Commit Graph

21020 Commits

Author SHA1 Message Date
Matthias Braun 08f4704ec8 IfConversion: Use references instead of pointers where possible; NFC
Also put some commonly used subexpressions into variables.

llvm-svn: 278895
2016-08-17 02:52:01 +00:00
Matthias Braun b1e0558df4 IfConversion: Use range based for; NFC
Also avoid some pointless use of auto! Because that's friendlier to
readers and avoids several types accidentally resolving to unnecessary
references here (MachineInstr *&, unsigned &).

llvm-svn: 278894
2016-08-17 02:51:59 +00:00
Matthias Braun 2c931798d6 IfConversion: Improve doxygen comments
llvm-svn: 278893
2016-08-17 02:51:57 +00:00
Duncan P. N. Exon Smith 00ec93da26 CodeGen: Avoid dereferencing end() in OptimizePHIs::OptimizeBB
llvm-svn: 278879
2016-08-17 00:43:59 +00:00
Duncan P. N. Exon Smith dcbce9c391 CodeGen: Avoid dereferencing end() when unconstifying iterators
Rather than doing a funny dance that relies on dereferencing end() not
crashing, add some API to MachineInstrBundleIterator to get a non-const
version of the iterator.

llvm-svn: 278870
2016-08-16 23:34:07 +00:00
Kyle Butt 07d61425e3 Codegen: Don't tail-duplicate blocks with un-analyzable fallthrough.
If AnalyzeBranch can't analyze a block and it is possible to
fallthrough, then duplicating the block doesn't make sense, as only one
block can be the layout predecessor for the un-analyzable fallthrough.

Submitted wit a test case, but NOTE: the test case doesn't currently
fail. However, the test case fails with D20505 and would have saved me
some time debugging.

llvm-svn: 278866
2016-08-16 22:56:14 +00:00
Duncan P. N. Exon Smith 41cf73ce16 CodeGen: Don't dereference end() in MachineBasicBlock::CorrectExtraCFGEdges
The current MachineBasicBlock might be the last block, so FallThru may
be past the end().  Use getNextNode(), which will convert to nullptr,
rather than &*++, which is invalid if we reach the end().

llvm-svn: 278858
2016-08-16 21:46:03 +00:00
Matt Arsenault b8037a1bd3 TailDuplicator: Use range loops
llvm-svn: 278847
2016-08-16 20:38:05 +00:00
Sjoerd Meijer 15c81b05ea [MBP] do not reorder and move up loop latch block
Do not reorder and move up a loop latch block before a loop header
when optimising for size because this will generate an extra 
unconditional branch.

Differential Revision: https://reviews.llvm.org/D22521

llvm-svn: 278840
2016-08-16 19:50:33 +00:00
Wolfgang Pieb 8df58f48dd When the inline spiller rematerializes an instruction, take the debug location from the instruction
that immediately follows the rematerialization point.

Patch by Andrea DiBiagio.

Differential Revision: http://reviews.llvm.org/D23539

llvm-svn: 278822
2016-08-16 17:12:50 +00:00
Ahmed Bougacha e4c03abddd [AArch64][GlobalISel] Select G_MUL.
llvm-svn: 278810
2016-08-16 14:37:46 +00:00
Brendon Cahoon 65b6ebccad [Pipeliner] Fix an asssert due to invalid Phi in the epilog
The pipeliner was generating an invalid Phi name for an operand
in the epilog block, which caused an assert in the live variable
analysis pass. The fix is to the code that generates new Phis
in the epilog block. In this case, there is an existing Phi that
needs to be reused rather than creating a new Phi instruction.

Differential Revision: https://reviews.llvm.org/D23513

llvm-svn: 278805
2016-08-16 14:29:24 +00:00
Ahmed Bougacha 0306b5ef07 [AArch64][GlobalISel] Select p0 G_FRAME_INDEX.
And mark it as legal.

llvm-svn: 278802
2016-08-16 14:02:42 +00:00
Pierre Gousseau 051db7d838 [x86] Refactor a PowerPC specific ctlz/srl transformation (NFC).
Following the discussion on D22038, this refactors a PowerPC specific setcc -> srl(ctlz) transformation so it can be used by other targets.

Differential Revision: https://reviews.llvm.org/D23445

llvm-svn: 278799
2016-08-16 13:53:53 +00:00
Tim Shen e0793db41d [ADT] Change PostOrderIterator to use NodeRef. NFC.
Reviewers: dblaikie

Subscribers: mzolotukhin, llvm-commits

Differential Revision: https://reviews.llvm.org/D23522

llvm-svn: 278752
2016-08-15 21:52:54 +00:00
Eli Friedman 98151d6440 Fix typo in lowering for fp128 ueq.
Regression from r259791.

Differential Revision: https://reviews.llvm.org/D23374

llvm-svn: 278750
2016-08-15 21:46:19 +00:00
Tim Northover 28fdc4272d GlobalISel: support loads and stores of strange types.
Before we mischaracterized structs and i1 types as a scalar with size 0 in
various ways.

llvm-svn: 278744
2016-08-15 21:13:17 +00:00
Wolfgang Pieb dfad9b20c9 Local variables whose address is taken and passed on to a call are described
in debug info using their stack slots instead of as an indirection of param reg + 0
offset. This is done by detecting FrameIndexSDNodes in SelectionDAG and generating
FrameIndexDbgValues for them. This ultimately generates DBG_VALUEs with stack
location operands.

Differential Revision: http://reviews.llvm.org/D23283

llvm-svn: 278703
2016-08-15 18:18:26 +00:00
Sjoerd Meijer 58156715b4 MachineLoop: add methods findLoopControlBlock and findLoopPreheader
This adds two new utility functions findLoopControlBlock and findLoopPreheader
to MachineLoop and MachineLoopInfo. These functions are refactored and taken
from the Hexagon target as they are target independent; thus this is intendend to
be a non-functional change.

Differential Revision: https://reviews.llvm.org/D22959

llvm-svn: 278661
2016-08-15 08:22:42 +00:00
Diana Picus 68be1eb885 Revert "CodeGen: If Convert blocks that would form a diamond when tail-merged."
This reverts commit r278287.

This commit broke the clang-cmake-thumbv7-a15-full-sh bot.
See https://llvm.org/bugs/show_bug.cgi?id=28949

llvm-svn: 278621
2016-08-14 02:10:18 +00:00
Diana Picus 35ccf53e75 Revert "Codegen: Don't tail-duplicate blocks with un-analyzable fallthrough."
This reverts commit r278288.

r278287 broke the clang-cmake-thumbv7-a15-full-sh bot.
Revert this so we can get to r278287.

llvm-svn: 278620
2016-08-14 02:10:12 +00:00
Justin Lebar cf56e92c50 Minor comment fix ("generate" --> "generates").
llvm-svn: 278578
2016-08-12 23:58:19 +00:00
Haicheng Wu 7c4535d1e7 Reapply [BranchFolding] Restrict tail merging loop blocks after MBP
Fixed a bug in the test case.

To fix PR28104, this patch restricts tail merging to blocks that belong to the
same loop after MBP.

llvm-svn: 278575
2016-08-12 23:13:38 +00:00
Haicheng Wu d9cbb1608f Revert "[BranchFolding] Restrict tail merging loop blocks after MBP"
This reverts commit r278463 because it hits the bot.

llvm-svn: 278484
2016-08-12 08:40:24 +00:00
Duncan P. N. Exon Smith f197b1f78f ADT: Remove all ilist_iterator => pointer casts, NFC
Remove all ilist_iterator to pointer casts.  There were two reasons for
casts:

  - Checking for an uninitialized (i.e., null) iterator.  I added
    MachineInstrBundleIterator::isValid() to check for that case.

  - Comparing an iterator against the underlying pointer value while
    avoiding converting the pointer value to an iterator.  This is
    occasionally necessary in MachineInstrBundleIterator, since there is
    an assertion in the constructors that the underlying MachineInstr is
    not bundled (but we don't care about that if we're just checking for
    pointer equality).

To support the latter case, I rewrote the == and != operators for
ilist_iterator and MachineInstrBundleIterator.

  - The implicit constructors now use enable_if to exclude
    const-iterator => non-const-iterator conversions from overload
    resolution (previously it was a compiler error on instantiation, now
    it's SFINAE).

  - The == and != operators are now global (friends), and are not
    templated.

  - MachineInstrBundleIterator has overloads to compare against both
    const_pointer and const_reference.  This avoids the implicit
    conversions to MachineInstrBundleIterator that assert, instead just
    checking the address (and I added unit tests to confirm this).

Notably, the only remaining uses of ilist_iterator::getNodePtrUnchecked
are in ilist.h, and no code outside of ilist*.h directly relies on this
UB end-iterator-to-pointer conversion anymore.  It's still needed for
ilist_*sentinel_traits, but I'll clean that up soon.

llvm-svn: 278478
2016-08-12 05:05:36 +00:00
David Majnemer c700490f48 Use the range variant of remove_if instead of unpacking begin/end
No functionality change is intended.

llvm-svn: 278475
2016-08-12 04:32:37 +00:00
David Majnemer 42531260b3 Use the range variant of find/find_if instead of unpacking begin/end
If the result of the find is only used to compare against end(), just
use is_contained instead.

No functionality change is intended.

llvm-svn: 278469
2016-08-12 03:55:06 +00:00
Haicheng Wu ea02372059 [BranchFolding] Restrict tail merging loop blocks after MBP
To fix PR28014, this patch restricts tail merging to blocks that belong to the
same loop after MBP.

Differential Revision: https://reviews.llvm.org/D23191

llvm-svn: 278463
2016-08-12 03:30:23 +00:00
David Majnemer 562e82945e Use the range variant of find_if instead of unpacking begin/end
No functionality change is intended.

llvm-svn: 278443
2016-08-12 00:18:03 +00:00
David Majnemer 0d955d0bf5 Use the range variant of find instead of unpacking begin/end
If the result of the find is only used to compare against end(), just
use is_contained instead.

No functionality change is intended.

llvm-svn: 278433
2016-08-11 22:21:41 +00:00
Tim Northover 8e0c53a018 GlobalISel: support 'null' constant in translation.
It's sharing the integer G_CONSTANT for now since I don't *think* it creates
any ambiguity (even on weird archs). If that turns out wrong we can create a
G_PTRCONSTANT or something.

llvm-svn: 278423
2016-08-11 21:40:55 +00:00
David Majnemer 0a16c22846 Use range algorithms instead of unpacking begin/end
No functionality change is intended.

llvm-svn: 278417
2016-08-11 21:15:00 +00:00
Duncan P. N. Exon Smith 38eea4a76f CodeGen: Avoid dereferencing end() in MachineScheduler
Check MachineInstr::isDebugValue for the same instruction as we're
calling isSchedBoundary, avoiding the possibility of dereferencing
end().

This is a functionality change even when I!=end().  Matthias had a look
and agrees this is the right resolution (as opposed to checking for
end()).

This is triggered by a huge number of tests, but they happen to
magically pass right now.  I found this because WIP patches for PR26753
convert them into crashes.

llvm-svn: 278394
2016-08-11 20:03:09 +00:00
Krzysztof Parzyszek a003b76391 If-conversion incorrectly calculates liveness of redefined registers
Differential Revision: https://reviews.llvm.org/D23207

llvm-svn: 278383
2016-08-11 18:42:06 +00:00
Dominic Chen 6ba19659cb Improve virtual register handling when computing debug information
Summary: Some backends, like WebAssembly, use virtual registers instead of physical registers. This crashes the DbgValueHistoryCalculator pass, which assumes that all registers are physical. Instead, skip virtual registers when iterating aliases, and assume that they are clobbered.

Reviewers: dexonsmith, dschuff, aprantl

Subscribers: yurydelendik, llvm-commits, jfb, sunfish

Differential Revision: https://reviews.llvm.org/D22590

llvm-svn: 278371
2016-08-11 17:52:40 +00:00
Michael Kuperstein e36d7716c3 Make TwoAddressInstructionPass::rescheduleMIBelowKill subreg-aware
This fixes PR28824.

Differential Revision: https://reviews.llvm.org/D23220

llvm-svn: 278370
2016-08-11 17:38:33 +00:00
Eugene Zelenko cdc7161281 Fix some Clang-tidy modernize and Include What You Use warnings.
Differential revision: https://reviews.llvm.org/D23291

llvm-svn: 278364
2016-08-11 17:20:18 +00:00
Tim Northover 0d51044b69 GlobalISel: clear vreg mapping after translating each function
Otherwise we only materialize (shared) constants in the first function they
appear in. This doesn't go well.

llvm-svn: 278351
2016-08-11 16:21:29 +00:00
Reid Kleckner 26f9e9ebc3 Remove FIXME about asserting on the end iterator
After machine block placement, MBBs may not have terminators, and it is
appropriate to check for the end iterator here. We can fold the check
into the next if, as well. This look is really just looking for BBs that
end in CATCHRET.

llvm-svn: 278350
2016-08-11 16:00:43 +00:00
Duncan P. N. Exon Smith 2e7af979b9 CodeGen: Check for a terminator in llvm::getFuncletMembership
Check for an end iterator from MachineBasicBlock::getFirstTerminator in
llvm::getFuncletMembership.  If this is turned into an assertion, it
fires in 48 X86 testcases (for example,
CodeGen/X86/regalloc-spill-at-ehpad.ll).

Since this is likely a latent bug (shouldn't all basic blocks end with a
terminator?) I've filed PR28938.

llvm-svn: 278344
2016-08-11 15:29:02 +00:00
Marina Yatsina 88f0c31f13 Avoid false dependencies of undef machine operands
This patch helps avoid false dependencies on undef registers by updating the machine instructions' undef operand to use a register that the instruction is truly dependent on, or use a register with clearance higher than Pref.

Pseudo example:

loop:
xmm0 = ...
xmm1 = vcvtsi2sdl eax, xmm0<undef>
... = inst xmm0
jmp loop

In this example, selecting xmm0 as the undef register creates false dependency between loop iterations.
This false dependency cannot be solved by inserting an xor before vcvtsi2sdl because xmm0 is alive at the point of the vcvtsi2sdl instruction.
Selecting a different register instead of xmm0, especially a register that is not used in the loop, will eliminate this problem.

Differential Revision: https://reviews.llvm.org/D22466

llvm-svn: 278321
2016-08-11 07:32:08 +00:00
Tim Northover 357f1be2ca GlobalISel: support same ConstantExprs as Instructions.
It's more than just inttoptr, but the others can't be tested until we have
support for non-trivial constants (they currently get unavoidably folded to a
ConstantInt).

llvm-svn: 278303
2016-08-10 23:02:41 +00:00
Tim Northover 406024a108 GlobalISel: implement simple function calls on AArch64.
We're still limited in the arguments we support, but this at least handles the
basic cases.

llvm-svn: 278293
2016-08-10 21:44:01 +00:00
Kyle Butt 81d32846b0 Codegen: Don't tail-duplicate blocks with un-analyzable fallthrough.
If AnalyzeBranch can't analyze a block and it is possible to
fallthrough, then duplicating the block doesn't make sense, as only one
block can be the layout predecessor for the un-analyzable fallthrough.

Submitted wit a test case, but NOTE: the test case doesn't currently
fail. However, the test case fails with D20505 and would have saved me
some time debugging.

llvm-svn: 278288
2016-08-10 21:03:27 +00:00
Kyle Butt e1c931b171 CodeGen: If Convert blocks that would form a diamond when tail-merged.
The following function currently relies on tail-merging for if
conversion to succeed. The common tail of cond_true and cond_false is
extracted, and this then forms a diamond pattern that can be
successfully if converted.

If this block does not get extracted, either because tail-merging is
disabled or the threshold is higher, we should still recognize this
pattern and if-convert it.

Fixed a regression in the original commit. Need to un-reverse branches after
reversing them, or other conversions go awry.

define i32 @t2(i32 %a, i32 %b) nounwind {
entry:
        %tmp1434 = icmp eq i32 %a, %b           ; <i1> [#uses=1]
        br i1 %tmp1434, label %bb17, label %bb.outer

bb.outer:               ; preds = %cond_false, %entry
        %b_addr.021.0.ph = phi i32 [ %b, %entry ], [ %tmp10, %cond_false ]
        %a_addr.026.0.ph = phi i32 [ %a, %entry ], [ %a_addr.026.0, %cond_false ]
        br label %bb

bb:             ; preds = %cond_true, %bb.outer
        %indvar = phi i32 [ 0, %bb.outer ], [ %indvar.next, %cond_true ]
        %tmp. = sub i32 0, %b_addr.021.0.ph
        %tmp.40 = mul i32 %indvar, %tmp.
        %a_addr.026.0 = add i32 %tmp.40, %a_addr.026.0.ph
        %tmp3 = icmp sgt i32 %a_addr.026.0, %b_addr.021.0.ph
        br i1 %tmp3, label %cond_true, label %cond_false

cond_true:              ; preds = %bb
        %tmp7 = sub i32 %a_addr.026.0, %b_addr.021.0.ph
        %tmp1437 = icmp eq i32 %tmp7, %b_addr.021.0.ph
        %indvar.next = add i32 %indvar, 1
        br i1 %tmp1437, label %bb17, label %bb

cond_false:             ; preds = %bb
        %tmp10 = sub i32 %b_addr.021.0.ph, %a_addr.026.0
        %tmp14 = icmp eq i32 %a_addr.026.0, %tmp10
        br i1 %tmp14, label %bb17, label %bb.outer

bb17:           ; preds = %cond_false, %cond_true, %entry
        %a_addr.026.1 = phi i32 [ %a, %entry ], [ %tmp7, %cond_true ], [ %a_addr.026.0, %cond_false ]
        ret i32 %a_addr.026.1
}

Without tail-merging or diamond-tail if conversion:
LBB1_1:                                 @ %bb
                                        @ =>This Inner Loop Header: Depth=1
        cmp     r0, r1
        ble     LBB1_3
@ BB#2:                                 @ %cond_true
                                        @   in Loop: Header=BB1_1 Depth=1
        subs    r0, r0, r1
        cmp     r1, r0
        it      ne
        cmpne   r0, r1
        bgt     LBB1_4
LBB1_3:                                 @ %cond_false
                                        @   in Loop: Header=BB1_1 Depth=1
        subs    r1, r1, r0
        cmp     r1, r0
        bne     LBB1_1
LBB1_4:                                 @ %bb17
        bx      lr

With diamond-tail if conversion, but without tail-merging:
@ BB#0:                                 @ %entry
        cmp     r0, r1
        it      eq
        bxeq    lr
LBB1_1:                                 @ %bb
                                        @ =>This Inner Loop Header: Depth=1
        cmp     r0, r1
        ite     le
        suble   r1, r1, r0
        subgt   r0, r0, r1
        cmp     r1, r0
        bne     LBB1_1
@ BB#2:                                 @ %bb17
        bx      lr

llvm-svn: 278287
2016-08-10 20:45:56 +00:00
Nicolai Haehnle 02d784172c LiveIntervalAnalysis: fix a crash in repairOldRegInRange
Summary:
See the new test case for one that was (non-deterministically) crashing
on trunk and deterministically hit the assertion that I added in D23302.
Basically, the machine function contains a sequence

     DS_WRITE_B32 %vreg4, %vreg14:sub0, ...
     DS_WRITE_B32 %vreg4, %vreg14:sub0, ...
     %vreg14:sub1<def> = COPY %vreg14:sub0

and SILoadStoreOptimizer::mergeWrite2Pair merges the two DS_WRITE_B32
instructions into one before calling repairIntervalsInRange.

Now repairIntervalsInRange wants to repair %vreg14, in particular, and
ends up trying to repair %vreg14:sub1 as well, but that only becomes
active _after_ the range that is to be repaired, hence the crash due
to LR.find(...) == LR.begin() at the start of repairOldRegInRange.

I believe that just skipping those subrange is fine, but again, not too
familiar with that code.

Reviewers: MatzeB, kparzysz, tstellarAMD

Subscribers: llvm-commits, MatzeB

Differential Revision: https://reviews.llvm.org/D23303

llvm-svn: 278268
2016-08-10 18:51:14 +00:00
Kyle Butt 71b1ca1be4 Codegen: Tail Merge: Be less aggressive with special cases.
This change makes it possible for tail-duplication and tail-merging to
be disjoint. By being less aggressive when merging during layout, there are no
overlapping cases between tail-duplication and tail-merging, provided the
thresholds are disjoint.

There is a remaining TODO to benchmark the succ_size() test for non-layout tail
merging.

llvm-svn: 278265
2016-08-10 18:36:18 +00:00
Tim Northover 7552ef5a00 GlobalISel: avoid inserting redundant COPYs for bitcasts.
If the value produced by the bitcast hasn't been referenced yet, we can simply
reuse the input register avoiding an unnecessary COPY instruction.

llvm-svn: 278245
2016-08-10 16:51:14 +00:00
Simon Pilgrim 85c7ea86ae [DAGCombine] Avoid INSERT_SUBVECTOR reinsertions (PR28678)
If the input vector to INSERT_SUBVECTOR is another INSERT_SUBVECTOR, and this inserted subvector replaces the last insertion, then insert into the common source vector.

i.e. 
INSERT_SUBVECTOR( INSERT_SUBVECTOR( Vec, SubOld, Idx ), SubNew, Idx ) --> INSERT_SUBVECTOR( Vec, SubNew, Idx )

Differential Revision: https://reviews.llvm.org/D23330

llvm-svn: 278211
2016-08-10 10:50:53 +00:00
Tim Northover d403a3d8ee GlobalISel: support 'undef' constant.
llvm-svn: 278174
2016-08-09 23:01:30 +00:00