The patch is generated using this command:
tools/clang/tools/extra/clang-tidy/tool/run-clang-tidy.py -fix \
-checks=-*,llvm-namespace-comment -header-filter='llvm/.*|clang/.*' \
llvm/lib/
Thanks to Eugene Kosov for the original patch!
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@240137 91177308-0d34-0410-b5e6-96231b3b80d8
on a per-function basis.
Previously some of the passes were conditionally added to ARM's pass pipeline
based on the target machine's subtarget. This patch makes changes to add those
passes unconditionally and execute them conditonally based on the predicate
functor passed to the pass constructors. This enables running different sets of
passes for different functions in the module.
rdar://problem/20542263
Differential Revision: http://reviews.llvm.org/D8717
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@239325 91177308-0d34-0410-b5e6-96231b3b80d8
Now that we can look at users, we can trivially do this: when we would
have otherwise disabled GlobalMerge (currently -O<3), we can just run
it for minsize functions, as it's usually a codesize win.
Differential Revision: http://reviews.llvm.org/D10054
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@239087 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
A later change that has RewriteStatepointsForGC change function
attributes throughout the module depends on this.
Reviewers: reames, pgavlin
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D10104
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@238882 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
This is a pass for speculative execution of instructions for simple if-then (triangle) control flow. It's aimed at GPUs, but could perhaps be used in other contexts. Enabling this pass gives us a 1.0% geomean improvement on Google benchmark suites, with one benchmark improving 33%.
Credit goes to Jingyue Wu for writing an earlier version of this pass.
Patched by Bjarke Roune.
Test Plan:
This patch adds a set of tests in test/Transforms/SpeculativeExecution/spec.ll
The pass is controlled by a flag which defaults to having the pass not run.
Reviewers: eliben, dberlin, meheff, jingyue, hfinkel
Reviewed By: jingyue, hfinkel
Subscribers: majnemer, jholewinski, llvm-commits
Differential Revision: http://reviews.llvm.org/D9360
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@237459 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
This implements the initial version as was proposed earlier this year
(http://lists.cs.uiuc.edu/pipermail/llvmdev/2015-January/080462.html).
Since then Loop Access Analysis was split out from the Loop Vectorizer
and was made into a separate analysis pass. Loop Distribution becomes
the second user of this analysis.
The pass is off by default and can be enabled
with -enable-loop-distribution. There is currently no notion of
profitability; if there is a loop with dependence cycles, the pass will
try to split them off from other memory operations into a separate loop.
I decided to remove the control-dependence calculation from this first
version. This and the issues with the PDT are actively discussed so it
probably makes sense to treat it separately. Right now I just mark all
terminator instruction required which keeps identical CFGs for each
distributed loop. This seems to be working pretty well for 456.hmmer
where even though there is an empty if-then block in the distributed
loop initially, it gets completely removed.
The pass keeps DominatorTree and LoopInfo updated. I've tested this
with -loop-distribute-verify with the testsuite where we distribute ~90
loops. SimplifyLoop is violated in some cases and I have a FIXME
covering this.
Reviewers: hfinkel, nadav, aschwaighofer
Reviewed By: aschwaighofer
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D8831
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@237358 91177308-0d34-0410-b5e6-96231b3b80d8
The pass doesn't actually modify the module outside of the function being processed. The only confusing piece is that it both inserts calls and then inlines the resulting calls. Given that, it definitely invalidates module level analysis results, but many FunctionPasses do that.
Differential Revision: http://reviews.llvm.org/D9590
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@237185 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
This transformation reassociates a n-ary add so that the add can partially reuse
existing instructions. For example, this pass can simplify
void foo(int a, int b) {
bar(a + b);
bar((a + 2) + b);
}
to
void foo(int a, int b) {
int t = a + b;
bar(t);
bar(t + 2);
}
saving one add instruction.
Fixes PR22357 (https://llvm.org/bugs/show_bug.cgi?id=22357).
Test Plan: nary-add.ll
Reviewers: broune, dberlin, hfinkel, meheff, sanjoy, atrick
Reviewed By: sanjoy, atrick
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D8950
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@234855 91177308-0d34-0410-b5e6-96231b3b80d8
This re-adds float2int to the tree, after fixing PR23038. It turns
out the argument to APSInt() is true-if-unsigned, rather than
true-if-signed :(. Added testcase and explanatory comment.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@233370 91177308-0d34-0410-b5e6-96231b3b80d8
This caused PR23008, compiles failing with: "Use still stuck around after Def is
destroyed: %.sroa.speculated"
Also reverting follow-up r233064.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@233105 91177308-0d34-0410-b5e6-96231b3b80d8
It is possible to have code that converts from integer to float, performs operations then converts back, and the result is provably the same as if integers were used.
This can come from different sources, but the most obvious is a helper function that uses floats but the arguments given at an inlined callsites are integers.
This pass considers all integers requiring a bitwidth less than or equal to the bitwidth of the mantissa of a floating point type (23 for floats, 52 for doubles) as exactly representable in floating point.
To reduce the risk of harming efficient code, the pass only attempts to perform complete removal of inttofp/fptoint operations, not just move them around.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@233062 91177308-0d34-0410-b5e6-96231b3b80d8
This pass interchanges loops to provide a more cache-friendly memory access.
For e.g. given a loop like -
for(int i=0;i<N;i++)
for(int j=0;j<N;j++)
A[j][i] = A[j][i]+B[j][i];
is interchanged to -
for(int j=0;j<N;j++)
for(int i=0;i<N;i++)
A[j][i] = A[j][i]+B[j][i];
This pass is currently disabled by default.
To give a brief introduction it consists of 3 stages-
LoopInterchangeLegality : Checks the legality of loop interchange based on Dependency matrix.
LoopInterchangeProfitability: A very basic heuristic has been added to check for profitibility. This will evolve over time.
LoopInterchangeTransform : Which does the actual transform.
LNT Performance tests shows improvement in Polybench/linear-algebra/kernels/mvt and Polybench/linear-algebra/kernels/gemver becnmarks.
TODO:
1) Add support for reductions and lcssa phi.
2) Improve profitability model.
3) Improve loop selection algorithm to select best loop for interchange. Currently the innermost loop is selected for interchange.
4) Improve compile time regression found in llvm lnt due to this pass.
5) Fix issues in Dependency Analysis module.
A special thanks to Hal for reviewing this code.
Review: http://reviews.llvm.org/D7499
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@231458 91177308-0d34-0410-b5e6-96231b3b80d8
It was previously using the subtarget to get values for the global
offset without actually checking each function as it was generating
code. Go ahead and solidify the current behavior and make the
existing FIXMEs more prominent.
As a note the ARM backend previously had a thumb1 and non-thumb1
set of defaults. Only the former was tested so I've changed the
behavior to only use that for now.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@230245 91177308-0d34-0410-b5e6-96231b3b80d8
This patch consists of a single pass whose only purpose is to visit previous inserted gc.statepoints which do not have gc.relocates inserted yet, and insert them. This can be used either immediately after IR generation to perform 'early safepoint insertion' or late in the pass order to perform 'late insertion'.
This patch is setting the stage for work to continue in tree. In particular, there are known naming and style violations in the current patch. I'll try to get those resolved over the next week or so. As I touch each area to make style changes, I need to make sure we have adequate testing in place. As part of the cleanup, I will be cleaning up a collection of test cases we have out of tree and submitting them upstream. The tests included in this change are very basic and mostly to provide examples of usage.
The pass has several main subproblems it needs to address:
- First, it has identify any live pointers. In the current code, the use of address spaces to distinguish pointers to GC managed objects is hard coded, but this will become parametrizable in the near future. Note that the current change doesn't actually contain a useful liveness analysis. It was seperated into a followup change as the code wasn't ready to be shared. Instead, the current implementation just considers any dominating def of appropriate pointer type to be live.
- Second, it has to identify base pointers for each live pointer. This is a fairly straight forward data flow algorithm.
- Third, the information in the previous steps is used to actually introduce rewrites. Rather than trying to do this by hand, we simply re-purpose the code behind Mem2Reg to do this for us.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@229945 91177308-0d34-0410-b5e6-96231b3b80d8
BDCE is a bit-tracking dead code elimination pass. It is based on ADCE (the
"aggressive DCE" pass), with the added capability to track dead bits of integer
valued instructions and remove those instructions when all of the bits are
dead.
Currently, it does not actually do this all-bits-dead removal, but rather
replaces the instruction's uses with a constant zero, and lets instcombine (and
the later run of ADCE) do the rest. Because we essentially get a run of ADCE
"for free" while tracking the dead bits, we also do what ADCE does and removes
actually-dead instructions as well (this includes instructions newly trivially
dead because all bits were dead, but not all such instructions can be removed).
The motivation for this is a case like:
int __attribute__((const)) foo(int i);
int bar(int x) {
x |= (4 & foo(5));
x |= (8 & foo(3));
x |= (16 & foo(2));
x |= (32 & foo(1));
x |= (64 & foo(0));
x |= (128& foo(4));
return x >> 4;
}
As it turns out, if you order the bit-field insertions so that all of the dead
ones come last, then instcombine will remove them. However, if you pick some
other order (such as the one above), the fact that some of the calls to foo()
are useless is not locally obvious, and we don't remove them (without this
pass).
I did a quick compile-time overhead check using sqlite from the test suite
(Release+Asserts). BDCE took ~0.4% of the compilation time (making it about
twice as expensive as ADCE).
I've not looked at why yet, but we eliminate instructions due to having
all-dead bits in:
External/SPEC/CFP2006/447.dealII/447.dealII
External/SPEC/CINT2006/400.perlbench/400.perlbench
External/SPEC/CINT2006/403.gcc/403.gcc
MultiSource/Applications/ClamAV/clamscan
MultiSource/Benchmarks/7zip/7zip-benchmark
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@229462 91177308-0d34-0410-b5e6-96231b3b80d8
This pass is responsible for figuring out where to place call safepoints and safepoint polls. It doesn't actually make the relocations explicit; that's the job of the RewriteStatepointsForGC pass (http://reviews.llvm.org/D6975).
Note that this code is not yet finalized. Its moving in tree for incremental development, but further cleanup is needed and will happen over the next few days. It is not yet part of the standard pass order.
Planned changes in the near future:
- I plan on restructuring the statepoint rewrite to use the functions add to the IRBuilder a while back.
- In the current pass, the function "gc.safepoint_poll" is treated specially but is not an intrinsic. I plan to make identifying the poll function a property of the GCStrategy at some point in the near future.
- As follow on patches, I will be separating a collection of test cases we have out of tree and submitting them upstream.
- It's not explicit in the code, but these two patches are introducing a new state for a statepoint which looks a lot like a patchpoint. There's no a transient form which doesn't yet have the relocations explicitly represented, but does prevent reordering of memory operations. Once this is in, I need to update actually make this explicit by reserving the 'unused' argument of the statepoint as a flag, updating the docs, and making the code explicitly check for such a thing. This wasn't really planned, but once I split the two passes - which was done for other reasons - the intermediate state fell out. Just reminds us once again that we need to merge statepoints and patchpoints at some point in the not that distant future.
Future directions planned:
- Identifying more cases where a backedge safepoint isn't required to ensure timely execution of a safepoint poll.
- Tweaking the insertion process to generate easier to optimize IR. (For example, investigating making SplitBackedge) the default.
- Adding opt-in flags for a GCStrategy to use this pass. Once done, add this pass to the actual pass ordering.
Differential Revision: http://reviews.llvm.org/D6981
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@228090 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
Straight-line strength reduction (SLSR) is implemented in GCC but not yet in
LLVM. It has proven to effectively simplify statements derived from an unrolled
loop, and can potentially benefit many other cases too. For example,
LLVM unrolls
#pragma unroll
foo (int i = 0; i < 3; ++i) {
sum += foo((b + i) * s);
}
into
sum += foo(b * s);
sum += foo((b + 1) * s);
sum += foo((b + 2) * s);
However, no optimizations yet reduce the internal redundancy of the three
expressions:
b * s
(b + 1) * s
(b + 2) * s
With SLSR, LLVM can optimize these three expressions into:
t1 = b * s
t2 = t1 + s
t3 = t2 + s
This commit is only an initial step towards implementing a series of such
optimizations. I will implement more (see TODO in the file commentary) in the
near future. This optimization is enabled for the NVPTX backend for now.
However, I am more than happy to push it to the standard optimization pipeline
after more thorough performance tests.
Test Plan: test/StraightLineStrengthReduce/slsr.ll
Reviewers: eliben, HaoLiu, meheff, hfinkel, jholewinski, atrick
Reviewed By: jholewinski, atrick
Subscribers: karthikthecool, jholewinski, llvm-commits
Differential Revision: http://reviews.llvm.org/D7310
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@228016 91177308-0d34-0410-b5e6-96231b3b80d8
IRCE eliminates range checks of the form
0 <= A * I + B < Length
by splitting a loop's iteration space into three segments in a way
that the check is completely redundant in the middle segment. As an
example, IRCE will convert
len = < known positive >
for (i = 0; i < n; i++) {
if (0 <= i && i < len) {
do_something();
} else {
throw_out_of_bounds();
}
}
to
len = < known positive >
limit = smin(n, len)
// no first segment
for (i = 0; i < limit; i++) {
if (0 <= i && i < len) { // this check is fully redundant
do_something();
} else {
throw_out_of_bounds();
}
}
for (i = limit; i < n; i++) {
if (0 <= i && i < len) {
do_something();
} else {
throw_out_of_bounds();
}
}
IRCE can deal with multiple range checks in the same loop (it takes
the intersection of the ranges that will make each of them redundant
individually).
Currently IRCE does not do any profitability analysis. That is a
TODO.
Please note that the status of this pass is *experimental*, and it is
not part of any default pass pipeline. Having said that, I will love
to get feedback and general input from people interested in trying
this out.
This pass was originally r226201. It was reverted because it used C++
features not supported by MSVC 2012.
Differential Revision: http://reviews.llvm.org/D6693
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226238 91177308-0d34-0410-b5e6-96231b3b80d8
The change used C++11 features not supported by MSVC 2012. I will fix
the change to use things supported MSVC 2012 and recommit shortly.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226216 91177308-0d34-0410-b5e6-96231b3b80d8
IRCE eliminates range checks of the form
0 <= A * I + B < Length
by splitting a loop's iteration space into three segments in a way
that the check is completely redundant in the middle segment. As an
example, IRCE will convert
len = < known positive >
for (i = 0; i < n; i++) {
if (0 <= i && i < len) {
do_something();
} else {
throw_out_of_bounds();
}
}
to
len = < known positive >
limit = smin(n, len)
// no first segment
for (i = 0; i < limit; i++) {
if (0 <= i && i < len) { // this check is fully redundant
do_something();
} else {
throw_out_of_bounds();
}
}
for (i = limit; i < n; i++) {
if (0 <= i && i < len) {
do_something();
} else {
throw_out_of_bounds();
}
}
IRCE can deal with multiple range checks in the same loop (it takes
the intersection of the ranges that will make each of them redundant
individually).
Currently IRCE does not do any profitability analysis. That is a
TODO.
Please note that the status of this pass is *experimental*, and it is
not part of any default pass pipeline. Having said that, I will love
to get feedback and general input from people interested in trying
this out.
Differential Revision: http://reviews.llvm.org/D6693
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226201 91177308-0d34-0410-b5e6-96231b3b80d8
If LowerGEP is enabled, it can lower a GEP with multiple indices into GEPs with a single index
or arithmetic operations. Lowering GEPs can always extract structure indices. Lowering GEPs can
also give use more optimization opportunities. It can benefit passes like CSE, LICM and CGP.
Reviewed in http://reviews.llvm.org/D5864
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@222328 91177308-0d34-0410-b5e6-96231b3b80d8
Summary:
This patch adds a threshold that controls the number of bonus instructions
allowed for folding branches with common destination. The original code allows
at most one bonus instruction. With this patch, users can customize the
threshold to allow multiple bonus instructions. The default threshold is still
1, so that the code behaves the same as before when users do not specify this
threshold.
The motivation of this change is that tuning this threshold significantly (up
to 25%) improves the performance of some CUDA programs in our internal code
base. In general, branch instructions are very expensive for GPU programs.
Therefore, it is sometimes worth trading more arithmetic computation for a more
straightened control flow. Here's a reduced example:
__global__ void foo(int a, int b, int c, int d, int e, int n,
const int *input, int *output) {
int sum = 0;
for (int i = 0; i < n; ++i)
sum += (((i ^ a) > b) && (((i | c ) ^ d) > e)) ? 0 : input[i];
*output = sum;
}
The select statement in the loop body translates to two branch instructions "if
((i ^ a) > b)" and "if (((i | c) ^ d) > e)" which share a common destination.
With the default threshold, SimplifyCFG is unable to fold them, because
computing the condition of the second branch "(i | c) ^ d > e" requires two
bonus instructions. With the threshold increased, SimplifyCFG can fold the two
branches so that the loop body contains only one branch, making the code
conceptually look like:
sum += (((i ^ a) > b) & (((i | c ) ^ d) > e)) ? 0 : input[i];
Increasing the threshold significantly improves the performance of this
particular example. In the configuration where both conditions are guaranteed
to be true, increasing the threshold from 1 to 2 improves the performance by
18.24%. Even in the configuration where the first condition is false and the
second condition is true, which favors shortcuts, increasing the threshold from
1 to 2 still improves the performance by 4.35%.
We are still looking for a good threshold and maybe a better cost model than
just counting the number of bonus instructions. However, according to the above
numbers, we think it is at least worth adding a threshold to enable more
experiments and tuning. Let me know what you think. Thanks!
Test Plan: Added one test case to check the threshold is in effect
Reviewers: nadav, eliben, meheff, resistor, hfinkel
Reviewed By: hfinkel
Subscribers: hfinkel, llvm-commits
Differential Revision: http://reviews.llvm.org/D5529
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@218711 91177308-0d34-0410-b5e6-96231b3b80d8
- BB duplication may not be desired on targets where there is no or small
branch penalty and code duplication needs restrict control.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@218375 91177308-0d34-0410-b5e6-96231b3b80d8
This adds a ScalarEvolution-powered transformation that updates load, store and
memory intrinsic pointer alignments based on invariant((a+q) & b == 0)
expressions. Many of the simple cases we can get with ValueTracking, but we
still need something like this for the more complicated cases (such as those
with an offset) that require some algebra. Note that gcc's
__builtin_assume_aligned's optional third argument provides exactly for this
kind of 'misalignment' offset for which this kind of logic is necessary.
The primary motivation is to fixup alignments for vector loads/stores after
vectorization (and unrolling). This pass is added to the optimization pipeline
just after the SLP vectorizer runs (which, admittedly, does not preserve SE,
although I imagine it could). Regardless, I actually don't think that the
preservation matters too much in this case: SE computes lazily, and this pass
won't issue any SE queries unless there are any assume intrinsics, so there
should be no real additional cost in the common case (SLP does preserve DT and
LoopInfo).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@217344 91177308-0d34-0410-b5e6-96231b3b80d8
Merges equivalent loads on both sides of a hammock/diamond
and hoists into into the header.
Merges equivalent stores on both sides of a hammock/diamond
and sinks it to the footer.
Can enable if conversion and tolerate better load misses
and store operand latencies.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@213396 91177308-0d34-0410-b5e6-96231b3b80d8
and via the command line, mirroring similar functionality in LoopUnroll. In
situations where clients used custom unrolling thresholds, their intent could
previously be foiled by LoopRotate having a hardcoded threshold.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@209617 91177308-0d34-0410-b5e6-96231b3b80d8
This optimization merges the common part of a group of GEPs, so we can compute
each pointer address by adding a simple offset to the common part.
The optimization is currently only enabled for the NVPTX backend, where it has
a large payoff on some benchmarks.
Review: http://reviews.llvm.org/D3462
Patch by Jingyue Wu.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@207783 91177308-0d34-0410-b5e6-96231b3b80d8
The generic (concatenation) loop unroller is currently placed early in the
standard optimization pipeline. This is a good place to perform full unrolling,
but not the right place to perform partial/runtime unrolling. However, most
targets don't enable partial/runtime unrolling, so this never mattered.
However, even some x86 cores benefit from partial/runtime unrolling of very
small loops, and follow-up commits will enable this. First, we need to move
partial/runtime unrolling late in the optimization pipeline (importantly, this
is after SLP and loop vectorization, as vectorization can drastically change
the size of a loop), while keeping the full unrolling where it is now. This
change does just that.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@205264 91177308-0d34-0410-b5e6-96231b3b80d8
This option caused LowerInvoke to generate code using SJLJ-based
exception handling, but there is no code left that interprets the
jmp_buf stack that the resulting code maintained (llvm.sjljeh.jblist).
This option has been obsolete for a while, and replaced by
SjLjEHPrepare.
This leaves the default behaviour of LowerInvoke, which is to convert
invokes to calls.
Differential Revision: http://llvm-reviews.chandlerc.com/D3136
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@204388 91177308-0d34-0410-b5e6-96231b3b80d8
DWARF discriminators are used to distinguish multiple control flow paths
on the same source location. When this happens, instructions across
basic block boundaries will share the same debug location.
This pass detects this situation and creates a new lexical scope to one
of the two instructions. This lexical scope is a child scope of the
original and contains a new discriminator value. This discriminator is
then picked up from MCObjectStreamer::EmitDwarfLocDirective to be
written on the object file.
This fixes http://llvm.org/bugs/show_bug.cgi?id=18270.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202752 91177308-0d34-0410-b5e6-96231b3b80d8
CodeGenPrepare uses extensively TargetLowering which is part of libLLVMCodeGen.
This is a layer violation which would introduce eventually a dependence on
CodeGen in ScalarOpts.
Move CodeGenPrepare into libLLVMCodeGen to avoid that.
Follow-up of <rdar://problem/15519855>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201912 91177308-0d34-0410-b5e6-96231b3b80d8
This reverts commit r200058 and adds the using directive for
ARMTargetTransformInfo to silence two g++ overload warnings.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200062 91177308-0d34-0410-b5e6-96231b3b80d8
This commit caused -Woverloaded-virtual warnings. The two new
TargetTransformInfo::getIntImmCost functions were only added to the superclass,
and to the X86 subclass. The other targets were not updated, and the
warning highlighted this by pointing out that e.g. ARMTTI::getIntImmCost was
hiding the two new getIntImmCost variants.
We could pacify the warning by adding "using TargetTransformInfo::getIntImmCost"
to the various subclasses, or turning it off, but I suspect that it's wrong to
leave the functions unimplemnted in those targets. The default implementations
return TCC_Free, which I don't think is right e.g. for ARM.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200058 91177308-0d34-0410-b5e6-96231b3b80d8
Retry commit r200022 with a fix for the build bot errors. Constant expressions
have (unlike instructions) module scope use lists and therefore may have users
in different functions. The fix is to simply ignore these out-of-function uses.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200034 91177308-0d34-0410-b5e6-96231b3b80d8
This pass identifies expensive constants to hoist and coalesces them to
better prepare it for SelectionDAG-based code generation. This works around the
limitations of the basic-block-at-a-time approach.
First it scans all instructions for integer constants and calculates its
cost. If the constant can be folded into the instruction (the cost is
TCC_Free) or the cost is just a simple operation (TCC_BASIC), then we don't
consider it expensive and leave it alone. This is the default behavior and
the default implementation of getIntImmCost will always return TCC_Free.
If the cost is more than TCC_BASIC, then the integer constant can't be folded
into the instruction and it might be beneficial to hoist the constant.
Similar constants are coalesced to reduce register pressure and
materialization code.
When a constant is hoisted, it is also hidden behind a bitcast to force it to
be live-out of the basic block. Otherwise the constant would be just
duplicated and each basic block would have its own copy in the SelectionDAG.
The SelectionDAG recognizes such constants as opaque and doesn't perform
certain transformations on them, which would create a new expensive constant.
This optimization is only applied to integer constants in instructions and
simple (this means not nested) constant cast experessions. For example:
%0 = load i64* inttoptr (i64 big_constant to i64*)
Reviewed by Eric
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@200022 91177308-0d34-0410-b5e6-96231b3b80d8
This adds a loop rerolling pass: the opposite of (partial) loop unrolling. The
transformation aims to take loops like this:
for (int i = 0; i < 3200; i += 5) {
a[i] += alpha * b[i];
a[i + 1] += alpha * b[i + 1];
a[i + 2] += alpha * b[i + 2];
a[i + 3] += alpha * b[i + 3];
a[i + 4] += alpha * b[i + 4];
}
and turn them into this:
for (int i = 0; i < 3200; ++i) {
a[i] += alpha * b[i];
}
and loops like this:
for (int i = 0; i < 500; ++i) {
x[3*i] = foo(0);
x[3*i+1] = foo(0);
x[3*i+2] = foo(0);
}
and turn them into this:
for (int i = 0; i < 1500; ++i) {
x[i] = foo(0);
}
There are two motivations for this transformation:
1. Code-size reduction (especially relevant, obviously, when compiling for
code size).
2. Providing greater choice to the loop vectorizer (and generic unroller) to
choose the unrolling factor (and a better ability to vectorize). The loop
vectorizer can take vector lengths and register pressure into account when
choosing an unrolling factor, for example, and a pre-unrolled loop limits that
choice. This is especially problematic if the manual unrolling was optimized
for a machine different from the current target.
The current implementation is limited to single basic-block loops only. The
rerolling recognition should work regardless of how the loop iterations are
intermixed within the loop body (subject to dependency and side-effect
constraints), but the significant restriction is that the order of the
instructions in each iteration must be identical. This seems sufficient to
capture all current use cases.
This pass is not currently enabled by default at any optimization level.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194939 91177308-0d34-0410-b5e6-96231b3b80d8
This adds a new scalar pass that reads a file with samples generated
by 'perf' during runtime. The samples read from the profile are
incorporated and emmited as IR metadata reflecting that profile.
The profile file is assumed to have been generated by an external
profile source. The profile information is converted into IR metadata,
which is later used by the analysis routines to estimate block
frequencies, edge weights and other related data.
External profile information files have no fixed format, each profiler
is free to define its own. This includes both the on-disk representation
of the profile and the kind of profile information stored in the file.
A common kind of profile is based on sampling (e.g., perf), which
essentially counts how many times each line of the program has been
executed during the run.
The SampleProfileLoader pass is organized as a scalar transformation.
On startup, it reads the file given in -sample-profile-file to
determine what kind of profile it contains. This file is assumed to
contain profile information for the whole application. The profile
data in the file is read and incorporated into the internal state of
the corresponding profiler.
To facilitate testing, I've organized the profilers to support two file
formats: text and native. The native format is whatever on-disk
representation the profiler wants to support, I think this will mostly
be bitcode files, but it could be anything the profiler wants to
support. To do this, every profiler must implement the
SampleProfile::loadNative() function.
The text format is mostly meant for debugging. Records are separated by
newlines, but each profiler is free to interpret records as it sees fit.
Profilers must implement the SampleProfile::loadText() function.
Finally, the pass will call SampleProfile::emitAnnotations() for each
function in the current translation unit. This function needs to
translate the loaded profile into IR metadata, which the analyzer will
later be able to use.
This patch implements the first steps towards the above design. I've
implemented a sample-based flat profiler. The format of the profile is
fairly simplistic. Each sampled function contains a list of relative
line locations (from the start of the function) together with a count
representing how many samples were collected at that line during
execution. I generate this profile using perf and a separate converter
tool.
Currently, I have only implemented a text format for these profiles. I
am interested in initial feedback to the whole approach before I send
the other parts of the implementation for review.
This patch implements:
- The SampleProfileLoader pass.
- The base ExternalProfile class with the core interface.
- A SampleProfile sub-class using the above interface. The profiler
generates branch weight metadata on every branch instructions that
matches the profiles.
- A text loader class to assist the implementation of
SampleProfile::loadText().
- Basic unit tests for the pass.
Additionally, the patch uses profile information to compute branch
weights based on instruction samples.
This patch converts instruction samples into branch weights. It
does a fairly simplistic conversion:
Given a multi-way branch instruction, it calculates the weight of
each branch based on the maximum sample count gathered from each
target basic block.
Note that this assignment of branch weights is somewhat lossy and can be
misleading. If a basic block has more than one incoming branch, all the
incoming branches will get the same weight. In reality, it may be that
only one of them is the most heavily taken branch.
I will adjust this assignment in subsequent patches.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194566 91177308-0d34-0410-b5e6-96231b3b80d8
As with the other loop unrolling parameters (the unrolling threshold, partial
unrolling, etc.) runtime unrolling can now also be controlled via the
constructor. This will be necessary for moving non-trivial unrolling late in
the pass manager (after loop vectorization).
No functionality change intended.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194027 91177308-0d34-0410-b5e6-96231b3b80d8
This pass was based on the previous (essentially unused) profiling
infrastructure and the assumption that by ordering the basic blocks at
the IR level in a particular way, the correct layout would happen in the
end. This sometimes worked, and mostly didn't. It also was a really
naive implementation of the classical paper that dates from when branch
predictors were primarily directional and when loop structure wasn't
commonly available. It also didn't factor into the equation
non-fallthrough branches and other machine level details.
Anyways, for all of these reasons and more, I wrote
MachineBlockPlacement, which completely supercedes this pass. It both
uses modern profile information infrastructure, and actually works. =]
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190748 91177308-0d34-0410-b5e6-96231b3b80d8
...so that it can be used for z too. Most of the code is the same.
The only real change is to use TargetTransformInfo to test when a sqrt
instruction is available.
The pass is opt-in because at the moment it only handles sqrt.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@189097 91177308-0d34-0410-b5e6-96231b3b80d8
