For historic reasons, the behavior of .align differs between targets.
Fortunately, there are alternatives, .p2align and .balign, which make the
interpretation of the parameter explicit, and which behave consistently across
targets.
This patch teaches MC to use .p2align instead of .align, so that people reading
code for multiple architectures don't have to remember which way each platform
does its .align directive.
Differential Revision: http://reviews.llvm.org/D16549
llvm-svn: 258750
The global entry point prologue currently assumes that the TOC
associated with a function is less than 2GB away from the function
entry point. This is always true when using the medium or small
code model, but may not be the case when using the large code model.
This patch adds a new variant of the ELFv2 global entry point prologue
that lifts the 2GB restriction when building with -mcmodel=large.
This works by emitting a quadword containing the distance from the
function entry point to its associated TOC immediately before the
entry point, and then using a prologue like:
ld r2,-8(r12)
add r2,r2,r12
Since creation of the entry point prologue is now split across two
separate routines (PPCLinuxAsmPrinter::EmitFunctionEntryLabel emits
the data word, PPCLinuxAsmPrinter::EmitFunctionBodyStart the prolog
code), I've switched to using named labels instead of just temporaries
to indicate the locations of the global and local entry points and the
new TOC offset data word.
These names are provided by new routines in PPCFunctionInfo modeled
after the existing PPCFunctionInfo::getPICOffsetSymbol.
Note that a corresponding change was committed to GCC here:
https://gcc.gnu.org/ml/gcc-patches/2015-12/msg00355.html
Reviewers: hfinkel
Differential Revision: http://reviews.llvm.org/D15500
llvm-svn: 257597
Only non-weighted predicates were handled in PPCInstrInfo::insertSelect. Handle
the weighted predicates as well.
This latent bug was triggered by r255398, because it added use of the
branch-weighted predicates.
While here, switch over an enum instead of an int to get the compiler to enforce
totality in the future.
llvm-svn: 257518
This is a fix for bug http://llvm.org/bugs/show_bug.cgi?id=25839.
For a PIC TLS variable access in a function, prologue (mflr followed by std and
stdu) gets scheduled after a tls_get_addr call. tls_get_addr messed up LR but
no one saves/restores it.
Also added a test for save/restore clobbered registers during calling __tls_get_addr.
Patch by Tim Shen
llvm-svn: 257137
This is the second in a set of patches for soft float support for ppc32,
it enables soft float operations.
Patch by Strahinja Petrovic.
Differential Revision: http://reviews.llvm.org/D13700
llvm-svn: 255516
This branch adds hints for highly biased branches on the PPC architecture. Even
in absence of profiling information, LLVM will mark code reaching unreachable
terminators and other exceptional control flow constructs as highly unlikely to
be reached.
Patch by Tom Jablin!
llvm-svn: 255398
Access to aligned globals gives us a chance to peephole optimize nonzero
offsets. If a struct is 4 byte aligned, then accesses to bytes 0-3 won't
overflow the available displacement. For example:
addis 3, 2, b4v@toc@ha
addi 4, 3, b4v@toc@l
lbz 5, b4v@toc@l(3) ; This is the result of the current peephole
lbz 6, 1(4) ; optimizer
lbz 7, 2(4)
lbz 8, 3(4)
If b4v is 4-byte aligned, we can skip using register 4 because we know
that b4v@toc@l+{1,2,3} won't overflow 32K, and instead generate:
addis 3, 2, b4v@toc@ha
lbz 4, b4v@toc@l(3)
lbz 5, b4v@toc@l+1(3)
lbz 6, b4v@toc@l+2(3)
lbz 7, b4v@toc@l+3(3)
Saving a register and an addition.
Larger alignments allow larger structures/arrays to be optimized.
llvm-svn: 255319
This was causing bad code gen and assembly that won't assemble, as
mixed altivec and vsx code would end up with a vsx high register
assigned to an altivec instruction, which won't work. Constraining the
classes allows the optimization to proceed.
llvm-svn: 255299
This patch corresponds to review:
http://reviews.llvm.org/D15286
LLVM IR frequently contains bitcast operations between floating point and
integer values of the same width. Doing this through memory operations is
quite expensive on PPC. This patch allows the use of direct register moves
between FPRs and GPRs for lowering bitcasts.
llvm-svn: 255246
AggressiveAntiDepBreaker was renaming registers specified by the user
for inline assembly. While this will work for compiler-specified
registers, it won't work for user-specified registers, and at the time
this runs, I don't currently see a way to distinguish them.
llvm-svn: 254532
The @llvm.get.dynamic.area.offset.* intrinsic family is used to get the offset
from native stack pointer to the address of the most recent dynamic alloca on
the caller's stack. These intrinsics are intendend for use in combination with
@llvm.stacksave and @llvm.restore to get a pointer to the most recent dynamic
alloca. This is useful, for example, for AddressSanitizer's stack unpoisoning
routines.
Patch by Max Ostapenko.
Differential Revision: http://reviews.llvm.org/D14983
llvm-svn: 254404
Re-enable shrink wrapping for PPC64 Little Endian.
One minor modification to PPCFrameLowering::findScratchRegister was necessary to handle fall-thru blocks (blocks with no terminator) correctly.
Tested with all LLVM test, clang tests, and the self-hosting build, with no problems found.
PHabricator: http://reviews.llvm.org/D14778
llvm-svn: 254314
The e500mc does not actually support the mfocrf instruction; update the
processor definitions to reflect that fact.
Patch by Tom Rix (with some test-case cleanup by me).
llvm-svn: 254064
The patch in http://reviews.llvm.org/D13745 is broken into four parts:
1. New interfaces without functional changes.
2. Use new interfaces in SelectionDAG, while in other passes treat probabilities
as weights.
3. Use new interfaces in all other passes.
4. Remove old interfaces.
This the second patch above. In this patch SelectionDAG starts to use
probability-based interfaces in MBB to add successors but other MC passes are
still using weight-based interfaces. Therefore, we need to maintain correct
weight list in MBB even when probability-based interfaces are used. This is
done by updating weight list in probability-based interfaces by treating the
numerator of probabilities as weights. This change affects many test cases
that check successor weight values. I will update those test cases once this
patch looks good to you.
Differential revision: http://reviews.llvm.org/D14361
llvm-svn: 253965
incorrect, as the chosen representative of the weak symbol may not live
with the code in question. Always indirect the access through the TOC
instead.
Patch by Kyle Butt!
llvm-svn: 253708
Note, this was reviewed (and more details are in) http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20151109/312083.html
These intrinsics currently have an explicit alignment argument which is
required to be a constant integer. It represents the alignment of the
source and dest, and so must be the minimum of those.
This change allows source and dest to each have their own alignments
by using the alignment attribute on their arguments. The alignment
argument itself is removed.
There are a few places in the code for which the code needs to be
checked by an expert as to whether using only src/dest alignment is
safe. For those places, they currently take the minimum of src/dest
alignments which matches the current behaviour.
For example, code which used to read:
call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 500, i32 8, i1 false)
will now read:
call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %dest, i8* align 8 %src, i32 500, i1 false)
For out of tree owners, I was able to strip alignment from calls using sed by replacing:
(call.*llvm\.memset.*)i32\ [0-9]*\,\ i1 false\)
with:
$1i1 false)
and similarly for memmove and memcpy.
I then added back in alignment to test cases which needed it.
A similar commit will be made to clang which actually has many differences in alignment as now
IRBuilder can generate different source/dest alignments on calls.
In IRBuilder itself, a new argument was added. Instead of calling:
CreateMemCpy(Dst, Src, getInt64(Size), DstAlign, /* isVolatile */ false)
you now call
CreateMemCpy(Dst, Src, getInt64(Size), DstAlign, SrcAlign, /* isVolatile */ false)
There is a temporary class (IntegerAlignment) which takes the source alignment and rejects
implicit conversion from bool. This is to prevent isVolatile here from passing its default
parameter to the source alignment.
Note, changes in future can now be made to codegen. I didn't change anything here, but this
change should enable better memcpy code sequences.
Reviewed by Hal Finkel.
llvm-svn: 253511
Several backends have instructions to reverse the order of bits in an integer. Conceptually matching such patterns is similar to @llvm.bswap, and it was mentioned in http://reviews.llvm.org/D14234 that it would be best if these patterns were matched in InstCombine instead of reimplemented in every different target.
This patch introduces an intrinsic @llvm.bitreverse.i* that operates similarly to @llvm.bswap. For plumbing purposes there is also a new ISD node ISD::BITREVERSE, with simple expansion and promotion support.
The intention is that InstCombine's BSWAP detection logic will be extended to support BITREVERSE too, and @llvm.bitreverse intrinsics emitted (if the backend supports lowering it efficiently).
llvm-svn: 252878
This patch adds a pass for doing PowerPC peephole optimizations at the
MI level while the code is still in SSA form. This allows for easy
modifications to the instructions while depending on a subsequent pass
of DCE. Both passes are very fast due to the characteristics of SSA.
At this time, the only peepholes added are for cleaning up various
redundancies involving the XXPERMDI instruction. However, I would
expect this will be a useful place to add more peepholes for
inefficiencies generated during instruction selection. The pass is
placed after VSX swap optimization, as it is best to let that pass
remove unnecessary swaps before performing any remaining clean-ups.
The utility of these clean-ups are demonstrated by changes to four
existing test cases, all of which now have tighter expected code
generation. I've also added Eric Schweiz's bugpoint-reduced test from
PR25157, for which we now generate tight code. One other test started
failing for me, and I've fixed it
(test/Transforms/PlaceSafepoints/finite-loops.ll) as well; this is not
related to my changes, and I'm not sure why it works before and not
after. The problem is that the CHECK-NOT: of "statepoint" from test1
fails because of the "statepoint" in test2, and so forth. Adding a
CHECK-LABEL in between keeps the different occurrences of that string
properly scoped.
llvm-svn: 252651
Under most circumstances, if SCEV can simplify X-Y to a constant, then it can
also simplify Y-X to a constant. However, there is no guarantee that this is
always true, and concensus is not to consider that a correctness bug in SCEV
(although it is undesirable).
PPCLoopPreIncPrep gathers pointers used to access memory (via loads, stores and
prefetches) into buckets, where in each bucket the relative pointer offsets are
constant. We used to keep each bucket as a multimap, where SCEV's subtraction
operation was used to define the ordering predicate. Instead, use a fixed SCEV
base expression for each bucket, record the constant offsets from that base
expression, and adjust it later, if desirable, once all pointers have been
collected.
Doing it this way should be more compile-time efficient than the previous
scheme (in addition to making the implementation less sensitive to SCEV
simplification quirks).
Fixes PR25170.
llvm-svn: 252417
Previously, subprograms contained a metadata reference to the function they
described. Because most clients need to get or set a subprogram for a given
function rather than the other way around, this created unneeded inefficiency.
For example, many passes needed to call the function llvm::makeSubprogramMap()
to build a mapping from functions to subprograms, and the IR linker needed to
fix up function references in a way that caused quadratic complexity in the IR
linking phase of LTO.
This change reverses the direction of the edge by storing the subprogram as
function-level metadata and removing DISubprogram's function field.
Since this is an IR change, a bitcode upgrade has been provided.
Fixes PR23367. An upgrade script for textual IR for out-of-tree clients is
attached to the PR.
Differential Revision: http://reviews.llvm.org/D14265
llvm-svn: 252219
This revision has introduced an issue that only affects bootstrapped compiler
when it is printing the ASM. It turns out that the new code path taken due to
legalizing a scalar_to_vector of i64 -> v2i64 exposes a missing check in a
micro optimization to change a load followed by a scalar_to_vector into a
load and splat instruction on PPC.
llvm-svn: 251798
We cannot form ctr-based loops around function calls, including calls to
__tls_get_addr used for PIC TLS variables. References to such TLS variables,
however, might be buried within constant expressions, and so we need to search
the entire constant expression to be sure that no references to such TLS
variables exist.
Fixes PR25256, reported by Eric Schweitz. This is a slightly-modified version
of the patch suggested by Eric in the bug report, and a test case I created.
llvm-svn: 251582
As a follow-up to r251566, do the same for the other optionally-supported
register classes (mostly for vector registers). Don't return an unavailable
register class (which would cause an assert later), but fail cleanly when
provided an unsupported inline asm constraint.
llvm-svn: 251575
cntlz is the old POWER mnemonic. cntlzw is the PowerPC mnemonic.
This change fixes an issue when -no-integrated-as: The opcode cntlz is
unrecognized by gas
Alias the POWER mnemonic cntlz[.] to the PowerPC mnemonic cntlzw[.]
This is done for because the POWER cntlz mnemonic has be used by LLVM for
a very long time. We need to make sure that assembly programs
that are using the cntlz[.] do not break with this change.
Change PowerPC tests to reflect the insn change from cntlz to cntlzw.
Add assembly test to verify cntlz[.] is encoded correctly.
Patch by Tom Rix!
llvm-svn: 251489
Summary:
Don't call `computeKnownBitsFromRangeMetadata` for extended loads --
this can cause a mismatch between the width of the !range metadata and
the width of the APInt's accumulating `KnownZero` (and `KnownOne` in the
future). This isn't a problem now, but will be after a future change.
Note: this can be made more aggressive in the future.
Reviewers: nlewycky
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D14107
llvm-svn: 251486
Recommit r250342: move coal-sections-powerpc.s to subdirectory for powerpc.
Some background on why we don't have to use *coal* sections anymore:
Long ago when C++ was new and "weak" had not been standardized, an attempt was
made in cctools to support C++ inlines that can be coalesced by putting them
into their own section (TEXT/textcoal_nt instead of TEXT/text).
The current macho linker supports the weak-def bit on any symbol to allow it to
be coalesced, but the compiler still puts weak-def functions/data into alternate
section names, which the linker must map back to the base section name.
This patch makes changes that are necessary to prevent the compiler from using
the "coal" sections and have it use the non-coal sections instead when the
target architecture is not powerpc:
TEXT/textcoal_nt instead use TEXT/text
TEXT/const_coal instead use TEXT/const
DATA/datacoal_nt instead use DATA/data
If the target is powerpc, we continue to use the *coal* sections since anyone
targeting powerpc is probably using an old linker that doesn't have support for
the weak-def bits.
Also, have the assembler issue a warning if it encounters a *coal* section in
the assembly file and inform the users to use the non-coal sections instead.
rdar://problem/14265330
Differential Revision: http://reviews.llvm.org/D13188
llvm-svn: 250370
Recommit r250342: add -arch=ppc32 to the RUN lines of powerpc tests.
Some background on why we don't have to use *coal* sections anymore:
Long ago when C++ was new and "weak" had not been standardized, an attempt was
made in cctools to support C++ inlines that can be coalesced by putting them
into their own section (TEXT/textcoal_nt instead of TEXT/text).
The current macho linker supports the weak-def bit on any symbol to allow it to
be coalesced, but the compiler still puts weak-def functions/data into alternate
section names, which the linker must map back to the base section name.
This patch makes changes that are necessary to prevent the compiler from using
the "coal" sections and have it use the non-coal sections instead when the
target architecture is not powerpc:
TEXT/textcoal_nt instead use TEXT/text
TEXT/const_coal instead use TEXT/const
DATA/datacoal_nt instead use DATA/data
If the target is powerpc, we continue to use the *coal* sections since anyone
targeting powerpc is probably using an old linker that doesn't have support for
the weak-def bits.
Also, have the assembler issue a warning if it encounters a *coal* section in
the assembly file and inform the users to use the non-coal sections instead.
rdar://problem/14265330
Differential Revision: http://reviews.llvm.org/D13188
llvm-svn: 250349
Some background on why we don't have to use *coal* sections anymore:
Long ago when C++ was new and "weak" had not been standardized, an attempt was
made in cctools to support C++ inlines that can be coalesced by putting them
into their own section (TEXT/textcoal_nt instead of TEXT/text).
The current macho linker supports the weak-def bit on any symbol to allow it to
be coalesced, but the compiler still puts weak-def functions/data into alternate
section names, which the linker must map back to the base section name.
This patch makes changes that are necessary to prevent the compiler from using
the "coal" sections and have it use the non-coal sections instead when the
target architecture is not powerpc:
TEXT/textcoal_nt instead use TEXT/text
TEXT/const_coal instead use TEXT/const
DATA/datacoal_nt instead use DATA/data
If the target is powerpc, we continue to use the *coal* sections since anyone
targeting powerpc is probably using an old linker that doesn't have support for
the weak-def bits.
Also, have the assembler issue a warning if it encounters a *coal* section in
the assembly file and inform the users to use the non-coal sections instead.
rdar://problem/14265330
Differential Revision: http://reviews.llvm.org/D13188
llvm-svn: 250342
PR25157 identifies a bug where a load plus a vector shuffle is
incorrectly converted into an LXVDSX instruction. That optimization
is only valid if the load is of a doubleword, and in the noted case,
it was not. This corrects that problem.
Joint patch with Eric Schweitz, who provided the bugpoint-reduced test
case.
llvm-svn: 250324
This patch corresponds to review:
http://reviews.llvm.org/D12032
This patch builds onto the patch that provided scalar to vector conversions
without stack operations (D11471).
Included in this patch:
- Vector element extraction for all vector types with constant element number
- Vector element extraction for v16i8 and v8i16 with variable element number
- Removal of some unnecessary COPY_TO_REGCLASS operations that ended up
unnecessarily moving things around between registers
Not included in this patch (will be in upcoming patch):
- Vector element extraction for v4i32, v4f32, v2i64 and v2f64 with
variable element number
- Vector element insertion for variable/constant element number
Testing is provided for all extractions. The extractions that are not
implemented yet are just placeholders.
llvm-svn: 249822
When AA is being used, non-aliasing stores are canonicalized to use the same
chain, and DAGCombiner::getStoreMergeAndAliasCandidates can take advantage of
this by looking only as users of a store's chain operand. However, user
iteration is not result-number specific, we need to check that the use is as a
chain operand, and not via some other operand. It is certainly possible to have
another potentially-aliasing store, which shares the first's base pointer, and
uses the first's chain's node via some other operand.
Failure to catch this situation caused, at least in the included test case, an
assert later because the relative sequence-number ordering caused later
replacement to create a cycle in the DAG.
llvm-svn: 248698
The algorithm would not modify the live-in list of blocks below the save
block point which is correct unless it happens to be a restore point at
the same time.
Also fixes the benign issue of live-in registers being added twice in
some cases.
The testcase is based on a test submitted by Kit Barton.
Differential Revision: http://reviews.llvm.org/D13176
llvm-svn: 248620
If storing multiple FP constants, some subset of the stores
would be replaced with integers due to visit order, so
MergeConsecutiveStores would only partially merge
these.
llvm-svn: 248169
- Strenghten the logic to be sure we hoist the restore point out of the current
loop. (The fixes a bug with infinite loop, added as part of the patch.)
- Walk over the exit blocks of the current loop to conver to the desired restore
point in one iteration of the update loop.
llvm-svn: 247958
When building LLVM as a (potentially dynamic) library that can be linked against
by multiple compilers, the default triple is not really meaningful.
We allow to explicitely set it to an empty string when configuring LLVM.
In this case, said "target independent" tests in the test suite that are using
the default triple are disabled by matching the newly available feature
"default_triple".
Reviewers: probinson, echristo
Differential Revision: http://reviews.llvm.org/D12660
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 247775
The changes in this patch are as follows:
1. Modify the emitPrologue and emitEpilogue methods to work properly when the prologue and epilogue blocks are not the first/last blocks in the function
2. Fix a bug in PPCEarlyReturn optimization caused by an empty entry block in the function
3. Override the runShrinkWrap PredicateFtor (defined in TargetMachine) to check whether shrink wrapping should run:
Shrink wrapping will run on PPC64 (Little Endian and Big Endian) unless -enable-shrink-wrap=false is specified on command line
A new test case, ppc-shrink-wrapping.ll was created based on the existing shrink wrapping tests for x86, arm, and arm64.
Phabricator review: http://reviews.llvm.org/D11817
llvm-svn: 247237
To commute a trivial rlwimi instructions (meaning one with a full mask and zero
shift), we'd need to ability to form an all-zero mask (instead of an all-one
mask) using rlwimi. We can't represent this, however, and we'll miscompile code
if we try.
The code quality problem that this highlights (that SDAG simplification can
lead to us generating an ISD::OR node with a constant zero LHS) will be fixed
as a follow-up.
Fixes PR24719.
llvm-svn: 246937
PPCISelDAGToDAG has a transformation that generates a rlwimi instruction from
an input pattern that looks like this:
and(or(x, c1), c2)
but the associated logic does not work if there are bits that are 1 in c1 but 0
in c2 (these are normally canonicalized away, but that can't happen if the 'or'
has other users. Make sure we abort the transformation if such bits are
discovered.
Fixes PR24704.
llvm-svn: 246900
When forming permutation-based unaligned vector loads, we need to know whether
it is valid to read ahead of the requested address by a full vector length.
Doing so is more efficient (and allows for more CSE with later loads), but
could trigger a page fault if invalid. To determine validity, we look for other
loads in the same block that access the relevant address range.
The relevant point here is that we need to do this as part of the process of
forming permutation-based vector loads, and this happens quite early in the
SDAG pipeline - specifically before many of the address calculations are fully
canonicalized. As a result, we need to try harder to recognize base+offset
address computations, because they still might appear as chain of adds
(base+offset+offset, for example). To account for this, we'll look through
chains of adds, accumulating the constant offsets.
llvm-svn: 246813
If you compute the MMO offset using unsigned arithmetic, you end up with a
large positive offset instead of a small negative one. In theory, this could
cause bad instruction-scheduling decisions later.
I noticed this by inspection from the debug output, and using that for the
regression test is the best I can do right now.
llvm-svn: 246805
I'm adding a regression test to better cover code generation for unaligned
vector loads and stores, but there's no functional change to the code
generation here. There is an improvement to the cost model for unaligned vector
loads and stores, mostly for QPX (for which we were not previously accounting
for the permutation-based loads), and the cost model implementation is cleaner.
llvm-svn: 246712
LowerVECTOR_SHUFFLE needs to decide whether to pass a vector shuffle off to the
TableGen-generated matching code, and it does this by testing the same
predicates used by the TableGen files. Unfortunately, when we added new
P8Altivec-only predicates, we started universally testing them in
LowerVECTOR_SHUFFLE, and if then matched when targeting a system prior to a P8,
we'd end up with a selection failure.
llvm-svn: 246675
SETCC is one of those special node types for which operation actions (legality,
etc.) is keyed off of an operand type, not the node's value type. This makes
sense because the value type of a legal SETCC node is determined by its
operands' value type (via the TLI function getSetCCResultType). When the
SDAGBuilder creates SETCC nodes, it either creates them with an MVT::i1 value
type, or directly with the value type provided by TLI.getSetCCResultType.
The first problem being fixed here is that DAGCombine had several places
querying TLI.isOperationLegal on SETCC, but providing the return of
getSetCCResultType, instead of the operand type directly. This does not mean
what the author thought, and "luckily", most in-tree targets have SETCC with
Custom lowering, instead of marking them Legal, so these checks return false
anyway.
The second problem being fixed here is that two of the DAGCombines could create
SETCC nodes with arbitrary (integer) value types; specifically, those that
would simplify:
(setcc a, b, op1) and|or (setcc a, b, op2) -> setcc a, b, op3
(which is possible for some combinations of (op1, op2))
If the operands of the and|or node are actual setcc nodes, then this is not an
issue (because the and|or must share the same type), but, the relevant code in
DAGCombiner::visitANDLike and DAGCombiner::visitORLike actually calls
DAGCombiner::isSetCCEquivalent on each operand, and that function will
recognise setcc-like select_cc nodes with other return types. And, thus, when
creating new SETCC nodes, we need to be careful to respect the value-type
constraint. This is even true before type legalization, because it is quite
possible for the SELECT_CC node to have a legal type that does not happen to
match the corresponding TLI.getSetCCResultType type.
To be explicit, there is nothing that later fixes the value types of SETCC
nodes (if the type is legal, but does not happen to match
TLI.getSetCCResultType). Creating SETCCs with an MVT::i1 value type seems to
work only because, either MVT::i1 is not legal, or it is what
TLI.getSetCCResultType returns if it is legal. Fixing that is a larger change,
however. For the time being, restrict the relevant transformations to produce
only SETCC nodes with a value type matching TLI.getSetCCResultType (or MVT::i1
prior to type legalization).
Fixes PR24636.
llvm-svn: 246507
AggressiveAntiDepBreaker was doing some EarlyClobber checking, but was not
checking that the register being potentially renamed was defined by an
early-clobber def where there was also a use, in that instruction, of the
register being considered as the target of the rename. Fixes PR24014.
llvm-svn: 246423
There were really two problems here. The first was that we had the truth tables
for signed i1 comparisons backward. I imagine these are not very common, but if
you have:
setcc i1 x, y, LT
this has the '0 1' and the '1 0' results flipped compared to:
setcc i1 x, y, ULT
because, in the signed case, '1 0' is really '-1 0', and the answer is not the
same as in the unsigned case.
The second problem was that we did not have patterns (at all) for the unsigned
comparisons select_cc nodes for i1 comparison operands. This was the specific
cause of PR24552. These had to be added (and a missing Altivec promotion added
as well) to make sure these function for all types. I've added a bunch more
test cases for these patterns, and there are a few FIXMEs in the test case
regarding code-quality.
Fixes PR24552.
llvm-svn: 246400
Add support for MIR serialization of PowerPC-specific operand target flags
(based on the generic infrastructure added in r244185 and r245383).
I won't even pretend that this is good test coverage, but this includes the
regression test associated with r246372. Adding an MIR test for that fix is far
superior to adding an IR-level test because particular instruction-scheduling
decisions are necessary in order to expose the bug, and using an MIR test we
can start the pipeline post-scheduling.
llvm-svn: 246373
As a follow-up to r246098, require `DISubprogram` definitions
(`isDefinition: true`) to be 'distinct'. Specifically, add an assembler
check, a verifier check, and bitcode upgrading logic to combat testcase
bitrot after the `DIBuilder` change.
While working on the testcases, I realized that
test/Linker/subprogram-linkonce-weak-odr.ll isn't relevant anymore. Its
purpose was to check for a corner case in PR22792 where two subprogram
definitions match exactly and share the same metadata node. The new
verifier check, requiring that subprogram definitions are 'distinct',
precludes that possibility.
I updated almost all the IR with the following script:
git grep -l -E -e '= !DISubprogram\(.* isDefinition: true' |
grep -v test/Bitcode |
xargs sed -i '' -e 's/= \(!DISubprogram(.*, isDefinition: true\)/= distinct \1/'
Likely some variant of would work for out-of-tree testcases.
llvm-svn: 246327
When combiner AA is enabled, look at stores on the same chain.
Non-aliasing stores are moved to the same chain so the existing
code fails because it expects to find an adajcent store on a consecutive
chain.
Because of how DAGCombiner tries these store combines,
MergeConsecutiveStores doesn't see the correct set of stores on the chain
when it visits the other stores. Each store individually has its chain
fixed before trying to merge consecutive stores, and then tries to merge
stores from that point before the other stores have been processed to
have their chains fixed. To fix this, attempt to use FindBetterChain
on any possibly neighboring stores in visitSTORE.
Suppose you have 4 32-bit stores that should be merged into 1 vector
store. One store would be visited first, fixing the chain. What happens is
because not all of the store chains have yet been fixed, 2 of the stores
are merged. The other 2 stores later have their chains fixed,
but because the other stores were already merged, they have different
memory types and merging the two different sized stores is not
supported and would be more difficult to handle.
llvm-svn: 246307
We might end up with a trivial copy as the addend, and if so, we should ignore
the corresponding FMA instruction. The trivial copy can be coalesced away later,
so there's nothing to do here. We should not, however, assert. Fixes PR24544.
llvm-svn: 245907
This patch fixes PR24546, which demonstrates a segfault during the VSX
swap removal pass. The problem is that debug value instructions were
not excluded from the list of instructions to be analyzed for webs of
related computation. I've added the test case from the PR as a crash
test in test/CodeGen/PowerPC.
llvm-svn: 245862
When PPCVSXFMAMutate would look at the input addend register, it would get its
input value number. This would fail, however, if the register was undef,
causing a segfault. Don't segfault (just skip such FMA instructions).
Fixes the test case from PR24542 (although that may have been over-reduced).
llvm-svn: 245741
XVCMPEQDP is used for VSX v2f64 equality comparisons, but the value type needs
to be v2i64 (as that's the corresponding SETCC type).
Fixes PR24225.
llvm-svn: 245535
This DAGCombine was creating custom SDAG nodes with an illegal ppc_fp128
operand type because it was triggering on f64/f32 int2fp(fp2int(ppc_fp128 x)),
but shouldn't (it should only apply to f32/f64 types). The result was a crash.
llvm-svn: 245530
This revision has introduced an issue that only affects bootstrapped compiler
when it is printing the ASM. I am working on resolving the issue, but in the
meantime, I'm disabling the legalization of scalar_to_vector operation for v2i64
and the associated testing until I can get this fixed.
llvm-svn: 245481
This patch corresponds to review:
http://reviews.llvm.org/D11471
It improves the code generated for converting a scalar to a vector value. With
direct moves from GPRs to VSRs, we no longer require expensive stack operations
for this. Subsequent patches will handle the reverse case and more general
operations between vectors and their scalar elements.
llvm-svn: 244921
frame setup instruction.
This commit ensures that the stack map lowering code in FastISel adds an
appropriate number of immediate operands to the frame setup instruction.
The previous code added just one immediate operand, which was fine for a target
like AArch64, but on X86 the ADJCALLSTACKDOWN64 instruction needs two explicit
operands. This caused the machine verifier to report an error when the old code
added just one.
Reviewers: Juergen Ributzka
Differential Revision: http://reviews.llvm.org/D11853
llvm-svn: 244508
I looked into adding a warning / error for this to FileCheck, but there doesn't
seem to be a good way to avoid it triggering on the instances of it in RUN lines.
llvm-svn: 244481
In r242277, I updated the MachineCombiner to work with itineraries, but I
missed a call that is scheduling-model-only (the opcode-only form of
computeInstrLatency). Using the form that takes an MI* allows this to work with
itineraries (and should be NFC for subtargets with scheduling models).
llvm-svn: 244020
Since r241097, `DIBuilder` has only created distinct `DICompileUnit`s.
The backend is liable to start relying on that (if it hasn't already),
so make uniquable `DICompileUnit`s illegal and automatically upgrade old
bitcode. This is a nice cleanup, since we can remove an unnecessary
`DenseSet` (and the associated uniquing info) from `LLVMContextImpl`.
Almost all the testcases were updated with this script:
git grep -e '= !DICompileUnit' -l -- test |
grep -v test/Bitcode |
xargs sed -i '' -e 's,= !DICompileUnit,= distinct !DICompileUnit,'
I imagine something similar should work for out-of-tree testcases.
llvm-svn: 243885
Remove the fake `DW_TAG_auto_variable` and `DW_TAG_arg_variable` tags,
using `DW_TAG_variable` in their place Stop exposing the `tag:` field at
all in the assembly format for `DILocalVariable`.
Most of the testcase updates were generated by the following sed script:
find test/ -name "*.ll" -o -name "*.mir" |
xargs grep -l 'DILocalVariable' |
xargs sed -i '' \
-e 's/tag: DW_TAG_arg_variable, //' \
-e 's/tag: DW_TAG_auto_variable, //'
There were only a handful of tests in `test/Assembly` that I needed to
update by hand.
(Note: a follow-up could change `DILocalVariable::DILocalVariable()` to
set the tag to `DW_TAG_formal_parameter` instead of `DW_TAG_variable`
(as appropriate), instead of having that logic magically in the backend
in `DbgVariable`. I've added a FIXME to that effect.)
llvm-svn: 243774
Given certain shuffle-vector masks, LLVM emits splat instructions
which splat the wrong bytes from the source register. The issue is
that the function PPC::isSplatShuffleMask() in PPCISelLowering.cpp
does not ensure that the splat pattern found is requesting bytes that
are aligned on an EltSize boundary. This patch detects this situation
as not a valid splat mask, resulting in a permute being generated
instead of a splat.
Patch and test case by Tyler Kenney, cleaned up a bit by me.
This is a simple bug fix that would be good to incorporate into 3.7.
llvm-svn: 243519
extension property we're requesting - zero or sign extended.
This fixes cases where we want to return a zero extended 32-bit -1
and not be sign extended for the entire register. Also updated the
already out of date comment with the current behavior.
llvm-svn: 243192
This makes one substantive change and a few stylistic changes to the
VSX swap optimization pass.
The substantive change is to permit LXSDX and LXSSPX instructions to
participate in swap optimization computations. The previous change to
insert a swap following a SUBREG_TO_REG widening operation makes this
almost trivial.
I experimented with also permitting STXSDX and STXSSPX instructions.
This can be done using similar techniques: we could insert a swap
prior to a narrowing COPY operation, and then permit these stores to
participate. I prototyped this, but discovered that the pattern of a
narrowing COPY followed by an STXSDX does not occur in any of our
test-suite code. So instead, I added commentary indicating that this
could be done.
Other TLC:
- I changed SH_COPYSCALAR to SH_COPYWIDEN to more clearly indicate
the direction of the copy.
- I factored the insertion of swap instructions into a separate
function.
Finally, I added a new test case to check that the scalar-to-vector
loads are working properly with swap optimization.
llvm-svn: 242838
I was looking at some vector code generation and kept seeing
unnecessary vector copies into the Altivec half of the VSX registers.
I discovered that we overlooked v4i32 when adding the register classes
for VSX; we only added v4f32 and v2f64. This means that anything that
canonicalizes into v4i32 (which is a LOT of stuff) ends up being
forced into VRRC on its way to VSRC.
The fix is one line. The rest of the patch is fixing up some test
cases whose code generation has changed as a result.
This seems like it would be a good candidate for backport to 3.7.
llvm-svn: 242442
This is a direct port of the code from the X86 backend (r239486/r240361), which
uses the MachineCombiner to reassociate (floating-point) adds/muls to increase
ILP, to the PowerPC backend. The rationale is the same.
There is a lot of copy-and-paste here between the X86 code and the PowerPC
code, and we should extract at least some of this into CodeGen somewhere.
However, I don't want to do that until this code is enhanced to handle FMAs as
well. After that, we'll be in a better position to extract the common parts.
llvm-svn: 242279
Follow-up r235483, with the corresponding support in PPC. We use a regular call
for symbolic targets (because they're much cheaper than indirect calls).
llvm-svn: 242239
We used to take the address specified as the direct target of the patchpoint
and did no TOC-pointer handling. This, however, as not all that useful,
because MCJIT tends to create a lot of modules, and they have their own TOC
sections. Thus, to call from the generated code to other generated code, you
really need to switch TOC pointers. Make this work as expected, and under
ELFv1, tread the address as the function descriptor address so that the correct
TOC pointer can be loaded.
llvm-svn: 242217
PowerPC uses itineraries to describe processor pipelines (and dispatch-group
restrictions for P7/P8 cores). Unfortunately, the target-independent
implementation of TII.getInstrLatency calls ItinData->getStageLatency, and that
looks for the largest cycle count in the pipeline for any given instruction.
This, however, yields the wrong answer for the PPC itineraries, because we
don't encode the full pipeline. Because the functional units are fully
pipelined, we only model the initial stages (there are no relevant hazards in
the later stages to model), and so the technique employed by getStageLatency
does not really work. Instead, we should take the maximum output operand
latency, and that's what PPCInstrInfo::getInstrLatency now does.
This caused some test-case churn, including two unfortunate side effects.
First, the new arrangement of copies we get from function parameters now
sometimes blocks VSX FMA mutation (a FIXME has been added to the code and the
test cases), and we have one significant test-suite regression:
SingleSource/Benchmarks/BenchmarkGame/spectral-norm
56.4185% +/- 18.9398%
In this benchmark we have a loop with a vectorized FP divide, and it with the
new scheduling both divides end up in the same dispatch group (which in this
case seems to cause a problem, although why is not exactly clear). The grouping
structure is hard to predict from the bottom of the loop, and there may not be
much we can do to fix this.
Very few other test-suite performance effects were really significant, but
almost all weakly favor this change. However, in light of the issues
highlighted above, I've left the old behavior available via a
command-line flag.
llvm-svn: 242188
This patch allows VSX swap optimization to succeed more frequently.
Specifically, it is concerned with common code sequences that occur
when copying a scalar floating-point value to a vector register. This
patch currently handles cases where the floating-point value is
already in a register, but does not yet handle loads (such as via an
LXSDX scalar floating-point VSX load). That will be dealt with later.
A typical case is when a scalar value comes in as a floating-point
parameter. The value is copied into a virtual VSFRC register, and
then a sequence of SUBREG_TO_REG and/or COPY operations will convert
it to a full vector register of the class required by the context. If
this vector register is then used as part of a lane-permuted
computation, the original scalar value will be in the wrong lane. We
can fix this by adding a swap operation following any widening
SUBREG_TO_REG operation. Additional COPY operations may be needed
around the swap operation in order to keep register assignment happy,
but these are pro forma operations that will be removed by coalescing.
If a scalar value is otherwise directly referenced in a computation
(such as by one of the many XS* vector-scalar operations), we
currently disable swap optimization. These operations are
lane-sensitive by definition. A MentionsPartialVR flag is added for
use in each swap table entry that mentions a scalar floating-point
register without having special handling defined.
A common idiom for PPC64LE is to convert a double-precision scalar to
a vector by performing a splat operation. This ensures that the value
can be referenced as V[0], as it would be for big endian, whereas just
converting the scalar to a vector with a SUBREG_TO_REG operation
leaves this value only in V[1]. A doubleword splat operation is one
form of an XXPERMDI instruction, which takes one doubleword from a
first operand and another doubleword from a second operand, with a
two-bit selector operand indicating which doublewords are chosen. In
the general case, an XXPERMDI can be permitted in a lane-swapped
region provided that it is properly transformed to select the
corresponding swapped values. This transformation is to reverse the
order of the two input operands, and to reverse and complement the
bits of the selector operand (derivation left as an exercise to the
reader ;).
A new test case that exercises the scalar-to-vector and generalized
XXPERMDI transformations is added as CodeGen/PowerPC/swaps-le-5.ll.
The patch also requires a change to CodeGen/PowerPC/swaps-le-3.ll to
use CHECK-DAG instead of CHECK for two independent instructions that
now appear in reverse order.
There are two small unrelated changes that are added with this patch.
First, the XXSLDWI instruction was incorrectly omitted from the list
of lane-sensitive instructions; this is now fixed. Second, I observed
that the same webs were being rejected over and over again for
different reasons. Since it's sufficient to reject a web only once, I
added a check for this to speed up the compilation time slightly.
llvm-svn: 242081
r238842 added the TargetRecip system for controlling use of reciprocal
estimates for sqrt and division using a set of parameters that can be set by
the frontend. Clang now supports a sophisticated -mrecip option, and this will
allow that option to effectively control the relevant code-generation
functionality of the PPC backend.
llvm-svn: 241985
This adds support for the 'nest' attribute, which allows the static chain
register to be set for functions calls under non-Darwin PPC/PPC64 targets. r11
is the chain register (which the PPC64 ELF ABI calls the "environment
pointer"). For indirect calls under PPC64 ELFv1, this would normally be loaded
from the function descriptor, but providing an explicit 'nest' parameter will
override that process and use the value provided.
This allows __builtin_call_with_static_chain to work as expected on PowerPC.
llvm-svn: 241984
In r241285, I removed the SUBREG_TO_REG restriction from VSX swap
removal, determining that this was overly conservative. We have
another form of the same restriction in that we check for the presence
of implicit subregs in vector operations. As with SUBREG_TO_REG for
partial register conversions, an implicit subreg is safe in and of
itself, provided no other operation makes a lane-sensitive assumption
about the result. This patch removes that restriction, by removing
the HasImplicitSubreg flag and all code that relies on it.
I've added a test case that fails to optimize before this patch is
applied, and optimizes properly with the patch. Test based on a
report from Anton Blanchard.
llvm-svn: 241290
With a previous patch, the VSX swap optimization is able to recognize
the doubleword load-splat idiom that can be implemented using lxvdsx.
However, that does not cover a doubleword splat where the source is a
register. We can implement this using xxspltd (a special form of
xxpermdi). This patch teaches the swap optimization pass about this
idiom.
As a prerequisite, it also permits swap optimization to succeed for
all forms of SUBREG_TO_REG. Previously we were conservative and only
allowed SUBREG_TO_REG when it copied a full register. However, on
reflection any form of SUBREG_TO_REG is safe in and of itself, so long
as an unsafe operation is not performed on its result. In particular,
a widening SUBREG_TO_REG often occurs as an input to a doubleword
splat idiom, particularly in auto-vectorized code.
The doubleword splat idiom is an XXPERMDI operation where both source
registers are identical, and the selection mask is either 0 (splat the
first element) or 3 (splat the second element). To determine whether
the registers are identical, we use the existing mechanism for looking
through "copy-like" operations. That mechanism has a side effect of
marking the XXPERMDI operation as using a physical register, which
would invalidate its presence in a swap-optimized region. This is
correct for the form of XXPERMDI that performs a swap and hence would
be removed, but is not what we want for a doubleword-splat variety of
XXPERMDI. Therefore we reset the physical-register flag on the
XXPERMDI when it represents a splat.
A simple test case is added to verify that we generate the splat and
that we also remove the xxswapd instructions that would otherwise be
associated with the load and store of another operand.
llvm-svn: 241285
When adding little-endian vector support for PowerPC last year, I
inadvertently disabled an optimization that recognizes a load-splat
idiom and generates the lxvdsx instruction. This patch moves the
offending logic so lxvdsx is once again generated.
This pattern is frequently generated by the vectorizer for scalar
loads of an effective constant. Previously the lxvdsx instruction was
wrongly listed as lane-sensitive for the VSX swap optimization (since
both doublewords are identical, swaps are safe). This patch fixes
this as well, so that vectorized code using lxvdsx can now have swaps
removed from the computation.
There is an existing test (@test50) in test/CodeGen/PowerPC/vsx.ll
that checks for the missing optimization. However, vsx.ll was only
being tested for POWER7 with big-endian code generation. I've added
a little-endian RUN statement and expected LE code generation for all
the tests in vsx.ll to give us a bit better VSX coverage, including
what's needed for this patch.
llvm-svn: 241183
This patch corresponds to review:
http://reviews.llvm.org/D10638
This is the back end portion of patch
http://reviews.llvm.org/D10637
It just adds the code gen and intrinsic functions necessary to support that patch to the back end.
llvm-svn: 240820
This patch adds support for the vector merge even word and vector merge odd word
instructions introduced in POWER8.
Phabricator review: http://reviews.llvm.org/D10704
llvm-svn: 240650
In a relocation target can take 3 basic forms
* A r_value in scattered relocations.
* A symbol in external relocations.
* A section is non-external relocations.
Have the dump reflect that. With this change we go from
CHECK-NEXT: Extern: 0
CHECK-NEXT: Type: X86_64_RELOC_SUBTRACTOR (5)
CHECK-NEXT: Symbol: 0x2
CHECK-NEXT: Scattered: 0
To just
// CHECK-NEXT: Type: X86_64_RELOC_SUBTRACTOR (5)
// CHECK-NEXT: Section: __data (2)
Since the relocation is with a section, we print the seciton name and don't
need to say that it is not scattered or external.
Someone motivated can add further special cases for things like
ARM64_RELOC_ADDEND and ARM_RELOC_PAIR.
llvm-svn: 240073
The personality routine currently lives in the LandingPadInst.
This isn't desirable because:
- All LandingPadInsts in the same function must have the same
personality routine. This means that each LandingPadInst beyond the
first has an operand which produces no additional information.
- There is ongoing work to introduce EH IR constructs other than
LandingPadInst. Moving the personality routine off of any one
particular Instruction and onto the parent function seems a lot better
than have N different places a personality function can sneak onto an
exceptional function.
Differential Revision: http://reviews.llvm.org/D10429
llvm-svn: 239940
The mftb instruction was incorrectly marked as deprecated in the PPC
Backend. Instead, it should not be treated as deprecated, but rather be
implemented using the mfspr instruction. A similar patch was put into GCC last
year. Details can be found at:
https://sourceware.org/ml/binutils/2014-11/msg00383.html.
This change will replace instances of the mftb instruction with the mfspr
instruction for all CPUs except 601 and pwr3. This will also be the default
behaviour.
Additional details can be found in:
https://llvm.org/bugs/show_bug.cgi?id=23680
Phabricator review: http://reviews.llvm.org/D10419
llvm-svn: 239827
This patch corresponds to review:
http://reviews.llvm.org/D10096
This is the back end portion of the patch related to D10095.
The patch adds the instructions and back end intrinsics for:
vbpermq
vgbbd
llvm-svn: 239505
This patch corresponds to review:
http://reviews.llvm.org/D9941
It adds the various FMA instructions introduced in the version 2.07 of
the ISA along with the testing for them. These are operations on single
precision scalar values in VSX registers.
llvm-svn: 238578
in POWER8:
vadduqm
vaddeuqm
vaddcuq
vaddecuq
vsubuqm
vsubeuqm
vsubcuq
vsubecuq
In addition to adding the instructions themselves, it also adds support for the
v1i128 type for intrinsics (Intrinsics.td, Function.cpp, and
IntrinsicEmitter.cpp).
http://reviews.llvm.org/D9081
llvm-svn: 238144
When the compare feeding a branch was in a different BB from the branch, we'd
try to "regenerate" the compare in the block with the branch, possibly trying
to make use of values not available there. Copy a page from AArch64's play book
here to fix the problem (at least in terms of correctness).
Fixes PR23640.
llvm-svn: 238097
My recent patch to add support for ISA 2.07 vector pack/unpack
instructions didn't properly check for availability of the vpkudum
instruction when recognizing it as a special vector shuffle case.
This causes us to leave the vector shuffle in place (rather than
converting it to a vector permute) so that it can be recognized later
as a vpkudum, but that pattern is invalid for processors prior to
POWER8. Thus LLVM crashes with an "unable to select" message. We
observed this since one of our buildbots is configured to generate
code for a POWER7.
This patch fixes the problem by checking for availability of the
vpkudum instruction during custom lowering of vector shuffles.
I've added a test case variant for the vpkudum pattern when the
instruction isn't available.
llvm-svn: 237952
http://reviews.llvm.org/D9891
Following up on the VSX single precision loads and stores added earlier, this
adds support for elementary arithmetic operations on single precision values
in VSX registers. These instructions utilize the new VSSRC register class.
Instructions added:
xsaddsp
xsdivsp
xsmulsp
xsresp
xsrsqrtesp
xssqrtsp
xssubsp
llvm-svn: 237937
If some commits are happy, and some commits are sad, this is a sad commit. It
is sad because it restricts instruction scheduling to work around a binutils
linker bug, and moreover, one that may never be fixed. On 2012-05-21, GCC was
updated not to produce code triggering this bug, and now we'll do the same...
When resolving an address using the ELF ABI TOC pointer, two relocations are
generally required: one for the high part and one for the low part. Only
the high part generally explicitly depends on r2 (the TOC pointer). And, so,
we might produce code like this:
.Ltmp526:
addis 3, 2, .LC12@toc@ha
.Ltmp1628:
std 2, 40(1)
ld 5, 0(27)
ld 2, 8(27)
ld 11, 16(27)
ld 3, .LC12@toc@l(3)
rldicl 4, 4, 0, 32
mtctr 5
bctrl
ld 2, 40(1)
And there is nothing wrong with this code, as such, but there is a linker bug
in binutils (https://sourceware.org/bugzilla/show_bug.cgi?id=18414) that will
misoptimize this code sequence to this:
nop
std r2,40(r1)
ld r5,0(r27)
ld r2,8(r27)
ld r11,16(r27)
ld r3,-32472(r2)
clrldi r4,r4,32
mtctr r5
bctrl
ld r2,40(r1)
because the linker does not know (and does not check) that the value in r2
changed in between the instruction using the .LC12@toc@ha (TOC-relative)
relocation and the instruction using the .LC12@toc@l(3) relocation.
Because it finds these instructions using the relocations (and not by
scanning the instructions), it has been asserted that there is no good way
to detect the change of r2 in between. As a result, this bug may never be
fixed (i.e. it may become part of the definition of the ABI). GCC was
updated to add extra dependencies on r2 to instructions using the @toc@l
relocations to avoid this problem, and we'll do the same here.
This is done as a separate pass because:
1. These extra r2 dependencies are not really properties of the
instructions, but rather due to a linker bug, and maybe one day we'll be
able to get rid of them when targeting linkers without this bug (and,
thus, keeping the logic centralized here will make that
straightforward).
2. There are ISel-level peephole optimizations that propagate the @toc@l
relocations to some user instructions, and so the exta dependencies do
not apply only to a fixed set of instructions (without undesirable
definition replication).
The test case was reduced with the help of bugpoint, with minimal cleaning. I'm
looking forward to our upcoming MI serialization support, and with that, much
better tests can be created.
llvm-svn: 237556
This patch adds support for the following new instructions in the
Power ISA 2.07:
vpksdss
vpksdus
vpkudus
vpkudum
vupkhsw
vupklsw
These instructions are available through the vec_packs, vec_packsu,
vec_unpackh, and vec_unpackl built-in interfaces. These are
lane-sensitive instructions, so the built-ins have different
implementations for big- and little-endian, and the instructions must
be marked as killing the vector swap optimization for now.
The first three instructions perform saturating pack operations. The
fourth performs a modulo pack operation, which means it can be
represented with a vector shuffle, and conversely the appropriate
vector shuffles may cause this instruction to be generated. The other
instructions are only generated via built-in support for now.
Appropriate tests have been added.
There is a companion patch to clang for the rest of this support.
llvm-svn: 237499
1) check whether the alignment of the memory is sufficient for the
*merged* store or load to be efficient.
Not doing so can result in some ridiculously poor code generation, if
merging creates a vector operation which must be aligned but isn't.
2) DON'T check that the alignment of each load/store is equal. If
you're merging 2 4-byte stores, the first *might* have 8-byte
alignment, but the second certainly will have 4-byte alignment. We do
want to allow those to be merged.
llvm-svn: 236850
This patch corresponds to review:
http://reviews.llvm.org/D9440
It adds a new register class to the PPC back end to contain single precision
values in VSX registers. Additionally, it adds scalar loads and stores for
VSX registers.
llvm-svn: 236755
The initial code drop for VSX swap optimization permitted the
optimization only when all operations in a web of related computation
are lane-insensitive. For some lane-sensitive operations, we can
still permit the optimization provided that we make adjustments to
those operations. This patch adds special handling for vector splats
so that their presence doesn't kill the optimization.
Vector splats are lane-sensitive since they identify by number a
vector element to be used as the source of a splat. When swap
optimizations take place, the desired vector element will move to the
opposite doubleword of the quadword vector. We thus replace the index
I by (I + N/2) % N, where N is the number of elements in the vector.
A new test case is added to test that swap optimization succeeds when
vector splats are present, and that the proper input element is used
as the source of the splat.
An ancillary change removes SH_BUILDVEC as one of the kinds of special
handling that may be required by VSX swap optimization. From
experience with GCC, I had expected to need some modifications for
vector build operations, but I did not find that to be the case.
llvm-svn: 236606
It adds v1i128 to the appropriate register classes and checks parameter passing
and return values.
This is related to http://reviews.llvm.org/D9081, which will add instructions
that exploit the v1i128 datatype.
Phabricator review: http://reviews.llvm.org/D9475
llvm-svn: 236503
Finish off PR23080 by renaming the debug info IR constructs from `MD*`
to `DI*`. The last of the `DIDescriptor` classes were deleted in
r235356, and the last of the related typedefs removed in r235413, so
this has all baked for about a week.
Note: If you have out-of-tree code (like a frontend), I recommend that
you get everything compiling and tests passing with the *previous*
commit before updating to this one. It'll be easier to keep track of
what code is using the `DIDescriptor` hierarchy and what you've already
updated, and I think you're extremely unlikely to insert bugs. YMMV of
course.
Back to *this* commit: I did this using the rename-md-di-nodes.sh
upgrade script I've attached to PR23080 (both code and testcases) and
filtered through clang-format-diff.py. I edited the tests for
test/Assembler/invalid-generic-debug-node-*.ll by hand since the columns
were off-by-three. It should work on your out-of-tree testcases (and
code, if you've followed the advice in the previous paragraph).
Some of the tests are in badly named files now (e.g.,
test/Assembler/invalid-mdcompositetype-missing-tag.ll should be
'dicompositetype'); I'll come back and move the files in a follow-up
commit.
llvm-svn: 236120
This patch adds a new SSA MI pass that runs on little-endian PPC64
code with VSX enabled. Loads and stores of 4x32 and 2x64 vectors
without alignment constraints are accomplished for little-endian using
lxvd2x/xxswapd and xxswapd/stxvd2x. The existence of the additional
xxswapd instructions hurts performance in comparison with big-endian
code, but they are necessary in the general case to support correct
semantics.
However, the general case does not apply to most vector code. Many
vector instructions are lane-insensitive; they do not "care" which
lanes the parallel computations are performed within, provided that
the resulting data is stored into the correct locations. Thus this
pass looks for computations that perform only lane-insensitive
operations, and remove the unnecessary swaps from loads and stores in
such computations.
Future improvements will allow computations using certain
lane-sensitive operations to also be optimized in this manner, by
modifying the lane-sensitive operations to account for the permuted
order of the lanes. However, this patch only adds the infrastructure
to permit this; no lane-sensitive operations are optimized at this
time.
This code is heavily exercised by the various vectorizing applications
in the projects/test-suite tree. For the time being, I have only added
one simple test case to demonstrate what the pass is doing. Although
it is quite simple, it provides coverage for much of the code,
including the special case handling of copies and subreg-to-reg
operations feeding the swaps. I plan to add additional tests in the
future as I fill in more of the "special handling" code.
Two existing tests were affected, because they expected the swaps to
be present, but they are now removed.
llvm-svn: 235910
TableGen had been nicely generating code to print a number of instructions using
shorter aliases (and PowerPC has plenty of short mnemonics), but we were not
calling it. For some of the aliases we support in the parser, TableGen can't
infer the "inverse" alias relationship, so there is still more to do.
Thus, after some hours of updating test cases...
llvm-svn: 235616
Third time's the charm. The previous commit was reverted as a
reverse for-loop in SelectionDAGBuilder::lowerWorkItem did 'I--'
on an iterator at the beginning of a vector, causing asserts
when using debugging iterators. This commit fixes that.
llvm-svn: 235608
This is a re-commit of r235101, which also fixes the problems with the previous patch:
- Switches with only a default case and non-fallthrough were handled incorrectly
- The previous patch tickled a bug in PowerPC Early-Return Creation which is fixed here.
> This is a major rewrite of the SelectionDAG switch lowering. The previous code
> would lower switches as a binary tre, discovering clusters of cases
> suitable for lowering by jump tables or bit tests as it went along. To increase
> the likelihood of finding jump tables, the binary tree pivot was selected to
> maximize case density on both sides of the pivot.
>
> By not selecting the pivot in the middle, the binary trees would not always
> be balanced, leading to performance problems in the generated code.
>
> This patch rewrites the lowering to search for clusters of cases
> suitable for jump tables or bit tests first, and then builds the binary
> tree around those clusters. This way, the binary tree will always be balanced.
>
> This has the added benefit of decoupling the different aspects of the lowering:
> tree building and jump table or bit tests finding are now easier to tweak
> separately.
>
> For example, this will enable us to balance the tree based on profile info
> in the future.
>
> The algorithm for finding jump tables is quadratic, whereas the previous algorithm
> was O(n log n) for common cases, and quadratic only in the worst-case. This
> doesn't seem to be major problem in practice, e.g. compiling a file consisting
> of a 10k-case switch was only 30% slower, and such large switches should be rare
> in practice. Compiling e.g. gcc.c showed no compile-time difference. If this
> does turn out to be a problem, we could limit the search space of the algorithm.
>
> This commit also disables all optimizations during switch lowering in -O0.
>
> Differential Revision: http://reviews.llvm.org/D8649
llvm-svn: 235560
This turned up after r235333, but was a pre-existing bug. The optimization
which transforms select(c, load, load) into a load of a select of the addresses
does not handle indexed loads (pre/post inc/dec). However, it did not check for
them either, leading to a crash if it tried to transform one of them.
llvm-svn: 235497
When an inline asm call has an output register marked as early-clobber, but
that same register is also an input operand, what should we do? GCC accepts
this, and is documented to accept this for read/write operands saying,
"Furthermore, if the earlyclobber operand is also a read/write operand, then
that operand is written only after it's used." For write-only operands, the
situation seems less clear, but I have at least one existing codebase that
assumes this will work, in part because it has syscall macros like this:
({ \
register uint64_t r0 __asm__ ("r0") = (__NR_ ## name); \
register uint64_t r3 __asm__ ("r3") = ((uint64_t) (arg0)); \
register uint64_t r4 __asm__ ("r4") = ((uint64_t) (arg1)); \
register uint64_t r5 __asm__ ("r5") = ((uint64_t) (arg2)); \
__asm__ __volatile__ \
("sc" \
: "=&r"(r0),"=&r"(r3),"=&r"(r4),"=&r"(r5) \
: "0"(r0), "1"(r3), "2"(r4), "3"(r5) \
: "r6","r7","r8","r9","r10","r11","r12","cr0","memory"); \
r3; \
})
Furthermore, with register aliases and subregister relationships that only the
backend knows about, rejecting this in the frontend seems like a difficult
proposition (if we wanted to do so). However, keeping the early-clobber flag on
the INLINEASM MI does not work for us, because it will cause the register's
live interval to end to soon (so it will not appear defined to be used as an
input).
Fortunately, fixing this does not seem hard: When forming the INLINEASM MI,
check to see if any of the early-clobber outputs are also inputs, and if so,
remove the early-clobber flag.
llvm-svn: 235283
See r230786 and r230794 for similar changes to gep and load
respectively.
Call is a bit different because it often doesn't have a single explicit
type - usually the type is deduced from the arguments, and just the
return type is explicit. In those cases there's no need to change the
IR.
When that's not the case, the IR usually contains the pointer type of
the first operand - but since typed pointers are going away, that
representation is insufficient so I'm just stripping the "pointerness"
of the explicit type away.
This does make the IR a bit weird - it /sort of/ reads like the type of
the first operand: "call void () %x(" but %x is actually of type "void
()*" and will eventually be just of type "ptr". But this seems not too
bad and I don't think it would benefit from repeating the type
("void (), void () * %x(" and then eventually "void (), ptr %x(") as has
been done with gep and load.
This also has a side benefit: since the explicit type is no longer a
pointer, there's no ambiguity between an explicit type and a function
that returns a function pointer. Previously this case needed an explicit
type (eg: a function returning a void() function was written as
"call void () () * @x(" rather than "call void () * @x(" because of the
ambiguity between a function returning a pointer to a void() function
and a function returning void).
No ambiguity means even function pointer return types can just be
written alone, without writing the whole function's type.
This leaves /only/ the varargs case where the explicit type is required.
Given the special type syntax in call instructions, the regex-fu used
for migration was a bit more involved in its own unique way (as every
one of these is) so here it is. Use it in conjunction with the apply.sh
script and associated find/xargs commands I've provided in rr230786 to
migrate your out of tree tests. Do let me know if any of this doesn't
cover your cases & we can iterate on a more general script/regexes to
help others with out of tree tests.
About 9 test cases couldn't be automatically migrated - half of those
were functions returning function pointers, where I just had to manually
delete the function argument types now that we didn't need an explicit
function type there. The other half were typedefs of function types used
in calls - just had to manually drop the * from those.
import fileinput
import sys
import re
pat = re.compile(r'((?:=|:|^|\s)call\s(?:[^@]*?))(\s*$|\s*(?:(?:\[\[[a-zA-Z0-9_]+\]\]|[@%](?:(")?[\\\?@a-zA-Z0-9_.]*?(?(3)"|)|{{.*}}))(?:\(|$)|undef|inttoptr|bitcast|null|asm).*$)')
addrspace_end = re.compile(r"addrspace\(\d+\)\s*\*$")
func_end = re.compile("(?:void.*|\)\s*)\*$")
def conv(match, line):
if not match or re.search(addrspace_end, match.group(1)) or not re.search(func_end, match.group(1)):
return line
return line[:match.start()] + match.group(1)[:match.group(1).rfind('*')].rstrip() + match.group(2) + line[match.end():]
for line in sys.stdin:
sys.stdout.write(conv(re.search(pat, line), line))
llvm-svn: 235145
This is a major rewrite of the SelectionDAG switch lowering. The previous code
would lower switches as a binary tre, discovering clusters of cases
suitable for lowering by jump tables or bit tests as it went along. To increase
the likelihood of finding jump tables, the binary tree pivot was selected to
maximize case density on both sides of the pivot.
By not selecting the pivot in the middle, the binary trees would not always
be balanced, leading to performance problems in the generated code.
This patch rewrites the lowering to search for clusters of cases
suitable for jump tables or bit tests first, and then builds the binary
tree around those clusters. This way, the binary tree will always be balanced.
This has the added benefit of decoupling the different aspects of the lowering:
tree building and jump table or bit tests finding are now easier to tweak
separately.
For example, this will enable us to balance the tree based on profile info
in the future.
The algorithm for finding jump tables is O(n^2), whereas the previous algorithm
was O(n log n) for common cases, and quadratic only in the worst-case. This
doesn't seem to be major problem in practice, e.g. compiling a file consisting
of a 10k-case switch was only 30% slower, and such large switches should be rare
in practice. Compiling e.g. gcc.c showed no compile-time difference. If this
does turn out to be a problem, we could limit the search space of the algorithm.
This commit also disables all optimizations during switch lowering in -O0.
Differential Revision: http://reviews.llvm.org/D8649
llvm-svn: 235101
Many of these predate llvm-readobj. With elf-dump we had to match
a relocation to symbol number and symbol number to symbol name or
section number.
llvm-svn: 235015
When I fixed these a couple of days ago to iterate over all loops, not just
depth == 1 loops, I inadvertently made it such that we'd only look at the first
top-level loop. Make sure that we really look at all of them.
llvm-svn: 234705
As it turns out, even though these are part of ISA 2.06, the P7 does not
support them (or, at least, not any P7s we're tested so far).
llvm-svn: 234686
This patch corresponds to review:
http://reviews.llvm.org/D8928
It adds direct move instructions to/from VSX registers to GPR's. These are
exploited for FP <-> INT conversions.
llvm-svn: 234682
This pass had the same problem as the data-prefetching pass: it was only
checking for depth == 1 loops in practice. Fix that, add some debugging
statements, and make sure that, when we grab an AddRec, it is for the loop we
expect.
llvm-svn: 234670
Iterating over loops from the LoopInfo instance only provides top-level loops.
We need to search the whole tree of loops to find the inner ones.
llvm-svn: 234603
When we have an instruction for this (and, thus, don't generate a runtime
call), we need to custom type legalize this (in a trivial way, just as we do
for fp_to_sint).
Fixes PR23173.
llvm-svn: 234561
This is the patch corresponding to review:
http://reviews.llvm.org/D8406
It adds some missing instructions from ISA 2.06 to the PPC back end.
llvm-svn: 234546
When enabling PPC64LE, I disabled some optimizations of BUILD_VECTOR
nodes for little endian because wrong results were produced. I've
subsequently investigated and found this is due to a call to
BuildVectorSDNode::isConstantSplat that was always specifying
big-endian. With this changed to correctly identify the target
endianness, the optimizations work as expected.
I found another case of a call to the same method with big-endian
hardcoded, in PPC::isAllNegativeZeroVector(). I discovered this was
an orphaned method with no callers, so I've just removed it.
The existing test/CodeGen/PowerPC/vec_constants.ll checks these
optimizations, so for testing I've just added a variant for little
endian.
llvm-svn: 234011
This patch attempts to fold the shuffling of 'scalar source' inputs - BUILD_VECTOR and SCALAR_TO_VECTOR nodes - if the shuffle node is the only user. This folds away a lot of unnecessary shuffle nodes, and allows quite a bit of constant folding that was being missed.
Differential Revision: http://reviews.llvm.org/D8516
llvm-svn: 234004
Under normal circumstances, use of CR bits is disabled when running at -O0, but
it is enabled by default otherwise, and if you have optnone functions, they'll
still generally be generated with crbits turned on (because nothing else turns
them off). FastISel can't handle most things dealing with i1 values when using
CR bits, and checks for that, but was not checking the return type on
functions; we can't fast-isel function calls with i1 return values either when
using CR bits for boolean values.
Fixes PR22664.
llvm-svn: 233775
Even at -O0, we fall back to SDAG when we hit intrinsics, and if the intrinsic
is a memset/memcpy/etc. we might normally use vector types. At -O0, this is
probably not a good idea (because, if there is a bug in the lowering code,
there would be no good way to turn it off). At -O0, only use scalar preferred
types.
Related to PR22754.
llvm-svn: 233755
The existing code in getMemsetValue only handled integer-preferred types when
the fill value was not a constant. Make this more robust in two ways:
1. If the preferred type is a floating-point value, do the mul-splat trick on
the corresponding integer type and then bitcast.
2. If the preferred type is a vector, do the mul-splat trick on one vector
element, and then build a vector out of them.
Fixes PR22754 (although, we should also turn off use of vector types at -O0).
llvm-svn: 233749
Fix debug info in these tests, which started failing with a WIP patch to
verify compile units and types. The problems look like they were all
caused by bitrot. They fell into these categories:
- Using `!{i32 0}` instead of `!{}`.
- Using `!{null}` instead of `!{}`.
- Using `!MDExpression()` instead of `!{}`.
- Using `!8` instead of `!{!8}`.
- `file:` references that pointed at `MDCompileUnit`s instead of the
same `MDFile` as the compile unit.
- `file:` references that were numerically off-by-one or (off-by-ten).
llvm-svn: 233415
"Fix the MachineScheduler's logic for updating ready times for in-order.
Now the scheduler updates a node's ready time as soon as it is
scheduled, before releasing dependent nodes."
This fix was only made in one variant of the ScheduleDAGMI driver.
Francois de Ferriere reported the issue in the other bit of code where
it was also needed.
I never got around to coming up with a test case, but it's an
obvious fix that shouldn't be delayed any longer.
I'll try to refactor this code a little better.
I did verify performance on a wide variety of targets and saw no
negative impact with this fix.
llvm-svn: 233366
This patch adds Hardware Transaction Memory (HTM) support supported by ISA 2.07
(POWER8). The intrinsic support is based on GCC one [1], but currently only the
'PowerPC HTM Low Level Built-in Function' are implemented.
The HTM instructions follows the RC ones and the transaction initiation result
is set on RC0 (with exception of tcheck). Currently approach is to create a
register copy from CR0 to GPR and comapring. Although this is suboptimal, since
the branch could be taken directly by comparing the CR0 value, it generates code
correctly on both test and branch and just return value. A possible future
optimization could be elimitate the MFCR instruction to branch directly.
The HTM usage requires a recently newer kernel with PPC HTM enabled. Tested on
powerpc64 and powerpc64le.
This is send along a clang patch to enabled the builtins and option switch.
[1] https://gcc.gnu.org/onlinedocs/gcc/PowerPC-Hardware-Transactional-Memory-Built-in-Functions.html
Phabricator Review: http://reviews.llvm.org/D8247
llvm-svn: 233204
Because the operands of a vector SETCC node can be of a different type from the
result (and often are), it can happen that even if we'd prefer to widen the
result type of the SETCC, the operands have been split instead. In this case,
the SETCC result also must be split. This mirrors what is done in
WidenVecRes_SELECT, and should be NFC elsewhere because if the operands are not
widened the following calls to GetWidenedVector will assert (which is what was
happening in the test case).
llvm-svn: 232935
This is very related to the bug fixed in r174431. The problem is that
SelectionDAG does not include alignment in the uniquing of loads and
stores. When an otherwise no-op DAGCombine would increase the alignment
of a load or store, the original node would be returned (with the
alignment increased), which would cause the node not to be processed by
any further DAGCombines.
I don't have a direct testcase for this that manifests on an in-tree
target, but I did see some noise in the tests for other targets and have
updated them for it.
llvm-svn: 232780
The VSX stores are sometimes generated with a undefined index register, causing %noreg to be used and R0 to be emitted later on. The semantics of the VSX store (e.g. stdsdx) requires R0 to be used as base if we want zero to be used in the computation of the effective address instead of the content of R0. This patch checks if no index register was generated and forces R0 to be used as base address.
llvm-svn: 232486
Similar to gep (r230786) and load (r230794) changes.
Similar migration script can be used to update test cases, which
successfully migrated all of LLVM and Polly, but about 4 test cases
needed manually changes in Clang.
(this script will read the contents of stdin and massage it into stdout
- wrap it in the 'apply.sh' script shown in previous commits + xargs to
apply it over a large set of test cases)
import fileinput
import sys
import re
rep = re.compile(r"(getelementptr(?:\s+inbounds)?\s*\()((<\d*\s+x\s+)?([^@]*?)(|\s*addrspace\(\d+\))\s*\*(?(3)>)\s*)(?=$|%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|zeroinitializer|<|\[\[[a-zA-Z]|\{\{)", re.MULTILINE | re.DOTALL)
def conv(match):
line = match.group(1)
line += match.group(4)
line += ", "
line += match.group(2)
return line
line = sys.stdin.read()
off = 0
for match in re.finditer(rep, line):
sys.stdout.write(line[off:match.start()])
sys.stdout.write(conv(match))
off = match.end()
sys.stdout.write(line[off:])
llvm-svn: 232184
Move the specialized metadata nodes for the new debug info hierarchy
into place, finishing off PR22464. I've done bootstraps (and all that)
and I'm confident this commit is NFC as far as DWARF output is
concerned. Let me know if I'm wrong :).
The code changes are fairly mechanical:
- Bumped the "Debug Info Version".
- `DIBuilder` now creates the appropriate subclass of `MDNode`.
- Subclasses of DIDescriptor now expect to hold their "MD"
counterparts (e.g., `DIBasicType` expects `MDBasicType`).
- Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp`
for printing comments.
- Big update to LangRef to describe the nodes in the new hierarchy.
Feel free to make it better.
Testcase changes are enormous. There's an accompanying clang commit on
its way.
If you have out-of-tree debug info testcases, I just broke your build.
- `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to
update all the IR testcases.
- Unfortunately I failed to find way to script the updates to CHECK
lines, so I updated all of these by hand. This was fairly painful,
since the old CHECKs are difficult to reason about. That's one of
the benefits of the new hierarchy.
This work isn't quite finished, BTW. The `DIDescriptor` subclasses are
almost empty wrappers, but not quite: they still have loose casting
checks (see the `RETURN_FROM_RAW()` macro). Once they're completely
gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I
also expect to make a few schema changes now that it's easier to reason
about everything.
llvm-svn: 231082
Straightforward patch to emit an alignment directive when emitting a
TOC entry. The test case was generated from the test in PR22711 that
demonstrated a misaligned .toc section. The object code is run
through llvm-readobj to verify that the correct alignment has been
applied to the .toc section.
Thanks to Ulrich Weigand for running down where the fix was needed.
llvm-svn: 230801
Essentially the same as the GEP change in r230786.
A similar migration script can be used to update test cases, though a few more
test case improvements/changes were required this time around: (r229269-r229278)
import fileinput
import sys
import re
pat = re.compile(r"((?:=|:|^)\s*load (?:atomic )?(?:volatile )?(.*?))(| addrspace\(\d+\) *)\*($| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$)")
for line in sys.stdin:
sys.stdout.write(re.sub(pat, r"\1, \2\3*\4", line))
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7649
llvm-svn: 230794
When using Altivec, we can use vector loads and stores for aligned memcpy and
friends. Starting with the P7 and VXS, we have reasonable unaligned vector
stores. Starting with the P8, we have fast unaligned loads too.
For QPX, we use vector loads are stores, but only for aligned memory accesses.
llvm-svn: 230788
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
Summary:
Currently fast-isel-abort will only abort for regular instructions,
and just warn for function calls, terminators, function arguments.
There is already fast-isel-abort-args but nothing for calls and
terminators.
This change turns the fast-isel-abort options into an integer option,
so that multiple levels of strictness can be defined.
This will help no being surprised when the "abort" option indeed does
not abort, and enables the possibility to write test that verifies
that no intrinsics are forgotten by fast-isel.
Reviewers: resistor, echristo
Subscribers: jfb, llvm-commits
Differential Revision: http://reviews.llvm.org/D7941
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 230775
LDtocL, and other loads that roughly correspond to the TOC_ENTRY SDAG node,
represent loads from the TOC, which is invariant. As a result, these loads can
be hoisted out of loops, etc. In order to do this, we need to generate
GOT-style MMOs for TOC_ENTRY, which requires treating it as a legitimate memory
intrinsic node type. Once this is done, the MMO transfer is automatically
handled for TableGen-driven instruction selection, and for nodes generated
directly in PPCISelDAGToDAG, we need to transfer the MMOs manually.
Also, we were not transferring MMOs associated with pre-increment loads, so do
that too.
Lastly, this fixes an exposed bug where R30 was not added as a defined operand of
UpdateGBR.
This problem was highlighted by an example (used to generate the test case)
posted to llvmdev by Francois Pichet.
llvm-svn: 230553
This adds support for the QPX vector instruction set, which is used by the
enhanced A2 cores on the IBM BG/Q supercomputers. QPX vectors are 256 bytes
wide, holding 4 double-precision floating-point values. Boolean values, modeled
here as <4 x i1> are actually also represented as floating-point values
(essentially { -1, 1 } for { false, true }). QPX shares many features with
Altivec and VSX, but is distinct from both of them. One major difference is
that, instead of adding completely-separate vector registers, QPX vector
registers are extensions of the scalar floating-point registers (lane 0 is the
corresponding scalar floating-point value). The operations supported on QPX
vectors mirrors that supported on the scalar floating-point values (with some
additional ones for permutations and logical/comparison operations).
I've been maintaining this support out-of-tree, as part of the bgclang project,
for several years. This is not the entire bgclang patch set, but is most of the
subset that can be cleanly integrated into LLVM proper at this time. Adding
this to the LLVM backend is part of my efforts to rebase bgclang to the current
LLVM trunk, but is independently useful (especially for codes that use LLVM as
a JIT in library form).
The assembler/disassembler test coverage is complete. The CodeGen test coverage
is not, but I've included some tests, and more will be added as follow-up work.
llvm-svn: 230413
The IBM BG/Q supercomputer's A2 cores have a hardware prefetching unit, the
L1P, but it does not prefetch directly into the A2's L1 cache. Instead, it
prefetches into its own L1P buffer, and the latency to access that buffer is
significantly higher than that to the L1 cache (although smaller than the
latency to the L2 cache). As a result, especially when multiple hardware
threads are not actively busy, explicitly prefetching data into the L1 cache is
advantageous.
I've been using this pass out-of-tree for data prefetching on the BG/Q for well
over a year, and it has worked quite well. It is enabled by default only for
the BG/Q, but can be enabled for other cores as well via a command-line option.
Eventually, we might want to add some TTI interfaces and move this into
Transforms/Scalar (there is nothing particularly target dependent about it,
although only machines like the BG/Q will benefit from its simplistic
strategy).
llvm-svn: 229966
Our register allocation has become better recently, it seems, and is now
starting to generate cross-block copies into inflated register classes. These
copies are not transformed into subregister insertions/extractions by the
PPCVSXCopy class, and so need to be handled directly by
PPCInstrInfo::copyPhysReg. The code to do this was *almost* there, but not
quite (it was unnecessarily restricting itself to only the direct
sub/super-register-class case (not copying between, for example, something in
VRRC and the lower-half of VSRC which are super-registers of F8RC).
Triggering this behavior manually is difficult; I'm including two
bugpoint-reduced test cases from the test suite.
llvm-svn: 229457
SimplifyCFG now knows how to speculate calls to intrinsic cttz/ctlz that are
'cheap' for the target. Therefore, some of the logic in CodeGenPrepare
that was originally added at revision 224899 can now be removed.
This patch is basically a no functional change. It removes the duplicated
logic in CodeGenPrepare and converts all the existing target specific tests
for cttz/ctlz into SimplifyCFG tests.
Differential Revision: http://reviews.llvm.org/D7608
llvm-svn: 229105
The PowerPC backend has long promoted some floating-point vector operations
(such as select) to integer vector operations. Unfortunately, this behavior was
broken by r216555. When using FP_EXTEND/FP_ROUND for promotions, we must check
that both the old and new types are floating-point types. Otherwise, we must
use BITCAST as we did prior to r216555 for everything.
llvm-svn: 228969
On PowerPC, which has a full set of logical operations on (its multiple sets
of) condition-register bits, it is not profitable to break of complex
conditions feeding a jump into multiple jumps. We can turn off this feature of
CGP/SDAGBuilder by marking jumps as "expensive".
P7 test-suite speedups (no regressions):
MultiSource/Benchmarks/FreeBench/pcompress2/pcompress2
-0.626647% +/- 0.323583%
MultiSource/Benchmarks/Olden/power/power
-18.2821% +/- 8.06481%
llvm-svn: 228895
See full discussion in http://reviews.llvm.org/D7491.
We now hide the add-immediate and call instructions together in a
separate pseudo-op, which is tagged to define GPR3 and clobber the
call-killed registers. The PPCTLSDynamicCall pass prior to RA now
expands this op into the two separate addi and call ops, with explicit
definitions of GPR3 on both instructions, and explicit clobbers on the
call instruction. The pass is now marked as requiring and preserving
the LiveIntervals and SlotIndexes analyses, and fixes these up after
the replacement sequences are introduced.
Self-hosting has been verified on LE P8 and BE P7 with various
optimization levels, etc. It has also been verified with the
--no-tls-optimize flag workaround removed.
llvm-svn: 228725
veqv (vector equivalence)
vnand
vorc
I increased the AddedComplexity for these instructions to 500 to ensure they are generated instead of issuing other VSX instructions.
Phabricator review: http://reviews.llvm.org/D7469
llvm-svn: 228580
If a loop predecessor has an invoke as its terminator, and the return value
from that invoke is used to determine the loop iteration space, then we can't
insert a computation based on that value in the loop predecessor prior to the
terminator (oops). If there's such an invoke, or just no predecessor for that
matter, insert a new loop preheader.
llvm-svn: 228488
Unfortunately, even with the workaround of disabling the linker TLS
optimizations in Clang restored (which has already been done), this still
breaks self-hosting on my P7 machine (-O3 -DNDEBUG -mcpu=native).
Bill is currently working on an alternate implementation to address the TLS
issue in a way that also fully elides the linker bug (which, unfortunately,
this approach did not fully), so I'm reverting this now.
llvm-svn: 228460
PowerPC supports pre-increment load/store instructions (except for Altivec/VSX
vector load/stores). Using these on embedded cores can be very important, but
most loops are not naturally set up to use them. We can often change that,
however, by placing loops into a non-canonical form. Generically, this means
transforming loops like this:
for (int i = 0; i < n; ++i)
array[i] = c;
to look like this:
T *p = array[-1];
for (int i = 0; i < n; ++i)
*++p = c;
the key point is that addresses accessed are pulled into dedicated PHIs and
"pre-decremented" in the loop preheader. This allows the use of pre-increment
load/store instructions without loop peeling.
A target-specific late IR-level pass (running post-LSR), PPCLoopPreIncPrep, is
introduced to perform this transformation. I've used this code out-of-tree for
generating code for the PPC A2 for over a year. Somewhat to my surprise,
running the test suite + externals on a P7 with this transformation enabled
showed no performance regressions, and one speedup:
External/SPEC/CINT2006/483.xalancbmk/483.xalancbmk
-2.32514% +/- 1.03736%
So I'm going to enable it on everything for now. I was surprised by this
because, on the POWER cores, these pre-increment load/store instructions are
cracked (and, thus, harder to schedule effectively). But seeing no regressions,
and feeling that it is generally easier to split instructions apart late than
it is to combine them late, this might be the better approach regardless.
In the future, we might want to integrate this functionality into LSR (but
currently LSR does not create new PHI nodes, so (for that and other reasons)
significant work would need to be done).
llvm-svn: 228328
PowerPC supports pre-increment floating-point load/store instructions, both r+r
and r+i, and we had patterns for them, but they were not marked as legal. Mark
them as legal (and add a test case).
llvm-svn: 228327
Patch by Kit Barton.
Add the vector count leading zeros instruction for byte, halfword,
word, and doubleword sizes. This is a fairly straightforward addition
after the changes made for vpopcnt:
1. Add the correct definitions for the various instructions in
PPCInstrAltivec.td
2. Make the CTLZ operation legal on vector types when using P8Altivec
in PPCISelLowering.cpp
Test Plan
Created new test case in test/CodeGen/PowerPC/vec_clz.ll to check the
instructions are being generated when the CTLZ operation is used in
LLVM.
Check the encoding and decoding in test/MC/PowerPC/ppc_encoding_vmx.s
and test/Disassembler/PowerPC/ppc_encoding_vmx.txt respectively.
llvm-svn: 228301
This patch is a third attempt to properly handle the local-dynamic and
global-dynamic TLS models.
In my original implementation, calls to __tls_get_addr were hidden
from view until the asm-printer phase, at which point the underlying
branch-and-link instruction was created with proper relocations. This
mostly worked well, but I used some repellent techniques to ensure
that the TLS_GET_ADDR nodes at the SD and MI levels correctly received
input from GPR3 and produced output into GPR3. This proved to work
badly in the presence of multiple TLS variable accesses, with the
copies to and from GPR3 being scheduled incorrectly and generally
creating havoc.
In r221703, I addressed that problem by representing the calls to
__tls_get_addr as true calls during instruction lowering. This had
the advantage of removing all of the bad hacks and relying on the
existing call machinery to properly glue the copies in place. It
looked like this was going to be the right way to go.
However, as a side effect of the recent discovery of problems with
linker optimizations for TLS, we discovered cases of suboptimal code
generation with this strategy. The problem comes when tls_get_addr is
called for the same address, and there is a resulting CSE
opportunity. It turns out that in such cases MachineCSE will common
the addis/addi instructions that set up the input value to
tls_get_addr, but will not common the calls themselves. MachineCSE
does not have any machinery to common idempotent calls. This is
perfectly sensible, since presumably this would be done at the IR
level, and introducing calls in the back end isn't commonplace. In
any case, we end up with two calls to __tls_get_addr when one would
suffice, and that isn't good.
I presumed that the original design would have allowed commoning of
the machine-specific nodes that hid the __tls_get_addr calls, so as
suggested by Ulrich Weigand, I went back to that design and cleaned it
up so that the copies were properly held together by glue
nodes. However, it turned out that this didn't work either...the
presence of copies to physical registers kept the machine-specific
nodes from being commoned also.
All of which leads to the design presented here. This is a return to
the original design, except that no attempt is made to introduce
copies to and from GPR3 during instruction lowering. Virtual registers
are used until prior to register allocation. At that point, a special
pass is run that identifies the machine-specific nodes that hide the
tls_get_addr calls and introduces the copies to and from GPR3 around
them. The register allocator then coalesces these copies away. With
this design, MachineCSE succeeds in commoning tls_get_addr calls where
possible, and we get nice optimal code generation (better than GCC at
the moment, which does not common these calls).
One additional problem must be dealt with: After introducing the
mentions of the physical register GPR3, the aggressive anti-dependence
breaker sees opportunities to improve scheduling by selecting a
different register instead. Flags must be used on the instruction
descriptions to tell the anti-dependence breaker to keep its hands in
its pockets.
One thing missing from the original design was recording a definition
of the link register on the GET_TLS_ADDR nodes. Doing this was found
to be insufficient to force a stack frame to be created, which led to
looping behavior because two different LR values were stored at the
same address. This appears to have been an oversight in
PPCFrameLowering::determineFrameLayout(), which is repaired here.
Because MustSaveLR() returns true for calls to builtin_return_address,
this changed the expected behavior of
test/CodeGen/PowerPC/retaddr2.ll, which now stacks a frame but
formerly did not. I've fixed the test case to reflect this.
There are existing TLS tests to catch regressions; the checks in
test/CodeGen/PowerPC/tls-store2.ll proved to be too restrictive in the
face of instruction scheduling with these changes, so I fixed that
up.
I've added a new test case based on the PrettyStackTrace module that
demonstrated the original problem. This checks that we get correct
code generation and that CSE of the calls to __get_tls_addr has taken
place.
llvm-svn: 227976
The VSX store instructions were also picking up an implicit "may read" from the
default pattern, which was an intrinsic (and we don't currently have a way of
specifying write-only intrinsics).
This was causing MI verification to fail for VSX spill restores.
llvm-svn: 227759
isel is actually a cracked instruction on the P7/P8, and must start a dispatch
group. The scheduling model should reflect this so that we don't bunch too many
of them together when possible.
Thanks to Bill Schmidt and Pat Haugen for helping to sort this out.
llvm-svn: 227758
The TOC base pointer is passed in r2, and we normally reserve this register so
that we can depend on it being there. However, for leaf functions, and
specifically those leaf functions that don't do any TOC access of their own
(which is generally due to accessing the constant pool, using TLS, etc.),
we can treat r2 as an ordinary callee-saved register (it must be callee-saved
because, for local direct calls, the linker will not insert any save/restore
code).
The allocation order has been changed slightly for PPC64/ELF systems to put r2
at the end of the list (while leaving it near the beginning for Darwin systems
to prevent unnecessary output changes). While r2 is allocatable, using it still
requires spill/restore traffic, and thus comes at the end of the list.
llvm-svn: 227745
Patch by Nemanja Ivanovic.
As was uncovered by the failing test case (when run on non-PPC
platforms), the feature set when compiling with -march=ppc64le was not
being picked up. This change ensures that if the -mcpu option is not
specified, the correct feature set is picked up regardless of whether
we are on PPC or not.
llvm-svn: 227455
It appears we have different behavior with and without -mcpu=pwr8 even
with ppc64le defaulting to POWER8. The failure appears as follows:
/home/bb/cmake-llvm-x86_64-linux/llvm-project/llvm/test/CodeGen/PowerPC/ppc64le-aggregates.ll:268:14: error: expected string not found in input
; CHECK-DAG: lfs 1, 0([[REG]])
^
<stdin>:497:11: note: scanning from here
ld 3, .LC1@toc@l(3)
^
<stdin>:497:11: note: with variable "REG" equal to "3"
ld 3, .LC1@toc@l(3)
^
<stdin>:514:2: note: possible intended match here
lfs 1, 0(4)
^
Reverting this particular test case change. Nemanja, please have a look
at the reason for the failure.
llvm-svn: 227055
Test by Nemanja Ivanovic.
Since ppc64le implies POWER8 as a minimum, it makes sense that the
same features are included. Since the pwr8 processor model will likely
be getting new features until the implementation is complete, I
created a new list to add these updates to. This will include them in
both pwr8 and ppc64le.
Furthermore, it seems that it would make sense to compose the feature
lists for other processor models (pwr3 and up). Per discussion in the
review, I will make this change in a subsequent patch.
In order to test the changes, I've added an additional run step to
test cases that specify -march=ppc64le -mcpu=pwr8 to omit the -mcpu
option. Since the feature lists are the same, the behaviour should be
unchanged.
llvm-svn: 227053
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 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
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
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
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
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
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
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 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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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
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 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
Determining the address of a TLS variable results in a function call in
certain TLS models. This means that a simple ICmpInst might actually
result in invalidating the CTR register.
In such cases, do not attempt to rely on the CTR register for loop
optimization purposes.
This fixes PR22034.
Differential Revision: http://reviews.llvm.org/D6786
llvm-svn: 224890
When materializing constant i1 values, they must be zero extended. We represent
i1 values as [0, 1], not [0, -1], in i32 registers. As it turns out, this code
path was dead for i1 values prior to r216006 (which is why this did not manifest in
miscompiles until recently).
Fixes -O0 self-hosting on PPC64/Linux.
llvm-svn: 224842
Dan Gohman