Summary:
Port rL265480, rL264754, rL265997 and rL266252 to SystemZ, in order to enable the Swift port on the architecture. SwiftSelf and SwiftError are assigned to R10 and R9, respectively, which are normally callee-saved registers. For more information, see:
RFC: Implementing the Swift calling convention in LLVM and Clang
https://groups.google.com/forum/#!topic/llvm-dev/epDd2w93kZ0
Reviewers: kbarton, manmanren, rjmccall, uweigand
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D19414
llvm-svn: 267823
For compilations with no explicit cpu specified, this exhibits
nice gains on Silvermont, with neutral performance on big cores.
Differential Revision: http://reviews.llvm.org/D19138
llvm-svn: 267809
The callseq_end node must be glued with the TLS calls, otherwise,
the generic code will miss the uses of the returned value and will
mark it dead.
Moreover, TLSCall 64-bit pseudo must not set an implicit-use on RDI,
the pseudo uses the symbol address at this point not RDI and the
lowering will do the right thing.
llvm-svn: 267797
transferSuccessors() would LoadCmpBB a successor of DoneBB,
whereas it should be a successor of the original MBB.
Follow-up to r266339.
Unfortunately, it's tricky to catch this in the verifier.
llvm-svn: 267779
transferSuccessors() would LoadCmpBB a successor of DoneBB, whereas
it should be a successor of the original MBB.
The testcase changes are caused by Thumb2SizeReduction, which
was previously confused by the broken CFG.
Follow-up to r266679.
Unfortunately, it's tricky to catch this in the verifier.
llvm-svn: 267778
Summary:
Currently the NVVMReflect pass is run at the beginning of our backend
passes. But really, it should be run as early as possible, as it's
simply resolving an "if" statement in code. So copy it into
TargetMachine::addEarlyAsPossiblePasses.
We still run it at the beginning of the backend passes, since it's
needed for correctness when lowering to nvptx.
(Specifically, NVVMReflect changes each call to the __nvvm_reflect
function or llvm.nvvm.reflect intrinsic into an integer constant, based
on the pass's configuration. Clearly we miss many optimization
opportunities if we perform this transformation at the beginning of
codegen.)
Reviewers: rnk
Subscribers: tra, llvm-commits, jholewinski
Differential Revision: http://reviews.llvm.org/D18616
llvm-svn: 267765
Added support of TTMP quads.
Reworked M0 exclusion machinery for SMRD and similar instructions
to enable usage of TTMP registers in those instructions as destinations.
Tests added.
Differential Revision: http://reviews.llvm.org/D19342
llvm-svn: 267733
Summary:
So it appears that to guarantee some of the ordering requirements of a GLSL
memoryBarrier() executed in the shader, we need to emit an s_waitcnt.
(We can't use an s_barrier, because memoryBarrier() may appear anywhere in
the shader, in particular it may appear in non-uniform control flow.)
Reviewers: arsenm, mareko, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D19203
llvm-svn: 267729
This change adds a new hook for estimating the cost of vector extracts followed
by zero- and sign-extensions. The motivating example for this change is the
SMOV and UMOV instructions on AArch64. These instructions move data from vector
to general purpose registers while performing the corresponding extension
(sign-extend for SMOV and zero-extend for UMOV) at the same time. For these
operations, TargetTransformInfo can assume the extensions are free and only
report the cost of the vector extract. The SLP vectorizer has been updated to
make use of the new hook.
Differential Revision: http://reviews.llvm.org/D18523
llvm-svn: 267725
Possibility to specify code of hardware register kept.
Disassemble to symbolic name, if name is known.
Tests updated/added.
Differential Revision: http://reviews.llvm.org/D19335
llvm-svn: 267724
We run after PEI, so we need to AddPristinesAndCSRs.
In practice, that makes no difference here, because we only ask about
liveness of super-registers of defined GR8/GR16 registers, so they
can't be pristine. Still, it's the correct thing to do.
Thanks to Quentin for noticing!
Follow-up to r267495.
llvm-svn: 267658
This effectively adds back the extractelt combine removed by r262358:
the direct case can still occur (because x86_mmx is special, see
r262446), but it's the indirect case that's now superseded by the
generic combine.
llvm-svn: 267651
the prologue.
Do not use basic blocks that have EFLAGS live-in as prologue if we need
to realign the stack. Realigning the stack uses AND instruction and this
clobbers EFLAGS.
An other alternative would have been to save and restore EFLAGS around
the stack realignment code, but this is likely inefficient.
Fixes PR27531.
llvm-svn: 267634
NVPTXLowerKernelArgs is required for correctness, so it should not be guarded
by CodeGenOpt::None.
NVPTXPeephole is optimization only, so it should be skipped when
CodeGenOpt::None.
llvm-svn: 267619
Support for SDWA instructions for VOP1 and VOP2 encoding.
Not done yet:
- converters for support optional operands and modifiers
- VOPC
- sext() modifier
- intrinsics
- VOP2b (see vop_dpp.s)
- V_MAC_F32 (see vop_dpp.s)
Differential Revision: http://reviews.llvm.org/D19360
llvm-svn: 267553
Handle MachineBasicBlock as a memory displacement operand in the LEA optimization pass.
Differential Revision: http://reviews.llvm.org/D19409
llvm-svn: 267551
print-stack-trace.cc test failure of compiler-rt has been fixed by
r266869 (http://reviews.llvm.org/D19148), so reenable sibling call
optimization on ppc64
Reviewers: nemanjai kbarton
llvm-svn: 267527
Summary:
We don't use MinLatency any more since r184032.
Reviewers: atrick, hfinkel, mcrosier
Differential Revision: http://reviews.llvm.org/D19474
llvm-svn: 267502
Kill-flags, which computeRegisterLiveness uses, are not reliable.
LivePhysRegs is.
Differential Revision: http://reviews.llvm.org/D19472
llvm-svn: 267495
The SparcV8 fneg and fabs instructions interestingly come only in a
single-float variant. Since the sign bit is always the topmost bit no
matter what size float it is, you simply operate on the high
subregister, as if it were a single float.
However, the layout of double-floats in the float registers is reversed
on little-endian CPUs, so that the high bits are in the second
subregister, rather than the first.
Thus, this expansion must check the endianness to use the correct
subregister.
llvm-svn: 267489
log2(Mask) is smaller than 32, we must use the 32-bit variant because the 64-bit
variant cannot encode it. Therefore, set the subreg part accordingly.
[AArch64] Fix optimizeCondBranch logic.
The opcode for the optimized branch does not depend on the size
of the activate bits in the AND masks, but the AND opcode itself.
Indeed, we need to use a X or W variant based on the AND variant
not based on whether the mask fits into the related variant.
Otherwise, we may end up using the W variant of the optimized branch
for 64-bit register inputs!
This fixes the last make check verifier issues for AArch64: PR27479.
llvm-svn: 267465
Use the operand for how long to wait. This is somewhat
distasteful, since it would be better to just emit s_nop
with the right argument in the first place. This would require
changing TII::insertNoop to emit N operands, which would be easy.
Slightly more problematic is the post-RA scheduler and hazard recognizer
represent nops as a single null node, and would require inventing
another way of representing N nops.
llvm-svn: 267456
Previously findClosestSuitableAluInstr was only considering the base register when checking the current instruction for suitability. Expand check to consider the offset if the offset is a register.
llvm-svn: 267424
Commit r266977 was reason for failing LLVM test suite with error message: fatal error: error in backend: Cannot select: t17: i32 = rotr t2, t11 ...
llvm-svn: 267418
Summary:
The expression is detected as a redundant expression.
Turn out, this is probably a bug.
```
/home/etienneb/llvm/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp:306:26: warning: both side of operator are equivalent [misc-redundant-expression]
if (isSMRD(*FirstLdSt) && isSMRD(*FirstLdSt)) {
```
Reviewers: rnk, tstellarAMD
Subscribers: arsenm, cfe-commits
Differential Revision: http://reviews.llvm.org/D19460
llvm-svn: 267415
Summary:
This patch adds support for the X asm constraint.
To do this, we lower the constraint to either a "w" or "r" constraint
depending on the operand type (both constraints are supported on ARM).
Fixes PR26493
Reviewers: t.p.northover, echristo, rengolin
Subscribers: joker.eph, jgreenhalgh, aemerson, rengolin, llvm-commits
Differential Revision: http://reviews.llvm.org/D19061
llvm-svn: 267411
ADD8TLS, a variant of add instruction used for initial-exec TLS,
currently accepts r0 as a source register. While add itself supports
r0 just fine, linker can relax it to a local-exec sequence, converting
it to addi - which doesn't support r0.
Differential Revision: http://reviews.llvm.org/D19193
llvm-svn: 267388
This corrects the MI annotations for the stack adjustment following the __chkstk
invocation. We were marking the original SP usage as a Def rather than Kill.
The (new) assigned value is the definition, the original reference is killed.
Adjust the ISelLowering to mark Kills and FrameSetup as well.
This partially resolves PR27480.
llvm-svn: 267361
We aren't currently making use of this in any successful mask decode and its actually incorrect as it inserts the wrong number of SM_SentinelUndef mask elements.
llvm-svn: 267350
Reused the ability to split constants of a type wider than the shuffle mask to work with masks generated from scalar constants transfered to xmm.
This fixes an issue preventing PSHUFB target shuffle masks decoding rematerialized scalar constants and also exposes the XOP VPPERM bug described in PR27472.
llvm-svn: 267343
This fixes PR22248 on s390x. The previous attempt at this was D19101,
which was before LOAD_STACK_GUARD existed. Compared to the previous
version, this always emits a rather ugly block of 4 instructions, involving
a thread pointer load that can't be shared with other potential users.
However, this is necessary for SSP - spilling the guard value (or thread
pointer used to load it) is counter to the goal, since it could be
overwritten along with the frame it protects.
Differential Revision: http://reviews.llvm.org/D19363
llvm-svn: 267340
The original patch caused crashes because it could derefence a null pointer
for SelectionDAGTargetInfo for targets that do not define it.
Evaluates fmul+fadd -> fmadd combines and similar code sequences in the
machine combiner. It adds support for float and double similar to the existing
integer implementation. The key features are:
- DAGCombiner checks whether it should combine greedily or let the machine
combiner do the evaluation. This is only supported on ARM64.
- It gives preference to throughput over latency: the heuristic used is
to combine always in loops. The targets decides whether the machine
combiner should optimize for throughput or latency.
- Supports for fmadd, f(n)msub, fmla, fmls patterns
- On by default at O3 ffast-math
llvm-svn: 267328
The option to control the emission of the new relocations
is -relax-relocations (blatantly copied from GNU as).
It can't be enabled by default because it breaks relatively
recent versions of ld.bfd/ld.gold (late 2015).
llvm-svn: 267307
The relative vtable ABI (PR26723) needs PLT relocations to refer to virtual
functions defined in other DSOs. The unnamed_addr attribute means that the
function's address is not significant, so we're allowed to substitute it
with the address of a PLT entry.
Also includes a bonus feature: addends for COFF image-relative references.
Differential Revision: http://reviews.llvm.org/D17938
llvm-svn: 267211
The opcode for the optimized branch does not depend on the size
of the activate bits in the AND masks, but the AND opcode itself.
Indeed, we need to use a X or W variant based on the AND variant
not based on whether the mask fits into the related variant.
Otherwise, we may end up using the W variant of the optimized branch
for 64-bit register inputs!
This fixes the last make check verifier issues for AArch64: PR27479.
llvm-svn: 267206
We used to simply set the kill flags to true when transforming a scalar
instruction to a vector one.
SrcScalar1 = copy SrcVector1
... = opScalar SrcScalar1
=>
SrcScalar1 = copy SrcVector1
... = opVector SrcVector1<kill>
This is obviously wrong. The proper update consists in:
1. Propagate the kill status from the copy to the new opVector
2. Reset the kill status on the copy, since the live-range of
SrcVector1 got extended.
This fixes some of the machine verifier errors for AArch64 with make check.
llvm-svn: 267180
- Switch few loops to range-based for loops
- Fix nop insertion at the end of BB
- Fix formatting
- Check for endpgm
Differential Revision: http://reviews.llvm.org/D19380
llvm-svn: 267167
Summary:
When generating assembly using -m16 we must explicitly mark it as
16-bit. Emit .code16 at beginning of file. Fixes wrong results when
using -fno-integrated-as.
Reviewers: dwmw2
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D19392
llvm-svn: 267152
When targetting MIPS64R6 some of the patterns for select were guarded by a
broken predicate. The predicate was supposed to test if a constant value
could fit in a 16 bit zero-extended field. Instead the value was tested to
fit in a 16 bit sign-extended field. For negative constants of native word
width this resulted in wrong code generation.
Reviewers: vkalintiris, dsanders
Differential Review: http://reviews.llvm.org/D19378
llvm-svn: 267151
Summary:
rL256194 transforms truncations between vectors of integers into PACKUS/PACKSS
operations during DAG combine. This generates better code for truncate, so cost
of truncate needs to be changed but looks like it got changed only in SSE2 table
Whereas this change is also applicable for SSE4.1, so the cost of truncate needs
to be changed for that as well. Cost of “TRUNCATE v16i32 to v16i8” & “TRUNCATE
v16i16 to v16i8” should be same in SSE4.1 & SSE2 table. Removing their cost from
SSE4.1, so it will fall back to SSE2.
Reviewers: Simon Pilgrim
llvm-svn: 267123
Specifically, itineraries for LEON processors has been added, along with several LEON processor Subtargets. Although currently all these targets are pretty much identical, support for features that will differ among these processors will be added in the very near future.
The different Instruction Itinerary Classes (IICs) added are sufficient to differentiate between the instruction timings used by LEON and, quite probably, by generic Sparc processors too, but the focus of the exercise has been for LEON processors, as the requirement of my project. If the IICs are not sufficient for other Sparc processor types and you want to add a new itinerary for one of those, it should be relatively trivial to adapt this.
As none of the LEON processors has Quad Floats, or is a Version 9 processor, none of those instructions have itinerary classes defined and revert to the default "NoItinerary" instruction itinerary.
Phabricator Review: http://reviews.llvm.org/D19359
llvm-svn: 267121
Summary:
This intrinsic returns true if the current thread belongs to a live pixel
and false if it belongs to a pixel that we are executing only for derivative
computation. It will be used by Mesa to implement gl_HelperInvocation.
Note that for pixels that are killed during the shader, this implementation
also returns true, but it doesn't matter because those pixels are always
disabled in the EXEC mask.
This unearthed a corner case in the instruction verifier, which complained
about a v_cndmask 0, 1, exec, exec<imp-use> instruction. That's stupid but
correct code, so make the verifier accept it as such.
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D19191
llvm-svn: 267102
Evaluates fmul+fadd -> fmadd combines and similar code sequences in the
machine combiner. It adds support for float and double similar to the existing
integer implementation. The key features are:
- DAGCombiner checks whether it should combine greedily or let the machine
combiner do the evaluation. This is only supported on ARM64.
- It gives preference to throughput over latency: the heuristic used is
to combine always in loops. The targets decides whether the machine
combiner should optimize for throughput or latency.
- Supports for fmadd, f(n)msub, fmla, fmls patterns
- On by default at O3 ffast-math
llvm-svn: 267098
WIN__DBZCHK will insert a CBZ instruction into the stream. This instruction
reserves 3 bits for the condition register (rn). As such, we must ensure that
we restrict the register to a low register. Use the tGPR class instead of GPR
to ensure that this is properly constrained. In debug builds, we would attempt
to use lr as a condition register which would silently get truncated with no
hint that the register selection was incorrect.
llvm-svn: 267080
We'd disabled them on x86 because back in the early days some host tools
couldn't handle the new load commands. This no longer holds: anyone capable of
deploying Clang should be able to deploy its copies of ar/ranlib/etc.
rdar://25254790
llvm-svn: 267075
Summary:
IntrReadWriteArgMem simply becomes IntrArgMemOnly.
So there are fewer intrinsic properties that express their orthogonality
better, and correspond more closely to the corresponding IR attributes.
Suggested by: Philip Reames
Reviewers: joker.eph, reames, tstellarAMD
Subscribers: jholewinski, arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D19291
llvm-svn: 267021
AArch64InstrInfo::optimizeCompareInstr has bug PR27158 which causes generation of incorrect code.
A compare instruction is substituted with another instruction which does not
produce the same flags as the original compare instruction.
This patch contains:
1. Fix of the bug.
2. A regression test in MIR.
3. A new test to check that SUBS is replaced by SUB.
Differential Revision: http://reviews.llvm.org/D18838
llvm-svn: 266969
CTTZ_ZERO_UNDEF can be custom lowered specially if CTLZ is supported. Otherwise CTTZ and CTTZ_ZERO_UNDEF are handled the same way by using CTPOP and bitmath.
llvm-svn: 266952
Differentiate between word and subword memory operations as they take different
amount of cycles to complete. This just adds a basic model of the subword
latency to the scheduler.
llvm-svn: 266898
Add ParseAMDGPURegister which can be invoked recursively for parsing lists.
Rename getRegForName to getSpecialRegForName.
Support legacy SP3 register list syntax: [s2,s3,s4,s5] or [flat_scratch_lo,flat_scratch_hi].
Add 64-bit registers TBA, TMA where missing.
Add some tests.
Differential Revision: http://reviews.llvm.org/D19163
llvm-svn: 266865
Because lowering of CMP_SWAP_64 occurs during type legalization, there can be
i64 types produced by more than just a BUILD_PAIR or similar. My initial tests
used just incoming function args.
llvm-svn: 266828
Summary:
This property is used to mark an intrinsic that only writes to memory, but
neither reads from memory nor has other side effects.
An example where this is useful is the llvm.amdgcn.buffer.store.format.*
intrinsic, which corresponds to a store instruction that goes through a special
buffer descriptor rather than through a plain pointer.
With this property, the intrinsic should still be handled as having side
effects at the LLVM IR level, but machine scheduling can make smarter
decisions.
Reviewers: tstellarAMD, arsenm, joker.eph, reames
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D18291
llvm-svn: 266826
Summary:
The added testcase, which triggered this, was derived from a shader-db case
via bugpoint. A separate question is why scalar branching wasn't used.
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D19208
llvm-svn: 266825
Summary:
A shader stored the live mask (initial exec mask) in an SGPR which was then
spilled during register allocation. The allocator quite reasonably
optimized turned the spill into
v_writelane_b32 %vgpr, exec_lo, N
v_writelane_b32 %vgpr, exec_hi, N+1
at the beginning of the shader, confusing the SGPR accounting.
No test case, because si-sgpr-spill.ll together with an upcoming patch for
WQM handling exhibits the problem.
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D19199
llvm-svn: 266824
Both AArch64 and ARM support llvm.<arch>.thread.pointer intrinsics that
just return the thread pointer. I have a pending patch that does the same
for SystemZ (D19054), and there are many more targets that could benefit
from one.
This patch merges the ARM and AArch64 intrinsics into a single target
independent one that will also be used by subsequent targets.
Differential Revision: http://reviews.llvm.org/D19098
llvm-svn: 266818
With this change, ideally IR pass can always generate llvm.stackguard
call to get the stack guard; but for now there are still IR form stack
guard customizations around (see getIRStackGuard()). Future SSP
customization should go through LOAD_STACK_GUARD.
There is a behavior change: stack guard values are not CSEed anymore,
since we should never reuse the value in case that it has been spilled (and
corrupted). See ssp-guard-spill.ll. This also cause the change of stack
size and codegen in X86 and AArch64 test cases.
Ideally we'd like to know if the guard created in llvm.stackprotector() gets
spilled or not. If the value is spilled, discard the value and reload
stack guard; otherwise reuse the value. This can be done by teaching
register allocator to know how to rematerialize LOAD_STACK_GUARD and
force a rematerialization (which seems hard), or check for spilling in
expandPostRAPseudo. It only makes sense when the stack guard is a global
variable, which requires more instructions to load. Anyway, this seems to go out
of the scope of the current patch.
llvm-svn: 266806
* Add lowering for SETCCE i32.
* Add test to check lowering of i64 compares uses SETCCE expansion (outside of EQ and NE).
* Fix select.ll test and immediate form selection for RI operations.
llvm-svn: 266802
in preparation for enabling the outgoing parameter store-to-push optimization
for 64-bit targets.
Differential Revision: http://reviews.llvm.org/D19222
llvm-svn: 266774
Using VPERMQ/VPERMPD allows memory folding of the (repeated) input where VINSERTI128/VINSERTF128 can not.
Differential Revision: http://reviews.llvm.org/D19228
llvm-svn: 266728
Summary:
The `"patchable-function"` attribute can be used by an LLVM client to
influence LLVM's code generation in ways that makes the generated code
easily patchable at runtime (for instance, to redirect control).
Right now only one patchability scheme is supported,
`"prologue-short-redirect"`, but this can be expanded in the future.
Reviewers: joker.eph, rnk, echristo, dberris
Subscribers: joker.eph, echristo, mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D19046
llvm-svn: 266715
The fast register-allocator cannot cope with inter-block dependencies without
spilling. This is fine for ldrex/strex loops coming from atomicrmw instructions
where any value produced within a block is dead by the end, but not for
cmpxchg. So we lower a cmpxchg at -O0 via a pseudo-inst that gets expanded
after regalloc.
Fortunately this is at -O0 so we don't have to care about performance. This
simplifies the various axes of expansion considerably: we assume a strong
seq_cst operation and ensure ordering via the always-present DMB instructions
rather than v8 acquire/release instructions.
Should fix the 32-bit part of PR25526.
llvm-svn: 266679
There's currently no raw_ostream &operator<<(SimpleValueType); provided by LLVM. It could be added by refactoring utils/TableGen/CodeGenTarget.cpp:getEnumName, but that's much more work than fixing the build.
llvm-svn: 266627
Also,
- Skip pass if machine module does not have debug info
- Minor comment changes
- Added test
Differential Revision: http://reviews.llvm.org/D19079
llvm-svn: 266626
Order should match the sp3 syntax, where destination (simm16 denoting the hwreg) is coming first.
Differential Revision: http://reviews.llvm.org/D19161
llvm-svn: 266617
Summary:
When clang is given -save-temps or -via-file-asm, any inline assembly in
the source is parsed twice. Once by the compiler, and again by the
assembler. We must take care to ensure that this doesn't lead to
double-filling delay slots.
Reviewers: sdardis, vkalintiris
Subscribers: dsanders, sdardis, llvm-commits
Differential Revision: http://reviews.llvm.org/D19166
llvm-svn: 266608
Summary:
This will allows us to eliminate some magic numbers from the offset operand of
branch instructions in favour of symbols and makes it possible to avoid
double-filling delay slots when clang is given -save-temps.
parseDirectiveCpRestore() is calling isIntegratedAssemblerRequired() for the
moment since correctly pushing the generation of these instructions into the
ELF target streamer is tricky enough to warrant a separate patch.
Reviewers: sdardis, vkalintiris
Subscribers: dsanders, llvm-commits, sdardis
Differential Revision: http://reviews.llvm.org/D19164
llvm-svn: 266602
Removed some unused headers, replaced some headers with forward class declarations.
Found using simple scripts like this one:
clear && ack --cpp -l '#include "llvm/ADT/IndexedMap.h"' | xargs grep -L 'IndexedMap[<]' | xargs grep -n --color=auto 'IndexedMap'
Patch by Eugene Kosov <claprix@yandex.ru>
Differential Revision: http://reviews.llvm.org/D19219
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266595
This required changing several places to print VT enums as strings instead of raw ints since the proper method to use to print became ambiguous. This is probably an improvement anyway.
This also appears to save ~8K from an x86 self host build of llc.
llvm-svn: 266562
no functional change.
ExtraLoad and WrapperKind are been used only if (OpFlags == X86II::MO_GOTPCREL).
Differential Revision: http://reviews.llvm.org/D18942
llvm-svn: 266557
This resolves more frame indexes early and folds
the immediate offsets into the scratch mubuf instructions.
This cleans up a lot of the mess that's currently emitted,
such as emitting add 0s and repeatedly initializing the same
register to 0 when spilling.
llvm-svn: 266508
Use the tryAddingSymbolicOperand callback to attempt to present immediate
values in symbolic form when disassembling. This is currently only used
for PC-relative immediates (which are most likely to be symbolic in the
SystemZ ISA). Add new DecodeMethod types to allow distinguishing between
branch and non-branch instructions.
llvm-svn: 266469
Divisions by a constant can be converted into multiplies which are usually
cheaper, but this isn't possible if the constant gets separated (particularly
in loops). Fix this by telling ConstantHoisting that the immediate in a DIV is
cheap.
I considered making the check generic, but neither AArch64 (strangely) nor x86
showed any benefit on the tests I had.
llvm-svn: 266464
This improves AA in the MI schduler when reason about paired instructions.
Phabricator Revision: http://reviews.llvm.org/D17098
PR26358
llvm-svn: 266462
Summary:
Without MMOs, the callee-save load/store instructions were treated as
volatile by the MI post-RA scheduler and AArch64LoadStoreOptimizer.
Reviewers: t.p.northover, mcrosier
Subscribers: aemerson, rengolin, mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D17661
llvm-svn: 266439
[PPC] Previously when casting generic loads to LXV2DX/ST instructions we
would leave the original load return type in place allowing for an
assertion failure when we merge two equivalent LXV2DX nodes with
different types.
This fixes PR27350.
Reviewers: nemanjai
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D19133
llvm-svn: 266438
Perform store clustering just like load clustering. This change add
StoreClusterMutation in machine-scheduler. To control StoreClusterMutation,
added enableClusterStores() in TargetInstrInfo.h. This is enabled only on
AArch64 for now.
This change also add support for unscaled stores which were not handled in
getMemOpBaseRegImmOfs().
llvm-svn: 266437
Summary:
In the added test-case, the atomic instruction feeds into a non-machine
CopyToReg node which hasn't been selected yet, so guard against
non-machine opcodes here.
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D19043
llvm-svn: 266433
Summary:
Calls on NVPTX are unusually expensive (for one thing, lots of state
needs to be saved to memory, which is slow), so make the inlininer much
more aggressive.
Reviewers: chandlerc
Subscribers: jholewinski, llvm-commits, tra
Differential Revision: http://reviews.llvm.org/D18561
llvm-svn: 266406
At the same time, fixes InstructionsTest::CastInst unittest: yes
you can leave the IR in an invalid state and exit when you don't
destroy the context (like the global one), no longer now.
This is the first part of http://reviews.llvm.org/D19094
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266379
Currently what comes out of instruction selection is a
register initialized to -1, and then copied to m0.
MachineCSE doesn't consider copies, but we want these
to be CSEed. This isn't much of a problem currently,
because SIFoldOperands is run immediately after.
This avoids regressions when SIFoldOperands is run later
from leaving all copies to m0.
llvm-svn: 266377
Some SIMD implementations are not IEEE-754 compliant, for example ARM's NEON.
This patch teaches the loop vectorizer to only allow transformations of loops
that either contain no floating-point operations or have enough allowance
flags supporting lack of precision (ex. -ffast-math, Darwin).
For that, the target description now has a method which tells us if the
vectorizer is allowed to handle FP math without falling into unsafe
representations, plus a check on every FP instruction in the candidate loop
to check for the safety flags.
This commit makes LLVM behave like GCC with respect to ARM NEON support, but
it stops short of fixing the underlying problem: sub-normals. Neither GCC
nor LLVM have a flag for allowing sub-normal operations. Before this patch,
GCC only allows it using unsafe-math flags and LLVM allows it by default with
no way to turn it off (short of not using NEON at all).
As a first step, we push this change to make it safe and in sync with GCC.
The second step is to discuss a new sub-normal's flag on both communitues
and come up with a common solution. The third step is to improve the FastMath
flags in LLVM to encode sub-normals and use those flags to restrict NEON FP.
Fixes PR16275.
llvm-svn: 266363
Summary:
This adds the necessary target code to be able to run the ir translator.
Lowering function arguments and returns is a nop and there is no support
for RegBankSelect.
Reviewers: arsenm, qcolombet
Subscribers: arsenm, joker.eph, vkalintiris, llvm-commits
Differential Revision: http://reviews.llvm.org/D19077
llvm-svn: 266356
MachineInstr.h and MachineInstrBuilder.h are very popular headers,
widely included across all LLVM backends. It turns out that there only a
handful of TUs that actually care about DI operands on MachineInstrs.
After this change, touching DebugInfoMetadata.h and rebuilding llc only
needs 112 actions instead of 542.
llvm-svn: 266351
Summary:
This fully solves the problem where the StructurizeCFG pass does not
consider the same branches as uniform as the SIAnnotateControlFlow pass.
The patch in D19013 helps with this problem, but is not sufficient
(and, interestingly, causes a "regression" with one of the existing
test cases).
No tests included here, because tests in D19013 already cover this.
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D19018
llvm-svn: 266346
Summary:
This pass is unnecessary and overly conservative. It was motivated by
situations like
def %vreg0:SGPR_32
...
if-block:
..
def %vreg1:SGPR_32
...
else-block:
...
use %vreg0:SGPR_32
...
and similar situations with uses after the non-uniform control flow, where
we are not allowed to assign %vreg0 and %vreg1 to the same physical register,
even though in the original, thread/workitem-based CFG, it looks like the
live ranges of these registers do not overlap.
However, by the time register allocation runs, we have moved to a wave-based
CFG that accurately represents the fact that the wave may run through both
the if- and the else-block. So the live ranges of %vreg0 and %vreg1 already
overlap even without the SIFixSGPRLiveRanges pass.
In addition to proving this change correct, I have tested it with Piglit
and a small number of other tests.
Reviewers: arsenm, tstellarAMD
Subscribers: MatzeB, arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D19041
llvm-svn: 266345
Summary:
I've been carrying this change around with me for a while, because the if ()
managed to confuse me while following the code. All callers ensure that the
assertion holds.
Reviewers: arsenm, tstellarAMD
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D19042
llvm-svn: 266344
FastRegAlloc works only at the basic-block level and spills all live-out
registers. Unfortunately for a stack-based cmpxchg near the spill slots, this
can perpetually clear the exclusive monitor, which means the cmpxchg will never
succeed.
I believe the only way to handle this within LLVM is by expanding the loop
post-regalloc. We don't want this in general because it severely limits the
optimisations that can be done, so we limit this to -O0 compilations.
It's an ugly hack, and about the one good point in the whole mess is that we
can treat all cmpxchg operations in the most naive way possible (seq_cst, no
clrex faff) without affecting correctness.
Should fix PR25526.
llvm-svn: 266339
Summary:
For GL_ARB_compute_shader we need to support workgroup sizes of at least 1024. However, if we want to allow large workgroup sizes, we may need to use less registers, as we have to run more waves per SIMD.
This patch adds an attribute to specify the maximum work group size the compiled program needs to support. It defaults, to 256, as that has no wave restrictions.
Reducing the number of registers available is done similarly to how the registers were reserved for chips with the sgpr init bug.
Reviewers: mareko, arsenm, tstellarAMD, nhaehnle
Subscribers: FireBurn, kerberizer, llvm-commits, arsenm
Differential Revision: http://reviews.llvm.org/D18340
Patch By: Bas Nieuwenhuizen
llvm-svn: 266337
Summary:
The code previously always used s1 as it was using the user + system SGPR
information for compute kernels. This is incorrect for Mesa shaders though,
The register should be the next SGPR after all user and system SGPR's.
We use that Mesa adds arguments for all input and system SGPR's and
take the next available SGPR for the scratch wave offset register.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewers: mareko, arsenm, nhaehnle, tstellarAMD
Subscribers: qcolombet, arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D18941
Patch By: Bas Nieuwenhuizen
llvm-svn: 266336
Alias 'jic $reg, 0' to 'jrc $reg' and 'jialc $reg, 0' to 'jalrc $reg' like
binutils.
This patch was previous committed as r266055 as seemed to have caused some spurious
test failures. They did not reappear after further local testing.
llvm-svn: 266301
This code was creating a new type in the global context, regardless
of which context the user is sitting in, what can possibly go wrong?
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266275
At some point, ARM stopped getting any benefit from ConstantHoisting because
the pass called a different variant of getIntImmCost. Reimplementing the
correct variant revealed some problems, however:
+ ConstantHoisting was modifying switch statements. This is simply invalid,
the cases must remain integer constants no matter the notional cost.
+ ConstantHoisting was mangling alloca instructions in the entry block. These
should be handled by FrameLowering, so constants actually have a cost of 0.
Worse, the resulting bitcasts meant they became dynamic allocas.
rdar://25707382
llvm-svn: 266260
It is very likely that the swiftself parameter is alive throughout most
functions function so putting it into a callee save register should
avoid spills for the callers with only a minimum amount of extra spills
in the callees.
Currently the generated code is correct but unnecessarily spills and
reloads arguments passed in callee save registers, I will address this
in upcoming patches.
This also adds a missing check that for tail calls the preserved value
of the caller must be the same as the callees parameter.
Differential Revision: http://reviews.llvm.org/D18901
llvm-svn: 266253
It is very likely that the swiftself parameter is alive throughout most
functions function so putting it into a callee save register should
avoid spills for the callers with only a minimum amount of extra spills
in the callees.
Currently the generated code is correct but unnecessarily spills and
reloads arguments passed in callee save registers, I will address this
in upcoming patches.
This also adds a missing check that for tail calls the preserved value
of the caller must be the same as the callees parameter.
Differential Revision: http://reviews.llvm.org/D18902
llvm-svn: 266252
It is very likely that the swiftself parameter is alive throughout most
functions function so putting it into a callee save register should
avoid spills for the callers with only a minimum amount of extra spills
in the callees.
Currently the generated code is correct but unnecessarily spills and
reloads arguments passed in callee save registers, I will address this
in upcoming patches.
This also adds a missing check that for tail calls the preserved value
of the caller must be the same as the callees parameter.
Differential Revision: http://reviews.llvm.org/D19007
llvm-svn: 266251
