This patch changes VPWidenRecipe to only store a single original IR
instruction. This is the first required step towards modeling it's
operands as VPValues and also towards breaking it up into a
VPInstruction.

Discussed as part of D74695.

Reviewers: Ayal, gilr, rengolin

Reviewed By: gilr

Differential Revision: https://reviews.llvm.org/D76988

We use a SmallPtrSet to track visited nodes, use a SmallVector
of the same size for the stack.

Replace the explicit isAtom() || isSLM() test with the more general (and more specific) slowTwoMemOps() check to avoid the use of the PUSHrmm push from memory case.

This is actually very tricky to test in anything but quite complex code, but the atomic-idempotent.ll tests seem to be the most straightforward to use.

Differential Revision: https://reviews.llvm.org/D76239

Summary:
On targets with different pointer sizes, -alignment-from-assumptions could attempt to create SCEV expressions which use different effective SCEV types. The provided test illustrates the issue.

In `getNewAlignment`, AASCEV would be the (only) alloca, which would have an effective SCEV type of i32. But PtrSCEV, the GEP in this case, due to being in the flat/default address space, will have an effective SCEV of i64.

This patch resolves the issue by truncating PtrSCEV to AASCEV's effective type.

Reviewers: hfinkel, jdoerfert

Reviewed By: jdoerfert

Subscribers: jvesely, nhaehnle, hiraditya, javed.absar, kerbowa, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D75471

Generalizes ad7199f3 (R_PPC_NONE/R_PPC64_NONE).

The __builtin_stdarg_start is the legacy spelling of __builtin_va_start.
It should behave exactly the same, but for the last 9 years it would
behave subtly different for diagnostics. Follow the change from
29ad95b2 to require custom type checking.

Found by msan.

Found by llvm::sort's expensive checks.

Dropping unreachable code may reduce use counts on other instructions,
so it's better to do this earlier rather than later.

NFC-ish, may only impact worklist order.

Merge AddReachableCodeToWorklist() into prepareICWorklistFromFunction().
It's one logical step, and this makes it easier to move code.

This is safe if the iterator type is a pointer and the comparator is
stateless. The enable_if pattern I'm adding here only uses
array_pod_sort for the default comparator (std::less).

Using array_pod_sort has a potential performance impact, but I didn't
notice anything when testing clang. Sorting doesn't seem to be on the
hot path anywhere in LLVM.

Shrinks Release+Asserts clang by 73k.

Found by the expensive checks in llvm::sort.

Currently, bpf does not specify 128bit alignment in its
layout spec. So for a structure like
  struct ipv6_key_t {
    unsigned pid;
    unsigned __int128 saddr;
    unsigned short lport;
  };
clang will generate IR type
  %struct.ipv6_key_t = type { i32, [12 x i8], i128, i16, [14 x i8] }
Additional padding is to ensure later IR->MIR can generate correct
stack layout with target layout spec.

But it is common practice for a tracing program to be
first compiled with target flag (e.g., x86_64 or aarch64) through
clang to generate IR and then go through llc to generate bpf
byte code. Tracing program often refers to kernel internal
data structures which needs to be compiled with non-bpf target.

But such a compilation model may cause a problem on aarch64.
The bcc issue https://github.com/iovisor/bcc/issues/2827
reported such a problem.

For the above structure, since aarch64 has "i128:128" in its
layout string, the generated IR will have
  %struct.ipv6_key_t = type { i32, i128, i16 }

Since bpf does not have "i128:128" in its spec string,
the selectionDAG assumes alignment 8 for i128 and
computes the stack storage size for the above is 32 bytes,
which leads incorrect code later.

The x86_64 does not have this issue as it does not have
"i128:128" in its layout spec as it does permits i128 to
be alignmented at 8 bytes at stack. Its IR type looks like
  %struct.ipv6_key_t = type { i32, [12 x i8], i128, i16, [14 x i8] }

The fix here is add i128 support in layout spec, the same as
aarch64. The only downside is we may have less optimal stack
allocation in certain cases since we require 16byte alignment
for i128 instead of 8. But this is probably fine as i128 is
not used widely and in most cases users should already
have proper alignment.

Differential Revision: https://reviews.llvm.org/D76587

There was already a test case for landingpads to handle this case, but I
had forgotten to consider PHI instructions preceding the EH_LABEL in the
landingpad.

PR45261

To make sure that replaced operands get DCEd. This drops one
iteration from gepphigep.ll, which is still not optimal.

This was the last test case performing more than 3 iterations.

NFC-ish, only worklist order should change.

The `NewGEP->setOperand(DI, NewPN)` call was duplicated, and the
insertion of NewGEP is the same in both if/else, so we can extract it.

Gets rid of a 150k static initializer (Release clang)

Tested on Linux with Clang 9, and on Windows with Visual Studio 2019 16.5.1 with -DLLVM_ENABLE_THREADS=ON and OFF.

Because this code does not use the IC-aware replaceInstUsesWith()
helper, we need to manually push users to the worklist.

This is NFC-ish, in that it may only change worklist order.

This particular case will stop needing multiple iterations in
a followup change.

This fixes PR# 45336.
Output sections described in a linker script as NOLOAD with no input sections would be marked as SHT_PROGBITS.

Reviewed By: MaskRay

Differential Revision: https://reviews.llvm.org/D76981

This statement

    if (ReplaceWith == S) ReplaceWith = UndefValue::get(S->getType());

is introduced in https://reviews.llvm.org/rG35609d97ae89b8e13f40f4e6b9b056954f8baa83
to fix a case where unreachable code can cause select instruction
simplification to fail. In https://reviews.llvm.org/rGd10480657527ffb44ea213460fb3676a6b1300aa,
we begin to perform a depth-first walk of basic blocks. This means
we will not visit unreachable blocks. So we do not need this the
special check any more.

Differential Revision: https://reviews.llvm.org/D76753

MC already knows how to emulate the .weak directive (with its ELF
semantics; i.e., an undefined weak symbol resolves to 0, and a defined
weak symbol has lower link precedence than a strong symbol of the same
name) using COFF weak externals. Plumb this through the ASM printer too,
so that definitions marked with __attribute__((weak)) at the language
level (which gets translated to weak linkage at the IR level) have the
corresponding .weak directive emitted. Note that declarations marked
with __attribute__((weak)) at the language level (which translates to
extern_weak at the IR level) already have .weak directives emitted.

Weak*/linkonce* symbols without an associated comdat (in particular, ones
generated with __attribute__((weak)) in C/C++) were earlier emitted as
normal unique globals, as the comdat is required to provide the linkonce
semantics. This change makes sure they are emitted as .weak instead,
allowing other symbols to override them.

Rename the existing coff-weak.ll test to coff-linkonce.ll. I'm not
quite sure what that test covers, since the behavior being tested in it
(the emission of a one_only section) is just a result of passing
-function-sections to llc; the linkonce_odr makes no difference.

Add a new coff-weak.ll which tests the new directive emission.

Based on an previous patch by Shoaib Meenai.

Differential Revision: https://reviews.llvm.org/D44543

The preprocessor reads the whole line even if the first condition of an and is false so this broke when compiling on older gcc versions which don't recognize `__has_builtin`

The LatticeVal alias was introduced to reduce the diff size for the
transition to ValueLatticeElement, which is done now.

This patch removes the unnecessary alias and updates some very verbose
type uses with auto.

LatticeVal is an alias for ValueLatticeElement and the function is not
used any longer.

Reviewers: sammccall

Subscribers: cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D75446

This seems to be used in some resource files, e.g.
https://github.com/wxWidgets/wxWidgets/blob/f3217573d7240411e7817c9d76d965b2452987a2/include/wx/msw/wx.rc#L28.

MSVC rc.exe and GNU windres both allow any value here, and silently
just truncate to uint16_t range. This just explicitly allows the
-1 value and errors out on others - the same was done for control
IDs in dialogs in c1a67857.

Differential Revision: https://reviews.llvm.org/D76951

Including some test coverage for PR38522