Skip to content
Snippets Groups Projects

Repository graph

You can move around the graph by using the arrow keys.
Select Git revision
  • aggressive-sroa
  • cpu_vector
  • cpu_vector_better2
  • debug_info
  • distributed
  • experimenting
  • fork-guard-elim-editor
  • forkify
  • frontend
  • gpu-cg
  • grape-initial
  • interpreter-fix
  • ir_debug
  • juno_macro
  • llama
  • main default protected
  • matmul_gpu_opt
  • matmul_paper_schedules
  • misc
  • opt_cfd
20 results
Created with Raphaël 2.2.07May30Apr29282723222116145231Mar2726241685432128Feb2726252422212019181615141312111098765432131Jan30292827262524232221201918191817161514131211109865130Dec292827232221161514131165126Nov25242322201817151211103131Oct29281487630Sep26252420Set up matmul bench, initial GPU schedulenum_cpus in more placesrestore matmulsaving herewe can get backing size out nowsome cleanupdc substitution workssamplesgrape-initialgrape-initialreduction on hostreduction mapsreassociation go brr part 2new reassoc exampleMerge remote-tracking branch 'origin/ryan-reassoc' into grape-initialfix multi returnreassociation go brrryan-reassocryan-reassocsome array paramsfunctional sim workingcreate reassociate pass that does nothinggrape functional simulatormore grapegrapeomg the delayed schedule workshigher order functions in progress!generate arguments string to rundon't overwrite unnecessarilycleanupOH MY GOD IT WORKSsuccessfully generate some jit launchergenerate more jsonno gcm for delayed functionsdelayed now generates no code weeehardcoded runtime JITcreate new rtc device which is same as asyncrustOnly use zero collection constants in Junozero-collection…zero-collection-constants2nd version of backendhorrible horribleconv1dprototype xdotno clue what i changedmost cursed algorithm