summaryrefslogtreecommitdiff
path: root/RadeonsiToDo.mdwn
diff options
context:
space:
mode:
Diffstat (limited to 'RadeonsiToDo.mdwn')
-rw-r--r--RadeonsiToDo.mdwn95
1 files changed, 95 insertions, 0 deletions
diff --git a/RadeonsiToDo.mdwn b/RadeonsiToDo.mdwn
new file mode 100644
index 0000000..6b4200c
--- /dev/null
+++ b/RadeonsiToDo.mdwn
@@ -0,0 +1,95 @@
+
+[[!toc ]]
+
+
+## Radeonsi To Do list
+
+And here is the radeonsi to do list; if you feel something is missing please add it. If you have done something and it no longer needs to be investigated, happily erase it from here :).
+
+
+### LLVM
+
+* **Turn on basic block vectorizer pass**
+ * _Difficulty:_ Medium
+ * _Who's working on it:_
+ * _Status:_ N/A
+ * _Description:_
+ * We might be able to use this pass to merge single dwords loads from constant memory into multiple dword loads.
+
+### Compute
+
+* **Add support for compute interface**
+ * _Difficulty:_ Medium
+ * _Who's working on it:_
+ * _Date Started: _
+ * _Status:_ WIP - Initial patch for non-compiler parts is [[here|http://people.freedesktop.org/~tstellar/0001-radeonsi-WIP-compute-patch.patch]]
+ * _Description:_
+ * Add support for emitting dispatch packets and related compute state
+* **Use asynchronous compute rings for compute **
+ * _Difficulty:_ Medium
+ * _Who's working on it:_ N/A
+ * _Date Started: _N/A
+ * _Status:_ N/A
+ * _Description:_
+ * SI supports two asynchronous compute rings. Use them for compute to avoid contention with gfx on the main gfx ring.
+
+### 3D Features
+
+* **TGSI opcodes DDX/DDY**
+ * _Difficulty:_ Medium
+ * _Who's working on it:_ Michel Dänzer
+ * _Status:_ WIP
+ * _Note:_
+* **MSAA support on SI**
+ * _Difficulty:_ Medium
+ * _Who's working on it:_
+ * _Status:_ N/A
+ * _Note:_ Works very similarly to cayman. Code from r600g can be ported over.
+* **MSAA with colorbuffer compression on SI**
+ * _Difficulty:_ Medium
+ * _Who's working on it:_
+ * _Status:_ N/A
+ * _Note:_ Works very similarly to cayman. Code from r600g can be ported over.
+* **Fast color clear for MSAA buffers on SI**
+ * _Difficulty:_ Medium
+ * _Who's working on it:_ N/A
+ * _Note:_ Works very similarly to cayman. Code from r600g can be ported over.
+* **MSAA on the window system framebuffer**
+ * _Difficulty:_ Medium
+ * _Who's working on it:_ N/A
+ * _Note:_ AFAIK, nothing has been done for it in st/dri. I think we should allocate private MSAA color/depth/stencil surfaces in Gallium and ask the DDX to only allocate the single-sampled front where we will resolve the MSAA colorbuffer on [[SwapBuffers|SwapBuffers]].
+* **transform feedback support**
+ * _Difficulty:_ Medium
+ * _Who's working on it:_ N/A
+ * _Note:_ Works very similarly to cayman. Code from r600g can be ported over.
+* **timer query support**
+ * _Difficulty:_ Easy
+ * _Who's working on it:_ N/A
+ * _Note:_ Works very similarly to r6xx-cayman. Code from r600g can be ported over.
+* **Polygon stippling**
+ * _Difficulty:_ Easy
+ * _Who's working on it:_ N/A
+ * _Note:_ Take advantage of util/u_pstipple. It contains the shader transformation code and can create a stipple texture.
+* **Polygon/line/point smoothing**
+ * _Difficulty:_ Medium
+ * _Who's working on it:_ N/A
+* **Texture buffer objects (read-only buffer resources in shaders)**
+ * _Difficulty:_ Medium
+ * _Who's working on it:_ N/A
+ * _Note:_: Core Mesa/Gallium support might need some work too.
+* **Uniform buffer objects (constant buffers)**
+ * _Difficulty:_ Medium
+ * _Who's working on it:_ N/A
+ * _Note:_: Core Mesa/Gallium support might need some work too.
+* **HyperZ support**
+ * _Difficulty:_ Hard
+ * _Who's working on it:_ N/A
+ * _Note:_: Works very similarly to cayman. Code for r600g can be ported over.
+* **Geometry shaders**
+ * _Difficulty:_ Medium
+ * _Who's working on it:_ N/A
+ * _Note:_: Core Mesa/Gallium support might need some work too.
+
+### Bugs
+
+* All open bugs assigned to [[Drivers/Gallium/radeonsi|https://bugs.freedesktop.org/buglist.cgi?bug_status=NEW&bug_status=ASSIGNED&bug_status=REOPENED&component=Drivers%2FGallium%2Fradeonsi&product=Mesa]] \ No newline at end of file