| Issue |
203378
|
| Summary |
[MIPS] assertion failure in `lowerVECTOR_SHUFFLE_VSHF` with `+msa` and `<4 x i128>`
|
| Labels |
backend:MIPS
|
| Assignees |
|
| Reporter |
folkertdev
|
Some code from the rust standard library hits an assertion failure when compiled with `+msa`
https://godbolt.org/z/7f7j4nxTP
```llvm
define i4 @foo(<4 x i128> %0) {
start:
%1 = icmp ne <4 x i128> %0, zeroinitializer
%2 = bitcast <4 x i1> %1 to i4
ret i4 %2
}
```
we get
```
llc: /root/llvm-project/llvm/lib/Target/Mips/MipsSEISelLowering.cpp:3135: llvm::SDValue lowerVECTOR_SHUFFLE_VSHF(llvm::SDValue, llvm::EVT, const llvm::SmallVector<int, 16>&, bool, llvm::SelectionDAG&): Assertion `Indices[0] >= 0 && "shuffle mask starts with an UNDEF, which is not expected"' failed.
PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace and instructions to reproduce the bug.
Stack dump:
0. Program arguments: /opt/compiler-explorer/clang-assertions-trunk/bin/llc -o /app/output.s -x86-asm-syntax=intel -mtriple=mipsisa64r6el-unknown-linux-gnuabi64 -mattr=+msa -O3 <source>
1. Running pass 'Function Pass Manager' on module '<source>'.
```
Without `+msa` this compiles just fine
```asm
foo: # @foo
or $1, $8, $9
or $3, $10, $11
or $2, $4, $5
sltu $1, $zero, $1
sltu $3, $zero, $3
sltu $2, $zero, $2
sll $1, $1, 2
sll $3, $3, 3
or $1, $3, $1
or $3, $6, $7
sltu $3, $zero, $3
sll $3, $3, 1
or $2, $2, $3
jr $ra
or $2, $2, $1
```
![]()
_______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
