RFR: 8347343: RISC-V: Unchecked zicntr csr reads

[Fei Yang]

On Fri, 10 Jan 2025 08:41:34 GMT, Robbin Ehn <rehn at openjdk.org> wrote:

>> src/hotspot/cpu/riscv/macroAssembler_riscv.cpp line 1338:
>> 
>>> 1336: 
>>> 1337: void MacroAssembler::csrr(Register Rd, unsigned csr) {
>>> 1338:   assert((csr != CSR_INSTRET && csr != CSR_CYCLE && csr != CSR_TIME) || VM_Version::ext_Zicntr.enabled(), "sanity");
>> 
>> Personally, I perfer to check a JVM option here (Say `UseZicntr` of diagnostic type). When we have hwprobe for this extension, we can further hook up this option to `VM_Version::ext_Zicntr`. And the user could still enable on command line before that. What do you think?
>
> As these instructions are unused, no need to enable them, I'm hesitant to add an NO-OP option.
> I figured that when we add uses fom them we can add that option, if needed.
> 
> Should I really add such dummy option?

Or simply remove this `VM_Version::ext_Zicntr.enabled()` check from the assert for now? We can add it back together with option `UseZicntr` when we have hwprobe for this.

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/23003#discussion_r1910183455