[Intel-gfx] [PATCH v3 4/4] drm/i915/icl: Add Multi-segmented gamma support
Ville Syrjälä
ville.syrjala at linux.intel.com
Tue May 7 14:27:50 UTC 2019
On Tue, May 07, 2019 at 07:26:44PM +0530, Shashank Sharma wrote:
> ICL introduces a new gamma correction mode in display engine, called
> multi-segmented-gamma mode. This mode allows users to program the
> darker region of the gamma curve with sueprfine precision. An
> example use case for this is HDR curves (like PQ ST-2084).
>
> If we plot a gamma correction curve from value range between 0.0 to 1.0,
> ICL's multi-segment has 3 different sections:
> - superfine segment: 9 values, ranges between 0 - 1/(128 * 256)
> - fine segment: 257 values, ranges between 0 - 1/(128)
> - corase segment: 257 values, ranges between 0 - 1
>
> This patch:
> - Changes gamma LUTs size for ICL/GEN11 to 262144 entries (8 * 128 * 256),
> so that userspace can program with highest precision supported.
> - Changes default gamma mode (non-legacy) to multi-segmented-gamma mode.
> - Adds functions to program/detect multi-segment gamma.
>
> V2: Addressed review comments from Ville
> - separate function for superfine and fine segments.
> - remove enum for segments.
> - reuse last entry of the LUT as gc_max value.
> - replace if() ....cond with switch...case in icl_load_luts.
> - add an entry variable, instead of 'word'
>
> V3: Addressed review comments from Ville
> - extra newline
> - s/entry/color/
> - remove LUT size checks
> - program ilk_lut_12p4_ldw value before ilk_lut_12p4_udw
> - Change the comments in description of fine and coarse segments,
> and try to make more sense.
> - use 8 * 128 instead of 1024
> - add 1 entry in LUT for GCMAX
>
> Cc: Ville Syrjälä <ville.syrjala at linux.intel.com>
> Cc: Maarten Lankhorst <maarten.lankhorst at linux.intel.com>
> Cc: Daniel Vetter <daniel.vetter at ffwll.ch>
>
> Suggested-by: Ville Syrjälä <ville.syrjala at linux.intel.com>
> Signed-off-by: Shashank Sharma <shashank.sharma at intel.com>
> Signed-off-by: Uma Shankar <uma.shankar at intel.com>
> ---
> drivers/gpu/drm/i915/i915_pci.c | 2 +-
> drivers/gpu/drm/i915/intel_color.c | 127 ++++++++++++++++++++++++++++-
> 2 files changed, 124 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_pci.c b/drivers/gpu/drm/i915/i915_pci.c
> index ffa2ee70a03d..2f99b585d44b 100644
> --- a/drivers/gpu/drm/i915/i915_pci.c
> +++ b/drivers/gpu/drm/i915/i915_pci.c
> @@ -749,7 +749,7 @@ static const struct intel_device_info intel_cannonlake_info = {
> GEN(11), \
> .ddb_size = 2048, \
> .has_logical_ring_elsq = 1, \
> - .color = { .degamma_lut_size = 33, .gamma_lut_size = 1024 }
> + .color = { .degamma_lut_size = 33, .gamma_lut_size = 262145 }
>
> static const struct intel_device_info intel_icelake_11_info = {
> GEN11_FEATURES,
> diff --git a/drivers/gpu/drm/i915/intel_color.c b/drivers/gpu/drm/i915/intel_color.c
> index 6c341bea514c..c1a9506fd069 100644
> --- a/drivers/gpu/drm/i915/intel_color.c
> +++ b/drivers/gpu/drm/i915/intel_color.c
> @@ -41,6 +41,8 @@
> #define CTM_COEFF_ABS(coeff) ((coeff) & (CTM_COEFF_SIGN - 1))
>
> #define LEGACY_LUT_LENGTH 256
> +#define ICL_GAMMA_MULTISEG_LUT_LENGTH (256 * 128 * 8)
Unused.
> +
> /*
> * Extract the CSC coefficient from a CTM coefficient (in U32.32 fixed point
> * format). This macro takes the coefficient we want transformed and the
> @@ -767,6 +769,116 @@ static void glk_load_luts(const struct intel_crtc_state *crtc_state)
> }
> }
>
> +/* ilk+ "12.4" interpolated format (high 10 bits) */
> +static u32 ilk_lut_12p4_ldw(const struct drm_color_lut *color)
> +{
> + return (color->red >> 6) << 20 | (color->green >> 6) << 10 |
> + (color->blue >> 6);
> +}
> +
> +/* ilk+ "12.4" interpolated format (low 6 bits) */
> +static u32 ilk_lut_12p4_udw(const struct drm_color_lut *color)
> +{
> + return (color->red & 0x3f) << 24 | (color->green & 0x3f) << 14 |
> + (color->blue & 0x3f);
Blue is missing the shift.
I'm not 100% sure if the ldw vs. udw are the right way around. The spec
has at times been inconsistent with the odd vs. even descriptions,
sometimes even contradicting itself. Also it never really defines
whether it starts counting dwords from from 0 or 1, so not sure what
odd and even actually mean. Can I presume someone has checked this
on actual hardware?
> +}
> +
> +static void
> +icl_load_gcmax(const struct intel_crtc_state *crtc_state,
> + const struct drm_color_lut *color)
> +{
> + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc);
> + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev);
> + enum pipe pipe = crtc->pipe;
> +
> + /* Fixme: LUT entries are 16 bit only, so we can prog 0xFFFF max */
> + I915_WRITE(PREC_PAL_GC_MAX(pipe, 0), color->red);
> + I915_WRITE(PREC_PAL_GC_MAX(pipe, 1), color->green);
> + I915_WRITE(PREC_PAL_GC_MAX(pipe, 2), color->blue);
> +}
> +
> +static void
> +icl_program_gamma_superfine_segment(const struct intel_crtc_state *crtc_state)
> +{
> + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc);
> + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev);
> + const struct drm_property_blob *blob = crtc_state->base.gamma_lut;
> + const struct drm_color_lut *lut = blob->data;
> + enum pipe pipe = crtc->pipe;
> + u32 i;
> +
> + /*
> + * Every entry in the multi-segment LUT is corresponding to a superfine
> + * segment step which is 1/(8 * 128 * 256).
> + *
> + * Superfine segment has 9 entries, corresponding to values
> + * 0, 1/(8 * 128 * 256), 2/(8 * 128 * 256) .... 8/(8 * 128 * 256).
> + */
> + I915_WRITE(PREC_PAL_MULTI_SEG_INDEX(pipe), PAL_PREC_AUTO_INCREMENT);
> +
> + for (i = 0; i < 9; i++) {
> + const struct drm_color_lut *entry = &lut[i];
> +
> + I915_WRITE(PREC_PAL_MULTI_SEG_DATA(pipe),
> + ilk_lut_12p4_ldw(entry));
> + I915_WRITE(PREC_PAL_MULTI_SEG_DATA(pipe),
> + ilk_lut_12p4_udw(entry));
> + }
> +}
> +
> +static void
> +icl_program_gamma_multi_segment(const struct intel_crtc_state *crtc_state)
> +{
> + struct intel_crtc *crtc = to_intel_crtc(crtc_state->base.crtc);
> + struct drm_i915_private *dev_priv = to_i915(crtc->base.dev);
> + const struct drm_property_blob *blob = crtc_state->base.gamma_lut;
> + const struct drm_color_lut *lut = blob->data;
> + const struct drm_color_lut *entry;
> + enum pipe pipe = crtc->pipe;
> + u32 i;
> +
> + /*
> + *
> + * Program Fine segment (let's call it seg2)...
> + *
> + * Fine segment's step is 1/(128 * 256) ie 1/(128 * 256), 2/(128*256)
> + * ... 256/(128*256). So in order to program fine segment of LUT we
> + * need to pick every 8'th entry in LUT, and program 256 indexes.
> + *
> + * PAL_PREC_INDEX[0] and PAL_PREC_INDEX[1] map to seg2[1],
> + * with seg2[0] being unused by the hardware.
> + */
> + I915_WRITE(PREC_PAL_INDEX(pipe), PAL_PREC_AUTO_INCREMENT);
> + for (i = 1; i < 257; i++) {
> + entry = &lut[i * 8];
> + I915_WRITE(PREC_PAL_DATA(pipe), ilk_lut_12p4_ldw(entry));
> + I915_WRITE(PREC_PAL_DATA(pipe), ilk_lut_12p4_udw(entry));
> + }
> +
> + /*
> + * Program Coarse segment (let's call it seg3)...
> + *
> + * Coarse segment's starts from index 0 and it's step is 1/256 ie 0,
> + * 1/256, 2/256 ...256/256. As per the description of each entry in LUT
> + * above, we need to pick every (8 * 128)th entry in LUT, and
> + * program 256 of those.
> + *
> + * Spec is not very clear about if entries seg3[0] and seg3[1] are
> + * being used or not, but we still need to program these to advance
> + * the index.
> + */
> + for (i = 0; i < 256; i++) {
> + entry = &lut[i * 8 * 128];
> + I915_WRITE(PREC_PAL_DATA(pipe), ilk_lut_12p4_ldw(entry));
> + I915_WRITE(PREC_PAL_DATA(pipe), ilk_lut_12p4_udw(entry));
> + }
> +
> + /* The last entry in the LUT is to be programmed in GCMAX */
> + entry = &lut[256 * 8 * 128 + 1];
The +1 shouldn't be here.
OK, mostly looks good. With the minor issues addressed this is
Reviewed-by: Ville Syrjälä <ville.syrjala at linux.intel.com>
> + icl_load_gcmax(crtc_state, entry);
> + ivb_load_lut_ext_max(crtc);
> +}
> +
> static void icl_load_luts(const struct intel_crtc_state *crtc_state)
> {
> const struct drm_property_blob *gamma_lut = crtc_state->base.gamma_lut;
> @@ -775,10 +887,17 @@ static void icl_load_luts(const struct intel_crtc_state *crtc_state)
> if (crtc_state->base.degamma_lut)
> glk_load_degamma_lut(crtc_state);
>
> - if ((crtc_state->gamma_mode & GAMMA_MODE_MODE_MASK) ==
> - GAMMA_MODE_MODE_8BIT) {
> + switch (crtc_state->gamma_mode & GAMMA_MODE_MODE_MASK) {
> + case GAMMA_MODE_MODE_8BIT:
> i9xx_load_luts(crtc_state);
> - } else {
> + break;
> +
> + case GAMMA_MODE_MODE_12BIT_MULTI_SEGMENTED:
> + icl_program_gamma_superfine_segment(crtc_state);
> + icl_program_gamma_multi_segment(crtc_state);
> + break;
> +
> + default:
> bdw_load_lut_10(crtc, gamma_lut, PAL_PREC_INDEX_VALUE(0));
> ivb_load_lut_ext_max(crtc);
> }
> @@ -1209,7 +1328,7 @@ static u32 icl_gamma_mode(const struct intel_crtc_state *crtc_state)
> crtc_state_is_legacy_gamma(crtc_state))
> gamma_mode |= GAMMA_MODE_MODE_8BIT;
> else
> - gamma_mode |= GAMMA_MODE_MODE_10BIT;
> + gamma_mode |= GAMMA_MODE_MODE_12BIT_MULTI_SEGMENTED;
>
> return gamma_mode;
> }
> --
> 2.17.1
--
Ville Syrjälä
Intel
More information about the Intel-gfx
mailing list