This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [Patch AArch64] Switch constant pools to separate rodata sections.
- From: Ramana Radhakrishnan <ramana dot gcc at googlemail dot com>
- To: Alan Lawrence <alan dot lawrence at arm dot com>
- Cc: Ramana Radhakrishnan <ramana dot radhakrishnan at foss dot arm dot com>, James Greenhalgh <james dot greenhalgh at arm dot com>, "gcc-patches at gcc dot gnu dot org" <gcc-patches at gcc dot gnu dot org>
- Date: Tue, 10 Nov 2015 16:46:56 +0000
- Subject: Re: [Patch AArch64] Switch constant pools to separate rodata sections.
- Authentication-results: sourceware.org; auth=none
- References: <5638E29D dot 6030704 at foss dot arm dot com> <20151103170952 dot GA5141 at arm dot com> <563A1597 dot 9060604 at foss dot arm dot com> <56421DCE dot 8090906 at arm dot com>
On Tue, Nov 10, 2015 at 4:39 PM, Alan Lawrence <alan.lawrence@arm.com> wrote:
> On 04/11/15 14:26, Ramana Radhakrishnan wrote:
>>
>>
>> True and I've just been reading more of the backend - We could now start
>> using blocks for constant pools as well. So let's do that.
>>
>> How does something like this look ?
>>
>> Tested on aarch64-none-elf - no regressions.
>>
>> 2015-11-04 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
>>
>> * config/aarch64/aarch64.c
>> (aarch64_can_use_per_function_literal_pools_p): New.
>> (aarch64_use_blocks_for_constant_p): Adjust declaration
>> and use aarch64_can_use_function_literal_pools_p.
>> (aarch64_select_rtx_section): Update.
>>
>
> Since r229878, I've been seeing
>
> FAIL: gcc.dg/attr-weakref-1.c (test for excess errors)
> UNRESOLVED: gcc.dg/attr-weakref-1.c compilation failed to produce executable
>
> (both previously passing) on aarch64-none-elf, aarch64_be-none-elf, and
> aarch64-none-linux-gnu. Here's a log from aarch64_be-none-elf (the others
> look similar):
>
> /work/alalaw01/build-aarch64_be-none-elf/obj/gcc2/gcc/xgcc
> -B/work/alalaw01/build-aarch64_be-none-elf/obj/gcc2/gcc/
> /work/alalaw01/src/gcc/gcc/testsuite/gcc.dg/attr-weakref-1.c
> -fno-diagnostics-show-caret -fdiagnostics-color=never -O2
> /work/alalaw01/src/gcc/gcc/testsuite/gcc.dg/attr-weakref-1a.c
> -specs=aem-validation.specs -lm -o ./attr-weakref-1.exe
> /tmp/ccEfngi6.o:(.rodata.cst8+0x30): undefined reference to `wv12'
> /tmp/ccEfngi6.o:(.rodata.cst8+0x38): undefined reference to `wv12'
> /tmp/ccEfngi6.o:(.rodata.cst8+0x60): undefined reference to `wf12'
> /tmp/ccEfngi6.o:(.rodata.cst8+0x68): undefined reference to `wf12'
> collect2: error: ld returned 1 exit status
> compiler exited with status 1
> output is:
> /tmp/ccEfngi6.o:(.rodata.cst8+0x30): undefined reference to `wv12'
> /tmp/ccEfngi6.o:(.rodata.cst8+0x38): undefined reference to `wv12'
> /tmp/ccEfngi6.o:(.rodata.cst8+0x60): undefined reference to `wf12'
> /tmp/ccEfngi6.o:(.rodata.cst8+0x68): undefined reference to `wf12'
> collect2: error: ld returned 1 exit status
>
> FAIL: gcc.dg/attr-weakref-1.c (test for excess errors)
>
Hmmm I'm surprised it failed in the first place as my testing didn't
show it - I need to check on that.
Nevertheless this fail has gone away in my testing with
https://gcc.gnu.org/ml/gcc-cvs/2015-11/msg00453.html in a bootstrap
and regression run on aarch64-none-linux-gnu. I see nothing triplet
specific in the testcase here for it to fail differently.
Is this something you see really with tip of trunk ?
regards
Ramana