A new pass DXILOpLowering was added.
It will scan all llvm intrinsics, create dxil op function if it can map to dxil op function.
Then translate call instructions on the intrinsic into call on dxil op function.
dxil op function will add i32 argument to the begining of args for dxil opcode.
So cannot use setCalledFunction to update the call instruction on intrinsic.
This commit only support sin to start the work.
Reviewed By: kuhar, beanz
Differential Revision: https://reviews.llvm.org/D124805
The patch make PseudoReadVL have the vtypes of the corresponding VLEFF/VLSEGFF.
It's useful to get the vtypes of locations of PseudoReadVL without finding the
corresponding VLEFF/VLSEGFF.
It could simplify optimizations in RISCVInsertVSETVLI like D123581.
Reviewed By: craig.topper
Differential Revision: https://reviews.llvm.org/D125199
This patch adds rvv codegen support for vp.fpext. The lowering of fp_round, vp.fptrunc, fp_extend and vp.fpext share most code so use a common lowering function to handle these four.
And this patch changes the intermediate cast from ISD::FP_EXTEND/ISD::FP_ROUND to the RVV VL version op RISCVISD::FP_EXTEND_VL and RISCVISD::FP_ROUND_VL for scalable vectors.
Reviewed By: frasercrmck
Differential Revision: https://reviews.llvm.org/D123975
Similar to change D125046.
If a programmer is able to compile and link a program that contains types that
are not yet supported by the runtime, it must be because they're not yet
implemented.
This change will make it easier to find unimplemented code in tests.
Differential Revision: https://reviews.llvm.org/D125267
Commit 8a63326150ee ("[BPF] Mark FI_ri as isPseudo to avoid
assertion during disassembly") added isPseudo to FI_ri insn
in BPFInstrInfo.td file. This patch added the missing test file.
Differential Revision: https://reviews.llvm.org/D125185
When a specific sequence of bytes is present in the file during
disassembly the disassembler fails with the following assertion:
...
0: 18 20 00 00 00 00 00 00 lea
... Assertion `idx < size()' failed.
...
llvm::SmallVectorTemplateCommon<...>::operator[](...) ...
llvm::MCInst::getOperand(unsigned int) ...
llvm::BPFInstPrinter::printOperand(...) ...
llvm::BPFInstPrinter::printInstruction() ...
llvm::BPFInstPrinter::printInst(...) ...
...
The byte sequence causing the error is (little endian):
18 20 00 00 00 00 00 00 00 00 00 00 00 00 00 00
The issue could be reproduced using the program bellow:
test.ir:
@G = constant
[16 x i8]
[i8 u0x18, i8 u0x20, i8 u0x00, i8 u0x00, i8 u0x00, i8 u0x00, i8 u0x00, i8 u0x00,
i8 u0x00, i8 u0x00, i8 u0x00, i8 u0x00, i8 u0x00, i8 u0x00, i8 u0x00, i8 u0x00],
section "foo", align 8
Compiled and disassembled as follows:
cat test.ir | llc -march=bpfel -filetype=obj -o - \
| llvm-objdump --arch=bpfel --section=foo -d -
This byte sequence corresponds to FI_ri instruction declared in the
BPFInstrInfo.td as follows:
def FI_ri
: TYPE_LD_ST<BPF_IMM.Value, BPF_DW.Value,
(outs GPR:$dst),
(ins MEMri:$addr),
"lea\t$dst, $addr",
[(set i64:$dst, FIri:$addr)]> {
// This is a tentative instruction, and will be replaced
// with MOV_rr and ADD_ri in PEI phase
let Inst{51-48} = 0;
let Inst{55-52} = 2;
let Inst{47-32} = 0;
let Inst{31-0} = 0;
let BPFClass = BPF_LD;
}
Notes:
- First byte (opcode) is formed as follows:
- BPF_IMM.Value is 0x00
- BPF_DW.Value is 0x18
- BPF_LD is 0x00
- Second byte (registers) is formed as follows:
- let Inst{55-52} = 2;
- let Inst{51-48} = 0;
The FI_ri instruction is always replaced by MOV_rr ADD_ri instructions
pair in the BPFRegisterInfo::eliminateFrameIndex method. Thus, this
instruction should be invisible to disassembler. This patch achieves
this by adding "isPseudo" flag for this instruction.
The bug was found by decompiling of one of the BPF tests from Linux
kernel (llvm-objdump -D tools/testing/selftests/bpf/bpf_iter_sockmap.o)
Differential Revision: https://reviews.llvm.org/D125185
Inquiries into the bounds, size, and length of local variables (and function results)
are acceptable specification expressions. A recent change allowed them for dummy
arguments that are not OPTIONAL or INTENT(OUT), but didn't address other object
entities.
Differential Revision: https://reviews.llvm.org/D125343
We're having a hard time booting the ARCH=i386 Linux kernel with clang
after removing -ffreestanding because instcombine was dropping inreg
from callers during libcall simplification, but not the callees defined
in different translation units. This led the callers and callees to have
wildly different calling conventions, which (predictably) blew up at
runtime.
Infer the inreg param attrs on function declarations from the module
metadata "NumRegisterParameters." This allows us to boot the ARCH=i386
Linux kernel (w/ -ffreestanding removed).
Fixes: https://github.com/llvm/llvm-project/issues/53645
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D125285
The original fix (commit 23ec5782c3cc) of
https://github.com/llvm/llvm-project/issues/52787 only adds `Function`s
that have `Instruction`s that directly use `BlockAddress`es into the
bitcode (`FUNC_CODE_BLOCKADDR_USERS`).
However, in either @rickyz's original reproducing code:
```
void f(long);
__attribute__((noinline)) static void fun(long x) {
f(x + 1);
}
void repro(void) {
fun(({
label:
(long)&&label;
}));
}
```
```
...
define dso_local void @repro() #0 {
entry:
br label %label
label: ; preds = %entry
tail call fastcc void @fun()
ret void
}
define internal fastcc void @fun() unnamed_addr #1 {
entry:
tail call void @f(i64 add (i64 ptrtoint (i8* blockaddress(@repro, %label) to i64), i64 1)) #3
ret void
}
...
```
or the xfs and overlayfs in the Linux kernel, `BlockAddress`es (e.g.,
`i8* blockaddress(@repro, %label)`) may first compose `ConstantExpr`s
(e.g., `i64 ptrtoint (i8* blockaddress(@repro, %label) to i64)`) and
then used by `Instruction`s. This case is not handled by the original
fix.
This patch adds *indirect* users of `BlockAddress`es, i.e., the
`Instruction`s using some `Constant`s which further use the
`BlockAddress`es, into the bitcode as well, by doing depth-first
searches.
Fixes: https://github.com/llvm/llvm-project/issues/52787
Fixes: 23ec5782c3cc ("[Bitcode] materialize Functions early when BlockAddress taken")
Reviewed By: nickdesaulniers
Differential Revision: https://reviews.llvm.org/D124878
Before dump, Insetad of switching to generic form silently after
verification failure. Print some debug logs to help identify why an op
may be printed in a different way.
Reviewed By: rriddle
Differential Revision: https://reviews.llvm.org/D125136
Move async copy operations to NVGPU as they only exist on NV target and are
designed to match ptx semantic. This allows us to also add more fine grain
caching hint attribute to the op.
Add hint to bypass L1 and hook it up to NVVM op.
Differential Revision: https://reviews.llvm.org/D125244
The current reordering scheme only checks the ordering of in-tree operands.
There are some cases, however, where we need to adjust the ordering based on
the ordering of a future SLP-tree who's instructions are not part of the
current tree, but are external users.
This patch is a simple implementation of this. We keep track of scalar stores
that are users of TreeEntries and if they look profitable to vectorize, then
we keep track of their ordering. During the reordering step we take this new
index order into account. This can remove some shuffles in cases like in the
lit test.
Differential Revision: https://reviews.llvm.org/D125111
This adds support for variable stride with the val, uval, and ref linear
modifiers. Previously only the no modifer type ls<argno> was supported.
val -> Ls<argno>
uval -> Us<argno>
ref -> Rs<argno>
Differential Revision: https://reviews.llvm.org/D125330
This patch adds cluster edges between independent MFMA instructions. Additionally, it propogates all predecessors of cluster insts to the root of the cluster(s), and all successors to the leaf(ves) of the cluster(s) -- this is done to remove the possibility that those insts will be interspersed within the cluster.
Reviewed By: kerbowa
Differential Revision: https://reviews.llvm.org/D124678
D98718 caused the order of Values/MemoryLocations we pass to alias() to
be significant due to storing the offset in the PartialAlias case. But
some callers weren't audited and were still passing swapped arguments,
causing the returned PartialAlias offset to be negative in some
cases. For example, the newly added unittests would return -1
instead of 1.
Fixes#55343, a miscompile.
Reviewed By: asbirlea, nikic
Differential Revision: https://reviews.llvm.org/D125328
The patch extends AArch64TTIImpl::instCombineIntrinsic to simplify
llvm.aarch64.neon.f{min,max}nm(a, a) -> a.
This helps with simplifying code written using the ACLE, e.g.
see https://godbolt.org/z/jYxsoc89c
Reviewed By: dmgreen
Differential Revision: https://reviews.llvm.org/D125234
The current implementation of `cloneWithNewYields` has a few issues
- It clones the loop body of the original loop to create a new
loop. This is very expensive.
- It performs `erase` operations which are incompatible when this
method is called from within a pattern rewrite. All erases need to
go through `PatternRewriter`.
To address these a new utility method `replaceLoopWithNewYields` is added
which
- moves the operations from the original loop into the new loop.
- replaces all uses of the original loop with the corresponding
results of the new loop
- use a call back to allow caller to generate the new yield values.
- the original loop is modified to just yield the basic block
arguments corresponding to the iter_args of the loop. This
represents a no-op loop. The loop itself is dead (since all its uses
are replaced), but is not removed. The caller is expected to erase
the op. Consequently, this method can be called from within a
`matchAndRewrite` method of a `PatternRewriter`.
The `cloneWithNewYields` could be replaces with
`replaceLoopWithNewYields`, but that seems to trigger a failure during
walks, potentially due to the operations being moved. That is left as
a TODO.
Differential Revision: https://reviews.llvm.org/D125147
CUDA/HIP programs use __noinline__ like a keyword e.g.
__noinline__ void foo() {} since __noinline__ is defined
as a macro __attribute__((noinline)) in CUDA/HIP runtime
header files.
However, gcc and clang supports __attribute__((__noinline__))
the same as __attribute__((noinline)). Some C++ libraries
use __attribute__((__noinline__)) in their header files.
When CUDA/HIP programs include such header files,
clang will emit error about invalid attributes.
This patch fixes this issue by supporting __noinline__ as
a keyword, so that CUDA/HIP runtime could remove
the macro definition.
Reviewed by: Aaron Ballman, Artem Belevich
Differential Revision: https://reviews.llvm.org/D124866
shuffle (cast X), (cast Y), Mask --> cast (shuffle X, Y, Mask)
This is similar to a recent transform with fneg ( b331a7ebc1e0 ),
but this is intentionally the most conservative first step to
try to avoid regressions in codegen. There are several
restrictions that could be removed as follow-up enhancements.
Note that a cast with a unary shuffle is currently canonicalized
in the other direction (shuffle after cast - D103038 ). We might
want to invert that to be consistent with this patch.
Summary:
We use the `--offload-new-driver` option to enable offload code
embedding. The check for when to do this was flawed and was enabling it
too early in the case of OpenMP, causing a segfault when dereferencing
the offloading toolchain.
Specifically for: !tbaa, !tbaa.struct, !annotation, !srcloc, !nosanitize.
The goal is to avoid test brittleness caused by hardcoded values.
Differential Revision: https://reviews.llvm.org/D123273
