Zoltán Gilián writes:
>> > Hmmm... So you only need it as padding? Wouldn't it be possible for
>> > you to declare samplers to be 0 bytes?
>> Maybe it can be done by changing the type of the sampler arg from i32
>> to an empty struct. I'll have to try this, I don't know if it will
>> work.
>
>
But samplers aren't necessary for the image attributes, so I'll break
this patch into two.
On Mon, Jul 27, 2015 at 10:25 AM, Zoltán Gilián wrote:
>> > Hmmm... So you only need it as padding? Wouldn't it be possible for
>> > you to declare samplers to be 0 bytes?
>> Maybe it can be done by chang
> > Hmmm... So you only need it as padding? Wouldn't it be possible for
> > you to declare samplers to be 0 bytes?
> Maybe it can be done by changing the type of the sampler arg from i32
> to an empty struct. I'll have to try this, I don't know if it will
> work.
But then it would only work for
> Why? Your module::argument::image_size and ::image_format cases don't
> touch the explicit_arg iterator at all AFAICT, so it will be left
> pointing one past the last general semantic argument
Ok, my mistake, I didn't think this through.
> Hmmm... So you only need it as padding? Wouldn't it
Zoltán Gilián writes:
>> auto img = dynamic_cast(**(explicit_arg - 1))
>
> Ok, so it should be (explicit_arg - 2) for the image format, I
> presume.
Why? Your module::argument::image_size and ::image_format cases don't
touch the explicit_arg iterator at all AFAICT, so it will be left
pointing o
> auto img = dynamic_cast(**(explicit_arg - 1))
Ok, so it should be (explicit_arg - 2) for the image format, I
presume. This will be incorrect, however, if the targets that need
implicit argument for format metadata are indeed a strict superset of
the ones that need dimension, as you mentioned bef
Zoltan Gilian writes:
> Read-only and write-only image arguments are recognized and
> distinguished.
> Attributes of the image arguments are passed to the kernel as implicit
> arguments.
> ---
> src/gallium/state_trackers/clover/core/kernel.cpp | 46 ++-
> src/gallium/state_trackers/clover
Read-only and write-only image arguments are recognized and
distinguished.
Attributes of the image arguments are passed to the kernel as implicit
arguments.
---
src/gallium/state_trackers/clover/core/kernel.cpp | 46 ++-
src/gallium/state_trackers/clover/core/kernel.hpp | 13 +-
src/galliu
Zoltán Gilián writes:
>> I have the suspicion
>> that it would simplify both the OpenCL front-end and compiler back-end
>> code if the image metadata was interleaved with images themselves.
>
> In fact this complicates the back-end, since the location of each
> argument following an image argume
> I have the suspicion
> that it would simplify both the OpenCL front-end and compiler back-end
> code if the image metadata was interleaved with images themselves.
In fact this complicates the back-end, since the location of each
argument following an image argument changes because of the metada
Zoltan Gilian writes:
> Read-only and write-only image arguments are recognized and
> distinguished.
> Attributes of the image arguments are passed to the kernel as implicit
> arguments.
Thanks, this looks much better. One thing that still seems kind of
unfortunate is the fact that you've added
Read-only and write-only image arguments are recognized and
distinguished.
Attributes of the image arguments are passed to the kernel as implicit
arguments.
---
src/gallium/state_trackers/clover/core/kernel.cpp | 27 ++
src/gallium/state_trackers/clover/core/kernel.hpp | 13 ++-
src/galliu
12 matches
Mail list logo
