amdgpu header files (was: Re: [git pull] drm for 6.7-rc1)

[Geert Uytterhoeven <geert@linux-m68k.org>]

On Tue, 31 Oct 2023, Dave Airlie wrote:
> This is the main drm pull request for 6.7.

> Highlights:
> - AMD adds some more upcoming HW platforms

> Alex Deucher (24):
>      drm/amdgpu: update to the latest GC 11.5 headers

> Candice Li (8):
>      drm/amd: Add umc v12_0_0 ip headers

> Lang Yu (57):
>      drm/amdgpu: add gc headers for gc 11.5.0
>      drm/amdgpu: add mmhub 3.3.0 headers
>      drm/amdgpu: add VPE 6.1.0 header files
>      drm/amdgpu: add UMSCH 4.0 register headers

> Li Ma (11):
>      drm/amdgpu: add header files for MP 14.0.0
>      drm/amdgpu: fix missing stuff in NBIO v7.11

> Qingqing Zhuo (38):
>      drm/amd/display: Add dcn35 register header files

> Saleemkhan Jamadar (9):
>      drm/amdgpu: add vcn 4_0_5 header files

> Yang Wang (16):
>      drm/amd/pm: add smu_13_0_6 mca dump support

> benl (3):
>      drm/amdgpu: add nbio 7.11 registers

> .../amd/include/asic_reg/dcn/dcn_3_5_0_offset.h    | 15255 +++++
> .../amd/include/asic_reg/dcn/dcn_3_5_0_sh_mask.h   | 53412 +++++++++++++++++
> .../drm/amd/include/asic_reg/gc/gc_11_5_0_offset.h | 10000 ++++
> .../amd/include/asic_reg/gc/gc_11_5_0_sh_mask.h    | 36579 ++++++++++++
> .../include/asic_reg/mmhub/mmhub_3_3_0_offset.h    |  1395 +
> .../include/asic_reg/mmhub/mmhub_3_3_0_sh_mask.h   |  6722 +++
> .../amd/include/asic_reg/mp/mp_13_0_6_sh_mask.h    |    28 +
> .../drm/amd/include/asic_reg/mp/mp_14_0_0_offset.h |   359 +
> .../amd/include/asic_reg/mp/mp_14_0_0_sh_mask.h    |   534 +
> .../amd/include/asic_reg/nbio/nbio_7_11_0_offset.h |  9400 +++
> .../include/asic_reg/nbio/nbio_7_11_0_sh_mask.h    | 57857 +++++++++++++++++++
> .../amd/include/asic_reg/umc/umc_12_0_0_offset.h   |    33 +
> .../amd/include/asic_reg/umc/umc_12_0_0_sh_mask.h  |    95 +
> .../amd/include/asic_reg/vcn/vcn_4_0_0_offset.h    |   422 +
> .../amd/include/asic_reg/vcn/vcn_4_0_0_sh_mask.h   |   882 +
> .../amd/include/asic_reg/vcn/vcn_4_0_5_offset.h    |  1797 +
> .../amd/include/asic_reg/vcn/vcn_4_0_5_sh_mask.h   |  8614 +++
> .../amd/include/asic_reg/vpe/vpe_6_1_0_offset.h    |  1553 +
> .../amd/include/asic_reg/vpe/vpe_6_1_0_sh_mask.h   |  4393 ++

These huge files can be reduced by 50%: all the *_SHIFT definitions are
redundant, as they can be derived from the corresponding *_MASK
definitions at compile-time, cfr. <linux/bitfield.h>.

E.g.:

#define AZCONTROLLER0_CORB_READ_POINTER__CORB_READ_POINTER__SHIFT        0x0
#define AZCONTROLLER0_CORB_READ_POINTER__CORB_READ_POINTER_RESET__SHIFT  0xf
#define AZCONTROLLER0_CORB_READ_POINTER__CORB_READ_POINTER_MASK          0x00FFL
#define AZCONTROLLER0_CORB_READ_POINTER__CORB_READ_POINTER_RESET_MASK    0x8000L

AZCONTROLLER0_CORB_READ_POINTER__CORB_READ_POINTER__SHIFT =
 	__bf_shf(AZCONTROLLER0_CORB_READ_POINTER__CORB_READ_POINTER_MASK)
AZCONTROLLER0_CORB_READ_POINTER__CORB_READ_POINTER_RESET__SHIFT =
 	__bf_shf(AZCONTROLLER0_CORB_READ_POINTER__CORB_READ_POINTER_RESET_MASK)

set_reg_field_value_masks() takes a shift and a mask, while it
could calculate the shift at run-time.
set_reg_field_values() takes pairs of shifts and masks, but the shifts
are not needed; lots of tables can be halved, etc...

Gr{oetje,eeting}s,

 						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
 							    -- Linus Torvalds