Strange ptx error on compile fatal error in ptx: Arguments mismatch for instruction 'shl'

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi,
I have no idea what to do with this error, any suggestion would be very appreciated.
I’m running windows xp 32 bits

1>------ Rebuild All started: Project: cppIntegration, Configuration: Debug Win32 ------
1>Deleting intermediate and output files for project ‘cppIntegration’, configuration ‘Debug|Win32’
1>Compiling with CUDA Build Rule…
1>“C:\CUDA\bin\nvcc.exe” -arch sm_10 -ccbin “C:\Program Files\Microsoft Visual Studio 8\VC\bin” -Xcompiler “/EHsc /W3 /nologo /Od /Zi /MTd " -I"C:\CUDA\include” -I"…/…/common/inc" -maxrregcount=32 --compile -o “Debug\cppIntegration.cu.obj” “c:\Documents and Settings\All Users\Application Data\NVIDIA Corporation\NVIDIA GPU Computing SDK\C\src\aes one\cppIntegration.cu”
1>cppIntegration.cu
1>tmpxft_00000c58_00000000-3_cppIntegration.cudafe1.gpu
1>tmpxft_00000c58_00000000-8_cppIntegration.cudafe2.gpu
1>ptxas C:\DOCUME~1\RESEAR~2\LOCALS~1\Temp/tmpxft_00000c58_00000000-4_cppIntegration.ptx, line 312; error : Arguments mismatch for instruction ‘shl’
1>ptxas C:\DOCUME~1\RESEAR~2\LOCALS~1\Temp/tmpxft_00000c58_00000000-4_cppIntegration.ptx, line 313; error : Arguments mismatch for instruction ‘shr’
1>ptxas fatal : Ptx assembly aborted due to errors
1>Compiling…
1>cppIntegration_gold.cpp
1>main.cpp
1>Generating Code…
1>Linking…
1>LINK : fatal error LNK1181: cannot open input file ‘.\Debug\cppIntegration.cu.obj’
1>Build log was saved at “file://c:\Documents and Settings\All Users\Application Data\NVIDIA Corporation\NVIDIA GPU Computing SDK\C\src\aes one\Debug\BuildLog.htm”
1>cppIntegration - 1 error(s), 0 warning(s)
========== Rebuild All: 0 succeeded, 1 failed, 0 skipped ==========

global void criptare_aesctr(unsigned char* TX,unsigned char* IV,unsigned char* RK,unsigned char* SB,unsigned int maxim)
{

 unsigned int i = (blockIdx.y*gridDim.x+blockIdx.x)*blockDim.x+threadIdx.x;
 int j,r,ix;
 unsigned int cx = i*16;
 unsigned char temp[4];
 unsigned char state[16];

.....

TX[cx+ix]=state[ix];    //this row generates the error

}

2

Since this does not seem to involve handcoded PTX code, it certainly looks like a bug in the compiler, which appears to generate an invalid PTX instruction. If this happens with the CUDA 4.2 toolchain, please file a bug, attaching a self-contained repro case. A link to the reporting bug reporting form can be found on the registered developer website. If you are using an older CUDA version, I would suggest upgrading to 4.2. You may also want to check out the CUDA 5.0 preview (fair warning: alpha release quality software) if you are blocked with 4.2.

Thank you for your help, and sorry for the inconvenience.

3

It seems like the compiler won’t let me copy the local unsigned char array to the global TX memory, but if i put to state[16], which is out of bounds it’s ok for him… no error.
How can i copy the state to TX please…

global void criptare_aesctr(unsigned char* TX,unsigned char* IV,unsigned char* RK,unsigned char* SB,unsigned int maxim)
{

 unsigned int i = (blockIdx.y*gridDim.x+blockIdx.x)*blockDim.x+threadIdx.x;
 int j,r,ix;
 unsigned int cx = i*16;
 unsigned char temp[4];
 unsigned char state[16];


				 for(ix=0;ix<16;ix++){
					 state[ix]=state[ix]^TX[cx+ix];
					 //TX[cx+ix]=state[ix]^TX[cx+ix]; error if uncomment
					 TX[cx+ix]=state[16];
					 //TX[cx+ix]=state[ix]; error if uncomment
				 }
 }

}

4

Please file a bug against the compiler. The compiler team may be able to suggest a workaround once they have determined at the root cause for the issue you observe.