Just a question:
is it possible to add a DP computation speed readout in the startup-messages for the GPU's?
The GFLOPS from the spec-sheets of the vendors do not really match the readout from BOINC, so this might help on some projects.

is it possible to add a DP computation speed readout in the startup-messages for the GPU's?
The GFLOPS from the spec-sheets of the vendors do not really match the readout from BOINC, so this might help on some projects.

I guess you are referring to something like this that you see on the boinc event log on GPU detection?

OpenCL: AMD/ATI GPU 0: Ellesmere (driver version 2117.7 (VM), device version OpenCL 1.2 AMD-APP (2117.7), 8098MB, 8098MB available, 995 GFLOPS peak)

EDIT:The information boinc extracts during GPU detection at startup is stored in coproc_info.xml so you might want to look at that file.

I'm not quite sure i understand what you mean by help. Do you mean help the project scheduler assign tasks?

Absolute correct.
In the same manner as the peak flops are reported the DP flops should be reported.

THX for the info, where mere infos are stored (copoc_info.xml)
BTW, what is the meaning of:
<target>20</target>

<half_fp_config>0</half_fp_config>

<single_fp_config>190</single_fp_config>

<double_fp_config>63</double_fp_config>

<endian_little>1</endian_little>

Absolute correct.
In the same manner as the peak flops are reported the DP flops should be reported.

Could be i guess.

THX for the info, where mere infos are stored (copoc_info.xml)
BTW, what is the meaning of:

Try looking around OpenCL documentation and AMD's OpenCL programming documentation.