This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Fix middle-end/17767


Richard Henderson wrote:

No, a *proper* solution is to perform the comparison element-by-element
as the real vector compare instruction does. Producing only 0 or only 0xffff_ffff_ffff_ffff is no good.




Please note that this code will be used only when (folded_arg0 == folded_arg1). In this special case the asumption that the resulting vector has all elements equal is valid.

[All other cases are protected by some kond of "if (const_arg1 == const0_rtx)" that is false for CONST_VECTOR].

I think you should just always skip vector compares for 4.0.


They are skipped, until someone defines VECTOR_STORE_FLAG_VALUE. Currently this macro is undefined for all architectures.

Uros.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]