On 13/01/2022 14:56, Christophe Lyon via Gcc-patches wrote:
The vmvnq_n* intrinsics and have [u]int[16|32]_t arguments, so use <V_elem> iterator instead of HI in mve_vmvnq_n_<supf><mode>. 2022-01-13 Christophe Lyon <christophe.l...@foss.st.com> gcc/ * config/arm/mve.md (mve_vmvnq_n_<supf><mode>): Use V_elem mode for operand 1. diff --git a/gcc/config/arm/mve.md b/gcc/config/arm/mve.md index 171dd384133..5c3b34dce3a 100644 --- a/gcc/config/arm/mve.md +++ b/gcc/config/arm/mve.md @@ -617,7 +617,7 @@ (define_insn "mve_vcvtaq_<supf><mode>" (define_insn "mve_vmvnq_n_<supf><mode>" [ (set (match_operand:MVE_5 0 "s_register_operand" "=w") - (unspec:MVE_5 [(match_operand:HI 1 "immediate_operand" "i")] + (unspec:MVE_5 [(match_operand:<V_elem> 1 "immediate_operand" "i")] VMVNQ_N)) ] "TARGET_HAVE_MVE"
While fixing this it might be good to fix the constraint and predicate inspired by "DL" and "neon_inv_logic_op2" respectively. This would avoid the compiler generating wrong assembly, and instead it would probably lead to the compiler using a load literal.
I kind of think it would be better to have the intrinsic refuse the immediate altogether, but it seems for NEON we also use the load literal approach.