This is the mail archive of the
mailing list for the binutils project.
Re: [PATCH] [MIPS] add MIPS64DSPR2 support.
- From: Richard Sandiford <rdsandiford at googlemail dot com>
- To: Jia Liu <proljc at gmail dot com>
- Cc: binutils at sourceware dot org
- Date: Sun, 01 Apr 2012 20:38:47 +0100
- Subject: Re: [PATCH] [MIPS] add MIPS64DSPR2 support.
- References: <CAJBMM-tV__HpCkAhjejhSbweRqfBUVSTsvoauefeUuy=CXtE9w@mail.gmail.com> <email@example.com> <CAJBMM-vbhyMv3j8Hi27jmjrW2FK6nm4gD1Mo_Ya3yk-mGtn91A@mail.gmail.com>
Jia Liu <firstname.lastname@example.org> writes:
> On Tue, Feb 7, 2012 at 3:57 AM, Richard Sandiford
> <email@example.com> wrote:
>> Thanks for the patch.
>> Liu <firstname.lastname@example.org> writes:
>>> I've added MIPS64DSPR2 support to binutils.
>>> Please review.
>> Following on from Chao-ying's question on the GCC list: what target
>> are you using? ÂIs it an emulated implementation of MIPS64 DSPr2
>> (e.g. QEMU)? ÂOr do you have a real hardware implementation?
>> Just curious :-)
>> Chao-ying, just to check: it sounded from your message on the GCC list
>> that some instructions had been removed from the ASE. ÂIs that right?
>> If so, is the latest document on the MIPS website up-to-date,
>> or have there been changes since then?
>> Liu: In the meantime, a patch of this size will need a copyright
>> assignment. ÂDo you already have one on file? ÂIf not, please send
>> a note to me privately and I can send you the form.
> Hi Richard
> Thank you.
> I've got and signed the FSF Assignment, and posted them to FSF.
> Please checkin my patch.
Thanks for your patience.
I'm really sorry about the confusion here, which is entirely my fault.
My understanding from Chao-Ying's message was that MIPS had effectively
withdrawn the current MIPS64 DSP r1/r2 spec. They are starting the
MIPS64 DSP spec from scratch for r3. This means that no processor has,
and I assume never will, support those extra MIPS64 DSP r1 and r2
instructions. It seems MIPS64 DSP r1/r2 is effectively dead.
So I think Chao-Ying's suggestion was that we should treat MIPS64 DSP
r1/r2 the same as MIPS32 DSP r1/r2. We should pretend that the separate
MIPS64 DSP r1 and r2 specs never existed:
> MIPS has an internal MIPS64 DSP spec verison 3.00, but hasn't released
> yet. The version 3.00 spec is quite different from the previous
> version. The best for now is to assume all MIPS32 DSP r1/DSP r2
> instructions appear in the MIPS64 DSP r1/DSP r2 spec, and to not
> support any new instructions in GCC/Binutils.
So my hope was that once version 3 is released, we'll see which
instructions are in MIPS64 DSP r3 and not in MIPS32 DSP r3.
Probably some of them will overlap r1 and r2 (but I'm just speculating
there). So I was hoping we could add DSP r3 support at that stage,
using your patch as a starting point.
Sorry again about this :-(