Age | Commit message (Collapse) | Author | Files | Lines |
|
|
|
Fixes #87102 .
|
|
Add annotated vtable GUID as referenced variables in per function
summary, and update bitcode writer to create value-ids for these
referenced vtables.
- This is the part3 of type profiling work, and described in the "Virtual Table Definition Import" [1] section of the
RFC.
[1] https://github.com/llvm/llvm-project/pull/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
|
|
The docstring for this function incorrectly specified when a widening is
not performed. This patch adds the additional specification for what
happens when the type size is a power of two but it is less than MinSize.
|
|
Finish the rename by #86604
|
|
This will eliminate some unnecessary REST API calls.
|
|
Use a seek offset that fits within the file size.
This was missed in presubmit because the FILE based stdio tests aren't
run in
overlay mode; fullbuild is not tested in presubmit.
WRITE_SIZE == 11, so using a value of 42 for offseto would cause the
expression
`WRITE_SIZE - offseto` to evaluate to -31 as an unsigned 64b integer
(18446744073709551585ULL).
Fixes #86928
|
|
|
|
We check if the next character after `N.` is `*` before we check its
length. Using `split` on the string is cleaner and less error prone than
using indices with `find` and `substr`.
Note: this does not make `N.` mean anything, it just prevents assertion
failures. `N.` is treated the same as an unrecognized breakpoint name:
```
(lldb) breakpoint enable 1
1 breakpoints enabled.
(lldb) breakpoint enable 1.*
1 breakpoints enabled.
(lldb) breakpoint enable 1.
0 breakpoints enabled.
(lldb) breakpoint enable xyz
0 breakpoints enabled.
```
Found via LLDB fuzzers.
|
|
Apple systems (#87219)
I was profiling our compiler and noticed that `llvm::get_threadid` was
at the top of the hotlist, taking up a surprising 5% (7 seconds) in the
profile trace. It seems that computing this on MacOS systems is
non-trivial, so cache the result in a thread_local.
Co-authored-by: Mehdi Amini <joker.eph@gmail.com>
|
|
(#79919)
As before we only skip instrumenting probe of `unwind`(`KnownColdBlock`)
block, this PR extends to skip the both EH flow from `invoke`, i.e. also
skip the `normal` dest. For more contexts: when doing call-to-invoke
conversion, the block is split by the `invoke` and two extra
blocks(`normal` and `unwind`) are added. With this PR, the
instrumentation is the same as the one before the call-to-invoke
conversion.
One significant benefit is this can help mitigate the "unstable IR"
issue(https://discourse.llvm.org/t/ipo-for-linkonce-odr-functions/69404),
the two versions now are on the same probe instrumentation, expected to
be the same checksum.
To achieve the same checksum, some tweaks is needed:
- Now it also skips incrementing the probe ID for the skipped probe.
- The checksum is also computed based on the CFG that skips the EH
edges.
We observed this fixes ~5% mismatched samples.
|
|
Extracted from #84858
|
|
Clean-up all of the calls and remove the redundant == 0 checks.
There is only one small visible change. For non-Android, the memalign
function will now fail if alignment is zero. Before this would have
passed.
|
|
This fixes #87227, a vbase can be placed below nvsize when empty members and/or bases are in play. We must account for that.
|
|
The ``dataclasses`` package makes sense for Python 3.6, becauses
``dataclasses`` is only included in the standard library with 3.7
version. Now, 3.6 has reached EOL, so all current supported versions of
Python (3.8, 3.9, 3.10, 3.11, 3.12) have this feature in their standard
libraries.
Therefore there's no need to install the ``dataclasses`` package now.
|
|
lvalues"
We don't have a document number for this, but the change was called out
explicitly in the editor's comments in the C99 foreword.
|
|
A precommit test case to show function summary and global values when a function has instructions annotated with vtable profiles and indirect call profiles.
- This is a precommit test for https://github.com/llvm/llvm-project/pull/79381
|
|
This patch moves OpenACC parts of `Sema` into a separate class
`SemaOpenACC` that is placed in a separate header `Sema/SemaOpenACC.h`.
This patch is intended to be a model of factoring things out of `Sema`,
so I picked a small OpenACC part.
Goals are the following:
1) Split `Sema` into manageable parts.
2) Make dependencies between parts visible.
3) Improve Clang development cycle by avoiding recompiling unrelated
parts of the compiler.
4) Avoid compile-time regressions.
5) Avoid notational regressions in the code that uses Sema.
|
|
(#87030)"
This reverts commit a4dec9d6bc67c4d8fbd4a4f54ffaa0399def9627
because the test failed in the following builder:
https://luci-milo.appspot.com/ui/p/fuchsia/builders/prod/clang-linux-x64/b8751864477467126481/overview
|
|
The content of a GOFF record is always dumped if NDEBUG is not defined,
which produces rather confusing output. This changes wrap the dumping
code in LLVM_DEBUG, so the dump is only done when debug output of this
module is requested.
|
|
Modeled after
https://github.com/llvm/llvm-project/commit/8513a681f7d8d1188706762e712168aebc3119dd#
According to
https://learn.microsoft.com/en-us/cpp/build/reference/fo-object-file-name?view=msvc-170,
`/Fo` accepts a trailing-colon variant. This is also tested in practice.
This allows clang-cl to parse this.
I just copied one of the existing tests, let me know if this is not the
best way to do this. I tested that the test does not pass beofre the
Options.td change, and that it does after.
See also #46065
|
|
|
|
We compute atan2f(y, x) in 2 stages:
- Fast step: perform computations in double precision , with relative
errors < 2^-50
- Accurate step: if the result from the Fast step fails Ziv's rounding
test, then we perform computations in double-double precision, with
relative errors < 2^-100.
On Ryzen 5900X, worst-case latency is ~ 200 clocks, compared to average
latency ~ 60 clocks, and average reciprocal throughput ~ 20 clocks.
|
|
|
|
Fixes #86928
|
|
The section discusses the reasons for the libraries inception more than
a decade ago. Now it discusses the progess libc++ has made, and the many
impressive acomplishments our contributors have brought it.
The initial section remains below.
---------
Co-authored-by: Louis Dionne <ldionne.2@gmail.com>
|
|
HLSL constant sized array function parameters do not decay to pointers.
Instead constant sized array types are preserved as unique types for
overload resolution, template instantiation and name mangling.
This implements the change by adding a new `ArrayParameterType` which
represents a non-decaying `ConstantArrayType`. The new type behaves the
same as `ConstantArrayType` except that it does not decay to a pointer.
Values of `ConstantArrayType` in HLSL decay during overload resolution
via a new `HLSLArrayRValue` cast to `ArrayParameterType`.
`ArrayParamterType` values are passed indirectly by-value to functions
in IR generation resulting in callee generated memcpy instructions.
The behavior of HLSL function calls is documented in the [draft language
specification](https://microsoft.github.io/hlsl-specs/specs/hlsl.pdf)
under the Expr.Post.Call heading.
Additionally the design of this implementation approach is documented in
[Clang's
documentation](https://clang.llvm.org/docs/HLSL/FunctionCalls.html)
Resolves #70123
|
|
|
|
Fixes: #85287
|
|
build errors (#87262)
`Linux/instrprof-vtable-value-prof.cpp` needs to be built for the test
to run. However, cpp compile & link failed with undefined-ABI error [1].
See original failure in
https://lab.llvm.org/buildbot/#/builders/18/builds/16429
[1]
```
FAIL: Profile-powerpc64 :: Linux/instrprof-vtable-value-prof.cpp (2406 of 2414)
******************** TEST 'Profile-powerpc64 :: Linux/instrprof-vtable-value-prof.cpp' FAILED ********************
Exit Code: 1
Command Output (stderr):
--
RUN: at line 3: /home/buildbots/llvm-external-buildbots/workers/ppc64be-sanitizer/sanitizer-ppc64be/build/build_debug/./bin/clang --driver-mode=g++ -m64 -ldl -fprofile-generate -fuse-ld=lld -O2 -g -fprofile-generate=. -mllvm -enable-vtable-value-profiling /home/buildbots/llvm-external-buildbots/workers/ppc64be-sanitizer/sanitizer-ppc64be/build/llvm-project/compiler-rt/test/profile/Linux/instrprof-vtable-value-prof.cpp -o /home/buildbots/llvm-external-buildbots/workers/ppc64be-sanitizer/sanitizer-ppc64be/build/build_debug/runtimes/runtimes-bins/compiler-rt/test/profile/Profile-powerpc64/Linux/Output/instrprof-vtable-value-prof.cpp.tmp-test
+ /home/buildbots/llvm-external-buildbots/workers/ppc64be-sanitizer/sanitizer-ppc64be/build/build_debug/./bin/clang --driver-mode=g++ -m64 -ldl -fprofile-generate -fuse-ld=lld -O2 -g -fprofile-generate=. -mllvm -enable-vtable-value-profiling /home/buildbots/llvm-external-buildbots/workers/ppc64be-sanitizer/sanitizer-ppc64be/build/llvm-project/compiler-rt/test/profile/Linux/instrprof-vtable-value-prof.cpp -o /home/buildbots/llvm-external-buildbots/workers/ppc64be-sanitizer/sanitizer-ppc64be/build/build_debug/runtimes/runtimes-bins/compiler-rt/test/profile/Profile-powerpc64/Linux/Output/instrprof-vtable-value-prof.cpp.tmp-test
ld.lld: error: /lib/../lib64/Scrt1.o: ABI version 1 is not supported
clang: error: linker command failed with exit code 1 (use -v to see invocation)
```
|
|
This includes algorithm.h directly to provide the definition for
`cpp:max` in parser.h. This will define `max(...)` in the libc namespace
for build systems that pull in parser.h explicitly.
|
|
Introduce 3 new optional attributes to the `transform.print` ops:
* `assume_verified`
* `use_local_scope`
* `skip_regions`
The primary motivation is to allow printing on large inputs that
otherwise take forever to print and verify. For the full context, see
this IREE issue: https://github.com/openxla/iree/issues/16901.
Also add some tests and fix the op description.
|
|
```
-----------------------------------------------------------------------------
Benchmark old new
-----------------------------------------------------------------------------
bm_mismatch_two_range_overload<char>/1 0.941 ns 1.88 ns
bm_mismatch_two_range_overload<char>/2 1.43 ns 2.15 ns
bm_mismatch_two_range_overload<char>/3 1.95 ns 2.55 ns
bm_mismatch_two_range_overload<char>/4 2.58 ns 2.90 ns
bm_mismatch_two_range_overload<char>/5 3.75 ns 3.31 ns
bm_mismatch_two_range_overload<char>/6 5.00 ns 3.83 ns
bm_mismatch_two_range_overload<char>/7 5.59 ns 4.35 ns
bm_mismatch_two_range_overload<char>/8 6.37 ns 4.84 ns
bm_mismatch_two_range_overload<char>/16 11.8 ns 6.72 ns
bm_mismatch_two_range_overload<char>/64 45.5 ns 2.59 ns
bm_mismatch_two_range_overload<char>/512 366 ns 12.6 ns
bm_mismatch_two_range_overload<char>/4096 2890 ns 91.6 ns
bm_mismatch_two_range_overload<char>/32768 23038 ns 758 ns
bm_mismatch_two_range_overload<char>/262144 142813 ns 6573 ns
bm_mismatch_two_range_overload<char>/1048576 366679 ns 26710 ns
bm_mismatch_two_range_overload<short>/1 0.934 ns 1.88 ns
bm_mismatch_two_range_overload<short>/2 1.30 ns 2.58 ns
bm_mismatch_two_range_overload<short>/3 1.76 ns 3.28 ns
bm_mismatch_two_range_overload<short>/4 2.24 ns 3.98 ns
bm_mismatch_two_range_overload<short>/5 2.80 ns 4.92 ns
bm_mismatch_two_range_overload<short>/6 3.58 ns 6.01 ns
bm_mismatch_two_range_overload<short>/7 4.29 ns 7.03 ns
bm_mismatch_two_range_overload<short>/8 4.67 ns 7.39 ns
bm_mismatch_two_range_overload<short>/16 9.86 ns 13.1 ns
bm_mismatch_two_range_overload<short>/64 38.9 ns 4.55 ns
bm_mismatch_two_range_overload<short>/512 348 ns 27.7 ns
bm_mismatch_two_range_overload<short>/4096 2881 ns 225 ns
bm_mismatch_two_range_overload<short>/32768 23111 ns 1715 ns
bm_mismatch_two_range_overload<short>/262144 184846 ns 14416 ns
bm_mismatch_two_range_overload<short>/1048576 742885 ns 57264 ns
bm_mismatch_two_range_overload<int>/1 0.838 ns 1.19 ns
bm_mismatch_two_range_overload<int>/2 1.19 ns 1.65 ns
bm_mismatch_two_range_overload<int>/3 1.83 ns 2.06 ns
bm_mismatch_two_range_overload<int>/4 2.38 ns 2.42 ns
bm_mismatch_two_range_overload<int>/5 3.60 ns 2.47 ns
bm_mismatch_two_range_overload<int>/6 3.68 ns 3.05 ns
bm_mismatch_two_range_overload<int>/7 4.32 ns 3.36 ns
bm_mismatch_two_range_overload<int>/8 5.18 ns 3.58 ns
bm_mismatch_two_range_overload<int>/16 10.6 ns 2.84 ns
bm_mismatch_two_range_overload<int>/64 39.0 ns 7.78 ns
bm_mismatch_two_range_overload<int>/512 247 ns 53.9 ns
bm_mismatch_two_range_overload<int>/4096 1927 ns 429 ns
bm_mismatch_two_range_overload<int>/32768 15569 ns 3393 ns
bm_mismatch_two_range_overload<int>/262144 125413 ns 28504 ns
bm_mismatch_two_range_overload<int>/1048576 504549 ns 112729 ns
```
|
|
(#66825)
(The profile format change is split into a standalone change into https://github.com/llvm/llvm-project/pull/81691)
* For InstrFDO value profiling, implement instrumentation and lowering for virtual table address.
* This is controlled by `-enable-vtable-value-profiling` and off by default.
* When the option is on, raw profiles will carry serialized `VTableProfData` structs and compressed vtables as payloads.
* Implement profile reader and writer support
* Raw profile reader is used by `llvm-profdata` but not compiler. Raw profile reader will construct InstrProfSymtab with symbol names, and map profiled runtime address to vtable symbols.
* Indexed profile reader is used by `llvm-profdata` and compiler. When initialized, the reader stores a pointer to the beginning of in-memory compressed vtable names and the length of string. When used in `llvm-profdata`, reader decompress the string to show symbols of a profiled site. When used in compiler, string decompression doesn't
happen since IR is used to construct InstrProfSymtab.
* Indexed profile writer collects the list of vtable names, and stores that to index profiles.
* Text profile reader and writer support are added but mostly follow the implementation for indirect-call value type.
* `llvm-profdata show -show-vtables <args> <profile>` is implemented.
rfc in
https://discourse.llvm.org/t/rfc-dynamic-type-profiling-and-optimizations-in-llvm/74600#pick-instrumentation-points-and-instrument-runtime-types-7
|
|
For more context on isa predicates, see:
https://github.com/llvm/llvm-project/pull/83753.
|
|
Store data is read later in the pipeline, so we use
SiFive7AnyToGPRBypass to model that a store instruction can begin some
cycles before that data is ready.
|
|
Need to drop skipping of the first zext/sext nodes, it leads to
incorrect and less profitable code.
|
|
element (#86974)
This is what SelectionDAG does. We'd like to reuse SelectionDAG
patterns.
|
|
Summary:
The previous code would potentially make it smaller if a device with a
lower ID touched it later. Also we should minimize changes to the state
for multi threaded reasons. This just sets up an owned slot for each at
initialization time.
|
|
non-contig… (#86781)
…uous function.
Fix from [thtsikas](https://github.com/thtsikas) based on a discussion
in
[slack](https://flang-compiler.slack.com/archives/C5C58TT32/p1711124374836079).
Example:
```
Program test
Integer, Pointer, Contiguous :: cont(:)
Interface
Function f()
Integer, Pointer :: f(:)
End Function
End Interface
cont => f()
Print *, cont(3)
End Program
Function f()
Integer, Pointer :: f(:)
Allocate (f(4),Source=[1,1,42,1])
! f => f(4:1:-1) !! not contiguous, runtime error
End Function f
```
Understanding is that the standard intended to allow this pattern. The
restriction 10.2.2.3 p6 Data pointer assignment "If the pointer object
has the CONTIGUOUS attribute, the pointer target shall be contiguous."
is not associated with a numbered constraint. If there is a mechanism
for injecting runtime checks, this would be a place to do it. Absent
that, a warning is the best we can do.
No other compiler treats contigPtr => func() as an error when func() is
not CONTIGUOUS, so a warning would probably be better for consistency.
https://godbolt.org/z/5cM6roeEE
|
|
Now that VPRecipeBase manages debug locations for recipes, use it in
VPWidenMemoryInstructionRecipe.
|
|
We are missing `[[maybe_unused]]`.
|
|
Inline the function into its single caller.
|
|
This suppresses warning `unused variable 'MRI' [-Wunused-variable]`
for those fusions that don't need `MRI`.
|
|
Add rounding mode attribute to `arith`. This attribute can be used in
different FP `arith` operations to control rounding mode. Rounding modes
correspond to IEEE 754-specified rounding modes. Use in `arith.truncf` folding.
As this is not supported in dialects other than LLVM, conversion should fail for
now in case this attribute is present.
---------
Signed-off-by: Victor Perez <victor.perez@codeplay.com>
|
|
(#86060)
1. Remove 'AllModes' and 'DefaultMode' suffixes for DecoderTables under
default HwMode.
2. Introduce a less aggressive suppression for HwMode DecoderTable, only
reduce necessary tables duplications. This allows encodings under
different HwModes to retain the original DecoderNamespace.
3. Change 'suppress-per-hwmode-duplicates' command option from bool type
to enum type, allowing users to choose what level of suppression to use.
|
|
decl offsets
This patch reorder the lexical block for the translation unit, visible update block for the TU and
the viisble upaete block for the extern C context after the type decl
offsets block.
This should be a NFC patch.
This is helpful for later optimizations for eliding unreachable
declarations in the global module fragment. See the comments in
https://github.com/llvm/llvm-project/pull/76930.
Simply, if we want to get the reachable sets of declaratins during the
writing process, we need to write the file-level context later than the
process of writing declarations (which is the main process to determine
the reachable set).
|
|
`RegisterClassInfo::getRegPressureSetLimit` has been changed to return a
smaller value than before so the limit may become negative in later
calculations. As a workaround, change to use
`TargetRegisterInfo::getRegPressureSetLimit`.
Also improve tests.
|
|
Fixes: #85243.
|
|
-- Convert `math.fpowi` to a series of `arith.mulf` operations.
-- If the power is negative, we divide the result by 1.
|