdar://16372710
Differential Revision: http://reviews.llvm.org/D6772
llvm-svn: 225181
Tag_compatibility takes two arguments, but before this patch it would
erroneously accept just one, it now produces an error in that case.
Change-Id: I530f918587620d0d5dfebf639944d6083871ef7d
llvm-svn: 225167
Claim conformance to version 2.09 of the ARM ABI.
This build attribute must be emitted first amongst the build attributes when
written to an object file. This is to simplify conformance detection by
consumers.
Change-Id: If9eddcfc416bc9ad6e5cc8cdcb05d0031af7657e
llvm-svn: 225166
This patch lowers patterns such as-
sub v0.4s, v0.4s, v1.4s
abs v0.4s, v0.4s
to
sabd v0.4s, v0.4s, v1.4s
on AArch64.
Review: http://reviews.llvm.org/D6781
llvm-svn: 225165
when all are being preserved.
We want to short-circuit this for a couple of reasons. One, I don't
really want passes to grow a dependency on actually receiving their
invalidate call when they've been preserved. I'm thinking about removing
this entirely. But more importantly, preserving everything is likely to
be the common case in a lot of scenarios, and it would be really good to
bypass all of the invalidation and preservation machinery there.
Avoiding calling N opaque functions to try to invalidate things that are
by definition still valid seems important. =]
This wasn't really inpsired by much other than seeing the spam in the
logging for analyses, but it seems better ot get it checked in rather
than forgetting about it.
llvm-svn: 225163
manager.
This starts to allow us to test analyses more easily, but it's really
only the beginning. Some of the code here is still untestable without
manual changes to create analysis passes, but I wanted to factor it into
a small of chunks as possible.
Next up in order to be able to test things are, in no particular order:
- No-op analyses passes so we don't have to use real ones to exercise
the pass maneger itself.
- Automatic way of generating dummy passes that require an analysis be
run, including a variant that calls a 'print' method on a pass to make
it even easier to print out the results of an analysis.
- Dummy passes that invalidate all analyses for their IR unit so we can
test invalidation and re-runs.
- Automatic way to print each analysis pass as it is re-run.
- Automatic but optional verification of analysis passes everywhere
possible.
I'm not claiming I'll get to all of these immediately, but that's what
is in the pipeline at some stage. I'm fleshing out exactly what I need
and what to prioritize by working on converting analyses and then trying
to test the conversion. =]
llvm-svn: 225162
{code}
// loop body
... = a[i] (1)
... = a[i+1] (2)
.......
a[i+1] = .... (3)
a[i] = ... (4)
{code}
The algorithm tries to collect memory access candidates from AliasSetTracker, and then check memory dependences one another. The memory accesses are unique in AliasSetTracker, and a single memory access in AliasSetTracker may map to multiple entries in AccessAnalysis, which could cover both 'read' and 'write'. Originally the algorithm only checked 'write' entry in Accesses if only 'write' exists. This is incorrect and the consequence is it ignored all read access, and finally some RAW and WAR dependence are missed.
For the case given above, if we ignore two reads, the dependence between (1) and (3) would not be able to be captured, and finally this loop will be incorrectly vectorized.
The fix simply inserts a new loop to find all entries in Accesses. Since it will skip most of all other memory accesses by checking the Value pointer at the very beginning of the loop, it should not increase compile-time visibly.
llvm-svn: 225159
PPC has an instruction for ctlz with defined zero behavior, and our lowering of
cttz (provided by DAGCombine) is also efficient and branchless, so speculating
these makes sense.
llvm-svn: 225150
assert out of the new pre-splitting in SROA.
This fix makes the code do what was originally intended -- when we have
a store of a load both dealing in the same alloca, we force them to both
be pre-split with identical offsets. This is really quite hard to do
because we can keep discovering problems as we go along. We have to
track every load over the current alloca which for any resaon becomes
invalid for pre-splitting, and go back to remove all stores of those
loads. I've included a couple of test cases derived from PR22093 that
cover the different ways this can happen. While that PR only really
triggered the first of these two, its the same fundamental issue.
The other challenge here is documented in a FIXME now. We end up being
quite a bit more aggressive for pre-splitting when loads and stores
don't refer to the same alloca. This aggressiveness comes at the cost of
introducing potentially redundant loads. It isn't clear that this is the
right balance. It might be considerably better to require that we only
do pre-splitting when we can presplit every load and store involved in
the entire operation. That would give more consistent if conservative
results. Unfortunately, it requires a non-trivial change to the actual
pre-splitting operation in order to correctly handle cases where we end
up pre-splitting stores out-of-order. And it isn't 100% clear that this
is the right direction, although I'm starting to suspect that it is.
llvm-svn: 225149
r225135 added the ability to materialize i64 constants using rotations in order
to reduce the instruction count. Sometimes we can use a rotation only with some
extra masking, so that we take advantage of the fact that generating a bunch of
extra higher-order 1 bits is easy using li/lis.
llvm-svn: 225147
The required functionality has been there for some time, but I never
managed to actually wire it into the command line registry of passes.
Let's do that.
llvm-svn: 225144
Materializing full 64-bit constants on PPC64 can be expensive, requiring up to
5 instructions depending on the locations of the non-zero bits. Sometimes
materializing a rotated constant, and then applying the inverse rotation, requires
fewer instructions than the direct method. If so, do that instead.
In r225132, I added support for forming constants using bit inversion. In
effect, this reverts that commit and replaces it with rotation support. The bit
inversion is useful for turning constants that are mostly ones into ones that
are mostly zeros (thus enabling a more-efficient shift-based materialization),
but the same effect can be obtained by using negative constants and a rotate,
and that is at least as efficient, if not more.
llvm-svn: 225135
Materializing full 64-bit constants on PPC64 can be expensive, requiring up to
5 instructions depending on the locations of the non-zero bits. Sometimes
materializing the bit-reversed constant, and then flipping the bits, requires
fewer instructions than the direct method. If so, do that instead.
llvm-svn: 225132
We assumed the output of a match was a Value, this would cause us to
assert because we would fail a cast<>. Instead, use a helper in the
Operator family to hide the distinction between Value and Constant.
This fixes PR22087.
llvm-svn: 225127
PHI nodes can have zero operands in the middle of a transform. It is
expected that utilities in Analysis don't freak out when this happens.
Note that it is considered invalid to allow these misshapen phi nodes to
make it to another pass.
This fixes PR22086.
llvm-svn: 225126
Weak externals are resolved statically, so we can actually generate the tail
call on PE/COFF targets without breaking the requirements. It is questionable
whether we want to propagate the current behaviour for MachO as the requirements
are part of the ARM ELF specifications, and it seems that prior to the SVN
r215890, we would have tail'ed the call. For now, be conservative and only
permit it on PE/COFF where the call will always be fully resolved.
llvm-svn: 225119
The existing code provided for specifying a global loop alignment preference.
However, the preferred loop alignment might depend on the loop itself. For
recent POWER cores, loops between 5 and 8 instructions should have 32-byte
alignment (while the others are better with 16-byte alignment) so that the
entire loop will fit in one i-cache line.
To support this, getPrefLoopAlignment has been made virtual, and can be
provided with an optional MachineLoop* so the target can inspect the loop
before answering the query. The default behavior, as before, is to return the
value set with setPrefLoopAlignment. MachineBlockPlacement now queries the
target for each loop instead of only once per function. There should be no
functional change for other targets.
llvm-svn: 225117
Most modern PowerPC cores prefer that functions and loops start on
16-byte-aligned boundaries (*), so instruct block placement, etc. to make this
happen. The branch selector has also been adjusted so account for the extra
nops that might now be inserted before loop headers.
(*) Some cores actually prefer other alignments for small loops, but that will
be addressed in a follow-up commit.
llvm-svn: 225115
Newer POWER cores, and the A2, support the cmpb instruction. This instruction
compares its operands, treating each of the 8 bytes in the GPRs separately,
returning a 'mask' result of 0 (for false) or -1 (for true) in each byte.
Code generation support is added, in the form of a PPCISelDAGToDAG
DAG-preprocessing routine, that recognizes patterns close to what the
instruction computes (either exactly, or related by a constant masking
operation), and generates the cmpb instruction (along with any necessary
constant masking operation). This can be expanded if use cases arise.
llvm-svn: 225106
This is necessary to allow the disassembler to be able to handle AdSize32 instructions in 64-bit mode when address size prefix is used.
Eventually we should probably also support 'addr32' and 'addr16' in the assembler to override the address size on some of these instructions. But for now we'll just use special operand types that will lookup the current mode size to select the right instruction.
llvm-svn: 225075
a pre-splitting pass over loads and stores.
Historically, splitting could cause enough problems that I hamstrung the
entire process with a requirement that splittable integer loads and
stores must cover the entire alloca. All smaller loads and stores were
unsplittable to prevent chaos from ensuing. With the new pre-splitting
logic that does load/store pair splitting I introduced in r225061, we
can now very nicely handle arbitrarily splittable loads and stores. In
order to fully benefit from these smarts, we need to mark all of the
integer loads and stores as splittable.
However, we don't actually want to rewrite partitions with all integer
loads and stores marked as splittable. This will fail to extract scalar
integers from aggregates, which is kind of the point of SROA. =] In
order to resolve this, what we really want to do is only do
pre-splitting on the alloca slices with integer loads and stores fully
splittable. This allows us to uncover all non-integer uses of the alloca
that would benefit from a split in an integer load or store (and where
introducing the split is safe because it is just memory transfer from
a load to a store). Once done, we make all the non-whole-alloca integer
loads and stores unsplittable just as they have historically been,
repartition and rewrite.
The result is that when there are integer loads and stores anywhere
within an alloca (such as from a memcpy of a sub-object of a larger
object), we can split them up if there are non-integer components to the
aggregate hiding beneath. I've added the challenging test cases to
demonstrate how this is able to promote to scalars even a case where we
have even *partially* overlapping loads and stores.
This restores the single-store behavior for small arrays of i8s which is
really nice. I've restored both the little endian testing and big endian
testing for these exactly as they were prior to r225061. It also forced
me to be more aggressive in an alignment test to actually defeat SROA.
=] Without the added volatiles there, we actually split up the weird i16
loads and produce nice double allocas with better alignment.
This also uncovered a number of bugs where we failed to handle
splittable load and store slices which didn't have a begininng offset of
zero. Those fixes are included, and without them the existing test cases
explode in glorious fireworks. =]
I've kept support for leaving whole-alloca integer loads and stores as
splittable even for the purpose of rewriting, but I think that's likely
no longer needed. With the new pre-splitting, we might be able to remove
all the splitting support for loads and stores from the rewriter. Not
doing that in this patch to try to isolate any performance regressions
that causes in an easy to find and revert chunk.
llvm-svn: 225074
stores.
When there are accesses to an entire alloca with an integer
load or store as well as accesses to small pieces of the alloca, SROA
splits up the large integer accesses. In order to do that, it uses bit
math to merge the small accesses into large integers. While this is
effective, it produces insane IR that can cause significant problems in
the rest of the optimizer:
- It can cause load and store mismatches with GVN on the non-alloca side
where we end up loading an i64 (or some such) rather than loading
specific elements that are stored.
- We can't always get rid of the integer bit math, which is why we can't
always fix the loads and stores to work well with GVN.
- This is especially bad when we have operations that mix poorly with
integer bit math such as floating point operations.
- It will block things like the vectorizer which might be able to handle
the scalar stores that underly the aggregate.
At the same time, we can't just directly split up these loads and stores
in all cases. If there is actual integer arithmetic involved on the
values, then using integer bit math is actually the perfect lowering
because we can often combine it heavily with the surrounding math.
The solution this patch provides is to find places where SROA is
partitioning aggregates into small elements, and look for splittable
loads and stores that it can split all the way to some other adjacent
load and store. These are uniformly the cases where failing to split the
loads and stores hurts the optimizer that I have seen, and I've looked
extensively at the code produced both from more and less aggressive
approaches to this problem.
However, it is quite tricky to actually do this in SROA. We may have
loads and stores to the same alloca, or other complex patterns that are
hard to handle. This complexity leads to the somewhat subtle algorithm
implemented here. We have to do this entire process as a separate pass
over the partitioning of the alloca, and split up all of the loads prior
to splitting the stores so that we can handle safely the cases of
overlapping, including partially overlapping, loads and stores to the
same alloca. We also have to reconstitute the post-split slice
configuration so we can avoid iterating again over all the alloca uses
(the slow part of SROA). But we also have to ensure that when we split
up loads and stores to *other* allocas, we *do* re-iterate over them in
SROA to adapt to the more refined partitioning now required.
With this, I actually think we can fix a long-standing TODO in SROA
where I avoided splitting as many loads and stores as probably should be
splittable. This limitation historically mitigated the fallout of all
the bad things mentioned above. Now that we have more intelligent
handling, I plan to remove the FIXME and more aggressively mark integer
loads and stores as splittable. I'll do that in a follow-up patch to
help with bisecting any fallout.
The net result of this change should be more fine-grained and accurate
scalars being formed out of aggregates. At the very least, Clang now
generates perfect code for this high-level test case using
std::complex<float>:
#include <complex>
void g1(std::complex<float> &x, float a, float b) {
x += std::complex<float>(a, b);
}
void g2(std::complex<float> &x, float a, float b) {
x -= std::complex<float>(a, b);
}
void foo(const std::complex<float> &x, float a, float b,
std::complex<float> &x1, std::complex<float> &x2) {
std::complex<float> l1 = x;
g1(l1, a, b);
std::complex<float> l2 = x;
g2(l2, a, b);
x1 = l1;
x2 = l2;
}
This code isn't just hypothetical either. It was reduced out of the hot
inner loops of essentially every part of the Eigen math library when
using std::complex<float>. Those loops would consistently and
pervasively hop between the floating point unit and the integer unit due
to bit math extraction and insertion of floating point values that were
"stored" in a 64-bit integer register around the loop backedge.
So far, this change has passed a bootstrap and I have done some other
testing and so far, no issues. That doesn't mean there won't be though,
so I'll be prepared to help with any fallout. If you performance swings
in particular, please let me know. I'm very curious what all the impact
of this change will be. Stay tuned for the follow-up to also split more
integer loads and stores.
llvm-svn: 225061
This is the second installment of improvements to instruction selection for "bit
permutation" instruction sequences. r224318 added logic for instruction
selection for 32-bit bit permutation sequences, and this adds lowering for
64-bit sequences. The 64-bit sequences are more complicated than the 32-bit
ones because:
a) the 64-bit versions of the 32-bit rotate-and-mask instructions
work by replicating the lower 32-bits of the value-to-be-rotated into the
upper 32 bits -- and integrating this into the cost modeling for the various
bit group operations is non-trivial
b) unlike the 32-bit instructions in 32-bit mode, the rotate-and-mask instructions
cannot, in one instruction, specify the
mask starting index, the mask ending index, and the rotation factor. Also,
forming arbitrary 64-bit constants is more complicated than in 32-bit mode
because the number of instructions necessary is value dependent.
Plus, support for 'late masking' was added: it is sometimes more efficient to
treat the overall value as if it had no mandatory zero bits when planning the
bit-group insertions, and then mask them in at the very end. Unfortunately, as
the structure of the bit groups is different in the two cases, the more
feasible implementation technique was to generate both instruction sequences,
and then pick the shorter one.
And finally, we now generate reasonable code for i64 bswap:
rldicl 5, 3, 16, 0
rldicl 4, 3, 8, 0
rldicl 6, 3, 24, 0
rldimi 4, 5, 8, 48
rldicl 5, 3, 32, 0
rldimi 4, 6, 16, 40
rldicl 6, 3, 48, 0
rldimi 4, 5, 24, 32
rldicl 5, 3, 56, 0
rldimi 4, 6, 40, 16
rldimi 4, 5, 48, 8
rldimi 4, 3, 56, 0
vs. what we used to produce:
li 4, 255
rldicl 5, 3, 24, 40
rldicl 6, 3, 40, 24
rldicl 7, 3, 56, 8
sldi 8, 3, 8
sldi 10, 3, 24
sldi 12, 3, 40
rldicl 0, 3, 8, 56
sldi 9, 4, 32
sldi 11, 4, 40
sldi 4, 4, 48
andi. 5, 5, 65280
andis. 6, 6, 255
andis. 7, 7, 65280
sldi 3, 3, 56
and 8, 8, 9
and 4, 12, 4
and 9, 10, 11
or 6, 7, 6
or 5, 5, 0
or 3, 3, 4
or 7, 9, 8
or 4, 6, 5
or 3, 3, 7
or 3, 3, 4
which is 12 instructions, instead of 25, and seems optimal (at least in terms
of code size).
llvm-svn: 225056
Rafael Espindola