Now that these instructions are no longer prototypes, we do not need to be
careful about keeping them opt-in and can use the standard LLVM infrastructure
for them. This commit removes the bespoke intrinsics we were using to represent
these operations in favor of the corresponding target-independent intrinsics.
The clang builtins are preserved because there is no standard way to easily
represent these operations in C/C++.
For consistency with the scalar codegen in the Wasm backend, the intrinsic used
to represent {f32x4,f64x2}.nearest is @llvm.nearbyint even though
@llvm.roundeven better captures the semantics of the underlying Wasm
instruction. Replacing our use of @llvm.nearbyint with use of @llvm.roundeven is
left to a potential future patch.
Differential Revision: https://reviews.llvm.org/D100411
Some of Microsoft's unit tests in D70631 fail because libc++'s
implementation of std::chars_format isn't a proper bitmask type. Adding
the required functions to make std::chars_format a proper bitmask type.
Implements parts of P0067: Elementary string conversions
Differential Revision: https://reviews.llvm.org/D97115
Multiple lines importing from the same URL can be merged:
import {X} from 'a';
import {Y} from 'a';
Merge to:
import {X, Y} from 'a';
This change implements this merge operation. It takes care not to merge in
various corner case situations (default imports, star imports).
Differential Revision: https://reviews.llvm.org/D100466
Consider the following set of files:
a.cc:
#include "a.h"
a.h:
#ifndef A_H
#define A_H
#include "b.h"
#include "c.h" // This gets "skipped".
#endif
b.h:
#ifndef B_H
#define B_H
#include "c.h"
#endif
c.h:
#ifndef C_H
#define C_H
void c();
#endif
And the output of the -H option:
$ clang -c -H a.cc
. ./a.h
.. ./b.h
... ./c.h
Note that the include of c.h in a.h is not shown in the output (GCC does the
same). This is because of the include guard optimization: clang knows c.h is
covered by an include guard which is already defined, so when it sees the
include in a.h, it skips it. The same would have happened if #pragma once were
used instead of include guards.
However, a.h *does* include c.h, and it may be useful to show that in the -H
output. This patch adds a flag for doing that.
Differential revision: https://reviews.llvm.org/D100480
This transformation is fundamentally broken when it comes to dominance,
it just happened to work when the source of the memcpy can be moved into
the place of the alloca. The bug shows up a lot more often since
077bff39d4 allows the source to be a
switch.
It would be possible to check dominance of the source and all its
operands, but that seems very heavy for instcombine.
Rename the name of "LDS lowering" pass from `amdgpu-disable-lower-module-lds` to
`amdgpu-enable-lower-module-lds` as later is consistent and reads better.
Reviewed By: JonChesterfield
Differential Revision: https://reviews.llvm.org/D100441
In the fold SHUFFLE(BINOP(X,Y),BINOP(Z,W)) -> BINOP(SHUFFLE(X,Z),SHUFFLE(Y,W)), check if both X/Z AND Y/W have at least one merge-able shuffle in which case the total number of shuffle should still fall.
Helps with instruction count regressions we saw while fixing PR48823
Only attempt to propagateIRFlags if we have both SelectInst - afaict we shouldn't have matched a min/max reduction without both SelectInst, but static analyzer doesn't know that.
The existing BTI placement pass avoids inserting "BTI c" when the
function has local linkage and is only directly called. However,
even in this case, there is a (small) chance that the linker later
adds a hunk with an indirect call to the function, e.g. if the
function is placed in a separate section and moved far away from
its callers. Make sure to add BTI for these functions too.
Differential Revision: https://reviews.llvm.org/D99417
This refactors SCCP and creates a SCCPSolver interface and class so that it can
be used by other passes and transformations. We will use this in D93838, which
adds a function specialisation pass.
This is based on an early version by Vinay Madhusudan.
Differential Revision: https://reviews.llvm.org/D93762
ICC permits this, and after some extensive testing it looks like we can
support this with very little trouble. We intentionally don't choose to
do this with attribute-target (despite it likely working as well!)
because GCC does not support that, and introducing said
incompatibility doesn't seem worth it.
Stepping through callstacks in the example from D99759 reveals
this potential compile-time improvement.
The savings come from avoiding ValueTracking's computing known
bits if we have already dealt with special-case patterns.
Further improvements in this direction seem possible.
This makes a degenerate test based on PR49785 about 40x faster
(25 sec -> 0.6 sec), but it does not address the larger question
of how to limit computeKnownBitsFromAssume(). Ie, the original
test there is still infinite-time for all practical purposes.
Differential Revision: https://reviews.llvm.org/D100408
Otherwise it reuses the same register for storing the stack slot
offset if the stack slot offset is big.
Differential Revision: https://reviews.llvm.org/D100461
The start value can't be null for something to be a non-zero
recurrence, so hoist that common check out of the switch.
Subsequent checks may be incomplete or over-specified as noted in:
D100408
These were added in 37935405ef,
but they fail on macOS (and on Windows with MSYS based tools, before
relanding D98859). Remove the tests that exercise "not not echo", as
the primary thing to test is the plain echo patterns above.
Aggregate types over 16 bytes are passed by reference.
Contrary to the x86_64 ABI, smaller structs with an odd (non power
of two) are padded and passed in registers.
Differential Revision: https://reviews.llvm.org/D100374
By checking for cpu and toolchain features ahead
of time we don't need the custom return codes.
Reviewed By: omjavaid
Differential Revision: https://reviews.llvm.org/D97684
After 077bff39d4,
isDereferenceableForAllocaSize() can recurse into selects,
which is causing a problem for the new test case,
reduced from https://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20210412/904154.html
because the replacement (the select) is defined after the first use
of an alloca, so we'd end up with a verifier error.
Now, this new check is too restrictive.
We likely can handle *some* cases, by trying to sink all uses of an alloca
to after the the def.
Extension to rG74f98391a7a4, we can also include any of the upper (known zero) bits in the comparison in the shuffle removal fold, just as long as we demand all the elements of the movmsk source vector.
This avoids breaking clang-tidy/infrastructure/validate-check-names.cpp
if 'not' is evaluated as a lit internal tool (making TestRunner
invoke 'grep' directly in that test, instead of invoking 'not', which
then invokes 'grep').
The quoting of arguments is still brittle if the executable is an
MSYS based tool though, as MSYS based tools incorrectly unescape
backslashes in quoted arguments (contrary to regular win32 argument
parsing rules), see D99406 and
https://github.com/msys2/msys2-runtime/issues/36 for more examples
of the issues.
Differential Revision: https://reviews.llvm.org/D99938
This fixes breakage on Windows/ARM64 after D94355.
Modelled after the corresponding code for X86; not entirely familiar
with those aspects of that layer otherwise.
Differential Revision: https://reviews.llvm.org/D99572
Also remove a superfluous semicolon after the braces for a switch
statement (that wasn't warned about).
Differential Revision: https://reviews.llvm.org/D100447
According to i386 System V ABI:
1. when __m256 are required to be passed on the stack, the stack pointer must be aligned on a 0 mod 32 byte boundary at the time of the call.
2. when __m512 are required to be passed on the stack, the stack pointer must be aligned on a 0 mod 64 byte boundary at the time of the call.
The current method of clang passing __m512 parameter are as follow:
1. when target supports avx512, passing it with 64 byte alignment;
2. when target supports avx, passing it with 32 byte alignment;
3. Otherwise, passing it with 16 byte alignment.
Passing __m256 parameter are as follow:
1. when target supports avx or avx512, passing it with 32 byte alignment;
2. Otherwise, passing it with 16 byte alignment.
This pach will passing __m128/__m256/__m512 following i386 System V ABI and
apply it to Linux only since other System V OS (e.g Darwin, PS4 and FreeBSD) don't
want to spend any effort dealing with the ramifications of ABI breaks at present.
Differential Revision: https://reviews.llvm.org/D78564
Sander de Smalen