[PATCH v2 01/12] drm/i915/xe3lpd: Update pmdemand programming
Gustavo Sousa
gustavo.sousa at intel.com
Wed Oct 23 22:16:34 UTC 2024
Quoting Clint Taylor (2024-10-23 18:46:50-03:00)
>From: Matt Roper <matthew.d.roper at intel.com>
>
>There are some minor changes to pmdemand handling on Xe3:
> - Active scalers are no longer tracked. We can simply skip the readout
> and programming of this field.
> - Active dbuf slices are no longer tracked. We should skip the readout
> and programming of this field and also make sure that it stays 0 in
> our software bookkeeping so that we won't erroneously return true
> from intel_pmdemand_needs_update() due to mismatches.
> - Even though there aren't enough pipes to utilize them, the size of
> the 'active pipes' field has expanded to four bits, taking over the
> register bits previously used for dbuf slices. Since the lower bits
> of the mask have moved, we need to update our reads/writes to handle
> this properly.
>
>v2: active pipes is no longer always max 3, add in the ability to go to
>4 for PTL.
>v3: use intel_display for display_ver check, use INTEL_NUM_PIPES
>v4: add a conditional for number of pipes macro vs using 3.
>v5: reverse conditional order of v4.
>
>Bspec: 68883, 69125
>Signed-off-by: Matt Roper <matthew.d.roper at intel.com>
>Signed-off-by: Matt Atwood <matthew.s.atwood at intel.com>
>Signed-off-by: Clint Taylor <Clinton.A.Taylor at intel.com>
>---
> drivers/gpu/drm/i915/display/intel_pmdemand.c | 68 +++++++++++++------
> drivers/gpu/drm/i915/display/intel_pmdemand.h | 4 +-
> drivers/gpu/drm/i915/i915_reg.h | 1 +
> 3 files changed, 50 insertions(+), 23 deletions(-)
>
>diff --git a/drivers/gpu/drm/i915/display/intel_pmdemand.c b/drivers/gpu/drm/i915/display/intel_pmdemand.c
>index ceaf9e3147da..746b547dddd0 100644
>--- a/drivers/gpu/drm/i915/display/intel_pmdemand.c
>+++ b/drivers/gpu/drm/i915/display/intel_pmdemand.c
>@@ -258,6 +258,7 @@ intel_pmdemand_connector_needs_update(struct intel_atomic_state *state)
>
> static bool intel_pmdemand_needs_update(struct intel_atomic_state *state)
> {
>+ struct intel_display *display = to_intel_display(state);
> const struct intel_bw_state *new_bw_state, *old_bw_state;
> const struct intel_cdclk_state *new_cdclk_state, *old_cdclk_state;
> const struct intel_crtc_state *new_crtc_state, *old_crtc_state;
>@@ -274,12 +275,16 @@ static bool intel_pmdemand_needs_update(struct intel_atomic_state *state)
> new_dbuf_state = intel_atomic_get_new_dbuf_state(state);
> old_dbuf_state = intel_atomic_get_old_dbuf_state(state);
> if (new_dbuf_state &&
>- (new_dbuf_state->active_pipes !=
>- old_dbuf_state->active_pipes ||
>- new_dbuf_state->enabled_slices !=
>- old_dbuf_state->enabled_slices))
>+ new_dbuf_state->active_pipes != old_dbuf_state->active_pipes)
> return true;
>
>+ if (DISPLAY_VER(display) < 30) {
>+ if (new_dbuf_state &&
>+ new_dbuf_state->enabled_slices !=
>+ old_dbuf_state->enabled_slices)
>+ return true;
>+ }
>+
> new_cdclk_state = intel_atomic_get_new_cdclk_state(state);
> old_cdclk_state = intel_atomic_get_old_cdclk_state(state);
> if (new_cdclk_state &&
>@@ -327,10 +332,15 @@ int intel_pmdemand_atomic_check(struct intel_atomic_state *state)
> if (IS_ERR(new_dbuf_state))
> return PTR_ERR(new_dbuf_state);
>
>- new_pmdemand_state->params.active_pipes =
>- min_t(u8, hweight8(new_dbuf_state->active_pipes), 3);
>- new_pmdemand_state->params.active_dbufs =
>- min_t(u8, hweight8(new_dbuf_state->enabled_slices), 3);
>+ if (DISPLAY_VER(i915) >= 30) {
>+ new_pmdemand_state->params.active_dbufs =
>+ min_t(u8, hweight8(new_dbuf_state->enabled_slices), 3);
Setting active_dbufs belongs to the "else" clause, as doing that is only
valid for pre-Xe3_LPD. (What I meant in my previous review was that the
selection of INTEL_NUM_PIPES(i915) vs 3 was the other way around).
>+ new_pmdemand_state->params.active_pipes =
>+ min_t(u8, hweight8(new_dbuf_state->active_pipes), INTEL_NUM_PIPES(i915));
Since this will require a new version, I think we would make stuff
simpler if we had a "max_active_pipes" variable that would hold
INTEL_NUM_PIPES(i915) for DISPLAY_VER(i915) >= 30 and 3 otherwise.
That way, this hunk could be turned into something like:
new_pmdemand_state->params.active_pipes =
min_t(u8, hweight8(new_dbuf_state->active_pipes), max_active_pipes);
if (DISPLAY_VER(i9) < 30)
new_pmdemand_state->params.active_dbufs =
min_t(u8, hweight8(new_dbuf_state->enabled_slices), 3);
(The above is using spaces instead of tabs).
>+ }
>+ else
>+ new_pmdemand_state->params.active_pipes =
>+ min_t(u8, hweight8(new_dbuf_state->active_pipes), 3);
>
> new_cdclk_state = intel_atomic_get_cdclk_state(state);
> if (IS_ERR(new_cdclk_state))
>@@ -395,27 +405,32 @@ intel_pmdemand_init_pmdemand_params(struct drm_i915_private *i915,
>
> reg2 = intel_de_read(i915, XELPDP_INITIATE_PMDEMAND_REQUEST(1));
>
>- /* Set 1*/
> pmdemand_state->params.qclk_gv_bw =
> REG_FIELD_GET(XELPDP_PMDEMAND_QCLK_GV_BW_MASK, reg1);
> pmdemand_state->params.voltage_index =
> REG_FIELD_GET(XELPDP_PMDEMAND_VOLTAGE_INDEX_MASK, reg1);
> pmdemand_state->params.qclk_gv_index =
> REG_FIELD_GET(XELPDP_PMDEMAND_QCLK_GV_INDEX_MASK, reg1);
>- pmdemand_state->params.active_pipes =
>- REG_FIELD_GET(XELPDP_PMDEMAND_PIPES_MASK, reg1);
>- pmdemand_state->params.active_dbufs =
>- REG_FIELD_GET(XELPDP_PMDEMAND_DBUFS_MASK, reg1);
> pmdemand_state->params.active_phys =
> REG_FIELD_GET(XELPDP_PMDEMAND_PHYS_MASK, reg1);
>
>- /* Set 2*/
> pmdemand_state->params.cdclk_freq_mhz =
> REG_FIELD_GET(XELPDP_PMDEMAND_CDCLK_FREQ_MASK, reg2);
> pmdemand_state->params.ddiclk_max =
> REG_FIELD_GET(XELPDP_PMDEMAND_DDICLK_FREQ_MASK, reg2);
>- pmdemand_state->params.scalers =
>- REG_FIELD_GET(XELPDP_PMDEMAND_SCALERS_MASK, reg2);
>+
>+ if (DISPLAY_VER(i915) >= 30) {
>+ pmdemand_state->params.active_pipes =
>+ REG_FIELD_GET(XE3_PMDEMAND_PIPES_MASK, reg1);
>+ } else {
>+ pmdemand_state->params.active_pipes =
>+ REG_FIELD_GET(XELPDP_PMDEMAND_PIPES_MASK, reg1);
>+ pmdemand_state->params.active_dbufs =
>+ REG_FIELD_GET(XELPDP_PMDEMAND_DBUFS_MASK, reg1);
>+
>+ pmdemand_state->params.scalers =
>+ REG_FIELD_GET(XELPDP_PMDEMAND_SCALERS_MASK, reg2);
>+ }
>
> unlock:
> mutex_unlock(&i915->display.pmdemand.lock);
>@@ -442,6 +457,10 @@ void intel_pmdemand_program_dbuf(struct drm_i915_private *i915,
> {
> u32 dbufs = min_t(u32, hweight8(dbuf_slices), 3);
>
>+ /* PM Demand only tracks active dbufs on pre-Xe3 platforms */
>+ if (DISPLAY_VER(i915) >= 30)
>+ return;
>+
> mutex_lock(&i915->display.pmdemand.lock);
> if (drm_WARN_ON(&i915->drm,
> !intel_pmdemand_check_prev_transaction(i915)))
>@@ -460,7 +479,8 @@ void intel_pmdemand_program_dbuf(struct drm_i915_private *i915,
> }
>
> static void
>-intel_pmdemand_update_params(const struct intel_pmdemand_state *new,
>+intel_pmdemand_update_params(struct drm_i915_private *i915,
Jani's feedback on other patches is that we should start using struct
intel_display for new variables and parameters.
--
Gustavo Sousa
>+ const struct intel_pmdemand_state *new,
> const struct intel_pmdemand_state *old,
> u32 *reg1, u32 *reg2, bool serialized)
> {
>@@ -495,16 +515,22 @@ intel_pmdemand_update_params(const struct intel_pmdemand_state *new,
> update_reg(reg1, qclk_gv_bw, XELPDP_PMDEMAND_QCLK_GV_BW_MASK);
> update_reg(reg1, voltage_index, XELPDP_PMDEMAND_VOLTAGE_INDEX_MASK);
> update_reg(reg1, qclk_gv_index, XELPDP_PMDEMAND_QCLK_GV_INDEX_MASK);
>- update_reg(reg1, active_pipes, XELPDP_PMDEMAND_PIPES_MASK);
>- update_reg(reg1, active_dbufs, XELPDP_PMDEMAND_DBUFS_MASK);
> update_reg(reg1, active_phys, XELPDP_PMDEMAND_PHYS_MASK);
>
> /* Set 2*/
> update_reg(reg2, cdclk_freq_mhz, XELPDP_PMDEMAND_CDCLK_FREQ_MASK);
> update_reg(reg2, ddiclk_max, XELPDP_PMDEMAND_DDICLK_FREQ_MASK);
>- update_reg(reg2, scalers, XELPDP_PMDEMAND_SCALERS_MASK);
> update_reg(reg2, plls, XELPDP_PMDEMAND_PLLS_MASK);
>
>+ if (DISPLAY_VER(i915) >= 30) {
>+ update_reg(reg1, active_pipes, XE3_PMDEMAND_PIPES_MASK);
>+ } else {
>+ update_reg(reg1, active_pipes, XELPDP_PMDEMAND_PIPES_MASK);
>+ update_reg(reg1, active_dbufs, XELPDP_PMDEMAND_DBUFS_MASK);
>+
>+ update_reg(reg2, scalers, XELPDP_PMDEMAND_SCALERS_MASK);
>+ }
>+
> #undef update_reg
> }
>
>@@ -529,7 +555,7 @@ intel_pmdemand_program_params(struct drm_i915_private *i915,
> reg2 = intel_de_read(i915, XELPDP_INITIATE_PMDEMAND_REQUEST(1));
> mod_reg2 = reg2;
>
>- intel_pmdemand_update_params(new, old, &mod_reg1, &mod_reg2,
>+ intel_pmdemand_update_params(i915, new, old, &mod_reg1, &mod_reg2,
> serialized);
>
> if (reg1 != mod_reg1) {
>diff --git a/drivers/gpu/drm/i915/display/intel_pmdemand.h b/drivers/gpu/drm/i915/display/intel_pmdemand.h
>index 128fd61f8f14..a1c49efdc493 100644
>--- a/drivers/gpu/drm/i915/display/intel_pmdemand.h
>+++ b/drivers/gpu/drm/i915/display/intel_pmdemand.h
>@@ -20,14 +20,14 @@ struct pmdemand_params {
> u8 voltage_index;
> u8 qclk_gv_index;
> u8 active_pipes;
>- u8 active_dbufs;
>+ u8 active_dbufs; /* pre-Xe3 only */
> /* Total number of non type C active phys from active_phys_mask */
> u8 active_phys;
> u8 plls;
> u16 cdclk_freq_mhz;
> /* max from ddi_clocks[] */
> u16 ddiclk_max;
>- u8 scalers;
>+ u8 scalers; /* pre-Xe3 only */
> };
>
> struct intel_pmdemand_state {
>diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
>index 405f409e9761..89e4381f8baa 100644
>--- a/drivers/gpu/drm/i915/i915_reg.h
>+++ b/drivers/gpu/drm/i915/i915_reg.h
>@@ -2696,6 +2696,7 @@
> #define XELPDP_PMDEMAND_QCLK_GV_BW_MASK REG_GENMASK(31, 16)
> #define XELPDP_PMDEMAND_VOLTAGE_INDEX_MASK REG_GENMASK(14, 12)
> #define XELPDP_PMDEMAND_QCLK_GV_INDEX_MASK REG_GENMASK(11, 8)
>+#define XE3_PMDEMAND_PIPES_MASK REG_GENMASK(7, 4)
> #define XELPDP_PMDEMAND_PIPES_MASK REG_GENMASK(7, 6)
> #define XELPDP_PMDEMAND_DBUFS_MASK REG_GENMASK(5, 4)
> #define XELPDP_PMDEMAND_PHYS_MASK REG_GENMASK(2, 0)
>--
>2.25.1
>
More information about the Intel-xe
mailing list