*We don't currently handle the edge case constants (min/max values), so it's not a complete
canonicalization.
To fully solve the motivating bugs, we need to enhance this to recognize a zero vector
too because that's a ConstantAggregateZero which is a ConstantData, not a ConstantVector
or a ConstantDataVector.
Differential Revision: http://reviews.llvm.org/D17859
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269426 91177308-0d34-0410-b5e6-96231b3b80d8
These tests were sitting in Phab for many months. They're good tests and should be in.
Patch by Charlie Turner.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269425 91177308-0d34-0410-b5e6-96231b3b80d8
Fix "Logic error" warnings of the type "Called C++ object pointer is
null" reported by Clang Static Analyzer on the following files:
lib/Analysis/ScalarEvolution.cpp,
lib/Analysis/LoopInfo.cpp.
Patch by Apelete Seketeli!
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269424 91177308-0d34-0410-b5e6-96231b3b80d8
I'm not going to regenerate these anytime soon but do have some diffs to apply that I'd like to do with update_llc_test_checks
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269420 91177308-0d34-0410-b5e6-96231b3b80d8
It's not entirely clear why R_MICROMIPS_(GOT|HI16|LO16) are evaluated
incorrectly in a small number of the LNT tests at this point. However, it's not
related to the STO_MIPS_MICROMIPS issue.
At this point all the microMIPS-related changes of r268900 have been reverted.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269410 91177308-0d34-0410-b5e6-96231b3b80d8
When we convert to the void Select interface, leaving unreferenced
nodes around won't be allowed anymore.
Part of llvm.org/pr26808.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269396 91177308-0d34-0410-b5e6-96231b3b80d8
This reverts commit r269388.
It caused some bots to fail, I'm reverting it until I investigate the
issue.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269395 91177308-0d34-0410-b5e6-96231b3b80d8
When we convert to the void Select interface, leaving unreferenced
nodes around won't be allowed anymore.
Part of llvm.org/pr26808.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269394 91177308-0d34-0410-b5e6-96231b3b80d8
- Where we were returning a node before, call ReplaceNode instead.
- Where we would return null to fall back to another selector, rename
the method to try* and return a bool for success.
- Where we were calling SelectNodeTo, just return afterwards.
Part of llvm.org/pr26808.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269393 91177308-0d34-0410-b5e6-96231b3b80d8
This should fix some compile-time regressions after r267672. Thanks to
Chris Matthews for bisecting it.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269392 91177308-0d34-0410-b5e6-96231b3b80d8
We only really need this to be true for SIFixSGPRCopies.
I'm not sure there's any way this could happen before that point.
Fixes a case where MachineCSE could introduce a cross block
scc use.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269391 91177308-0d34-0410-b5e6-96231b3b80d8
a sequence of values.
It increments through the values in the half-open range: [Begin, End),
producing those values when indirecting the iterator. It should support
integers, iterators, and any other type providing these basic arithmetic
operations.
This came up in the C++ standards committee meeting, and it seemed like
a useful construct that LLVM might want as well, and I wanted to
understand how easily we could solve it. I suspect this can be used to
write simpler counting loops even in LLVM along the lines of:
for (int i : seq(0, v.size())) {
...
};
As part of this, I had to fix the lack of a proxy object returned from
the operator[] in our iterator facade.
Differential Revision: http://reviews.llvm.org/D17870
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269390 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
...loop after the last iteration.
This is really hard to do correctly. The core problem is that we need to
model liveness through the induction PHIs from iteration to iteration in
order to get the correct results, and we need to correctly de-duplicate
the common subgraphs of instructions feeding some subset of the
induction PHIs. All of this can be driven either from a side effect at
some iteration or from the loop values used after the loop finishes.
This patch implements this by storing the forward-propagating analysis
of each instruction in a cache to recall whether it was free and whether
it has become live and thus counted toward the total unroll cost. Then,
at each sink for a value in the loop, we recursively walk back through
every value that feeds the sink, including looping back through the
iterations as needed, until we have marked the entire input graph as
live. Because we cache this, we never visit instructions more than twice
-- once when we analyze them and put them into the cache, and once when
we count their cost towards the unrolled loop. Also, because the cache
is only two bits and because we are dealing with relatively small
iteration counts, we can store all of this very densely in memory to
avoid this from becoming an excessively slow analysis.
The code here is still pretty gross. I would appreciate suggestions
about better ways to factor or split this up, I've stared too long at
the algorithmic side to really have a good sense of what the design
should probably look at.
Also, it might seem like we should do all of this bottom-up, but I think
that is a red herring. Specifically, the simplification power is *much*
greater working top-down. We can forward propagate very effectively,
even across strange and interesting recurrances around the backedge.
Because we use data to propagate, this doesn't cause a state space
explosion. Doing this level of constant folding, etc, would be very
expensive to do bottom-up because it wouldn't be until the last moment
that you could collapse everything. The current solution is essentially
a top-down simplification with a bottom-up cost accounting which seems
to get the best of both worlds. It makes the simplification incremental
and powerful while leaving everything dead until we *know* it is needed.
Finally, a core property of this approach is its *monotonicity*. At all
times, the current UnrolledCost is a conservatively low estimate. This
ensures that we will never early-exit from the analysis due to exceeding
a threshold when if we had continued, the cost would have gone back
below the threshold. These kinds of bugs can cause incredibly hard to
track down random changes to behavior.
We could use a techinque similar (but much simpler) within the inliner
as well to avoid considering speculated code in the inline cost.
Reviewers: chandlerc
Subscribers: sanjoy, mzolotukhin, llvm-commits
Differential Revision: http://reviews.llvm.org/D11758
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269388 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
Currently we consider such instructions as simplified, which is incorrect,
because if their user isn't simplified, we can't actually simplify them too.
This biases our estimates of profitability: for instance the analyzer expects
much more gains from unrolling memcpy loops than there actually are.
Reviewers: hfinkel, chandlerc
Subscribers: mzolotukhin, llvm-commits
Differential Revision: http://reviews.llvm.org/D17365
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269387 91177308-0d34-0410-b5e6-96231b3b80d8
This change is the gold side of the change made in D17115 and clang
patch r261045 to add a ThinLTO specific pipeline that moves more of
the optimization to the backends.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269386 91177308-0d34-0410-b5e6-96231b3b80d8
In verbose mode, we emit a warning if the DWOId of a skeleton CU
mismatches the DWOId of the referenced module. This patch updates the
cached DWOId after a module has been loaded to the DWOId of the module
on disk (instead of storing the DWOId we expected to load). This
allows us to correctly emit the mismatch warning for all subsequent
object files that want to import the same module. This patch also
ensures both warnings are only emitted in verbose mode.
rdar://problem/26214027
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269383 91177308-0d34-0410-b5e6-96231b3b80d8
Having the MachO enums in a def file instead of inline will allow us to write utilities and encoding/decoding methods for load commands without having to write a lot of mechanically repeated code.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269380 91177308-0d34-0410-b5e6-96231b3b80d8
This one has a lot of code churn, but it's all mechanical and
straightforward.
- Where we were returning a node before, call ReplaceNode instead.
- Where we would return null to fall back to another selector, rename
the method to try* and return a bool for success.
- Where we were calling SelectNodeTo, just return afterwards.
Part of llvm.org/pr26808.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269379 91177308-0d34-0410-b5e6-96231b3b80d8
Ported DA to the new PM by splitting the former DependenceAnalysis Pass
into a DependenceInfo result type and DependenceAnalysisWrapperPass type
and adding a new PM-style DependenceAnalysis analysis pass returning the
DependenceInfo.
Patch by Philip Pfaffe, most of the review by Justin.
Differential Revision: http://reviews.llvm.org/D18834
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269370 91177308-0d34-0410-b5e6-96231b3b80d8
Dynamic table when printed shows uppercase tag/values.
This changes it to lower case when printing in GNU style
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269368 91177308-0d34-0410-b5e6-96231b3b80d8
- Where we were returning a node before, call ReplaceNode instead.
- Where we were calling SelectNodeTo, just return afterwards.
Part of llvm.org/pr26808.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269364 91177308-0d34-0410-b5e6-96231b3b80d8
- Where we were returning a node before, call ReplaceNode instead.
- Where we had already replaced all uses and we returned a node, just
remove the dead node instead.
- Where we would return null to fall back to another selector, rename
the method to try* and return a bool for success.
Part of llvm.org/pr26808.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269358 91177308-0d34-0410-b5e6-96231b3b80d8
When we convert to the void Select interface, leaving unreferenced
nodes around won't be allowed anymore.
Part of llvm.org/pr26808.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269355 91177308-0d34-0410-b5e6-96231b3b80d8
This change implements the transformation in processInstruction() for the
LDR rt, =expression to MOV rt, expression when the expression can be evaluated
and can fit into the immediate field of the MOV or a MVN.
Across the ARM and Thumb instruction sets there are several cases to consider,
each with a different range of representatble constants.
In ARM we have:
* Modified immediate (All ARM architectures)
* MOVW (v6t2 and above)
In Thumb we have:
* Modified immediate (v6t2, v7m and v8m.mainline)
* MOVW (v6t2, v7m, v8.mainline and v8m.baseline)
* Narrow Thumb MOV that can be used in an IT block (non flag-setting)
If the immediate fits any of the available alternatives then we make the transformation.
Fixes 25722.
Patch by Peter Smith.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269354 91177308-0d34-0410-b5e6-96231b3b80d8
Alter instances in the test-suite that use immediates that can be represented
in the immediate field of a MOV. The reason for doing this is that when the
LDR rt,=imm transformation to MOV rt, imm the existing tests do not need to
be modified.
Required by the patch that fixes PR25722.
Patch by Peter Smith.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269353 91177308-0d34-0410-b5e6-96231b3b80d8
This change adds a new constant pool kind to ARMOperand. When parsing the
operand for =immediate we create an instance of this operand rather than
creating a constant pool entry and rewriting the operand.
As the new operand kind is only created for ldr rt,= we can make ldr rt,=
an explicit pseudo instruction in ARM, Thumb and Thumb2
The pseudo instruction is expanded in processInstruction(). This creates the
constant pool and transforms the pseudo instruction into a pc-relative ldr to
the constant pool.
There are no functional changes and no modifications needed to existing tests.
Required by the patch that fixes PR25722.
Patch by Peter Smith.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269352 91177308-0d34-0410-b5e6-96231b3b80d8
- Where we were returning a node before, call ReplaceNode instead.
- Where we were calling SelectNodeTo, just return afterwards.
Part of llvm.org/pr26808.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269350 91177308-0d34-0410-b5e6-96231b3b80d8
- Where we were returning a node before, call ReplaceNode instead.
- Where we would return null to fall back to another selector, rename
the method to try* and return a bool for success.
- Where we were calling SelectNodeTo, just return afterwards.
Part of llvm.org/pr26808.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269349 91177308-0d34-0410-b5e6-96231b3b80d8
Split FCMP//ICMP/SEL from the basic arithmetic cost functions. They were not sharing any notable code path (just the return) and were repeatedly testing the opcode.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269348 91177308-0d34-0410-b5e6-96231b3b80d8
When we convert to the void Select interface, leaving unreferenced
nodes around won't be allowed anymore.
Part of llvm.org/pr26808.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@269345 91177308-0d34-0410-b5e6-96231b3b80d8
