As reported in Bug 42535, `clang` doesn't inline atomic ops on 32-bit
Sparc, unlike `gcc` on Solaris. In a 1-stage build with `gcc`, only two
testcases are affected (currently `XFAIL`ed), while in a 2-stage build more
than 100 tests `FAIL` due to this issue.
The reason for this `gcc`/`clang` difference is that `gcc` on 32-bit
Solaris/SPARC defaults to `-mpcu=v9` where atomic ops are supported, unlike
with `clang`'s default of `-mcpu=v8`. This patch changes `clang` to use
`-mcpu=v9` on 32-bit Solaris/SPARC, too.
Doing so uncovered two bugs:
`clang -m32 -mcpu=v9` chokes with any Solaris system headers included:
/usr/include/sys/isa_defs.h:461:2: error: "Both _ILP32 and _LP64 are defined"
#error "Both _ILP32 and _LP64 are defined"
While `clang` currently defines `__sparcv9` in a 32-bit `-mcpu=v9`
compilation, neither `gcc` nor Studio `cc` do. In fact, the Studio 12.6
`cc(1)` man page clearly states:
These predefinitions are valid in all modes:
[...]
__sparcv8 (SPARC)
__sparcv9 (SPARC -m64)
At the same time, the patch defines `__GCC_HAVE_SYNC_COMPARE_AND_SWAP_[1248]`
for a 32-bit Sparc compilation with any V9 cpu. I've also changed
`MaxAtomicInlineWidth` for V9, matching what `gcc` does and the Oracle
Developer Studio 12.6: C User's Guide documents (Ch. 3, Support for Atomic
Types, 3.1 Size and Alignment of Atomic C Types).
The two testcases that had been `XFAIL`ed for Bug 42535 are un-`XFAIL`ed
again.
Tested on `sparcv9-sun-solaris2.11` and `amd64-pc-solaris2.11`.
Differential Revision: https://reviews.llvm.org/D86621
The LinalgTilingPattern class dervied from the base deletes the
original operation. This allows for the use case where the more
transformations are necessary on the original operation after
tiling. In such cases the pattern can derive from
LinalgBaseTilingPattern instead of LinalgTilingPattern.
Differential Revision: https://reviews.llvm.org/D87308
When inlining functions containing allocas of scalable vectors we
cannot specify the size in the lifetime markers, since we don't
know this at compile time.
Added new test here:
test/Transforms/Inline/AArch64/sve-alloca-merge.ll
Differential Revision: https://reviews.llvm.org/D87139
Check that all passes, which report they preserve CFG,
are really preserving CFG.
A new standard instrumentation is introduced. It can be
switched on/off by the flag verify-cfg-preserved, which
is on by default for debug builds.
Reviewers: kuhar, fedor.sergeev
Differential Revision: https://reviews.llvm.org/D81558
This gives a pretty substantial size reduction; for a 6.5 MB
DLL with 300 KB .xdata, the .xdata shrinks by 66 KB.
Differential Revision: https://reviews.llvm.org/D87369
Convert 2-byte opcodes to equivalent 1-byte ones.
Adjust the existing exhaustive testcase to avoid being altered by
the simplification rules (to keep that test exercising all individual
opcodes).
Fix the assembler parser limits for register pairs; for .seh_save_regp
and .seh_save_regp_x, we can allow up to x29, for a x29+x30 pair
(which gets remapped to the UOP_SaveFPLR(X) opcodes), for .seh_save_fregp
and .seh_save_fregpx, allow up to d14+d15.
Not creating .seh_save_next for float register pairs, as the
actual unwinder implementation in current versions of Windows is buggy
for that case.
This gives a minimal but measurable size reduction. (For a 6.5 MB
DLL with 300 KB .xdata, the .xdata shrinks by 48 bytes. The opcode
sequences are padded to a 4 byte boundary, so very small improvements
might not end up mattering directly.)
Differential Revision: https://reviews.llvm.org/D87367
This patch updates the documentation about `__builtin_memcpy_inline` and reorders the sections so it is more consitent and understandable.
Differential Revision: https://reviews.llvm.org/D87458
This is to fix CodeView build failure https://bugs.llvm.org/show_bug.cgi?id=47287
after DIsSubrange upgrade D80197
Assert condition is now removed and Count is calculated in case LowerBound
is absent or zero and Count or UpperBound is constant. If Count is unknown
it is later handled as VLA (currently Count is set to zero).
Reviewed By: rnk
Differential Revision: https://reviews.llvm.org/D87406
In addition to calculate hash consistently by swapping SELECT's
operands, we also need to inverse the select pattern favor to match the
original logic.
[EarlyCSE] Equivalent SELECTs should hash equally
DenseMap<SimpleValue> assumes that, if its isEqual method returns true
for two elements, then its getHashValue method must return the same value
for them. This invariant is broken when one SELECT node is a min/max
operation, and the other can be transformed into an equivalent min/max by
inverting its predicate and swapping its operands. This patch fixes an
assertion failure that would occur intermittently while compiling the
following IR:
define i32 @t(i32 %i) {
%cmp = icmp sle i32 0, %i
%twin1 = select i1 %cmp, i32 %i, i32 0
%cmpinv = icmp sgt i32 0, %i
%twin2 = select i1 %cmpinv, i32 0, i32 %i
%sink = add i32 %twin1, %twin2
ret i32 %sink
}
Differential Revision: https://reviews.llvm.org/D86843
Update both thread and stack.
Update thread and stack as atomic operation.
Keep all 32bit of TID as now we have enough bits.
Depends on D87135.
Reviewed By: morehouse
Differential Revision: https://reviews.llvm.org/D87217
An upcoming change to Scudo will change how we use the TLS slot
in tsd_shared.h, which will be a little easier to deal with if
we can remove the code path that calls pthread_getspecific and
pthread_setspecific. The only known user of this code path is Fuchsia.
We can't eliminate this code path by making Fuchsia use ELF TLS
because although Fuchsia supports ELF TLS, it is not supported within
libc itself. To address this, Roland McGrath on the Fuchsia team has
proposed that Scudo will optionally call a platform-provided function
to access a TLS slot reserved for Scudo. Android also has a reserved
TLS slot, but the code that accesses the TLS slot lives in Scudo.
We can eliminate some complexity and duplicated code by having Android
use the same mechanism that was proposed for Fuchsia, which is what
this change does. A separate change to Android implements it.
Differential Revision: https://reviews.llvm.org/D87420
This patch adds support for dumping the .debug_ranges section to
elf2yaml.
Reviewed By: jhenderson
Differential Revision: https://reviews.llvm.org/D87429
With optimizations we leave the decision to eliminate fallthrough branches to
bock placement, but at -O0 we should do it in the selector to save code size.
This regressed -O0 with a recent change to a combiner.
Halide users reported this here: https://llvm.org/pr46176
I reported the issue to MSVC here:
https://developercommunity.visualstudio.com/content/problem/1179643/msvc-copies-overaligned-non-trivially-copyable-par.html
This codepath is apparently not covered by LLVM's unit tests, so I added
coverage in a unit test.
If we want to support this configuration going forward, it means that is
in general not safe to pass a SmallVector<T, N> by value if alignof(T)
is greater than 4. This doesn't appear to come up often because passing
a SmallVector by value is inefficient and not idiomatic: it copies the
inline storage. In this case, the SmallVector<LLT,4> is captured by
value by a lambda, and the lambda is passed by value into std::function,
and that's how we hit the bug.
Differential Revision: https://reviews.llvm.org/D87475
When we define a derived type that extends another derived type, we can then
create a structure constructor that contains values for the fields of both the
child type and its parent. The compiler's internal representation of that
value contains the name of the parent type where a component name would
normally appear. This caused an assert during contant folding.
There are three cases for components that appear in structure constructors.
The first is the normal case of a component appearing in a structure
constructor for its type.
The second is a component of the parent (or grandparent) type appearing in a
structure constructor for the child type.
The third is the parent type component, which can appear in the structure
constructor of its child.
There are also cases where the component can be arrays.
I created the test case folding12.f90 that covers all of these cases and
modified the code to handle them.
Most of my changes were to the "Find()" method of the type
"StructureConstructor" where I added code to cover the second and third cases
described above. To handle these cases, I needed to create a
"StructureConstructor" for the parent type component and return it. To handle
returning a newly created "StructureConstructor", I changed the return type of
"Find()" to be "std::optional" rather than an ordinary pointer.
This change supersedes D86172.
Differential Revision: https://reviews.llvm.org/D87151
The tests have been updated and I plan to move them from the MSSA
directory up.
Some end-to-end tests needed small adjustments. One difference to the
legacy DSE is that legacy DSE also deletes trivially dead instructions
that are unrelated to memory operations. Because MemorySSA-backed DSE
just walks the MemorySSA, we only visit/check memory instructions. But
removing unrelated dead instructions is not really DSE's job and other
passes will clean up.
One noteworthy change is in llvm/test/Transforms/Coroutines/ArgAddr.ll,
but I think this comes down to legacy DSE not handling instructions that
may throw correctly in that case. To cover this with MemorySSA-backed
DSE, we need an update to llvm.coro.begin to treat it's return value to
belong to the same underlying object as the passed pointer.
There are some minor cases MemorySSA-backed DSE currently misses, e.g. related
to atomic operations, but I think those can be implemented after the switch.
This has been discussed on llvm-dev:
http://lists.llvm.org/pipermail/llvm-dev/2020-August/144417.html
For the MultiSource/SPEC2000/SPEC2006 the number of eliminated stores
goes from ~17500 (legayc DSE) to ~26300 (MemorySSA-backed). More numbers
and details in the thread on llvm-dev.
Impact on CTMark:
```
Legacy Pass Manager
exec instrs size-text
O3 + 0.60% - 0.27%
ReleaseThinLTO + 1.00% - 0.42%
ReleaseLTO-g. + 0.77% - 0.33%
RelThinLTO (link only) + 0.87% - 0.42%
RelLO-g (link only) + 0.78% - 0.33%
```
http://llvm-compile-time-tracker.com/compare.php?from=3f22e96d95c71ded906c67067d75278efb0a2525&to=ae8be4642533ff03803967ee9d7017c0d73b0ee0&stat=instructions
```
New Pass Manager
exec instrs. size-text
O3 + 0.95% - 0.25%
ReleaseThinLTO + 1.34% - 0.41%
ReleaseLTO-g. + 1.71% - 0.35%
RelThinLTO (link only) + 0.96% - 0.41%
RelLO-g (link only) + 2.21% - 0.35%
```
http://195.201.131.214:8000/compare.php?from=3f22e96d95c71ded906c67067d75278efb0a2525&to=ae8be4642533ff03803967ee9d7017c0d73b0ee0&stat=instructions
Reviewed By: asbirlea, xbolva00, nikic
Differential Revision: https://reviews.llvm.org/D87163
DenseMap<SimpleValue> assumes that, if its isEqual method returns true
for two elements, then its getHashValue method must return the same value
for them. This invariant is broken when one SELECT node is a min/max
operation, and the other can be transformed into an equivalent min/max by
inverting its predicate and swapping its operands. This patch fixes an
assertion failure that would occur intermittently while compiling the
following IR:
define i32 @t(i32 %i) {
%cmp = icmp sle i32 0, %i
%twin1 = select i1 %cmp, i32 %i, i32 0
%cmpinv = icmp sgt i32 0, %i
%twin2 = select i1 %cmpinv, i32 0, i32 %i
%sink = add i32 %twin1, %twin2
ret i32 %sink
}
Differential Revision: https://reviews.llvm.org/D86843
Making MaterializationResponsibility instances immovable allows their
associated VModuleKeys to be updated by the ExecutionSession while the
responsibility is still in-flight. This will be used in the upcoming
removable code feature to enable safe merging of resource keys even if
there are active compiles using the keys being merged.
Add DemandedBits / BDCE support for min/max intrinsics: If the low
bits are not demanded in the result, they also aren't demanded in
the operands.
Differential Revision: https://reviews.llvm.org/D87161
Currently AMDGPU does not support sanitizer. Disable
sanitizer options for now until they are supported.
Differential Revision: https://reviews.llvm.org/D87461
The PointerReg arg was passed into the dependence function for an
assertion which no longer exists. So, this patch updates the dependence
functions to avoid the PointerReg in the signature.
Tests-Run: make check
Bail from maskIsAllZeroOrUndef and maskIsAllOneOrUndef prior to iterating over the number of
elements for scalable vectors.
Assert that the mask type is not scalable in possiblyDemandedEltsInMask .
Assert that the types are correct in all three functions.
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D87424
Currently the global operator!=(bool, bool) is selected due to the implicit bool
conversion operator. Since this is never the desired semantics, we give it a
standard operator!= and make the bool conversion explicit.
Depends On D86809
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D86810
This allows Dialect to follow the MLIR style of nullable objects, and in fact is expected by `Dialect::operator bool() const` which already tests whether `def == nullptr`. This just wasn't a reachable situation, because the constructor was dereferencing the pointer unconditionally.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D86807
Also make the behavior of getting a dialect more forgiving, in the case where
there isn't a dialect associated with an attribute.
Depends On D86807
Reviewed By: mehdi_amini
Differential Revision: https://reviews.llvm.org/D86809
Currently, using llvm-objdump to disassemble a function containing
unreachable will trigger an assertion while decoding the opcode, since both
unreachable and debug_unreachable have the same encoding. To avoid this, set
unreachable as the canonical decoding.
Differential Revision: https://reviews.llvm.org/D87431
Previously we could match fcmp+select to a reduction if the fcmp had
the nonans fast math flag. But if the select had the nonans fast
math flag, InstCombine would turn it into a fminnum/fmaxnum intrinsic
before SLP gets to it. Seems fairly likely that if one of the
fcmp+select pair have the fast math flag, they both would.
My plan is to start vectorizing the fmaxnum/fminnum version soon,
but I wanted to get this code out as it had some of the strangest
fast math flag behaviors.
Rainer Orth