The old 'switcher' mechanism (a program that determined the CPU capabilities and chose an application binary out of three to execute) on Windows has been replaced by the plan class mechanism (the client reports the CPU features and the scheduler decides which App to send). The main reason is that the switcher app caused problems for an increasing number of people that use Windows 7.
einstein_S5R6_3.01_windows_intelx86__S5R6sse2.exe is identical to einstein_S5R6_3.01_windows_intelx86_2.exe of the old app bundle.
The old 'switcher' mechanism (a program that determined the CPU capabilities and chose an application binary out of three to execute) on Windows has been replaced by the plan class mechanism (the client reports the CPU features and the scheduler decides which App to send). The main reason is that the switcher app caused problems for an increasing number of people that use Windows 7.
einstein_S5R6_3.01_windows_intelx86__S5R6sse2.exe is identical to einstein_S5R6_3.01_windows_intelx86_2.exe of the old app bundle.
BM
Bernd,
why does it state "sse2" even though CPUZ reports my processors are "sse4.1"?
One very beneficial side-effect will be substantially reduced downloads (over 20MB saved) for new users/hosts.
A much more modest benefit to folks who watch their task lists using something such as Process Explorer will be the end of the excess ap per core running Einstein that was just the switcher. I'll appreciate that.
New apps for S5R6?
)
Where are you seeing this? The applications page shows nothing new since 12 October 2009.
12.01.2010
)
12.01.2010 15:47:09 Einstein@Home Started download of einstein_S5R6_3.01_windows_intelx86__S5R6sse2.exe
And in the taskstab of BOINC the app has the tag "(S5R6sse2)"
RE: 12.01.2010
)
I see the same thing. Wonder why "sse2" since my computer is capable of sse4.1???
The old 'switcher' mechanism
)
The old 'switcher' mechanism (a program that determined the CPU capabilities and chose an application binary out of three to execute) on Windows has been replaced by the plan class mechanism (the client reports the CPU features and the scheduler decides which App to send). The main reason is that the switcher app caused problems for an increasing number of people that use Windows 7.
einstein_S5R6_3.01_windows_intelx86__S5R6sse2.exe is identical to einstein_S5R6_3.01_windows_intelx86_2.exe of the old app bundle.
BM
BM
RE: The old 'switcher'
)
Bernd,
why does it state "sse2" even though CPUZ reports my processors are "sse4.1"?
RE: why does it state
)
There are only different apps and plan classes for SSE and SSE2. What you see there is the name of the plan class, not that of your CPU capabilities.
BM
BM
Ooooh, I've got one
)
Ooooh, I've got one too.
You can see the change in client_state.xml: it's gone from
einstein_S5R6 301 windows_intelx86 1.000000 1.000000 2644463532.873237 einstein_S5R6_3.01_windows_intelx86.exe einstein_S5R6_3.01_windows_intelx86_0.exe einstein_S5R6_3.01_windows_intelx86_0.exe einstein_S5R6_3.01_windows_intelx86_1.exe einstein_S5R6_3.01_windows_intelx86_1.exe einstein_S5R6_3.01_windows_intelx86_2.exe einstein_S5R6_3.01_windows_intelx86_2.exe einstein_S5R6_3.01_graphics_windows_intelx86.exe graphics_appto
einstein_S5R6 301 windows_intelx86 1.000000 1.000000 4231141652.597179 S5R6sse2 6.7.4 einstein_S5R6_3.01_windows_intelx86__S5R6sse2.exe einstein_S5R6_3.01_graphics_windows_intelx86.exe graphics_appOne very beneficial side-effect will be substantially reduced downloads (over 20MB saved) for new users/hosts.
RE: One very beneficial
)
A much more modest benefit to folks who watch their task lists using something such as Process Explorer will be the end of the excess ap per core running Einstein that was just the switcher. I'll appreciate that.
Does that imply that Windows
)
Does that imply that Windows 7 ( ie. the API ) wasn't reporting the CPU capabilities sufficiently well enough for the needs of the switcher ?
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
RE: Does that imply that
)
Did the switcher rely on external tools/reports to identify the CPU, or did it test CPUID directly itself?