[PATCH v5 0/4] lib/gpgpu_shader: simplify load/store shaders and add Xe3 support

[Hajda, Andrzej]

W dniu 25.11.2024 o 15:32, Andrzej Hajda pisze:
> The patchset tries to simplify and unify iga64 load/store assemblies.
> Moreover it adds support for Xe3 platforms (PTL).
>
> v2:
>    - addressed comments to v1, details in patches
> v3:
>    - added passing surface descriptor via COMPUTE_WALKER/inline_data,
>      required for stateless load/stores in Xe3.
>    - added support for Xe3 platforms.
> v4:
>    - fixed placement of gpgpu_alloc_gpu_addr changes(Dominik),
>    - pass vm_id in gpgpu_alloc_gpu_addr (Dominik),
>    - added R-B
> v5:
>    - updated comment in the last patch (Dominik),
>    - removed extra lines (Dominik)
>    - added R-B,
>    - rebased
>
> Signed-off-by: Andrzej Hajda <andrzej.hajda at intel.com>
> ---
> - Link to v1: https://lore.kernel.org/r/20241114-gpgpu_send_rework-v1-0-e0914e09e7b2@intel.com
> - Link to v2: https://lore.kernel.org/r/20241115-gpgpu_send_rework-v2-0-d1c84135aaaf@intel.com
> - Link to v3: https://lore.kernel.org/r/20241121-gpgpu_send_rework-v3-0-0b6aa48ab006@intel.com
> - Link to v4: https://lore.kernel.org/r/20241125-gpgpu_send_rework-v4-0-c16b568a1f3d@intel.com


Merged.

Thx for reviews.

Andrzej


>
> ---
> Andrzej Hajda (4):
>        scripts/generate_iga64_codes: add iga64_macros.h to checksum calculation
>        lib/gpgpu_shader: simplify load/store shaders
>        lib/gpgpu_shader: pass surface desription to shaders via inline data
>        lib/gpgpu_shader: add support for Xe3 platforms
>
>   lib/gpgpu_shader.c           | 188 +++++++------------------
>   lib/iga64_generated_codes.c  | 329 +++++++++++++++++++++----------------------
>   lib/iga64_macros.h           |  72 +++++++++-
>   lib/meson.build              |   1 +
>   scripts/generate_iga64_codes |   2 +-
>   5 files changed, 280 insertions(+), 312 deletions(-)
> ---
> base-commit: 027caf37991f4d39cb41beaf82dc7c920d976db6
> change-id: 20241106-gpgpu_send_rework-817a3e4f464e
>
> Best regards,