================
@@ -3264,17 +3268,28 @@ CIRGenFunction::emitAArch64BuiltinExpr(unsigned
builtinID, const CallExpr *expr,
case NEON::BI__builtin_neon_vcvtm_u64_v:
case NEON::BI__builtin_neon_vcvtmq_u64_v:
case NEON::BI__builtin_neon_vcvtn_s16_f16:
- case NEON::BI__builtin_neon_vcvtn_s32_v:
case NEON::BI__builtin_neon_vcvtnq_s16_f16:
- case NEON::BI__builtin_neon_vcvtnq_s32_v:
case NEON::BI__builtin_neon_vcvtn_u16_f16:
- case NEON::BI__builtin_neon_vcvtn_u32_v:
case NEON::BI__builtin_neon_vcvtnq_u16_f16:
+ cgm.errorNYI(expr->getSourceRange(),
+ std::string("unimplemented AArch64 builtin call: ") +
+ getContext().BuiltinInfo.getName(builtinID));
+ return mlir::Value{};
+
+ case NEON::BI__builtin_neon_vcvtn_s32_v:
----------------
banach-space wrote:
Yes, it's just about testing and keeping track of what's implemented and tested
:)
The number of variants for the conversion intrinsics is rather large (non-fp16
ones: https://arm-software.github.io/acle/neon_intrinsics/advsimd.html), so I
would rather focus on non-fp16 for now. Also, for FP16 variants we need
`-target-feature +fullfp16`, which means a different test file.
I am basically splitting this into small PRs so that I can keep track.
https://github.com/llvm/llvm-project/pull/216757
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
