Re: [Qemu-devel] [PATCH V4 00/11] Qemu MIPS ASE DSP Support

[Jia Liu]

On Fri, Mar 30, 2012 at 7:42 PM, Richard Henderson <address@hidden> wrote:
> On 03/29/2012 11:17 PM, Jia Liu wrote:
>> This is MIPS ASE DSP instructions support for QEMU. These instructions
>> are grouped according to "Chapter 4. MIPS DSP ASE Instruction Summary"
>> in MIPS ASE DSP manual [1].
>>
>> [1] MIPS32® Architecture for Programmers VolumeIV-e: The MIPS® DSP
>>     Application-Specific Extension to the MIPS32®Architecture
>>     
>> http://www.mips.com/products/product-materials/processor/mips-architecture/
>
> Any plans to support the mips64 version?
>
> It also occurred to me to actually try this code, which shows you haven't
> thought about mips64 at all.  Configure with --enable-debug:
>

I planed when mips32dsp is OK, I'll working on mips64dsp, so I'm not
think about mips64 too much yet.

>
>>   CC    mips64-softmmu/translate.o
>> /home/rth/work/qemu/qemu/target-mips/translate.c: In function 
>> 'gen_compute_branch':
>> /home/rth/work/qemu/qemu/target-mips/translate.c:2989:9: error: incompatible 
>> type for argument 1 of 'tcg_gen_andi_i32'
>> /home/rth/work/qemu/qemu/tcg/tcg-op.h:521:20: note: expected 'TCGv_i32' but 
>> argument is of type 'TCGv_i64'
>> /home/rth/work/qemu/qemu/target-mips/translate.c:2989:9: error: incompatible 
>> type for argument 2 of 'tcg_gen_andi_i32'
>> /home/rth/work/qemu/qemu/tcg/tcg-op.h:521:20: note: expected 'TCGv_i32' but 
>> argument is of type 'TCGv_i64'
>
>
> All of your helpers, whereever they receive or return a register value, need 
> to use
> type target_ulong and the DEF_HELPER lines need to use "tl".  All of the 
> mips32
> helpers you've written so far that return values need to make sure to 
> sign-extend
> their results from 32-bits.

I'll check the return values.

And, do you mean replace all "i32" "int" in DEF_HELPER with "tl", all
"uint32_t" "int" in dsp_hlper.c with "target_ulong"?

>
> With that, the compilation errors should go away, and you should produce 
> correct
> results for the subset of the mips64 dsp ase that overlaps with the mips32 
> dsp ase.
>

Thanks, I'll make sure the subset of the mips64dsp that overlaps with
the mips32dsp get correct results.

>
> r~

Regards,
Jia.