[riscv-port-jdk17u:riscv-port] RFR: 8295270: RISC-V: Clean up and refactoring for assembler functions
Xiaolin Zheng
xlinzheng at openjdk.org
Mon Apr 10 03:43:11 UTC 2023
On Sat, 8 Apr 2023 03:55:31 GMT, Gui Cao <gcao at openjdk.org> wrote:
> Please review this backport to riscv-port-jdk17u.
> Backport of [JDK-8295270](https://bugs.openjdk.org/browse/JDK-8295270).
> [JDK-8295270](https://bugs.openjdk.org/browse/JDK-8295270) not clean,origin patch has no eden_allocate, which removed with [JDK-8290706](https://bugs.openjdk.org/browse/JDK-8290706)
>
> Testing:
>
> Tier1-3 passed without new failure on unmacthed (release).
src/hotspot/cpu/riscv/assembler_riscv.hpp line 496:
> 494: "only internal_word_type relocs make sense here"); \
> 495: relocate(InternalAddress(dest).rspec()); \
> 496: IncompressibleRegion ir(this); /* relocations */
Here we have an `IncompressibleRegion` thing to keep instructions in `INSN_ENTRY_RELOC` from being compressed by the C extension, for relocations are here. But after moving to different locations in this patch, the `IncompressibleRegion` is missing. Please see the below comments.
src/hotspot/cpu/riscv/macroAssembler_riscv.hpp line 491:
> 489: guarantee(rtype == relocInfo::internal_word_type, \
> 490: "only internal_word_type relocs make sense here"); \
> 491: relocate(InternalAddress(dest).rspec());
Suggestion:
relocate(InternalAddress(dest).rspec()); \
IncompressibleRegion ir(this); /* relocations */
src/hotspot/cpu/riscv/macroAssembler_riscv.hpp line 653:
> 651: guarantee(rtype == relocInfo::internal_word_type, \
> 652: "only internal_word_type relocs make sense here"); \
> 653: relocate(InternalAddress(dest).rspec());
Suggestion:
relocate(InternalAddress(dest).rspec()); \
IncompressibleRegion ir(this); /* relocations */
-------------
PR Review Comment: https://git.openjdk.org/riscv-port-jdk17u/pull/38#discussion_r1161407899
PR Review Comment: https://git.openjdk.org/riscv-port-jdk17u/pull/38#discussion_r1161408118
PR Review Comment: https://git.openjdk.org/riscv-port-jdk17u/pull/38#discussion_r1161408202
More information about the riscv-port-dev
mailing list