drm/amd/powerplay: using fixed vram address for table transferring
Subsystems (2)
PATCH 1/2 drm/amd/powerplay: using fixed vram address for table transferring
By this, we can avoid to pass in the vram address on every table transferring.
--- drivers/gpu/drm/amd/powerplay/amdgpu_smu.c | 42 ++++++++++++------- drivers/gpu/drm/amd/powerplay/arcturus_ppt.c | 1 + .../gpu/drm/amd/powerplay/inc/amdgpu_smu.h | 2 + drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h | 2 + drivers/gpu/drm/amd/powerplay/inc/smu_v12_0.h | 2 + drivers/gpu/drm/amd/powerplay/navi10_ppt.c | 1 + drivers/gpu/drm/amd/powerplay/renoir_ppt.c | 1 + drivers/gpu/drm/amd/powerplay/smu_internal.h | 2 + drivers/gpu/drm/amd/powerplay/smu_v11_0.c | 18 ++++++++ drivers/gpu/drm/amd/powerplay/smu_v12_0.c | 18 ++++++++ drivers/gpu/drm/amd/powerplay/vega20_ppt.c | 1 + 11 files changed, 76 insertions(+), 14 deletions(-)
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +28 -14
@@ -490,26 +490,19 @@ int smu_update_table(struct smu_context *smu, enum smu_table_id table_index, int {struct smu_table_context *smu_table = &smu->smu_table;struct amdgpu_device *adev = smu->adev; -struct smu_table *table =NULL ; -int ret =0 ; +struct smu_table *table = smu_table->driver_table;int table_id =smu_table_get_index (smu, table_index); +uint32_t table_size; +int ret =0 ;if (!table_data || table_id >=SMU_TABLE_COUNT || table_id <0 )return -EINVAL ; - table = &smu_table->tables[table_index]; + table_size = smu_table->tables[table_index].size;if (drv2smu) -memcpy (table->cpu_addr, table_data, table->size); +memcpy (table->cpu_addr, table_data, table_size); - ret =smu_send_smc_msg_with_param (smu, SMU_MSG_SetDriverDramAddrHigh, -upper_32_bits (table->mc_address)); -if (ret) -return ret; - ret =smu_send_smc_msg_with_param (smu, SMU_MSG_SetDriverDramAddrLow, -lower_32_bits (table->mc_address)); -if (ret) -return ret; ret = smu_send_smc_msg_with_param(smu, drv2smu ? SMU_MSG_TransferTableDram2Smu : SMU_MSG_TransferTableSmu2Dram,
@@ -521,7 +514,7 @@ int smu_update_table(struct smu_context *smu, enum smu_table_id table_index, int adev->nbio.funcs->hdp_flush (adev,NULL );if (!drv2smu) -memcpy (table_data, table->cpu_addr, table->size); +memcpy (table_data, table->cpu_addr, table_size);return ret; }
@@ -948,7 +941,9 @@ static int smu_init_fb_allocations(struct smu_context *smu)struct amdgpu_device *adev = smu->adev;struct smu_table_context *smu_table = &smu->smu_table;struct smu_table *tables = smu_table->tables; -int ret, i; +struct smu_table **driver_table = &(smu_table->driver_table); +uint32_t max_table_size =0 ; +int ret, i, index =0 ;for (i =0 ; i <SMU_TABLE_COUNT ; i++) {if (tables[i].size ==0 )
@@ -964,6 +959,21 @@ static int smu_init_fb_allocations(struct smu_context *smu) goto failed; } +for (i =0 ; i <SMU_TABLE_COUNT ; i++) { +if (tables[i].size ==0 ) +continue ; + +if (i ==SMU_TABLE_PMSTATUSLOG ) +continue ; + +if (max_table_size < tables[i].size) { + max_table_size = tables[i].size; + index = i; + } + } + + *driver_table = &tables[index]; +return 0 ; failed:while (--i >=0 ) {
@@ -1063,6 +1073,10 @@ static int smu_smc_table_hw_init(struct smu_context *smu,/* smu_dump_pptable(smu); */ + ret =smu_set_driver_table_location (smu); +if (ret) +return ret; +/* * Copy pptable bo in the vram to smc with SMU MSGs such as * SetDriverDramAddr and TransferTableDram2Smu.
drivers/gpu/drm/amd/powerplay/arcturus_ppt.c +1 -0
@@ -2261,6 +2261,7 @@ static const struct pptable_funcs arcturus_ppt_funcs = { .check_fw_version = smu_v11_0_check_fw_version, .write_pptable = smu_v11_0_write_pptable, .set_min_dcef_deep_sleep = smu_v11_0_set_min_dcef_deep_sleep, + .set_driver_table_location = smu_v11_0_set_driver_table_location, .set_tool_table_location = smu_v11_0_set_tool_table_location, .notify_memory_pool_location = smu_v11_0_notify_memory_pool_location, .system_features_control = smu_v11_0_system_features_control,
drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h +2 -0
@@ -259,6 +259,7 @@ struct smu_table_contextstruct smu_bios_boot_up_values boot_values;void *driver_pptable;struct smu_table *tables; +struct smu_table *driver_table;struct smu_table memory_pool;uint8_t thermal_controller_type;
@@ -497,6 +498,7 @@ struct pptable_funcs {int (*set_gfx_cgpg)(struct smu_context *smu,bool enable);int (*write_pptable)(struct smu_context *smu);int (*set_min_dcef_deep_sleep)(struct smu_context *smu); +int (*set_driver_table_location)(struct smu_context *smu);int (*set_tool_table_location)(struct smu_context *smu);int (*notify_memory_pool_location)(struct smu_context *smu);int (*set_last_dcef_min_deep_sleep_clk)(struct smu_context *smu);
drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h +2 -0
@@ -170,6 +170,8 @@ int smu_v11_0_write_pptable(struct smu_context *smu);int smu_v11_0_set_min_dcef_deep_sleep (struct smu_context *smu); +int smu_v11_0_set_driver_table_location (struct smu_context *smu); +int smu_v11_0_set_tool_table_location (struct smu_context *smu);int smu_v11_0_notify_memory_pool_location (struct smu_context *smu);
drivers/gpu/drm/amd/powerplay/inc/smu_v12_0.h +2 -0
@@ -90,4 +90,6 @@ int smu_v12_0_mode2_reset(struct smu_context *smu);int smu_v12_0_set_soft_freq_limited_range (struct smu_context *smu,enum smu_clk_type clk_type,uint32_t min,uint32_t max); +int smu_v12_0_set_driver_table_location (struct smu_context *smu); +#endif
drivers/gpu/drm/amd/powerplay/navi10_ppt.c +1 -0
@@ -2116,6 +2116,7 @@ static const struct pptable_funcs navi10_ppt_funcs = { .check_fw_version = smu_v11_0_check_fw_version, .write_pptable = smu_v11_0_write_pptable, .set_min_dcef_deep_sleep = smu_v11_0_set_min_dcef_deep_sleep, + .set_driver_table_location = smu_v11_0_set_driver_table_location, .set_tool_table_location = smu_v11_0_set_tool_table_location, .notify_memory_pool_location = smu_v11_0_notify_memory_pool_location, .system_features_control = smu_v11_0_system_features_control,
drivers/gpu/drm/amd/powerplay/renoir_ppt.c +1 -0
@@ -916,6 +916,7 @@ static const struct pptable_funcs renoir_ppt_funcs = { .get_dpm_ultimate_freq = smu_v12_0_get_dpm_ultimate_freq, .mode2_reset = smu_v12_0_mode2_reset, .set_soft_freq_limited_range = smu_v12_0_set_soft_freq_limited_range, + .set_driver_table_location = smu_v12_0_set_driver_table_location, };void renoir_set_ppt_funcs (struct smu_context *smu)
drivers/gpu/drm/amd/powerplay/smu_internal.h +2 -0
@@ -61,6 +61,8 @@ ((smu)->ppt_funcs->write_pptable ? (smu)->ppt_funcs->write_pptable((smu)) : 0)#define smu_set_min_dcef_deep_sleep (smu) \ ((smu)->ppt_funcs->set_min_dcef_deep_sleep ? (smu)->ppt_funcs->set_min_dcef_deep_sleep((smu)) : 0) +#define smu_set_driver_table_location (smu) \ + ((smu)->ppt_funcs->set_driver_table_location ? (smu)->ppt_funcs->set_driver_table_location((smu)) : 0)#define smu_set_tool_table_location (smu) \ ((smu)->ppt_funcs->set_tool_table_location ? (smu)->ppt_funcs->set_tool_table_location((smu)) : 0)#define smu_notify_memory_pool_location (smu) \
drivers/gpu/drm/amd/powerplay/smu_v11_0.c +18 -0
@@ -774,6 +774,24 @@ int smu_v11_0_set_min_dcef_deep_sleep(struct smu_context *smu)return smu_v11_0_set_deep_sleep_dcefclk (smu, table_context->boot_values.dcefclk /100 ); } +int smu_v11_0_set_driver_table_location (struct smu_context *smu) +{ +struct smu_table *driver_table = smu->smu_table.driver_table; +int ret =0 ; + +if (driver_table->mc_address) { + ret =smu_send_smc_msg_with_param (smu, + SMU_MSG_SetDriverDramAddrHigh, +upper_32_bits (driver_table->mc_address)); +if (!ret) + ret =smu_send_smc_msg_with_param (smu, + SMU_MSG_SetDriverDramAddrLow, +lower_32_bits (driver_table->mc_address)); + } + +return ret; +} +int smu_v11_0_set_tool_table_location (struct smu_context *smu) {int ret =0 ;
drivers/gpu/drm/amd/powerplay/smu_v12_0.c +18 -0
@@ -514,3 +514,21 @@ int smu_v12_0_set_soft_freq_limited_range(struct smu_context *smu, enum smu_clk_return ret; } + +int smu_v12_0_set_driver_table_location (struct smu_context *smu) +{ +struct smu_table *driver_table = smu->smu_table.driver_table; +int ret =0 ; + +if (driver_table->mc_address) { + ret =smu_send_smc_msg_with_param (smu, + SMU_MSG_SetDriverDramAddrHigh, +upper_32_bits (driver_table->mc_address)); +if (!ret) + ret =smu_send_smc_msg_with_param (smu, + SMU_MSG_SetDriverDramAddrLow, +lower_32_bits (driver_table->mc_address)); + } + +return ret; +}
drivers/gpu/drm/amd/powerplay/vega20_ppt.c +1 -0
@@ -3232,6 +3232,7 @@ static const struct pptable_funcs vega20_ppt_funcs = { .check_fw_version = smu_v11_0_check_fw_version, .write_pptable = smu_v11_0_write_pptable, .set_min_dcef_deep_sleep = smu_v11_0_set_min_dcef_deep_sleep, + .set_driver_table_location = smu_v11_0_set_driver_table_location, .set_tool_table_location = smu_v11_0_set_tool_table_location, .notify_memory_pool_location = smu_v11_0_notify_memory_pool_location, .system_features_control = smu_v11_0_system_features_control,
Signature · 6 lines
2.24.0 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 2/2 drm/amd/powerplay: allocate one piece of VRAM for all tables transferring
Simplify the table transferring between driver and SMU and use less VRAM.
--- drivers/gpu/drm/amd/powerplay/amdgpu_smu.c | 101 ++++++++++-------- drivers/gpu/drm/amd/powerplay/arcturus_ppt.c | 2 +- .../gpu/drm/amd/powerplay/inc/amdgpu_smu.h | 3 +- drivers/gpu/drm/amd/powerplay/navi10_ppt.c | 4 + drivers/gpu/drm/amd/powerplay/renoir_ppt.c | 4 + drivers/gpu/drm/amd/powerplay/smu_v11_0.c | 4 +- drivers/gpu/drm/amd/powerplay/smu_v12_0.c | 10 +- drivers/gpu/drm/amd/powerplay/vega20_ppt.c | 4 + 8 files changed, 73 insertions(+), 59 deletions(-)
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +54 -47
@@ -490,7 +490,7 @@ int smu_update_table(struct smu_context *smu, enum smu_table_id table_index, int {struct smu_table_context *smu_table = &smu->smu_table;struct amdgpu_device *adev = smu->adev; -struct smu_table *table = smu_table->driver_table; +struct smu_table *table = &smu_table->driver_table;int table_id =smu_table_get_index (smu, table_index);uint32_t table_size;int ret =0 ;
@@ -941,24 +941,26 @@ static int smu_init_fb_allocations(struct smu_context *smu)struct amdgpu_device *adev = smu->adev;struct smu_table_context *smu_table = &smu->smu_table;struct smu_table *tables = smu_table->tables; -struct smu_table **driver_table = &(smu_table->driver_table); +struct smu_table *driver_table = &(smu_table->driver_table);uint32_t max_table_size =0 ; -int ret, i, index =0 ; +int ret, i; -for (i =0 ; i <SMU_TABLE_COUNT ; i++) { -if (tables[i].size ==0 ) -continue ; +/* VRAM allocation for tool table */ +if (tables[SMU_TABLE_PMSTATUSLOG ].size) { ret =amdgpu_bo_create_kernel (adev, - tables[i].size, - tables[i].align, - tables[i].domain, - &tables[i].bo, - &tables[i].mc_address, - &tables[i].cpu_addr); -if (ret) - goto failed; + tables[SMU_TABLE_PMSTATUSLOG ].size, + tables[SMU_TABLE_PMSTATUSLOG ].align, + tables[SMU_TABLE_PMSTATUSLOG ].domain, + &tables[SMU_TABLE_PMSTATUSLOG ].bo, + &tables[SMU_TABLE_PMSTATUSLOG ].mc_address, + &tables[SMU_TABLE_PMSTATUSLOG ].cpu_addr); +if (ret) { +pr_err ("VRAM allocation for tool table failed!\n" ); +return ret; + } } +/* VRAM allocation for driver table */ for (i =0 ; i <SMU_TABLE_COUNT ; i++) {if (tables[i].size ==0 )continue ;
@@ -966,24 +968,29 @@ static int smu_init_fb_allocations(struct smu_context *smu)if (i ==SMU_TABLE_PMSTATUSLOG )continue ; -if (max_table_size < tables[i].size) { +if (max_table_size < tables[i].size) max_table_size = tables[i].size; - index = i; - } } - *driver_table = &tables[index]; - -return 0 ; -failed: -while (--i >=0 ) { -if (tables[i].size ==0 ) -continue ; -amdgpu_bo_free_kernel (&tables[i].bo, - &tables[i].mc_address, - &tables[i].cpu_addr); - + driver_table->size = max_table_size; + driver_table->align =PAGE_SIZE ; + driver_table->domain =AMDGPU_GEM_DOMAIN_VRAM ; + + ret =amdgpu_bo_create_kernel (adev, + driver_table->size, + driver_table->align, + driver_table->domain, + &driver_table->bo, + &driver_table->mc_address, + &driver_table->cpu_addr); +if (ret) { +pr_err ("VRAM allocation for driver table failed!\n" ); +if (tables[SMU_TABLE_PMSTATUSLOG ].mc_address) +amdgpu_bo_free_kernel (&tables[SMU_TABLE_PMSTATUSLOG ].bo, + &tables[SMU_TABLE_PMSTATUSLOG ].mc_address, + &tables[SMU_TABLE_PMSTATUSLOG ].cpu_addr); } +return ret; }
@@ -991,18 +998,19 @@ static int smu_fini_fb_allocations(struct smu_context *smu) {struct smu_table_context *smu_table = &smu->smu_table;struct smu_table *tables = smu_table->tables; -uint32_t i =0 ; +struct smu_table *driver_table = &(smu_table->driver_table);if (!tables)return 0 ; -for (i =0 ; i <SMU_TABLE_COUNT ; i++) { -if (tables[i].size ==0 ) -continue ; -amdgpu_bo_free_kernel (&tables[i].bo, - &tables[i].mc_address, - &tables[i].cpu_addr); - } +if (tables[SMU_TABLE_PMSTATUSLOG ].mc_address) +amdgpu_bo_free_kernel (&tables[SMU_TABLE_PMSTATUSLOG ].bo, + &tables[SMU_TABLE_PMSTATUSLOG ].mc_address, + &tables[SMU_TABLE_PMSTATUSLOG ].cpu_addr); + +amdgpu_bo_free_kernel (&driver_table->bo, + &driver_table->mc_address, + &driver_table->cpu_addr);return 0 ; }
@@ -1913,26 +1921,25 @@ int smu_set_df_cstate(struct smu_context *smu,int smu_write_watermarks_table (struct smu_context *smu) { -int ret =0 ; -struct smu_table_context *smu_table = &smu->smu_table; -struct smu_table *table =NULL ; +void *watermarks_table = smu->smu_table.watermarks_table; - table = &smu_table->tables[SMU_TABLE_WATERMARKS ]; - -if (!table->cpu_addr) +if (!watermarks_table)return -EINVAL ; - ret =smu_update_table (smu,SMU_TABLE_WATERMARKS ,0 , table->cpu_addr, +return smu_update_table (smu, +SMU_TABLE_WATERMARKS , +0 , + watermarks_table, true); - -return ret; }int smu_set_watermarks_for_clock_ranges (struct smu_context *smu,struct dm_pp_wm_sets_with_clock_ranges_soc15 *clock_ranges) { -struct smu_table *watermarks = &smu->smu_table.tables[SMU_TABLE_WATERMARKS ]; -void *table = watermarks->cpu_addr; +void *table = smu->smu_table.watermarks_table; + +if (!table) +return -EINVAL ;mutex_lock (&smu->mutex);
drivers/gpu/drm/amd/powerplay/arcturus_ppt.c +1 -1
@@ -2022,7 +2022,7 @@ static int arcturus_i2c_eeprom_read_data(struct i2c_adapter *control,SwI2cRequest_t req;struct amdgpu_device *adev =to_amdgpu_device (control);struct smu_table_context *smu_table = &adev->smu.smu_table; -struct smu_table *table = &smu_table->tables[SMU_TABLE_I2C_COMMANDS ]; +struct smu_table *table = &smu_table->driver_table;memset (&req,0 ,sizeof (req));arcturus_fill_eeprom_i2c_req (&req, false, address, numbytes, data);
drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h +2 -1
@@ -254,12 +254,13 @@ struct smu_table_contextunsigned long metrics_time;void *metrics_table;void *clocks_table; +void *watermarks_table;void *max_sustainable_clocks;struct smu_bios_boot_up_values boot_values;void *driver_pptable;struct smu_table *tables; -struct smu_table *driver_table; +struct smu_table driver_table;struct smu_table memory_pool;uint8_t thermal_controller_type;
drivers/gpu/drm/amd/powerplay/navi10_ppt.c +4 -0
@@ -555,6 +555,10 @@ static int navi10_tables_init(struct smu_context *smu, struct smu_table *tables)return -ENOMEM ; smu_table->metrics_time =0 ; + smu_table->watermarks_table =kzalloc (sizeof (Watermarks_t),GFP_KERNEL ); +if (!smu_table->watermarks_table) +return -ENOMEM ; +return 0 ; }
drivers/gpu/drm/amd/powerplay/renoir_ppt.c +4 -0
@@ -209,6 +209,10 @@ static int renoir_tables_init(struct smu_context *smu, struct smu_table *tables)return -ENOMEM ; smu_table->metrics_time =0 ; + smu_table->watermarks_table =kzalloc (sizeof (Watermarks_t),GFP_KERNEL ); +if (!smu_table->watermarks_table) +return -ENOMEM ; +return 0 ; }
drivers/gpu/drm/amd/powerplay/smu_v11_0.c +3 -1
@@ -450,8 +450,10 @@ int smu_v11_0_fini_smc_tables(struct smu_context *smu)kfree (smu_table->tables);kfree (smu_table->metrics_table); +kfree (smu_table->watermarks_table); smu_table->tables =NULL ; smu_table->metrics_table =NULL ; + smu_table->watermarks_table =NULL ; smu_table->metrics_time =0 ; ret =smu_v11_0_fini_dpm_context (smu);
@@ -776,7 +778,7 @@ int smu_v11_0_set_min_dcef_deep_sleep(struct smu_context *smu)int smu_v11_0_set_driver_table_location (struct smu_context *smu) { -struct smu_table *driver_table = smu->smu_table.driver_table; +struct smu_table *driver_table = &smu->smu_table.driver_table;int ret =0 ;if (driver_table->mc_address) {
drivers/gpu/drm/amd/powerplay/smu_v12_0.c +1 -9
@@ -318,14 +318,6 @@ int smu_v12_0_fini_smc_tables(struct smu_context *smu)int smu_v12_0_populate_smc_tables (struct smu_context *smu) {struct smu_table_context *smu_table = &smu->smu_table; -struct smu_table *table =NULL ; - - table = &smu_table->tables[SMU_TABLE_DPMCLOCKS ]; -if (!table) -return -EINVAL ; - -if (!table->cpu_addr) -return -EINVAL ;return smu_update_table (smu,SMU_TABLE_DPMCLOCKS ,0 , smu_table->clocks_table, false); }
@@ -517,7 +509,7 @@ int smu_v12_0_set_soft_freq_limited_range(struct smu_context *smu, enum smu_clk_int smu_v12_0_set_driver_table_location (struct smu_context *smu) { -struct smu_table *driver_table = smu->smu_table.driver_table; +struct smu_table *driver_table = &smu->smu_table.driver_table;int ret =0 ;if (driver_table->mc_address) {
drivers/gpu/drm/amd/powerplay/vega20_ppt.c +4 -0
@@ -338,6 +338,10 @@ static int vega20_tables_init(struct smu_context *smu, struct smu_table *tables)return -ENOMEM ; smu_table->metrics_time =0 ; + smu_table->watermarks_table =kzalloc (sizeof (Watermarks_t),GFP_KERNEL ); +if (!smu_table->watermarks_table) +return -ENOMEM ; +return 0 ; }
Signature · 6 lines
2.24.0 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Re: [PATCH 2/2] drm/amd/powerplay: allocate one piece of VRAM for all tables transferring
Quoted 102 lines · Evan Quan
Simplify the table transferring between driver and SMU and use less VRAM.
Change-Id I3f9b54fd9b8c0bcaeb533fc1a07bb06050fefbd8Signed-off-by Evan Quan <[email protected]>--- drivers/gpu/drm/amd/powerplay/amdgpu_smu.c | 101 ++++++++++-------- drivers/gpu/drm/amd/powerplay/arcturus_ppt.c | 2 +- .../gpu/drm/amd/powerplay/inc/amdgpu_smu.h | 3 +- drivers/gpu/drm/amd/powerplay/navi10_ppt.c | 4 + drivers/gpu/drm/amd/powerplay/renoir_ppt.c | 4 + drivers/gpu/drm/amd/powerplay/smu_v11_0.c | 4 +- drivers/gpu/drm/amd/powerplay/smu_v12_0.c | 10 +- drivers/gpu/drm/amd/powerplay/vega20_ppt.c | 4 + 8 files changed, 73 insertions(+), 59 deletions(-)drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +27
-28@@ -490,7 +490,7 @@ int smu_update_table(struct smu_context *smu, enum smu_table_id table_index, int {struct smu_table_context *smu_table = &smu->smu_table;struct amdgpu_device *adev = smu->adev; -struct smu_table *table = smu_table->driver_table; +struct smu_table *table = &smu_table->driver_table;int table_id =smu_table_get_index (smu, table_index);uint32_t table_size;int ret =0 ;@@ -941,24 +941,26 @@ static int smu_init_fb_allocations(struct smu_context *smu)struct amdgpu_device *adev = smu->adev;struct smu_table_context *smu_table = &smu->smu_table;struct smu_table *tables = smu_table->tables; -struct smu_table **driver_table = &(smu_table->driver_table); +struct smu_table *driver_table = &(smu_table->driver_table);uint32_t max_table_size =0 ; -int ret, i, index =0 ; +int ret, i; -for (i =0 ; i <SMU_TABLE_COUNT ; i++) { -if (tables[i].size ==0 ) -continue ; +/* VRAM allocation for tool table */ +if (tables[SMU_TABLE_PMSTATUSLOG ].size) { ret =amdgpu_bo_create_kernel (adev, - tables[i].size, - tables[i].align, - tables[i].domain, - &tables[i].bo, - &tables[i].mc_address, - &tables[i].cpu_addr); -if (ret) - goto failed; + tables[SMU_TABLE_PMSTATUSLOG ].size, + tables[SMU_TABLE_PMSTATUSLOG ].align, + tables[SMU_TABLE_PMSTATUSLOG ].domain, + &tables[SMU_TABLE_PMSTATUSLOG ].bo, + &tables[SMU_TABLE_PMSTATUSLOG ].mc_address, + &tables[SMU_TABLE_PMSTATUSLOG ].cpu_addr); +if (ret) { +pr_err ("VRAM allocation for tool table failed!\n" ); +return ret; + } } +/* VRAM allocation for driver table */ for (i =0 ; i <SMU_TABLE_COUNT ; i++) {if (tables[i].size ==0 )continue ;@@ -966,24 +968,29 @@ static int smu_init_fb_allocations(struct smu_context *smu)if (i ==SMU_TABLE_PMSTATUSLOG )continue ; -if (max_table_size < tables[i].size) { +if (max_table_size < tables[i].size) max_table_size = tables[i].size; - index = i; - } } - *driver_table = &tables[index]; - -return 0 ; -failed: -while (--i >=0 ) { -if (tables[i].size ==0 ) -continue ; -amdgpu_bo_free_kernel (&tables[i].bo, - &tables[i].mc_address, - &tables[i].cpu_addr); - + driver_table->size = max_table_size; + driver_table->align =PAGE_SIZE ; + driver_table->domain =AMDGPU_GEM_DOMAIN_VRAM ; + + ret = amdgpu_bo_create_kernel(adev, + driver_table->size, + driver_table->align, + driver_table->domain, + &driver_table->bo, + &driver_table->mc_address,
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c static int smu_init_fb_allocations(struct smu_context *smu) · line 994
Quoted 10 lines
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +8
-0+ &driver_table->cpu_addr); +if (ret) { +pr_err ("VRAM allocation for driver table failed!\n" ); +if (tables[SMU_TABLE_PMSTATUSLOG ].mc_address) +amdgpu_bo_free_kernel (&tables[SMU_TABLE_PMSTATUSLOG ].bo, + &tables[SMU_TABLE_PMSTATUSLOG ].mc_address, + &tables[SMU_TABLE_PMSTATUSLOG ].cpu_addr); } +return ret;
Shouldn't this change be merged into the previous patch? Otherwise, we'll break I think.
Quoted 85 lines
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +19
-19}@@ -991,18 +998,19 @@ static int smu_fini_fb_allocations(struct smu_context *smu) {struct smu_table_context *smu_table = &smu->smu_table;struct smu_table *tables = smu_table->tables; -uint32_t i =0 ; +struct smu_table *driver_table = &(smu_table->driver_table);if (!tables)return 0 ; -for (i =0 ; i <SMU_TABLE_COUNT ; i++) { -if (tables[i].size ==0 ) -continue ; -amdgpu_bo_free_kernel (&tables[i].bo, - &tables[i].mc_address, - &tables[i].cpu_addr); - } +if (tables[SMU_TABLE_PMSTATUSLOG ].mc_address) +amdgpu_bo_free_kernel (&tables[SMU_TABLE_PMSTATUSLOG ].bo, + &tables[SMU_TABLE_PMSTATUSLOG ].mc_address, + &tables[SMU_TABLE_PMSTATUSLOG ].cpu_addr); + +amdgpu_bo_free_kernel (&driver_table->bo, + &driver_table->mc_address, + &driver_table->cpu_addr);return 0 ; }@@ -1913,26 +1921,25 @@ int smu_set_df_cstate(struct smu_context *smu,int smu_write_watermarks_table (struct smu_context *smu) { -int ret =0 ; -struct smu_table_context *smu_table = &smu->smu_table; -struct smu_table *table =NULL ; +void *watermarks_table = smu->smu_table.watermarks_table; - table = &smu_table->tables[SMU_TABLE_WATERMARKS ]; - -if (!table->cpu_addr) +if (!watermarks_table)return -EINVAL ; - ret =smu_update_table (smu,SMU_TABLE_WATERMARKS ,0 , table->cpu_addr, +return smu_update_table (smu, +SMU_TABLE_WATERMARKS , +0 , + watermarks_table, true); - -return ret; }int smu_set_watermarks_for_clock_ranges (struct smu_context *smu,struct dm_pp_wm_sets_with_clock_ranges_soc15 *clock_ranges) { -struct smu_table *watermarks = &smu->smu_table.tables[SMU_TABLE_WATERMARKS ]; -void *table = watermarks->cpu_addr; +void *table = smu->smu_table.watermarks_table; + +if (!table) +return -EINVAL ;mutex_lock (&smu->mutex);drivers/gpu/drm/amd/powerplay/arcturus_ppt.c +1
-1@@ -2022,7 +2022,7 @@ static int arcturus_i2c_eeprom_read_data(struct i2c_adapter *control,SwI2cRequest_t req;struct amdgpu_device *adev =to_amdgpu_device (control);struct smu_table_context *smu_table = &adev->smu.smu_table; -struct smu_table *table = &smu_table->tables[SMU_TABLE_I2C_COMMANDS ]; +struct smu_table *table = &smu_table->driver_table;memset (&req,0 ,sizeof (req));arcturus_fill_eeprom_i2c_req (&req, false, address, numbytes, data);drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h +0
-0
drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h struct smu_table_context · line 257
Quoted 4 lines
drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h +1
-0@@ -254,12 +254,13 @@ struct smu_table_contextunsigned long metrics_time;void *metrics_table;void *clocks_table; +void *watermarks_table;
Can you split out the watermarks change as a separate patch or at least explain why you are changing how it's handled?
Alex
Quoted 110 lines
drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h +1
-1void *max_sustainable_clocks;struct smu_bios_boot_up_values boot_values;void *driver_pptable;struct smu_table *tables; -struct smu_table *driver_table; +struct smu_table driver_table;struct smu_table memory_pool;uint8_t thermal_controller_type;drivers/gpu/drm/amd/powerplay/navi10_ppt.c +4
-0@@ -555,6 +555,10 @@ static int navi10_tables_init(struct smu_context *smu, struct smu_table *tables)return -ENOMEM ; smu_table->metrics_time =0 ; + smu_table->watermarks_table =kzalloc (sizeof (Watermarks_t),GFP_KERNEL ); +if (!smu_table->watermarks_table) +return -ENOMEM ; +return 0 ; }drivers/gpu/drm/amd/powerplay/renoir_ppt.c +4
-0@@ -209,6 +209,10 @@ static int renoir_tables_init(struct smu_context *smu, struct smu_table *tables)return -ENOMEM ; smu_table->metrics_time =0 ; + smu_table->watermarks_table =kzalloc (sizeof (Watermarks_t),GFP_KERNEL ); +if (!smu_table->watermarks_table) +return -ENOMEM ; +return 0 ; }drivers/gpu/drm/amd/powerplay/smu_v11_0.c +3
-1@@ -450,8 +450,10 @@ int smu_v11_0_fini_smc_tables(struct smu_context *smu)kfree (smu_table->tables);kfree (smu_table->metrics_table); +kfree (smu_table->watermarks_table); smu_table->tables =NULL ; smu_table->metrics_table =NULL ; + smu_table->watermarks_table =NULL ; smu_table->metrics_time =0 ; ret =smu_v11_0_fini_dpm_context (smu);@@ -776,7 +778,7 @@ int smu_v11_0_set_min_dcef_deep_sleep(struct smu_context *smu)int smu_v11_0_set_driver_table_location (struct smu_context *smu) { -struct smu_table *driver_table = smu->smu_table.driver_table; +struct smu_table *driver_table = &smu->smu_table.driver_table;int ret =0 ;if (driver_table->mc_address) {drivers/gpu/drm/amd/powerplay/smu_v12_0.c +1
-9@@ -318,14 +318,6 @@ int smu_v12_0_fini_smc_tables(struct smu_context *smu)int smu_v12_0_populate_smc_tables (struct smu_context *smu) {struct smu_table_context *smu_table = &smu->smu_table; -struct smu_table *table =NULL ; - - table = &smu_table->tables[SMU_TABLE_DPMCLOCKS ]; -if (!table) -return -EINVAL ; - -if (!table->cpu_addr) -return -EINVAL ;return smu_update_table (smu,SMU_TABLE_DPMCLOCKS ,0 , smu_table->clocks_table, false); }@@ -517,7 +509,7 @@ int smu_v12_0_set_soft_freq_limited_range(struct smu_context *smu, enum smu_clk_int smu_v12_0_set_driver_table_location (struct smu_context *smu) { -struct smu_table *driver_table = smu->smu_table.driver_table; +struct smu_table *driver_table = &smu->smu_table.driver_table;int ret =0 ;if (driver_table->mc_address) {drivers/gpu/drm/amd/powerplay/vega20_ppt.c +4
-1@@ -338,6 +338,10 @@ static int vega20_tables_init(struct smu_context *smu, struct smu_table *tables)return -ENOMEM ; smu_table->metrics_time =0 ; + smu_table->watermarks_table =kzalloc (sizeof (Watermarks_t),GFP_KERNEL ); +if (!smu_table->watermarks_table) +return -ENOMEM ; +return 0 ; } --2.24.0
Quoted 4 lines
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
RE: [PATCH 2/2] drm/amd/powerplay: allocate one piece of VRAM for all tables transferring
Hi Alex,
Please check the new patch sets.
https://lists.freedesktop.org/archives/amd-gfx/2019-December/044352.html
https://lists.freedesktop.org/archives/amd-gfx/2019-December/044353.html
Regards,
Evan
Quoted 364 lines
-----Original Message-----
From: Alex Deucher <[email protected]>
Sent: Monday, December 30, 2019 11:22 PM
To: Quan, Evan <[email protected]>
Cc: amd-gfx list <[email protected]>
Subject: Re: [PATCH 2/2] drm/amd/powerplay: allocate one piece of VRAM for all tables transferringdrivers/gpu/drm/amd/powerplay/amdgpu_smu.c int smu_update_table(struct smu_context *smu,
Quoted 23 lines · Evan Quan
Simplify the table transferring between driver and SMU and use less VRAM.
Change-Id I3f9b54fd9b8c0bcaeb533fc1a07bb06050fefbd8Signed-off-by Evan Quan <[email protected]>--- drivers/gpu/drm/amd/powerplay/amdgpu_smu.c | 101 ++++++++++-------- drivers/gpu/drm/amd/powerplay/arcturus_ppt.c | 2 +- .../gpu/drm/amd/powerplay/inc/amdgpu_smu.h | 3 +- drivers/gpu/drm/amd/powerplay/navi10_ppt.c | 4 + drivers/gpu/drm/amd/powerplay/renoir_ppt.c | 4 + drivers/gpu/drm/amd/powerplay/smu_v11_0.c | 4 +- drivers/gpu/drm/amd/powerplay/smu_v12_0.c | 10 +- drivers/gpu/drm/amd/powerplay/vega20_ppt.c | 4 + 8 files changed, 73 insertions(+), 59 deletions(-)drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +0
-0@@ -490,7 +490,7 @@ int smu_update_table(struct smu_context *smu,enum
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c static int smu_init_fb_allocations(struct
Quoted 9 lines
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +1
-1smu_table_id table_index, int {struct smu_table_context *smu_table = &smu->smu_table;struct amdgpu_device *adev = smu->adev; -struct smu_table *table = smu_table->driver_table; +struct smu_table *table = &smu_table->driver_table;int table_id =smu_table_get_index (smu, table_index);uint32_t table_size;int ret =0 ;@@ -941,24 +941,26 @@ static int smu_init_fb_allocations(struct
smu_context*smu)Quoted 18 lines
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +4
-8struct amdgpu_device *adev = smu->adev;struct smu_table_context *smu_table = &smu->smu_table;struct smu_table *tables = smu_table->tables; -struct smu_table **driver_table = &(smu_table->driver_table); +struct smu_table *driver_table = &(smu_table->driver_table);uint32_t max_table_size =0 ; -int ret, i, index =0 ; +int ret, i; -for (i =0 ; i <SMU_TABLE_COUNT ; i++) { -if (tables[i].size ==0 ) -continue ; +/* VRAM allocation for tool table */ +if (tables[SMU_TABLE_PMSTATUSLOG ].size) { ret = amdgpu_bo_create_kernel(adev, - tables[i].size, - tables[i].align, - tables[i].domain,drivers/gpu/drm/amd/powerplay/amdgpu_smu.c static int smu_init_fb_allocations(struct smu_context *smu)
Quoted 10 lines
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +5
-5- &tables[i].bo, - &tables[i].mc_address, - &tables[i].cpu_addr); -if (ret) - goto failed; + tables[SMU_TABLE_PMSTATUSLOG ].size, + tables[SMU_TABLE_PMSTATUSLOG ].align, + tables[SMU_TABLE_PMSTATUSLOG ].domain, + &tables[SMU_TABLE_PMSTATUSLOG].bo, +&tables[SMU_TABLE_PMSTATUSLOG].mc_address,
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c static int smu_init_fb_allocations(struct
Quoted 12 lines
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +6
-0+ &tables[SMU_TABLE_PMSTATUSLOG].cpu_addr); +if (ret) { +pr_err ("VRAM allocation for tool table failed!\n" ); +return ret; + } } +/* VRAM allocation for driver table */ for (i =0 ; i <SMU_TABLE_COUNT ; i++) {if (tables[i].size ==0 )continue ;@@ -966,24 +968,29 @@ static int smu_init_fb_allocations(struct
smu_context*smu)Quoted 26 lines
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +6
-14if (i ==SMU_TABLE_PMSTATUSLOG )continue ; -if (max_table_size < tables[i].size) { +if (max_table_size < tables[i].size) max_table_size = tables[i].size; - index = i; - } } - *driver_table = &tables[index]; - -return 0 ; -failed: -while (--i >=0 ) { -if (tables[i].size ==0 ) -continue ; -amdgpu_bo_free_kernel (&tables[i].bo, - &tables[i].mc_address, - &tables[i].cpu_addr); - + driver_table->size = max_table_size; + driver_table->align =PAGE_SIZE ; + driver_table->domain =AMDGPU_GEM_DOMAIN_VRAM ; + + ret = amdgpu_bo_create_kernel(adev,drivers/gpu/drm/amd/powerplay/amdgpu_smu.c static int smu_init_fb_allocations(struct smu_context *smu) · line 993
Quoted 10 lines
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +10
-0+ driver_table->size, + driver_table->align, + driver_table->domain, + &driver_table->bo, + &driver_table->mc_address, + &driver_table->cpu_addr); +if (ret) { +pr_err ("VRAM allocation for driver table failed!\n" ); +if (tables[SMU_TABLE_PMSTATUSLOG ].mc_address) +amdgpu_bo_free_kernel(&tables[SMU_TABLE_PMSTATUSLOG].bo,
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c static int smu_init_fb_allocations(struct · line 990
Quoted 1 line
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +1
-0+&tables[SMU_TABLE_PMSTATUSLOG].mc_address,
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c static int smu_init_fb_allocations(struct smu_context *smu) · line 994
Quoted 5 lines
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +3
-0+ + &tables[SMU_TABLE_PMSTATUSLOG].cpu_addr); } +return ret;Shouldn't this change be merged into the previous patch? Otherwise, we'll break I think.
Review
Quoted 70 lines
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +20
-19}@@ -991,18 +998,19 @@ static int smu_fini_fb_allocations(struct smu_context *smu) {struct smu_table_context *smu_table = &smu->smu_table;struct smu_table *tables = smu_table->tables; -uint32_t i =0 ; +struct smu_table *driver_table = &(smu_table->driver_table);if (!tables)return 0 ; -for (i =0 ; i <SMU_TABLE_COUNT ; i++) { -if (tables[i].size ==0 ) -continue ; -amdgpu_bo_free_kernel (&tables[i].bo, - &tables[i].mc_address, - &tables[i].cpu_addr); - } +if (tables[SMU_TABLE_PMSTATUSLOG ].mc_address) +amdgpu_bo_free_kernel (&tables[SMU_TABLE_PMSTATUSLOG ].bo, + &tables[SMU_TABLE_PMSTATUSLOG ].mc_address, + + &tables[SMU_TABLE_PMSTATUSLOG ].cpu_addr); + +amdgpu_bo_free_kernel (&driver_table->bo, + &driver_table->mc_address, + &driver_table->cpu_addr);return 0 ; }@@ -1913,26 +1921,25 @@ int smu_set_df_cstate(struct smu_context *smu,int smu_write_watermarks_table (struct smu_context *smu) { -int ret =0 ; -struct smu_table_context *smu_table = &smu->smu_table; -struct smu_table *table =NULL ; +void *watermarks_table = smu->smu_table.watermarks_table; - table = &smu_table->tables[SMU_TABLE_WATERMARKS ]; - -if (!table->cpu_addr) +if (!watermarks_table)return -EINVAL ; - ret =smu_update_table (smu,SMU_TABLE_WATERMARKS ,0 , table- cpu_addr, + return smu_update_table(smu, +SMU_TABLE_WATERMARKS , +0 , + watermarks_table, true); - -return ret; }int smu_set_watermarks_for_clock_ranges (struct smu_context *smu,struct dm_pp_wm_sets_with_clock_ranges_soc15 *clock_ranges) { -struct smu_table *watermarks = &smu- smu_table.tables[SMU_TABLE_WATERMARKS ]; -void *table = watermarks->cpu_addr; +void *table = smu->smu_table.watermarks_table; + +if (!table) +return -EINVAL ;mutex_lock (&smu->mutex);(quoted diff) +0
-0
b/drivers/gpu/drm/amd/powerplay/arcturus_ppt.cdrivers/gpu/drm/amd/powerplay/arcturus_ppt.c static int arcturus_i2c_eeprom_read_data(struct
Quoted 4 lines
index
e89965e5fdcb..064b5201a8a7100644drivers/gpu/drm/amd/powerplay/arcturus_ppt.c +0
-0@@ -2022,7 +2022,7 @@ static int arcturus_i2c_eeprom_read_data(struct
i2c_adapter*control,Review
Quoted 10 lines
drivers/gpu/drm/amd/powerplay/arcturus_ppt.c +1
-1SwI2cRequest_t req;struct amdgpu_device *adev =to_amdgpu_device (control);struct smu_table_context *smu_table = &adev->smu.smu_table; -struct smu_table *table = &smu_table- tables[SMU_TABLE_I2C_COMMANDS ]; +struct smu_table *table = &smu_table->driver_table;memset (&req,0 ,sizeof (req));arcturus_fill_eeprom_i2c_req(&req, false, address, numbytes, data);(quoted diff) +0
-0
b/drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.hdrivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h struct smu_table_context · line 257
Quoted 8 lines
index
ed193adc881c..121bf81eced5100644drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h +1
-0@@ -254,12 +254,13 @@ struct smu_table_contextunsigned long metrics_time;void *metrics_table;void *clocks_table; +void *watermarks_table;Can you split out the watermarks change as a separate patch or at least explain why you are changing how it's handled?
Alex
Review
Quoted 11 lines
drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h +1
-1void *max_sustainable_clocks;struct smu_bios_boot_up_values boot_values;void *driver_pptable;struct smu_table *tables; -struct smu_table *driver_table; +struct smu_table driver_table;struct smu_table memory_pool;uint8_t thermal_controller_type;(quoted diff) +0
-0
b/drivers/gpu/drm/amd/powerplay/navi10_ppt.cdrivers/gpu/drm/amd/powerplay/navi10_ppt.c static int navi10_tables_init(struct smu_context
Quoted 4 lines
index
ed147dd51325..aa206bde619b100644drivers/gpu/drm/amd/powerplay/navi10_ppt.c +0
-0@@ -555,6 +555,10 @@ static int navi10_tables_init(struct smu_context*smu, struct
smu_table*tables)drivers/gpu/drm/amd/powerplay/navi10_ppt.c static int navi10_tables_init(struct smu_context *smu, struct smu_table *tables)
Quoted 4 lines
drivers/gpu/drm/amd/powerplay/navi10_ppt.c +1
-0return -ENOMEM ; smu_table->metrics_time =0 ; + smu_table->watermarks_table = kzalloc(sizeof(Watermarks_t),
GFP_KERNEL);Review
Quoted 7 lines
drivers/gpu/drm/amd/powerplay/navi10_ppt.c +3
-0+if (!smu_table->watermarks_table) +return -ENOMEM ; +return 0 ; }(quoted diff) +0
-0
b/drivers/gpu/drm/amd/powerplay/renoir_ppt.cdrivers/gpu/drm/amd/powerplay/renoir_ppt.c static int renoir_tables_init(struct smu_context
Quoted 4 lines
index
c4b5984c86d9..861e6410363b100644drivers/gpu/drm/amd/powerplay/renoir_ppt.c +0
-0@@ -209,6 +209,10 @@ static int renoir_tables_init(struct smu_context*smu, struct
smu_table*tables)drivers/gpu/drm/amd/powerplay/renoir_ppt.c static int renoir_tables_init(struct smu_context · line 212
Quoted 4 lines
drivers/gpu/drm/amd/powerplay/renoir_ppt.c +1
-0return -ENOMEM ; smu_table->metrics_time =0 ; + smu_table->watermarks_table = kzalloc(sizeof(Watermarks_t),
GFP_KERNEL);Review
Quoted 7 lines
drivers/gpu/drm/amd/powerplay/renoir_ppt.c +3
-0+if (!smu_table->watermarks_table) +return -ENOMEM ; +return 0 ; }(quoted diff) +0
-0
b/drivers/gpu/drm/amd/powerplay/smu_v11_0.cdrivers/gpu/drm/amd/powerplay/smu_v11_0.c int smu_v11_0_fini_smc_tables(struct smu_context
Quoted 4 lines
index
79a63edcd7ba..962e97860fe8100644drivers/gpu/drm/amd/powerplay/smu_v11_0.c +0
-0@@ -450,8 +450,10 @@ int smu_v11_0_fini_smc_tables(struct smu_context*smu)
drivers/gpu/drm/amd/powerplay/smu_v11_0.c int smu_v11_0_set_min_dcef_deep_sleep(struct
Quoted 11 lines
drivers/gpu/drm/amd/powerplay/smu_v11_0.c +2
-0kfree (smu_table->tables);kfree (smu_table->metrics_table); +kfree (smu_table->watermarks_table); smu_table->tables =NULL ; smu_table->metrics_table =NULL ; + smu_table->watermarks_table =NULL ; smu_table->metrics_time =0 ; ret =smu_v11_0_fini_dpm_context (smu);@@ -776,7 +778,7 @@ int smu_v11_0_set_min_dcef_deep_sleep(struct
smu_context*smu)Review
Quoted 9 lines
drivers/gpu/drm/amd/powerplay/smu_v11_0.c +1
-1int smu_v11_0_set_driver_table_location (struct smu_context *smu) { -struct smu_table *driver_table = smu->smu_table.driver_table; +struct smu_table *driver_table = &smu->smu_table.driver_table;int ret =0 ;if (driver_table->mc_address) {(quoted diff) +0
-0
b/drivers/gpu/drm/amd/powerplay/smu_v12_0.cdrivers/gpu/drm/amd/powerplay/smu_v12_0.c int smu_v12_0_fini_smc_tables(struct smu_context
Quoted 4 lines
index
cd2be9fb2513..2b1ef9eb0db6100644drivers/gpu/drm/amd/powerplay/smu_v12_0.c +0
-0@@ -318,14 +318,6 @@ int smu_v12_0_fini_smc_tables(struct smu_context*smu)
drivers/gpu/drm/amd/powerplay/smu_v12_0.c int smu_v12_0_set_soft_freq_limited_range(struct
Quoted 16 lines
drivers/gpu/drm/amd/powerplay/smu_v12_0.c +0
-8int smu_v12_0_populate_smc_tables (struct smu_context *smu) {struct smu_table_context *smu_table = &smu->smu_table; -struct smu_table *table =NULL ; - - table = &smu_table->tables[SMU_TABLE_DPMCLOCKS ]; -if (!table) -return -EINVAL ; - -if (!table->cpu_addr) -return -EINVAL ;return smu_update_table (smu,SMU_TABLE_DPMCLOCKS ,0 , smu_table- clocks_table, false);}
drivers/gpu/drm/amd/powerplay/smu_v12_0.c +0
-0@@ -517,7 +509,7 @@ int smu_v12_0_set_soft_freq_limited_range(struct
smu_context*smu, enumsmu_clk_Review
Quoted 9 lines
drivers/gpu/drm/amd/powerplay/smu_v12_0.c +1
-1int smu_v12_0_set_driver_table_location (struct smu_context *smu) { -struct smu_table *driver_table = smu->smu_table.driver_table; +struct smu_table *driver_table = &smu->smu_table.driver_table;int ret =0 ;if (driver_table->mc_address) {(quoted diff) +0
-0
b/drivers/gpu/drm/amd/powerplay/vega20_ppt.cdrivers/gpu/drm/amd/powerplay/vega20_ppt.c static int vega20_tables_init(struct smu_context
Quoted 4 lines
index
12f97956058b..38febd5ca4da100644drivers/gpu/drm/amd/powerplay/vega20_ppt.c +0
-0@@ -338,6 +338,10 @@ static int vega20_tables_init(struct smu_context*smu, struct
smu_table*tables)drivers/gpu/drm/amd/powerplay/vega20_ppt.c static int vega20_tables_init(struct smu_context · line 341
Quoted 4 lines
drivers/gpu/drm/amd/powerplay/vega20_ppt.c +1
-0return -ENOMEM ; smu_table->metrics_time =0 ; + smu_table->watermarks_table = kzalloc(sizeof(Watermarks_t),
GFP_KERNEL);Quoted 9 lines
drivers/gpu/drm/amd/powerplay/vega20_ppt.c +3
-1+if (!smu_table->watermarks_table) +return -ENOMEM ; +return 0 ; } --2.24.0
Quoted 4 lines
_______________________________________________
amd-gfx mailing list
[email protected]
Quoted 6 lines
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.free
desktop.org%2Fmailman%2Flistinfo%2Famd- gfx&data=02%7C01%7Cevan.quan%40amd.com%7Cdd3a4f9d66ba4e7085 ba08d78d3c0c38%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637 133161355995859&sdata=KYYcYqB%2Bd3vWoIHVYVSed8bXtqrfg1dWrlFT0 QMFhRs%3D&reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Quoted 85 lines · Evan Quan
By this, we can avoid to pass in the vram address on every table transferring.
Change-Id Ia8a3dbe923bc562286ab102a5830392a95dcf28cSigned-off-by Evan Quan <[email protected]>--- drivers/gpu/drm/amd/powerplay/amdgpu_smu.c | 42 ++++++++++++------- drivers/gpu/drm/amd/powerplay/arcturus_ppt.c | 1 + .../gpu/drm/amd/powerplay/inc/amdgpu_smu.h | 2 + drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h | 2 + drivers/gpu/drm/amd/powerplay/inc/smu_v12_0.h | 2 + drivers/gpu/drm/amd/powerplay/navi10_ppt.c | 1 + drivers/gpu/drm/amd/powerplay/renoir_ppt.c | 1 + drivers/gpu/drm/amd/powerplay/smu_internal.h | 2 + drivers/gpu/drm/amd/powerplay/smu_v11_0.c | 18 ++++++++ drivers/gpu/drm/amd/powerplay/smu_v12_0.c | 18 ++++++++ drivers/gpu/drm/amd/powerplay/vega20_ppt.c | 1 + 11 files changed, 76 insertions(+), 14 deletions(-)drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +14
-14@@ -490,26 +490,19 @@ int smu_update_table(struct smu_context *smu, enum smu_table_id table_index, int {struct smu_table_context *smu_table = &smu->smu_table;struct amdgpu_device *adev = smu->adev; -struct smu_table *table =NULL ; -int ret =0 ; +struct smu_table *table = smu_table->driver_table;int table_id =smu_table_get_index (smu, table_index); +uint32_t table_size; +int ret =0 ;if (!table_data || table_id >=SMU_TABLE_COUNT || table_id <0 )return -EINVAL ; - table = &smu_table->tables[table_index]; + table_size = smu_table->tables[table_index].size;if (drv2smu) -memcpy (table->cpu_addr, table_data, table->size); +memcpy (table->cpu_addr, table_data, table_size); - ret =smu_send_smc_msg_with_param (smu, SMU_MSG_SetDriverDramAddrHigh, -upper_32_bits (table->mc_address)); -if (ret) -return ret; - ret =smu_send_smc_msg_with_param (smu, SMU_MSG_SetDriverDramAddrLow, -lower_32_bits (table->mc_address)); -if (ret) -return ret; ret = smu_send_smc_msg_with_param(smu, drv2smu ? SMU_MSG_TransferTableDram2Smu : SMU_MSG_TransferTableSmu2Dram,@@ -521,7 +514,7 @@ int smu_update_table(struct smu_context *smu, enum smu_table_id table_index, int adev->nbio.funcs->hdp_flush (adev,NULL );if (!drv2smu) -memcpy (table_data, table->cpu_addr, table->size); +memcpy (table_data, table->cpu_addr, table_size);return ret; }@@ -948,7 +941,9 @@ static int smu_init_fb_allocations(struct smu_context *smu)struct amdgpu_device *adev = smu->adev;struct smu_table_context *smu_table = &smu->smu_table;struct smu_table *tables = smu_table->tables; -int ret, i; +struct smu_table **driver_table = &(smu_table->driver_table); +uint32_t max_table_size =0 ; +int ret, i, index =0 ;for (i =0 ; i <SMU_TABLE_COUNT ; i++) {if (tables[i].size ==0 )@@ -964,6 +959,21 @@ static int smu_init_fb_allocations(struct smu_context *smu) goto failed; } +for (i =0 ; i <SMU_TABLE_COUNT ; i++) { +if (tables[i].size ==0 ) +continue ; + +if (i ==SMU_TABLE_PMSTATUSLOG )
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c static int smu_init_fb_allocations(struct smu_context *smu) · line 976
Quoted 10 lines
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +10
-0+continue ; + +if (max_table_size < tables[i].size) { + max_table_size = tables[i].size; + index = i; + } + } + + *driver_table = &tables[index]; +
Will this work? I don't think we have any guarantee that the memory for each table is contiguous.
Quoted 4 lines
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +0
-0return 0 ; failed:while (--i >=0 ) {
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c static int smu_smc_table_hw_init(struct smu_context *smu, · line 1082
Quoted 10 lines
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +4
-0@@ -1063,6 +1073,10 @@ static int smu_smc_table_hw_init(struct smu_context *smu,/* smu_dump_pptable(smu); */ + ret =smu_set_driver_table_location (smu); +if (ret) +return ret; +/* * Copy pptable bo in the vram to smc with SMU MSGs such as * SetDriverDramAddr and TransferTableDram2Smu.
Fix this comment to match the code changes.
Quoted 163 lines
drivers/gpu/drm/amd/powerplay/arcturus_ppt.c +1
-0@@ -2261,6 +2261,7 @@ static const struct pptable_funcs arcturus_ppt_funcs = { .check_fw_version = smu_v11_0_check_fw_version, .write_pptable = smu_v11_0_write_pptable, .set_min_dcef_deep_sleep = smu_v11_0_set_min_dcef_deep_sleep, + .set_driver_table_location = smu_v11_0_set_driver_table_location, .set_tool_table_location = smu_v11_0_set_tool_table_location, .notify_memory_pool_location = smu_v11_0_notify_memory_pool_location, .system_features_control = smu_v11_0_system_features_control,drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h +2
-0@@ -259,6 +259,7 @@ struct smu_table_contextstruct smu_bios_boot_up_values boot_values;void *driver_pptable;struct smu_table *tables; +struct smu_table *driver_table;struct smu_table memory_pool;uint8_t thermal_controller_type;@@ -497,6 +498,7 @@ struct pptable_funcs {int (*set_gfx_cgpg)(struct smu_context *smu,bool enable);int (*write_pptable)(struct smu_context *smu);int (*set_min_dcef_deep_sleep)(struct smu_context *smu); +int (*set_driver_table_location)(struct smu_context *smu);int (*set_tool_table_location)(struct smu_context *smu);int (*notify_memory_pool_location)(struct smu_context *smu);int (*set_last_dcef_min_deep_sleep_clk)(struct smu_context *smu);drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h +2
-0@@ -170,6 +170,8 @@ int smu_v11_0_write_pptable(struct smu_context *smu);int smu_v11_0_set_min_dcef_deep_sleep (struct smu_context *smu); +int smu_v11_0_set_driver_table_location (struct smu_context *smu); +int smu_v11_0_set_tool_table_location (struct smu_context *smu);int smu_v11_0_notify_memory_pool_location (struct smu_context *smu);drivers/gpu/drm/amd/powerplay/inc/smu_v12_0.h +2
-0@@ -90,4 +90,6 @@ int smu_v12_0_mode2_reset(struct smu_context *smu);int smu_v12_0_set_soft_freq_limited_range (struct smu_context *smu,enum smu_clk_type clk_type,uint32_t min,uint32_t max); +int smu_v12_0_set_driver_table_location (struct smu_context *smu); +#endif drivers/gpu/drm/amd/powerplay/navi10_ppt.c +1
-0@@ -2116,6 +2116,7 @@ static const struct pptable_funcs navi10_ppt_funcs = { .check_fw_version = smu_v11_0_check_fw_version, .write_pptable = smu_v11_0_write_pptable, .set_min_dcef_deep_sleep = smu_v11_0_set_min_dcef_deep_sleep, + .set_driver_table_location = smu_v11_0_set_driver_table_location, .set_tool_table_location = smu_v11_0_set_tool_table_location, .notify_memory_pool_location = smu_v11_0_notify_memory_pool_location, .system_features_control = smu_v11_0_system_features_control,drivers/gpu/drm/amd/powerplay/renoir_ppt.c +1
-0@@ -916,6 +916,7 @@ static const struct pptable_funcs renoir_ppt_funcs = { .get_dpm_ultimate_freq = smu_v12_0_get_dpm_ultimate_freq, .mode2_reset = smu_v12_0_mode2_reset, .set_soft_freq_limited_range = smu_v12_0_set_soft_freq_limited_range, + .set_driver_table_location = smu_v12_0_set_driver_table_location, };void renoir_set_ppt_funcs (struct smu_context *smu)drivers/gpu/drm/amd/powerplay/smu_internal.h +2
-0@@ -61,6 +61,8 @@ ((smu)->ppt_funcs->write_pptable ? (smu)->ppt_funcs->write_pptable((smu)) : 0)#define smu_set_min_dcef_deep_sleep (smu) \ ((smu)->ppt_funcs->set_min_dcef_deep_sleep ? (smu)->ppt_funcs->set_min_dcef_deep_sleep((smu)) : 0) +#define smu_set_driver_table_location (smu) \ + ((smu)->ppt_funcs->set_driver_table_location ? (smu)->ppt_funcs->set_driver_table_location((smu)) : 0)#define smu_set_tool_table_location (smu) \ ((smu)->ppt_funcs->set_tool_table_location ? (smu)->ppt_funcs->set_tool_table_location((smu)) : 0)#define smu_notify_memory_pool_location (smu) \drivers/gpu/drm/amd/powerplay/smu_v11_0.c +18
-0@@ -774,6 +774,24 @@ int smu_v11_0_set_min_dcef_deep_sleep(struct smu_context *smu)return smu_v11_0_set_deep_sleep_dcefclk (smu, table_context->boot_values.dcefclk /100 ); } +int smu_v11_0_set_driver_table_location (struct smu_context *smu) +{ +struct smu_table *driver_table = smu->smu_table.driver_table; +int ret =0 ; + +if (driver_table->mc_address) { + ret =smu_send_smc_msg_with_param (smu, + SMU_MSG_SetDriverDramAddrHigh, +upper_32_bits (driver_table->mc_address)); +if (!ret) + ret =smu_send_smc_msg_with_param (smu, + SMU_MSG_SetDriverDramAddrLow, +lower_32_bits (driver_table->mc_address)); + } + +return ret; +} +int smu_v11_0_set_tool_table_location (struct smu_context *smu) {int ret =0 ;drivers/gpu/drm/amd/powerplay/smu_v12_0.c +18
-0@@ -514,3 +514,21 @@ int smu_v12_0_set_soft_freq_limited_range(struct smu_context *smu, enum smu_clk_return ret; } + +int smu_v12_0_set_driver_table_location (struct smu_context *smu) +{ +struct smu_table *driver_table = smu->smu_table.driver_table; +int ret =0 ; + +if (driver_table->mc_address) { + ret =smu_send_smc_msg_with_param (smu, + SMU_MSG_SetDriverDramAddrHigh, +upper_32_bits (driver_table->mc_address)); +if (!ret) + ret =smu_send_smc_msg_with_param (smu, + SMU_MSG_SetDriverDramAddrLow, +lower_32_bits (driver_table->mc_address)); + } + +return ret; +}drivers/gpu/drm/amd/powerplay/vega20_ppt.c +1
-1@@ -3232,6 +3232,7 @@ static const struct pptable_funcs vega20_ppt_funcs = { .check_fw_version = smu_v11_0_check_fw_version, .write_pptable = smu_v11_0_write_pptable, .set_min_dcef_deep_sleep = smu_v11_0_set_min_dcef_deep_sleep, + .set_driver_table_location = smu_v11_0_set_driver_table_location, .set_tool_table_location = smu_v11_0_set_tool_table_location, .notify_memory_pool_location = smu_v11_0_notify_memory_pool_location, .system_features_control = smu_v11_0_system_features_control, --2.24.0
Quoted 4 lines
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx