Quellcodebibliothek Statistik Leitseite products/sources/formale Sprachen/C/Linux/drivers/gpu/drm/amd/display/dc/dml2/   (Open Source Betriebssystem Version 6.17.9©)  Datei vom 24.10.2025 mit Größe 12 kB image not shown  

Quelle  dml2_policy.c   Sprache: C

 
/* SPDX-License-Identifier: MIT */
/*
 * Copyright 2023 Advanced Micro Devices, Inc.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
 * OTHER DEALINGS IN THE SOFTWARE.
 *
 * Authors: AMD
 *
 */


#include "dml2_policy.h"

static void get_optimal_ntuple(
 const struct soc_bounding_box_st *socbb,
 struct soc_state_bounding_box_st *entry)
{
 if (entry->dcfclk_mhz > 0) {
  float bw_on_sdp = (float)(entry->dcfclk_mhz * socbb->return_bus_width_bytes * ((float)socbb->pct_ideal_sdp_bw_after_urgent / 100));

  entry->fabricclk_mhz = bw_on_sdp / (socbb->return_bus_width_bytes * ((float)socbb->pct_ideal_fabric_bw_after_urgent / 100));
  entry->dram_speed_mts = bw_on_sdp / (socbb->num_chans *
   socbb->dram_channel_width_bytes * ((float)socbb->pct_ideal_dram_bw_after_urgent_pixel_only / 100));
 } else if (entry->fabricclk_mhz > 0) {
  float bw_on_fabric = (float)(entry->fabricclk_mhz * socbb->return_bus_width_bytes * ((float)socbb->pct_ideal_fabric_bw_after_urgent / 100));

  entry->dcfclk_mhz = bw_on_fabric / (socbb->return_bus_width_bytes * ((float)socbb->pct_ideal_sdp_bw_after_urgent / 100));
  entry->dram_speed_mts = bw_on_fabric / (socbb->num_chans *
   socbb->dram_channel_width_bytes * ((float)socbb->pct_ideal_dram_bw_after_urgent_pixel_only / 100));
 } else if (entry->dram_speed_mts > 0) {
  float bw_on_dram = (float)(entry->dram_speed_mts * socbb->num_chans *
   socbb->dram_channel_width_bytes * ((float)socbb->pct_ideal_dram_bw_after_urgent_pixel_only / 100));

  entry->fabricclk_mhz = bw_on_dram / (socbb->return_bus_width_bytes * ((float)socbb->pct_ideal_fabric_bw_after_urgent / 100));
  entry->dcfclk_mhz = bw_on_dram / (socbb->return_bus_width_bytes * ((float)socbb->pct_ideal_sdp_bw_after_urgent / 100));
 }
}

static float calculate_net_bw_in_mbytes_sec(const struct soc_bounding_box_st *socbb,
 struct soc_state_bounding_box_st *entry)
{
 float memory_bw_mbytes_sec = (float)(entry->dram_speed_mts *  socbb->num_chans *
  socbb->dram_channel_width_bytes * ((float)socbb->pct_ideal_dram_bw_after_urgent_pixel_only / 100));

 float fabric_bw_mbytes_sec = (float)(entry->fabricclk_mhz * socbb->return_bus_width_bytes * ((float)socbb->pct_ideal_fabric_bw_after_urgent / 100));

 float sdp_bw_mbytes_sec = (float)(entry->dcfclk_mhz * socbb->return_bus_width_bytes * ((float)socbb->pct_ideal_sdp_bw_after_urgent / 100));

 float limiting_bw_mbytes_sec = memory_bw_mbytes_sec;

 if (fabric_bw_mbytes_sec < limiting_bw_mbytes_sec)
  limiting_bw_mbytes_sec = fabric_bw_mbytes_sec;

 if (sdp_bw_mbytes_sec < limiting_bw_mbytes_sec)
  limiting_bw_mbytes_sec = sdp_bw_mbytes_sec;

 return limiting_bw_mbytes_sec;
}

static void insert_entry_into_table_sorted(const struct soc_bounding_box_st *socbb,
 struct soc_states_st *table,
 struct soc_state_bounding_box_st *entry)
{
 int index = 0;
 int i = 0;
 float net_bw_of_new_state = 0;

 get_optimal_ntuple(socbb, entry);

 if (table->num_states == 0) {
  index = 0;
 } else {
  net_bw_of_new_state = calculate_net_bw_in_mbytes_sec(socbb, entry);
  while (net_bw_of_new_state > calculate_net_bw_in_mbytes_sec(socbb, &table->state_array[index])) {
   index++;
   if (index >= (int) table->num_states)
    break;
  }

  for (i = table->num_states; i > index; i--) {
   table->state_array[i] = table->state_array[i - 1];
  }
  //ASSERT(index < MAX_CLK_TABLE_SIZE);
 }

 table->state_array[index] = *entry;
 table->state_array[index].dcfclk_mhz = (int)entry->dcfclk_mhz;
 table->state_array[index].fabricclk_mhz = (int)entry->fabricclk_mhz;
 table->state_array[index].dram_speed_mts = (int)entry->dram_speed_mts;
 table->num_states++;
}

static void remove_entry_from_table_at_index(struct soc_states_st *table,
 unsigned int index)
{
 int i;

 if (table->num_states == 0)
  return;

 for (i = index; i < (int) table->num_states - 1; i++) {
  table->state_array[i] = table->state_array[i + 1];
 }
 memset(&table->state_array[--table->num_states], 0, sizeof(struct soc_state_bounding_box_st));
}

int dml2_policy_build_synthetic_soc_states(struct dml2_policy_build_synthetic_soc_states_scratch *s,
 struct dml2_policy_build_synthetic_soc_states_params *p)
{
 int i, j;
 unsigned int min_fclk_mhz = p->in_states->state_array[0].fabricclk_mhz;
 unsigned int min_dcfclk_mhz = p->in_states->state_array[0].dcfclk_mhz;
 unsigned int min_socclk_mhz = p->in_states->state_array[0].socclk_mhz;

 int max_dcfclk_mhz = 0, max_dispclk_mhz = 0, max_dppclk_mhz = 0,
  max_phyclk_mhz = 0, max_dtbclk_mhz = 0, max_fclk_mhz = 0,
  max_uclk_mhz = 0, max_socclk_mhz = 0;

 int num_uclk_dpms = 0, num_fclk_dpms = 0;

 for (i = 0; i < __DML_MAX_STATE_ARRAY_SIZE__; i++) {
  if (p->in_states->state_array[i].dcfclk_mhz > max_dcfclk_mhz)
   max_dcfclk_mhz = (int) p->in_states->state_array[i].dcfclk_mhz;
  if (p->in_states->state_array[i].fabricclk_mhz > max_fclk_mhz)
   max_fclk_mhz = (int) p->in_states->state_array[i].fabricclk_mhz;
  if (p->in_states->state_array[i].socclk_mhz > max_socclk_mhz)
   max_socclk_mhz = (int) p->in_states->state_array[i].socclk_mhz;
  if (p->in_states->state_array[i].dram_speed_mts > max_uclk_mhz)
   max_uclk_mhz = (int) p->in_states->state_array[i].dram_speed_mts;
  if (p->in_states->state_array[i].dispclk_mhz > max_dispclk_mhz)
   max_dispclk_mhz = (int) p->in_states->state_array[i].dispclk_mhz;
  if (p->in_states->state_array[i].dppclk_mhz > max_dppclk_mhz)
   max_dppclk_mhz = (int) p->in_states->state_array[i].dppclk_mhz;
  if (p->in_states->state_array[i].phyclk_mhz > max_phyclk_mhz)
   max_phyclk_mhz = (int)p->in_states->state_array[i].phyclk_mhz;
  if (p->in_states->state_array[i].dtbclk_mhz > max_dtbclk_mhz)
   max_dtbclk_mhz = (int)p->in_states->state_array[i].dtbclk_mhz;

  if (p->in_states->state_array[i].fabricclk_mhz > 0)
   num_fclk_dpms++;
  if (p->in_states->state_array[i].dram_speed_mts > 0)
   num_uclk_dpms++;
 }

 if (!max_dcfclk_mhz || !max_dispclk_mhz || !max_dppclk_mhz || !max_phyclk_mhz || !max_dtbclk_mhz)
  return -1;

 p->out_states->num_states = 0;

 s->entry = p->in_states->state_array[0];

 s->entry.dispclk_mhz = max_dispclk_mhz;
 s->entry.dppclk_mhz = max_dppclk_mhz;
 s->entry.dtbclk_mhz = max_dtbclk_mhz;
 s->entry.phyclk_mhz = max_phyclk_mhz;

 s->entry.dscclk_mhz = max_dispclk_mhz / 3;
 s->entry.phyclk_mhz = max_phyclk_mhz;
 s->entry.dtbclk_mhz = max_dtbclk_mhz;

 // Insert all the DCFCLK STAs first
 for (i = 0; i < p->num_dcfclk_stas; i++) {
  s->entry.dcfclk_mhz = p->dcfclk_stas_mhz[i];
  s->entry.fabricclk_mhz = 0;
  s->entry.dram_speed_mts = 0;
  if (i > 0)
   s->entry.socclk_mhz = max_socclk_mhz;

  insert_entry_into_table_sorted(p->in_bbox, p->out_states, &s->entry);
 }

 // Insert the UCLK DPMS
 for (i = 0; i < num_uclk_dpms; i++) {
  s->entry.dcfclk_mhz = 0;
  s->entry.fabricclk_mhz = 0;
  s->entry.dram_speed_mts = p->in_states->state_array[i].dram_speed_mts;
  if (i == 0) {
   s->entry.socclk_mhz = min_socclk_mhz;
  } else {
   s->entry.socclk_mhz = max_socclk_mhz;
  }

  insert_entry_into_table_sorted(p->in_bbox, p->out_states, &s->entry);
 }

 // Insert FCLK DPMs (if present)
 if (num_fclk_dpms > 2) {
  for (i = 0; i < num_fclk_dpms; i++) {
   s->entry.dcfclk_mhz = 0;
   s->entry.fabricclk_mhz = p->in_states->state_array[i].fabricclk_mhz;
   s->entry.dram_speed_mts = 0;

  insert_entry_into_table_sorted(p->in_bbox, p->out_states, &s->entry);
  }
 }
 // Add max FCLK
 else {
  s->entry.dcfclk_mhz = 0;
  s->entry.fabricclk_mhz = p->in_states->state_array[num_fclk_dpms - 1].fabricclk_mhz;
  s->entry.dram_speed_mts = 0;

  insert_entry_into_table_sorted(p->in_bbox, p->out_states, &s->entry);
 }

 // Remove states that require higher clocks than are supported
 for (i = p->out_states->num_states - 1; i >= 0; i--) {
  if (p->out_states->state_array[i].dcfclk_mhz > max_dcfclk_mhz ||
   p->out_states->state_array[i].fabricclk_mhz > max_fclk_mhz ||
   p->out_states->state_array[i].dram_speed_mts > max_uclk_mhz)
   remove_entry_from_table_at_index(p->out_states, i);
 }

 // At this point, the table contains all "points of interest" based on
 // DPMs from PMFW, and STAs. Table is sorted by BW, and all clock
 // ratios (by derate, are exact).

 // Round up UCLK to DPMs
 for (i = p->out_states->num_states - 1; i >= 0; i--) {
  for (j = 0; j < num_uclk_dpms; j++) {
   if (p->in_states->state_array[j].dram_speed_mts >= p->out_states->state_array[i].dram_speed_mts) {
    p->out_states->state_array[i].dram_speed_mts = p->in_states->state_array[j].dram_speed_mts;
    break;
   }
  }
 }

 // If FCLK is coarse grained, round up to next DPMs
 if (num_fclk_dpms > 2) {
  for (i = p->out_states->num_states - 1; i >= 0; i--) {
   for (j = 0; j < num_fclk_dpms; j++) {
    if (p->in_states->state_array[j].fabricclk_mhz >= p->out_states->state_array[i].fabricclk_mhz) {
     p->out_states->state_array[i].fabricclk_mhz = p->in_states->state_array[j].fabricclk_mhz;
     break;
    }
   }
  }
 }

 // Clamp to min FCLK/DCFCLK
 for (i = p->out_states->num_states - 1; i >= 0; i--) {
  if (p->out_states->state_array[i].fabricclk_mhz < min_fclk_mhz) {
   p->out_states->state_array[i].fabricclk_mhz = min_fclk_mhz;
  }
  if (p->out_states->state_array[i].dcfclk_mhz < min_dcfclk_mhz) {
   p->out_states->state_array[i].dcfclk_mhz = min_dcfclk_mhz;
  }
 }

 // Remove duplicate states, note duplicate states are always neighbouring since table is sorted.
 i = 0;
 while (i < (int) p->out_states->num_states - 1) {
  if (p->out_states->state_array[i].dcfclk_mhz == p->out_states->state_array[i + 1].dcfclk_mhz &&
   p->out_states->state_array[i].fabricclk_mhz == p->out_states->state_array[i + 1].fabricclk_mhz &&
   p->out_states->state_array[i].dram_speed_mts == p->out_states->state_array[i + 1].dram_speed_mts)
   remove_entry_from_table_at_index(p->out_states, i);
 else
  i++;
 }

 return 0;
}

void build_unoptimized_policy_settings(enum dml_project_id project, struct dml_mode_eval_policy_st *policy)
{
 for (int i = 0; i < __DML_NUM_PLANES__; i++) {
  policy->MPCCombineUse[i] = dml_mpc_as_needed_for_voltage; // TOREVIEW: Is this still needed?  When is MPCC useful for pstate given CRB?
  policy->ODMUse[i] = dml_odm_use_policy_combine_as_needed;
  policy->ImmediateFlipRequirement[i] = dml_immediate_flip_required;
  policy->AllowForPStateChangeOrStutterInVBlank[i] = dml_prefetch_support_uclk_fclk_and_stutter_if_possible;
 }

 /* Change the default policy initializations as per spreadsheet. We might need to
 * review and change them later on as per Jun's earlier comments.
 */

 policy->UseUnboundedRequesting = dml_unbounded_requesting_enable;
 policy->UseMinimumRequiredDCFCLK = false;
 policy->DRAMClockChangeRequirementFinal = true// TOREVIEW: What does this mean?
 policy->FCLKChangeRequirementFinal = true// TOREVIEW: What does this mean?
 policy->USRRetrainingRequiredFinal = true;
 policy->EnhancedPrefetchScheduleAccelerationFinal = true// TOREVIEW: What does this mean?
 policy->NomDETInKByteOverrideEnable = false;
 policy->NomDETInKByteOverrideValue = 0;
 policy->DCCProgrammingAssumesScanDirectionUnknownFinal = true;
 policy->SynchronizeTimingsFinal = true;
 policy->SynchronizeDRRDisplaysForUCLKPStateChangeFinal = true;
 policy->AssumeModeSupportAtMaxPwrStateEvenDRAMClockChangeNotSupported = true// TOREVIEW: What does this mean?
 policy->AssumeModeSupportAtMaxPwrStateEvenFClockChangeNotSupported = true// TOREVIEW: What does this mean?
 if (project == dml_project_dcn35 ||
  project == dml_project_dcn36 ||
  project == dml_project_dcn351) {
  policy->DCCProgrammingAssumesScanDirectionUnknownFinal = false;
  policy->EnhancedPrefetchScheduleAccelerationFinal = 0;
  policy->AllowForPStateChangeOrStutterInVBlankFinal = dml_prefetch_support_uclk_fclk_and_stutter_if_possible; /*new*/
  policy->UseOnlyMaxPrefetchModes = 1;
 }
}

Messung V0.5
C=94 H=87 G=90

¤ Dauer der Verarbeitung: 0.1 Sekunden  (vorverarbeitet)  ¤

*© Formatika GbR, Deutschland






Wurzel

Suchen

Beweissystem der NASA

Beweissystem Isabelle

NIST Cobol Testsuite

Cephes Mathematical Library

Wiener Entwicklungsmethode

Haftungshinweis

Die Informationen auf dieser Webseite wurden nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit, noch Qualität der bereit gestellten Informationen zugesichert.

Bemerkung:

Die farbliche Syntaxdarstellung und die Messung sind noch experimentell.