[jdk17u-dev] RFR: 8291550: RISC-V: jdk uses misaligned memory access when AvoidUnalignedAccess enabled
Olga Mikhaltsova
omikhaltcova at openjdk.org
Fri Oct 6 14:44:30 UTC 2023
On Fri, 6 Oct 2023 13:52:14 GMT, Vladimir Kempik <vkempik at openjdk.org> wrote:
>> I'd like to backport JDK-8291550 to jdk17u.
>>
>> The patch applies not cleanly. It differs from the original one in the following:
>>
>> **1.** **src/hotspot/cpu/riscv/c2_MacroAssembler_riscv.cpp**
>>
>> `#undef VFCVT_SAFE` was removed because there is no `#define VFCVT_SAFE(VFLOATCVT)` (JDK-8306966)
>>
>> **2.** **src/hotspot/cpu/riscv/gc/shared/barrierSetNMethod_riscv.cpp**
>>
>> The difference is in context due to method signature change by JDK-8299229
>> `void NativeNMethodBarrier::verify() const {..`
>> `bool NativeNMethodBarrier::check_barrier(err_msg& msg) const {..`
>>
>> `msg.print(..)` was replaced with `tty->print_cr(..)`
>>
>> **3.** **src/hotspot/cpu/riscv/interp_masm_riscv.cpp**
>>
>> The changes were not applied to
>> `void InterpreterMacroAssembler::load_resolved_indy_entry(Register cache, Register index) {`
>> This method doesn't exist, it was added by JDK-8301995.
>>
>> **4.** **src/hotspot/cpu/riscv/macroAssembler_riscv.cpp**
>>
>> The difference is in context due to the method signature change by JDK-8298075
>> `address MacroAssembler::get_target_of_li32(address insn_addr) {`
>> `static address get_target_of_li32(address insn_addr) `{
>>
>> `NULL` was replaced with `nullptr` by JDK-8301496.
>>
>> **5.** **src/hotspot/cpu/riscv/nativeInst_riscv.cpp**
>> **src/hotspot/cpu/riscv/nativeInst_riscv.hpp**
>>
>> `NULL` was replaced with `nullptr` by JDK-8301496
>
> src/hotspot/cpu/riscv/macroAssembler_riscv.cpp line 4102:
>
>> 4100: // base: Address of a buffer to be zeroed, 8 bytes aligned.
>> 4101: // cnt: Immediate count in HeapWords.
>> 4102: void MacroAssembler::zero_words(Register base, u_int64_t cnt) {
>
> looks like this edit was accidentially added to the backport, other than that - looks good
Thx! That's right, this line break was inserted by mistake. Fixed.
-------------
PR Review Comment: https://git.openjdk.org/jdk17u-dev/pull/1852#discussion_r1348798966
More information about the jdk-updates-dev
mailing list