Nintendo Switch Brew

GPU: Difference between revisions

← Older editNewer edit →
this is good stuff
No edit summary
Line 1: Line 1:
==Mapping Memory==
== Mapping Memory ==


First, to map a memory region on the GPU Address Space, caching needs to be disabled by using [[SVC#svcSetMemoryAttribute|svcSetMemoryAttribute]]. The Address passed is the Virtual Address of the region that will be mapped, the size is the region size, and State0/1 are both set to 8 to disable caching of the memory region. This is done to ensure that the GPU can actually "see" the data written there, and it doesn't get stuck on some cache.
First, to map a memory region on the GPU Address Space, caching needs to be disabled by using [[SVC#svcSetMemoryAttribute|svcSetMemoryAttribute]]. The Address passed is the Virtual Address of the region that will be mapped, the size is the region size, and State0/1 are both set to 8 to disable caching of the memory region. This is done to ensure that the GPU can actually "see" the data written there, and it doesn't get stuck on some cache.
Line 7: Line 7:
The above process is used to map all data that will be used by the GPU, like Textures, Command Lists (a.k.a. Push Buffers), Vertex/Index buffers and Shaders. They usually have their own mapping, but Command Lists can share the same mapping.
The above process is used to map all data that will be used by the GPU, like Textures, Command Lists (a.k.a. Push Buffers), Vertex/Index buffers and Shaders. They usually have their own mapping, but Command Lists can share the same mapping.


==Commands Submission==
== FIFO Commands ==


Commands are sent to the GPU through [[NV_services#NVGPU_IOCTL_CHANNEL_SUBMIT_GPFIFO|NVGPU_IOCTL_CHANNEL_SUBMIT_GPFIFO]]. This IoCtl command accepts various GpFifo entries, and each GpFifo entry points to a Command List. The GPU Command List is composed of 32-bits words, which usually are Command/Argument pairs.
The GPU implements a variation of Tegra's push buffer format for it's PFIFO engine. PFIFO is a special engine responsible for receiving user command lists and routing them to the appropriate engines (2D, 3D, DMA).


====Command Word Structure====
Commands are submitted to the GPU's PFIFO engine through [[NV_services#NVGPU_IOCTL_CHANNEL_SUBMIT_GPFIFO|NVGPU_IOCTL_CHANNEL_SUBMIT_GPFIFO]].
 
This ioctl takes an array of gpfifo entries where each entry points to a FIFO command list. This list is composed of alternating 32-bit words containing FIFO commands and their respective arguments.
 
=== Command Structure ===


{| class="wikitable"
{| class="wikitable"
Line 19: Line 23:
|-
|-
|12-0
|12-0
|Command/Register Id
|Method
|-
|-
|15-13
|15-13
|Sub Channel
|Subchannel
|-
|-
|28-16
|28-16
|Arguments Count (in 32-bits Words) or Inline Data (see below)
|Argument count (in 32-bits Words) or inline data (see below)
|-
|-
|31-29
|31-29
|Mode
|[[#Submission_mode|Submission mode]]
|}
|}


====Command Mode====
Note: Methods are treated as 4-byte addressable locations, and hence their numbers are written down multiplied by 4.
 
Note: The command's arguments, when present, follow the command word immediately.
 
==== Submission mode ====


{| class="wikitable"
{| class="wikitable"
Line 38: Line 46:
! scope="col"| Description
! scope="col"| Description
! scope="col"| Offical name
! scope="col"| Offical name
|-
|0
|Increasing mode (old)
|
|-
|-
|1
|1
|Sequential Mode - Reads "Argument Count" arguments, while automatically incrementing the Register Id. So, each argument is written to a different register.
|Increasing mode - Tells PFIFO to read as much arguments as specified by '''argument count''', while automatically incrementing the '''method''' value. This means that each argument will be written to a different method location.
|INCR
|INCR
|-
|2
|Non-increasing mode (old)
|
|-
|-
|3
|3
|Normal Mode - This is a Command with multiple arguments. Reads "Argument Count" arguments, all belonging to the same Command.
|Non-increasing mode - Tells PFIFO to read as much arguments as specified by '''argument count'''. However, all arguments will be written to the same method location.
|NONINCR
|NONINCR
|-
|-
|4
|4
|Inline Mode - Bits 28-16 of the Command Word (where Inline Data is located) contains the Value of the argument written to the register. The next Word is another Command.
|Inline mode - Tells PFIFO to read '''inline data''' from bits 28-16 of the command word, thus eliminating the need to pass additional words for the arguments.
|IMM
|IMM
|-
|-
|5
|5
|Unobserved, but is valid too.
|Increase-once mode - Tells PFIFO to read as much arguments as specified by '''argument count''' and automatically increments the '''method''' value once only.
|
|
|}
|}


TODO: Find a better name for the "Normal Mode" and figure out what mode 5 is.
=== Command List ===
 
All methods with values < 0x100 are special and executed by the PFIFO's DMA puller. The others are forwarded to the engine object currently bound to a given subchannel.
 
{| class="wikitable" border="1"
|-
! Command || Method || Subchannel || Arg Count || Mode || Name
|-
| 0x2001?000 || 0x000 || Variable || 1 || 1 || [[#BindObject|BindObject]]
|-
| 0xA0020E00 || 0xE00 || 0 || 2 || 5 || BeginTransformFeedback
|-
| 0xA0030E30 || 0xE30 || 0 || 3 || 5 || DrawArrays
|-
| 0xA0050E36 || 0xE36 || 0 || 5 || 5 || DrawElements
|-
| 0xA0020E2E || 0xE2E || 0 || 2 || 5 || PopDebugGroupId
|-
| 0xA0040E2C || 0xE2C || 0 || 4 || 5 || PushDebugGroup
|-
| 0x2001054C || 0x54C || 0 || 1 || 1 || ResetCounter
|-
| 0x8001047F || 0x47F || 0 || 1 || 4 || ResolveDepthBuffer
|-
| 0x200104C4 || 0x4C4 || 0 || 1 || 1 || SetAlphaRef
|-
| 0x200404C7 || 0x4C7 || 0 || 4 || 1 || SetBlendColor
|-
| 0x2001064F || 0x6F4 || 0 || 1 || 1 || SetDepthClamp
|-
| 0x200200CD || 0xCD || 0 || 2 || 1 || SetInnerTessellationLevels
|-
| 0x200204EC || 0x4EC || 0 || 2 || 1 || SetLineWidth
|-
| 0x200400C9 || 0xC9 || 0 || 4 || 1 || SetOuterTessellationLevels
|-
| 0x8???0373 || 0x373 || 0 || Variable || 4 || SetPatchSize
|-
| 0x20010546 || 0x546 || 0 || 1 || 1 || SetPointSize
|-
| 0x20030554 || 0x554 || 0 || 3 || 1 || SetRenderEnableConditional
|-
| 0x200403EF || 0x3EF || 0 || 4 || 1 || SetSampleMask
|-
| 0x200103D9 || 0x3D9 || 0 || 1 || 1 || SetTiledCacheTileSize
|}


Other mode values are unobserved.
Note: These still need to be heavily verified and ''could'' be wrong.


Note: All Commands/Register Id values are multiples of 4, so they are divided by 4 when packing, and multiplied by 4 when unpacking.
=== BindObject ===


==Sub Channel binding==
In order to bind an engine object to a specific subchannel, method 0 (BindObject) must be used first. The target subchannel is specified in bits 15-13 of the command word.


All Command Id values < 0x100 are special and aren't fowarded to the engines. The command 0 is used to bind engines to Sub Channels, and needs to be used before commands are submited to the engines.
After the engine object is bound to the desired subchannel, setting it's value in bits 15-13 of any subsequent command word will make PFIFO forward the command to the target engine.


The command 0 only has one argument, the Engine Id.
This method only takes one argument, an [[#Engine_IDs|engine ID]].


====Engine Ids====
==== Engine IDs ====


{| class="wikitable"
{| class="wikitable"
|-
|-
! scope="col"| Id
! scope="col"| ID
! scope="col"| Engine
! scope="col"| Engine
|-
|-
|0x902d
|0x902D
|2D
|FERMI_TWOD_A (2D)
|-
|-
|0xb197
|0xB197
|3D
|MAXWELL_B (3D)
|-
|-
|0xb1c0
|0xB1C0
|Compute
|MAXWELL_COMPUTE_B
|-
|-
|0xa140
|0xA140
|Kepler
|KEPLER_INLINE_TO_MEMORY_B
|-
|-
|0xb0b5
|0xB0B5
|DMA
|MAXWELL_DMA_COPY_A (DMA)
|}
|}


The bits 15-13 of the Command Word contains the Sub Channel index that should be bound.
=== Fences ===
 
After binding the required Sub Channels, then the respective values can be used on the "Sub Channel" field of the Command Word to talk with the respective Engines.
 
==Fences==


Command Lists can contain fences to ensure that commands are executed on the correct order, and subsequent commands are only sent when the previously sent commands were already processed by the GPU. Fences uses the QUERY_* commands, and works like this:
Command lists can contain fences to ensure that commands are executed on the correct order, and subsequent commands are only sent when the previously sent commands were already processed by the GPU. Fences uses the QUERY_* commands, and works like this:


* First, QUERY_ADDRESS_HIGH and QUERY_ADDRESS_LOW commands are added to the Command List, with the High/Low 32 bits part of the 64-bits GPU Virtual Address where the fence is located. This GPU Virtual Address needs to be mapped to the process Virtual Address beforehand.
* First, QUERY_ADDRESS_HIGH and QUERY_ADDRESS_LOW commands are added to the Command List, with the High/Low 32 bits part of the 64-bits GPU Virtual Address where the fence is located. This GPU Virtual Address needs to be mapped to the process Virtual Address beforehand.
Line 103: Line 160:
* Finally, QUERY_GET is added and contains the mode and other unknown data.
* Finally, QUERY_GET is added and contains the mode and other unknown data.


The above commands are added using the Sequential Mode, since the Ids for all those 4 registers are sequential.
The above commands are added using the [[#Submission_mode|increasing mode]], since the Ids for all those 4 registers are sequential.


====QUERY_GET Structure====
==== QUERY_GET Structure ====


{| class="wikitable"
{| class="wikitable"
Line 122: Line 179:
|}
|}


====QUERY_GET Mode====
==== QUERY_GET Mode ====


{| class="wikitable"
{| class="wikitable"
Line 149: Line 206:
On the CPU side, the game code should wait until the value at the address pointed to by QUERY_ADDRESS is >= to the last written SEQUENCE value. Official code waits for this condition to be true on a loop, and won't send any further commands before that.
On the CPU side, the game code should wait until the value at the address pointed to by QUERY_ADDRESS is >= to the last written SEQUENCE value. Official code waits for this condition to be true on a loop, and won't send any further commands before that.


==Vertex Data Submission==
== Vertex Data Submission ==


Note: This is a observation on how the game Puyo Puyo Tetris sends textured squares to the GPU.
Note: This is a observation on how the game Puyo Puyo Tetris sends textured squares to the GPU.
Line 162: Line 219:
# VERTEX_END_GL is used with value 0 (currently unknown what this value means).
# VERTEX_END_GL is used with value 0 (currently unknown what this value means).


==Command List==
== References ==
 
These still need to be heavily verified and ''could'' be wrong
{| class="wikitable" border="1"
|-
! Command || ID/Register || Sub Channel || Arg Count || Mode || Command Name
|-
| 0xA0020E00 || 0xE00 || 0 || 2 || 5 || BeginTransformFeedback
|-
| 0xA0030E30 || 0xE30 || 0 || 3 || 5 || DrawArrays
|-
| 0xA0050E36 || 0xE36 || 0 || 5 || 5 || DrawElements
|-
| 0xA0020E2E || 0xE2E || 0 || 2 || 5 || PopDebugGroupId
|-
| 0xA0040E2C || 0xE2C || 0 || 4 || 5 || PushDebugGroup
|-
| 0x2001054C || 0x54C || 0 || 1 || 1 || ResetCounter
|-
| 0x8001047F || 0x47F || 0 || 1 || 4 || ResolveDepthBuffer
|-
| 0x200104C4 || 0x4C4 || 0 || 1 || 1 || SetAlphaRef
|-
| 0x200404C7 || 0x4C7 || 0 || 4 || 1 || SetBlendColor
|-
| 0x2001064F || 0x6F4 || 0 || 1 || 1 || SetDepthClamp
|-
| 0x200200CD || 0xCD || 0 || 2 || 1 || SetInnerTessellationLevels
|-
| 0x200204EC || 0x4EC || 0 || 2 || 1 || SetLineWidth
|-
| 0x200400C9 || 0xC9 || 0 || 4 || 1 || SetOuterTessellationLevels
|-
| 0x8???0373 || 0x373 || 0 || Variable || 4 || SetPatchSize
|-
| 0x20010546 || 0x546 || 0 || 1 || 1 || SetPointSize
|-
| 0x20030554 || 0x554 || 0 || 3 || 1 || SetRenderEnableConditional
|-
| 0x200403EF || 0x3EF || 0 || 4 || 1 || SetSampleMask
|-
| 0x200103D9 || 0x3D9 || 0 || 1 || 1 || SetTiledCacheTileSize
|}
 
==References==


Check out those pages for more useful data.
FIFO engine overview:
[https://envytools.readthedocs.io/en/latest/hw/fifo/intro.html]


Register Id values from the Fermi family GPU (a bit older than the Tegra X1, but values seems to be mostly the same):
Method values from the Fermi family GPU (a bit older than the Tegra X1, but values seems to be mostly the same):
[https://github.com/envytools/envytools/blob/master/rnndb/graph/gf100_3d.xml]
[https://github.com/envytools/envytools/blob/master/rnndb/graph/gf100_3d.xml]


Line 216: Line 230:
[https://github.com/envytools/envytools/blob/master/rnndb/graph/nv_3ddefs.xml]
[https://github.com/envytools/envytools/blob/master/rnndb/graph/nv_3ddefs.xml]


Command Word packing code used on Mesa3d:
Command word packing code used on Mesa3d:
[https://cgit.freedesktop.org/mesa/mesa/tree/src/gallium/drivers/nouveau/nvc0/nvc0_winsys.h]
[https://cgit.freedesktop.org/mesa/mesa/tree/src/gallium/drivers/nouveau/nvc0/nvc0_winsys.h]
Retrieved from "https://switchbrew.org/wiki/GPU"