When unrolling, the exit values in LCSSA phis will get updated.
Invalidate cached SCEV values for those phis in case SCEV looked through
a exit phi.
Fixes#58340.
Loop peeling currently requires that a) the latch is exiting
b) a branch and c) other exits are unreachable/deopt. This patch
removes all of these limitations, and adds the necessary branch
weight updating support. It essentially works the same way as
before with latch -> exiting terminator and
loop trip count -> per exit trip count.
It's worth noting that there are still other limitations in
profitability heuristics: This patch enables peeling of loops to
make conditions invariant (which is pretty much always highly
profitable if possible), while peeling to make loads dereferenceable
still checks that non-latch exits are unreachable and PGO-based
peeling has even more conditions. Those checks could be relaxed
later if we consider those cases profitable.
The motivation for this change is that loops using iterator adaptors
in Rust often optimize very badly, and end up with a loop phi of the
form phi(true, false) in the final result. Peeling eliminates that
phi and conditions based on it, which enables a lot of follow-on
simplification.
Differential Revision: https://reviews.llvm.org/D134803
At the moment, LoopAccessAnalysis is a loop analysis for the new pass
manager. The issue with that is that LAI caches SCEV expressions and
modifications in a loop may impact SCEV expressions in other loops, but
we do not have a convenient way to invalidate LAI for other loops
withing a loop pipeline.
To avoid this issue, turn it into a function analysis which returns a
manager object that keeps track of the individual LAI objects per loop.
Fixes#50940.
Fixes#51669.
Reviewed By: aeubanks
Differential Revision: https://reviews.llvm.org/D134606
SimplifyCFG folds
bool foo() {
if (cond1) return false;
if (cond2) return false;
return true;
}
as
bool foo() {
if (cond1 | cond2) return false
return true;
}
'cond2' is called 'bonus insts' in branch folding since they introduce overhead
since the original CFG could do early exit but the folded CFG always executes
them. SimplifyCFG calculates the costs of 'bonus insts' of a folding a BB into
its predecessor BB which shares the destination. If it is below bonus-inst-threshold,
SimplifyCFG will fold that BB into its predecessor and cond2 will always be executed.
When SimplifyCFG calculates the cost of 'bonus insts', it only consider 'bonus' insts
in the current BB to be considered for folding. This causes issue for unrolled loops
which share destinations, e.g.
bool foo(int *a) {
for (int i = 0; i < 32; i++)
if (a[i] > 0) return false;
return true;
}
After unrolling, it becomes
bool foo(int *a) {
if(a[0]>0) return false
if(a[1]>0) return false;
//...
if(a[31]>0) return false;
return true;
}
SimplifyCFG will merge each BB with its predecessor BB,
and ends up with 32 'bonus insts' which are always executed, which
is much slower than the original CFG.
The root cause is that SimplifyCFG does not consider the
accumulated cost of 'bonus insts' which are folded from
different BB's.
This patch fixes that by introducing a ValueMap to track
costs of 'bonus insts' coming from different BB's into
the same BB, and cuts off if the accumulated cost
exceeds a threshold.
Reviewed by: Artem Belevich, Florian Hahn, Nikita Popov, Matt Arsenault
Differential Revision: https://reviews.llvm.org/D132408
Summary:
The code for generating a name for loops for various reporting scenarios
created a name by serializing the loop into a string. This may result in
a very large name for a loop containing many blocks. Use the getName()
function on the loop instead.
Author: Jamie Schmeiser <schmeise@ca.ibm.com>
Reviewed By: Whitney (Whitney Tsang), aeubanks (Arthur Eubanks)
Differential Revision: https://reviews.llvm.org/D133587
This updates the naming for the LAA printing pass to be in line with
most other analysis printing passes.
The old name has come up as confusing multiple times already, e.g. in
D131924.
The previous implementation translated from names like sifive-7-series
to sifive-7-rv32 or sifive-7-rv64. This also required sifive-7-rv32
and sifive-7-rv64 to be valid CPU names. As those are not real
CPUs it doesn't make sense to accept them in -mcpu.
This patch does away with the translation and adds sifive-7-series
directly to RISCV.td. Removing sifive-7-rv32 and sifive-7-rv64.
sifive-7-series is only allowed in -mtune.
I've also added "rocket" to RISCV.td but have not removed rocket-rv32
or rocket-rv64.
To prevent -mcpu=sifive-7-series or -mcpu=rocket being used with llc,
I've added a Feature32Bit to all rv32 CPUs. And made it an error to
have an rv32 triple without Feature32Bit. sifive-7-series and rocket
do not have Feature32Bit or Feature64Bit set so the user would need
to provide -mattr=+32bit or -mattr=+64bit along with the -mcpu to
avoid the error.
SiFive no longer names their newer products with 3, 5, or 7 series.
Instead we have p200 series, x200 series, p500 series, and p600 series.
Following the previous behavior would require a sifive-p500-rv32 and
sifive-p500-rv64 in order to support -mtune=sifive-p500-series. There
is currently no p500 product, but it could start getting confusing if
there was in the future.
I'm open to hearing alternatives for how to achieve my main goal
of removing sifive-7-rv32/rv64 as a CPU name.
Reviewed By: reames
Differential Revision: https://reviews.llvm.org/D131708
This reverts commit 354fa0b480.
Returning as is. The patch was reverted due to a miscompile, but
this patch is not causing it. This patch made it possible to infer
some nuw flags in code guarded by `false` condition, and then someone
else to managed to propagate the flag from dead code outside.
Returning the patch to be able to reproduce the issue.
This reverts commit 34ae308c73.
Our internal testing found a miscompile. Not sure if it's caused by
this patch or it revealed something else. Reverting while investigating.
Sometimes SCEV cannot infer nuw/nsw from something as simple as
```
len in [0, MAX_INT]
...
iv = phi(0, iv.next)
guard(iv <s len)
guard(iv <u len)
iv.next = iv + 1
```
just because flag strenthening only relies on definition and does not use local facts.
This patch adds support for the simplest case: inference of flags of `add(x, constant)`
if we can contextually prove that `x <= max_int - constant`.
In case if it has negative CT impact, we can add an option to switch it off. I woudln't
expect that though.
Differential Revision: https://reviews.llvm.org/D129643
Reviewed By: apilipenko
Following some recent discussions, this changes the representation
of callbrs in IR. The current blockaddress arguments are replaced
with `!` label constraints that refer directly to callbr indirect
destinations:
; Before:
%res = callbr i8* asm "", "=r,r,i"(i8* %x, i8* blockaddress(@test8, %foo))
to label %asm.fallthrough [label %foo]
; After:
%res = callbr i8* asm "", "=r,r,!i"(i8* %x)
to label %asm.fallthrough [label %foo]
The benefit of this is that we can easily update the successors of
a callbr, without having to worry about also updating blockaddress
references. This should allow us to remove some limitations:
* Allow unrolling/peeling/rotation of callbr, or any other
clone-based optimizations
(https://github.com/llvm/llvm-project/issues/41834)
* Allow duplicate successors
(https://github.com/llvm/llvm-project/issues/45248)
This is just the IR representation change though, I will follow up
with patches to remove limtations in various transformation passes
that are no longer needed.
Differential Revision: https://reviews.llvm.org/D129288
ConnectProlog adds new incoming values to exit phi nodes which can
change the SCEV for the phi after 20d798bd47.
Fix is analog to cfc741bc0e.
Fixes#56286.
ConnectEpilog adds new incoming values to exit phi nodes which can
change the SCEV for the phi after 20d798bd47.
Fix is analog to cfc741bc0e.
Fixes#56282.
LoopPeel add new incoming values to exit phi nodes which can change the
SCEV for the phi after 20d798bd47.
Forget SCEVs for such phis.
Fixes#56044.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D128164
Teach the unroller(s) how to handle an invalid cost. This avoids crashes when the backend can't provide a cost due to either a fundemental limitation or an unimplemented cost model case.
Differential Revision: https://reviews.llvm.org/D127305
There are a few places where we use report_fatal_error when the input is broken.
Currently, this function always crashes LLVM with an abort signal, which
then triggers the backtrace printing code.
I think this is excessive, as wrong input shouldn't give a link to
LLVM's github issue URL and tell users to file a bug report.
We shouldn't print a stack trace either.
This patch changes report_fatal_error so it uses exit() rather than
abort() when its argument GenCrashDiag=false.
Reviewed by: nikic, MaskRay, RKSimon
Differential Revision: https://reviews.llvm.org/D126550
This is a followup to D125754. We introduce two branches, one
before the unrolled loop and one before the epilogue (and similar
for the prologue case). The previous patch only froze the
condition on the first branch.
Rather than independently freezing the second condition, this patch
instead freezes TripCount and bases BECount on it. These are the
two quantities involved in the conditions, and this ensures that
both work on a consistent, non-poisonous trip count.
Differential Revision: https://reviews.llvm.org/D125896
When performing runtime unrolling with multiple exits, one of the
earlier (non-latch) exits may exit the loop on the first iteration,
such that we never branch on the latch exit condition. As such, we
need to freeze the condition of the new branch that is introduced
before the loop, as it now executes unconditionally.
Differential Revision: https://reviews.llvm.org/D125754
IMO when user provide unroll pragma, compiler should always respect it.
It is not clear to me why loop unroll pass currently ensure that the
unrolled loop size is limited by PragmaUnrollThreshold.
Reviewed By: Meinersbur
Differential Revision: https://reviews.llvm.org/D119148
This patch tries to sink instructions when they are only used in a successor block.
This is a further enhancement patch based on Anna's commit:
D109700, which allows sinking an instruction having multiple uses in a single user.
In this patch, sink instructions with multiple users in a single successor block will be supported.
It could fix a known issue from rust:
https://github.com/rust-lang/rust/issues/51346#issuecomment-394443610
Reviewed By: nikic, reames
Differential Revision: https://reviews.llvm.org/D121585
LICM will speculatively hoist code outside of loops. This requires removing information, like alias analysis (https://github.com/llvm/llvm-project/issues/53794), range information (https://bugs.llvm.org/show_bug.cgi?id=50550), among others. Prior to https://reviews.llvm.org/D99249 , LICM would only be run after LoopRotate. Running Loop Rotate prior to LICM prevents a instruction hoist from being speculative, if it was conditionally executed by the iteration (as is commonly emitted by clang and other frontends). Adding the additional LICM pass first, however, forces all of these instructions to be considered speculative, even if they are not speculative after LoopRotate. This destroys information, resulting in performance losses for discarding this additional information.
This PR modifies LICM to accept a ``speculative'' parameter which allows LICM to be set to perform information-loss speculative hoists or not. Phase ordering is then modified to not perform the information-losing speculative hoists until after loop rotate is performed, preserving this additional information.
Reviewed By: lebedev.ri
Differential Revision: https://reviews.llvm.org/D119965
In D115311, we're looking to modify clang to emit i constraints rather
than X constraints for callbr's indirect destinations. Prior to doing
so, update all of the existing tests in llvm/ to match.
Reviewed By: void, jyknight
Differential Revision: https://reviews.llvm.org/D115410
If we have an exit which is controlled by a loop invariant condition and which dominates the latch, we know only the copy in the first unrolled iteration can be taken. All other copies are dead.
The change itself is pretty straight forward, but let me add two points of context:
* I'd have expected other transform passes to catch this after unrolling, but I'm seeing multiple examples where we get to the end of O2/O3 without simplifying.
* I'd like to do a stronger change which did CSE during unroll and accounted for invariant expressions (as defined by SCEV instead of trivial ones from LoopInfo), but that doesn't fit cleanly into the current code structure.
Differential Revision: https://reviews.llvm.org/D116496
Commit 150681f increases
cost of producing MMA types (vector pair and quad).
However, it increases the cost for getUserCost() which is
used in unrolling. As a result, loops that contain these
types already (from the user code) cannot be unrolled
(even with the user's unroll pragma). This was an unintended
sideeffect. Reverting that portion of the commit to allow
unrolling such loops.
Differential revision: https://reviews.llvm.org/D115424
Both these preference helper functions have initial support with
this change. The loop unrolling preferences are set with initial
settings to control thresholds, size and attributes of loops to
unroll with some tuning done. The peeling preferences may need
some tuning as well as the initial support looks much like what
other architectures utilize.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D113798
This change allows us to estimate trip count from profile metadata for all multiple exit loops. We still do the estimate only from the latch, but that's fine as it causes us to over estimate the trip count at worst.
Reviewing the uses of the API, all but one are cases where we restrict a loop transformation (unroll, and vectorize respectively) when we know the trip count is short enough. So, as a result, the change makes these passes strictly less aggressive. The test change illustrates a case where we'd previously have runtime unrolled a loop which ran fewer iterations than the unroll factor. This is definitely unprofitable.
The one case where an upper bound on estimate trip count could drive a more aggressive transform is peeling, and I duplicated the logic being removed from the generic estimation there to keep it the same. The resulting heuristic makes no sense and should probably be immediately removed, but we can do that in a separate change.
This was noticed when analyzing regressions on D113939.
I plan to come back and incorporate estimated trip counts from other exits, but that's a minor improvement which can follow separately.
Differential Revision: https://reviews.llvm.org/D115362
Both these preference helper functions have initial support with
this change. The loop unrolling preferences are set with initial
settings to control thresholds, size and attributes of loops to
unroll with some tuning done. The peeling preferences may need
some tuning as well as the initial support looks much like what
other architectures utilize.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D113798
The basic problem we have is that we're trying to reuse an instruction which is mapped to some SCEV. Since we can have multiple such instructions (potentially with different flags), this is analogous to our need to drop flags when performing CSE. A trivial implementation would simply drop flags on any instruction we decided to reuse, and that would be correct.
This patch is almost that trivial patch except that we preserve flags on the reused instruction when existing users would imply UB on overflow already. Adding new users can, at most, refine this program to one which doesn't execute UB which is valid.
In practice, this fixes two conceptual problems with the previous code: 1) a binop could have been canonicalized into a form with different opcode or operands, or 2) the inbounds GEP case which was simply unhandled.
On the test changes, most are pretty straight forward. We loose some flags (in some cases, they'd have been dropped on the next CSE pass anyways). The one that took me the longest to understand was the ashr-expansion test. What's happening there is that we're considering reuse of the mul, previously we disallowed it entirely, now we allow it with no flags. The surrounding diffs are all effects of generating the same mul with a different operand order, and then doing simple DCE.
The loss of the inbounds is unfortunate, but even there, we can recover most of those once we actually treat branch-on-poison as immediate UB.
Differential Revision: https://reviews.llvm.org/D112734
There are multiple possible ways to represent the X - urem X, Y pattern. SCEV was not canonicalizing, and thus, depending on which you analyzed, you could get different results. The sub representation appears to produce strictly inferior results in practice, so I decided to canonicalize to the Y * X/Y version.
The motivation here is that runtime unroll produces the sub X - (and X, Y-1) pattern when Y is a power of two. SCEV is thus unable to recognize that an unrolled loop exits because we don't figure out that the new unrolled step evenly divides the trip count of the unrolled loop. After instcombine runs, we convert the the andn form which SCEV recognizes, so essentially, this is just fixing a nasty pass ordering dependency.
The ARM loop hardware interaction in the test diff is opague to me, but the comments in the review from others knowledge of the infrastructure appear to indicate these are improvements in loop recognition, not regressions.
Differential Revision: https://reviews.llvm.org/D114018
This is one of those wonderful "in theory X doesn't matter, but in practice is does" changes. In this particular case, we shift the IVs inserted by the runtime unroller to clamp iteration count of the loops* from decrementing to incrementing.
Why does this matter? A couple of reasons:
* SCEV doesn't have a native subtract node. Instead, all subtracts (A - B) are represented as A + -1 * B and drops any flags invalidated by such. As a result, SCEV is slightly less good at reasoning about edge cases involving decrementing addrecs than incrementing ones. (You can see this in the inferred flags in some of the test cases.)
* Other parts of the optimizer produce incrementing IVs, and they're common in idiomatic source language. We do have support for reversing IVs, but in general if we produce one of each, the pair will persist surprisingly far through the optimizer before being coalesced. (You can see this looking at nearby phis in the test cases.)
Note that if the hardware prefers decrementing (i.e. zero tested) loops, LSR should convert back immediately before codegen.
* Mostly irrelevant detail: The main loop of the prolog case is handled independently and will simple use the original IV with a changed start value. We could in theory use this scheme for all iteration clamping, but that's a larger and more invasive change.
The unrolling code was previously inserting new cloned blocks at the end of the function. The result of this with typical loop structures is that the new iterations are placed far from the initial iteration.
With unrolling, the general assumption is that the a) the loop is reasonable hot, and b) the first Count-1 copies of the loop are rarely (if ever) loop exiting. As such, placing Count-1 copies out of line is a fairly poor code placement choice. We'd much rather fall through into the hot (non-exiting) path. For code with branch profiles, later layout would fix this, but this may have a positive impact on non-PGO compiled code.
However, the real motivation for this change isn't performance. Its readability and human understanding. Having to jump around long distances in an IR file to trace an unrolled loop structure is error prone and tedious.
Florian Hahn