Recent shouldAssumeDSOLocal changes (introduced by 961f31d8ad14c66)
do not take in consideration the relocation model anymore. The ARM
fast-isel pass uses the function return to set whether a global symbol
is loaded indirectly or not, and without the expected information
llvm now generates an extra load for following code:
```
$ cat test.ll
@__asan_option_detect_stack_use_after_return = external global i32
define dso_local i32 @main(i32 %argc, i8** %argv) #0 {
entry:
%0 = load i32, i32* @__asan_option_detect_stack_use_after_return,
align 4
%1 = icmp ne i32 %0, 0
br i1 %1, label %2, label %3
2:
ret i32 0
3:
ret i32 1
}
attributes #0 = { noinline optnone }
$ lcc test.ll -o -
[...]
main:
.fnstart
[...]
movw r0, :lower16:__asan_option_detect_stack_use_after_return
movt r0, :upper16:__asan_option_detect_stack_use_after_return
ldr r0, [r0]
ldr r0, [r0]
cmp r0, #0
[...]
```
And without 'optnone' it produces:
```
[...]
main:
.fnstart
[...]
movw r0, :lower16:__asan_option_detect_stack_use_after_return
movt r0, :upper16:__asan_option_detect_stack_use_after_return
ldr r0, [r0]
clz r0, r0
lsr r0, r0, #5
bx lr
[...]
```
This triggered a lot of invalid memory access in sanitizers for
arm-linux-gnueabihf. I checked this patch both a stage1 built with
gcc and a stage2 bootstrap and it fixes all the Linux sanitizers
issues.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D95379
239cfbccb0509da1a08d9e746706013b732e646b add support for legalizing
i8/i16 UDIV/UREM/SDIV to use *W instructions. So we need to truncate
to i8/i16 if we're legalizing one of those.
The initial problem with the remaining bot config was resolved.
We can now use Python3. Let's use `os.cpu_count()` to cleanup this
helper.
Differential Revision: https://reviews.llvm.org/D94734
37510f69b4cb8d76064f108d57bebe95984a23ae tried to fix GCC 5.x compilation
by making the enum which is used as a unordered_map key unscoped. However it
seems that in GCC 5.x, enum keys are not supported *at all* in unordered_maps
(at least that's what some trial&error on godbolt tells me). This updates the
workaround to just use an int until GCC 5.x support is dropped.
The current context is thread-local state, and in preparation of GPU async execution (on multiple threads) we need to set the context before calling API that create resources.
Reviewed By: herhut
Differential Revision: https://reviews.llvm.org/D94495
If a function has stack objects, and a call, we require an FP. If we
did not initially have any stack objects, and only introduced them
during PrologEpilogInserter for CSR VGPR spills, SILowerSGPRSpills
would end up spilling the FP register as if it were a normal
register. This would result in an assert in a debug build, or
redundant handling of the FP register in a release build.
Try to predict that we will have an FP later, although this is ugly.
From this patch (plus some landed patches), `deviceRTLs` is taken as a regular OpenMP program with just `declare target` regions. In this way, ideally, `deviceRTLs` can be written in OpenMP directly. No CUDA, no HIP anymore. (Well, AMD is still working on getting it work. For now AMDGCN still uses original way to compile) However, some target specific functions are still required, but they're no longer written in target specific language. For example, CUDA parts have all refined by replacing CUDA intrinsic and builtins with LLVM/Clang/NVVM intrinsics.
Here're a list of changes in this patch.
1. For NVPTX, `DEVICE` is defined empty in order to make the common parts still work with AMDGCN. Later once AMDGCN is also available, we will completely remove `DEVICE` or probably some other macros.
2. Shared variable is implemented with OpenMP allocator, which is defined in `allocator.h`. Again, this feature is not available on AMDGCN, so two macros are redefined properly.
3. CUDA header `cuda.h` is dropped in the source code. In order to deal with code difference in various CUDA versions, we build one bitcode library for each supported CUDA version. For each CUDA version, the highest PTX version it supports will be used, just as what we currently use for CUDA compilation.
4. Correspondingly, compiler driver is also updated to support CUDA version encoded in the name of bitcode library. Now the bitcode library for NVPTX is named as `libomptarget-nvptx-cuda_[cuda_version]-sm_[sm_number].bc`, such as `libomptarget-nvptx-cuda_80-sm_20.bc`.
With this change, there are also multiple features to be expected in the near future:
1. CUDA will be completely dropped when compiling OpenMP. By the time, we also build bitcode libraries for all supported SM, multiplied by all supported CUDA version.
2. Atomic operations used in `deviceRTLs` can be replaced by `omp atomic` if OpenMP 5.1 feature is fully supported. For now, the IR generated is totally wrong.
3. Target specific parts will be wrapped into `declare variant` with `isa` selector if it can work properly. No target specific macro is needed anymore.
4. (Maybe more...)
Reviewed By: JonChesterfield
Differential Revision: https://reviews.llvm.org/D94745
This is a common symbol that's missing from our mapping because
cppreference yields multiple headers.
Add it manually by picking cstddef to prevent insertion of some stdlib-internal
headers instead.
Fixes https://github.com/clangd/clangd/issues/666.
Differential Revision: https://reviews.llvm.org/D95423
This follows up on the introduction of C API for the same object and is similar
to AffineExpr and AffineMap.
Reviewed By: stellaraccident
Differential Revision: https://reviews.llvm.org/D95437
The existing test has less FMF than we might expect if
our FMF was fixed (on all FP values), so this additional
test is intended to check propagation in a more "normal"
example.
The switch must set the predicate correctly; anything else
should lead to unreachable/assert.
I'm trying to fix FMF propagation here and the callers,
so this is a preliminary cleanup.
This patch adds additional checks to avoid partial unswitching
in cases where it won't be profitable, e.g. because the path directly
exits the loop anyways.
HasModifiers should be true if at least one modifier is used.
This should make the use of this field bit more consistent.
Differential Revision: https://reviews.llvm.org/D94795
Loop peeling removes conditions from loop bodies that become invariant
after a small number of iterations. When triggered, this leads to fewer
compares and possibly PHIs in loop bodies, enabling further
optimizations. The current cost-model of loop peeling should be quite
conservative/safe, i.e. only peel if a condition in the loop becomes
known after peeling.
For example, see PR47671, where loop peeling enables vectorization by
removing a PHI the vectorizer does not understand. Granted, the
loop-vectorizer could also be taught about constant PHIs, but loop
peeling is likely to enable other optimizations as well.
This has an impact on quite a few benchmarks from
MultiSource/SPEC2000/SPEC2006 on X86 with -O3 -flto, for example
Same hash: 186 (filtered out)
Remaining: 51
Metric: loop-vectorize.LoopsVectorized
Program base patch diff
test-suite...ve-susan/automotive-susan.test 8.00 9.00 12.5%
test-suite...nal/skidmarks10/skidmarks.test 35.00 31.00 -11.4%
test-suite...lications/sqlite3/sqlite3.test 41.00 43.00 4.9%
test-suite...s/ASC_Sequoia/AMGmk/AMGmk.test 25.00 26.00 4.0%
test-suite...006/450.soplex/450.soplex.test 88.00 89.00 1.1%
test-suite...TimberWolfMC/timberwolfmc.test 120.00 119.00 -0.8%
test-suite.../CINT2006/403.gcc/403.gcc.test 215.00 216.00 0.5%
test-suite...006/447.dealII/447.dealII.test 957.00 958.00 0.1%
test-suite...ternal/HMMER/hmmcalibrate.test 75.00 75.00 0.0%
Same hash: 186 (filtered out)
Remaining: 51
Metric: loop-vectorize.LoopsAnalyzed
Program base patch diff
test-suite...ks/Prolangs-C/agrep/agrep.test 440.00 434.00 -1.4%
test-suite...nal/skidmarks10/skidmarks.test 312.00 308.00 -1.3%
test-suite...marks/7zip/7zip-benchmark.test 6399.00 6323.00 -1.2%
test-suite...lications/minisat/minisat.test 134.00 135.00 0.7%
test-suite...rks/FreeBench/pifft/pifft.test 295.00 297.00 0.7%
test-suite...TimberWolfMC/timberwolfmc.test 1879.00 1869.00 -0.5%
test-suite...pplications/treecc/treecc.test 689.00 691.00 0.3%
test-suite...T2000/300.twolf/300.twolf.test 1593.00 1597.00 0.3%
test-suite.../Benchmarks/Bullet/bullet.test 1394.00 1392.00 -0.1%
test-suite...ications/JM/ldecod/ldecod.test 1431.00 1429.00 -0.1%
test-suite...6/464.h264ref/464.h264ref.test 2229.00 2230.00 0.0%
test-suite...lications/sqlite3/sqlite3.test 2590.00 2589.00 -0.0%
test-suite...ications/JM/lencod/lencod.test 2732.00 2733.00 0.0%
test-suite...006/453.povray/453.povray.test 3395.00 3394.00 -0.0%
Note the -11% regression in number of loops vectorized for skidmarks. I
suspect this corresponds to the fact that those loops are gone now (see
the reduction in number of loops analyzed by LV).
Reviewed By: lebedev.ri
Differential Revision: https://reviews.llvm.org/D88471
Compilers may insert new definitions during compilation, E.g. EH personality
function pointers, or named constant pool entries. This commit causes
ObjectLinkingLayer to attempt to claim responsibility for all weak definitions
in objects as they're linked. This is always safe (first claimant for each
symbol is granted responsibility, subsequent claims are rejected without error)
and prevents compiler-injected symbols from being dead-stripped (which they
will be if they remain unclaimed by anyone).
This change was motivated by errors seen by an out-of-tree client while testing
eh-frame support in JITLink ELF/x86-64: IR containing exceptions didn't define
DW.ref.__gxx_personality_v0 (since it's added by CodeGen), and this caused
DW.ref.__gxx_personality_v0 to be dead-stripped leading to linker failures.
No test case yet: We won't have a way to test in-tree until we enable JITLink
for lli on Linux.
The previous implementation required that `-maltivec` be specified when using either `-mabi=vec-extabi` or `-mabi=vec-default`, this patch removes that requirement.
Reviewed By: cebowleratibm
Differential Revision: https://reviews.llvm.org/D94986
This is required for ELF where PCRel32 doesn't implicitly subtract 4.
No test case yet: I haven't figured out a good way to test stub
generation -- this may required extensions to jitlink-check.
Depends On D95000
Move async.execute outlining and async -> async.runtime lowering into the separate Async transformation pass
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D95311
There were some right-shift tests in
CodeGen/AArch64/sve-int-arith-imm.ll
that were being folded away because we were shifting all the bits
out to the right. I've updated the tests to ensure this doesn't
happen.
We already set the `sh_entsize` field in a single place
for all non-implicit sections.
This patch reorders the logic slightly and with it
we finally have the only one place where the `sh_entsize` is set.
obj2yaml will not dump the `EntSize` key for `SHT_DYNSYM/SHT_SYMTAB` sections anymore,
when the value of `sh_entsize` is equal to `sizeof(Elf_Sym)`
Note that this also seems revealed an issue in llvm-objcopy:
Previously yaml2obj set the `sh_entsize` for the `.symtab` section to 0x18,
now we it sets it for `SHT_SYMTAB` sections, i.e. by type.
But the `llvm-objcopy/ELF/only-keep-debug.test` has a `.symtab` section of type `SHT_STRTAB`,
and now yaml2obj sets the `sh_entsize` to 0 for it.
I had to update the corresponding check lines for `ES`, but the behavior of
`llvm-objcopy` should be fixed instead I think.
I've added a TODO and a comment.
Differential revision: https://reviews.llvm.org/D95364
D95354 started to set the sh_link field for SHT_SYMTAB sections.
Previously it was set for symbol tables basing on their names (e.g. ".symtab").
This test now crashes see:
http://lab.llvm.org:8011/#/builders/68/builds/5911
I updated it to restore the old behavior.
Port some GPU-related language options to the marshalling system for automatic command line parsing and generation.
Reviewed By: dexonsmith
Differential Revision: https://reviews.llvm.org/D95345
This refactors the logic that sets the `sh_link` field.
With this patch we set it in a single place for all sections.
Differential revision: https://reviews.llvm.org/D95354
