Is this something you're calling in the code, or is the top of the call stack when the app seg faults. Also, is this an app included with the SDK or is this something you wrote? I installed the SDK on XP64 and the apps that came with the SDK seemed to still work. However, I've gotten it to work in Ubuntu (with some hacking of course), so I've moved over to that in favor of the GNU development platform.
Hi!, I'm using Brook+ 1.3 on WinXP 64 (Visual Studio 2005, Radeon 4850), the
only problem I've found is that you can't build 32 bit aplications.
Have you installed driver 8.12? Which version of Visual Studio are you using?
Do the prebuilt binaries that come with the SDK work? (BROOK/samples/bin)
Originally posted by: Ceq Hi!, I'm using Brook+ 1.3 on WinXP 64 (Visual Studio 2005, Radeon 4850), the only problem I've found is that you can't build 32 bit aplications. Have you installed driver 8.12? Which version of Visual Studio are you using? Do the prebuilt binaries that come with the SDK work? (BROOK/samples/bin)
Yeah, I never got 32 bit apps to work either.
Brook+ doesn't ship 32 bit binaries with 64-bit installer.
Thank you all for your replies.
Actually this is a simple stream kernel that sets a=b. I was using it for testing the SDK flow.
I was using SDK1.2.1 without any issues.
The thing I did not mention is that I do not use VS for build. I use shell files to keep portability to Linux. The code compiles OK. I got the same problem with trying to build hello_brook.
I have a Phenom quad core,
I am using Microsoft CL "version 14.00.50727.42 for x64" with XP x64 SP2, with the following options:
" -c -Gd -Od -D_WIN64 -D_CRT_SECURE_NO_DEPRECATE -DWIN64 -D_AMD64_ -D_DEBUG -EHsc -MTd -w -nologo -Zi -Tp"
and Linker options "-MACHINE:AMD64 -DEBUG"
I think it is an issue with combination of brook_d.lib/brook.lib with the MTd/MDd options. I tried all options permutations.
Again, I had no problems with SDk 1.2.1, both for CAL & Brook+
I am using this with StreamFire 9170 with the latest 8.561 driver.
BTW, I got the brook + of SDK1.3 to work on 64b Linux (Scientific Linux 5.2) , but with slightly degraded kernel overheads
I also have the precompiled binary Brook+/sdk1.3 samples running OK
The probable reason of this faillure is that Brook+ libraries was built using /MT for release and /MD for debug. You might want to try changing these flags for your application.
Sorry, I was wrong they are using /MTd for debug. Try rebuilding Brook+ runtime on local system and see if it helps.
You mean building tle brook_d.lib & brook.lib again? This will not be easy on XP x64, since I do not use vs.
As you see, I am using the /MTd flag already in my code
Hi bayoumi, I answer you in this thread, this is my Visual Studio 2005 compiler configuration:
/O2 /Ob2 /Oi /Ot /GL /I "." /I "source" /I "source\include" /D "WIN32" /D "NDEBUG" /D "_CONSOLE" /FD /EHsc /MD /fp:fast /Fo"x64\Release\\" /Fd"x64\Release\vc80.pdb" /W3 /nologo /c /Wp64 /Zi /TP /errorReport: Prompt
And this is my linker configuration:
/OUT:"x64\Release\ShWa1L.exe" /INCREMENTAL /NOLOGO /LIBPATH:"." /LIBPATH:"E:\hd1\BROOK\\sdk\lib" /MANIFEST /MANIFESTFILE:"x64\Release\ShWa1L.exe.intermediate.manifest" /SUBSYSTEM:CONSOLE /HEAP:134217728,134217728 /STACK:134217728,134217728 /OPT:REF /OPT:ICF /LTCG /MACHINE:X64 /ERRORREPORT: PROMPT brook.lib kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib
If you find something wrong please let me know, I just used the same configuration from one old example.
I didn't need to rebuild Brook+ (altough I did it to find and report a bug)
Thank you for your reply. I was away of the office. I will try it and get back to you
I am still getting the same kind of errors. It is surprising the defining WIN64 instead of WIN32 makes the errors seem immediate, as opposed to WIN32 where I get the errors after few seconds. The runtime libraries seem to be 32b?
In all cases, I plan to switch back to sdk 1.2.1 till AMD releases a patch or an improved version