In our real world application, we found the following optimization is missed in DAGCombiner
(zext (and/or/xor (shl/shr (load x), cst), cst)) -> (and/or/xor (shl/shr (zextload x), (zext cst)), (zext cst))
If the user of original zext is an add, it may enable further lea optimization on x86.
This patch add a new function CombineZExtLogicopShiftLoad to do this optimization.
Differential Revision: https://reviews.llvm.org/D44402
llvm-svn: 328252
As in SystemZ backend, correctly propagate node ids when inserting new
unselected nodes into the DAG during instruction Seleciton for X86
target.
Fixes PR36865.
Reviewers: jyknight, craig.topper
Subscribers: hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D44797
llvm-svn: 328233
Summary:
When building the selection DAG we sometimes need to postpone
the handling of a dbg.value until the value it should refer to
is created. This is done by using the DanglingDebugInfoMap.
In the past this map has been limited to hold one dangling
dbg.value per value. This patch removes that restriction.
Reviewers: aprantl, rnk, probinson, vsk
Reviewed By: aprantl
Subscribers: Ka-Ka, llvm-commits, JDevlieghere
Tags: #debug-info
Differential Revision: https://reviews.llvm.org/D44610
llvm-svn: 328084
I'm not entirely sure these hacks are still needed. If you remove the hacks completely, the name of the library call that gets generated doesn't match the grep the test previously had. So the test wasn't really checking anything.
If the hack is still needed it belongs in PPC specific code. I believe the FP_TO_SINT code here is the only place in the tree where a FP_ROUND_INREG node is created today. And I don't think its even being used correctly because the legalization returned a BUILD_PAIR with the same value twice. That doesn't seem right to me. By moving the code entirely to PPC we can avoid creating the FP_ROUND_INREG at all.
I replaced the grep in the existing test with full checks generated by hacking update_llc_test_check.py to support ppc32 just long enough to generate it.
Differential Revision: https://reviews.llvm.org/D44061
llvm-svn: 328017
Summary:
DbgValue nodes were not transferred when integer DAG nodes were promoted. For example, if an i32 add node was promoted to an i64 add node by DAGTypeLegalizer::PromoteIntegerResult(), its DbgValue node was not transferred to the new node. The simple fix is to update SetPromotedInteger() to transfer DbgValues.
Add AArch64/dbg-value-i8.ll to test this change and fix ARM/debug-info-d16-reg.ll which had the wrong DILocalVariable nodes with arg numbers even though they are not for function parameters.
Patch by Se Jong Oh!
Reviewers: vsk, JDevlieghere, aprantl
Reviewed By: JDevlieghere
Subscribers: javed.absar, kristof.beyls, llvm-commits
Tags: #debug-info
Differential Revision: https://reviews.llvm.org/D44546
llvm-svn: 327919
X86 Supports Indirect Branch Tracking (IBT) as part of Control-Flow Enforcement Technology (CET).
IBT instruments ENDBR instructions used to specify valid targets of indirect call / jmp.
The `nocf_check` attribute has two roles in the context of X86 IBT technology:
1. Appertains to a function - do not add ENDBR instruction at the beginning of the function.
2. Appertains to a function pointer - do not track the target function of this pointer by adding nocf_check prefix to the indirect-call instruction.
This patch implements `nocf_check` context for Indirect Branch Tracking.
It also auto generates `nocf_check` prefixes before indirect branchs to jump tables that are guarded by range checks.
Differential Revision: https://reviews.llvm.org/D41879
llvm-svn: 327767
The BITCAST handling in computeKnownBits() previously only worked for little
endian.
This patch reverses the iteration over elements for a big endian target which
allows this to work in this case also.
SystemZ test case.
Review: Eli Friedman
https://reviews.llvm.org/D44249
llvm-svn: 327764
Previously if getSetccResultType returned an illegal type we just fell back to using the default promoted type. This appears to have been to handle the case where for vectors getSetccResultType returns the input type, but the input type itself isn't legal and will need to be promoted. Without the legality check we would never reach a legal type.
But just picking the promoted type to be the setcc type can create strange setccs where the result type is 128 bits and the operand type is 256 bits. If for example the result type was promoted to v8i16 from v8i1, but the input type was promoted from v8i23 to v8i32. We currently handle this with custom lowering code in X86.
This legality check also caused us reject the getSetccResultType when the input type needed to be widened or split. Even though that result wouldn't have caused legalization to get stuck.
This patch tries to fix this by detecting the getSetccResultType needs to be promoted. If its input type also needs to be promoted we'll try a ask for a new setcc result type based on its eventual promoted value. Otherwise we fall back to default type to promote to.
For any other illegal values we might get back from the initial call to getSetccResultType we just keep and allow it to be re-legalized later via splitting or widening or scalarizing.
llvm-svn: 327683
Summary:
Local values are constants, global addresses, and stack addresses that
can't be folded into the instruction that uses them. For example, when
storing the address of a global variable into memory, we need to
materialize that address into a register.
FastISel doesn't want to materialize any given local value more than
once, so it generates all local value materialization code at
EmitStartPt, which always dominates the current insertion point. This
allows it to maintain a map of local value registers, and it knows that
the local value area will always dominate the current insertion point.
The downside is that local value instructions are always emitted without
a source location. This is done to prevent jumpy line tables, but it
means that the local value area will be considered part of the previous
statement. Consider this C code:
call1(); // line 1
++global; // line 2
++global; // line 3
call2(&global, &local); // line 4
Today we end up with assembly and line tables like this:
.loc 1 1
callq call1
leaq global(%rip), %rdi
leaq local(%rsp), %rsi
.loc 1 2
addq $1, global(%rip)
.loc 1 3
addq $1, global(%rip)
.loc 1 4
callq call2
The LEA instructions in the local value area have no source location and
are treated as being on line 1. Stepping through the code in a debugger
and correlating it with the assembly won't make much sense, because
these materializations are only required for line 4.
This is actually problematic for the VS debugger "set next statement"
feature, which effectively assumes that there are no registers live
across statement boundaries. By sinking the local value code into the
statement and fixing up the source location, we can make that feature
work. This was filed as https://bugs.llvm.org/show_bug.cgi?id=35975 and
https://crbug.com/793819.
This change is obviously not enough to make this feature work reliably
in all cases, but I felt that it was worth doing anyway because it
usually generates smaller, more comprehensible -O0 code. I measured a
0.12% regression in code generation time with LLC on the sqlite3
amalgamation, so I think this is worth doing.
There are some special cases worth calling out in the commit message:
1. local values materialized for phis
2. local values used by no-op casts
3. dead local value code
Local values can be materialized for phis, and this does not show up as
a vreg use in MachineRegisterInfo. In this case, if there are no other
uses, this patch sinks the value to the first terminator, EH label, or
the end of the BB if nothing else exists.
Local values may also be used by no-op casts, which adds the register to
the RegFixups table. Without reversing the RegFixups map direction, we
don't have enough information to sink these instructions.
Lastly, if the local value register has no other uses, we can delete it.
This comes up when fastisel tries two instruction selection approaches
and the first materializes the value but fails and the second succeeds
without using the local value.
Reviewers: aprantl, dblaikie, qcolombet, MatzeB, vsk, echristo
Subscribers: dotdash, chandlerc, hans, sdardis, amccarth, javed.absar, zturner, llvm-commits, hiraditya
Differential Revision: https://reviews.llvm.org/D43093
llvm-svn: 327581
Get rid of the "; mem:" suffix and use the one we use in MIR: ":: (load 2)".
rdar://38163529
Differential Revision: https://reviews.llvm.org/D42377
llvm-svn: 327580
I had to modify the bswap recognition to allow unshrunk masks to make this work.
Fixes PR36689.
Differential Revision: https://reviews.llvm.org/D44442
llvm-svn: 327530
BUILD_VECTORs aren't themselves legalized until LegalizeDAG so we should still be able to create an "illegal" one before that. This helps combine with BUILD_VECTORS that are introduced during LegalizeVectorOps due to unrolling.
llvm-svn: 327446
Under some circumstances the divrems won't have been combined together before getting to this code.
So replace the assertion with a if() guard to not expand to X-((X/C)*C) to give the other combine chance to happen.
Reduced from OSS-Fuzz #6883https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=6883
llvm-svn: 327424
Summary:
This change is part of step five in the series of changes to remove alignment argument from
memcpy/memmove/memset in favour of alignment attributes. In particular, this changes the
SelectionDAGBuilder to cease using the old getAlignment() API of MemoryIntrinsic in favour of getting
source & dest specific alignments through the new API.
Steps:
Step 1) Remove alignment parameter and create alignment parameter attributes for
memcpy/memmove/memset. ( rL322965, rC322964, rL322963 )
Step 2) Expand the IRBuilder API to allow creation of memcpy/memmove with differing
source and dest alignments. ( rL323597 )
Step 3) Update Clang to use the new IRBuilder API. ( rC323617 )
Step 4) Update Polly to use the new IRBuilder API. ( rL323618 )
Step 5) Update LLVM passes that create memcpy/memmove calls to use the new IRBuilder API,
and those that use use MemIntrinsicInst::[get|set]Alignment() to use [get|set]DestAlignment()
and [get|set]SourceAlignment() instead. ( rL323886, rL323891, rL324148, rL324273, rL324278,
rL324384, rL324395, rL324402, rL324626, rL324642, rL324653, rL324654, rL324773, rL324774,
rL324781, rL324784, rL324955, rL324960, rL325816, rL327398 )
Step 6) Remove the single-alignment IRBuilder API for memcpy/memmove, and the
MemIntrinsicInst::[get|set]Alignment() methods.
Reference
http://lists.llvm.org/pipermail/llvm-dev/2015-August/089384.htmlhttp://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20151109/312083.html
llvm-svn: 327421
Summary:
1) Make sure to discard dangling debug info if the variable (or
variable fragment) is mapped to something new before we had a
chance to resolve the dangling debug info.
2) When resolving debug info, make sure to bump the associated
SDNodeOrder to ensure that the DBG_VALUE is emitted after the
instruction that defines the value used in the DBG_VALUE.
This will avoid a debug-use before def scenario as seen in
https://bugs.llvm.org/show_bug.cgi?id=36417.
The new test case, test/DebugInfo/X86/sdag-dangling-dbgvalue.ll,
show some other limitations in how dangling debug info is
handled in the SelectionDAG. Since we currently only support
having one dangling dbg.value per Value, we will end up dropping
debug info when there are more than one variable that is described
by the same "dangling value".
Reviewers: aprantl
Reviewed By: aprantl
Subscribers: aprantl, eraman, llvm-commits, JDevlieghere
Tags: #debug-info
Differential Revision: https://reviews.llvm.org/D44369
llvm-svn: 327303
r327171 "Improve Dependency analysis when doing multi-node Instruction Selection"
r328170 "[DAG] Enforce stricter NodeId invariant during Instruction selection"
Reverting patch as NodeId invariant change is causing pathological
increases in compile time on PPC
llvm-svn: 327197
Relanding after fixing NodeId Invariant.
Cleanup cycle/validity checks in ISel (IsLegalToFold,
HandleMergeInputChains) and X86 (isFusableLoadOpStore). Now do a full
search for cycles / dependencies pruning the search when topological
property of NodeId allows.
As part of this propogate the NodeId-based cutoffs to narrow
hasPreprocessorHelper searches.
Reviewers: craig.topper, bogner
Subscribers: llvm-commits, hiraditya
Differential Revision: https://reviews.llvm.org/D41293
llvm-svn: 327171
Instruction Selection makes use of the topological ordering of nodes
by node id (a node's operands have smaller node id than it) when doing
cycle detection. During selection we may violate this property as a
selection of multiple nodes may induce a use dependence (and thus a
node id restriction) between two unrelated nodes. If a selected node
has an unselected successor this may allow us to miss a cycle in
detection an invalid selection.
This patch fixes this by marking all unselected successors of a
selected node have negated node id. We avoid pruning on such negative
ids but still can reconstruct the original id for pruning.
In-tree targets have been updated to replace DAG-level replacements
with ISel-level ones which enforce this property.
This preemptively fixes PR36312 before triggering commit r324359 relands
Reviewers: craig.topper, bogner, jyknight
Subscribers: arsenm, nhaehnle, javed.absar, llvm-commits, hiraditya
Differential Revision: https://reviews.llvm.org/D43198
llvm-svn: 327170
The retpoline mitigation for variant 2 of CVE-2017-5715 inhibits the
branch predictor, and as a result it can lead to a measurable loss of
performance. We can reduce the performance impact of retpolined virtual
calls by replacing them with a special construct known as a branch
funnel, which is an instruction sequence that implements virtual calls
to a set of known targets using a binary tree of direct branches. This
allows the processor to speculately execute valid implementations of the
virtual function without allowing for speculative execution of of calls
to arbitrary addresses.
This patch extends the whole-program devirtualization pass to replace
certain virtual calls with calls to branch funnels, which are
represented using a new llvm.icall.jumptable intrinsic. It also extends
the LowerTypeTests pass to recognize the new intrinsic, generate code
for the branch funnels (x86_64 only for now) and lay out virtual tables
as required for each branch funnel.
The implementation supports full LTO as well as ThinLTO, and extends the
ThinLTO summary format used for whole-program devirtualization to
support branch funnels.
For more details see RFC:
http://lists.llvm.org/pipermail/llvm-dev/2018-January/120672.html
Differential Revision: https://reviews.llvm.org/D42453
llvm-svn: 327163
The code to match and produce more x86 vector blends was enabled for all
architectures even though the transform may pessimize the code for other
architectures that do not provide a vector blend instruction.
Added an aarch64 testcase to check that a VZIP instruction is generated instead
of byte movs.
Differential Revision: https://reviews.llvm.org/D44118
llvm-svn: 327132
This patch is a fix for PR36642.
While legalizing long vector types, make sure the smaller types get the
flags of the wider type.
bugzilla link: https://bugs.llvm.org/show_bug.cgi?id=36642
Change-Id: I0c2829639f094c862c10a6b51b342d4c2563e1fa
llvm-svn: 327079
Loading a constant into a k-register in AVX512 requires a bitcast from a scalar constant. In the test case here we have a k-register store that gets split into multiple parts of KNL. MergeConsecutiveStores sees each of these pieces as a consecutive store and looks through the bitcast to find the underly scalar constant. But when we went to create the combined store we didn't look through the same bitcast.
llvm-svn: 326677
X86 considers v1i1 a legal type under AVX512 and as such a truncate from a v1iX type to v1i1 can be turned into a scalar truncate plus a conversion to v1i1. We would much prefer a v1i1 SCALAR_TO_VECTOR over a one element BUILD_VECTOR.
During lowering we were detecting the v1i1 BUILD_VECTOR as a splat BUILD_VECTOR like we try to do for v2i1/v4i1/etc. In this case we create (select i1 splat_elt, v1i1 all-ones, v1i1 all-zeroes). That goes through some more legalization and we end up with a CMOV choosing between 0 and 1 in scalar and a scalar_to_vector.
Arguably we could detect the v1i1 BUILD_VECTOR and do this better in X86 target code. But just using a SCALAR_TO_VECTOR in legalization is much easier.
llvm-svn: 326637
The fast/linear DAG scheduler doesn't lower DBG_VALUEs except for
function entry nodes.
Patch by Joshua Cranmer!
Differential Revision: https://reviews.llvm.org/D43028
llvm-svn: 326631
Masking first, prevents the extend from being combine with loads. Its also interfering with some vXi1 extraction code.
Differential Revision: https://reviews.llvm.org/D42679
llvm-svn: 326500
This supports things like
(setcc ugt X, 0) -> (setcc ne X, 0)
I've restricted to only make changes to vectors before legalize ops because I doubt all targets have accurate condition code legality information for vectors given how little we did before.
Differential Revision: https://reviews.llvm.org/D42948
llvm-svn: 326495
AVX512 used to promote v32i1 to v32i8 during legalization when BWI was disabled. So this code was added to improve legalization of v32i1 concat_vectors of v16i1 by extending the v16i1 to v16i8 to avoid scalarization.
X86 has since switched to legalizing v32i1 by splitting to v16i1 instead. This has rendered this code unnecessary and its no longer exercised.
llvm-svn: 326153
Summary:
There are transformation that change setcc into other constructs, and transform that try to reconstruct a setcc from the brcond condition. Depending on what order these transform are done, the end result differs.
Most of the time, it is preferable to get a setcc as a brcond argument (and this is why brcond try to recreate the setcc in the first place) so we ensure this is done every time by also doing it at the setcc level when the only user is a brcond.
Reviewers: spatel, hfinkel, niravd, craig.topper
Subscribers: nhaehnle, llvm-commits
Differential Revision: https://reviews.llvm.org/D41235
llvm-svn: 325892
isCondCodeLegal internally checked Legal or Custom which is misleading. Though no targets set any cond code action to Custom today.
So I've renamed isCondCodeLegal to isCondCodeLegalOrCustom and added a real isCondCodeLegal that only checks Legal.
I've changed legalization code to use isCondCodeLegalOrCustom and left things reachable via DAG combine as isCondCodeLegal. I've also changed some places that called getCondCodeAction and compared to Legal to just use isCondCodeLegal.
I'm looking at trying to keep SETCC all the way to isel for the AVX512 integer comparisons and I suspect I'll need to make some condition codes Custom to stop DAG combine from changing things post LegalizeOps. Prior to this only Expand stopped DAG combine, but that causes LegalizeOps to try to swap operands or invert rather than calling our Custom handler.
Differential Revision: https://reviews.llvm.org/D43607
llvm-svn: 325829
This patch reverts r325440 and r325438 because it triggers an
assertion in SelectionDAGBuilder.cpp. Also having debug enabled
may unintentionally affect code-gen. The patch is reverted until
we find a better solution.
llvm-svn: 325825
This allows us to improve vector constant matching in more DAG code (backends, TargetLowering etc.).
Differential Revision: https://reviews.llvm.org/D43466
llvm-svn: 325815
We looked through a BITCAST, but the bitcast might be a from a scalar type rather than a vector.
I don't have a test case. I stumbled onto it while prototyping another change that isn't ready yet.
llvm-svn: 325750
This is split off from D42948 and includes just the cases that constant fold to true or false. It also includes some refactoring to keep predicate checks together.
This supports things like
(setcc uge X, 0) -> true
Differential Revision: https://reviews.llvm.org/D43489
llvm-svn: 325627
DAGCombiner and SimplifySetCC both use getPointerTy for shift amounts pre-legalization. DAGCombiner uses a single helper function to hide this. SimplifySetCC does it in multiple places.
This patch adds a defaulted parameter to getShiftAmountTy that can make it return getPointerTy for scalar types. Use this parameter to simplify the SimplifySetCC and DAGCombiner.
Additionally, there were two places in SimplifySetCC that were creating shifts using the target's preferred shift amount pre-legalization. If the target uses a narrow type and the type is illegal, this can cause SimplfiySetCC to create a shift with an amount that can't represent all possible shift values for the type. To fix this we should use pointer type there too.
Alternatively we could make getScalarShiftAmountTy for each target return a safe value for large types as proposed in D43445. And maybe we should still do that, but fixing the SimplifySetCC code keeps other targets from tripping over this in the future.
Fixes PR36250.
Differential Revision: https://reviews.llvm.org/D43449
llvm-svn: 325602
ExpandUINT_TO_FLOAT can accept vXi32 or vXi64 inputs, so we need to use a uint64_t shift to generate the 2^(BW/2) constant.
No test case unfortunately as no upstream target uses this, but its affecting a downstream target.
llvm-svn: 325578
If we have a clamp pattern, SMIN(SMAX(X, LO),HI) or SMAX(SMIN(X, HI),LO) then we can deduce that the number of signbits (zeros/ones) will be at least the minimum of the LO and HI constants.
ComputeKnownBits equivalent of D43338.
Differential Revision: https://reviews.llvm.org/D43463
llvm-svn: 325521
If we have a clamp pattern, SMIN(SMAX(X, LO),HI) or SMAX(SMIN(X, HI),LO) then we can deduce that the number of signbits will be at least the minimum of the LO and HI constants.
I haven't bothered with the UMIN/UMAX equivalent as (1) we don't have any current use cases and (2) I wonder if we'd be better off immediately falling back for ComputeKnownBits for UMIN/UMAX which already has optimization patterns useful for unsigned cases.
Differential Revision: https://reviews.llvm.org/D43338
llvm-svn: 325450
Summary:
https://llvm.org/PR36263 shows that when compiling at -O0 a dbg.value()
instruction (that remains from an original dbg.declare()) is dropped
by FastISel. Since FastISel selects instructions by iterating a basic
block backwards, it drops the dbg.value if one of its operands is not
yet instantiated by a previously selected instruction.
Instead of calling 'lookUpRegForValue()' we can call 'getRegForValue()'
instead that will insert a placeholder for the operand to be filled in
when continuing the instruction selection.
Reviewers: aprantl, dblaikie, probinson
Reviewed By: aprantl
Subscribers: llvm-commits, dstenb, JDevlieghere
Differential Revision: https://reviews.llvm.org/D43386
llvm-svn: 325438
Sadly, r324359 caused at least PR36312. There is a patch out for review
but it seems to be taking a bit and we've already had these crashers in
tree for too long. We're hitting this PR in real code now and are
blocked on shipping new compilers as a consequence so I'm reverting us
back to green.
Sorry for the churn due to the stacked changes that I had to revert. =/
llvm-svn: 325420
Based off the DemandedElts mask the and UNDEF elements returned from the SimplifyDemandedVectorElts calls to the shuffle operands, we can attempt to simplify the shuffle mask.
I had to be very conservative here as accepting post-legalized shuffle masks could cause problems for targets that legalize UNDEF mask elements back to inrange values (PowerPC), similarly combining to identity shuffle masks could cause too much UNDEF information to disappear for later combines.
llvm-svn: 325354
Summary:
Currently when expanding a SETCC node into a SELECT_CC, LLVM uses
an incorrect type for determining BooleanContent of the result. This
patch fixes the issue.
Fixes PR36079.
Reviewers: rogfer01, javed.absar, efriedma
Reviewed By: efriedma
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D43282
llvm-svn: 325325
Same for the sign extend case.
Currently we check for multiple uses on the binop. Then we call ExtendUsesToFormExtLoad to capture SetCCs that use the load. So we only end up finding any setccs when the and has additional uses and the load is used by a setcc. I don't think the and having multiple uses is relevant here. I think we should only be checking for the load having multiple uses.
This changes an NVPTX test because we now find that the load has a second use by a truncate, but ExtendUsesToFormExtLoad only looks at setccs it can extend. All other operations just check isTruncateFree. Maybe we should allow widening of an existing truncate even if its not free?
Differential Revision: https://reviews.llvm.org/D43063
llvm-svn: 325289
This is mainly a move of simplifyShuffleOperands from DAGCombiner::visitVECTOR_SHUFFLE to create a more general purpose TargetLowering::SimplifyDemandedVectorElts implementation.
Further features can be moved/added in future patches.
Differential Revision: https://reviews.llvm.org/D42896
llvm-svn: 325232
When creating high MachineMemOperand for MSTORE/MLOAD we supply
it with the original PointerInfo, while the pointer itself had been incremented.
The patch adds the proper offset to the PointerInfo.
llvm-svn: 325135
Making a width of GEP Index, which is used for address calculation, to be one of the pointer properties in the Data Layout.
p[address space]:size:memory_size:alignment:pref_alignment:index_size_in_bits.
The index size parameter is optional, if not specified, it is equal to the pointer size.
Till now, the InstCombiner normalized GEPs and extended the Index operand to the pointer width.
It works fine if you can convert pointer to integer for address calculation and all registered targets do this.
But some ISAs have very restricted instruction set for the pointer calculation. During discussions were desided to retrieve information for GEP index from the Data Layout.
http://lists.llvm.org/pipermail/llvm-dev/2018-January/120416.html
I added an interface to the Data Layout and I changed the InstCombiner and some other passes to take the Index width into account.
This change does not affect any in-tree target. I added tests to cover data layouts with explicitly specified index size.
Differential Revision: https://reviews.llvm.org/D42123
llvm-svn: 325102
Summary:
If the and has an additional use we shouldn't invert it. That creates an additional instruction.
While there add a one use check to the transform above that looked similar.
Reviewers: spatel, RKSimon
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D43225
llvm-svn: 325019
The bug has been lying dormant, but apparently was never exposed, until
after rL324941 because we didn't return the correct result
for shifts with undef operands.
llvm-svn: 325010
This started by noticing that scalar and vector types were producing different results with div ops in PR36305:
https://bugs.llvm.org/show_bug.cgi?id=36305
...but the problem is bigger. I couldn't keep it straight without a table, so I'm attaching that as a PDF to
the review. The x86 tests in undef-ops.ll correspond to that table.
Green means that instsimplify and the DAG agree on the result for all types.
Red means the DAG was returning undef when IR was not.
Yellow means the DAG was returning a non-undef result when IR returned undef.
This patch assumes that we're currently doing the right thing in IR.
Note: I couldn't find any problems with lowering vector constants as the code comments were warning,
but those comments were written long ago in rL36413 .
Differential Revision: https://reviews.llvm.org/D43141
llvm-svn: 324941
Armv8.1-A added an atomic load-clear instruction (which performs bitwise
and with the complement of it's operand), but not a load-and
instruction. Our current code-generation for atomic load-and always
inserts an MVN instruction to invert its argument, even if it could be
folded into a constant or another instruction.
This adds lowering early in selection DAG to convert a load-and
operation into an xor with -1 and a load-clear, allowing the normal DAG
optimisations to work on it.
To do this, I've had to add a new ISD opcode, ATOMIC_LOAD_CLR. I don't
see any easy way to do this with an AArch64-specific ISD node, because
the code-generation for atomic operations assumes the SDNodes are of
type AtomicSDNode.
I've left the old tablegen patterns in because they are still needed for
global isel.
Differential revision: https://reviews.llvm.org/D42478
llvm-svn: 324908
This reverses instcombine's demanded bits' transform which always tries to clear bits in constants.
As noted in PR35792 and shown in the test diffs:
https://bugs.llvm.org/show_bug.cgi?id=35792
...we can do better in codegen by trying to form -1. The x86 sub test shows a missed opportunity.
I did investigate changing instcombine's behavior, but it would be more work to change
canonicalization in IR. Clearing bits / shrinking constants can allow killing instructions,
so we'd have to figure out how to not regress those cases.
Differential Revision: https://reviews.llvm.org/D42986
llvm-svn: 324839
This allows us to recognise more saturation patterns and also simplify some MINMAX codegen that was failing to combine CMPGE comparisons to a legal CMPGT.
Differential Revision: https://reviews.llvm.org/D43014
llvm-svn: 324837
SelectionDAG::getBoolConstant was recently introduced. At the time I didn't know getConstTrueVal existed, but I think getBoolConstant is better as it will use the source VT to make sure it can properly detect floating point if it is configured differently.
llvm-svn: 324832
When adding operands to machine instructions in case of
RegisterSDNodes, generate a COPY node in case the register class
does not match the one in the instruction definition.
Differental Revision: https://reviews.llvm.org/D35561
llvm-svn: 324733
Many in SimplifySetCC and FoldSetCC try to create true or false constants. Some of them query getBooleanContents to figure out whether to use all ones or just 1 for true. But many places do not check and just use 1 without ensuring the VT has an i1 scalar type. Note sure if those places only trigger before type legalization so they only see an i1
type?
To cleanup the inconsistency and reduce some duplicated code, this patch adds a getBoolConstant method to SelectionDAG that takes are of querying getBooleanContents and doing the right thing.
Differential Revision: https://reviews.llvm.org/D43037
llvm-svn: 324634
We're passing the binary op that uses the load instead of the load.
Noticed by inspection. Not sure how to test this because this just prevents the introduction of an extend that will later be truncated and will probably be combined out.
llvm-svn: 324568
The truncate is being used to replace other users of of the load, but we checked that the load only has one use so there are no other uses to replace.
llvm-svn: 324567
The truncate is only needed if the load has additional users. It used to get passed to extendSetCCUses so was created early, but that's no longer the case.
llvm-svn: 324562
Travel all chains paths to first non-tokenfactor node can be
exponential work. Add simple redundency check to avoid this.
Fixes PR36264.
llvm-svn: 324491
X86 currently has a late DAG combine after cttz/ctlz are turned into BSR+BSF+CMOV to detect this and remove the CMOV. But we should be able to do this much earlier and avoid creating the cmov all together.
For the changed AMDGPU test case it appears that previously the i8 cttz was type legalized to i16 which introduced an OR with 256 in order to limit the result to 8 on the widened type. At this point the result is known to never be zero, but nothing checked that. Then operation legalization is told to promote all i16 cttz to i32. This introduces an extend and a truncate and another OR with 65536 to limit the result to 16. With the DAG combiner change we are able to prevent the creation of the second OR since the opcode will have been changed to cttz_zero_undef after the first OR. I the lack of the OR caused the instruction to change to v_ffbl_b32_sdwa
Differential Revision: https://reviews.llvm.org/D42985
llvm-svn: 324427
Instruction Selection
Cleanup cycle/validity checks in ISel (IsLegalToFold,
HandleMergeInputChains) and X86 (isFusableLoadOpStore). Now do a full
search for cycles / dependencies pruning the search when topological
property of NodeId allows.
As part of this propogate the NodeId-based cutoffs to narrow
hasPreprocessorHelper searches.
Reviewers: craig.topper, bogner
Subscribers: llvm-commits, hiraditya
Differential Revision: https://reviews.llvm.org/D41293
llvm-svn: 324359
Summary:
This method is trying to use the truncate node to find which SETCC operand should be replaced directly with the extended load.
This used to work correctly because all uses of the original load were replaced by the truncate before this function was called. So this was used to effectively bypass the truncate and find the load under it.
All but one of the callers now call this before the truncate has replaced the laod so the setcc doesn't yet use the truncate. To account for this we should pass the original load instead.
I changed the order of that one caller to make this work there too.
I don't have a test case because this is probably hidden by later DAG combines causing the extend and truncate to cancel out. I assume this way is a little more efficient and matches what was originally intended.
Reviewers: RKSimon, spatel, niravd
Reviewed By: niravd
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D42878
llvm-svn: 324311
Summary:
If the load is already an extended load we should be using the memory VT for the legality check, not just the VT of the current extension.
I don't have a test case, just noticed it while investigating some load extension improvements.
Reviewers: RKSimon, spatel, niravd
Reviewed By: niravd
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D42783
llvm-svn: 324181
Better to assume that any value type may be commuted, not just MVTs.
No test case right now, but discovered while investigating possible shuffle combines.
llvm-svn: 324179
When handling vectors with non byte-sized elements, reverse the order of the
elements in the built integer if the target is Big-Endian.
SystemZ tests updated.
Review: Eli Friedman, Ulrich Weigand.
https://reviews.llvm.org/D42786
llvm-svn: 324063
When getNode() is called to create an EXTRACT_VECTOR_ELT, assert that
the result VT is at least as wide as the vector element type.
Review: Eli Friedman
llvm-svn: 324061
We were only checking the element count, but not the total width. This could cause illegal bitcasts to be created if for example the output was 512-bits, but N1 is 256 bits, and the extraction size was 128-bits.
Fixes PR36199
Differential Revision: https://reviews.llvm.org/D42809
llvm-svn: 324002
As shown in the example in PR34994:
https://bugs.llvm.org/show_bug.cgi?id=34994
...we can return a very wrong answer (inf instead of 0.0) for square root when
using a reciprocal square root estimate instruction.
Here, I've conditionalized the filtering out of denorms based on the function
having "denormal-fp-math"="ieee" in its attributes. The other options for this
attribute are 'preserve-sign' and 'positive-zero'.
So we don't generate this extra code by default with just '-ffast-math' (because
then there's no denormal attribute string at all), but it works if you specify
'-ffast-math -fdenormal-fp-math=ieee' from clang.
As noted in the review, there may be other problems in clang that affect the
results depending on platform (Linux x86 at least), but this should allow
creating the desired codegen.
Differential Revision: https://reviews.llvm.org/D42323
llvm-svn: 323981
Summary:
In Instruction Selection UpdateChains replaces all matched Nodes'
chain references including interior token factors and deletes them.
This may allow nodes which depend on these interior nodes but are not
part of the set of matched nodes to be left with a dangling dependence.
Avoid this by doing the replacement for matched non-TokenFactor nodes.
Fixes PR36164.
Reviewers: jonpa, RKSimon, bogner
Subscribers: llvm-commits, hiraditya
Differential Revision: https://reviews.llvm.org/D42754
llvm-svn: 323977
Summary:
This change expands the amount of registers stashed by the entry and
`__xray_CustomEvent` trampolines.
We've found that since the `__xray_CustomEvent` trampoline calls can show up in
situations where the scratch registers are being used, and since we don't
typically want to affect the code-gen around the disabled
`__xray_customevent(...)` intrinsic calls, that we need to save and restore the
state of even the scratch registers in the handling of these custom events.
Reviewers: pcc, pelikan, dblaikie, eizan, kpw, echristo, chandlerc
Reviewed By: echristo
Subscribers: chandlerc, echristo, hiraditya, davide, dblaikie, llvm-commits
Differential Revision: https://reviews.llvm.org/D40894
llvm-svn: 323940
Summary:
Instruction Selection preserves relative orders of all nodes save
TokenFactors which we treat specially. As a result Node Ids for
TokenFactors may violate the topological ordering and should not be
considered as valid pruning candidates in predecessor search.
Fixes PR35316.
Reviewers: RKSimon, hfinkel
Subscribers: hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D42701
llvm-svn: 323880
In Thumb 1, with the new ADDCARRY / SUBCARRY the scheduler may need to do
copies CPSR ↔ GPR but not all Thumb1 targets implement them.
The schedule can attempt, before attempting a copy, to clone the instructions
but it does not currently do that for nodes with input glue. In this patch we
introduce a target-hook to let the hook decide if a glued machinenode is still
eligible for copying. In this case these are ARM::tADCS and ARM::tSBCS .
As a follow-up of this change we should actually implement the copies for the
Thumb1 targets that do implement them and restrict the hook to the targets that
can't really do such copy as these clones are not ideal.
This change fixes PR35836.
Differential Revision: https://reviews.llvm.org/D42051
llvm-svn: 323857
PR36061 showed that during the expansion of ISD::FPOWI, that there
was an incorrect zero extension of the integer argument which for
MIPS64 would then give incorrect results. Address this with the
existing mechanism for correcting sign extensions.
This resolves PR36061.
Thanks to James Cowgill for reporting the issue!
Reviewers: atanasyan, hfinkel
Differential Revision: https://reviews.llvm.org/D42537
llvm-svn: 323781
When a function return value can't be directly lowered, such as
returning an i128 on WebAssembly, as indicated by the CanLowerReturn
target hook, SelectionDAGBuilder can translate it to return the
value through a hidden sret-like argument.
If such a function has an argument with the "returned" attribute,
the attribute can't be automatically lowered, because the function
no longer has a normal return value. For now, just discard the
"returned" attribute.
This fixes PR36128.
llvm-svn: 323715
Summary:
There's a check in the code to only check getSetCCResultType after LegalOperations or if the type is MVT::i1. But the i1 check is only allowing scalar types through. I think it should check that the scalar type is MVT::i1 so that it will work for vectors.
The changed test already does this combine with AVX512VL where getSetCCResultType returns vXi1. But with avx512f and no VLX getSetCCResultType returns a type matching the width of the input type.
Reviewers: spatel, RKSimon
Reviewed By: spatel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D42619
llvm-svn: 323631
The code was using getValueSizeInBits and combining with the result of a call to DAG.ComputeNumSignBits. But for vector types getValueSizeInBits returns the width of the full vector while ComputeNumSignBits is going to give a number no larger than the width of a single element. So we should be using getScalarValueSizeInBits to get the element width.
llvm-svn: 323583
Previously some targets printed their own message at the start of Select to indicate what they were selecting. For the targets that didn't, it means there was no print of the root node before any custom handling in the target executed. So if the target did something custom and never called SelectNodeCommon, no print would be made. For the targets that did print a message in Select, if they didn't custom handle a node SelectNodeCommon would reprint the root node before walking the isel table.
It seems better to just print the message before the call to Select so all targets behave the same. And then remove the root node printing from SelectNodeCommon and just leave a message that says we're starting the table search.
There were also some oddities in blank line behavior. Usually due to a \n after a call to SelectionDAGNode::dump which already inserted a new line.
llvm-svn: 323551
Apparently checking the pass structure isn't enough to ensure that we don't fall
back to FastISel, as it's set up as part of the SelectionDAGISel.
llvm-svn: 323369
For the included test case, the DAG transformation
concat_vectors(scalar, undef) -> scalar_to_vector(sclr)
would attempt to create a v2i32 vector for a v9i8
concat_vector. Bail out to avoid creating a bitcast with
mismatching sizes later on.
Differential Revision: https://reviews.llvm.org/D42379
llvm-svn: 323312
This was completely broken, but hopefully fixed by this patch.
In cases where it is needed, a vector with non byte-sized elements is stored
by extracting, zero-extending, shift:ing and or:ing the elements into an
integer of the same width as the vector, which is then stored.
Review: Eli Friedman, Ulrich Weigand
https://reviews.llvm.org/D42100#inline-369520https://bugs.llvm.org/show_bug.cgi?id=35520
llvm-svn: 323042
The second return value of ATOMIC_CMP_SWAP_WITH_SUCCESS is known to be a
boolean, and should therefore be treated by computeKnownBits just like
the second return values of SMULO / UMULO.
Differential Revision: https://reviews.llvm.org/D42067
llvm-svn: 322985
Follow-up to r322120 which can cause assertions for AArch64 because
v1f64 and v1i64 are legal types.
Differential Revision: https://reviews.llvm.org/D42097
llvm-svn: 322823
For example, a build_vector of i64 bitcasted from v2i32 can be turned into a concat_vectors of the v2i32 vectors with a bitcast to a vXi64 type
Differential Revision: https://reviews.llvm.org/D42090
llvm-svn: 322811
The code wasn't zero-extending correctly, so the comparison could
spuriously fail.
Adds some AArch64 tests to cover this case.
Inspired by D41791.
Differential Revision: https://reviews.llvm.org/D41798
llvm-svn: 322767
Summary:
In preparation for https://reviews.llvm.org/D41675 this NFC changes this
prototype of MemIntrinsicInst::setAlignment() to accept an unsigned instead
of a Constant.
llvm-svn: 322403
Summary:
Fold cases such as:
(v8i8 truncate (v8i32 extract_subvector (v16i32 sext (v16i8 V), Idx)))
->
(v8i8 extract_subvector (v16i8 V), Idx)
This can be generalized to cases where the truncate and extend do not
fully cancel each other out, but it may require querying the target
about profitability.
Reviewers: RKSimon, craig.topper, spatel, efriedma
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D41927
llvm-svn: 322300
The code that is supposed to "Round odd types to the next pow of two" seems
broken and as well completely unused (untested). It also seems that
ExpandStore really shouldn't ever change the memory VT, which this in fact
does.
As a first step in fixing the broken handling of vector stores (of irregular
types, e.g. an i1 vector), this code is removed. For discussion, see
https://bugs.llvm.org/show_bug.cgi?id=35520.
Review: Eli Friedman
llvm-svn: 322275
Currently we infer the scale at isel time by analyzing whether the base is a constant 0 or not. If it is we assume scale is 1, else we take it from the element size of the pass thru or stored value. This seems a little weird and I think it makes more sense to make it explicit in the DAG rather than doing tricky things in the backend.
Most of this patch is just making sure we copy the scale around everywhere.
Differential Revision: https://reviews.llvm.org/D40055
llvm-svn: 322210
Prefetches used to always be chained between any previous and following
memory accesses. The problem with this was that later optimizations, such as
folding of a load into the user instruction, got disrupted.
This patch relaxes the chaining of prefetches in order to remedy this.
Reveiw: Hal Finkel
https://reviews.llvm.org/D38886
llvm-svn: 322163
Summary:
In the case of an fp_extend of v1f16 to v1f32 where the v1f16 is the
result of a bitcast from i16, avoid creating an illegal fp16_to_fp where
the input is not a vector and the result is a v1f32.
V2: The fix is now to avoid vector scalarization creating a v1->scalar
bitcast.
Reviewers: srhines, t.p.northover
Subscribers: nhaehnle, llvm-commits, dstuttard, t-tye, yaxunl, wdng, kzhuravl, arsenm
Differential Revision: https://reviews.llvm.org/D41126
llvm-svn: 322120
Ingredients in this patch:
1. Add HANDLE_LIBCALL defs for finite mathlib functions that correspond to LLVM intrinsics.
2. Plumbing to send TargetLibraryInfo down to SelectionDAGLegalize.
3. Relaxed math and library checking in SelectionDAGLegalize::ConvertNodeToLibcall() to choose finite libcalls.
There was a bug about determining the availability of the finite calls that should be fixed with:
rL322010
Not in this patch:
This doesn't resolve the question/bug of clang creating the intrinsic IR in the first place.
There's likely follow-up work needed to support the long double variants better.
There's room for improvement to reduce the code duplication.
Create finite calls that don't originate from a corresponding intrinsic or DAG node?
Differential Revision: https://reviews.llvm.org/D41338
llvm-svn: 322087
Allow SimplifyDemandedBits to use TargetLoweringOpt::computeKnownBits to look through bitcasts. This can help simplifying in some cases where bitcasts of constants generated during or after legalization can't be folded away, and thus didn't get picked up by SimplifyDemandedBits. This fixes PR34620, where a redundant pand created during legalization from lowering and lshr <16xi8> wasn't being simplified due to the presence of a bitcasted build_vector as an operand.
Committed on the behalf of @sameconrad (Sam Conrad)
Differential Revision: https://reviews.llvm.org/D41643
llvm-svn: 321969
Summary:
There are few oddities that occur due to v1i1, v8i1, v16i1 being legal without v2i1 and v4i1 being legal when we don't have VLX. Particularly during legalization of v2i32/v4i32/v2i64/v4i64 masked gather/scatter/load/store. We end up promoting the mask argument to these during type legalization and then have to widen the promoted type to v8iX/v16iX and truncate it to get the element size back down to v8i1/v16i1 to use a 512-bit operation. Since need to fill the upper bits of the mask we have to fill with 0s at the promoted type.
It would be better if we could just have the v2i1/v4i1 types as legal so they don't undergo any promotion. Then we can just widen with 0s directly in a k register. There are no real v4i1/v2i1 instructions anyway. Everything is done on a larger register anyway.
This also fixes an issue that we couldn't implement a masked vextractf32x4 from zmm to xmm properly.
We now have to support widening more compares to 512-bit to get a mask result out so new tablegen patterns got added.
I had to hack the legalizer for widening the operand of a setcc a bit so it didn't try create a setcc returning v4i32, extract from it, then try to promote it using a sign extend to v2i1. Now we create the setcc with v4i1 if the original setcc's result type is v2i1. Then extract that and don't sign extend it at all.
There's definitely room for improvement with some follow up patches.
Reviewers: RKSimon, zvi, guyblank
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D41560
llvm-svn: 321967
Summary:
I believe legalization is really expecting that ReplaceNodeResults will return something with the same type as the thing that's being legalized. Ultimately, it uses the output to replace the uses in the DAG so the type should match to make that work.
There are two relevant cases here. When crbits are enabled, then i1 is a legal type and getSetCCResultType should return i1. In this case, the truncate will be between i1 and i1 and should be removed (SelectionDAG::getNode does this). Otherwise, getSetCCResultType will be i32 and the legalizer will promote the truncate to be i32 -> i32 which will be similarly removed.
With this fixed we can remove some code from PromoteIntRes_SETCC that seemed to only exist to deal with the intrinsic being replaced with a larger type without changing the other operand. With the truncate being used for connectivity this doesn't happen anymore.
Reviewers: hfinkel
Reviewed By: hfinkel
Subscribers: nemanjai, llvm-commits, kbarton
Differential Revision: https://reviews.llvm.org/D41654
llvm-svn: 321959
Handle this in DAGCombiner::visitEXTRACT_VECTOR_ELT the same as we already do in SelectionDAG::getNode and use APInt instead of getZExtValue.
This should also fix oss-fuzz #4910
llvm-svn: 321767
Our internal testing has revealed has discovered bugs in PPC builds.
I have forward reproduction instructions to the original author (Nirav).
llvm-svn: 321649
Currently the promotion for these ignores the normal getTypeToPromoteTo and instead just tries to double the element width. This is because the default behavior of getTypeToPromote to just adds 1 to the SimpleVT, which has the affect of increasing the element count while keeping the scalar size the same.
If multiple steps are required to get to a legal operation type, int_to_fp will be promoted multiple times. And fp_to_int will keep trying wider types in a loop until it finds one that works.
getTypeToPromoteTo does have the ability to query a promotion map to get the type and not do the increasing behavior. It seems better to just let the target specify the promotion type in the map explicitly instead of letting the legalizer iterate via widening.
FWIW, it's worth I think for any other vector operations that need to be promoted, we have to specify the type explicitly because the default behavior of getTypeToPromote isn't useful for vectors. The other types of promotion already require either the element count is constant or the total vector width is constant, but neither happens by incrementing the SimpleVT enum.
Differential Revision: https://reviews.llvm.org/D40664
llvm-svn: 321629
Summary:
I have been getting rather difficult to reproduce SIGBUS crashes when
compiling certain FreeBSD sources, and their stack traces pointed
squarely at `SelectionDAG::salvageDebugInfo()`:
```
Core was generated by `/usr/obj/share/dim/src/freebsd/clang600-import/amd64.amd64/tmp/usr/bin/cc -cc1 -'.
Program terminated with signal SIGBUS, Bus error.
#0 isInvalidated () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/CodeGen/SelectionDAG/SDNodeDbgValue.h:115
115 bool isInvalidated() const { return Invalid; }
(gdb) bt
#0 isInvalidated () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/CodeGen/SelectionDAG/SDNodeDbgValue.h:115
#1 salvageDebugInfo () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp:7116
#2 0x00000000033b2516 in operator() () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp:3595
#3 __invoke<(lambda at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp:3593:59) &, llvm::SDNode *, llvm::SDNode *> () at /usr/include/c++/v1/type_traits:4323
#4 __call<(lambda at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp:3593:59) &, llvm::SDNode *, llvm::SDNode *> () at /usr/include/c++/v1/__functional_base:349
#5 operator() () at /usr/include/c++/v1/functional:1562
#6 0x00000000033b0817 in operator() () at /usr/include/c++/v1/functional:1916
#7 NodeDeleted () at /share/dim/src/freebsd/clang600-import/contrib/llvm/include/llvm/CodeGen/SelectionDAG.h:293
#8 0x0000000003529dde in RemoveDeadNodes () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp:610
#9 0x00000000035556df in MorphNodeTo () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp:6794
#10 0x00000000033a9acc in MorphNode () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp:2594
#11 0x00000000033ac80b in SelectCodeCommon () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp:3601
#12 0x00000000023d464b in SelectCode () at /usr/obj/share/dim/src/freebsd/clang600-import/amd64.amd64/tmp/obj-tools/lib/clang/libllvm/X86GenDAGISel.inc:282902
#13 Select () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp:3072
#14 0x00000000033a5afa in DoInstructionSelection () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp:988
#15 0x00000000033a4e1a in CodeGenAndEmitDAG () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp:868
#16 0x00000000033a2643 in SelectAllBasicBlocks () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp:1624
#17 0x000000000339f158 in runOnMachineFunction () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp:466
#18 0x00000000023d03c4 in runOnMachineFunction () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp:175
#19 0x00000000035cc8c2 in runOnFunction () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/CodeGen/MachineFunctionPass.cpp:62
#20 0x00000000030dca9a in runOnFunction () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/IR/LegacyPassManager.cpp:1520
#21 0x00000000030dccf3 in runOnModule () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/IR/LegacyPassManager.cpp:1541
#22 0x00000000030dd228 in runOnModule () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/IR/LegacyPassManager.cpp:1597
#23 run () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/IR/LegacyPassManager.cpp:1700
#24 0x00000000014db578 in EmitAssembly () at /share/dim/src/freebsd/clang600-import/contrib/llvm/tools/clang/lib/CodeGen/BackendUtil.cpp:815
#25 EmitBackendOutput () at /share/dim/src/freebsd/clang600-import/contrib/llvm/tools/clang/lib/CodeGen/BackendUtil.cpp:1181
#26 0x00000000014d5b26 in HandleTranslationUnit () at /share/dim/src/freebsd/clang600-import/contrib/llvm/tools/clang/lib/CodeGen/CodeGenAction.cpp:292
#27 0x0000000001c4c332 in ParseAST () at /share/dim/src/freebsd/clang600-import/contrib/llvm/tools/clang/lib/Parse/ParseAST.cpp:159
#28 0x00000000015d546c in Execute () at /share/dim/src/freebsd/clang600-import/contrib/llvm/tools/clang/lib/Frontend/FrontendAction.cpp:897
#29 0x0000000001cec311 in ExecuteAction () at /share/dim/src/freebsd/clang600-import/contrib/llvm/tools/clang/lib/Frontend/CompilerInstance.cpp:991
#30 0x00000000014b4f81 in ExecuteCompilerInvocation () at /share/dim/src/freebsd/clang600-import/contrib/llvm/tools/clang/lib/FrontendTool/ExecuteCompilerInvocation.cpp:252
#31 0x00000000014aa73f in cc1_main () at /share/dim/src/freebsd/clang600-import/contrib/llvm/tools/clang/tools/driver/cc1_main.cpp:221
#32 0x00000000014b2928 in ExecuteCC1Tool () at /share/dim/src/freebsd/clang600-import/contrib/llvm/tools/clang/tools/driver/driver.cpp:309
#33 main () at /share/dim/src/freebsd/clang600-import/contrib/llvm/tools/clang/tools/driver/driver.cpp:388
(gdb) frame 1
#1 salvageDebugInfo () at /share/dim/src/freebsd/clang600-import/contrib/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp:7116
7116 if (DV->isInvalidated())
(gdb) disassemble
Dump of assembler code for function salvageDebugInfo():
[...]
0x0000000003557348 <+744>: nopl 0x0(%rax,%rax,1)
0x0000000003557350 <+752>: mov (%r12),%r13
=> 0x0000000003557354 <+756>: cmpb $0x0,0x31(%r13)
0x0000000003557359 <+761>: jne 0x35573b0 <salvageDebugInfo()+848>
(gdb) info registers
[...]
r13 0x5a5a5a5a5a5a5a5a 6510615555426900570
```
The `0x5a5a5a5a5a5a5a5a` value in `r13` indicates the memory was either
uninitialized, or already freed.
Unfortunately I do not have a simple self-contained test case for this.
However, it seems pretty clear that the call to `AddDbgValue()` in
`salvageDebugInfo()` causes the problems, since it modifies
`SelectionDag::DbgInfo` while looping through one of its DenseMaps:
```
void SelectionDAG::salvageDebugInfo(SDNode &N) {
[...]
for (auto DV : GetDbgValues(&N)) {
if (DV->isInvalidated())
continue;
[...]
AddDbgValue(Clone, N0.getNode(), false);
[...]
}
}
```
At least, if I comment out the `AddDbgValue()` call, the crashes go
away. I propose to change this function slightly, similar to the
`SelectionDAG::transferDbgValues()` function just above it, to save the
cloned SDDbgValues in a separate SmallVector, and only call
AddDbgValue() on them after the for loop is done.
Reviewers: aprantl, bogner, bkramer, davide
Reviewed By: davide
Subscribers: davide, krytarowski, JDevlieghere, emaste, llvm-commits
Differential Revision: https://reviews.llvm.org/D41589
llvm-svn: 321545
For example, float operations may fail to constant fold under certain circumstances (inf/nan/denormal creation etc.)
Reduced from oss-fuzz #4802 test case
llvm-svn: 321488
This moves the combine for turning ANDs into shuffle with zero out of SimplifyVBinOps and places it only in visitAND below the reassociate handling. This fixes the specific case I noticed where we failed to combine two ands with constants.
llvm-svn: 321417
getOperand returns an SDValue that contains the node and the result number. There is no guarantee that the result number if 0. By using the -> operator we are calling SDNode::getValueType rather than SDValue::getValueType. This requires supplying a result number and we shouldn't assume it was 0.
I don't have a test case. Just noticed while cleaning up some other code and saw that it occurred in other places.
llvm-svn: 321397
BaseIndexOffset supercedes findBaseOffset analysis save only Constant
Pool addresses. Migrate analysis to BaseIndexOffset.
Relanding after correcting base address matching check.
llvm-svn: 321389
This seems to improve X86's ability to match this into an address computation. Otherwise the other operand gets assigned to the base register and the stack pointer + frame index ends up in the index register. But index registers can't encode ESP/RSP so we end up having to move it into another register to meet the constraint.
I could try to improve the address matcher in X86, but swapping the producer seemed easier. Several other places already have the operands in this order so this is at least consistent.
llvm-svn: 321370
The knownbits_mask_or_shuffle_uitofp change is interesting - shuffle combines manage to kick in, removing the AND constant mask load. For targets with fast-variable-shuffle this should reduce further to VPOR+VPSHUFB+VCVTDQ2PS.
llvm-svn: 321279
If the SRL node is only used by an AND, we may be able to set the
ExtVT to the width of the mask, making the AND redundant. To support
this, another check has been added in isLegalNarrowLoad which queries
whether the load is valid.
Differential Revision: https://reviews.llvm.org/D41350
llvm-svn: 321259
When intrinsics are allowed to have mem operands, there
are two ways this can happen. First is an intrinsic
that is marked has having a mem operand, but is not handled
by getTgtMemIntrinsic.
The second way can occur even for intrinsics which do not
have a mem operand. It seems the selector table does
some kind of sorting based on the opcode, and the
mem ref recording can happen in the same scope for
intrinsics that both do and do not have mem refs.
I haven't been able to figure out exactly why this happens
(although it happens even with the matcher optimizations disabled).
I'm not sure if it's worth trying to avoid hitting this for
these nodes since I think it's still reasonable to handle
this in case getTgtMemIntrinic is not implemented.
llvm-svn: 321208
These functions simply call their counterparts in the associated SDNode,
which do take an optional SelectionDAG. This change makes the legalization
debug trace a little easier to read, since target-specific nodes will
now have their names shown instead of "Unknown node #123".
llvm-svn: 321180
Summary:
Extend overlapping store elision to handle overwrites of stores by
larger stores.
Nontemporal tests have been modified to add memory dependencies to
prevent store elision.
Reviewers: craig.topper, rnk, t.p.northover
Subscribers: javed.absar, hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D40969
llvm-svn: 321089
Search from AND nodes to find whether they can be propagated back to
loads, so that the AND and load can be combined into a narrow load.
We search through OR, XOR and other AND nodes and all bar one of the
leaves are required to be loads or constants. The exception node then
needs to be masked off meaning that the 'and' isn't removed, but the
loads(s) are narrowed still.
Differential Revision: https://reviews.llvm.org/D41177
llvm-svn: 320962
Summary:
Currently we don't handle v32i1/v64i1 insert_vector_elt correctly as we fail to look at the number of elements closely and assume it can only be v16i1 or v8i1.
We also can't type legalize v64i1 insert_vector_elt correctly on KNL due to the type not being byte addressable as required by the legalizing through memory accesses path requires.
For the first issue, the patch now tries to pick a 512-bit register with the correct number of elements and promotes to that.
For the second issue, we now extend the vector to a byte addressable type, do the stores to memory, load the two halves, and then truncate the halves back to the original type. Technically since we changed the type, we may not need two loads, but actually checking that is more work and for the v64i1 case we do need them.
Reviewers: RKSimon, delena, spatel, zvi
Reviewed By: RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D40942
llvm-svn: 320849
This makes it work better with some build_vector and concat_vectors creations.
Adjust the NewSDValueDbgMsg in getConstant to avoid duplicating the print when it calls getSplatBuildVector since getSplatBuildVector didn't trigger a print before.
llvm-svn: 320783
While investigating LLVM PR22316 (http://llvm.org/bugs/show_bug.cgi?id=22316)
I started wondering if it were not always preferable to emit the
initial DBG_VALUEs for stack arguments as FI locations instead of
describing the first register they get copied into. The advantage of
doing this is that the arguments will be available as soon as the
stack is setup. As illustrated by the testcase in the PR, the first
copy of the FI into a register may be sunk by MachineSink.cpp into a
later basic block. By describing the argument on the stack, we nicely
circumvent this problem.
<rdar://problem/19583723>
Differential Revision: https://reviews.llvm.org/D41135
llvm-svn: 320758
Most of the -Wsign-compare warnings are due to the fact that
enums are signed by default in the MS ABI, while the
tautological comparison warnings trigger on x86 builds where
sizeof(size_t) is 4 bytes, so N > numeric_limits<unsigned>::max()
is always false.
Differential Revision: https://reviews.llvm.org/D41256
llvm-svn: 320750
Rather than adding more bits to express every
MMO flag you could want, just directly use the
MMO flags. Also fixes using a bunch of bool arguments to
getMemIntrinsicNode.
On AMDGPU, buffer and image intrinsics should always
have MODereferencable set, but currently there is no
way to do that directly during the initial intrinsic
lowering.
llvm-svn: 320746
Recommitting rL319773, which was reverted due to a recursive issue
causing timeouts. This happened because I failed to check whether
the discovered loads could be narrowed further. In the case of a tree
with one or more narrow loads, that could not be further narrowed, as
well as a node that would need masking, an AND could be introduced
which could then be visited and recombined again with the same load.
This could again create the masking load, with would be combined
again... We now check that the load can be narrowed so that this
process stops.
Original commit message:
Search from AND nodes to find whether they can be propagated back to
loads, so that the AND and load can be combined into a narrow load.
We search through OR, XOR and other AND nodes and all bar one of the
leaves are required to be loads or constants. The exception node then
needs to be masked off meaning that the 'and' isn't removed, but the
loads(s) are narrowed still.
Differential Revision: https://reviews.llvm.org/D41177
llvm-svn: 320679
A v32i1 CONCAT_VECTORS of v16i1 uses promotion to v32i8 to legalize the v32i1. This results in a bunch of extract_vector_elts and a build_vector that ultimately gets scalarized.
This patch checks to see if v16i8 is legal and inserts a any_extend to that so that we can concat v16i8 to v32i8 and avoid creating the extracts.
llvm-svn: 320674
If so go ahead and get the promoted input vector to extract from. Previously, we would create a bunch of any_extends of extract_vector_elts with illegal input type that needs to be promoted. The legalization of those extract_vector_elts would then potentially introduce a truncate. So now we have a bunch of any_extends of truncates. By legalizing both parts together we avoid creating these extra nodes.
The test changes seem to be because we were previously combining the build_vector with the any_extend before the any_extend got combined with the truncate.
llvm-svn: 320669
At first, I tried to thread the x86 needle and use a target hook (isVectorShiftByScalarCheap())
to disable the transform only for non-splat pow-of-2 constants, but not AVX2, but only some
element types, but...it's difficult.
Here we just avoid the loop with the x86 vector transform that conflicts with the general DAG
combine and preserve all of the existing behavior AFAICT otherwise.
Some tests that will probably fail if someone does try to restrict this in a more targeted way
for x86-only may be found in:
test/CodeGen/X86/combine-mul.ll
test/CodeGen/X86/vector-mul.ll
test/CodeGen/X86/widen_arith-5.ll
This should prevent the infinite looping seen with:
https://bugs.llvm.org/show_bug.cgi?id=35579
Differential Revision: https://reviews.llvm.org/D41040
llvm-svn: 320374
This commit is the first part of https://reviews.llvm.org/D40348.
In order to allow target combines to be performed on newly combined
indexed loads, add them back to the worklist. The remainder of the
above patch will be committed in subsequent revisions and will use
this. Test cases will be included with those follow-up commits.
llvm-svn: 320365
This is a preparatory step for D34515.
This change:
- makes nodes ISD::ADDCARRY and ISD::SUBCARRY legal for i32
- lowering is done by first converting the boolean value into the carry flag
using (_, C) ← (ARMISD::ADDC R, -1) and converted back to an integer value
using (R, _) ← (ARMISD::ADDE 0, 0, C). An ARMISD::ADDE between the two
operations does the actual addition.
- for subtraction, given that ISD::SUBCARRY second result is actually a
borrow, we need to invert the value of the second operand and result before
and after using ARMISD::SUBE. We need to invert the carry result of
ARMISD::SUBE to preserve the semantics.
- given that the generic combiner may lower ISD::ADDCARRY and
ISD::SUBCARRYinto ISD::UADDO and ISD::USUBO we need to update their lowering
as well otherwise i64 operations now would require branches. This implies
updating the corresponding test for unsigned.
- add new combiner to remove the redundant conversions from/to carry flags
to/from boolean values (ARMISD::ADDC (ARMISD::ADDE 0, 0, C), -1) → C
- fixes PR34045
- fixes PR34564
- fixes PR35103
Differential Revision: https://reviews.llvm.org/D35192
llvm-svn: 320355
Introduces the AddrFI "addressing mode", which is necessary simply because
it's not possible to write a pattern that directly matches a frameindex.
Ensure callee-saved registers are accessed relative to the stackpointer. This
is necessary as callee-saved register spills are performed before the frame
pointer is set.
Move HexagonDAGToDAGISel::isOrEquivalentToAdd to SelectionDAGISel, so we can
make use of it in the RISC-V backend.
Differential Revision: https://reviews.llvm.org/D39848
llvm-svn: 320353
Summary:
This relaxes an assertion inside SelectionDAGBuilder which is overly
restrictive on targets which have no concept of alignment (such as AVR).
In these architectures, all types are aligned to 8-bits.
After this, LLVM will only assert that accesses are aligned on targets
which actually require alignment.
This patch follows from a discussion on llvm-dev a few months ago
http://llvm.1065342.n5.nabble.com/llvm-dev-Unaligned-atomic-load-store-td112815.html
Reviewers: bogner, nemanjai, joerg, efriedma
Reviewed By: efriedma
Subscribers: efriedma, cactus, llvm-commits
Differential Revision: https://reviews.llvm.org/D39946
llvm-svn: 320243
I noticed this pattern in D38316 / D38388. We failed to combine a shuffle that is either
repeating a scalar insertion at the same position in a vector or translated to a different
element index.
Like the earlier patch, this could be an instcombine too, but since we opted to make this
a DAG transform earlier, I've made this one a DAG patch too.
We do not need any legality checking because the new insert is identical to the existing
insert except that it may have a different constant insertion operand.
The constant insertion test in test/CodeGen/X86/vector-shuffle-combining.ll was the
motivation for D38756.
Differential Revision: https://reviews.llvm.org/D40209
llvm-svn: 320050
If the mask needs to be promoted that should occur by the legalizer detecting the mask operand needs to be promoted not as a side effect of another action.
llvm-svn: 319852
The mask will be promoted if necessary when operands are promoted. It's possible the mask type is legal, but the setcc result type is a different. We shouldn't promote to the setcc result type unless the mask needs to be promoted.
llvm-svn: 319850
The patch originally broke Chromium (crbug.com/791714) due to its failing to
specify that the new pseudo instructions clobber EFLAGS. This commit fixes
that.
> Summary: This strengthens the guard and matches MSVC.
>
> Reviewers: hans, etienneb
>
> Subscribers: hiraditya, JDevlieghere, vlad.tsyrklevich, llvm-commits
>
> Differential Revision: https://reviews.llvm.org/D40622
llvm-svn: 319824
There's no such thing as a setcc with vector operands and scalar result. And if we're trying to widen the result we would have to already be looking at a vector result type.
So this patch renames the VSETCC function as the SETCC function and delete the original SETCC function.
llvm-svn: 319799
Search from AND nodes to find whether they can be propagated back to
loads, so that the AND and load can be combined into a narrow load.
We search through OR, XOR and other AND nodes and all bar one of the
leaves are required to be loads or constants. The exception node then
needs to be masked off meaning that the 'and' isn't removed, but the
loads(s) are narrowed still.
Differential Revision: https://reviews.llvm.org/D39604
llvm-svn: 319773
Summary:
Found out, at code inspection, that there was a fault in
DAGCombiner::CombineConsecutiveLoads for big-endian targets.
A BUILD_PAIR is always having the least significant bits of
the composite value in element 0. So when we are doing the checks
for consecutive loads, for big endian targets, we should check
if the load to elt 1 is at the lower address and the load
to elt 0 is at the higher address.
Normally this bug only resulted in missed oppurtunities for
doing the load combine. I guess that in some rare situation it
could lead to faulty combines, but I've not seen that happen.
Note that this patch actually will trigger load combine for
some big endian regression tests.
One example is test/CodeGen/PowerPC/anon_aggr.ll where we now get
t76: i64,ch = load<LD8[FixedStack-9]
instead of
t37: i32,ch = load<LD4[FixedStack-10]>
t35: i32,ch = load<LD4[FixedStack-9]>
t41: i64 = build_pair t37, t35
before legalization. Then the legalization will split the LD8
into two loads, so the end result is the same. That should
verify that the transfomation is correct now.
Reviewers: niravd, hfinkel
Reviewed By: niravd
Subscribers: nemanjai, llvm-commits
Differential Revision: https://reviews.llvm.org/D40444
llvm-svn: 319771
Pull the checks upon the load out from ReduceLoadWidth into their own
function.
Differential Revision: https://reviews.llvm.org/D40833
llvm-svn: 319766
The CONCAT_VECTORS operand get its type from getSetCCResultType, but if the mask type and the setcc have different scalar sizes this creates an illegal CONCAT_VECTORS operation. The concat type should be 2x the mask type, and then an extend should be added if needed.
llvm-svn: 319744
MatchRotate assumes the types of the types of LHS and RHS are equal,
which is always the case then they come from an OR node, but here
we're getting them from two different TRUNC nodes, so we have to check
the types.
llvm-svn: 319695
If the truncation has been pushed past the or-node, look through it and
truncate afterwards.
Differential revision: https://reviews.llvm.org/D40792
llvm-svn: 319692
As part of the unification of the debug format and the MIR format, print
MBB references as '%bb.5'.
The MIR printer prints the IR name of a MBB only for block definitions.
* find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g'
* find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g'
* find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g'
* grep -nr 'BB#' and fix
Differential Revision: https://reviews.llvm.org/D40422
llvm-svn: 319665
Both LoadedVT and NarrowLoad are passed as references and neither
of them are used by any of its callers.
Differential Revision: https://reviews.llvm.org/D40713
llvm-svn: 319645
If we have a non-splat constant shift amount, the minimum shift amount can be used to infer the number of zero upper bits of the result. There's probably a lot more that we can do here, but this
fixes a case where I wanted to infer the sign bit as zero when all the shift amounts are non-zero.
llvm-svn: 319639
SelectionDAGISel::LowerArguments assumes sret addr space is 0, which is
not true for amdgcn---amdgiz target.
This patch fixes that.
Differential Revision: https://reviews.llvm.org/D40255
llvm-svn: 319630
Two issues found when doing codegen for splitting vector with non-zero alloca addr space:
DAGTypeLegalizer::SplitVecRes_INSERT_VECTOR_ELT/SplitVecOp_EXTRACT_VECTOR_ELT uses dummy pointer info for creating
SDStore. Since one pointer operand contains multiply and add, InferPointerInfo is unable to
infer the correct pointer info, which ends up with a dummy pointer info for the target to lower
store and results in isel failure. The fix is to introduce MachinePointerInfo::getUnknownStack to
represent MachinePointerInfo which is known in alloca address space but without other information.
TargetLowering::getVectorElementPointer uses value type of pointer in addr space 0 for
multiplication of index and then add it to the pointer. However the pointer may be in an addr
space which has different size than addr space 0. The fix is to use the pointer value type for
index multiplication.
Differential Revision: https://reviews.llvm.org/D39758
llvm-svn: 319622
Summary:
1/ Operand folding during complex pattern matching for LEAs has been extended, such that it promotes Scale to
accommodate similar operand appearing in the DAG e.g.
T1 = A + B
T2 = T1 + 10
T3 = T2 + A
For above DAG rooted at T3, X86AddressMode will now look like
Base = B , Index = A , Scale = 2 , Disp = 10
2/ During OptimizeLEAPass down the pipeline factorization is now performed over LEAs so that if there is an opportunity
then complex LEAs (having 3 operands) could be factored out e.g.
leal 1(%rax,%rcx,1), %rdx
leal 1(%rax,%rcx,2), %rcx
will be factored as following
leal 1(%rax,%rcx,1), %rdx
leal (%rdx,%rcx) , %edx
3/ Aggressive operand folding for AM based selection for LEAs is sensitive to loops, thus avoiding creation of any complex LEAs within a loop.
4/ Simplify LEA converts (lea (BASE,1,INDEX,0) --> add (BASE, INDEX) which offers better through put.
PR32755 will be taken care of by this pathc.
Previous patch revisions : r313343 , r314886
Reviewers: lsaba, RKSimon, craig.topper, qcolombet, jmolloy, jbhateja
Reviewed By: lsaba, RKSimon, jbhateja
Subscribers: jmolloy, spatel, igorb, llvm-commits
Differential Revision: https://reviews.llvm.org/D35014
llvm-svn: 319543
Type promotion makes no guarantee about the contents of the promoted bits. Since the gather/scatter instruction will use the bits to calculate addresses, we need to ensure they aren't garbage.
llvm-svn: 319520
These command line options are not intended for public use, and often
don't even make sense in the context of a particular tool anyway. About
90% of them are already hidden, but when people add new options they
forget to hide them, so if you were to make a brand new tool today, link
against one of LLVM's libraries, and run tool -help you would get a
bunch of junk that doesn't make sense for the tool you're writing.
This patch hides these options. The real solution is to not have
libraries defining command line options, but that's a much larger effort
and not something I'm prepared to take on.
Differential Revision: https://reviews.llvm.org/D40674
llvm-svn: 319505
output
As part of the unification of the debug format and the MIR format,
always use `printReg` to print all kinds of registers.
Updated the tests using '_' instead of '%noreg' until we decide which
one we want to be the default one.
Differential Revision: https://reviews.llvm.org/D40421
llvm-svn: 319445
visitAND attempts to narrow the width of extending loads that are
then masked off. ReduceLoadWidth already exists for a similar purpose
and handles shifts, so I've moved the code to handle AND nodes there.
Differential Revision: https://reviews.llvm.org/D39595
llvm-svn: 319421
If we put in an assertsext/zext here, we're able to generate better truncate code using pack on pre-avx512 targets.
Similar is already done during type legalization. This is the equivalent for op legalization
Differential Revision: https://reviews.llvm.org/D40591
llvm-svn: 319368
The object can't straddle the address space
wrap around, so I think it's OK to assume any
offsets added to the base object pointer can't
overflow. Similar logic already appears to be
applied in SelectionDAGBuilder when lowering
aggregate returns.
llvm-svn: 319272
Previously we had an isel pattern to add the truncate. Instead use Promote to add the truncate to the DAG before isel.
The Promote legalization code had to be updated to prevent an infinite loop if promotion took multiple steps because it wasn't remembering the previously tried value.
llvm-svn: 319259
Summary:
Recommitting this with the correct sorting predicate. The Low field of Clusters is a ConstantInt and
cannot be directly compared. So we needed to invoke slt (signed less than) to compare correctly.
This fixes failures in the following tests uncovered by D39245:
LLVM :: CodeGen/ARM/ifcvt3.ll
LLVM :: CodeGen/ARM/switch-minsize.ll
LLVM :: CodeGen/X86/switch.ll
LLVM :: CodeGen/X86/switch-bt.ll
LLVM :: CodeGen/X86/switch-density.ll
Reviewers: hans, fhahn
Reviewed By: hans
Subscribers: aemerson, llvm-commits, kristof.beyls
Differential Revision: https://reviews.llvm.org/D40541
llvm-svn: 319210
This is needed for cases when the memory access is not as big as the width of
the data type. For instance, storing i1 (1 bit) would be done in a byte (8
bits).
Using 'BitSize >> 3' (or '/ 8') would e.g. give the memory access of an i1 a
size of 0, which for instance makes alias analysis return NoAlias even when
it shouldn't.
There are no tests as this was done as a follow-up to the bugfix for the case
where this was discovered (r318824). This handles more similar cases.
Review: Björn Petterson
https://reviews.llvm.org/D40339
llvm-svn: 319173
LLVM Coding Standards:
Function names should be verb phrases (as they represent actions), and
command-like function should be imperative. The name should be camel
case, and start with a lower case letter (e.g. openFile() or isFoo()).
Differential Revision: https://reviews.llvm.org/D40416
llvm-svn: 319168
Unoptimized IR can have linear sequences of stores to an array, where the
initial GEP for the first store is formed from the pointer to the array, and the
GEP for each store after the first is formed from the previous GEP with some
offset in an inductive fashion.
The (large) resulting DAG when analyzed by DAGCombine undergoes an excessive
number of combines as each store node is examined every time its' offset node
is combined with any child of the offset. One of the transformations is
findBetterNeighborChains which assists MergeConsecutiveStores. The former
relies on repeated chain walking to do its' work, however MergeConsecutiveStores
is disabled at O0 which makes the transformation redundant.
Any optimization level other than O0 would invoke InstCombine which would
resolve the chain of GEPs into flat base + offset GEP for each store which
does not exhibit the repeated examination of each store to the array.
Disabling this optimization fixes an excessive compile time issue (30~ minutes
for the test case provided) at O0.
Reviewers: niravd, craig.topper, t.p.northover
Differential Revision: https://reviews.llvm.org/D40193
llvm-svn: 319142
With AVX512 vXi1 types are legal so we shouldn't be extending them.
This change is similar to existing code in the zext(setcc) combine.
llvm-svn: 319120
Summary:
Currently ScalarizeVecRes_SETCC checks for the result type being a vector and jumps to ScalarizeVecRes_VSETCC. But if we're scalarizing a vector result, aren't we guaranteed to be looking at a vector type?
This patch deletes the current ScalarizeVecRes_SETCC and renames ScalarizeVecRes_VSETCC to ScalarizeVecRes_SETCC.
Reviewers: RKSimon, arsenm, eladcohen, zvi
Reviewed By: RKSimon
Subscribers: wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D40452
llvm-svn: 318982
This patch reverts change to X86TargetLowering::getScalarShiftAmountTy in
rL318727 and move the logic to DAGTypeLegalizer::SplitInteger.
The reason is that getScalarShiftAmountTy returns a shift amount type that
is suitable for common use cases in CodeGen. DAGTypeLegalizer::SplitInteger
is a rare situation which requires a shift amount type larger than what
getScalarShiftAmountTy. In this case, it is more reasonable to do special
handling of shift amount type in DAGTypeLegalizer::SplitInteger only. If
similar situations arises the logic may be moved to a separate function.
Differential Revision: https://reviews.llvm.org/D40320
llvm-svn: 318890
Since i1 is a legal type, this:
NumBytes = Op1->getMemoryVT().getSizeInBits() >> 3;
is wrong and should be instead
NumBytes = Op0->getMemoryVT().getStoreSize();
There seems to be more places where this should be fixed outside DAGCombiner.
Review: Hal Finkel
https://bugs.llvm.org/show_bug.cgi?id=35366
llvm-svn: 318824
DAGTypeLegalizer::SplitInteger uses default pointer size as shift amount constant type,
which causes less performant ISA in amdgcn---amdgiz target since the default pointer
type is i64 whereas the desired shift amount type is i32.
This patch fixes that by using TLI.getScalarShiftAmountTy in DAGTypeLegalizer::SplitInteger.
The X86 change is necessary since splitting i512 requires shifting amount of 256, which
cannot be held by i8.
Differential Revision: https://reviews.llvm.org/D40148
llvm-svn: 318727
Normally this would be cleaned up by promoting the condition operand next. But in the attached case we promoted the result from v2i48 to v2i64 and the condition from v2i1 to v2i48. Then we tried to "promote" the v2i48 condition back to v2i1 because that's what the SetCC result type for v2i64 is on X86 with VLX. But promote is either a NOP or SIGN_EXTEND and this would need a truncation.
With the change here we now get the SetCC type of v2i1 when we're handling the result promotion and the operand no longer needs to be promoted itself.
Fixes PR35272.
llvm-svn: 318706
Previously we were assuming all results were vectors and calling SetWidenedVector, but if its a chain result we should just replace uses instead.
This fixes an error found by expensive checks after r318368.
llvm-svn: 318509
TransferDbgValues (capital 'T') is wired into ReplaceAllUsesWith, and
transferDbgValues (lowercase 't') is used elsewhere (e.g in Legalize).
Both functions should be doing the exact same thing. This patch
consolidates the logic into one place.
This was reverted in r318455 because some newly introduced asserts,
which I thought were NFC, were firing. I filed PR35338. For now I've
weakened the asserts.
Testing: check-llvm, check-clang, and a stage2 Rel+Deb build of clang
Differential Revision: https://reviews.llvm.org/D40104
llvm-svn: 318498
All these headers already depend on CodeGen headers so moving them into
CodeGen fixes the layering (since CodeGen depends on Target, not the
other way around).
llvm-svn: 318490
The sign extend might be from an i16 or i8 type and was inserted by InstCombine to match the pointer width. X86 gather legalization isn't currently detecting this to reinsert a sign extend to make things legal.
It's a bit weird for the SelectionDAGBuilder to do this kind of optimization in the first place. With this removed we can at least lean on InstCombine somewhat to ensure the index is i32 or i64.
I'll work on trying to recover some of the test cases by removing sign extends in the backend when its safe to do so with an understanding of the current legalizer capabilities.
This should fix PR30690.
llvm-svn: 318466
TransferDbgValues (capital 'T') is wired into ReplaceAllUsesWith, and
transferDbgValues (lowercase 't') is used elsewhere (e.g in Legalize).
Both functions should be doing the exact same thing. This patch
consolidates the logic into one place.
Differential Revision: https://reviews.llvm.org/D40104
llvm-svn: 318448
SelectionDAGBuilder::visitAlloca assumes alloca address space is 0, which is
incorrect for triple amdgcn---amdgiz and causes isel failure.
This patch fixes that.
Differential Revision: https://reviews.llvm.org/D40095
llvm-svn: 318392
Change the calculation for the desired ValueType for non-sign
extending loads, as in those cases we don't care about the
higher bits. This creates a smaller ExtVT and allows for such
combinations as:
(srl (zextload i16, [addr]), 8) -> (zextload i8, [addr + 1])
Differential Revision: https://reviews.llvm.org/D40034
llvm-svn: 318390
For example, this is currently reachable by X86 if you use a masked store intrinsic with a v1iX type.
Using a fatal error seems like a better user experience if someone were to encounter this on a release build. There are several other similar places that have been converted from unreachable to fatal error previously.
llvm-svn: 318379
processDbgDeclares assumes pointer size is the same for different addr spaces.
It uses pointer size for addr space 0 for all pointers, which causes assertion
in stripAndAccumulateInBoundsConstantOffsets for amdgcn---amdgiz since
pointer in addr space 5 has different size than in addr space 0.
This patch fixes that.
Differential Revision: https://reviews.llvm.org/D40085
llvm-svn: 318370
Due to integer precision, we might have numerator greater than denominator in
the branch probability scaling. Add a check to prevent this from happening.
llvm-svn: 318353
This patch peels off the top case in switch statement into a branch if the
probability exceeds a threshold. This will help the branch prediction and
avoids the extra compares when lowering into chain of branches.
Differential Revision: http://reviews.llvm.org/D39262
llvm-svn: 318202
TargetLowering::LowerCallTo assumes that sret value type corresponds to a
pointer in default address space, which is incorrect, since sret value type
should correspond to a pointer in alloca address space, which may not
be the default address space. This causes assertion for amdgcn target
in amdgiz environment.
This patch fixes that.
Differential Revision: https://reviews.llvm.org/D39996
llvm-svn: 318167
This is a fix for a bug in r317947. We were supposed to check that all the indices are are constant 0, but instead we're only make sure that indices that are constant are 0. Non-constant indices are being ignored.
llvm-svn: 317950
Currently we can only get a uniform base from a simple GEP with 2 operands. This causes us to miss address folding opportunities for simple global array accesses as the test case shows.
This patch adds support for larger GEPs if the other indices are 0 since those don't require any additional computations to be inserted.
We may also want to handle constant splats of zero here, but I'm leaving that for future work when I have a real world example.
Differential Revision: https://reviews.llvm.org/D39911
llvm-svn: 317947
This patch implements Chandler's idea [0] for supporting languages that
require support for infinite loops with side effects, such as Rust, providing
part of a solution to bug 965 [1].
Specifically, it adds an `llvm.sideeffect()` intrinsic, which has no actual
effect, but which appears to optimization passes to have obscure side effects,
such that they don't optimize away loops containing it. It also teaches
several optimization passes to ignore this intrinsic, so that it doesn't
significantly impact optimization in most cases.
As discussed on llvm-dev [2], this patch is the first of two major parts.
The second part, to change LLVM's semantics to have defined behavior
on infinite loops by default, with a function attribute for opting into
potential-undefined-behavior, will be implemented and posted for review in
a separate patch.
[0] http://lists.llvm.org/pipermail/llvm-dev/2015-July/088103.html
[1] https://bugs.llvm.org/show_bug.cgi?id=965
[2] http://lists.llvm.org/pipermail/llvm-dev/2017-October/118632.html
Differential Revision: https://reviews.llvm.org/D38336
llvm-svn: 317729
In 2010 a commit with no testcase and no further explanation
explicitly disabled the handling of inlined variables in
EmitFuncArgumentDbgValue(). I don't think there is a good reason for
this any more and re-enabling this adds debug locations for variables
associated with an LLVM function argument in functions that are
inlined into the first basic block. The only downside of doing this is
that we may insert a DBG_VALUE before the inlined scope, but (1) this
could be filtered out later, and (2) LiveDebugValues will not
propagate it into subsequent basic blocks if they don't dominate the
variable's lexical scope, so this seems like a small price to pay.
rdar://problem/26228128
llvm-svn: 317702
Some of the AMDGPU stack addressing modes require knowing the sign
bit is zero. We used to accomplish this by custom lowering
frame indexes, and then putting an AssertZext around a
TargetFrameIndex. This required specifically looking for
the AssextZext + frame index pattern which was moderately
disgusting. The same could probably be accomplished
with a target specific node, but would still
require special handling of frame indexes.
llvm-svn: 317671
This header includes CodeGen headers, and is not, itself, included by
any Target headers, so move it into CodeGen to match the layering of its
implementation.
llvm-svn: 317647
As discussed on llvm-dev:
http://lists.llvm.org/pipermail/llvm-dev/2016-November/107104.html
and again more recently:
http://lists.llvm.org/pipermail/llvm-dev/2017-October/118118.html
...this is a step in cleaning up our fast-math-flags implementation in IR to better match
the capabilities of both clang's user-visible flags and the backend's flags for SDNode.
As proposed in the above threads, we're replacing the 'UnsafeAlgebra' bit (which had the
'umbrella' meaning that all flags are set) with a new bit that only applies to algebraic
reassociation - 'AllowReassoc'.
We're also adding a bit to allow approximations for library functions called 'ApproxFunc'
(this was initially proposed as 'libm' or similar).
...and we're out of bits. 7 bits ought to be enough for anyone, right? :) FWIW, I did
look at getting this out of SubclassOptionalData via SubclassData (spacious 16-bits),
but that's apparently already used for other purposes. Also, I don't think we can just
add a field to FPMathOperator because Operator is not intended to be instantiated.
We'll defer movement of FMF to another day.
We keep the 'fast' keyword. I thought about removing that, but seeing IR like this:
%f.fast = fadd reassoc nnan ninf nsz arcp contract afn float %op1, %op2
...made me think we want to keep the shortcut synonym.
Finally, this change is binary incompatible with existing IR as seen in the
compatibility tests. This statement:
"Newer releases can ignore features from older releases, but they cannot miscompile
them. For example, if nsw is ever replaced with something else, dropping it would be
a valid way to upgrade the IR."
( http://llvm.org/docs/DeveloperPolicy.html#ir-backwards-compatibility )
...provides the flexibility we want to make this change without requiring a new IR
version. Ie, we're not loosening the FP strictness of existing IR. At worst, we will
fail to optimize some previously 'fast' code because it's no longer recognized as
'fast'. This should get fixed as we audit/squash all of the uses of 'isFast()'.
Note: an inter-dependent clang commit to use the new API name should closely follow
commit.
Differential Revision: https://reviews.llvm.org/D39304
llvm-svn: 317488
This header already includes a CodeGen header and is implemented in
lib/CodeGen, so move the header there to match.
This fixes a link error with modular codegeneration builds - where a
header and its implementation are circularly dependent and so need to be
in the same library, not split between two like this.
llvm-svn: 317379
When splitting a large load to smaller legally-typed loads, the last load should be padded to reach the size of the previous one so a CONCAT_VECTORS node could reunite them again.
The code currently pads the last load to reach the size of the first load (instead of the previous).
Differential Revision: https://reviews.llvm.org/D38495
Change-Id: Ib60b55ed26ce901fabf68108daf52683fbd5013f
llvm-svn: 317206
Summary:
For reference, see: http://lists.llvm.org/pipermail/llvm-dev/2017-August/116589.html
This patch fleshes out the instruction class hierarchy with respect to atomic and
non-atomic memory intrinsics. With this change, the relevant part of the class
hierarchy becomes:
IntrinsicInst
-> MemIntrinsicBase (methods-only class)
-> MemIntrinsic (non-atomic intrinsics)
-> MemSetInst
-> MemTransferInst
-> MemCpyInst
-> MemMoveInst
-> AtomicMemIntrinsic (atomic intrinsics)
-> AtomicMemSetInst
-> AtomicMemTransferInst
-> AtomicMemCpyInst
-> AtomicMemMoveInst
-> AnyMemIntrinsic (both atomicities)
-> AnyMemSetInst
-> AnyMemTransferInst
-> AnyMemCpyInst
-> AnyMemMoveInst
This involves some class renaming:
ElementUnorderedAtomicMemCpyInst -> AtomicMemCpyInst
ElementUnorderedAtomicMemMoveInst -> AtomicMemMoveInst
ElementUnorderedAtomicMemSetInst -> AtomicMemSetInst
A script for doing this renaming in downstream trees is included below.
An example of where the Any* classes should be used in LLVM is when reasoning
about the effects of an instruction (ex: aliasing).
---
Script for renaming AtomicMem* classes:
PREFIXES="[<,([:space:]]"
CLASSES="MemIntrinsic|MemTransferInst|MemSetInst|MemMoveInst|MemCpyInst"
SUFFIXES="[;)>,[:space:]]"
REGEX="(${PREFIXES})ElementUnorderedAtomic(${CLASSES})(${SUFFIXES})"
REGEX2="visitElementUnorderedAtomic(${CLASSES})"
FILES=$( grep -E "(${REGEX}|${REGEX2})" -r . | tr ':' ' ' | awk '{print $1}' | sort | uniq )
SED_SCRIPT="s~${REGEX}~\1Atomic\2\3~g"
SED_SCRIPT2="s~${REGEX2}~visitAtomic\1~g"
for f in $FILES; do
echo "Processing: $f"
sed -i ".bak" -E "${SED_SCRIPT};${SED_SCRIPT2};${EA_SED_SCRIPT};${EA_SED_SCRIPT2}" $f
done
Reviewers: sanjoy, deadalnix, apilipenko, anna, skatkov, mkazantsev
Reviewed By: sanjoy
Subscribers: hfinkel, jholewinski, arsenm, sdardis, nhaehnle, JDevlieghere, javed.absar, llvm-commits
Differential Revision: https://reviews.llvm.org/D38419
llvm-svn: 316950
Introduce a isConstOrDemandedConstSplat helper function that can recognise a constant splat build vector for at least the demanded elts we care about.
llvm-svn: 316866
For cases where we know the floating point representations match the bitcasted integer equivalent, allow bitcasting to these types.
This is especially useful for the X86 floating point compare results which return all/zero bits but as a floating point type.
Differential Revision: https://reviews.llvm.org/D39289
llvm-svn: 316831
In function DAGCombiner::visitSIGN_EXTEND_INREG, sext can be combined with extload even if sextload is not supported by target, then
if sext is the only user of extload, there is no big difference, no harm no benefit.
if extload has more than one user, the combined sextload may block extload from combining with other zext, causes extra zext instructions generated. As demonstrated by the attached test case.
This patch add the constraint that when sextload is not supported by target, sext can only be combined with extload if it is the only user of extload.
Differential Revision: https://reviews.llvm.org/D39108
llvm-svn: 316802
Not having the subclass data on an MemIntrinsicSDNodes means it was possible
to try to fold 2 nodes with the same operands but differing MMO flags. This
would trip an assertion when trying to refine the alignment between the 2
MachineMemOperands.
Differential Revision: https://reviews.llvm.org/D38898
llvm-svn: 316737
This code added in r297930 assumed that it could create
a select with a condition type that is just an integer
bitcast of the selected type. For AMDGPU any vselect is
going to be scalarized (although the vector types are legal),
and all select conditions must be i1 (the same as getSetCCResultType).
This logic doesn't really make sense to me, but there's
never really been a consistent policy in what the select
condition mask type is supposed to be. Try to extend
the logic for skipping the transform for condition types
that aren't setccs. It doesn't seem quite right to me though,
but checking conditions that seem more sensible (like whether the
vselect is going to be expanded) doesn't work since this
seems to depend on that also.
llvm-svn: 316554
Similar to how llvm::salvagDebugInfo hooks into InstCombine, this adds
a hook that can be invoked before an SDNode that is associated with an
SDDbgValue is erased to capture the effect of the deleted node in a
DIExpression.
The motivating example is an SDDebugValue attached to an ADD operation
that gets folded into a LOAD+OFFSET operation.
rdar://problem/32121503
llvm-svn: 316525
This fixes a bug where we'd crash given code like the test-case from
https://bugs.llvm.org/show_bug.cgi?id=30792 . Instead, we let the
offending clobber silently slide through.
This doesn't fully fix said bug, since the assembler will still complain
the moment it sees a crypto/fp/vector op, and we still don't diagnose
calls that require vector regs.
Differential Revision: https://reviews.llvm.org/D39030
llvm-svn: 316374
combineShuffleOfScalars is very conservative about shuffled BUILD_VECTORs that can be combined together.
This patch adds one additional case - if both BUILD_VECTORs represent splats of the same scalar value but with different UNDEF elements, then we should create a single splat BUILD_VECTOR, sharing only the UNDEF elements defined by the shuffle mask.
Differential Revision: https://reviews.llvm.org/D38696
llvm-svn: 316331
We don't need to do any additional recursion, we just need to analyze the APInt stored in the node. This matches what the ValueTracking versions do for IR.
llvm-svn: 316256
Summary:
We shouldn't recurse any further but it doesn't mean we shouldn't be able to give the known bits for a constant. The caller would probably like that we always return the right answer for a constant RHS. This matches what InstCombine does in this case.
I don't have a test case because this showed up while trying to revive D31724.
Reviewers: RKSimon, spatel
Reviewed By: RKSimon
Subscribers: arsenm, llvm-commits
Differential Revision: https://reviews.llvm.org/D38967
llvm-svn: 316255
Minor addition and follow up of r314773 and r311533: this adds more
debug messages to the type legalizer. For each node, it dumps
legalization info for results and operands nodes, rather than just the
final legalized node.
Differential Revision: https://reviews.llvm.org/D38726
llvm-svn: 315904
This is only currently used for mad/fma transforms.
This is the only case where it should be used for AMDGPU,
so add an opcode to be sure.
llvm-svn: 315740
I don't know if we ever hit this case or not. Turning it into an assert only fired on expanding some atomic operation in a SystemZ lit test.
llvm-svn: 315648
Summary:
Add LLVM_FORCE_ENABLE_DUMP cmake option, and use it along with
LLVM_ENABLE_ASSERTIONS to set LLVM_ENABLE_DUMP.
Remove NDEBUG and only use LLVM_ENABLE_DUMP to enable dump methods.
Move definition of LLVM_ENABLE_DUMP from config.h to llvm-config.h so
it'll be picked up by public headers.
Differential Revision: https://reviews.llvm.org/D38406
llvm-svn: 315590
Eg:
insert v4i32 V, (v2i16 X), 2 --> shuffle v8i16 V', X', {0,1,2,3,8,9,6,7}
This is a generalization of the IR fold in D38316 to handle insertion into a non-undef vector.
We may want to abandon that one if we can't find value in squashing the more specific pattern sooner.
We're using the existing legal shuffle target hook to avoid AVX512 horror with vXi1 shuffles.
There may be room for improvement in the shuffle lowering here, but that would be follow-up work.
Differential Revision: https://reviews.llvm.org/D38388
llvm-svn: 315460
The NumFixedArgs field of CallLoweringInfo is used by
TargetLowering::LowerCallTo to determine whether a given argument is passed
using the vararg calling convention or not (specifically, to set IsFixed for
each ISD::OutputArg).
Firstly, CallLoweringInfo::setLibCallee and CallLoweringInfo::setCallee both
incorrectly set NumFixedArgs based on the _previous_ args list. Secondly,
TargetLowering::LowerCallTo failed to increment NumFixedArgs when modifying
the argument list so a pointer is passed for the return value.
If your backend uses the IsFixed property or directly accesses NumFixedArgs,
it is _possible_ this change could result in codegen changes (although the
previous behaviour would have been incorrect). No such cases have been
identified during code review for any in-tree architecture.
Differential Revision: https://reviews.llvm.org/D37898
llvm-svn: 315457
Summary:
See https://llvm.org/PR33743 for more details
It seems that for non-power of 2 vector sizes, the algorithm can produce
non-matching sizes for input and result causing an assert.
This usually isn't a problem as the isAnyExtend check will weed these out, but
in some cases (most often with lots of undefined values for the mask indices) it
can pass this check for non power of 2 vectors.
Adding in an extra check that ensures that bit size will match for the result
and input (as required)
Subscribers: nhaehnle
Differential Revision: https://reviews.llvm.org/D35241
llvm-svn: 315307
Summary:
FastISel::hasTrivialKill() was the only user of the "IntPtrTy" version of
Cast::isNoopCast(). According to review comments in D37894 we could instead
use the "DataLayout" version of the method, and thus get rid of the
"IntPtrTy" versions of isNoopCast() completely.
With the above done, the remaining isNoopCast() could then be simplified
a bit more.
Reviewers: arsenm
Reviewed By: arsenm
Subscribers: wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D38497
llvm-svn: 314969
It broke the Chromium / SQLite build; see PR34830.
> Summary:
> 1/ Operand folding during complex pattern matching for LEAs has been
> extended, such that it promotes Scale to accommodate similar operand
> appearing in the DAG.
> e.g.
> T1 = A + B
> T2 = T1 + 10
> T3 = T2 + A
> For above DAG rooted at T3, X86AddressMode will no look like
> Base = B , Index = A , Scale = 2 , Disp = 10
>
> 2/ During OptimizeLEAPass down the pipeline factorization is now performed over LEAs
> so that if there is an opportunity then complex LEAs (having 3 operands)
> could be factored out.
> e.g.
> leal 1(%rax,%rcx,1), %rdx
> leal 1(%rax,%rcx,2), %rcx
> will be factored as following
> leal 1(%rax,%rcx,1), %rdx
> leal (%rdx,%rcx) , %edx
>
> 3/ Aggressive operand folding for AM based selection for LEAs is sensitive to loops,
> thus avoiding creation of any complex LEAs within a loop.
>
> Reviewers: lsaba, RKSimon, craig.topper, qcolombet, jmolloy
>
> Reviewed By: lsaba
>
> Subscribers: jmolloy, spatel, igorb, llvm-commits
>
> Differential Revision: https://reviews.llvm.org/D35014
llvm-svn: 314919
Summary:
1/ Operand folding during complex pattern matching for LEAs has been
extended, such that it promotes Scale to accommodate similar operand
appearing in the DAG.
e.g.
T1 = A + B
T2 = T1 + 10
T3 = T2 + A
For above DAG rooted at T3, X86AddressMode will no look like
Base = B , Index = A , Scale = 2 , Disp = 10
2/ During OptimizeLEAPass down the pipeline factorization is now performed over LEAs
so that if there is an opportunity then complex LEAs (having 3 operands)
could be factored out.
e.g.
leal 1(%rax,%rcx,1), %rdx
leal 1(%rax,%rcx,2), %rcx
will be factored as following
leal 1(%rax,%rcx,1), %rdx
leal (%rdx,%rcx) , %edx
3/ Aggressive operand folding for AM based selection for LEAs is sensitive to loops,
thus avoiding creation of any complex LEAs within a loop.
Reviewers: lsaba, RKSimon, craig.topper, qcolombet, jmolloy
Reviewed By: lsaba
Subscribers: jmolloy, spatel, igorb, llvm-commits
Differential Revision: https://reviews.llvm.org/D35014
llvm-svn: 314886
Summary:
Take the target's endianness into account when splitting the
debug information in DAGTypeLegalizer::SetExpandedInteger.
This patch fixes so that, for big-endian targets, the fragment
expression corresponding to the high part of a split integer
value is placed at offset 0, in order to correctly represent
the memory address order.
I have attached a PPC32 reproducer where the resulting DWARF
pieces for a 64-bit integer were incorrectly reversed.
Original patch was reverted due to using -stop-after=isel in
the test case (but that is only working when AMDGPU target
is included in the llc build). The test case has now been
updated to use -stop-before=expand-isel-pseudos instead.
Patch by: dstenb
Reviewers: JDevlieghere, aprantl, dblaikie
Reviewed By: JDevlieghere, aprantl, dblaikie
Subscribers: nemanjai
Differential Revision: https://reviews.llvm.org/D38172
llvm-svn: 314781
This adds some more debug messages to the type legalizer and functions
like PromoteNode, ExpandNode, ExpandLibCall in an attempt to make
the debug messages a little bit more informative and useful.
Differential Revision: https://reviews.llvm.org/D38450
llvm-svn: 314773
Summary:
Take the target's endianness into account when splitting the
debug information in DAGTypeLegalizer::SetExpandedInteger.
This patch fixes so that, for big-endian targets, the fragment
expression corresponding to the high part of a split integer
value is placed at offset 0, in order to correctly represent
the memory address order.
I have attached a PPC32 reproducer where the resulting DWARF
pieces for a 64-bit integer were incorrectly reversed.
Patch by: dstenb
Reviewers: JDevlieghere, aprantl, dblaikie
Reviewed By: JDevlieghere, aprantl, dblaikie
Subscribers: nemanjai
Differential Revision: https://reviews.llvm.org/D38172
llvm-svn: 314666
Currently expandUnalignedLoad/Store uses place holder pointer info for temporary memory operand
in stack, which does not have correct address space. This causes unaligned private double16 load/store to be
lowered to flat_load instead of buffer_load for amdgcn target.
This fixes failures of OpenCL conformance test basic/vload_private/vstore_private on target amdgcn---amdgizcl.
Differential Revision: https://reviews.llvm.org/D35361
llvm-svn: 314566
Without this, we could end up trying to get the Nth (0-indexed) element
from a subvector of size N.
Differential Revision: https://reviews.llvm.org/D37880
llvm-svn: 314380
This teach simplifyDemandedBits to handle constant splat vector shifts.
This required changing some uses of getZExtValue to getLimitedValue since we can't rely on legalization using getShiftAmountTy for the shift amount.
I believe there may have been a bug in the ((X << C1) >>u ShAmt) handling where we didn't check if the inner shift was too large. I've fixed that here.
I had to add new patterns to ARM because the zext/sext the patterns were trying to look for got turned into an any_extend with this patch. Happy to split that out too, but not sure how to test without this change.
Differential Revision: https://reviews.llvm.org/D37665
llvm-svn: 314139
David Blaikie