lam@windchime.arc.nasa.gov (03/30/91)
I am trying to tune my program on the VGX and the profiling statistics returns with: %time seconds cum % cum sec procedure (file) 34.0 79.6100 34.0 79.61 gl_impucode (impcode.c) 9.8 22.9100 43.8 102.52 draw_object (sg_object.c) 4.8 11.2000 48.6 113.72 v3f (./STAPUFTIP7GT.O/gl_if.s) 4.4 10.3400 53.0 124.06 gl_i_n3f (sgl2.s) 4.2 9.8000 57.2 133.86 gl_i_v3f (sgl2.s) My question is what does "gl_impucode" do? Can someone who knows explain? Thanks, Tony Lam
rad@home.asd.sgi.com (Bob Drebin) (04/02/91)
In article <1991Mar30.010020.6227@riacs.edu>, lam@windchime.arc.nasa.gov writes: |> I am trying to tune my program on the VGX and the profiling statistics |> returns with: |> |> %time seconds cum % cum sec procedure (file) |> |> 34.0 79.6100 34.0 79.61 gl_impucode (impcode.c) |> 9.8 22.9100 43.8 102.52 draw_object (sg_object.c) |> 4.8 11.2000 48.6 113.72 v3f (./STAPUFTIP7GT.O/gl_if.s) |> 4.4 10.3400 53.0 124.06 gl_i_n3f (sgl2.s) |> 4.2 9.8000 57.2 133.86 gl_i_v3f (sgl2.s) |> |> My question is what does "gl_impucode" do? Can someone who knows explain? |> |> Thanks, |> |> Tony Lam gl_impucode is a gl-internal subroutine to download microcode to the Image Memory Processors (IMPs). The subroutine is called whenever the pixel microcode needed for the current gl mode settings is not resident in the IMPs. If gl_impucode REALLY is your bottleneck, then you must be changing modes extremely often (maybe every polygon?). It should be mentioned that measuring the percentage of cpu tine spent in each of the gl subroutines is generally not an effective method for tuning gl applications. Most of the performance critical work of the gl happens in the graphics pipeline, not on the cpu. --Bob Drebin