Summary:
Currently clang emits these instructions via inline (volatile) asm in
the CUDA headers. Switching to intrinsics will let the optimizer reason
across calls to these intrinsics.
Reviewers: tra
Subscribers: llvm-commits, jholewinski
Differential Revision: http://reviews.llvm.org/D21160
llvm-svn: 272298
As suggested by clang-tidy's performance-unnecessary-copy-initialization.
This can easily hit lifetime issues, so I audited every change and ran the
tests under asan, which came back clean.
llvm-svn: 272126
NVVMIntrRange adds !range metadata to calls of NVVM intrinsics
that return values within known limited range.
This allows LLVM to generate optimal code for indexing arrays
based on tid/ctaid which is a frequently used pattern in CUDA code.
Differential Revision: http://reviews.llvm.org/D20644
llvm-svn: 270872
Having an enum member named Default is quite confusing: Is it distinct
from the others?
This patch removes that member and instead uses Optional<Reloc> in
places where we have a user input that still hasn't been maped to the
default value, which is now clear has no be one of the remaining 3
options.
llvm-svn: 269988
- Where we were returning a node before, call ReplaceNode instead.
- Where we would return null to fall back to another selector, rename
the method to try* and return a bool for success.
Part of llvm.org/pr26808.
llvm-svn: 269483
Many files include Passes.h but only a fraction needs to know about the
TargetPassConfig class. Move it into an own header. Also rename
Passes.cpp to TargetPassConfig.cpp while we are at it.
llvm-svn: 269011
This is a step towards removing the rampant undefined behaviour in
SelectionDAG, which is a part of llvm.org/PR26808.
We rename SelectionDAGISel::Select to SelectImpl and update targets to
match, and then change Select to return void and consolidate the
sketchy behaviour we're trying to get away from there.
Next, we'll update backends to implement `void Select(...)` instead of
SelectImpl and eventually drop the base Select implementation.
llvm-svn: 268693
Summary:
We don't have sign-/zero-extending ldg/ldu instructions defined,
so we need to emulate them with explicit CVTs. We were originally
handling the i8 case, but not any other cases.
Fixes PR26185
Reviewers: jingyue, jlebar
Subscribers: jholewinski
Differential Revision: http://reviews.llvm.org/D19615
llvm-svn: 268272
Summary:
Currently the NVVMReflect pass is run at the beginning of our backend
passes. But really, it should be run as early as possible, as it's
simply resolving an "if" statement in code. So copy it into
TargetMachine::addEarlyAsPossiblePasses.
We still run it at the beginning of the backend passes, since it's
needed for correctness when lowering to nvptx.
(Specifically, NVVMReflect changes each call to the __nvvm_reflect
function or llvm.nvvm.reflect intrinsic into an integer constant, based
on the pass's configuration. Clearly we miss many optimization
opportunities if we perform this transformation at the beginning of
codegen.)
Reviewers: rnk
Subscribers: tra, llvm-commits, jholewinski
Differential Revision: http://reviews.llvm.org/D18616
llvm-svn: 267765
NVPTXLowerKernelArgs is required for correctness, so it should not be guarded
by CodeGenOpt::None.
NVPTXPeephole is optimization only, so it should be skipped when
CodeGenOpt::None.
llvm-svn: 267619
Removed some unused headers, replaced some headers with forward class declarations.
Found using simple scripts like this one:
clear && ack --cpp -l '#include "llvm/ADT/IndexedMap.h"' | xargs grep -L 'IndexedMap[<]' | xargs grep -n --color=auto 'IndexedMap'
Patch by Eugene Kosov <claprix@yandex.ru>
Differential Revision: http://reviews.llvm.org/D19219
From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 266595
Summary:
Calls on NVPTX are unusually expensive (for one thing, lots of state
needs to be saved to memory, which is slow), so make the inlininer much
more aggressive.
Reviewers: chandlerc
Subscribers: jholewinski, llvm-commits, tra
Differential Revision: http://reviews.llvm.org/D18561
llvm-svn: 266406
Add a common parent class for ConstantArray, ConstantVector, and
ConstantStruct called ConstantAggregate. These are the aggregate
subclasses of Constant that take operands.
This is mainly a cleanup, adding common `isa` target and removing
duplicated code. However, it also simplifies caching which constants
point transitively at `GlobalValue` (a possible future direction).
llvm-svn: 265466
Summary:
Previously, we were running afoul of the assertion
EVT(CLI.Ins[i].VT) == InVals[i].getValueType() && "LowerCall emitted a value with the wrong type!"
in SelectionDAGBuilder.cpp when running the NVPTX/i8-param.ll test.
This is because our backend (for some reason) treats small return values
as i32, but it wasn't ever truncating the i32 back down to the expected
width in the DAG.
Unclear to me whether this fixes any actual bugs -- in this test, at
least, the generated code is unchanged.
Reviewers: jingyue
Subscribers: llvm-commits, tra, jholewinski
Differential Revision: http://reviews.llvm.org/D17872
llvm-svn: 265091
Summary:
Previously the NVVMReflect pass would read its configuration from
command-line flags or a static configuration given to the pass at
instantiation time.
This doesn't quite work for clang's use-case. It needs to pass a value
for __CUDA_FTZ down on a per-module basis. We use a module flag for
this, so the NVVMReflect pass needs to be updated to read said flag.
Reviewers: tra, rnk
Subscribers: cfe-commits, jholewinski
Differential Revision: http://reviews.llvm.org/D18672
llvm-svn: 265090
Summary:
Tablegen tries to infer this from the selection DAG patterns defined for
the instructions, but it can't always.
An instructive example is CLZr64. CLZr32 is correctly inferred to have
no side-effects, but the selection DAG pattern for CLZr64 is slightly
more complicated, and in particular the ctlz DAG node is not at the root
of the pattern. Thus tablegen can't infer that CLZr64 has no
side-effects.
Reviewers: jholewinski
Subscribers: jholewinski, tra, llvm-commits
Differential Revision: http://reviews.llvm.org/D17472
llvm-svn: 265089
This will become necessary in a subsequent change to make this method
merge adjacent stack adjustments, i.e. it might erase the previous
and/or next instruction.
It also greatly simplifies the calls to this function from Prolog-
EpilogInserter. Previously, that had a bunch of logic to resume iteration
after the call; now it just continues with the returned iterator.
Note that this changes the behaviour of PEI a little. Previously,
it attempted to re-visit the new instruction created by
eliminateCallFramePseudoInstr(). That code was added in r36625,
but I can't see any reason for it: the new instructions will obviously
not be pseudo instructions, they will not have FrameIndex operands,
and we have already accounted for the stack adjustment.
Differential Revision: http://reviews.llvm.org/D18627
llvm-svn: 265036
Summary:
Currently it's a module pass. Make it a function pass so that we can
move it to PassManagerBuilder's EP_EarlyAsPossible extension point,
which only accepts function passes.
Reviewers: rnk
Subscribers: tra, llvm-commits, jholewinski
Differential Revision: http://reviews.llvm.org/D18615
llvm-svn: 264919
MachineFunctionProperties represents a set of properties that a MachineFunction
can have at particular points in time. Existing examples of this idea are
MachineRegisterInfo::isSSA() and MachineRegisterInfo::tracksLiveness() which
will eventually be switched to use this mechanism.
This change introduces the AllVRegsAllocated property; i.e. the property that
all virtual registers have been allocated and there are no VReg operands
left.
With this mechanism, passes can declare that they require a particular property
to be set, or that they set or clear properties by implementing e.g.
MachineFunctionPass::getRequiredProperties(). The MachineFunctionPass base class
verifies that the requirements are met, and handles the setting and clearing
based on the delcarations. Passes can also directly query and update the current
properties of the MF if they want to have conditional behavior.
This change annotates the target-independent post-regalloc passes; future
changes will also annotate target-specific ones.
Reviewers: qcolombet, hfinkel
Differential Revision: http://reviews.llvm.org/D18421
llvm-svn: 264593
This reserves an MDKind for !llvm.loop, which allows callers to avoid a
string-based lookup. I'm not sure why it was missing.
There should be no functionality change here, just a small compile-time
speedup.
llvm-svn: 264371
Summary:
The old address space inference pass (NVPTXFavorNonGenericAddrSpaces) is unable
to convert the address space of a pointer induction variable. This patch adds a
new pass called NVPTXInferAddressSpaces that overcomes that limitation using a
fixed-point data-flow analysis (see the file header comments for details).
The new pass is experimental and not enabled by default. Users can turn
it on by setting the -nvptx-use-infer-addrspace flag of llc.
Reviewers: jholewinski, tra, jlebar
Subscribers: jholewinski, llvm-commits
Differential Revision: http://reviews.llvm.org/D17965
llvm-svn: 263916
tests to run GVN in both modes.
This is mostly the boring refactoring just like SROA and other complex
transformation passes. There is some trickiness in that GVN's
ValueNumber class requires hand holding to get to compile cleanly. I'm
open to suggestions about a better pattern there, but I tried several
before settling on this. I was trying to balance my desire to sink as
much implementation detail into the source file as possible without
introducing overly many layers of abstraction.
Much like with SROA, the design of this system is made somewhat more
cumbersome by the need to support both pass managers without duplicating
the significant state and logic of the pass. The same compromise is
struck here.
I've also left a FIXME in a doxygen comment as the GVN pass seems to
have pretty woeful documentation within it. I'd like to submit this with
the FIXME and let those more deeply familiar backfill the information
here now that we have a nice place in an interface to put that kind of
documentaiton.
Differential Revision: http://reviews.llvm.org/D18019
llvm-svn: 263208
Summary:
Tablegen was unable to determine that param loads/stores were actually
reading or writing from memory. I think this isn't a problem in
practice for param stores, because those occur in a block right before
we make our call. But param loads don't have to at the very beginning
of a function, so should be annotated as mayLoad so we don't incorrectly
optimize them.
Reviewers: jholewinski
Subscribers: jholewinski, llvm-commits
Differential Revision: http://reviews.llvm.org/D17471
llvm-svn: 262381
Summary: Looks like this was caused by a typo.
Reviewers: jholewinski
Subscribers: jholewinski, llvm-commits, tra
Differential Revision: http://reviews.llvm.org/D17357
llvm-svn: 262380
Summary:
Calls sometimes need to be convergent. This is already handled at the
LLVM IR level, but it also needs to be handled at the MI level.
Ideally we'd propagate convergence from instructions, down through the
selection DAG, and into MIs. But this is Hard, and would affect
optimizations in the SDNs -- right now only SDNs with two operands have
any flags at all.
Instead, here's a much simpler hack: Add new opcodes for NVPTX for
convergent calls, and generate these when lowering convergent LLVM
calls.
Reviewers: jholewinski
Subscribers: jholewinski, chandlerc, joker.eph, jhen, tra, llvm-commits
Differential Revision: http://reviews.llvm.org/D17423
llvm-svn: 262373
Summary:
Also simplify some of the embedded C++ logic.
No functional changes.
Reviewers: jholewinski
Subscribers: llvm-commits, tra, jholewinski
Differential Revision: http://reviews.llvm.org/D17354
llvm-svn: 262371
Change TargetInstrInfo API to take `MachineInstr&` instead of
`MachineInstr*` in the functions related to predicated instructions
(I'll try to come back later and get some of the rest). All of these
functions require non-null parameters already, so references are more
clear. As a bonus, this happens to factor away a host of implicit
iterator => pointer conversions.
No functionality change intended.
llvm-svn: 261605
Summary:
Previously the machine instructions for bar.sync &co. were not marked as
convergent. This resulted in some MI passes (such as TailDuplication,
fixed in an upcoming patch) doing unsafe things to these instructions.
Reviewers: jingyue
Subscribers: llvm-commits, tra, jholewinski, hfinkel
Differential Revision: http://reviews.llvm.org/D17318
llvm-svn: 261115
Summary:
Otherwise we'll try to do unsafe optimizations on these MIs, such as
sinking loads below calls.
(I suspect that this is not the only bug in the NVPTX instruction
tablegen files; I need to comb through them.)
Reviewers: jholewinski, tra
Subscribers: jingyue, jhen, llvm-commits
Differential Revision: http://reviews.llvm.org/D17315
llvm-svn: 261113
Summary:
This patch is provided in preparation for removing autoconf on 1/26. The proposal to remove autoconf on 1/26 was discussed on the llvm-dev thread here: http://lists.llvm.org/pipermail/llvm-dev/2016-January/093875.html
"I felt a great disturbance in the [build system], as if millions of [makefiles] suddenly cried out in terror and were suddenly silenced. I fear something [amazing] has happened."
- Obi Wan Kenobi
Reviewers: chandlerc, grosbach, bob.wilson, tstellarAMD, echristo, whitequark
Subscribers: chfast, simoncook, emaste, jholewinski, tberghammer, jfb, danalbert, srhines, arsenm, dschuff, jyknight, dsanders, joker.eph, llvm-commits
Differential Revision: http://reviews.llvm.org/D16471
llvm-svn: 258861
Summary:
Previously, we would just output "foo = bar" in the assembly, and then
ptxas would choke. Now we die before emitting any invalid code.
Reviewers: echristo
Subscribers: jholewinski, llvm-commits, jhen, tra
Differential Revision: http://reviews.llvm.org/D16490
llvm-svn: 258638
These are redundant pairs of nodes defined for
INSERT_VECTOR_ELEMENT/EXTRACT_VECTOR_ELEMENT.
insertelement/extractelement are slightly closer to the corresponding
C++ node name, and has stricter type checking so prefer it.
Update targets to only use these nodes where it is trivial to do so.
AArch64, ARM, and Mips all have various type errors on simple replacement,
so they will need work to fix.
Example from AArch64:
def : Pat<(sext_inreg (vector_extract (v16i8 V128:$Rn), VectorIndexB:$idx), i8),
(i32 (SMOVvi8to32 V128:$Rn, VectorIndexB:$idx))>;
Which is trying to do sext_inreg i8, i8.
llvm-svn: 255359
If a section is rw, it is irrelevant if the dynamic linker will write to
it or not.
It looks like llvm implemented this because gcc was doing it. It looks
like gcc implemented this in the hope that it would put all the
relocated items close together and speed up the dynamic linker.
There are two problem with this:
* It doesn't work. Both bfd and gold will map .data.rel to .data and
concatenate the input sections in the order they are seen.
* If we want a feature like that, it can be implemented directly in the
linker since it knowns where the dynamic relocations are.
llvm-svn: 253436
This extends the work done in r233995 so that now getFragment (in addition to
getSection) also works for variable symbols.
With that the existing logic to decide if a-b can be computed works even if
a or b are variables. Given that, the expression evaluation can avoid expanding
variables as aggressively and that in turn lets the relocation code see the
original variable.
In order for this to work with the asm streamer, there is now a dummy fragment
per section. It is used to assign a section to a symbol when no other fragment
exists.
This patch is a joint work by Maxim Ostapenko andy myself.
llvm-svn: 249303
Summary:
This is the first patch in the series to migrate Triple's (which are ambiguous)
to TargetTuple's (which aren't).
For the moment, TargetTuple simply passes all requests to the Triple object it
holds. Once it has replaced Triple, it will start to implement the interface in
a more suitable way.
This change makes some changes to the public C++ API. In particular,
InitMCSubtargetInfo(), createMCRelocationInfo(), and createMCSymbolizer()
now take TargetTuples instead of Triples. The other public C++ API's have
been left as-is for the moment to reduce patch size.
This commit also contains a trivial patch to clang to account for the C++ API
change. Thanks go to Pavel Labath for fixing LLDB for me.
Reviewers: rengolin
Subscribers: jyknight, dschuff, arsenm, rampitec, danalbert, srhines, javed.absar, dsanders, echristo, emaste, jholewinski, tberghammer, ted, jfb, llvm-commits, rengolin
Differential Revision: http://reviews.llvm.org/D10969
llvm-svn: 247692
Summary:
This is the first patch in the series to migrate Triple's (which are ambiguous)
to TargetTuple's (which aren't).
For the moment, TargetTuple simply passes all requests to the Triple object it
holds. Once it has replaced Triple, it will start to implement the interface in
a more suitable way.
This change makes some changes to the public C++ API. In particular,
InitMCSubtargetInfo(), createMCRelocationInfo(), and createMCSymbolizer()
now take TargetTuples instead of Triples. The other public C++ API's have
been left as-is for the moment to reduce patch size.
This commit also contains a trivial patch to clang to account for the C++ API
change.
Reviewers: rengolin
Subscribers: jyknight, dschuff, arsenm, rampitec, danalbert, srhines, javed.absar, dsanders, echristo, emaste, jholewinski, tberghammer, ted, jfb, llvm-commits, rengolin
Differential Revision: http://reviews.llvm.org/D10969
llvm-svn: 247683
Summary: This fixes a variety of typos in docs, code and headers.
Subscribers: jholewinski, sanjoy, arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D12626
llvm-svn: 247495
The pass is needed to remove __nvvm_reflect calls when we link in
libdevice bitcode that comes with CUDA.
Differential Revision: http://reviews.llvm.org/D11663
llvm-svn: 247072
Summary:
Let NVPTX backend detect integer min and max patterns during isel and emit intrinsics that enable hardware support.
Reviewers: jholewinski, meheff, jingyue
Subscribers: arsenm, llvm-commits, meheff, jingyue, eliben, jholewinski
Differential Revision: http://reviews.llvm.org/D12377
llvm-svn: 246107
Summary:
__shared__ variable may now emit undef value as initializer, do not
throw error on that.
Test Plan: test/CodeGen/NVPTX/global-addrspace.ll
Patch by Xuetian Weng
Reviewers: jholewinski, tra, jingyue
Subscribers: llvm-commits, jholewinski
Differential Revision: http://reviews.llvm.org/D12242
llvm-svn: 245785
Summary:
Add an LSR test that exercises isTruncateFree. Without this change, LSR creates
another indvar representing the truncated value.
Reviewers: jholewinski, eliben
Subscribers: jholewinski, llvm-commits
Differential Revision: http://reviews.llvm.org/D12058
llvm-svn: 245611
For NVPTX, try to use 32-bit division instead of 64-bit division when the dividend and divisor
fit in 32 bits. This speeds up some internal benchmarks significantly. The underlying reason
is that many index computations are carried out in 64-bits but never actually exceed the
capacity of a 32-bit word.
llvm-svn: 244684
More specifically, make NVPTXISelDAGToDAG able to emit cached loads (LDG) for pointer induction variables.
Also fix latent bug where LDG was not restricted to kernel functions. I believe that this could not be triggered so far since we do not currently infer that a pointer is global outside a kernel function, and only loads of global pointers are considered for cached loads.
llvm-svn: 244166
rather than 'unsigned' for their costs.
For something like costs in particular there is a natural "negative"
value, that of savings or saved cost. As a consequence, there is a lot
of code that subtracts or creates negative values based on cost, all of
which is prone to awkwardness or bugs when dealing with an unsigned
type. Similarly, we *never* want these values to wrap, as that would
cause Very Bad code generation (likely percieved as an infinite loop as
we try to emit over 2^32 instructions or some such insanity).
All around 'int' seems a much better fit for these basic metrics. I've
added asserts to ensure that at least the TTI interface never returns
negative numbers here. If we ever have a use case for negative numbers,
we can remove this, but this way a bug where someone used '-1' to
produce a 'very large' cost will be caught by the assert.
This passes all tests, and is also UBSan clean.
No functional change intended.
Differential Revision: http://reviews.llvm.org/D11741
llvm-svn: 244080
Summary:
For example, in
struct S {
int *x;
int *y;
};
__global__ void foo(S s) {
int *b = s.y;
// use b
}
"b" is guaranteed to point to global. NVPTX should emit ld.global/st.global for
accessing "b".
Reviewers: jholewinski
Subscribers: llvm-commits, jholewinski
Differential Revision: http://reviews.llvm.org/D11505
llvm-svn: 243790
Summary:
MCRegAliasIterator only works for physical registers. So, do not run it
on virtual registers.
With this issue fixed, we can resurrect the BranchFolding pass in NVPTX
backend.
Reviewers: jholewinski, bkramer
Subscribers: henryhu, meheff, llvm-commits, jholewinski
Differential Revision: http://reviews.llvm.org/D11174
llvm-svn: 242871
Summary:
[NVPTX] make load on global readonly memory to use ldg
Summary:
As describe in [1], ld.global.nc may be used to load memory by nvcc when
__restrict__ is used and compiler can detect whether read-only data cache
is safe to use.
This patch will try to check whether ldg is safe to use and use them to
replace ld.global when possible. This change can improve the performance
by 18~29% on affected kernels (ratt*_kernel and rwdot*_kernel) in
S3D benchmark of shoc [2].
Patched by Xuetian Weng.
[1] http://docs.nvidia.com/cuda/kepler-tuning-guide/#read-only-data-cache
[2] https://github.com/vetter/shoc
Test Plan: test/CodeGen/NVPTX/load-with-non-coherent-cache.ll
Reviewers: jholewinski, jingyue
Subscribers: jholewinski, llvm-commits
Differential Revision: http://reviews.llvm.org/D11314
llvm-svn: 242713
Justin Lebar