Martin Storsjö [Fri, 14 Apr 2023 20:58:34 +0000 (20:58 +0000)]
[compiler-rt] [test] Don't override config.unsupported in TestCases/Windows
This is a leftover from when these tests were added in
2bdca2006941a3a7de8e16ea9af9a758b367ee32 in 2014 (when the toplevel
asan/lit.cfg set "config.unsupported = True" on Windows). When the
common tests were included on Windows in
89d994367a597134fa7e1724314044e217718165 in 2015, that commit should
have removed the now unnecessary "config.unsupported = False" line here.
Differential Revision: https://reviews.llvm.org/D148442
Martin Storsjö [Wed, 5 Apr 2023 12:25:09 +0000 (15:25 +0300)]
[libcxx] [ci] Run the libcxxabi and libunwind tests in mingw configurations
The check-runtimes function runs check-cxx, check-cxxabi and check-unwind.
Differential Revision: https://reviews.llvm.org/D148267
Yonghong Song [Mon, 17 Apr 2023 18:11:59 +0000 (11:11 -0700)]
[TTI][BPF] Ensure ArgumentPromotion Not Exceeding Target MaxArgs
With LLVM patch https://reviews.llvm.org/D148269, we hit a linux kernel
bpf selftest compilation failure like below:
...
progs/test_xdp_noinline.c:739:8: error: too many args to t8: i64 = GlobalAddress<ptr @encap_v4> 0, progs/test_xdp_noinline.c:739:8
if (!encap_v4(xdp, cval, &pckt, dst, pkt_bytes))
^
...
progs/test_xdp_noinline.c:321:6: error: defined with too many args
bool encap_v4(struct xdp_md *xdp, struct ctl_value *cval,
^
...
Note that bpf selftests are compiled with -O2 which is
the recommended flag for bpf community.
The bpf backend calling convention is only allowing 5
parameters in registers and does not allow pass arguments
through stacks. In the above case, ArgumentPromotionPass
replaced parameter '&pckt' as two parameters, so the total
number of arguments after ArgumentPromotion pass becomes 6
and this caused later compilation failure during instruction
selection phase.
This patch added a TargetTransformInfo hook getMaxNumArgs()
which returns 5 for BPF and UINT_MAX for other targets.
Differential Revision: https://reviews.llvm.org/D148551
David Sherwood [Tue, 11 Apr 2023 12:10:33 +0000 (12:10 +0000)]
[NFC][AArch64] Add cost model tests for extending loads
Differential Revision: https://reviews.llvm.org/D148122
Mark de Wever [Tue, 18 Apr 2023 17:19:48 +0000 (19:19 +0200)]
[libc++] Adds missing includes.
This patch makes are code less dependant on transitive includes.
This was part of D145800. This patch will be abandoned, but these
changes are still useful. I manually verified declarations of the new
includes are used in these files.
Reviewed By: #libc, philnik
Differential Revision: https://reviews.llvm.org/D148645
Joe Nash [Mon, 17 Apr 2023 21:22:43 +0000 (17:22 -0400)]
[AMDGPU] NFC. Refactor GISel for cmp intrinsics
Combine the logic for fcmp and icmp intrinsics and use operand presence
instead.
Reviewed By: kosarev, foad
Differential Revision: https://reviews.llvm.org/D148716
Jay Foad [Wed, 19 Apr 2023 14:26:32 +0000 (15:26 +0100)]
[AMDGPU] Fix some check prefixes
Jay Foad [Wed, 19 Apr 2023 14:23:55 +0000 (15:23 +0100)]
[AMDGPU] Remove unused check lines from tests
Arash Taheri-Dezfouli [Wed, 19 Apr 2023 15:09:29 +0000 (15:09 +0000)]
[MLIR] Fix missing ::mlir namespace
Add a missing ::mlir namespace for MixedContainerType
that only triggers if the user is not using namespace mlir.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D147956
Scott Linder [Wed, 19 Apr 2023 15:02:08 +0000 (15:02 +0000)]
[dwarfdump][AMDGPU] Support EF_AMDGPU_MACH_NONE
A quirk of the AMDGPU backend is EF_AMDGPU_MACH_NONE, which is not
specific to the r600 or amdgcn architecture, but can be combined with
either.
AMDGPU ELF code objects with this mach value cannot be mapped back to a
Triple architecture, as it could be either r600 or amdgcn. For
llvm-dwarfdump this means the normal method of inspecting
ObjectFile::getArch to determine how to handle relocations is
insufficient.
This patch adds an extra check for ELF code objects which would
otherwise be categorized as UnknownArch, making it possible to use
llvm-dwarfdump with them. For completeness it also adds support for
known r600 machines.
This also adds specific tests for llvm-dwarfdump for amdgcn and r600,
both with known and unknown mach values.
Reviewed By: jhenderson
Differential Revision: https://reviews.llvm.org/D144301
Razvan Lupusoru [Tue, 18 Apr 2023 22:42:16 +0000 (15:42 -0700)]
[flang][openacc] Add OpenACC pointer interface to FIR pointer types
The OpenACC dialect data clauses require for variables appearing in data
operations to implement the `PointerLikeType` interface (similarly to
OpenMP dialect). Thus, this interface needs attached to FIR types.
Two of the FIR types, HeapType and LLVMPointerType, did not provide
getElementType - now they do by calling getEleTy. This cleans up the
interface attachment for both OpenACC and OpenMP dialects.
Reviewed By: clementval, jeanPerier
Differential Revision: https://reviews.llvm.org/D148560
Dinar Temirbulatov [Wed, 19 Apr 2023 14:17:32 +0000 (14:17 +0000)]
[AArch64][CodeGen] Allow vectors larger than hardware support to use SVE's load zero/sign-extend for fixed vectors.
Prefer to fold LOAD + SIGN/ZEROEXTEND to SVE load and sign extend instructions
for fixed-length-vectors even if a type is not representable on a hardware.
Differential Revision: https://reviews.llvm.org/D147533
OCHyams [Wed, 19 Apr 2023 14:14:18 +0000 (15:14 +0100)]
Remove unused ValueTracking.h include from DebugInfo.cpp
Buildbot: https://buildkite.com/llvm-project/upstream-bazel/builds/
59967#
01879985-8d44-4041-9cd0-
a1e41371208e
See https://reviews.llvm.org/D148536
Jay Foad [Wed, 19 Apr 2023 14:12:50 +0000 (15:12 +0100)]
[AMDGPU] Remove unused check lines from GlobalISel IR tests
Sam McCall [Wed, 19 Apr 2023 14:11:58 +0000 (16:11 +0200)]
[dataflow] add missing clangLex dep after
a443b3d18ef4d01e7
Benjamin Kramer [Wed, 19 Apr 2023 13:31:34 +0000 (15:31 +0200)]
[SCF] Clean up ForOpTensorCastFolder and harden it against nop tensor casts
The code was inserting a new cast, discarding it, then inserting it
again.
The self-cast issue is the root of #62135 because it would end up
dropping the loop and inserting an invalid cast to itself. As far as I
can tell tensor.cast with the same src and dst types is not invalid but
it can't really be tested in isolation as it's immediately folded.
Fixes #62135
Differential Revision: https://reviews.llvm.org/D148714
Jay Foad [Wed, 19 Apr 2023 13:54:50 +0000 (14:54 +0100)]
[AMDGPU] Remove unused check lines from GlobalISel MIR tests
Sam McCall [Wed, 8 Mar 2023 15:13:28 +0000 (16:13 +0100)]
[dataflow] add HTML logger: browse code/cfg/analysis timeline/state
With -dataflow-log=/dir we will write /dir/0.html etc for each
function analyzed.
These files show the function's code and CFG, and the path through
the CFG taken by the analysis. At each analysis point we can see the
lattice state.
Currently the lattice state dump is not terribly useful but we can
improve this: showing values associated with the current Expr,
simplifying flow condition, highlighting changes etc.
(Trying not to let this patch scope-creep too much, so I ripped out the
half-finished features)
Demo: https://htmlpreview.github.io/?https://gist.githubusercontent.com/sam-mccall/
1746985bf13406bd19181af281aea9ff/raw/
9718fdd48406dabccb3092acd983b4bd55da9dfa/analysis.html
Differential Revision: https://reviews.llvm.org/D146591
Andrew Gozillon [Wed, 19 Apr 2023 13:21:24 +0000 (08:21 -0500)]
[OpenMP][MLIR] Add lowering of omp::FlagsAttr to LLVM-IR
The omp::FlagsAttr contains OpenMP RTL flags
given by a user to the compiler and a frontend
(flang currently) then populates the omp::FlagsAttr
which will then lower these to LLVM globals
which are utilised by the OpenMP runtime.
Reviewers: jdoerfert
Differential Revision: https://reviews.llvm.org/D147344
David Green [Wed, 19 Apr 2023 13:26:27 +0000 (14:26 +0100)]
[AArch64] Remove dead tryMLAV64LaneV128 and tryMULLV64LaneV128 code.
As far as I can tell this code is never used, as the pattern recognised by
checkHighLaneIndex (an duplane with insert_subvec and extract_subvec) will not
be generated any more. There are no tests that change from removing it
(including the clang neon tests), and it didn't appear to come up in any
benchmarks I ran. There are already existing tablegen patterns for MLA with
index and s/umull with index.
Removing it also prevents it from causing problems for SVE, as in #62151.
Differential Revision: https://reviews.llvm.org/D148646
Dinar Temirbulatov [Wed, 19 Apr 2023 13:24:58 +0000 (13:24 +0000)]
[DAGCombine][AArch64] Allow transformable to legal vectors to be used for MULH lowering.
It looks like it is still profitable to accept a transformable to a legal vector
type, not just a legal vector, as long as vector elements are the same between
two of those types.
Differential Revision: https://reviews.llvm.org/D148229
OCHyams [Wed, 19 Apr 2023 12:59:08 +0000 (13:59 +0100)]
Reapply D146987 "[Assignment Tracking] Enable by default"
This reverts commit
7674ae7231c4e286b93eb399e475765630666f8b which
reverts D146987.
See https://reviews.llvm.org/D146987 for issues.
Joseph Huber [Mon, 17 Apr 2023 20:08:59 +0000 (15:08 -0500)]
[libc] Add the '--threads' and '--blocks' option to the GPU loaders
We will want to test the GPU `libc` with multiple threads in the future.
This patch adds the `--threads` and `--blocks` option to set the `x`
dimension of the kernel. Using CUDA terminology instead of OpenCL for
familiarity.
Depends on D148288 D148342
Reviewed By: jdoerfert, sivachandra, tra
Differential Revision: https://reviews.llvm.org/D148485
Joseph Huber [Tue, 18 Apr 2023 14:44:27 +0000 (09:44 -0500)]
[libc] Add a support library for GPU utilities
The GPU has many features that can only be accessed through builtin or
intrinsic functions. Furthermore, these functions are unique for each
GPU target. This patch outlines an interface to create a common `libc`
interface to access these. Currently I only implement a function for the
CUDA equivalent of `blockIdx.x`. More will be added in the future.
Reviewed By: sivachandra
Differential Revision: https://reviews.llvm.org/D148635
Alexey Bataev [Fri, 14 Apr 2023 16:35:03 +0000 (09:35 -0700)]
[SLP]Fix cost estimation for buildvectors with extracts and/or constants.
If the partial matching is found and some other scalars must be
inserted, need to account the cost of the extractelements, transformed
to shuffles, and/or reused entries and calculate the cost of inserting
constants properly into the non-poison vectors.
Also, fixed the cost calculation for final gather/buildvector sequence.
Differential Revision: https://reviews.llvm.org/D148362
Marco Elver [Wed, 19 Apr 2023 12:46:42 +0000 (14:46 +0200)]
[SanitizerBinaryMetadata] Respect no_sanitize("thread") function attribute
To avoid false positives, respect no_sanitize("thread") when atomics
metadata is enabled.
Reviewed By: dvyukov
Differential Revision: https://reviews.llvm.org/D148694
LLVM GN Syncbot [Wed, 19 Apr 2023 12:38:32 +0000 (12:38 +0000)]
[gn build] Port
bf47c1ed8556
OCHyams [Wed, 19 Apr 2023 12:11:22 +0000 (13:11 +0100)]
Reapply "[Assignment Tracking] Fix fragment error for some DSE-shortened stores"
This reverts commit
6db6ab4815a44bfcaabfcdd84a0ff458394f6f52 which reverts
D148536.
Build issues addressed in D148698.
Sam McCall [Mon, 17 Apr 2023 18:35:20 +0000 (20:35 +0200)]
[dataflow] Extract arena for Value/StorageLocation out of DataflowAnalysisContext
DataflowAnalysisContext has a few too many responsibilities, this narrows them.
It also allows the Arena to be shared with analysis steps, which need to create
Values, without exposing the whole DACtx API (flow conditions etc).
This means Environment no longer needs to proxy all these methods.
(For now it still does, because there are many callsites to update, and maybe
if we separate bool formulas from values we can avoid churning them twice)
In future, if we untangle the concepts of Values from boolean formulas/atoms,
Arena would also be responsible for creating formulas and managing atom IDs.
Differential Revision: https://reviews.llvm.org/D148554
Matt Arsenault [Thu, 13 Apr 2023 01:33:02 +0000 (21:33 -0400)]
ValueTracking: Handle shufflevector in computeKnownFPClass
Matt Arsenault [Tue, 18 Apr 2023 14:28:39 +0000 (10:28 -0400)]
ValueTracking: Add baseline tests for shufflevector computeKnownFPClass handling
Matt Arsenault [Wed, 12 Apr 2023 23:21:49 +0000 (19:21 -0400)]
ValueTracking: Handle insertelement in computeKnownFPClass
Matt Arsenault [Thu, 13 Apr 2023 13:48:08 +0000 (09:48 -0400)]
ValueTracking: sitofp cannot return -0
Matt Arsenault [Sat, 8 Apr 2023 23:29:53 +0000 (19:29 -0400)]
ValueTracking: Add baseline test for computeKnownFPClass for minnum/maxnum
Matt Arsenault [Mon, 17 Apr 2023 23:12:18 +0000 (19:12 -0400)]
ValueTracking: Add more fpext and fptrunc tests
David Spickett [Wed, 19 Apr 2023 11:42:14 +0000 (11:42 +0000)]
[lldb][RISCV] Add missing flags pointer to register infos
efd64c2f2506df3e2ed6d9da68e49f03fc31763c added these.
Enna1 [Wed, 19 Apr 2023 11:27:26 +0000 (19:27 +0800)]
[hwasan] provide a runtime flag for printing remaining threads in error report as an extra information
This patch adds a runtime flag `print_live_threads_info`, which defaults to true, controls whether or not prints remaining threads in error report as an extra information.
We(ByteDance) are in the process of enabling hwasan for our server-side applications on AArch64/Linux, these server-side applications have thousands of threads, so it is nice to have this option.
Reviewed By: fmayer
Differential Revision: https://reviews.llvm.org/D148513
OCHyams [Wed, 19 Apr 2023 10:59:37 +0000 (11:59 +0100)]
[NFC] Rename isPointerOffset to getPointerOffsetFrom and move to Value.h
Linking LLVMCore failed when building D148536 with shared libs enabled:
https://lab.llvm.org/buildbot/#/builders/121/builds/29766
Make isPointerOffset a Value method and rename it to getPointerOffsetFrom.
Reviewed By: jmorse
Differential Revision: https://reviews.llvm.org/D148698
Martin Storsjö [Thu, 30 Mar 2023 11:40:17 +0000 (14:40 +0300)]
[libcxxabi] [test] Fix the mingw test config
Don't link libc++abi separately in addition to the main -lc++; in
mingw build configs, libc++abi is always bundled into libc++
(via LIBCXX_ENABLE_STATIC_ABI_LIBRARY).
In the case of a shared linked libc++, linking a separate static
libc++abi leads to linker errors.
Differential Revision: https://reviews.llvm.org/D147638
Martin Storsjö [Sat, 1 Apr 2023 22:45:07 +0000 (01:45 +0300)]
[libcxxabi] [test] Avoid mingw warnings about missing a return statement
With current versions of mingw-w64 headers, code following
assert(false) isn't considered unreachable - thus add a dummy
"return nullptr;", to avoid warnings (treated as errors) for a
missing return statement.
The root cause does get fixed further upstream in mingw-w64 in
https://github.com/mingw-w64/mingw-w64/commit/
1690994f515910a31b9fb7c7bd3a52d4ba987abe
though.
Differential Revision: https://reviews.llvm.org/D147860
Martin Storsjö [Sat, 15 Apr 2023 21:58:51 +0000 (21:58 +0000)]
[libcxxabi] Omit dllimport in public headers in MinGW mode
This matches the corresponding change to libcxx headers in
dfa88927ae1411ccc3b248b7e624f2acf623d947.
This avoids needing to define _LIBCXXABI_DISABLE_VISIBILITY_ANNOTATIONS
when building libcxxabi tests, for two reasons:
- It defaults to normal linkage (as opposed to dllimport), fixing
linking of a static library version of libcxx/libcxxabi
- It avoids issues with using dllimport declarations on symbols when
including cxxabi implementations into the testcase itself
Differential Revision: https://reviews.llvm.org/D148441
Hassnaa Hamdi [Wed, 19 Apr 2023 09:23:13 +0000 (09:23 +0000)]
[AArch64][CostModel]: Add costs for zero/sign extend.
Add cost for extending to illegal scalable vector types.
Add testing file for the extend operations.
Reviewed By: sdesmalen
Differential Revision: https://reviews.llvm.org/D142456
Martin Liska [Wed, 19 Apr 2023 09:47:43 +0000 (11:47 +0200)]
Make use of '#pragma clang diagnostic' only with __clang__.
Otherwise one gets the following warning with GCC:
warning: ignoring ‘#pragma clang diagnostic’ [-Wunknown-pragmas]
Differential Revision: https://reviews.llvm.org/D148695
pvanhout [Tue, 18 Apr 2023 12:22:45 +0000 (14:22 +0200)]
[AMDGPU] Add more verbose logs to PromoteAlloca
More specifically make it more talkative when it's looking at the users of
an alloca to promote it to a vector.
A common failure point of the pass is unknown or weird users of the alloca.
While debugging issues related to this pass one of the first thing I usually
did was to add logs to see how the users were being handled.
Having such logs in directly seems to be a nice addition.
Reviewed By: arsenm, rampitec
Differential Revision: https://reviews.llvm.org/D148629
David Spickett [Mon, 17 Apr 2023 13:36:03 +0000 (13:36 +0000)]
[compiler-rt][ASAN] Replace find_executable with shtuil.which in script
distutils is deprecated and shutil.which is the suggested
replacement for this function.
https://peps.python.org/pep-0632/#migration-advice
https://docs.python.org/3/library/shutil.html#shutil.which
which was added in 3.3 (https://docs.python.org/3/library/shutil.html#shutil.which)
and LLVM requires at least 3.6 (https://llvm.org/docs/GettingStarted.html#software).
Reviewed By: delcypher, vitalybuka
Differential Revision: https://reviews.llvm.org/D148528
OCHyams [Wed, 19 Apr 2023 09:02:24 +0000 (10:02 +0100)]
Revert "[Assignment Tracking] Fix fragment error for some DSE-shortened stores"
This reverts commit
fca3e8e024f0015604d21e6f76f3e199345679c5.
Buildbot: https://lab.llvm.org/buildbot/#/builders/121/builds/29766
OCHyams [Wed, 19 Apr 2023 09:01:40 +0000 (10:01 +0100)]
Revert D146987 "[Assignment Tracking] Enable by default"
This reverts commit
107b307bd61e51aa66ea41f1a2c92c1459b39c7e due
to build errors in a parent commit
fca3e8e024f0015604d21e6f76f3e199345679c5
Jay Foad [Wed, 12 Apr 2023 13:22:40 +0000 (14:22 +0100)]
[AMDGPU] Don't transform illegal intrinsics to V_ILLEGAL
This reverts parts of D123693. The functionality of allowing unsupported
intrinsics to select has been superseded by D139000 "Remove function
with incompatible features".
Retain assembler/disassembler support for v_illegal on GFX10+ only,
where it is documented.
Differential Revision: https://reviews.llvm.org/D148127
terrydang [Wed, 19 Apr 2023 08:45:05 +0000 (09:45 +0100)]
[docs] Fix the CMAKE_BUILD_TYPE option in the cmake command in GettingStarted.rst
The cmake command contained a duplicate CMAKE_BUILD_TYPE option in the
section "Compiling the LLVM Suite Source Code".
Guillaume Chatelet [Tue, 18 Apr 2023 15:16:20 +0000 (15:16 +0000)]
[reland][libc] handle memset sequence as a separate struct
These sequence of calls don't really make sense for head_tail and loop_and_tail.
Luke Lau [Thu, 6 Apr 2023 14:26:31 +0000 (15:26 +0100)]
[RISCV] Combine concat_vectors of loads into strided loads
If we're concatenating several smaller loads separated by a stride, we
can try and increase the element size and perform a strided load.
For example:
```
concat_vectors (load v4i8, p+0), (load v4i8, p+n), (load v4i8, p+n*2), (load v4i8, p+n*3)
=>
vlse32 p, stride=n, VL=4
```
This pattern can be produced by the SLP vectorizer.
A special case is when the stride is exactly equal to the width of the
vector, in which case it can be converted into a single consecutive
vector load. For example:
```
concat_vectors (load v4i8, p), (load v4i8, p+4), (load v4i8, p+8), (load v4i8, p+12)
=>
vle8 p, VL=16
```
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D147713
Luke Lau [Thu, 6 Apr 2023 13:48:37 +0000 (14:48 +0100)]
[RISCV] Add tests for concats of vectors that could become strided loads
These patterns of concat_vector nodes of loads can be combined into
widened vector loads or a strided vector loads.
Co-authored-by: Philip Reames <preames@rivosinc.com>
Differential Revision: https://reviews.llvm.org/D147712
OCHyams [Wed, 19 Apr 2023 08:20:58 +0000 (09:20 +0100)]
Reapply D146987 "[Assignment Tracking] Enable by default"
This reverts commit
efc8b52cbd942f4bd5ffe8f64da5fb8a3b7adc32
which reverts D146987.
See https://reviews.llvm.org/D146987 for issues.
OCHyams [Wed, 19 Apr 2023 07:43:02 +0000 (08:43 +0100)]
[Assignment Tracking] Fix fragment error for some DSE-shortened stores
`shortenAssignment` inserts dbg.assigns with fragments describing the dead part
of a shortened store after each dbg.assign linked to the store.
Without this patch it doesn't take into account that the dead part of a
shortened store may be outside the bounds of a variable of a linked
dbg.assign. It also doesn't correctly account for a non-zero offset in the
address modifying `DIExpression` of the dbg.assign (which is possible for
fragments now even though whole variables currently cannot have a non-zero
offset in their alloca).
Fix this by moving the dead slice into variable-space and performing an
intersect of that adjusted slice with the existing fragment.
This fixes a verifier error reported when building fuchsia with assignment
tracking enabled:
https://ci.chromium.org/ui/p/fuchsia/builders/ci/
clang_toolchain.ci.core.x64-release/
b8784000953022145169/overview
Reviewed By: jmorse
Differential Revision: https://reviews.llvm.org/D148536
Krasimir Georgiev [Wed, 19 Apr 2023 08:26:38 +0000 (08:26 +0000)]
Revert "Reapply [SimplifyCFG][LICM] Preserve nonnull, range and align metadata when speculating"
This reverts commit
6f7e5c0f1ac6cc3349a2e1479ac4208465b272c6.
Seems to expose a miscompile in rust, possibly exposing a bug in LLVM
somewhere. Investigation thread over at:
https://rust-lang.zulipchat.com/#narrow/stream/187780-t-compiler.2Fwg-llvm/topic/LLVM.
20D146629.20breakage
Emmmer [Mon, 6 Feb 2023 08:17:14 +0000 (16:17 +0800)]
[LLDB][RISCV] Add RVV register infos
RVV stands for "RISC-V V Extension", which adds 32 vector registers, and seven unprivileged CSRs (vstart, vxsat, vxrm, vcsr, vtype, vl, vlenb) to a base scalar RISC-V ISA.
The base vector extension is intended to provide general support for data-parallel execution within the 32-bit instruction encoding space, with later vector extensions supporting richer functionality for certain domains.
This patch adds the definitions of RVV registers in `RegisterInfos_riscv64.h`, whose purpose is to provide support (such as reading, writing, and calculating the offsets) for future register-related functions.
Reviewed By: kito-cheng
Differential Revision: https://reviews.llvm.org/D143374
Pavel Kosov [Wed, 19 Apr 2023 07:59:07 +0000 (10:59 +0300)]
[llvm-exegesis] Allow setting dump file name
This will be used for writing test cases.
~~
Huawei RRI, OS Lab
Reviewed By: courbet
Differential Revision: https://reviews.llvm.org/D147700
Jean Perier [Wed, 19 Apr 2023 07:00:51 +0000 (09:00 +0200)]
[flang][hlfir] Apply component lower bounds in hlfir.designate codegen
The array component indices in the "path" of a fir.slice are zero based
because FIR does not know about the component lower bounds.
When lowering hlfir.designate to FIR for `array%x(i, j)`, convert `i` and
`j` to zero based indices before generating the fir.slice.
Differential Revision: https://reviews.llvm.org/D148627
Chen Zheng [Mon, 9 Jan 2023 04:07:52 +0000 (04:07 +0000)]
[GlobalISelEmitter] handle operand without MVT/class
There are some patterns in td files without MVT/class set
for some operands in target pattern that are from the source
pattern. This prevents GlobalISelEmitter from adding them as
a valid rule, because the target child operand is an
unsupported kind operand. For now, for a leaf child, only
IntInit and DefInit are handled in GlobalISelEmitter.
This issue can be workaround by adding MVT/class to the
patterns in the td files, like the workarounds for patterns
anyext and setcc in PPCInstrInfo.td in D140878.
To avoid adding the same workarounds for other patterns in
td files, this patch tries to handle the UnsetInit case in
GlobalISelEmitter.
Adding the new handling allows us to remove the workarounds
in the td files and also generates many selection rules for
PPC target.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D141247
Matthias Springer [Wed, 19 Apr 2023 06:49:20 +0000 (15:49 +0900)]
[mlir][linalg] ValueBoundsOpInterface: Add support for linalg.index
Differential Revision: https://reviews.llvm.org/D148598
Timm Bäder [Wed, 19 Apr 2023 06:10:13 +0000 (08:10 +0200)]
[clang][Interp][NFC] Small State.cpp refactoring
Alex Bradbury [Wed, 19 Apr 2023 06:03:57 +0000 (07:03 +0100)]
[RISCV] Fix canonical ordering of s* vs z* extensions in RISCVISAInfo
As noted in https://reviews.llvm.org/D148315, the ordering logic for
OrderedExtensionMap currently puts s* before z* extensions, but per the
ISA manual the correct order should be z* and then s* (with the
exception of zxm*, which are ordered after s*).
This patch fixes the ordering and adds a TODO for zxm*. The changes are
visible in the test case added in
a35e67fc5be654a7efdfa6125343b90f8960a487 which also demonstrates an
issue with the ordering of single letter extensions (which isn't
addressed in this patch).
This ordering matches the one used by GCC/binutils as well.
Differential Revision: https://reviews.llvm.org/D148615
Adrian Kuegel [Wed, 19 Apr 2023 05:56:28 +0000 (07:56 +0200)]
[clang][Bazel] Add missing dependency after
310ee08d5d4
Alex Bradbury [Wed, 19 Apr 2023 05:43:40 +0000 (06:43 +0100)]
[docs][RISCV] Use anonymous references in RISCVUsage to avoid warnings
2a5661c8415876be3fbd56ce90c2031e89ba0ef3 added a new external link with
the link text "0.2 draft specification". Surprisingly, as multiple links
have this same text but different targets this causes a warning, which
causes a failure on the llvm-sphinx-docs builder (which treats warnings
as errors). As suggested in
<https://github.com/sphinx-doc/sphinx/issues/3921>, this commit moves to
using anonymous references for the links in the experimental extensions
section.
Alex Bradbury [Wed, 19 Apr 2023 05:31:10 +0000 (06:31 +0100)]
[RISCV][NFC] Consistently use Opcode local variable in RISCVAsmParser::validateInstruction
As we already do `unsigned Opcode = Inst.getOpcode();`, we may as well
use that variable.
Alex Bradbury [Wed, 19 Apr 2023 05:24:53 +0000 (06:24 +0100)]
[RISCV] Bump Zfa version to 0.2 and correct RISCVUsage description
As of
https://github.com/riscv/riscv-isa-manual/commit/
1f038182810727f5feca311072e630d6baac51da
in the riscv-isa-manual, Zfa is at version 0.2. Reviewing the commit
history for
zfa.tex
<https://github.com/riscv/riscv-isa-manual/commits/master/src/zfa.tex>
there are no relevant changes since 0.1. As such, we can simply
increment the version number.
This change also removes the claim in RISCVUsage that we implement a
"subset of" Zfa, as I believe this is no longer true. That sentence
previously incorrectly claimed we didn't implement fli.{h,s,d} (I
[corrected this a couple of weeks
ago](https://reviews.llvm.org/rG3d969191b277)) but I think should have
removed the "subset of" wording too.
As was noted during the review, we never added Zfa to the release notes.
This is corrected in this patch.
Differential Revision: https://reviews.llvm.org/D148634
Serguei Katkov [Wed, 19 Apr 2023 05:05:29 +0000 (12:05 +0700)]
[InstSimplify] Update tests for max(min,max) to cover more types.
chenglin.bi [Wed, 19 Apr 2023 03:15:14 +0000 (11:15 +0800)]
Revert "[AMDGPU] Ressociate patterns with sub to use SALU"
The patch will caused dead loop because of DAGCombiner's canonicalization:
// (x + C) - y -> (x - y) + C
// y - (x + C) -> (y - x) - C
// (x - C) - y -> (x - y) - C
// (C - x) - y -> C - (x + y)
This reverts commit
b3529b5bf3ba2cd7f38665de16450afefb263c9b.
Yeting Kuo [Tue, 18 Apr 2023 08:47:33 +0000 (16:47 +0800)]
[RISCV] Customed lower vector nearbyint and rint in RISC-V.
The patch lowers vector rint/nearbyint like vp.rint/nearbyint.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D148619
wangpc [Wed, 19 Apr 2023 02:45:10 +0000 (10:45 +0800)]
[RISCV] Remove SEW=8 case for floating-point
For floating-point instructions, SEW won't be 8. So we don't need
to generate scheduling resources for it.
Reviewed By: michaelmaitland
Differential Revision: https://reviews.llvm.org/D148317
Matthias Springer [Wed, 19 Apr 2023 02:28:53 +0000 (11:28 +0900)]
[mlir][NFC] Minor cleanups around ShapedType
* Remove unnecessary casts.
* Use concrete shaped types (e.g., `MemRefType`, `RankedTensorType`) instead of `ShapedType` when possible.
* Minor documentation cleanups.
Differential Revision: https://reviews.llvm.org/D148488
Matthias Springer [Wed, 19 Apr 2023 02:00:48 +0000 (11:00 +0900)]
[mlir][IR] Remove ShapedType::getSizeInBits
This function returns incorrect values for memrefs and vectors due to "widening".
Differential Revision: https://reviews.llvm.org/D148501
Aart Bik [Wed, 19 Apr 2023 01:18:34 +0000 (18:18 -0700)]
bazel patch for OpenACC changes
Reviewed By: razvanlupusoru, anlunx
Differential Revision: https://reviews.llvm.org/D148678
Lang Hames [Wed, 19 Apr 2023 01:02:19 +0000 (18:02 -0700)]
Revert "[ExecutionEngine] Allow JIT tests to run on PowerPC."
This reverts commit
d771f54107c4889cde449d2bf5ba13d193017716 due to builder
failures, e.g. https://lab.llvm.org/buildbot#builders/93/builds/14488.
Alex Brachet [Wed, 19 Apr 2023 00:54:32 +0000 (00:54 +0000)]
[libc] Add fuchsia/io.h for TestLogger
b3c696fb86 started using TestLogger in tests. This class depends on
write_to_stderr which was previously only provided for Linux and
GPU's. Note, this function is used on those platforms in the libc
proper, though for Fuchsia we only ever want to provide this for
TestLogger.
No CMake files are changed because this fuchsia/ directory isn't
useful in any CMake build.
Differential Revision: https://reviews.llvm.org/D148675
Michael Jones [Mon, 10 Apr 2023 20:10:27 +0000 (13:10 -0700)]
[libc] move strerror and strsignal to OS msg maps
Other OSes may have different mappings from error number to message.
This creates a system to allow new platforms to define their own
mappings.
Reviewed By: sivachandra
Differential Revision: https://reviews.llvm.org/D147967
Razvan Lupusoru [Tue, 18 Apr 2023 23:28:18 +0000 (16:28 -0700)]
[mlir][openacc] Fix build after D148389
Buildbot reported undefined references to LLVM dialect and Memref
dialect. The issue is that OpenACC dialect now depends on those
(since it attaches interface to the types) but the cmake file
did not explicitly add those dependencies.
Reviewed By: clementval, vzakhari
Differential Revision: https://reviews.llvm.org/D148673
Peiming Liu [Tue, 18 Apr 2023 21:38:49 +0000 (21:38 +0000)]
[mlir][sparse] introduce a new compressed(hi) dimension level type
`compressed(hi)` is similar to `compressed`, but instead of reusing the previous position high as the current position low, it uses a pair of positions for each sparse index.
The patch only introduces the definition (syntax) but does not provide codegen implementation.
Reviewed By: aartbik
Differential Revision: https://reviews.llvm.org/D148664
Razvan Lupusoru [Tue, 18 Apr 2023 00:01:51 +0000 (17:01 -0700)]
[mlir][openacc] Add separate acc data operations for OpenACC data clauses
As outlined in [1], data clauses are now implemented as separate operations
from the constructs that they belong to. Some of the highlighted benefits:
- Correctly represent dataflow of data operations
- Easier to track debugging information
- Friendlier to add attributes and to optimize operations
For now, all of the other operand lists are being kept until all references
to them in LLVM can be removed (such as those in flang lowering)
[1] https://discourse.llvm.org/t/rfc-openacc-dialect-data-operation-improvements/69825
Reviewed By: clementval, vzakhari
Differential Revision: https://reviews.llvm.org/D148389
Craig Topper [Tue, 18 Apr 2023 22:13:54 +0000 (15:13 -0700)]
[TableGen] Reduce the amount of storage space used for AddrSpaces in TypeSetByHwMode. NFC
We reserved 16 AddrSpaces in every TypeSetByHwMode. But we only ever
use the first one on targets that make use of the AddrSpace feature.
The vector was populated by pushing for each entry in the ArrayRef
passed to the TypeSetByHwMode constructor. Each entry is a
ValueTypeByHwMode that stores one VT for each HwMode.
The vector is accessed by a loop in TypeSetByHwMode::getValueTypeByHwMode.
That loop is over HwModes with in the TypeSetByHwMode. This is
unrelated to how the vector was created. The entries in the vector
don't represent HwModes.
The targets that use AddrSpace don't make use of HwModes so the
loop in getValueTypeByHwMode will only run 1 iteration. So we only
the first entry in the vector is meaningful used.
This patch simplifies things by storing only 1 AddrSpace in
TypeSetByMode. Reducing the memory used by TypeSetByHwMode.
More work will be needed to support HwModes with AddrSpace if we
need a different AddrSpace for each HwMode.
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D148194
LLVM GN Syncbot [Tue, 18 Apr 2023 22:24:22 +0000 (22:24 +0000)]
[gn build] Port
c2114bd8e1f0
Valentin Churavy [Tue, 18 Apr 2023 22:21:48 +0000 (18:21 -0400)]
Revert "Non-debuginfo JITLink perf jitdump support"
This reverts commit
76e1521b0acff739c0425d0fcbb9360fc17f1af8.
Nathan Sidwell [Fri, 14 Apr 2023 21:12:05 +0000 (17:12 -0400)]
[ELF] hash fn return type
Be explicit about the 32bitness of the return type.
Differential Revision: https://reviews.llvm.org/D148379
Nathan Sidwell [Fri, 14 Apr 2023 12:51:05 +0000 (08:51 -0400)]
[BOLT] Don't enable runtime when not building X86
Noah Goldstein [Tue, 18 Apr 2023 21:34:56 +0000 (16:34 -0500)]
Recommit "[ValueTracking] Apply the isKnownNonZero techniques in `ashr`/`lshl` to `shl` and vice-versa" (2nd Try)
Wasn't related to the bug it was original thought to be causing.
Noah Goldstein [Tue, 18 Apr 2023 21:34:17 +0000 (16:34 -0500)]
Recommit "[InstCombine] Add transforms for `(icmp {u|s}ge/le (xor X, Y), X)`" (2nd Try)
Wasn't related to the bug it was original thought to be causing.
Noah Goldstein [Tue, 18 Apr 2023 21:34:02 +0000 (16:34 -0500)]
[InstCombine] Fix buggy `(mul X, Y)` -> `(shl X, Log2(Y))` transform PR62175
Bug was because we recognized patterns like `(shl 4, Z)` as a power of
2 we could take Log2 of (`2 + Z`), but doing `(shl X, (2 + Z))` can
cause a poison shift.
https://alive2.llvm.org/ce/z/yuJm_k
The fix is to verify that `Log2(Y)` will be a non-poisonous shift
amount. We can do this with:
`nsw` flag:
- https://alive2.llvm.org/ce/z/yyyJBr
- https://alive2.llvm.org/ce/z/YgubD_
`nuw` flag:
- https://alive2.llvm.org/ce/z/-4mpyV
- https://alive2.llvm.org/ce/z/a6ik6r
Prove `Y != 0`:
- https://alive2.llvm.org/ce/z/ced4su
- https://alive2.llvm.org/ce/z/X-JJHb
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D148609
Lang Hames [Tue, 18 Apr 2023 22:11:58 +0000 (15:11 -0700)]
[ExecutionEngine] Allow JIT tests to run on PowerPC.
These tests passed on my ppc64le test machine. If they survive testing by the
buildbots then we can leave them enabled, and this will allow us to land the
new ppc64 JITLink backend (https://reviews.llvm.org/D148192).
Fangrui Song [Tue, 18 Apr 2023 22:01:36 +0000 (15:01 -0700)]
[AsmPrinter] Fix placement of function entry comments
The placement is currently wrong in the presence of function entry related
instrumentations (prefixdata, -fpatchable-function-entry=, -fsanitize=kcfi,
etc).
Caslyn Tonelli [Thu, 30 Mar 2023 00:03:05 +0000 (00:03 +0000)]
[gwp_asan] Employ EXPECT_DEATH for zxtest compatibility
Employ a similar tactic introduced by https://reviews.llvm.org/D94362
for gwp_asan tests. zxtest `ASSERT_DEATH` syntax differs from gtest in
that it expects a lambda.
zxtest does not have `EXPECT_DEATH`, so it introduced for Fuchsia builds
and wraps the expression with a lambda to create a compatible syntax
between zxtest and gtest for death tests.
An example of where this compatiblity is needed is in
`never_allocated.cpp`.
Differential Revision: https://reviews.llvm.org/D147189
Rahman Lavaee [Tue, 18 Apr 2023 21:19:28 +0000 (21:19 +0000)]
Refactor BasicBlockSectionsProfileReader::getBBClusterInfoForFunction.
LLVM GN Syncbot [Tue, 18 Apr 2023 21:16:29 +0000 (21:16 +0000)]
[gn build] Port
76e1521b0acf
Prem Chintalapudi [Tue, 18 Apr 2023 21:15:32 +0000 (17:15 -0400)]
Non-debuginfo JITLink perf jitdump support
This patch ports PerfJITEventListener to a JITLink plugin, but adds unwind record support and drops debuginfo support temporarily. Debuginfo can be enabled in the future by providing a way to obtain a DWARFContext from a LinkGraph.
See D146060 for an experimental implementation that adds debuginfo parsing.
Reviewed By: lhames
Differential Revision: https://reviews.llvm.org/D146169
Lei Zhang [Tue, 18 Apr 2023 20:57:35 +0000 (13:57 -0700)]
[mlir][spirv][vector] Enable vector3 when converting to dot product
It's common to see such cases for contraction from convolution with
input channel as 3. Although we aren't utilizing all 4 lanes for
dot product, it should still be better than performing the multiply
and reduction separately.
Reviewed By: kuhar
Differential Revision: https://reviews.llvm.org/D148642
Bill Wendling [Mon, 17 Apr 2023 23:31:42 +0000 (16:31 -0700)]
[Clang][NFC] Rename methods/vars to reflect their real usage
The "getField" method is a bit confusing considering we also have a
"getFieldName" method. Instead, use "getFieldDecl" rather than
"getField".
Differential Revision: https://reviews.llvm.org/D147743
Jason Molenda [Tue, 18 Apr 2023 20:01:49 +0000 (13:01 -0700)]
Remove hardcoded address bits in ABIMacOS_arm64
A default number of addressing bits was hardcoded in
ABIMacOSX_arm64::FixAddress while we updated different
environments to fetch the value dynamically. Remove
the old hardcoded value.
Differential Revision: https://reviews.llvm.org/D148603
rdar://
108068497
Vitaly Buka [Tue, 18 Apr 2023 19:43:13 +0000 (12:43 -0700)]
[NFC][lsan] Rename test function
Vitaly Buka [Tue, 18 Apr 2023 19:41:43 +0000 (12:41 -0700)]
[lsan] Disabled D148281 test on Darwin
For unknown reason it fails to link.
Siva Chandra Reddy [Fri, 31 Mar 2023 19:19:58 +0000 (12:19 -0700)]
[libc] Add two more recipes to do a cross build.
Reviewed By: jeffbailey
Differential Revision: https://reviews.llvm.org/D147350
Alexey Bataev [Thu, 13 Apr 2023 18:28:54 +0000 (11:28 -0700)]
[SLP]Add final resize to ShuffleCostEstimator::finalize member function and basic add member functions.
Implemented the reshuffling in finalize member function + add basic
support for add member functions, used during vector build.
Part of D110978
Differential Revision: https://reviews.llvm.org/D148279