[AArch64] compile crash on "Call operand has unhandled type"

[vfdff]

• test: https://gcc.godbolt.org/z/1fsW7cjG9

• report error: clang -march=armv8.3-a+sve -O3 -S reduced.ll

AArch64ISelLowering.cpp:7014: 
void analyzeCallOperands(const llvm::AArch64TargetLowering&, const llvm::AArch64Subtarget*, const llvm::TargetLowering::CallLoweringInfo&, llvm::CCState&): 
Assertion `!Res && "Call operand has unhandled type"' failed

[llvmbot]

@llvm/issue-subscribers-backend-aarch64

Author: Allen (vfdff)

* test: https://gcc.godbolt.org/z/1fsW7cjG9

• report error: clang -march=armv8.3-a+sve -O3 -S reduced.ll

AArch64ISelLowering.cpp:7014: 
void analyzeCallOperands(const llvm::AArch64TargetLowering&, const llvm::AArch64Subtarget*, const llvm::TargetLowering::CallLoweringInfo&, llvm::CCState&): 
Assertion `!Res && "Call operand has unhandled type"' failed

[dtcxzyw]

The original Godbolt link is broken.
Godbolt: https://godbolt.org/z/bo339Wjrh

[vfdff]

Thanks for your update, and here is a more simple case reduced by llvm-reduce, https://godbolt.org/z/n66zcKMbc

[vfdff]

maybe fixed by #70034

[paulwalker-arm]

I've raised #70847 (currently a WIP just to show intent) that fixes the core code generation problem by ensuring we still honour the existing NEON ABI (be that documented or historical) even when SVE is used to allow better code generation for fixed-length vectors bigger than 128-bit.

[vfdff]

Thanks for the great helping @paulwalker-arm

[vfdff]

Fixed with #70847