Advertisement
Not a member of Pastebin yet?
Sign Up,
it unlocks many cool features!
- Here there be dragons!
- Along with debugging options (see below for a hopefully pretty complete list of possible options),
- additional features can be enabled on Radeon cards using the "R600_DEBUG" env var. (r600/radeonsi/amdgpu)
- The value of this variable can be set e.g. per individual user via the "~/.profile" config file, or for the entire machine via the "/etc/environment" config file.
- Example contents of the most useful to the average user:
- R600_DEBUG=sbcl,hyperz,llvm,sisched,forcedma
- As of late Jan 2016, Radeonsi and AMDGPU now have the "SI Scheduler" which can be enabled by the user (rather than solely by the package maintainer) by setting "R600_DEBUG=sisched".
- There may be more than this list includes. For better descriptions and to see what is enabled by default for your card, see the man pages. Certain values will require specific (usually later) versions of the software in order to work (for example, the si scheduler requires llvm 3.8+ in order to do anything)!
- The new SI scheduler is not enabled by default, but may be worth enabling if you like, since the option is available.
- Tests on phoronix have shown improved performance, although admittedly sometimes only marginal, depending on the usecase.
- Such features may be subject to slight hiccups during future tweaking to the source (as expected of bleeding edge software). (What works now might not work after a patch, but then suddenly work much better...who knows!)
- The usual risks apply when using this software, and devs cannot be held liable. Bugs are more likely to occur with these new values, but they also present a good way to help debug them! (Double edged sword!)
- Please excuse any errors, this file is intended to help curious mesa users get the most out of their AMD GPU(s).
- -LOGGING-
- "tex" "Print texture info"
- "texmip" Print texture info (mipmapped only)"
- "compute" Print compute info
- "vm" Print virtual addresses when creating resources
- "trace_cs" Trace cs and write rlockup_<csid>.c file with faulty cs
- "info" Print driver information
- -SHADERS- (debugging)
- "fs" Print fetch shaders
- "vs" Print vertex shaders
- "gs" Print geometry shaders
- "ps" Print pixel shaders
- "cs" Print compute shaders
- "tcs" Print tessellation control shaders
- "tes" Print tessellation evaluation shaders
- "noir" Don't print the LLVM IR
- "notgsi" Don't print the TGSI
- "noasm" Don't print disassembled shaders
- -FEATURES-
- "nodma" Disable asynchronous DMA
- "nohyperz" Disable Hyper-Z
- "noinvalrange" (GL uses the word INVALIDATE, gallium uses DISCARD) Disable handling of INVALIDATE_RANGE map flags
- "no2d" Disable 2D tiling
- "notiling" Disable tiling
- "switch_on_eop" Program WD/IA to switch on end-of-packet
- "precompile" Compile one shader variant at shader creation
- "nowc" Disable GTT write combining
- "check_vm" Check VM faults and dump debug information
- "llvm" Enable the LLVM shader compiler
- "sisched" Enable the SI scheduler in the LLVM backend
- "forcedma" Use asynchronous DMA for all operations when possible
- "nocpdma" Disable CP DMA
- -SHADER BACKEND-
- "nosb" Disable sb backend for graphics shaders
- "sbcl" Enable sb backend for compute shaders
- "sbdry" Don't use optimised bytecode (just print the dumps)
- "sbstat" Print optimisation statistics for shaders
- "sbdump" Print IR dumps after some optimisation passes
- "sbnofallback" Abort on errors instead of fallback
- "sbdisasm" Use sb disassembler for shader dumps
- "sbsafemath" Disable unsafe math optimisations
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement