[Mesa-dev] [PATCH 05/11] intel/compiler: Use null destination register for memory fence messages

[Matt Turner]

On Wed, Mar 21, 2018 at 3:08 PM, Francisco Jerez <currojerez at riseup.net> wrote:
> Matt Turner <mattst88 at gmail.com> writes:
>
>> On Wed, Mar 21, 2018 at 2:56 PM, Francisco Jerez <currojerez at riseup.net> wrote:
>>> Matt Turner <mattst88 at gmail.com> writes:
>>>
>>>> From Message Descriptor section in gfxspecs:
>>>>
>>>>   "Memory fence messages without Commit Enable set do not return
>>>>    anything to the thread (response length is 0 and destination
>>>>    register is null)."
>>>>
>>>> This fixes a GPU hang in simulation in the piglit test
>>>> arb_shader_image_load_store-shader-mem-barrier
>>>>
>>>
>>> On what platform?
>>
>> I'm pretty sure Anuj found this in ICL. I'll revert this patch from my
>> branch and try to confirm.
>
> That sounds pretty bogus, this patch cannot possibly have any effect on
> ICL because brw_memory_fence() is already setting commit_enable
> unconditionally on Gen10+, so the destination register won't ever be
> null regardless.

I just ran the associated piglit test through simulation without this
patch, and everything passed. I'll drop the patch.