Nouveau - Mar 2016 - [PATCH 4/4] pmu/fuc: movw is somewhat weird on gk208, use mov instead

On 26/02/16 17:19, Karol Herbst wrote:
> currently there is no change, because nobody uses those macros yet, but
they
> shouldn't stay broken
>
> Signed-off-by: Karol Herbst <nouveau at karolherbst.de>
> ---
>   drm/nouveau/nvkm/subdev/pmu/fuc/macros.fuc | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drm/nouveau/nvkm/subdev/pmu/fuc/macros.fuc
b/drm/nouveau/nvkm/subdev/pmu/fuc/macros.fuc
> index 0d5cbeb..bb59eb4 100644
> --- a/drm/nouveau/nvkm/subdev/pmu/fuc/macros.fuc
> +++ b/drm/nouveau/nvkm/subdev/pmu/fuc/macros.fuc
> @@ -252,12 +252,12 @@
>   #endif
>   
>   #define st(size, addr, reg) /*
> -*/	movw $r0 addr /*
> +*/	mov $r0 addr /*
First of all, I know it is annoying, but we *need* to understand exactly 
what movw is now doing.

Secondly, I seem to remember that a 32 bit mov was not added until fuc3 
or something. Have you tried assembling this code on older fuc versions? 
Pretty sure it will fail.
>   */	st size D[$r0] reg /*
>   */	clear b32 $r0
>   
>   #define ld(size, reg, addr) /*
> -*/	movw $r0 addr /*
> +*/	mov $r0 addr /*
>   */	ld size reg D[$r0] /*
>   */	clear b32 $r0
>

On Tue, Mar 1, 2016 at 4:36 PM, Martin Peres <martin.peres at free.fr>
wrote:
> On 26/02/16 17:19, Karol Herbst wrote:
>>
>> currently there is no change, because nobody uses those macros yet, but
>> they
>> shouldn't stay broken
>>
>> Signed-off-by: Karol Herbst <nouveau at karolherbst.de>
>> ---
>>   drm/nouveau/nvkm/subdev/pmu/fuc/macros.fuc | 4 ++--
>>   1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/drm/nouveau/nvkm/subdev/pmu/fuc/macros.fuc
>> b/drm/nouveau/nvkm/subdev/pmu/fuc/macros.fuc
>> index 0d5cbeb..bb59eb4 100644
>> --- a/drm/nouveau/nvkm/subdev/pmu/fuc/macros.fuc
>> +++ b/drm/nouveau/nvkm/subdev/pmu/fuc/macros.fuc
>> @@ -252,12 +252,12 @@
>>   #endif
>>     #define st(size, addr, reg) /*
>> -*/     movw $r0 addr /*
>> +*/     mov $r0 addr /*
>
>
> First of all, I know it is annoying, but we *need* to understand exactly
> what movw is now doing.
>
> Secondly, I seem to remember that a 32 bit mov was not added until fuc3 or
fuc5
> something. Have you tried assembling this code on older fuc versions?
Pretty
> sure it will fail.
That's what the imm32() macro is for.
>
>>   */    st size D[$r0] reg /*
>>   */    clear b32 $r0
>>     #define ld(size, reg, addr) /*
>> -*/     movw $r0 addr /*
>> +*/     mov $r0 addr /*
>>   */    ld size reg D[$r0] /*
>>   */    clear b32 $r0
>>
>
>
> _______________________________________________
> Nouveau mailing list
> Nouveau at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/nouveau

On 01/03/16 23:38, Ilia Mirkin wrote:
> On Tue, Mar 1, 2016 at 4:36 PM, Martin Peres <martin.peres at
free.fr> wrote:
>> On 26/02/16 17:19, Karol Herbst wrote:
>>> currently there is no change, because nobody uses those macros yet,
but
>>> they
>>> shouldn't stay broken
>>>
>>> Signed-off-by: Karol Herbst <nouveau at karolherbst.de>
>>> ---
>>>    drm/nouveau/nvkm/subdev/pmu/fuc/macros.fuc | 4 ++--
>>>    1 file changed, 2 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drm/nouveau/nvkm/subdev/pmu/fuc/macros.fuc
>>> b/drm/nouveau/nvkm/subdev/pmu/fuc/macros.fuc
>>> index 0d5cbeb..bb59eb4 100644
>>> --- a/drm/nouveau/nvkm/subdev/pmu/fuc/macros.fuc
>>> +++ b/drm/nouveau/nvkm/subdev/pmu/fuc/macros.fuc
>>> @@ -252,12 +252,12 @@
>>>    #endif
>>>      #define st(size, addr, reg) /*
>>> -*/     movw $r0 addr /*
>>> +*/     mov $r0 addr /*
>>
>> First of all, I know it is annoying, but we *need* to understand
exactly
>> what movw is now doing.
>>
>> Secondly, I seem to remember that a 32 bit mov was not added until fuc3
or
> fuc5
Ah, right, fuc3 was on the GT2xx. Thanks!
>
>> something. Have you tried assembling this code on older fuc versions?
Pretty
>> sure it will fail.
> That's what the imm32() macro is for.
Right, with the mov replace with the imm32 macro, this patch will have 
my R-b!