The new ORC remote-JITing support provides a superset of the old code's
functionality, so we can replace the old stuff. As a bonus, a couple of
previously XFAILed tests have started passing.
llvm-svn: 257343
Summary:
It actually takes an offset into the current PC-region.
This fixes the 'expr' command in lldb.
Reviewers: vkalintiris, jaydeep, bhushan
Subscribers: dsanders, llvm-commits
Differential Revision: http://reviews.llvm.org/D16054
llvm-svn: 257339
In the OptimizeLEA pass keep instructions' positions in the basic block saved and use them for calculation of the distance between two instructions instead of std::distance. This reduces complexity of the pass from O(n^3) to O(n^2) and thus the compile time.
Differential Revision: http://reviews.llvm.org/D15692
llvm-svn: 257328
This is a recommit of r257253 which was reverted in r257270.
Previous testcase can make failure on some targets due to using opt with O3 option.
Original Summary:
Merge MBBICommon and MBBI's MMOs.
Differential Revision: http://reviews.llvm.org/D15990
llvm-svn: 257317
This patch adds utilities to ORC for managing a remote JIT target. It consists
of:
1. A very primitive RPC system for making calls over a byte-stream. See
RPCChannel.h, RPCUtils.h.
2. An RPC API defined in the above system for managing memory, looking up
symbols, creating stubs, etc. on a remote target. See OrcRemoteTargetRPCAPI.h.
3. An interface for creating high-level JIT components (memory managers,
callback managers, stub managers, etc.) that operate over the RPC API. See
OrcRemoteTargetClient.h.
4. A helper class for building servers that can handle the RPC calls. See
OrcRemoteTargetServer.h.
The system is designed to work neatly with the existing ORC components and
functionality. In particular, the ORC callback API (and consequently the
CompileOnDemandLayer) is supported, enabling lazy compilation of remote code.
Assuming this doesn't trigger any builder failures, a follow-up patch will be
committed which tests these utilities by using them to replace LLI's existing
remote-JITing demo code.
llvm-svn: 257305
This is a more generic version of the MCJITMemoryManager::notifyObjectLoaded
method: It provides only a RuntimeDyld reference (rather than an
ExecutionEngine), and so can be used with ORC JIT stacks.
llvm-svn: 257296
RuntimeDyld::MemoryManager.
The RuntimeDyld::MemoryManager::reserveAllocationSpace method is called when
object files are loaded, and gives clients a chance to pre-allocate memory for
all segments. Previously only the size of each segment (code, ro-data, rw-data)
was supplied but not the alignment. This hasn't caused any problems so far, as
most clients allocate via the MemoryBlock interface which returns page-aligned
blocks. Adding alignment arguments enables finer grained allocation while still
satisfying alignment restrictions.
llvm-svn: 257294
In r255760, I optimized the SectionMemoryManager to make better use
of virtual memory on platforms where the allocation granularity was
bigger than the protection granularity. As part of this, fixing up
the free list became more complicated and was moved into
`applyMemoryGroupPermissions`. Unfortunately, I forgot to actually
remove the call that drops the free list for RO memory (I did
remove the corresponding one for RX memory), defeating the whole
optimization.
llvm-svn: 257293
Summary:
Use proper dataflow ordering to speed convergence.
This will converge the testcase on bug 26055 in 2 iterations.
(data structures speedups to come to make even that faster)
Reviewers: kcc, samsonov, echristo, dblaikie, tvvikram
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D16039
llvm-svn: 257292
llvm\unittests\ExecutionEngine\Orc\ObjectLinkingLayerTest.cpp(115) : error C2327: 'llvm::OrcExecutionTest::TM' : is not a type name, static, or enumerator
llvm\unittests\ExecutionEngine\Orc\ObjectLinkingLayerTest.cpp(115) : error C2065: 'TM' : undeclared identifier
FYI, "this->TM" was valid even before moving class SectionMemoryManagerWrapper.
llvm-svn: 257290
MSVC seems to have problems looking up Value inside of the template. Not
really sure whether that's a bug there or Clang and GCC being too
permissive.
llvm-svn: 257288
type.
This makes it easy and safe to use a set of flags as one elmenet of
a tagged union with pointers. There is quite a bit of code that has
historically done this by casting arbitrary integers to "pointers" and
assuming that this was safe and reliable. It is neither, and has started
to rear its head by triggering safety asserts in various abstractions
like PointerLikeTypeTraits when the integers chosen are invariably poor
choices for *some* platform and *some* situation. Not to mention the
(hopefully unlikely) prospect of one of these integers actually getting
allocated!
With this, it will be straightforward to build type safe abstractions
like this without being error prone. The abstraction itself is also
remarkably simple thanks to the implicit conversion.
This use case and pattern was also independently created by the folks
working on Swift, and they're going to incrementally add any missing
functionality they find.
Differential Revision: http://reviews.llvm.org/D15844
llvm-svn: 257284
This is a much more general and powerful form of PointerUnion. It
provides a reasonably complete sum type (from type theory) for
pointer-like types. It has several significant advantages over the
existing PointerUnion infrastructure:
1) It allows more than two pointer types to participate without awkward
nesting structures.
2) It directly exposes the tag so that it is convenient to write
switches over the possible members.
3) It can re-use the same type for multiple tag values, something that
has been worked around by either abusing PointerIntPair or defining
nonce types and doing unsafe pointer casting.
4) It supports customization of the PointerLikeTypeTraits used for
specific member types. This means it could (in theory) be used even
with types that are over-aligned on allocation to expose larger
numbers of bits to the tag.
All in all, I think it is at least complimentary to the existing
infrastructure, and a strict improvement for some use cases.
Differential Revision: http://reviews.llvm.org/D15843
llvm-svn: 257282
JumpThreading's runOnFunction is supposed to return true if it made any
changes. JumpThreading has a call to removeUnreachableBlocks which may
result in changes to the IR but runOnFunction didn't appropriate account
for this possibility, leading to badness.
While we are here, make sure to call LazyValueInfo::eraseBlock in
removeUnreachableBlocks; JumpThreading preserves LVI.
This fixes PR26096.
llvm-svn: 257279
Summary:
This is a fix of D13718. D13718 was committed but then reverted because of the following bug:
https://llvm.org/bugs/show_bug.cgi?id=25299
This patch fixes the issue shown in the bug.
Reviewers: majnemer, reames
Subscribers: jevinskie, llvm-commits
Differential Revision: http://reviews.llvm.org/D14308
llvm-svn: 257277
Summary:
The code was simply ensuring that the catchpad's pred is its catchswitch,
which was letting cases slip through where the flow edge was the unwind
edge of the catchswitch rather than one of its catch clauses.
Reviewers: andrew.w.kaylor, rnk, majnemer
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D16011
llvm-svn: 257275
Summary:
Funclet-based EH personalities/tables likely can't handle these, and they
can't be generated at source, so make them officially illegal in IR as
well.
Reviewers: andrew.w.kaylor, rnk, majnemer
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D15963
llvm-svn: 257274
Summary:
A funclet EH pad may be exited by an unwind edge, which may be a
cleanupret exiting its cleanuppad, an invoke exiting a funclet, or an
unwind out of a nested funclet transitively exiting its parent. Funclet
EH personalities require all such exceptional exits from a given funclet to
have the same unwind destination, and EH preparation / state numbering /
table generation implicitly depends on this. Formalize it as a rule of
the IR in the LangRef and verifier.
Reviewers: rnk, majnemer, andrew.w.kaylor
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D15962
llvm-svn: 257273
Summary:
Funclet EH personalities require a tree-like nesting among funclets
(enforced by the ParentPad linkage in the IR), and also require that
unwind edges conform to certain rules with respect to the tree:
- An unwind edge may exit 0 or more ancestor pads
- An unwind edge must enter exactly one EH pad, which must be distinct
from any exited pads
- A cleanupret's edge must exit its cleanuppad
Describe these rules in the LangRef, and enforce them in the verifier.
Reviewers: rnk, majnemer, andrew.w.kaylor
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D15961
llvm-svn: 257272
This reverts r257221.
This caused several build bot failures
* It looks like some of the tests don't work correctly under Windows
* It looks like the lit per test timeout tests fail
So I'm reverting for now. Once the above failures are fixed running
lit's tests can be enabled again.
llvm-svn: 257268
AVX1 v8i32/v4i64 shuffles are bitcasted to v8f32/v4f64, this patch peeks through any bitcast to check for a load node to allow broadcasts to occur.
This is a re-commit of r257055 after r257264 fixed 32-bit broadcast loads of i64 scalars.
llvm-svn: 257266
managers.
Prior to this patch, recursive finalization (where finalization of one
RuntimeDyld instance triggers finalization of another instance on which the
first depends) could trigger memory access failures: When the inner (dependent)
RuntimeDyld instance and its memory manager are finalized, memory allocated
(but not yet relocated) by the outer instance is locked, and relocation in the
outer instance fails with a memory access error.
This patch adds a latch to the RuntimeDyld::MemoryManager base class that is
checked by a new method: RuntimeDyld::finalizeWithMemoryManagerLocking, ensuring
that shared memory managers are only finalized by the outermost RuntimeDyld
instance.
This allows ORC clients to supply the same memory manager to multiple calls to
addModuleSet. In particular it enables the use of user-supplied memory managers
with the CompileOnDemandLayer which must reuse the supplied memory manager for
each function that is lazily compiled.
llvm-svn: 257263
Summary:
This is analogous to r256079, which removed an overly strong assertion, and
r256812, which simplified the code by replacing three conditionals by one.
Reviewers: reames
Subscribers: sanjoy, llvm-commits
Differential Revision: http://reviews.llvm.org/D16019
llvm-svn: 257250
This patch teaches rewrite-statepoints-for-gc to relocate vector-of-pointers directly rather than trying to split them. This builds on the recent lowering/IR changes to allow vector typed gc.relocates.
The motivation for this is that we recently found a bug in the vector splitting code where depending on visit order, a vector might not be relocated at some safepoint. Specifically, the bug is that the splitting code wasn't updating the side tables (live vector) of other safepoints. As a result, a vector which was live at two safepoints might not be updated at one of them. However, if you happened to visit safepoints in post order over the dominator tree, everything worked correctly. Weirdly, it turns out that post order is actually an incredibly common order to visit instructions in in practice. Frustratingly, I have not managed to write a test case which actually hits this. I can only reproduce it in large IR files produced by actual applications.
Rather than continue to make this code more complicated, we can remove all of the complexity by just representing the relocation of the entire vector natively in the IR.
At the moment, the new functionality is hidden behind a flag. To use this code, you need to pass "-rs4gc-split-vector-values=0". Once I have a chance to stress test with this option and get feedback from other users, my plan is to flip the default and remove the original splitting code. I would just remove it now, but given the rareness of the bug, I figured it was better to leave it in place until the new approach has been stress tested.
Differential Revision: http://reviews.llvm.org/D15982
llvm-svn: 257244
directy with ``make check-lit`` and are run as part of
``make check-all``.
In principle we should run lit's testsuite before testing LLVM using lit
so that any problems with lit get discovered before testing LLVM so we
can bail out early. However this implementation (``check-all`` runs all
tests together) seemed simpler and will still report failing lit tests.
Note that the tests and the configured ``lit.site.cfg`` have to be
copied into the build directory to avoid polluting the source tree.
llvm-svn: 257221
Look for PHI/Select in the same BB of the form
bb:
%p = phi [false, %bb1], [true, %bb2], [false, %bb3], [true, %bb4], ...
%s = select p, trueval, falseval
And expand the select into a branch structure. This later enables
jump-threading over bb in this pass.
Using the similar approach of SimplifyCFG::FoldCondBranchOnPHI(), unfold
select if the associated PHI has at least one constant. If the unfolded
select is not jump-threaded, it will be folded again in the later
optimizations.
llvm-svn: 257198
It's strange that LoopInfo mostly owns the Loop objects, but that it
defers deleting them to the loop pass manager. Instead, change the
oddly named "updateUnloop" to "markAsRemoved" and have it queue the
Loop object for deletion. We can't delete the Loop immediately when we
remove it, since we need its pointer identity still, so we'll mark the
object as "invalid" so that clients can see what's going on.
llvm-svn: 257191
Summary:
r255334 matches bit-reverse pattern in InstCombine and generates calls to Instrinsic::bitreverse.
RBIT instruction is only available for ARMv6t2 and above. This patch has the intrinsic expanded during legalization for ARMv4 and ARMv5.
Patch by Z. Zheng <zhaoshiz@codeaurora.org>
Reviewers: apazos, jmolloy, weimingz
Subscribers: aemerson, rengolin, llvm-commits
Differential Revision: http://reviews.llvm.org/D15932
llvm-svn: 257188
Summary: In ARMConstantIslandPass, which runs after Shrink Wrap pass, long jumps will be fixed up as BL (tBfar) which depends on spilling LR in epilogue. However, shrink-wrap may remove the LR, which causes issues when the function returns.
Reviewers: qcolombet, rengolin
Subscribers: aemerson, rengolin
Differential Revision: http://reviews.llvm.org/D15984
llvm-svn: 257187
Summary:
During legalization if i16, do not ASSERTZEXT the result of FP_TO_FP16.
Directly return an FP_TO_FP16 node with return type as the
promote-to-type of i16.
This patch also removes extraneous length check. This legalization
should be valid even if integer and float types are of different
lengths.
This patch breaks a hard-float test for fp16 args. The test is changed
to allow a vmov to zero-out the top bits, and also ensure that the
return value is in an FP register.
Reviewers: ab, jmolloy
Subscribers: srhines, llvm-commits
Differential Revision: http://reviews.llvm.org/D15438
llvm-svn: 257184
StackColoring rewrites the frame indicies of operations involving
allocas if it can find that the life time of two objects do not overlap.
MSVC EH needs to be kept aware of this if happens in the event that a
catch object has moved around. However, we represent the non-existance
of a catch object with a sentinel frame index (INT_MAX). This sentinel
also happens to be the EmptyKey of the SlotRemap DenseMap. Testing for
whether or not we need to translate the frame index fails in this case
because we call the count method on the DenseMap with the EmptyKey,
leading to assertions. Instead, check if it is our sentinel value
before trying to look into the DenseMap.
This fixes PR26073.
llvm-svn: 257182
Due to the new in-place ThinLTO symbol handling support added in
r257174, we now invoke renameModuleForThinLTO on the current
module from within the FunctionImport pass.
Additionally, renameModuleForThinLTO no longer needs to return the
Module as it is performing the renaming in place on the one provided.
This commit will be immediately preceeded by a companion clang patch to
remove its invocation of renameModuleForThinLTO.
llvm-svn: 257181
Summary:
Move ThinLTO global value processing functions out of ModuleLinker and
into a new ThinLTOGlobalProcessor class, which performs any necessary
linkage and naming changes on the given module in place.
As a result, renameModuleForThinLTO no longer needs to create a new
Module when performing any necessary local to global promotion on a
module that we are possibly exporting from during a ThinLTO backend
compilation.
During function importing the ThinLTO processing is still invoked from
the ModuleLinker (via the new class), as it needs to perform renaming and
linkage changes on the source module, e.g. in order to get the correct
renaming during local to global promotion.
Reviewers: joker.eph
Subscribers: davidxl, llvm-commits, joker.eph
Differential Revision: http://reviews.llvm.org/D15696
llvm-svn: 257174
The function importer was still materializing metadata when modules were
loaded for function importing. We only want to materialize it when we
are going to invoke the metadata linking postpass. Materializing it
before function importing is not only unnecessary, but also causes
metadata referenced by imported functions to be mapped in early, and
then not connected to the rest of the module level metadata when it is
ultimately linked in.
Augmented the test case to specifically check for the metadata being
properly connected, which it wasn't before this fix.
llvm-svn: 257171
This patch corresponds to review:
http://reviews.llvm.org/D15930
Moves to and from CR fields depend on shifts/masks that depend on the
target/source CR field. Thus, post-ra anti-dep breaking must not later
change that CR register assignment.
llvm-svn: 257168
In setInsertionPoint if the value is not a PHI, Instruction or
Argument it should be a Constant, not a ConstantExpr.
Original commit message:
[InstCombine] Look through PHIs, GEPs, IntToPtrs and PtrToInts to expose more constants when comparing GEPs
Summary:
When comparing two GEP instructions which have the same base pointer
and one of them has a constant index, it is possible to only compare
indices, transforming it to a compare with a constant. This removes
one use for the GEP instruction with the constant index, can reduce
register pressure and can sometimes lead to removing the comparisson
entirely.
InstCombine was already doing this when comparing two GEPs if the base
pointers were the same. However, in the case where we have complex
pointer arithmetic (GEPs applied to GEPs, PHIs of GEPs, conversions to
or from integers, etc) the value of the original base pointer will be
hidden to the optimizer and this transformation will be disabled.
This change detects when the two sides of the comparison can be
expressed as GEPs with the same base pointer, even if they don't
appear as such in the IR. The transformation will convert all the
pointer arithmetic to arithmetic done on indices and all the relevant
uses of GEPs to GEPs with a common base pointer. The GEP comparison
will be converted to a comparison done on indices.
Reviewers: majnemer, jmolloy
Subscribers: hfinkel, jevinskie, jmolloy, aadg, llvm-commits
Differential Revision: http://reviews.llvm.org/D15146
llvm-svn: 257164
a top-down manner into a true top-down or RPO pass over the call graph.
There are specific patterns of function attributes, notably the
norecurse attribute, which are most effectively propagated top-down
because all they us caller information.
Walk in RPO over the call graph SCCs takes the form of a module pass run
immediately after the CGSCC pass managers postorder walk of the SCCs,
trying again to deduce norerucrse for each singular SCC in the call
graph.
This removes a very legacy pass manager specific trick of using a lazy
revisit list traversed during finalization of the CGSCC pass. There is
no analogous finalization step in the new pass manager, and a lazy
revisit list is just trying to produce an RPO iteration of the call
graph. We can do that more directly if more expensively. It seems
unlikely that this will be the expensive part of any compilation though
as we never examine the function bodies here. Even in an LTO run over
a very large module, this should be a reasonable fast set of operations
over a reasonably small working set -- the function call graph itself.
In the future, if this really is a compile time performance issue, we
can look at building support for both post order and RPO traversals
directly into a pass manager that builds and maintains the PO list of
SCCs.
Differential Revision: http://reviews.llvm.org/D15785
llvm-svn: 257163
Windows EH keeping track of which frame index corresponds to a catchpad
in order to inform the runtime where the catch parameter should be
initialized. LLVM's optimizations are able to prove that the memory
used by the catch parameter can be reused with another memory
optimization, changing it's frame index.
We need to keep WinEHFuncInfo up to date with respect to this or we will
miscompile/assert.
This fixes PR26069.
llvm-svn: 257158
Done in InstrProfWriter to eliminate the need for client
code to do the sorting. The operation is done once and reused
many times so it is more efficient. Update unit test to remove
sorting. Also update expected output of affected tests.
llvm-svn: 257145
For a new record with weight != 1, only edge profiling
counters are scaled, VP data is not properly scaled.
This patch refactors the code and fixes the problem.
Also added sort by count interface (for follow up patch).
llvm-svn: 257143
This remove the need for locking when deleting a function.
Differential Revision: http://reviews.llvm.org/D15988
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 257139
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
The early return seems to be missed. This causes a radical and wrong loop
optimization on powerpc. It isn't reproducible on x86_64, because
"UseInterleaved" is false.
Patch by Tim Shen.
llvm-svn: 257134
Limit this transform to a basic block and guard against PHIs.
Hopefully, this fixes the remaining failures in PR25999:
https://llvm.org/bugs/show_bug.cgi?id=25999
llvm-svn: 257133
The new leader is known anyway so we can return it for some micro
optimization in code where it is easy to pass along the result to the
next join().
llvm-svn: 257130
.zero is confusing when used with two arguments. Documentation:
This directive emits SIZE 0-valued bytes. SIZE must be an absolute
expression. This directive is actually an alias for the '.skip'
directive so in can take an optional second argument of the value to
store in the bytes instead of zero. Using '.zero' in this way would be
confusing however.
Ref: https://sourceware.org/bugzilla/show_bug.cgi?id=18353
Hexagon and Sparc do the same, and it's all the same to WebAssembly so
let's pick the less confusing of the two.
llvm-svn: 257111
Summary:
Teach the Verifier to make sure that the storage size given to llvm.dbg.declare
or the value size given to llvm.dbg.value agree with what is declared in
DebugInfo. This is implicitly assumed in a number of passes (e.g. in SROA).
Additionally this catches a number of common mistakes, such as passing a
pointer when a value was intended or vice versa.
One complication comes from stack coloring which modifies the original IR when
it merges allocas in order to make sure that if AA falls back to the IR it gets
the correct result. However, given this new invariant, indiscriminately
replacing one alloca by a different (differently sized one) is no longer valid.
Fix this by just undefing out any use of the alloca in a dbg.declare in this
case.
Additionally, I had to fix a number of test cases. Of particular note:
- I regenerated dbg-changes-codegen-branch-folding.ll from the given source as
it was affected by the bug fixed in r256077
- two-cus-from-same-file.ll was changed to avoid having a variable-typed debug
variable as that would depend on the target, even though this test is
supposed to be generic
- I had to manually declared size/align for reference type. See also the
discussion for D14275/r253186.
- fpstack-debuginstr-kill.ll required changing `double` to `long double`
- most others were just a question of adding OP_deref
Reviewers: aprantl
Differential Revision: http://reviews.llvm.org/D14276
llvm-svn: 257105
Summary:
In rL242338, debugger tuning was introduced, and the tuning for FreeBSD
was set to lldb by default. However, for the foreseeable future we
still need to default to gdb tuning, since lldb is not ready for all of
FreeBSD's architectures, and some system tools (like objcopy, etc) have
not yet been adapted to cope with the lldb tuned format, which has
.apple sections.
Therefore, let FreeBSD use gdb by default for now.
Reviewers: emaste, probinson
Subscribers: llvm-commits, emaste
Differential Revision: http://reviews.llvm.org/D15966
llvm-svn: 257103
We marked values which are 'undef' as constant instead of undefined
which violates SCCP's invariants. If we can figure out that a
computation results in 'undef', leave it in the undefined state.
This fixes PR16052.
llvm-svn: 257102
Fix PR24852 (crash with -debug -instcombine)
Patch by Than McIntosh <thanm@google.com>
Summary:
Add guards to the asm writer to prevent crashing
when dumping an instruction that has no basic
block.
Differential Revision: http://reviews.llvm.org/D15798
From: Than McIntosh <thanm@google.com>
llvm-svn: 257094
Coverage mapping data may reference names of functions
that are skipped by FE (e.g, unused inline functions). Since
those functions are skipped, normal instr-prof function lowering
pass won't put those names in the right section, so special
handling is needed to walk through coverage mapping structure
and recollect the references.
With this patch, only names that are skipped are processed. This
simplifies the lowering code and it no longer needs to make
assumptions coverage mapping data layout. It should also be
more efficient.
llvm-svn: 257091
The fix for PR23999 made us mark loads of null as producing the constant
undef which upsets the lattice. Instead, keep the load as "undefined".
This fixes PR26044.
llvm-svn: 257087
Previously we only supported putting the FI into memory operand offset
fields if there was nothing there already. Now combine them.
Differential Revision: http://reviews.llvm.org/D15941
llvm-svn: 257084
The MC assembler doesn't like using the empty string as a private label
prefix because then it treats all labels as private. This commit reverts
back to the default prefix, which is .L, which is common in ELF targets
and consistent with the LLVM name mangler.
llvm-svn: 257083
Summary:
Multi-dword constant loads generated unnecessary moves from SGPRs into VGPRs,
increasing the code size and VGPR pressure. These moves are now folded away.
Note that this lack of operand folding was not a problem for VMEM loads,
because COPY nodes from VReg_Nnn to VGPR32 are eliminated by the register
coalescer.
Some tests are updated, note that the fsub.ll test explicitly checks that
the move is elided.
With the IR generated by current Mesa, the changes are obviously relatively
minor:
7063 shaders in 3531 tests
Totals:
SGPRS: 351872 -> 352560 (0.20 %)
VGPRS: 199984 -> 200732 (0.37 %)
Code Size: 9876968 -> 9881112 (0.04 %) bytes
LDS: 91 -> 91 (0.00 %) blocks
Scratch: 1779712 -> 1767424 (-0.69 %) bytes per wave
Wait states: 295164 -> 295337 (0.06 %)
Totals from affected shaders:
SGPRS: 65784 -> 66472 (1.05 %)
VGPRS: 38064 -> 38812 (1.97 %)
Code Size: 1993828 -> 1997972 (0.21 %) bytes
LDS: 42 -> 42 (0.00 %) blocks
Scratch: 795648 -> 783360 (-1.54 %) bytes per wave
Wait states: 54026 -> 54199 (0.32 %)
Reviewers: tstellarAMD, arsenm, mareko
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15875
llvm-svn: 257074
Summary:
Somehow, I first interpreted the docs as saying space for xnack_mask is only
reserved when XNACK is enabled via SH_MEM_CONFIG. I felt uneasy about this and
went back to actually test what is happening, and it turns out that xnack_mask
is always reserved at least on Tonga and Carrizo, in the sense that flat_scr
is always fixed below the SGPRs that are used to implement xnack_mask, whether
or not they are actually used.
I confirmed this by writing a shader using inline assembly to tease out the
aliasing between flat_scratch and regular SGPRs. For example, on Tonga, where
we fix the number of SGPRs to 80, s[74:75] aliases flat_scratch (so
xnack_mask is s[76:77] and vcc is s[78:79]).
This patch changes both the calculation of the total number of SGPRs and the
various register reservations to account for this.
It ought to be possible to use the gap left by xnack_mask when the feature
isn't used, but this patch doesn't try to do that. (Note that the same applies
to vcc.)
Note that previously, even before my earlier change in r256794, the SGPRs that
alias to xnack_mask could end up being used as well when flat_scr was unused
and the total number of SGPRs happened to fall on the right alignment
(e.g. highest regular SGPR being used s29 and VCC used would lead to number
of SGPRs being 32, where s28 and s29 alias with xnack_mask). So if there
were some conflict due to such aliasing, we should have noticed that already.
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15898
llvm-svn: 257073
Summary:
When comparing two GEP instructions which have the same base pointer
and one of them has a constant index, it is possible to only compare
indices, transforming it to a compare with a constant. This removes
one use for the GEP instruction with the constant index, can reduce
register pressure and can sometimes lead to removing the comparisson
entirely.
InstCombine was already doing this when comparing two GEPs if the
base pointers were the same. However, in the case where we have
complex pointer arithmetic (GEPs applied to GEPs, PHIs of GEPs,
conversions to or from integers, etc) the value of the original
base pointer will be hidden to the optimizer and this transformation
will be disabled.
This change detects when the two sides of the comparison can be
expressed as GEPs with the same base pointer, even if they don't
appear as such in the IR. The transformation will convert all the
pointer arithmetic to arithmetic done on indices and all the
relevant uses of GEPs to GEPs with a common base pointer. The
GEP comparison will be converted to a comparison done on indices.
Reviewers: majnemer, jmolloy
Subscribers: hfinkel, jevinskie, jmolloy, aadg, llvm-commits
Differential Revision: http://reviews.llvm.org/D15146
llvm-svn: 257064
See PR25822 for a more full summary, but we were conflating the concepts of "capture" and "escape". We were proving nocapture and using that proof to infer noescape, which is not true. Escaped-ness is a function-local property - as soon as a value is used in a call argument it escapes. Capturedness is a related but distinct property. It implies a *temporally limited* escape. Consider:
static int a;
int b;
int g(int * nocapture arg);
int f() {
a = 2; // Even though a escapes to g, it is not captured so can be treated as non-escaping here.
g(&a); // But here it must be treated as escaping.
g(&b); // Now that g(&a) has returned we know it was not captured so we can treat it as non-escaping again.
}
The original commit did not sufficiently understand this nuance and so caused PR25822 and PR26046.
r248576 included both a performance improvement (which has been backed out) and a related conformance fix (which has been kept along with its testcase).
llvm-svn: 257058
AVX1 v8i32/v4i64 shuffles are bitcasted to v8f32/v4f64, this patch peeks through bitcasts to check for a load node to allow broadcasts to occur.
Follow up to D15310
llvm-svn: 257055
Change Triple::get32BitArchVariant to return arm/armeb as the 32bit
variant of aarch64/aarch64_be and do the same change for the oppoiste
direction in Triple::get64BitArchVariant.
Differential revision: http://reviews.llvm.org/D15529
llvm-svn: 257048
Follow up to D15378, added INSERTPS to the list of decodable target shuffles and enabled XFormVExtractWithShuffleIntoLoad to handle target shuffles with SentinelZero and tested this with INSERTPS.
llvm-svn: 257046
Darwin TLS accesses most closely resemble ELF's general-dynamic situation,
since they have to be able to handle all possible situations. The descriptors
and so on are obviously slightly different though.
llvm-svn: 257039
Serialize will perform a hardware serialization operation, and is
acting as a memory barrier. Therefore it must have the hasSideEffects
flag set so it will be treated as a global memory object.
Reviewed by Ulrich Weigand
llvm-svn: 257036
Unlike my comment in 257022 said, it turns out we do handle constant vectors in the statepoint lowering, but only because SelectionDAG doesn't actually produce constants for them. Add a couple of tests which show this working.
Also, add a triple to the same test file to hopefully fix a failing bot.
It turns out we do han
llvm-svn: 257025
Currently, we try to split vectors of pointers back into their component pointer elements during rewrite-statepoints-for-gc. This is less than ideal since presumably the vectorizer chose to vectorize for a reason. :) It's also been a source of bugs - in particular, the relocation logic as currently implemented was recently discovered to be wrong.
The alternate approach is to allow gc.relocates of vector-of-pointer type and update the backend to handle them. That's what this patch tries to do. This won't actually enable vector-of-pointers in practice - there are some RS4GC changes needed - but the lowering is standalone and testable so it makes sense to separate.
Note that there are some known cases around vector constants which this patch does not handle. Once this is in, I'll send another patch with individual fixes and test cases.
Differential Revision: http://reviews.llvm.org/D15632
llvm-svn: 257022
Follow-up to r257000: DIImportedEntity can reach a DISubprogram via
its entity, but also via its scope. Handle the latter case as well.
PR26037.
llvm-svn: 257019
At the moment, this is essentially a diangostic option so that I can start collecting failing test cases, but we will eventually migrate to removing the vector splitting code entirely.
llvm-svn: 257015
We need to know whether or not a given basic block is in a loop for the analysis
to be correct.
Loop information may be incomplete on irreducible CFGs, therefore we may
generate incorrect code if we use it in those situations.
This fixes PR25988.
llvm-svn: 257012
Summary:
r252532 added support for reporting the monolithic library
when LLVM_BUILD_LLVM_DYLIB is used. This would only be done
if the individual components were not found, and the dynamic
library is found.
This diff extends this as follows:
- If LLVM_LINK_LLVM_DYLIB is set, then prefer the shared
library, even if all component libraries exist.
- Two flags, --link-shared and --link-static are introduced
to provide explicit guidance. If --link-shared is passed
and the shared library does not exist, an error results.
Additionally, changed the expected shared library names from
(e.g.) LLVM-3.8.0 to LLVM-3.8. The former exists only in an
installation (and then only in CMake builds I think?), and not
in the build tree; this breaks usage of llvm-config during
builds, e.g. by llvm-go.
Reviewers: DiamondLovesYou, beanz
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D15033
llvm-svn: 257003
It is illegal to have a null entity in a DIImportedEntity, so
we must link in a DISubprogram metadata node referenced by one,
even if the associated function is not linked in or inlined anywhere.
Fixes PR26037.
llvm-svn: 257000
getTargetShuffleMask may return shuffle masks with SM_SentinelZero (-2) values (currently just for PSHUFB but VPERM2X128 as well with this patch). Although some calling functions can make use of this (mainly for shuffle combining), others can not and their inclusion makes shuffle mask comparisons more difficult.
This patch adds a flag to getTargetShuffleMask to indicate if the calling function can't handle SM_SentinelZero; getTargetShuffleMask will then return false if it occurs to make handling much easier.
I've tidied up some uses of getTargetShuffleMask to better indicate what is going on - more could be done but at present I don't have test cases to demonstrate it.
Some upcoming patches will make use of this to both support more uses where SM_SentinelZero is not permitted (e.g. combineShuffleToAddSub), and also will allow us to add INSERTPS support to getTargetShuffleMask as part of better zero handling discussed in D14261.
Differential Revision: http://reviews.llvm.org/D15378
llvm-svn: 256992
With r256990, bogner introduced comprehensive tests for constant arrays
and vectors. We no longer need the existing ones because they are
redundant.
llvm-svn: 256991
I added a couple of tests in r256982, but vedantk suggested that they
fit better into compatibility.ll, since they could catch format breaks
later on there.
llvm-svn: 256990
In r254991 I allowed ConstantDataVectors to contain elements of
HalfTy, but I missed updating the bitcode reader and writer to handle
this, so now we crash if we try to emit bitcode on programs that have
constant vectors of half.
This fixes the issue and adds test coverage for reading and writing
constant sequences in bitcode.
llvm-svn: 256982
Summary:
This is admittedly something that you could only run into by manually
playing around with shader assembly because the SITypeWriter pass is
skipped for compute.
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15902
llvm-svn: 256980
Summary: This patch adds a check in SplitLandingPadPredecessors to see if the original landingpad instruction has any uses. If not, we don't need to create a PHINode for it in the joint block since it's gonna be a dead code anyway. The motivation for this patch is that we found a bug that SplitLandingPadPredecessors created a PHINode of token type landingpad, which failed the verifier since PHINode can not be token type. However, the created PHINode will never be used in our code pattern. This patch will workaround this bug, and we might add supports in SplitLandingPadPredecessors to handle token type landingpad with uses in the future.
Reviewers: reames
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D15835
llvm-svn: 256972
Summary: As per title. This will allow the optimizer to pick up on it.
Reviewers: craig.topper, spatel, dexonsmith, Prazek, chandlerc, joker.eph, majnemer
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D15923
llvm-svn: 256969
Move the logic from BranchFolding to use the shared infrastructure for merging MMOs introduced in 256909. This has the effect of making BranchFolding more capable.
In the process, fix a latent bug. The existing handling for merging didn't handle the case where one of the instructions being merged had overflowed and dropped MemRefs. This was a latent bug in the places the code was commoned from, but potentially reachable in BranchFolding.
Once this is in, we're left with a single place to consider implementing MMO unique-ing as proposed in http://reviews.llvm.org/D15230.
Differential Revision: http://reviews.llvm.org/D15913
llvm-svn: 256966
The functionality that calculateCatchReturnSuccessorColors provides was
once non-trivial: it was a computation layered on top of funclet
coloring.
These days, LLVM IR directly encodes what
calculateCatchReturnSuccessorColors computed, obsoleting the need for
it.
No functionality change is intended.
llvm-svn: 256965
LLVM_ENABLE_TIMESTAMPS controls if timestamps are embedded into llvm's
binaries. Turning it off is useful for deterministic builds.
r246905 made it so that the define suddenly also controls if the binaries that
the llvm binaries _create_ embed timestamps or not – but this shouldn't be a
configure-time option. r256203/r256204 added a driver option to toggle this on
and off, so this patch now passes this driver option in LLVM_ENABLE_TIMESTAMPS
builds so that if LLVM_ENABLE_TIMESTAMPS is set, the build of LLVM is
deterministic – but the built clang can still write timestamps into other
executables when requested.
This also allows removing some of the test machinery added in r292012 to work
around this problem.
See PR24740 for background.
http://reviews.llvm.org/D15783
llvm-svn: 256958
FindIDom() can fail in two different ways - it can either return nullptr or the
block itself, depending on the circumstances. Some users of FindIDom() check
one error condition, while others check the other.
Change it to always return nullptr on failure.
This fixes PR26004.
Differential Revision: http://reviews.llvm.org/D15847
llvm-svn: 256955
The first instruction in a block is what the rend() iterator points to, so
if it moves, we need to re-evaluate rend() so that we continue to iterate
through the rest of the instructions.
llvm-svn: 256953
Summary:
This patch implements "-print-funcs" option to support function filtering for IR printing like -print-after-all, -print-before etc.
Examples:
-print-after-all -print-funcs=foo,bar
Reviewers: mcrosier, joker.eph
Subscribers: tejohnson, joker.eph, llvm-commits
Differential Revision: http://reviews.llvm.org/D15776
llvm-svn: 256952
Summary: The example in desc should match with actual option name
Reviewers: jmolloy
Differential Revision: http://reviews.llvm.org/D15800
llvm-svn: 256951
Summary:
In buildSchedGraph(), when adding memory dependencies for loads, move
the call to adjustChainDeps() after the call to
addChainDependency(AliasChain) to handle the case where
addChainDependency(AliasChain) ends up not adding a dependency and
instead putting the SU on the RejectMemNodes list. The call to
adjustChainDeps() must be done after the call to addChainDependency() in
order to process the SU added to the RejectMemNodes list to create
memory dependencies for it.
Reviewers: hfinkel, atrick, jonpa, resistor
Subscribers: mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D15927
llvm-svn: 256950
Since writeonly is the only missing attribute and special case left for the memset/memcpy family of intrinsics, rearrange the code to make that much more clear.
llvm-svn: 256949
In general, disabling comments in the output reduces the chances of a
CHECK line accidentally matching a comment instead of its intended text.
llvm-svn: 256946
This adds a unittest for the support added in r256648 to add
a flag that can be used to prevent RAUW on temporary metadata
used as a map key.
llvm-svn: 256938
Summary:
This reverts commit 5a9e526f29cf8510ab5c3d566fbdcf47ac24e1e9.
As per discussion in D15665
This also add a test case so that regression introduced by that diff are not reintroduced.
Reviewers: vaivaswatha, jmolloy, hfinkel, reames
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D15919
llvm-svn: 256932
This patch prevents us from unintentionally creating entries in the reductions
map for PHIs that are not actually reductions. This is currently not an issue
since we bail out if we encounter PHIs other than inductions or reductions.
However the behavior could become problematic as we add support for additional
recurrence types.
llvm-svn: 256930
Summary: It turns out that if we don't try to do it at the store location, we can do it before any operation that alias the load, as long as no operation alias the store.
Reviewers: craig.topper, spatel, dexonsmith, Prazek, chandlerc, joker.eph
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D15903
llvm-svn: 256923
As discussed on D15378, move the mask.empty() tests to after the switch statement and consider any shuffle decode where the extracted target shuffle mask is empty as a failure.
llvm-svn: 256921
...and mark it as merely an input_iterator rather than a forward_iterator,
since it is destructive. And then rewrite == to take advantage of that.
Patch by Alex Denisov!
llvm-svn: 256913
If we replace one call-site with another, be sure to move over any
operand bundles that lingered on the old call-site.
This fixes PR26036.
llvm-svn: 256912
Most of the properties of memset_pattern16 can be now covered by the generic attributes and inferred by InferFunctionAttrs. The only exceptions are:
- We don't yet have a writeonly attribute for the first argument.
- We don't have an attribute for modeling the access size facts encoded in MemoryLocation.cpp.
Differential Revision: http://reviews.llvm.org/D15879
llvm-svn: 256911
We only need to describe the writeonly property of one of the arguments. All of the rest of the semantics are nicely described by existing attributes in Intrinsics.td.
Differential Revision: http://reviews.llvm.org/D15880
llvm-svn: 256910
In the discussion on http://reviews.llvm.org/D15730, Andy pointed out we had a utility function for merging MMO lists. Since it turned we actually had two copies and there's another review in progress (http://reviews.llvm.org/D15230) which needs the same, extract it into a utility function and clean up the interfaces to make it easier to use with a MachineInstBuilder.
I introduced a pair here to track size and allocation together. I think we should probably move in the direction of the MachineOperandsRef helper class, but I'm leaving that for further work. I want to get the poison state introduced before I make major changes to the interface.
Differential Revision: http://reviews.llvm.org/D15757
llvm-svn: 256909
confused with what version of mingw is actually installed on the buildbot, and
for now I will just assume this is an unknown version which does not ship with
VersionHelpers.h.
llvm-svn: 256902
This patch is similar to the Python issue#11395. We need to cap the output
size to 32767 on Windows to work around the size limit of WriteConsole().
Reference: https://bugs.python.org/issue11395
Writing a test for this bug turns out to be harder than I thought. I am
still working on it (see phabricator review D15705).
Differential Revision: http://reviews.llvm.org/D15553
llvm-svn: 256892
In an inbounds getelementptr, when an index produces a constant non-negative
offset to add to the base, the add can be assumed to not have unsigned overflow.
This relies on the assumption that addresses can't occupy more than half the
address space, which isn't possible in C because it wouldn't be possible to
represent the difference between the start of the object and one-past-the-end
in a ptrdiff_t.
Setting the NoUnsignedWrap flag is theoretically useful in general, and is
specifically useful to the WebAssembly backend, since it permits stronger
constant offset folding.
Differential Revision: http://reviews.llvm.org/D15544
llvm-svn: 256890
Due to the SGPR init bug, every program claims to use the same number
of SGPRs anyway, so there's no point in trying to shift those registers
down from their initial spot of reservation.
Add a test that uses VGPR spilling and blocks most SGPRs from being used for
the scratch resource register. Previously, this would run into an assertion.
Differential Revision: http://reviews.llvm.org/D15724
llvm-svn: 256870
Lang Hames