The penalty is currently getting applied in a bunch of places where it
doesn't make sense, like bitcasts (which are free) and calls (which
were getting the call penalty applied twice). Instead, just apply the
penalty to binary operators and floating-point casts.
While I'm here, also fix getFPOpCost() to do the right thing in more
cases, so we don't have to dig into function attributes.
Differential Revision: https://reviews.llvm.org/D41522
llvm-svn: 321332
SROA analysis of InlineCost can figure out that some stores can be removed
after inlining and then the repeated loads clobbered by these stores are also
free. This patch finds these clobbered loads and adjust the inline cost
accordingly.
Differential Revision: https://reviews.llvm.org/D33946
llvm-svn: 320814
This patch fix this FIXME in visitPHI()
FIXME: We should potentially be tracking values through phi nodes,
especially when they collapse to a single value due to deleted CFG edges
during inlining.
Differential Revision: https://reviews.llvm.org/D38594
llvm-svn: 320699
Summary:
Add LLVM_FORCE_ENABLE_DUMP cmake option, and use it along with
LLVM_ENABLE_ASSERTIONS to set LLVM_ENABLE_DUMP.
Remove NDEBUG and only use LLVM_ENABLE_DUMP to enable dump methods.
Move definition of LLVM_ENABLE_DUMP from config.h to llvm-config.h so
it'll be picked up by public headers.
Differential Revision: https://reviews.llvm.org/D38406
llvm-svn: 315590
parameterized emit() calls
Summary: This is not functional change to adopt new emit() API added in r313691.
Reviewed By: anemet
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D38285
llvm-svn: 315476
The code responsible for analysis of inbounds GEPs is extracted into a separate
function: CallAnalyzer::canFoldInboundsGEP. With the patch SROA
enabling/disabling code is localized at one place instead of spreading across
the code of CallAnalyzer::visitGetElementPtr.
Differential Revision: https://reviews.llvm.org/D38233
llvm-svn: 314787
InlineCost can understand Select IR now. This patch finds free Select IRs and
continue the propagation of SimplifiedValues, ConstantOffsetPtrs, and
SROAArgValues.
Differential Revision: https://reviews.llvm.org/D37198
llvm-svn: 314307
Summary:
Full inline cost is computed when -inline-cost-full is true or ORE is
non-null. This patch adds another way to compute full inline cost by
adding a field to InlineParams. This will be used by SampleProfileLoader
to check legality of inlining a callee that it wants to inline.
Reviewers: danielcdh, haicheng
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D37819
llvm-svn: 313185
Change the early exit condition from Cost > Threshold to Cost >= Threshold
because the inline condition is Cost < Threshold.
Differential Revision: https://reviews.llvm.org/D37087
llvm-svn: 311791
Currently, the inline cost model will bail once the inline cost exceeds the
inline threshold in order to avoid unnecessary compile-time. However, when
debugging it is useful to compute the full cost, so this command line option
is added to override the default behavior.
I took over this work from Chad Rosier (mcrosier@codeaurora.org).
Differential Revision: https://reviews.llvm.org/D35850
llvm-svn: 311371
localized to the code that uses those analyses.
Technically, this can change behavior as we no longer require the
existence of the ProfileSummaryInfo analysis to use local profile
information via BFI. We didn't actually require the PSI to have an
interesting profile though, so this only really impacts the behavior in
non-default pass pipelines.
IMO, this makes it substantially less surprising how everything works --
before an analysis that wasn't actually used had to exist to trigger
*any* profile aware inlining. I think the new organization makes it more
obvious where various checks for profile signals happen.
Differential Revision: https://reviews.llvm.org/D36710
llvm-svn: 310888
Summary:
This increases the inlining threshold for hot callsites. Hotness is
defined in terms of block frequency of the callsite relative to the
caller's entry block's frequency. Since this requires BFI in the
inliner, this only affects the new PM pipeline. This is enabled by
default at -O3.
This improves the performance of some internal benchmarks. Notably, an
internal benchmark for Gipfeli compression
(https://github.com/google/gipfeli) improves by ~7%. Povray in SPEC2006
improves by ~2.5%. I am running more experiments and will update the
thread if other benchmarks show improvement/regression.
In terms of text size, LLVM test-suite shows an 1.22% text size
increase. Diving into the results, 13 of the benchmarks in the
test-suite increases by > 10%. Most of these are small, but
Adobe-C++/loop_unroll (17.6% increases) and tramp3d(20.7% size increase)
have >250K text size. On a large application, the text size increases by
2%
Reviewers: chandlerc, davidxl
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D36199
llvm-svn: 309994
Summary:
Inlining threshold is increased by application of bonuses when the
callee has a single reachable basic block or is rich in vector
instructions. Similarly, inlining cost is reduced by applying a large
bonus when the last call to a static function is considered for
inlining. This patch disables the application of these bonuses when the
callsite or the callee is cold. The intention here is to prevent a large
cold callsite from being inlined to a non-cold caller that could prevent
the caller from being inlined. This is especially important when the
cold callsite is a last call to a static since the associated bonus is
very high.
Reviewers: chandlerc, davidxl
Subscribers: danielcdh, llvm-commits
Differential Revision: https://reviews.llvm.org/D35823
llvm-svn: 309441
Currently CallAnalyzer::isGEPFree uses TTI::getGEPCost to check if GEP is free.
TTI::getGEPCost cannot handle cases when GEPs participate in Def-Use dependencies
(see https://reviews.llvm.org/D31186 for example).
There is TTI::getUserCost which can calculate the cost more accurately by
taking dependencies into account.
Differential Revision: https://reviews.llvm.org/D33685
llvm-svn: 309268
Summary: visitSwitchInst should not take INT_MAX when Cost is negative. Instead of INT_MAX , we also use a valid upperbound cost when overflow occurs in Cost.
Reviewers: hans, echristo, dmgreen
Reviewed By: dmgreen
Subscribers: mcrosier, javed.absar, llvm-commits, eraman
Differential Revision: https://reviews.llvm.org/D34436
llvm-svn: 306118
Summary:
This is to enable the new switch inline cost heuristic (r301649) by removing the
old heuristic as well as the flag itself.
In my experiment for LLVM test suite and spec2000/2006, +17.82% performance and
8% code size reduce was observed in spec2000/vertex with O3 LTO in AArch64.
No significant code size / performance regression was found in O3/O2/Os. No
significant complain was reported from the llvm-dev thread.
Reviewers: hans, chandlerc, eraman, haicheng, mcrosier, bmakam, eastig, ddibyend, echristo
Reviewed By: echristo
Subscribers: javed.absar, kristof.beyls, echristo, aemerson, rengolin, mehdi_amini
Differential Revision: https://reviews.llvm.org/D32653
llvm-svn: 304594
Update threshold based on callee's hotness only when BFI is not available.
Otherwise use only callsite's hotness. This makes it easier to reason about
hotness related threshold updates.
Differential revision: https://reviews.llvm.org/D33157
llvm-svn: 303210
I ran the test-suite (including SPEC 2006) in PGO mode comparing cold
thresholds of 225 and 45. Here are some stats on the text size:
Out of 904 tests that ran, 197 see a change in text size. The average
text size reduction (of all the 904 binaries) is 1.07%. Of the 197
binaries, 19 see a text size increase, as high as 18%, but most of them
are small single source benchmarks. There are 3 multisource benchmarks
with a >0.5% size increase (0.7, 1.3 and 2.1 are their % increases). On
the other side of the spectrum, 31 benchmarks see >10% size reduction
and 6 of them are MultiSource.
I haven't run the test-suite with other values of inlinecold-threshold.
Since we have a cold callsite threshold of 45, I picked this value.
Differential revision: https://reviews.llvm.org/D33106
llvm-svn: 302829
Summary:
The motivation example is like below which has 13 cases but only 2 distinct targets
```
lor.lhs.false2: ; preds = %if.then
switch i32 %Status, label %if.then27 [
i32 -7012, label %if.end35
i32 -10008, label %if.end35
i32 -10016, label %if.end35
i32 15000, label %if.end35
i32 14013, label %if.end35
i32 10114, label %if.end35
i32 10107, label %if.end35
i32 10105, label %if.end35
i32 10013, label %if.end35
i32 10011, label %if.end35
i32 7008, label %if.end35
i32 7007, label %if.end35
i32 5002, label %if.end35
]
```
which is compiled into a balanced binary tree like this on AArch64 (similar on X86)
```
.LBB853_9: // %lor.lhs.false2
mov w8, #10012
cmp w19, w8
b.gt .LBB853_14
// BB#10: // %lor.lhs.false2
mov w8, #5001
cmp w19, w8
b.gt .LBB853_18
// BB#11: // %lor.lhs.false2
mov w8, #-10016
cmp w19, w8
b.eq .LBB853_23
// BB#12: // %lor.lhs.false2
mov w8, #-10008
cmp w19, w8
b.eq .LBB853_23
// BB#13: // %lor.lhs.false2
mov w8, #-7012
cmp w19, w8
b.eq .LBB853_23
b .LBB853_3
.LBB853_14: // %lor.lhs.false2
mov w8, #14012
cmp w19, w8
b.gt .LBB853_21
// BB#15: // %lor.lhs.false2
mov w8, #-10105
add w8, w19, w8
cmp w8, #9 // =9
b.hi .LBB853_17
// BB#16: // %lor.lhs.false2
orr w9, wzr, #0x1
lsl w8, w9, w8
mov w9, #517
and w8, w8, w9
cbnz w8, .LBB853_23
.LBB853_17: // %lor.lhs.false2
mov w8, #10013
cmp w19, w8
b.eq .LBB853_23
b .LBB853_3
.LBB853_18: // %lor.lhs.false2
mov w8, #-7007
add w8, w19, w8
cmp w8, #2 // =2
b.lo .LBB853_23
// BB#19: // %lor.lhs.false2
mov w8, #5002
cmp w19, w8
b.eq .LBB853_23
// BB#20: // %lor.lhs.false2
mov w8, #10011
cmp w19, w8
b.eq .LBB853_23
b .LBB853_3
.LBB853_21: // %lor.lhs.false2
mov w8, #14013
cmp w19, w8
b.eq .LBB853_23
// BB#22: // %lor.lhs.false2
mov w8, #15000
cmp w19, w8
b.ne .LBB853_3
```
However, the inline cost model estimates the cost to be linear with the number
of distinct targets and the cost of the above switch is just 2 InstrCosts.
The function containing this switch is then inlined about 900 times.
This change use the general way of switch lowering for the inline heuristic. It
etimate the number of case clusters with the suitability check for a jump table
or bit test. Considering the binary search tree built for the clusters, this
change modifies the model to be linear with the size of the balanced binary
tree. The model is off by default for now :
-inline-generic-switch-cost=false
This change was originally proposed by Haicheng in D29870.
Reviewers: hans, bmakam, chandlerc, eraman, haicheng, mcrosier
Reviewed By: hans
Subscribers: joerg, aemerson, llvm-commits, rengolin
Differential Revision: https://reviews.llvm.org/D31085
llvm-svn: 301649
This avoids the confusing 'CS.paramHasAttr(ArgNo + 1, Foo)' pattern.
Previously we were testing return value attributes with index 0, so I
introduced hasReturnAttr() for that use case.
llvm-svn: 300367
and to expose a handle to represent the actual case rather than having
the iterator return a reference to itself.
All of this allows the iterator to be used with common STL facilities,
standard algorithms, etc.
Doing this exposed some missing facilities in the iterator facade that
I've fixed and required some work to the actual iterator to fully
support the necessary API.
Differential Revision: https://reviews.llvm.org/D31548
llvm-svn: 300032
Summary: Because SamplePGO passes will be invoked twice in ThinLTO build: once at compile phase, the other at backend. We want to make sure the IR at the 2nd phase matches the hot part in profile, thus we do not want to inline hot callsites in the first phase.
Reviewers: tejohnson, eraman
Reviewed By: tejohnson
Subscribers: mehdi_amini, llvm-commits, Prazek
Differential Revision: https://reviews.llvm.org/D31201
llvm-svn: 298428
Several visitors check if operands to the instruction are constants,
either as it is or after looking up SimplifiedValues, check if the
result is a constant and update the SimplifiedValues map. This
refactoring splits it into a common function that does the checking of
whether the operands are constants and updating of the SimplifiedValues
table, and an instruction specific part that is implemented by each
instruction visitor as a lambda and passed to the common function.
Differential revision: https://reviews.llvm.org/D30104
llvm-svn: 295552
This adds the following to the new PM based inliner in PGO mode:
* Use block frequency analysis to derive callsite's profile count and use
that to adjust thresholds of hot and cold callsites.
* Incrementally update the BFI of the caller after a callee gets inlined
into it. This incremental update is only within an invocation of the run
method - BFI is not preserved across calls to run.
Update the function entry count of the callee after inlining it into a
caller.
* I've tuned the thresholds for the hot and cold callsites using a hacked
up version of the old inliner that explicitly computes BFI on a set of
internal benchmarks and spec. Once the new PM based pipeline stabilizes
(IIRC Chandler mentioned there are known issues) I'll benchmark this
again and adjust the thresholds if required.
Inliner PGO support.
Differential revision: https://reviews.llvm.org/D28331
llvm-svn: 292666
This is the third attemp to recommit r292526.
The original summary:
Currently, a GEP is considered free only if its indices are all constant.
TTI::getGEPCost() can give target-specific more accurate analysis. TTI is
already used for the cost of many other instructions.
llvm-svn: 292633
This is the second attemp to recommit r292526.
The original summary:
Currently, a GEP is considered free only if its indices are all constant.
TTI::getGEPCost() can give target-specific more accurate analysis. TTI is
already used for the cost of many other instructions.
llvm-svn: 292616
This recommits r292526 which is reverted in r292529 after fixing the test case.
The original summary:
Currently, a GEP is considered free only if its indices are all constant.
TTI::getGEPCost() can give target-specific more accurate analysis. TTI is
already used for the cost of many other instructions.
llvm-svn: 292570
Currently, a GEP is considered free only if its indices are all constant.
TTI::getGEPCost() can give target-specific more accurate analysis. TTI is
already used for the cost of many other instructions.
Differential Revision: https://reviews.llvm.org/D28693
llvm-svn: 292526
Functional change: Previously, if a callee is cold, we used ColdThreshold if it minimizes the existing threshold. This was irrespective of whether we were optimizing for minsize (-Oz) or not. But -Oz uses very low threshold to begin with and the inlining with -Oz is expected to be tuned for lowering code size, so there is no good reason to set an even lower threshold for cold callees. We now lower the threshold for cold callees only when -Oz is not used. For default values of -inlinethreshold and -inlinecold-threshold, this change has no effect and this simplifies the code.
NFC changes: Group all threshold updates that are guarded by !Caller->optForMinSize() and within that group threshold updates that require profile summary info.
Differential revision: https://reviews.llvm.org/D28369
llvm-svn: 291487
This doesn't implement *every* feature of the existing inliner, but
tries to implement the most important ones for building a functional
optimization pipeline and beginning to sort out bugs, regressions, and
other problems.
Notable, but intentional omissions:
- No alloca merging support. Why? Because it isn't clear we want to do
this at all. Active discussion and investigation is going on to remove
it, so for simplicity I omitted it.
- No support for trying to iterate on "internally" devirtualized calls.
Why? Because it adds what I suspect is inappropriate coupling for
little or no benefit. We will have an outer iteration system that
tracks devirtualization including that from function passes and
iterates already. We should improve that rather than approximate it
here.
- Optimization remarks. Why? Purely to make the patch smaller, no other
reason at all.
The last one I'll probably work on almost immediately. But I wanted to
skip it in the initial patch to try to focus the change as much as
possible as there is already a lot of code moving around and both of
these *could* be skipped without really disrupting the core logic.
A summary of the different things happening here:
1) Adding the usual new PM class and rigging.
2) Fixing minor underlying assumptions in the inline cost analysis or
inline logic that don't generally hold in the new PM world.
3) Adding the core pass logic which is in essence a loop over the calls
in the nodes in the call graph. This is a bit duplicated from the old
inliner, but only a handful of lines could realistically be shared.
(I tried at first, and it really didn't help anything.) All told,
this is only about 100 lines of code, and most of that is the
mechanics of wiring up analyses from the new PM world.
4) Updating the LazyCallGraph (in the new PM) based on the *newly
inlined* calls and references. This is very minimal because we cannot
form cycles.
5) When inlining removes the last use of a function, eagerly nuking the
body of the function so that any "one use remaining" inline cost
heuristics are immediately refined, and queuing these functions to be
completely deleted once inlining is complete and the call graph
updated to reflect that they have become dead.
6) After all the inlining for a particular function, updating the
LazyCallGraph and the CGSCC pass manager to reflect the
function-local simplifications that are done immediately and
internally by the inline utilties. These are the exact same
fundamental set of CG updates done by arbitrary function passes.
7) Adding a bunch of test cases to specifically target CGSCC and other
subtle aspects in the new PM world.
Many thanks to the careful review from Easwaran and Sanjoy and others!
Differential Revision: https://reviews.llvm.org/D24226
llvm-svn: 290161
