The combine that forms extloads used to be disabled on vector types,
because "None of the supported targets knows how to perform load and
sign extend on vectors in one instruction."
That's not entirely true, since at least SSE4.1 X86 knows how to do
those sextloads/zextloads (with PMOVS/ZX).
But there are several aspects to getting this right.
First, vector extloads are controlled by a profitability callback.
For instance, on ARM, several instructions have folded extload forms,
so it's not always beneficial to create an extload node (and trying to
match extloads is a whole 'nother can of worms).
The interesting optimization enables folding of s/zextloads to illegal
(splittable) vector types, expanding them into smaller legal extloads.
It's not ideal (it introduces some legalization-like behavior in the
combine) but it's better than the obvious alternative: form illegal
extloads, and later try to split them up. If you do that, you might
generate extloads that can't be split up, but have a valid ext+load
expansion. At vector-op legalization time, it's too late to generate
this kind of code, so you end up forced to scalarize. It's better to
just avoid creating egregiously illegal nodes.
This optimization is enabled unconditionally on X86.
Note that the splitting combine is happy with "custom" extloads. As
is, this bypasses the actual custom lowering, and just unrolls the
extload. But from what I've seen, this is still much better than the
current custom lowering, which does some kind of unrolling at the end
anyway (see for instance load_sext_4i8_to_4i64 on SSE2, and the added
FIXME).
Also note that the existing combine that forms extloads is now also
enabled on legal vectors. This doesn't have a big effect on X86
(because sext+load is usually combined to sext_inreg+aextload).
On ARM it fires on some rare occasions; that's for a separate commit.
Differential Revision: http://reviews.llvm.org/D6904
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@228325 91177308-0d34-0410-b5e6-96231b3b80d8
The node is still defined oddly so that the
address spaces are not operands and not accessible
from tablegen, but as-is this can now be used to write
a ComplexPattern with an addrspacecast root node.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@228270 91177308-0d34-0410-b5e6-96231b3b80d8
This moves the transformation introduced in r223757 into a separate MI pass.
This allows it to cover many more cases (not only cases where there must be a
reserved call frame), and perform rudimentary call folding. It still doesn't
have a heuristic, so it is enabled only for optsize/minsize, with stack
alignment <= 8, where it ought to be a fairly clear win.
(Re-commit of r227728)
Differential Revision: http://reviews.llvm.org/D6789
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227752 91177308-0d34-0410-b5e6-96231b3b80d8
TargetIRAnalysis access path directly rather than implementing getTTI.
This even removes getTTI from the interface. It's more efficient for
each target to just register a precise callback that creates their
specific TTI.
As part of this, all of the targets which are building their subtargets
individually per-function now build their TTI instance with the function
and thus look up the correct subtarget and cache it. NVPTX, R600, and
XCore currently don't leverage this functionality, but its trivial for
them to add it now.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227735 91177308-0d34-0410-b5e6-96231b3b80d8
This moves the transformation introduced in r223757 into a separate MI pass.
This allows it to cover many more cases (not only cases where there must be a
reserved call frame), and perform rudimentary call folding. It still doesn't
have a heuristic, so it is enabled only for optsize/minsize, with stack
alignment <= 8, where it ought to be a fairly clear win.
Differential Revision: http://reviews.llvm.org/D6789
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227728 91177308-0d34-0410-b5e6-96231b3b80d8
produce it.
This adds a function to the TargetMachine that produces this analysis
via a callback for each function. This in turn faves the way to produce
a *different* TTI per-function with the correct subtarget cached.
I've also done the necessary wiring in the opt tool to thread the target
machine down and make it available to the pass registry so that we can
construct this analysis from a target machine when available.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227721 91177308-0d34-0410-b5e6-96231b3b80d8
base which it adds a single analysis pass to, to instead return the type
erased TargetTransformInfo object constructed for that TargetMachine.
This removes all of the pass variants for TTI. There is now a single TTI
*pass* in the Analysis layer. All of the Analysis <-> Target
communication is through the TTI's type erased interface itself. While
the diff is large here, it is nothing more that code motion to make
types available in a header file for use in a different source file
within each target.
I've tried to keep all the doxygen comments and file boilerplate in line
with this move, but let me know if I missed anything.
With this in place, the next step to making TTI work with the new pass
manager is to introduce a really simple new-style analysis that produces
a TTI object via a callback into this routine on the target machine.
Once we have that, we'll have the building blocks necessary to accept
a function argument as well.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227685 91177308-0d34-0410-b5e6-96231b3b80d8
type erased interface and a single analysis pass rather than an
extremely complex analysis group.
The end result is that the TTI analysis can contain a type erased
implementation that supports the polymorphic TTI interface. We can build
one from a target-specific implementation or from a dummy one in the IR.
I've also factored all of the code into "mix-in"-able base classes,
including CRTP base classes to facilitate calling back up to the most
specialized form when delegating horizontally across the surface. These
aren't as clean as I would like and I'm planning to work on cleaning
some of this up, but I wanted to start by putting into the right form.
There are a number of reasons for this change, and this particular
design. The first and foremost reason is that an analysis group is
complete overkill, and the chaining delegation strategy was so opaque,
confusing, and high overhead that TTI was suffering greatly for it.
Several of the TTI functions had failed to be implemented in all places
because of the chaining-based delegation making there be no checking of
this. A few other functions were implemented with incorrect delegation.
The message to me was very clear working on this -- the delegation and
analysis group structure was too confusing to be useful here.
The other reason of course is that this is *much* more natural fit for
the new pass manager. This will lay the ground work for a type-erased
per-function info object that can look up the correct subtarget and even
cache it.
Yet another benefit is that this will significantly simplify the
interaction of the pass managers and the TargetMachine. See the future
work below.
The downside of this change is that it is very, very verbose. I'm going
to work to improve that, but it is somewhat an implementation necessity
in C++ to do type erasure. =/ I discussed this design really extensively
with Eric and Hal prior to going down this path, and afterward showed
them the result. No one was really thrilled with it, but there doesn't
seem to be a substantially better alternative. Using a base class and
virtual method dispatch would make the code much shorter, but as
discussed in the update to the programmer's manual and elsewhere,
a polymorphic interface feels like the more principled approach even if
this is perhaps the least compelling example of it. ;]
Ultimately, there is still a lot more to be done here, but this was the
huge chunk that I couldn't really split things out of because this was
the interface change to TTI. I've tried to minimize all the other parts
of this. The follow up work should include at least:
1) Improving the TargetMachine interface by having it directly return
a TTI object. Because we have a non-pass object with value semantics
and an internal type erasure mechanism, we can narrow the interface
of the TargetMachine to *just* do what we need: build and return
a TTI object that we can then insert into the pass pipeline.
2) Make the TTI object be fully specialized for a particular function.
This will include splitting off a minimal form of it which is
sufficient for the inliner and the old pass manager.
3) Add a new pass manager analysis which produces TTI objects from the
target machine for each function. This may actually be done as part
of #2 in order to use the new analysis to implement #2.
4) Work on narrowing the API between TTI and the targets so that it is
easier to understand and less verbose to type erase.
5) Work on narrowing the API between TTI and its clients so that it is
easier to understand and less verbose to forward.
6) Try to improve the CRTP-based delegation. I feel like this code is
just a bit messy and exacerbating the complexity of implementing
the TTI in each target.
Many thanks to Eric and Hal for their help here. I ended up blocked on
this somewhat more abruptly than I expected, and so I appreciate getting
it sorted out very quickly.
Differential Revision: http://reviews.llvm.org/D7293
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227669 91177308-0d34-0410-b5e6-96231b3b80d8
incarnation of target transform info.
This is in preparation for starting to redesign TTI to be amenable to
the new PM world.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227525 91177308-0d34-0410-b5e6-96231b3b80d8
derived classes.
Since global data alignment, layout, and mangling is often based on the
DataLayout, move it to the TargetMachine. This ensures that global
data is going to be layed out and mangled consistently if the subtarget
changes on a per function basis. Prior to this all targets(*) have
had subtarget dependent code moved out and onto the TargetMachine.
*One target hasn't been migrated as part of this change: R600. The
R600 port has, as a subtarget feature, the size of pointers and
this affects global data layout. I've currently hacked in a FIXME
to enable progress, but the port needs to be updated to either pass
the 64-bitness to the TargetMachine, or fix the DataLayout to
avoid subtarget dependent features.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227113 91177308-0d34-0410-b5e6-96231b3b80d8
While the term "Target" is in the name, it doesn't really have to do
with the LLVM Target library -- this isn't an abstraction which LLVM
targets generally need to implement or extend. It has much more to do
with modeling the various runtime libraries on different OSes and with
different runtime environments. The "target" in this sense is the more
general sense of a target of cross compilation.
This is in preparation for porting this analysis to the new pass
manager.
No functionality changed, and updates inbound for Clang and Polly.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226078 91177308-0d34-0410-b5e6-96231b3b80d8
utils/sort_includes.py.
I clearly haven't done this in a while, so more changed than usual. This
even uncovered a missing include from the InstrProf library that I've
added. No functionality changed here, just mechanical cleanup of the
include order.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@225974 91177308-0d34-0410-b5e6-96231b3b80d8
A pass that adds random noops to X86 binaries to introduce diversity with the goal of increasing security against most return-oriented programming attacks.
Command line options:
-noop-insertion // Enable noop insertion.
-noop-insertion-percentage=X // X% of assembly instructions will have a noop prepended (default: 50%, requires -noop-insertion)
-max-noops-per-instruction=X // Randomly generate X noops per instruction. ie. roll the dice X times with probability set above (default: 1). This doesn't guarantee X noop instructions.
In addition, the following 'quick switch' in clang enables basic diversity using default settings (currently: noop insertion and schedule randomization; it is intended to be extended in the future).
-fdiversify
This is the llvm part of the patch.
clang part: D3393
http://reviews.llvm.org/D3392
Patch by Stephen Crane (@rinon)
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@225908 91177308-0d34-0410-b5e6-96231b3b80d8
This requires a new hook to prevent expanding sqrt in terms
of rsqrt and reciprocal. v_rcp_f32, v_rsq_f32, and v_sqrt_f32 are
all the same rate, so this expansion would just double the number
of instructions and cycles.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@225828 91177308-0d34-0410-b5e6-96231b3b80d8
While, generally speaking, the process of lowering arguments for a patchpoint
is the same as lowering a regular indirect call, on some targets it may not be
exactly the same. Targets may not, for example, want to add additional register
dependencies that apply only to making cross-DSO calls through linker stubs,
may not want to load additional registers out of function descriptors, and may
not want to add additional side-effect-causing instructions that cannot be
removed later with the call itself being generated.
The PowerPC target will use this in a future commit (for all of the reasons
stated above).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@225806 91177308-0d34-0410-b5e6-96231b3b80d8
Some targets, PowerPC for example, have pseudo-registers (such as that used to
represent the rounding mode), that don't have DWARF register numbers or a
register class. These are used only for internal dependency tracking, and
should not appear in the recorded live-outs. This adds a callback allowing the
target to pre-process the live-out mask in order to remove these kinds of
registers so that the StackMaps code does not complain about them and/or
attempt to include them in the output.
This will be used by the PowerPC target in a future commit.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@225805 91177308-0d34-0410-b5e6-96231b3b80d8
Peephole optimizer is scanning a basic block forward. At some point it
needs to answer the question "given a pointer to an MI in the current
BB, is it located before or after the current instruction".
To perform this, it keeps a set of the MIs already seen during the scan,
if a MI is not in the set, it is assumed to be after.
It means that newly created MIs have to be inserted in the set as well.
This commit passes the set as an argument to the target-dependent
optimizeSelect() so that it can properly update the set with the
(potentially) newly created MIs.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@225772 91177308-0d34-0410-b5e6-96231b3b80d8
These intrinsics allow multiple functions to share a single stack
allocation from one function's call frame. The function with the
allocation may only perform one allocation, and it must be in the entry
block.
Functions accessing the allocation call llvm.recoverframeallocation with
the function whose frame they are accessing and a frame pointer from an
active call frame of that function.
These intrinsics are very difficult to inline correctly, so the
intention is that they be introduced rarely, or at least very late
during EH preparation.
Reviewers: echristo, andrew.w.kaylor
Differential Revision: http://reviews.llvm.org/D6493
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@225746 91177308-0d34-0410-b5e6-96231b3b80d8
This adds two new fields to the RegisterOperand TableGen class:
string OperandNamespace = "MCOI";
string OperandType = "OPERAND_REGISTER";
These fields can be used to specify a target specific operand type,
which will be stored in the OperandType member of the MCOperandInfo
object.
This can be useful for targets that need to store some extra information
about operands that cannot be expressed using the target independent
types. For example, in the R600 backend, there are operands which
can take either registers or immediates and it is convenient to be able
to specify this in the TableGen definitions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@225661 91177308-0d34-0410-b5e6-96231b3b80d8
PEI tries to keep track of how much starting or ending a call sequence adjusts the stack pointer by, so that it can resolve frame-index references. Currently, it takes a very simplistic view of how SP adjustments are done - both FrameStartOpcode and FrameDestroyOpcode adjust it exactly by the amount written in its first argument.
This view is in fact incorrect for some targets (e.g. due to stack re-alignment, or because it may want to adjust the stack pointer in multiple steps). However, that doesn't cause breakage, because most targets (the only in-tree exception appears to be 32-bit ARM) rely on being able to simplify the call frame pseudo-instructions earlier, so this code is never hit.
Moving the computation into TargetInstrInfo allows targets to override the way the adjustment is computed if they need to have a non-zero SPAdj.
Differential Revision: http://reviews.llvm.org/D6863
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@225437 91177308-0d34-0410-b5e6-96231b3b80d8
type (in addition to the memory type).
The *LoadExt* legalization handling used to only have one type, the
memory type. This forced users to assume that as long as the extload
for the memory type was declared legal, and the result type was legal,
the whole extload was legal.
However, this isn't always the case. For instance, on X86, with AVX,
this is legal:
v4i32 load, zext from v4i8
but this isn't:
v4i64 load, zext from v4i8
Whereas v4i64 is (arguably) legal, even without AVX2.
Note that the same thing was done a while ago for truncstores (r46140),
but I assume no one needed it yet for extloads, so here we go.
Calls to getLoadExtAction were changed to add the value type, found
manually in the surrounding code.
Calls to setLoadExtAction were mechanically changed, by wrapping the
call in a loop, to match previous behavior. The loop iterates over
the MVT subrange corresponding to the memory type (FP vectors, etc...).
I also pulled neighboring setTruncStoreActions into some of the loops;
those shouldn't make a difference, as the additional types are illegal.
(e.g., i128->i1 truncstores on PPC.)
No functional change intended.
Differential Revision: http://reviews.llvm.org/D6532
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@225421 91177308-0d34-0410-b5e6-96231b3b80d8
Now that we have MVT::FIRST_VALUETYPE (r225362), we can provide a method
checking that the MVT is valid, that is, it's in
[FIRST_VALUETYPE, LAST_VALUETYPE[.
This commit also uses it in a few asserts, that would previously accept
invalid MVTs, such as the default constructed -1. In that case,
the code following those asserts would do an out-of-bounds array access.
Using MVT::isValid, those assertions fail as expected when passed
invalid MVTs.
It feels clunky to have such a validity checking function, but it's
at least better than the alternative of broken manual checks.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@225411 91177308-0d34-0410-b5e6-96231b3b80d8
This is affecting the behavior of some ObjC++ / AArch64 test cases on Darwin.
Reverting to get the bots green while I track down the source of the changed
behavior.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@225311 91177308-0d34-0410-b5e6-96231b3b80d8
The existing code provided for specifying a global loop alignment preference.
However, the preferred loop alignment might depend on the loop itself. For
recent POWER cores, loops between 5 and 8 instructions should have 32-byte
alignment (while the others are better with 16-byte alignment) so that the
entire loop will fit in one i-cache line.
To support this, getPrefLoopAlignment has been made virtual, and can be
provided with an optional MachineLoop* so the target can inspect the loop
before answering the query. The default behavior, as before, is to return the
value set with setPrefLoopAlignment. MachineBlockPlacement now queries the
target for each loop instead of only once per function. There should be no
functional change for other targets.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@225117 91177308-0d34-0410-b5e6-96231b3b80d8
If the control flow is modelling an if-statement where the only instruction in
the 'then' basic block (excluding the terminator) is a call to cttz/ctlz,
CodeGenPrepare can try to speculate the cttz/ctlz call and simplify the control
flow graph.
Example:
\code
entry:
%cmp = icmp eq i64 %val, 0
br i1 %cmp, label %end.bb, label %then.bb
then.bb:
%c = tail call i64 @llvm.cttz.i64(i64 %val, i1 true)
br label %end.bb
end.bb:
%cond = phi i64 [ %c, %then.bb ], [ 64, %entry]
\code
In this example, basic block %then.bb is taken if value %val is not zero.
Also, the phi node in %end.bb would propagate the size-of in bits of %val
only if %val is equal to zero.
With this patch, CodeGenPrepare will try to hoist the call to cttz from %then.bb
into basic block %entry only if cttz is cheap to speculate for the target.
Added two new hooks in TargetLowering.h to let targets customize the behavior
(i.e. decide whether it is cheap or not to speculate calls to cttz/ctlz). The
two new methods are 'isCheapToSpeculateCtlz' and 'isCheapToSpeculateCttz'.
By default, both methods return 'false'.
On X86, method 'isCheapToSpeculateCtlz' returns true only if the target has
LZCNT. Method 'isCheapToSpeculateCttz' only returns true if the target has BMI.
Differential Revision: http://reviews.llvm.org/D6728
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@224899 91177308-0d34-0410-b5e6-96231b3b80d8
of the abi we should be using. For targets that don't use the
option there's no change, otherwise this allows external users
to set the ABI via string and avoid some of the -backend-option
pain in clang.
Use this option to move the ABI for the ARM port from the
Subtarget to the TargetMachine and update the testcases
accordingly since it's no longer valid to set via -mattr.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@224492 91177308-0d34-0410-b5e6-96231b3b80d8
This handles the case of a BUILD_VECTOR being constructed out of elements extracted from a vector twice the size of the result vector. Previously this was always scalarized. Now, we try to construct a shuffle node that feeds on extract_subvectors.
This fixes PR15872 and provides a partial fix for PR21711.
Differential Revision: http://reviews.llvm.org/D6678
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@224429 91177308-0d34-0410-b5e6-96231b3b80d8
The type promotion helper does not support vector type, so when make
such it does not kick in in such cases.
Original commit message:
[CodeGenPrepare] Move sign/zero extensions near loads using type promotion.
This patch extends the optimization in CodeGenPrepare that moves a sign/zero
extension near a load when the target can combine them. The optimization may
promote any operations between the extension and the load to make that possible.
Although this optimization may be beneficial for all targets, in particular
AArch64, this is enabled for X86 only as I have not benchmarked it for other
targets yet.
** Context **
Most targets feature extended loads, i.e., loads that perform a zero or sign
extension for free. In that context it is interesting to expose such pattern in
CodeGenPrepare so that the instruction selection pass can form such loads.
Sometimes, this pattern is blocked because of instructions between the load and
the extension. When those instructions are promotable to the extended type, we
can expose this pattern.
** Motivating Example **
Let us consider an example:
define void @foo(i8* %addr1, i32* %addr2, i8 %a, i32 %b) {
%ld = load i8* %addr1
%zextld = zext i8 %ld to i32
%ld2 = load i32* %addr2
%add = add nsw i32 %ld2, %zextld
%sextadd = sext i32 %add to i64
%zexta = zext i8 %a to i32
%addza = add nsw i32 %zexta, %zextld
%sextaddza = sext i32 %addza to i64
%addb = add nsw i32 %b, %zextld
%sextaddb = sext i32 %addb to i64
call void @dummy(i64 %sextadd, i64 %sextaddza, i64 %sextaddb)
ret void
}
As it is, this IR generates the following assembly on x86_64:
[...]
movzbl (%rdi), %eax # zero-extended load
movl (%rsi), %es # plain load
addl %eax, %esi # 32-bit add
movslq %esi, %rdi # sign extend the result of add
movzbl %dl, %edx # zero extend the first argument
addl %eax, %edx # 32-bit add
movslq %edx, %rsi # sign extend the result of add
addl %eax, %ecx # 32-bit add
movslq %ecx, %rdx # sign extend the result of add
[...]
The throughput of this sequence is 7.45 cycles on Ivy Bridge according to IACA.
Now, by promoting the additions to form more extended loads we would generate:
[...]
movzbl (%rdi), %eax # zero-extended load
movslq (%rsi), %rdi # sign-extended load
addq %rax, %rdi # 64-bit add
movzbl %dl, %esi # zero extend the first argument
addq %rax, %rsi # 64-bit add
movslq %ecx, %rdx # sign extend the second argument
addq %rax, %rdx # 64-bit add
[...]
The throughput of this sequence is 6.15 cycles on Ivy Bridge according to IACA.
This kind of sequences happen a lot on code using 32-bit indexes on 64-bit
architectures.
Note: The throughput numbers are similar on Sandy Bridge and Haswell.
** Proposed Solution **
To avoid the penalty of all these sign/zero extensions, we merge them in the
loads at the beginning of the chain of computation by promoting all the chain of
computation on the extended type. The promotion is done if and only if we do not
introduce new extensions, i.e., if we do not degrade the code quality.
To achieve this, we extend the existing “move ext to load” optimization with the
promotion mechanism introduced to match larger patterns for addressing mode
(r200947).
The idea of this extension is to perform the following transformation:
ext(promotableInst1(...(promotableInstN(load))))
=>
promotedInst1(...(promotedInstN(ext(load))))
The promotion mechanism in that optimization is enabled by a new TargetLowering
switch, which is off by default. In other words, by default, the optimization
performs the “move ext to load” optimization as it was before this patch.
** Performance **
Configuration: x86_64: Ivy Bridge fixed at 2900MHz running OS X 10.10.
Tested Optimization Levels: O3/Os
Tests: llvm-testsuite + externals.
Results:
- No regression beside noise.
- Improvements:
CINT2006/473.astar: ~2%
Benchmarks/PAQ8p: ~2%
Misc/perlin: ~3%
The results are consistent for both O3 and Os.
<rdar://problem/18310086>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@224402 91177308-0d34-0410-b5e6-96231b3b80d8
This patch extends the optimization in CodeGenPrepare that moves a sign/zero
extension near a load when the target can combine them. The optimization may
promote any operations between the extension and the load to make that possible.
Although this optimization may be beneficial for all targets, in particular
AArch64, this is enabled for X86 only as I have not benchmarked it for other
targets yet.
** Context **
Most targets feature extended loads, i.e., loads that perform a zero or sign
extension for free. In that context it is interesting to expose such pattern in
CodeGenPrepare so that the instruction selection pass can form such loads.
Sometimes, this pattern is blocked because of instructions between the load and
the extension. When those instructions are promotable to the extended type, we
can expose this pattern.
** Motivating Example **
Let us consider an example:
define void @foo(i8* %addr1, i32* %addr2, i8 %a, i32 %b) {
%ld = load i8* %addr1
%zextld = zext i8 %ld to i32
%ld2 = load i32* %addr2
%add = add nsw i32 %ld2, %zextld
%sextadd = sext i32 %add to i64
%zexta = zext i8 %a to i32
%addza = add nsw i32 %zexta, %zextld
%sextaddza = sext i32 %addza to i64
%addb = add nsw i32 %b, %zextld
%sextaddb = sext i32 %addb to i64
call void @dummy(i64 %sextadd, i64 %sextaddza, i64 %sextaddb)
ret void
}
As it is, this IR generates the following assembly on x86_64:
[...]
movzbl (%rdi), %eax # zero-extended load
movl (%rsi), %es # plain load
addl %eax, %esi # 32-bit add
movslq %esi, %rdi # sign extend the result of add
movzbl %dl, %edx # zero extend the first argument
addl %eax, %edx # 32-bit add
movslq %edx, %rsi # sign extend the result of add
addl %eax, %ecx # 32-bit add
movslq %ecx, %rdx # sign extend the result of add
[...]
The throughput of this sequence is 7.45 cycles on Ivy Bridge according to IACA.
Now, by promoting the additions to form more extended loads we would generate:
[...]
movzbl (%rdi), %eax # zero-extended load
movslq (%rsi), %rdi # sign-extended load
addq %rax, %rdi # 64-bit add
movzbl %dl, %esi # zero extend the first argument
addq %rax, %rsi # 64-bit add
movslq %ecx, %rdx # sign extend the second argument
addq %rax, %rdx # 64-bit add
[...]
The throughput of this sequence is 6.15 cycles on Ivy Bridge according to IACA.
This kind of sequences happen a lot on code using 32-bit indexes on 64-bit
architectures.
Note: The throughput numbers are similar on Sandy Bridge and Haswell.
** Proposed Solution **
To avoid the penalty of all these sign/zero extensions, we merge them in the
loads at the beginning of the chain of computation by promoting all the chain of
computation on the extended type. The promotion is done if and only if we do not
introduce new extensions, i.e., if we do not degrade the code quality.
To achieve this, we extend the existing “move ext to load” optimization with the
promotion mechanism introduced to match larger patterns for addressing mode
(r200947).
The idea of this extension is to perform the following transformation:
ext(promotableInst1(...(promotableInstN(load))))
=>
promotedInst1(...(promotedInstN(ext(load))))
The promotion mechanism in that optimization is enabled by a new TargetLowering
switch, which is off by default. In other words, by default, the optimization
performs the “move ext to load” optimization as it was before this patch.
** Performance **
Configuration: x86_64: Ivy Bridge fixed at 2900MHz running OS X 10.10.
Tested Optimization Levels: O3/Os
Tests: llvm-testsuite + externals.
Results:
- No regression beside noise.
- Improvements:
CINT2006/473.astar: ~2%
Benchmarks/PAQ8p: ~2%
Misc/perlin: ~3%
The results are consistent for both O3 and Os.
<rdar://problem/18310086>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@224351 91177308-0d34-0410-b5e6-96231b3b80d8
Add an option to disable optimization to shrink truncated larger type
loads to smaller type loads. On SI this prevents using scalar load
instructions in some cases, since there are no scalar extloads.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@224084 91177308-0d34-0410-b5e6-96231b3b80d8
This works like the composeSubRegisterIndices() function but transforms
a subregister lane mask instead of a subregister index.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223874 91177308-0d34-0410-b5e6-96231b3b80d8
Let tablegen compute the combination of subregister lanemasks for all
subregisters in a register/register class. This is preparation for further
work subregister allocation
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223873 91177308-0d34-0410-b5e6-96231b3b80d8
I'm recommiting the codegen part of the patch.
The vectorizer part will be send to review again.
Masked Vector Load and Store Intrinsics.
Introduced new target-independent intrinsics in order to support masked vector loads and stores. The loop vectorizer optimizes loops containing conditional memory accesses by generating these intrinsics for existing targets AVX2 and AVX-512. The vectorizer asks the target about availability of masked vector loads and stores.
Added SDNodes for masked operations and lowering patterns for X86 code generator.
Examples:
<16 x i32> @llvm.masked.load.v16i32(i8* %addr, <16 x i32> %passthru, i32 4 /* align */, <16 x i1> %mask)
declare void @llvm.masked.store.v8f64(i8* %addr, <8 x double> %value, i32 4, <8 x i1> %mask)
Scalarizer for other targets (not AVX2/AVX-512) will be done in a separate patch.
http://reviews.llvm.org/D6191
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223348 91177308-0d34-0410-b5e6-96231b3b80d8
This is the second patch in a small series. This patch contains the MachineInstruction and x86-64 backend pieces required to lower Statepoints. It does not include the code to actually generate the STATEPOINT machine instruction and as a result, the entire patch is currently dead code. I will be submitting the SelectionDAG parts within the next 24-48 hours. Since those pieces are by far the most complicated, I wanted to minimize the size of that patch. That patch will include the tests which exercise the functionality in this patch. The entire series can be seen as one combined whole in http://reviews.llvm.org/D5683.
The STATEPOINT psuedo node is generated after all gc values are explicitly spilled to stack slots. The purpose of this node is to wrap an actual call instruction while recording the spill locations of the meta arguments used for garbage collection and other purposes. The STATEPOINT is modeled as modifing all of those locations to prevent backend optimizations from forwarding the value from before the STATEPOINT to after the STATEPOINT. (Doing so would break relocation semantics for collectors which wish to relocate roots.)
The implementation of STATEPOINT is closely modeled on PATCHPOINT. Eventually, much of the code in this patch will be removed. The long term plan is to merge the functionality provided by statepoints and patchpoints. Merging their implementations in the backend is likely to be a good starting point.
Reviewed by: atrick, ributzka
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223085 91177308-0d34-0410-b5e6-96231b3b80d8
This reverts commit r222632 (and follow-up r222636), which caused a host
of LNT failures on an internal bot. I'll respond to the commit on the
list with a reproduction of one of the failures.
Conflicts:
lib/Target/X86/X86TargetTransformInfo.cpp
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@222936 91177308-0d34-0410-b5e6-96231b3b80d8
