Guys, I would appreciate it if someone with a 5xxx and 6xxx radeon could test my modified kernel and tell how did his hps change. That would help to understand if the optimal code is different for VLIW and GCN architectures and implement it accordingly.
It tested your new code with a HD 5850 over night:
old: 91Mh/s (estimate on share ratio) @ 67°C
new:88 Mh/s(estimate on share ratio) @ 63°C
core running at 900 Mhz, memory at 400 Mhz
So it seems to give comparable hashrates, but the work load of my gpu seems to be less. I will repeat this with a different memory clock speed hopefully today ( I want to install the temp sensing module before doing so)
P.s.: nice work!