Ulrich Weigand [Wed, 1 Aug 2018 11:57:58 +0000 (11:57 +0000)]
[SystemZ, TableGen] Fix shift count handling
The DAG combiner logic to simplify AND masks in shift counts is invalid.
While it is true that the SystemZ shift instructions ignore all but the
low 6 bits of the shift count, it is still invalid to simplify the AND
masks while the DAG still uses the standard shift operators (which are
*not* defined to match the SystemZ instruction behavior).
Instead, this patch performs equivalent operations during instruction
selection. For completely removing the AND, this now happens via
additional DAG match patterns implemented by a multi-alternative
PatFrags. For simplifying a 32-bit AND to a 16-bit AND, the existing DAG
patterns were already mostly OK, they just needed an output XForm to
actually truncate the immediate value.
Unfortunately, the latter change also exposed a bug in TableGen: it
seems XForms are currently only handled correctly for direct operands of
the outermost operation node. This patch also fixes that bug by simply
recurring through the whole pattern. This should be NFC for all other
targets.
[DebugInfo] Have custom std::reverse_iterator<DWARFDie>
The DWARFDie is a lightweight utility wrapper that stores a pointer to a
compile unit and a debug info entry. Currently, its iterator (used for
walking over its children) stores a DWARFDie and returns a const
reference when dereferencing it.
When the iterator is modified (by incrementing or decrementing it), this
reference becomes invalid. This was happening when calling reverse on
it, because the std::reverse_iterator is keeping a temporary copy of the
iterator (see
https://en.cppreference.com/w/cpp/iterator/reverse_iterator for a good
illustration).
When dereferencing the reverse iterator, we decrement and return a
reference to a DWARFDie stored in the stack frame of this function,
resulting in UB at runtime.
This patch specifies the std::reverse_iterator for DWARFDie to do the
right thing.
David Bolvansky [Wed, 1 Aug 2018 07:37:16 +0000 (07:37 +0000)]
Enrich inline messages
Summary:
This patch improves Inliner to provide causes/reasons for negative inline decisions.
1. It adds one new message field to InlineCost to report causes for Always and Never instances. All Never and Always instantiations must provide a simple message.
2. Several functions that used to return the inlining results as boolean are changed to return InlineResult which carries the cause for negative decision.
3. Changed remark priniting and debug output messages to provide the additional messages and related inline cost.
4. Adjusted tests for changed printing.
Victor Leschuk [Wed, 1 Aug 2018 05:48:06 +0000 (05:48 +0000)]
[DWARF] Basic support for producing DWARFv5 .debug_addr section
This revision implements support for generating DWARFv5 .debug_addr section.
The implementation is pretty straight-forward: we just check the dwarf version
and emit section header if needed.
Hiroshi Inoue [Wed, 1 Aug 2018 04:40:32 +0000 (04:40 +0000)]
[InstSimplify] fold extracting from std::pair (1/2)
This patch intends to enable jump threading when a method whose return type is std::pair<int, bool> or std::pair<bool, int> is inlined.
For example, jump threading does not happen for the if statement in func.
std::pair<int, bool> callee(int v) {
int a = dummy(v);
if (a) return std::make_pair(dummy(v), true);
else return std::make_pair(v, v < 0);
}
int func(int v) {
std::pair<int, bool> rc = callee(v);
if (rc.second) {
// do something
}
SROA executed before the method inlining replaces std::pair by i64 without splitting in both callee and func since at this point no access to the individual fields is seen to SROA.
After inlining, jump threading fails to identify that the incoming value is a constant due to additional instructions (like or, and, trunc).
This series of patch add patterns in InstructionSimplify to fold extraction of members of std::pair. To help jump threading, actually we need to optimize the code sequence spanning multiple BBs.
These patches does not handle phi by itself, but these additional patterns help NewGVN pass, which calls instsimplify to check opportunities for simplifying instructions over phi, apply phi-of-ops optimization to result in successful jump threading.
SimplifyDemandedBits in InstCombine, can do more general optimization but this patch aims to provide opportunities for other optimizers by supporting a simple but common case in InstSimplify.
This first patch in the series handles code sequences that merges two values using shl and or and then extracts one value using lshr.
[x86] Fix a really subtle miscompile due to a somewhat glaring bug in
EFLAGS copy lowering.
If you have a branch of LLVM, you may want to cherrypick this. It is
extremely unlikely to hit this case empirically, but it will likely
manifest as an "impossible" branch being taken somewhere, and will be
... very hard to debug.
Hitting this requires complex conditions living across complex control
flow combined with some interesting memory (non-stack) initialized with
the results of a comparison. Also, because you have to arrange for an
EFLAGS copy to be in *just* the right place, almost anything you do to
the code will hide the bug. I was unable to reduce anything remotely
resembling a "good" test case from the place where I hit it, and so
instead I have constructed synthetic MIR testing that directly exercises
the bug in question (as well as the good behavior for completeness).
The issue is that we would mistakenly assume any SETcc with a valid
condition and an initial operand that was a register and a virtual
register at that to be a register *defining* SETcc...
It isn't though....
This would in turn cause us to test some other bizarre register,
typically the base pointer of some memory. Now, testing this register
and using that to branch on doesn't make any sense. It even fails the
machine verifier (if you are running it) due to the wrong register
class. But it will make it through LLVM, assemble, and it *looks*
fine... But wow do you get a very unsual and surprising branch taken in
your actual code.
The fix is to actually check what kind of SETcc instruction we're
dealing with. Because there are a bunch of them, I just test the
may-store bit in the instruction. I've also added an assert for sanity
that ensure we are, in fact, *defining* the register operand. =D
[x86/slh] Add unwind info to several tests to make it more obvious that
we aren't incorrectly generating any of it when doing SLH.
There was a bug that only occured with SLH that very much looked like it
could be caused by bad unwind info, and so this was a prime suspect.
Turns out that everything is fine, but this way we'll *see* if we end
up, for example, putting things we shouldn't inside the prolog.
Hsiangkai Wang [Wed, 1 Aug 2018 02:18:06 +0000 (02:18 +0000)]
[DebugInfo] Generate fixups as emitting DWARF .debug_line.
It is necessary to generate fixups in .debug_line as relaxation is
enabled due to the address delta may be changed after relaxation.
DWARF will record the mappings of lines and addresses in
.debug_line section. It will encode the information using special
opcodes, standard opcodes and extended opcodes in Line Number
Program. I use DW_LNS_fixed_advance_pc to encode fixed length
address delta and DW_LNE_set_address to encode absolute address
to make it possible to generate fixups in .debug_line section.
Amara Emerson [Wed, 1 Aug 2018 02:17:42 +0000 (02:17 +0000)]
[GlobalISel][IRTranslator] Use RPO traversal when visiting blocks to translate.
Previously we were just visiting the blocks in the function in IR order, which
is rather arbitrary. Therefore we wouldn't always visit defs before uses, but
the translation code relies on this assumption in some places.
Only codegen change seen in tests is an elision of a redundant copy.
Eli Friedman [Wed, 1 Aug 2018 00:37:20 +0000 (00:37 +0000)]
[MachineOutliner] Clean up subtarget handling.
Call shouldOutlineFromFunctionByDefault, isFunctionSafeToOutlineFrom,
getOutliningType, and getMachineOutlinerMBBFlags using the correct
TargetInstrInfo. And don't create a MachineFunction for a function
declaration.
The call to getOutliningCandidateInfo is still a little weird, but at
least the weirdness is explicitly called out.
Fangrui Song [Tue, 31 Jul 2018 21:26:35 +0000 (21:26 +0000)]
[llvm-objcopy] Make --strip-debug strip .gdb_index
Summary:
See binutils-gdb/bfd/elf.c, GNU objcopy also strips .stab* (STABS)
.line* (DWARF 1) .gnu.linkonce.wi.* (linkonce section for .debug_info) but
I'm not sure we need to be compatible with it.
Matt Arsenault [Tue, 31 Jul 2018 19:05:14 +0000 (19:05 +0000)]
AMDGPU: Scalarize vector argument types to calls
When lowering calling conventions, prefer to decompose vectors
into the constitute register types. This avoids artifical constraints
to satisfy a wide super-register.
This improves code quality because now optimizations don't need to
deal with the super-register constraint. For example the immediate
folding code doesn't deal with 4 component reg_sequences, so by
breaking the register down earlier the existing immediate folding
code is able to work.
This also avoids the need for the shader input processing code
to manually split vector types.
Simon Pilgrim [Tue, 31 Jul 2018 18:05:56 +0000 (18:05 +0000)]
[X86][SSE] Use ISD::MULHU for constant/non-zero ISD::SRL lowering (PR38151)
As was done for vector rotations, we can efficiently use ISD::MULHU for vXi8/vXi16 ISD::SRL lowering.
Shift-by-zero cases are still problematic (mainly on v32i8 due to extra AND/ANDN/OR or VPBLENDVB blend masks but v8i16/v16i16 aren't great either if PBLENDW fails) so I've limited this first patch to known non-zero cases if we can't easily use PBLENDW.
This patch does the same thing as r338153 for COFF.
Note that this patch affects only the order of log messages.
The output file is already deterministic.
Summary:
Similar to D49636, but for PMADDUBSW. This instruction has the additional complexity that the addition of the two products saturates to 16-bits rather than wrapping around. And one operand is treated as signed and the other as unsigned.
A C example that triggers this pattern
```
static const int N = 128;
int8_t A[2*N];
uint8_t B[2*N];
int16_t C[N];
void foo() {
for (int i = 0; i != N; ++i)
C[i] = MIN(MAX((int16_t)A[2*i]*(int16_t)B[2*i] + (int16_t)A[2*i+1]*(int16_t)B[2*i+1], -32768), 32767);
}
```
[X86] Preserve more liveness information in emitStackProbeInline
This commit fixes two issues with the liveness information after the
call:
1) The code always spills RCX and RDX if InProlog == true, which results
in an use of undefined phys reg.
2) FinalReg, JoinReg, RoundedReg, SizeReg are not added as live-ins to
the basic blocks that use them, therefore they are seen undefined.
Hsiangkai Wang [Tue, 31 Jul 2018 16:22:09 +0000 (16:22 +0000)]
[DebugInfo] Fix build failed in 'clang-cmake-armv8-full'.
Builder clang-cmake-armv8-full failed due to the assembly 'comment'
notation is not '#' in the target. So, I use CHECK-SAME to avoid to
check the comment notation in the same line in the test case.
Jakub Kuderski [Tue, 31 Jul 2018 15:53:10 +0000 (15:53 +0000)]
[Dominators] Make slow walks shorter
Summary:
When DFS numbers are not yet calculated for a dominator tree, we have to walk it up to say whether one node dominates some other.
This patch makes the slow walks shorter by only walking until the level of the node we check against is reached. This is because a node cannot possibly dominate something higher in its tree.
When running opt with -O3, the patch results in:
* 25% fewer loop iterations for `opt` (fullLTO)
* 30% fewer loop iterations for sqlite
Workaround bug where the InstCombine pass was asserting on the IR added in lit
test, where we have a bitcast instruction after a GEP from an addrspace cast.
The second bitcast in the test was getting combined into
`bitcast <16 x i32>* %0 to <16 x i32> addrspace(3)*`, which looks like it should
be an addrspace cast instruction instead. Otherwise if control flow is allowed
to continue as it is now we create a GEP instruction
`<badref> = getelementptr inbounds <16 x i32>, <16 x i32>* %0, i32 0`. However
because the type of this instruction doesn't match the address space we hit an
assert when replacing the bitcast with that GEP.
```
void llvm::Value::doRAUW(llvm::Value*, bool): Assertion `New->getType() == getType() && "replaceAllUses of value with new value of different type!"' failed.
```
We will collect label information from DBG_LABEL. Before every DBG_LABEL,
we will generate a temporary symbol to denote the location of the label.
The symbol could be used to get DW_AT_low_pc afterwards. So, we create a
mapping between 'inlined label' and DBG_LABEL MachineInstr in DebugHandlerBase.
The DBG_LABEL in the mapping is used to query the symbol before it.
The AbstractLabels in DwarfCompileUnit is used to process labels in inlined
functions.
We also keep a mapping between scope and labels in DwarfFile to help to
generate correct tree structure of DIEs.
It also generates label debug information under global isel.
David Bolvansky [Tue, 31 Jul 2018 14:25:24 +0000 (14:25 +0000)]
Enrich inline messages
Summary:
This patch improves Inliner to provide causes/reasons for negative inline decisions.
1. It adds one new message field to InlineCost to report causes for Always and Never instances. All Never and Always instantiations must provide a simple message.
2. Several functions that used to return the inlining results as boolean are changed to return InlineResult which carries the cause for negative decision.
3. Changed remark priniting and debug output messages to provide the additional messages and related inline cost.
4. Adjusted tests for changed printing.
Andrea Di Biagio [Tue, 31 Jul 2018 14:23:49 +0000 (14:23 +0000)]
[llvm-mca] Remove README.txt
A detailed description of the tool has been recently added by Matt to
CommandGuide/llvm-mca.rst. File README.txt is now redundant and can be removed;
all the relevant user-guide information has been improved and then moved to
llvm-mca.rst.
In future, we should add another .rst for the "llvm-mca developer manual" to
provide infromation about:
- llvm-mca internals.
- How to add custom stages to the simulated pipeline.
- How to provide extra processor info in the scheduling model to improve the
analysis performed by llvm-mca.
John Brawn [Tue, 31 Jul 2018 14:19:29 +0000 (14:19 +0000)]
[MemDep] Use PhiValuesAnalysis to improve alias analysis results
This is being done in order to make GVN able to better optimize certain inputs.
MemDep doesn't use PhiValues directly, but does need to notifiy it when things
get invalidated.
[SLP] Fix PR38339: Instruction does not dominate all uses!
Summary:
If the ExtractElement instructions can be optimized out during the
vectorization and we need to reshuffle the parent vector, this
ShuffleInstruction may be inserted in the wrong place causing compiler
to produce incorrect code.
Matt Arsenault [Tue, 31 Jul 2018 13:34:31 +0000 (13:34 +0000)]
AMDGPU: Fold undef fcanonicalize to qNaN
We could choose a free 0 for this, but this
matches the behavior for fmul undef, 1.0. Also,
the NaN use is more useful for folding use operations
although if it's not eliminated it is more expensive
in terms of code size.
Peter Smith [Tue, 31 Jul 2018 13:24:49 +0000 (13:24 +0000)]
[ARM] Complete enumeration values for Tag_ABI_VFP_args
The LLD implementation of Tag_ABI_VFP_args needs to check the rarely seen
values of 3 (toolchain specific) and 4 compatible with both Base and VFP.
Add the missing enumeration values so that LLD can refer to them without
having to use the raw numbers.
Andrea Di Biagio [Tue, 31 Jul 2018 13:21:43 +0000 (13:21 +0000)]
[llvm-mca][BtVer2] Teach how to identify dependency-breaking idioms.
This patch teaches llvm-mca how to identify dependency breaking instructions on
btver2.
An example of dependency breaking instructions is the zero-idiom XOR (example:
`XOR %eax, %eax`), which always generates zero regardless of the actual value of
the input register operands.
Dependency breaking instructions don't have to wait on their input register
operands before executing. This is because the computation is not dependent on
the inputs.
Not all dependency breaking idioms are also zero-latency instructions. For
example, `CMPEQ %xmm1, %xmm1` is independent on
the value of XMM1, and it generates a vector of all-ones.
That instruction is not eliminated at register renaming stage, and its opcode is
issued to a pipeline for execution. So, the latency is not zero.
This patch adds a new method named isDependencyBreaking() to the MCInstrAnalysis
interface. That method takes as input an instruction (i.e. MCInst) and a
MCSubtargetInfo.
The default implementation of isDependencyBreaking() conservatively returns
false for all instructions. Targets may override the default behavior for
specific CPUs, and return a value which better matches the subtarget behavior.
In future, we should teach to Tablegen how to automatically generate the body of
isDependencyBreaking from scheduling predicate definitions. This would allow us
to expose the knowledge about dependency breaking instructions to the machine
schedulers (and, potentially, other codegen passes).
Peter Smith [Tue, 31 Jul 2018 13:03:54 +0000 (13:03 +0000)]
[ELF][ARM] Add Arm ABI names for float ABI ELF Header flags
The ELF for the Arm architecture document defines, for EF_ARM_EABI_VER5 and
above, the flags EF_ARM_ABI_FLOAT_HARD and EF_ARM_ABI_FLOAT_SOFT. These
have been defined to be compatible with the existing EF_ARM_VFP_FLOAT and
EF_ARM_SOFT_FLOAT used by gcc for EF_ARM_EABI_UNKNOWN.
This patch adds the flags in addition to the existing ones so that any code
depending on the old names will still work.
Jonas Paulsson [Tue, 31 Jul 2018 13:00:42 +0000 (13:00 +0000)]
[SystemZ] Improve decoding in case of instructions with four register operands.
Since z13, the max group size will be 2 if any μop has more than 3 register
sources.
This has been ignored sofar in the SystemZHazardRecognizer, but is now
handled by recognizing those instructions and adjusting the tracking of
decoding and the cost heuristic for grouping.
[InstCombine] simplify code for A & (A ^ B) --> A & ~B
This fold was written in an odd way and tried to avoid
an endless loop by bailing out on all constants instead
of the supposedly problematic case of -1. But (X & -1)
should always be simplified before we reach here, so I'm
not sure how that is a problem.
There were no tests for the commuted patterns, so I added
those at rL338364.
Martin Storsjo [Tue, 31 Jul 2018 09:27:01 +0000 (09:27 +0000)]
[ARM] Support the .inst directive for MachO and COFF targets
Contrary to ELF, we don't add any markers that distinguish data generated
with .short/.long from normal instructions, so the .inst directive only
adds compatibility with assembly that uses it.
Martin Storsjo [Tue, 31 Jul 2018 09:26:52 +0000 (09:26 +0000)]
[AArch64] Support the .inst directive for MachO and COFF targets
Contrary to ELF, we don't add any markers that distinguish data generated
with .long from normal instructions, so the .inst directive only adds
compatibility with assembly that uses it.
Diego Caballero [Tue, 31 Jul 2018 01:57:29 +0000 (01:57 +0000)]
[VPlan] Introduce VPLoopInfo analysis.
The patch introduces loop analysis (VPLoopInfo/VPLoop) for VPBlockBases.
This analysis will be necessary to perform some H-CFG transformations and
detect and introduce regions representing a loop in the H-CFG.
[X86] Stop accidentally running the Bonnell LEA fixup path on Goldmont.
In one place we checked X86Subtarget.slowLEA() to decide if the pass should run. But to decide what the pass should we only check isSLM. This resulted in Goldmont going down the Bonnell path.
[MS Demangler] Better demangling of template arguments.
This patch fixes demangling of template aliases as template-template
arguments, and also fixes function pointers and references as
not type template parameters. All of these can be properly
demangled now, so I've ported over the test
clang/test/CodeGenCXX/ms-template-callbacks.cpp. All of these
tests pass
projects / llvm / log