[Intel-gfx] [PATCH 1/6] drm/i915: Reorder gen3/4 swizzle detection logic
Tvrtko Ursulin
tvrtko.ursulin at linux.intel.com
Tue Mar 19 13:15:49 UTC 2019
On 18/03/2019 16:56, Ville Syrjala wrote:
> From: Ville Syrjälä <ville.syrjala at linux.intel.com>
>
> g33/i964g/g45 are the exceptional cases when it comes to
> the swizzle detectiong. Let's reorder the code to handle
> them first and let everything else be handled by the
> else branch. This allows us to unset .is_mobile for the
> desktop PNV variant (which supposedly must follow the
> "mobile" path here).
>
> Cc: Tvrtko Ursulin <tvrtko.ursulin at linux.intel.com>
> Signed-off-by: Ville Syrjälä <ville.syrjala at linux.intel.com>
> ---
> drivers/gpu/drm/i915/i915_gem_fence_reg.c | 65 +++++++++++------------
> 1 file changed, 32 insertions(+), 33 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_gem_fence_reg.c b/drivers/gpu/drm/i915/i915_gem_fence_reg.c
> index 65624b8e4d15..9418ad499b7e 100644
> --- a/drivers/gpu/drm/i915/i915_gem_fence_reg.c
> +++ b/drivers/gpu/drm/i915/i915_gem_fence_reg.c
> @@ -585,8 +585,38 @@ i915_gem_detect_bit_6_swizzle(struct drm_i915_private *dev_priv)
> */
> swizzle_x = I915_BIT_6_SWIZZLE_NONE;
> swizzle_y = I915_BIT_6_SWIZZLE_NONE;
> - } else if (IS_MOBILE(dev_priv) ||
> - IS_I915G(dev_priv) || IS_I945G(dev_priv)) {
> + } else if (IS_G45(dev_priv) || IS_I965G(dev_priv) || IS_G33(dev_priv)) {
> + /* The 965, G33, and newer, have a very flexible memory
> + * configuration. It will enable dual-channel mode
> + * (interleaving) on as much memory as it can, and the GPU
> + * will additionally sometimes enable different bit 6
> + * swizzling for tiled objects from the CPU.
> + *
> + * Here's what I found on the G965:
> + * slot fill memory size swizzling
> + * 0A 0B 1A 1B 1-ch 2-ch
> + * 512 0 0 0 512 0 O
> + * 512 0 512 0 16 1008 X
> + * 512 0 0 512 16 1008 X
> + * 0 512 0 512 16 1008 X
> + * 1024 1024 1024 0 2048 1024 O
> + *
> + * We could probably detect this based on either the DRB
> + * matching, which was the case for the swizzling required in
> + * the table above, or from the 1-ch value being less than
> + * the minimum size of a rank.
> + *
> + * Reports indicate that the swizzling actually
> + * varies depending upon page placement inside the
> + * channels, i.e. we see swizzled pages where the
> + * banks of memory are paired and unswizzled on the
> + * uneven portion, so leave that as unknown.
> + */
> + if (I915_READ16(C0DRB3) == I915_READ16(C1DRB3)) {
> + swizzle_x = I915_BIT_6_SWIZZLE_9_10;
> + swizzle_y = I915_BIT_6_SWIZZLE_9;
> + }
> + } else {
> u32 dcc;
>
> /* On 9xx chipsets, channel interleave by the CPU is
> @@ -636,37 +666,6 @@ i915_gem_detect_bit_6_swizzle(struct drm_i915_private *dev_priv)
> swizzle_x = I915_BIT_6_SWIZZLE_UNKNOWN;
> swizzle_y = I915_BIT_6_SWIZZLE_UNKNOWN;
> }
> - } else {
> - /* The 965, G33, and newer, have a very flexible memory
> - * configuration. It will enable dual-channel mode
> - * (interleaving) on as much memory as it can, and the GPU
> - * will additionally sometimes enable different bit 6
> - * swizzling for tiled objects from the CPU.
> - *
> - * Here's what I found on the G965:
> - * slot fill memory size swizzling
> - * 0A 0B 1A 1B 1-ch 2-ch
> - * 512 0 0 0 512 0 O
> - * 512 0 512 0 16 1008 X
> - * 512 0 0 512 16 1008 X
> - * 0 512 0 512 16 1008 X
> - * 1024 1024 1024 0 2048 1024 O
> - *
> - * We could probably detect this based on either the DRB
> - * matching, which was the case for the swizzling required in
> - * the table above, or from the 1-ch value being less than
> - * the minimum size of a rank.
> - *
> - * Reports indicate that the swizzling actually
> - * varies depending upon page placement inside the
> - * channels, i.e. we see swizzled pages where the
> - * banks of memory are paired and unswizzled on the
> - * uneven portion, so leave that as unknown.
> - */
> - if (I915_READ16(C0DRB3) == I915_READ16(C1DRB3)) {
> - swizzle_x = I915_BIT_6_SWIZZLE_9_10;
> - swizzle_y = I915_BIT_6_SWIZZLE_9;
> - }
> }
>
> if (swizzle_x == I915_BIT_6_SWIZZLE_UNKNOWN ||
>
Old/new/future behaviour checks out as far as I can see.
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
Regards,
Tvrtko
More information about the Intel-gfx
mailing list