RFR: 8348186: C1: Purge fpu_stack_size infrastructure

Aleksey Shipilev shade at openjdk.org
Tue Jan 21 19:25:45 UTC 2025


On Tue, 21 Jan 2025 14:11:27 GMT, Aleksey Shipilev <shade at openjdk.org> wrote:

> There is currently a build failure for x86_32 in mainline due to [JDK-8346038](https://bugs.openjdk.org/browse/JDK-8346038) moving things around. The failure looks like this:
> 
> 
> * For target hotspot_variant-server_libjvm_objs_c1_LinearScan_x86.o:
> src/hotspot/cpu/x86/c1_LinearScan_x86.cpp: In member function 'void FpuStackAllocator::handle_op1(LIR_Op1*)':
> src/hotspot/cpu/x86/c1_LinearScan_x86.cpp:652:12: error: 'class LIR_Op1' has no member named 'set_fpu_stack_size'
>   652 | op1->set_fpu_stack_size(sim()->stack_size());
>       | ^~~~~~~~~~~~~~~~~~
> 
> 
> This does not warrant a fix, given the code is in deprecated x86_32 parts. But we do not need the whole `set_fpu_stack_size` infra to begin with: the last uses were removed by [JDK-8143353](https://bugs.openjdk.org/browse/JDK-8143353) and [JDK-8152907](https://bugs.openjdk.org/browse/JDK-8152907). So we can clean up this part for both x86_64 and x86_32 at once, and fix an easy build failure meanwhile.
> 
> Additional testing:
>  - [x] linux-x86-server-fastdebug build
>  - [x] linux-x86_64-server-fastdebug build

Thanks!

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

PR Comment: https://git.openjdk.org/jdk/pull/23215#issuecomment-2605557822


More information about the hotspot-compiler-dev mailing list