Archives Discussions

[Raistmer]

SKA shows no scratchshould be used but profiler uses scratch registers.

For same kernel SKA shows:
Name,Scratch Reg,GPR,Min,Max,Avg,Est Cycles,Est Cycles(Bi),ALU:Fetch(Bi),BottleNeck(Bi),%s\Clock(Bi),Throughput(Bi)
Radeon HD 4870,0,119,244.90,36705.76,6919.27,6919.27,6919.27,1.19,ALU Ops,0.00,2 M Threads\Sec

0 scratch registers and 119 GPR.
And indeed, HD4870 assembly doesn't contain scratch read/write instructions.

But profiler shows 121 GPR and scratch registers usage!
And its assembly shows scratch instructions.
MaxScratchRegsNeeded = 98
SQ_PGM_RESOURCES:NUM_GPRS = 121


Why disagreement in so important thing for performance as scratch registers usage?

[bpurnomo]

Are you using the same Catalyst version for SKA (set in the tool) and profiler (installed in the system)?

[ryta1203]

Originally posted by: bpurnomo Are you using the same Catalyst version for SKA (set in the tool) and profiler (installed in the system)?

Shouldn't need to...

If the SKA is reporting incorrect results because it's using an "old" driver then that essentially makes the SKA useless...

...I honestly don't see why the SKA can't use the most recent driver that the user has on his/her system, it's not like the SKA is made by some 3rd party.... or is it?

[Raistmer]

Well, I have same suspiction...
Catalyst installed - 10.4 [08/05/2010 11:14:34 ATI GPU 0: ATI Radeon HD 4700/4800 (RV740/RV770) (CAL version 1.4.635, 512MB, 1200 GFLOPS peak) ]
Catalyst used by SKA - 10.3. It reports 10.3 as "last available" [Use Latest Available (CAL 10.3) - v1.54.2407 ]- definitely it's untrue. I could switch to older catalyst version in SKA but not on really installed in the system.

BTW, it means that new Catalyst 10.4 treats my kernel much worse than prev one! It issues scratch register using instruction that should be slower...