Simplify `HeaderSearch::LookupFile`. Instead of deconstructing a
`FileEntryRef` into a name and `FileEntry` and then rebuilding it later,
use it as is. This helps to unblock making the constructor of
`FileEntryRef` private to `FileManager`.
Differential Revision:
In `ReplayPreamble::replay`, use `getFileRef` instead of `getFile`, and
then use that `FileEntryRef` later to avoid needing
`FileEntryRef::FileEntryRef`. The latter is going to become private to
`FileManager` in a later commit.
This dependency was already existing indirectly, but is now more direct
since the registration relies on a inline function. This fixes the
link of the tools with BFD.
Duplicated callsites share the same callee profile if the original callsite was inlined. The sharing also causes the profile of callee's callee to be shared. This breaks the assert introduced ealier by D84997 in a tricky way.
To illustrate, I'm using an abstract example. Say we have three functions `A`, `B` and `C`. A calls B twice and B calls C once. Some optimize performed prior to the sample profile loader duplicates first callsite to `B` and the program may look like
```
A()
{
B(); // with nested profile B1 and C1
B(); // duplicated, with nested profile B1 and C1
B(); // with nested profile B2 and C2
}
```
For some reason, the sample profile loader inliner then decides to only inline the first callsite in `A` and transforms `A` into
```
A()
{
C(); // with nested profile C1
B(); // duplicated, with nested profile B1 and C1
B(); // with nested profile B2 and C2.
}
```
Here is what happens next:
1. Failing to inline the callsite `C()` results in `C1`'s samples returned to `C`'s base (outlined) profile. In the meantime, `C1`'s head samples are updated to `C1`'s entry sample. This also affects the profile of the middle callsite which shares `C1` with the first callsite.
2. Failing to inline the middle callsite results in `B1` returned to `B`'s base profile, which in turn will cause `C1` merged into `B`'s base profile. Note that the nest `C` profile in `B`'s base has a non-zero head sample count now. The value actually equals to `C1`'s entry count.
3. Failing to inline last callsite results in `B2` returned to `B`'s base profile. Note that the nested `C` profile in `B`'s base now has an entry count equal to the sum of that of `C1` and `C2`, with the head count equal to that of `C1`. This will trigger the assert later on.
4. Compiling `B` using `B`'s base profile. Failing to inline `C` there triggers the returning of the nested `C` profile. Since the nested `C` profile has a non-zero head count, the returning doesn't go through. Instead, the assert goes off.
It's good that `C1` is only returned once, based on using a non-zero head count to ensure an inline profile is only returned once. However C2 is never returned. While it seems hard to solve this perfectly within the current framework, I'm just removing the broken assert. This should be reasonably fixed by the upcoming CSSPGO work where counts returning is based on context-sensitivity and a distribution factor for callsite probes.
The simple example is extracted from one of our internal services. In reality, why the original callsite `B()` and duplicate one having different inline behavior is a magic. It has to do with imperfect counts in profile and extra complicated inlining that makes the hotness for them different.
Reviewed By: wenlei
Differential Revision: https://reviews.llvm.org/D90056
As mentioned in the comment inside the code, the Intel documentation
states that the internal CPU buffer is flushed out to RAM only when tracing is
disabled. Otherwise, the buffer on RAM might be stale.
This diff disables tracing when the trace buffer is going to be read. This is a
quite safe operation, as the reading is done when the inferior is paused at a
breakpoint, so we are not losing any packets because there's no code being
executed.
After the reading is finished, tracing is enabled back.
It's a bit hard to write a test for this now, but Greg Clayton and I will
refactor the PT support and writing tests for it will be easier. However
I tested it manually by doing a script that automates
the following flow
```
(lldb) b main
Breakpoint 1: where = a.out`main + 15 at main.cpp:4:7, address = 0x000000000040050f
(lldb) r
Process 3078226 stopped
* thread #1, name = 'a.out', stop reason = breakpoint 1.1
frame #0: 0x000000000040050f a.out`main at main.cpp:4:7
(lldb) processor-trace start
(lldb) b 5
Breakpoint 2: where = a.out`main + 22 at main.cpp:5:12, address = 0x0000000000400516
(lldb) c
Process 3078226 resuming
Process 3078226 stopped
* thread #1, name = 'a.out', stop reason = breakpoint 2.1
frame #0: 0x0000000000400516 a.out`main at main.cpp:5:12
(lldb) processor-trace show-instr-log
thread #1: tid=3078226
0x40050f <+15>: movl $0x0, -0x8(%rbp)
>>> Before, some runs of the script up to this point lead to empty traces
(lldb) b 6
Breakpoint 3: where = a.out`main + 42 at main.cpp:6:14, address = 0x000000000040052a
(lldb) c
Process 3092991 resuming
Process 3092991 stopped
* thread #1, name = 'a.out', stop reason = breakpoint 3.1
frame #0: 0x000000000040052a a.out`main at main.cpp:6:14
(lldb) processor-trace show-instr-log thread #1: tid=3092991
0x40050f <+15>: movl $0x0, -0x8(%rbp)
0x400516 <+22>: movl $0x0, -0xc(%rbp)
0x40051d <+29>: cmpl $0x2710, -0xc(%rbp) ; imm = 0x2710
0x400524 <+36>: jge 0x400546 ; <+70> at main.cpp
0x400524 <+36>: jge 0x400546 ; <+70> at main.cpp
>>> The trace was re-enabled correctly and includes the instruction of the
first reading.
```
Those instructions correspond to these lines
```
3 int main() {
4 int z = 0;
5 for (int i = 0; i < 10000; i++) {
6 z += fun(z)
...
```
Differential Revision: https://reviews.llvm.org/D85241
lambda-expression's captures.
The built-in structured binding rules for classes require that all
fields can be accessed by name, and the fields introduced for lambda
captures are unnamed, so decomposing a capturing lambda is ill-formed.
-print_full_coverage=1 produces a detailed branch coverage dump when run on a single file.
Uses same infrastructure as -print_coverage flag, but prints all branches (regardless of coverage status) in an easy-to-parse format.
Usage: For internal use with machine learning fuzzing models which require detailed coverage information on seed files to generate mutations.
Differential Revision: https://reviews.llvm.org/D85928
Reverts the XFAIL added in b67a2aef8a,
which had no effect.
Adjust the test to make sure all output is dumped to stderr, so that
hopefully I can get a better idea of where/why this is failing.
Remove some redundant checking while here.
This doesn't support -structurizecfg-skip-uniform-regions since that
would require porting LegacyDivergenceAnalysis.
The NPM doesn't support adding a non-analysis pass as a dependency of
another, so I had to add -lowerswitch to some tests or pin them to the
legacy PM.
This is the only RegionPass in tree, so I simply copied the logic for
finding all Regions from the legacy PM's RGManager into
StructurizeCFG::run().
Reviewed By: arsenm
Differential Revision: https://reviews.llvm.org/D89026
This makes it pass under the NPM.
The legacy PM pass ran passes on SCCs in a different order, causing
argpromotion to not trigger on @bar().
Reviewed By: rnk
Differential Revision: https://reviews.llvm.org/D89889
This diff adds the option -prepend_rpath which inserts an rpath as
the first rpath in the binary.
Test plan: make check-all
Differential revision: https://reviews.llvm.org/D89605
This change introduces a GC parseable lowering for element atomic
memcpy/memmove intrinsics. This way runtime can provide an
implementation which can take a safepoint during copy operation.
See "GC-parseable element atomic memcpy/memmove" thread on llvm-dev
for the background and details:
https://groups.google.com/g/llvm-dev/c/NnENHzmX-b8/m/3PyN8Y2pCAAJ
Differential Revision: https://reviews.llvm.org/D88861
I added a test to verify that the associated symbol did not have errors before
doing the anaylsis of a call to a component ref along with a test that
triggers the original problem.
Differential Revision: https://reviews.llvm.org/D90074
The current pattern for vector unrolling takes the native shape to
unroll to at pattern instantiation time, but the native shape might
defer based on the types of the operand. Introduce a
UnrollVectorOptions struct which allows for using a function that will
return the native shape based on the operation. Move other options of
unrolling like `filterConstraints` into this struct.
Differential Revision: https://reviews.llvm.org/D89744
While implementing inline stack traces on Windows I noticed that the stack
traces in many asan tests included an inlined frame that shouldn't be there.
Currently we get the PC and then do a stack unwind and use the PC to
find the beginning of the stack trace.
In the failing tests the first thing in the stack trace is inside an inline
call site that shouldn't be in the stack trace, so replace it with the PC.
Differential Revision: https://reviews.llvm.org/D89996
For performance reasons the reproducers don't copy the files captured by
the file collector eagerly, but wait until the reproducer needs to be
generated.
This is a problematic when LLDB crashes and we have to do all this
signal-unsafe work in the signal handler. This patch uses a similar
trick to clang, which has the driver invoke a new cc1 instance to do all
this work out-of-process.
This patch moves the writing of the mapping file as well as copying over
the reproducers into a separate process spawned when lldb crashes.
Differential revision: https://reviews.llvm.org/D89600
Add folder for the case where ExtractStridedSliceOp source comes from a chain
of InsertStridedSliceOp. Also add a folder for the trivial case where the
ExtractStridedSliceOp is a no-op.
Differential Revision: https://reviews.llvm.org/D89850
Use `LineOffsetMapping:get` directly and remove/inline the helper
`ComputeLineNumbers`, simplifying the callers.
Differential Revision: https://reviews.llvm.org/D89922
It's currently ambiguous in IR whether the source language explicitly
did not want a stack a stack protector (in C, via function attribute
no_stack_protector) or doesn't care for any given function.
It's common for code that manipulates the stack via inline assembly or
that has to set up its own stack canary (such as the Linux kernel) would
like to avoid stack protectors in certain functions. In this case, we've
been bitten by numerous bugs where a callee with a stack protector is
inlined into an __attribute__((__no_stack_protector__)) caller, which
generally breaks the caller's assumptions about not having a stack
protector. LTO exacerbates the issue.
While developers can avoid this by putting all no_stack_protector
functions in one translation unit together and compiling those with
-fno-stack-protector, it's generally not very ergonomic or as
ergonomic as a function attribute, and still doesn't work for LTO. See also:
https://lore.kernel.org/linux-pm/20200915172658.1432732-1-rkir@google.com/https://lore.kernel.org/lkml/20200918201436.2932360-30-samitolvanen@google.com/T/#u
Typically, when inlining a callee into a caller, the caller will be
upgraded in its level of stack protection (see adjustCallerSSPLevel()).
By adding an explicit attribute in the IR when the function attribute is
used in the source language, we can now identify such cases and prevent
inlining. Block inlining when the callee and caller differ in the case that one
contains `nossp` when the other has `ssp`, `sspstrong`, or `sspreq`.
Fixes pr/47479.
Reviewed By: void
Differential Revision: https://reviews.llvm.org/D87956
We were returning the default constructed unique_pointer from
TypeSystem.h for which the compiler does not have a definition. Move the
implementation into the cpp file.
On AIX, to support vector types, which should always be 16 bytes aligned,
we set alloca to return 16 bytes aligned memory space.
Differential Revision: https://reviews.llvm.org/D89910
This does not change anything at the moment, but needed for
D89170. In that change I am probing a physical SGPR to see if
it is legal. RC is SReg_32, but DRC for scratch instructions
is SReg_32_XEXEC_HI and test fails.
That is sufficient just to check if DRC contains a register
here in case of physreg. Physregs also do not use subregs
so the subreg handling below is irrelevant for these.
Differential Revision: https://reviews.llvm.org/D90064
The change in 0ba9843397 changed the behaviour of the build when
using an XL build compiler because `-G` is not a pure linker option:
it also implies `-shared`. This was accounted for in the base CMake
configuration, so an analysis of the change from 0ba9843397 in
relation to a build using Clang (where `-shared` is introduced by CMake)
would not identify the issue. This patch resolves this particular issue
by adding `-shared` alongside `-Wl,-G`.
At the same time, the investigation reveals that several aspects of the
various build configurations are not operating in the manner originally
intended.
The other issue related to the `-G` linker option in the build is that
the removal of it (to avoid unnecessary use of run-time linking) is not
effective for the build using the Clang compiler. This patch addresses
this by adjusting the regular expressions used to remove the broadly-
applied `-G`.
Finally, the issue of specifying the export list with `-Wl,` instead of
a compiler option is flagged with a FIXME comment.
Reviewed By: daltenty, amyk
Differential Revision: https://reviews.llvm.org/D90041