[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH v5 29/35] target/arm: Implement SVE fp complex m
From: |
Peter Maydell |
Subject: |
Re: [Qemu-devel] [PATCH v5 29/35] target/arm: Implement SVE fp complex multiply add |
Date: |
Tue, 26 Jun 2018 16:17:42 +0100 |
On 26 June 2018 at 16:04, Richard Henderson
<address@hidden> wrote:
> On 06/26/2018 06:29 AM, Peter Maydell wrote:
>>> + e2 = (flip ? ni : nr);
>>> + e1 = (flip ? mi : mr) ^ neg_real;
>>> + e4 = e2;
>>> + e3 = (flip ? mr : mi) ^ neg_imag;
>>
>> These don't seem to match up with the pseudocode, which
>> applies the neg_real or neg_imag negations to element2,
>> not element1/3. I think the operations are correct but the
>> variable names are confusingly swapped.
>
> Yes, I know. They have different names between the SVE document and the main
> ARM ARM. You swapped them to match the ARM when you applied the ARMv8.3-FCMA
> patch set, and I copied code from vec_helper.c to write this.
>
> Which is least confusing?
Oh, right. Let's use the same numbering in vec_helper.c as this
code, so we're at least internally consistent.
thanks
-- PMM
[Qemu-devel] [PATCH v5 34/35] target/arm: Enable SVE for aarch64-linux-user, Richard Henderson, 2018/06/20
[Qemu-devel] [PATCH v5 32/35] target/arm: Implement SVE dot product (vectors), Richard Henderson, 2018/06/20