Skip to content

L
linux
Commit 1b5d0e7e authored by Ilya Bakoulin's avatar Ilya Bakoulin Committed by Alex Deucher
Browse files
Options

drm/amd/display: Speed up DML fast_validate path 

[Why]
Iterating over every voltage state when we need to validate thousands of
configurations all at once (i.e. display hotplug) can take a significant
amount of time.

[How]
Check just the highest voltage state when fast_validate is true to
verify whether the configuration can work at all, then do a proper
validation including all voltage states later when fast_validate is false.
Reviewed-by: default avatarJun Lei <Jun.Lei@amd.com>
Acked-by: default avatarJasdeep Dhillon <jdhillon@amd.com>
Signed-off-by: default avatarIlya Bakoulin <Ilya.Bakoulin@amd.com>
Signed-off-by: default avatarAlex Deucher <alexander.deucher@amd.com>
parent b8ff7e08
Hide whitespace changes
Inline Side-by-side
Showing with 23 additions and 17 deletions
drivers/gpu/drm/amd/display/dc/dml/dcn32/dcn32_fpu.c
View file @ 1b5d0e7e
......@@ -1551,6 +1551,7 @@ bool dcn32_internal_validate_bw(struct dc *dc,
context->bw_ctx.dml.soc.allow_for_pstate_or_stutter_in_vblank_final =
dm_prefetch_support_fclk_and_stutter;
context->bw_ctx.dml.validate_max_state = fast_validate;
vlevel = dml_get_voltage_level(&context->bw_ctx.dml, pipes, pipe_cnt);
/* Last attempt with Prefetch mode 2 (dm_prefetch_support_stutter == 3) */
......@@ -1559,6 +1560,7 @@ bool dcn32_internal_validate_bw(struct dc *dc,
dm_prefetch_support_stutter;
vlevel = dml_get_voltage_level(&context->bw_ctx.dml, pipes, pipe_cnt);
}
context->bw_ctx.dml.validate_max_state = false;
if (vlevel < context->bw_ctx.dml.soc.num_states) {
memset(split, 0, sizeof(split));
......
drivers/gpu/drm/amd/display/dc/dml/dcn32/display_mode_vba_32.c
View file @ 1b5d0e7e
......@@ -1707,7 +1707,7 @@ static void mode_support_configuration(struct vba_vars_st *v,
void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_lib)
{
struct vba_vars_st *v = &mode_lib->vba;
int i, j;
int i, j, start_state;
unsigned int k, m;
unsigned int MaximumMPCCombine;
unsigned int NumberOfNonCombinedSurfaceOfMaximumBandwidth;
......@@ -1720,7 +1720,10 @@ void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l
#endif
/*MODE SUPPORT, VOLTAGE STATE AND SOC CONFIGURATION*/
if (mode_lib->validate_max_state)
start_state = v->soc.num_states - 1;
else
start_state = 0;
/*Scale Ratio, taps Support Check*/
mode_lib->vba.ScaleRatioAndTapsSupport = true;
......@@ -2009,7 +2012,7 @@ void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l
mode_lib->vba.MPCCombineMethodIncompatible = v->dummy_vars.dml32_ModeSupportAndSystemConfigurationFull.MPCCombineMethodAsNeededForPStateChangeAndVoltage
&& v->dummy_vars.dml32_ModeSupportAndSystemConfigurationFull.MPCCombineMethodAsPossible;
for (i = 0; i < v->soc.num_states; i++) {
for (i = start_state; i < v->soc.num_states; i++) {
for (j = 0; j < 2; j++) {
mode_lib->vba.TotalNumberOfActiveDPP[i][j] = 0;
mode_lib->vba.TotalAvailablePipesSupport[i][j] = true;
......@@ -2286,7 +2289,7 @@ void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l
}
}
for (i = 0; i < v->soc.num_states; ++i) {
for (i = start_state; i < v->soc.num_states; ++i) {
mode_lib->vba.ExceededMultistreamSlots[i] = false;
for (k = 0; k < mode_lib->vba.NumberOfActiveSurfaces; ++k) {
if (mode_lib->vba.OutputMultistreamEn[k] == true && mode_lib->vba.OutputMultistreamId[k] == k) {
......@@ -2386,7 +2389,7 @@ void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l
}
}
for (i = 0; i < v->soc.num_states; ++i) {
for (i = start_state; i < v->soc.num_states; ++i) {
mode_lib->vba.DTBCLKRequiredMoreThanSupported[i] = false;
for (k = 0; k < mode_lib->vba.NumberOfActiveSurfaces; ++k) {
if (mode_lib->vba.BlendingAndTiming[k] == k
......@@ -2403,7 +2406,7 @@ void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l
}
}
for (i = 0; i < v->soc.num_states; ++i) {
for (i = start_state; i < v->soc.num_states; ++i) {
mode_lib->vba.ODMCombine2To1SupportCheckOK[i] = true;
mode_lib->vba.ODMCombine4To1SupportCheckOK[i] = true;
for (k = 0; k < mode_lib->vba.NumberOfActiveSurfaces; ++k) {
......@@ -2421,7 +2424,7 @@ void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l
}
}
for (i = 0; i < v->soc.num_states; i++) {
for (i = start_state; i < v->soc.num_states; i++) {
mode_lib->vba.DSCCLKRequiredMoreThanSupported[i] = false;
for (k = 0; k <= mode_lib->vba.NumberOfActiveSurfaces - 1; k++) {
if (mode_lib->vba.BlendingAndTiming[k] == k) {
......@@ -2458,7 +2461,7 @@ void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l
/* Check DSC Unit and Slices Support */
v->dummy_vars.dml32_ModeSupportAndSystemConfigurationFull.TotalDSCUnitsRequired = 0;
for (i = 0; i < v->soc.num_states; ++i) {
for (i = start_state; i < v->soc.num_states; ++i) {
mode_lib->vba.NotEnoughDSCUnits[i] = false;
mode_lib->vba.NotEnoughDSCSlices[i] = false;
v->dummy_vars.dml32_ModeSupportAndSystemConfigurationFull.TotalDSCUnitsRequired = 0;
......@@ -2493,7 +2496,7 @@ void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l
}
/*DSC Delay per state*/
for (i = 0; i < v->soc.num_states; ++i) {
for (i = start_state; i < v->soc.num_states; ++i) {
for (k = 0; k < mode_lib->vba.NumberOfActiveSurfaces; ++k) {
mode_lib->vba.DSCDelayPerState[i][k] = dml32_DSCDelayRequirement(
mode_lib->vba.RequiresDSC[i][k], mode_lib->vba.ODMCombineEnablePerState[i][k],
......@@ -2520,7 +2523,7 @@ void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l
//Calculate Swath, DET Configuration, DCFCLKDeepSleep
//
for (i = 0; i < (int) v->soc.num_states; ++i) {
for (i = start_state; i < (int) v->soc.num_states; ++i) {
for (j = 0; j <= 1; ++j) {
for (k = 0; k < mode_lib->vba.NumberOfActiveSurfaces; ++k) {
mode_lib->vba.RequiredDPPCLKThisState[k] = mode_lib->vba.RequiredDPPCLK[i][j][k];
......@@ -2655,7 +2658,7 @@ void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l
mode_lib->vba.SurfaceSizeInMALL,
&mode_lib->vba.ExceededMALLSize);
for (i = 0; i < v->soc.num_states; i++) {
for (i = start_state; i < v->soc.num_states; i++) {
for (j = 0; j < 2; j++) {
for (k = 0; k <= mode_lib->vba.NumberOfActiveSurfaces - 1; k++) {
mode_lib->vba.swath_width_luma_ub_this_state[k] =
......@@ -2882,7 +2885,7 @@ void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l
}
//Calculate Return BW
for (i = 0; i < (int) v->soc.num_states; ++i) {
for (i = start_state; i < (int) v->soc.num_states; ++i) {
for (j = 0; j <= 1; ++j) {
for (k = 0; k <= mode_lib->vba.NumberOfActiveSurfaces - 1; k++) {
if (mode_lib->vba.BlendingAndTiming[k] == k) {
......@@ -2961,7 +2964,7 @@ void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l
&mode_lib->vba.MinPrefetchMode,
&mode_lib->vba.MaxPrefetchMode);
for (i = 0; i < (int) v->soc.num_states; ++i) {
for (i = start_state; i < (int) v->soc.num_states; ++i) {
for (j = 0; j <= 1; ++j)
mode_lib->vba.DCFCLKState[i][j] = mode_lib->vba.DCFCLKPerState[i];
}
......@@ -3083,7 +3086,7 @@ void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l
mode_lib->vba.DCFCLKState);
} // UseMinimumRequiredDCFCLK == true
for (i = 0; i < (int) v->soc.num_states; ++i) {
for (i = start_state; i < (int) v->soc.num_states; ++i) {
for (j = 0; j <= 1; ++j) {
mode_lib->vba.ReturnBWPerState[i][j] = dml32_get_return_bw_mbps(&mode_lib->vba.soc, i,
mode_lib->vba.HostVMEnable, mode_lib->vba.DCFCLKState[i][j],
......@@ -3092,7 +3095,7 @@ void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l
}
//Re-ordering Buffer Support Check
for (i = 0; i < (int) v->soc.num_states; ++i) {
for (i = start_state; i < (int) v->soc.num_states; ++i) {
for (j = 0; j <= 1; ++j) {
if ((mode_lib->vba.ROBBufferSizeInKByte - mode_lib->vba.PixelChunkSizeInKByte) * 1024
/ mode_lib->vba.ReturnBWPerState[i][j]
......@@ -3114,7 +3117,7 @@ void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l
+ mode_lib->vba.ReadBandwidthChroma[k];
}
for (i = 0; i < (int) v->soc.num_states; ++i) {
for (i = start_state; i < (int) v->soc.num_states; ++i) {
for (j = 0; j <= 1; ++j) {
mode_lib->vba.MaxTotalVerticalActiveAvailableBandwidth[i][j] =
dml_min3(mode_lib->vba.ReturnBusWidth * mode_lib->vba.DCFCLKState[i][j]
......@@ -3138,7 +3141,7 @@ void dml32_ModeSupportAndSystemConfigurationFull(struct display_mode_lib *mode_l
/* Prefetch Check */
for (i = 0; i < (int) v->soc.num_states; ++i) {
for (i = start_state; i < (int) v->soc.num_states; ++i) {
for (j = 0; j <= 1; ++j) {
mode_lib->vba.TimeCalc = 24 / mode_lib->vba.ProjectedDCFCLKDeepSleep[i][j];
......
drivers/gpu/drm/amd/display/dc/dml/display_mode_lib.h
View file @ 1b5d0e7e
......@@ -91,6 +91,7 @@ struct display_mode_lib {
struct dal_logger *logger;
struct dml_funcs funcs;
struct _vcs_dpi_display_e2e_pipe_params_st dml_pipe_state[6];
bool validate_max_state;
};
void dml_init_instance(struct display_mode_lib *lib,
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7