Small APUs(i.e., consumer, embedded products) usually have a small
carveout device memory which can't satisfy most compute workloads
memory allocation requirements.
We can't even run a Basic MNIST Example with a default 512MB carveout.
https://github.com/pytorch/examples/tree/main/mnist. Error Log:
"torch.cuda.OutOfMemoryError: HIP out of memory. Tried to allocate
84.00 MiB. GPU 0 has a total capacity of 512.00 MiB of which 0 bytes
is free. Of the allocated memory 103.83 MiB is allocated by PyTorch,
and 22.17 MiB is reserved by PyTorch but unallocated"
Though we can change BIOS settings to enlarge carveout size,
which is inflexible and may bring complaint. On the other hand,
the memory resource can't be effectively used between host and device.
The solution is MI300A approach, i.e., let VRAM allocations go to GTT.
Then device and host can flexibly and effectively share memory resource.
v2: Report local_mem_size_private as 0. (Felix)
Signed-off-by: Lang Yu <Lang.Yu@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This version brings along following fixes:
- Fix some problems reported by Coverity
- Fix idle optimization checks for multi-display and dual eDP
- Fix incorrect size calculation for loop
- Fix DSC-re-computing
- Add Replay capability and state in debugfs
- Refactor DCCG into component folder
- Refactor input mode programming for DIG FIFO
- Reset DSC clock in post unlock update
- Clean-up recout calculation for visual confirm
- Enable urgent latency adjustments for DCN35
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
- Implement command interface to query ABM SW algorithm and
HW caps. This is primarily intended as a debugging interface
- Add new definitions for max number of histogram bins and ABM
curve segments available in hardware
- Add structures to retrieve caps to describe ABM HW caps
since not all ASICs have the same number of cure segments and
bins
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Anthony Koo <anthony.koo@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[why]
Switching between DSC clock or disable DSC block are not double buffered update.
Corruption is observed if these updates happen before DSC double buffered
disconnection.
[how]
Move DSC disable and refclk reset to post unlock update. Wait for DSC double buffered
disconnection and all mpccs are disconnected before reset DSC clock.
Reviewed-by: Samson Tam <samson.tam@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Wenjing Liu <wenjing.liu@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
Some older MST hubs do not report DPCD registers according to
specification.
[How]
This change re-applies commit c536555451 ("drm/amd/display: dsc mst
re-compute pbn for changes on hub").
With an additional check for these older MST devices.
Reviewed-by: Swapnil Patel <swapnil.patel@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
This fixes a bug introduced by commit c536555451 ("drm/amd/display: dsc
mst re-compute pbn for changes on hub").
The change caused light-up issues with a second display that required
DSC on some MST docks.
[How]
Use Virtual DPCD for DSC caps in MST case.
[Limitations]
This change only affects MST DSC devices that follow specifications
additional changes are required to check for old MST DSC devices such as
ones which do not check for Virtual DPCD registers.
Reviewed-by: Swapnil Patel <swapnil.patel@amd.com>
Reviewed-by: Hersen Wu <hersenxs.wu@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Agustin Gutierrez <agustin.gutierrez@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
Underflow occurs when running Netflix in a 4k144 eDP + 4k60 HDMI FRL
setup. It is caused by latency varying based on the DCFCLK/FCLK state.
[How]
Enable urgent latency adjustment and match the reference to existing
ASIC that also see increased latency at low FCLK.
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Nicholas Susanto <nicholas.susanto@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[WHY]
Input mode for the DIG FIFO should be programmed as part of stream
encoder setup.
[HOW]
Pre-calculate the pixels per cycle as part of the pixel clock params,
and program as part of stream encoder setup.
Reviewed-by: Wenjing Liu <wenjing.liu@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Dillon Varone <dillon.varone@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
Providing a mechanism to manipulate the pwm frequency on the
individual GPUs and monitor the transition during the switch.
[How]
Added a variable in dc_debug_options data structure. Using
this variable to call the newly added command on the firmware.
Reviewed-by: Harry Vanzylldejong <harry.vanzylldejong@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Sreeja Golui <sreeja.golui@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why & How]
User can get the panel replay capability and state for debug.
sudo cat /sys/kernel/debug/dri/0/eDP-1/replay_capability
"Sink support: no" - if panel doesn't support Replay
"Sink support: yes" - if panel supports Replay
"Driver support: no\n" - if driver doesn't support Replay
"Driver support: yes\n" - if driver supports Replay
sudo cat /sys/kernel/debug/dri/0/eDP-1/replay_state
It will return current panel replay state
Reviewed-by: Sun peng Li <sunpeng.li@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Tom Chung <chiahsuan.chung@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[WHY & HOW]
The assignment of disp_cfg_index_max is missed and should be
assigned to __DML2_WRAPPER_MAX_STREAMS_PLANES__.
Fixes: 55ec7679e6 ("drm/amd/display: Limit array index according to architecture")
Reviewed-by: Hersen Wu <hersenxs.wu@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[WHY & HOW]
Make sure vmid0p72_idx, vnom0p8_idx and vmax0p9_idx calculation will
never overflow and exceess array size.
This fixes 3 OVERRUN and 1 INTEGER_OVERFLOW issues reported by Coverity.
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
dc_clk_table->entries has size of MAX_NUM_DPM_LVL(=8), but the loop
counter i can go up to DML_MAX_CLK_TABLE_SIZE(=20) - 1.
[How]
The loop should be min(DML_MAX_CLK_TABLE_SIZE, MAX_NUM_DPM_LVL) - 1
instead.
This fixes 21 OVERRUN issues reported by Coverity.
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
For addtion (uint8_t) variable + constant 1,
coverity generates message below:
Truncation due to cast operation on "cur_idx + 1" from
32 to 8 bits.
Then Coverity assume result is 32 bits value be saved into
8 bits variable. When result is used as index to access
array, Coverity suspects index invalid.
[How]
Change varaible type to uint32_t.
Reviewed-by: Alex Hung <alex.hung@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Hersen Wu <hersenxs.wu@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[WHY & HOW]
ctx->architecture determine array sizes of ODMMode and DPPPerSurface
arrays to __DML2_WRAPPER_MAX_STREAMS_PLANES__ or __DML_NUM_PLANES__,
and these array index should be checked before used
This fixes 2 OVERRUN issues reported by Coverity.
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
Coverity reports OVERRUN warning for
CalculateSwathAndDETConfiguration_params->hw_debug5
= &s->dummy_boolean[2].
bool dummy_boolean[2] is defined within
struct dml2_core_calcs_mode_support_locals.
[How]
Change array size from 2 to 3.
Reviewed-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Hersen Wu <hersenxs.wu@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[WHY & HOW]
dc->links[] has max size of MAX_LINKS and NULL is return when trying to
access with out-of-bound index.
This fixes 3 OVERRUN and 1 RESOURCE_LEAK issues reported by Coverity.
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[WHY]
fe_clk_en has size of 5 but sizeof(fe_clk_en) has byte size 20 which is
lager than the array size.
[HOW]
Divide byte size 20 by its element size.
This fixes 2 OVERRUN issues reported by Coverity.
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
Coverity report OVERRUN warning. There are
only max_links elements within dc->links. link
count could up to AMDGPU_DM_MAX_DISPLAY_INDEX 31.
[How]
Make sure link count less than max_links.
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Acked-by: Tom Chung <chiahsuan.chung@amd.com>
Signed-off-by: Hersen Wu <hersenxs.wu@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
