Closed GoogleCodeExporter closed 9 years ago
sudo make GPU=true
cat Cuda/CudaCore.cu > Cuda.cu
cat Cuda/CudaPkcs5.cu >> Cuda.cu
cat Cuda/CudaRmd160.cu >> Cuda.cu
cat Cuda/CudaAes.cu >> Cuda.cu
cat Cuda/CudaXts.cu >> Cuda.cu
/usr/local/cuda/bin/nvcc -c Cuda.cu -I./Common/ -I./Crypto/ -I./Cuda/ -I./Main/
-I./ -lm -D_GPU_ --opencc-options -OPT:Olimit=0 -o Cuda.o
./Cuda.cu(189): Warning: Cannot tell what pointer points to, assuming global
memory space
./Cuda.cu(196): Warning: Cannot tell what pointer points to, assuming global
memory space
./Cuda.cu(189): Warning: Cannot tell what pointer points to, assuming global
memory space
./Cuda.cu(196): Warning: Cannot tell what pointer points to, assuming global
memory space
./Cuda.cu(1164): Warning: Cannot tell what pointer points to, assuming global
memory space
/usr/local/cuda/bin/nvcc -o truecrack Cuda.o Volumes.o Crc.o Endian.o Core.o
Charset.o Utils.o Main.o -I./Common/ -I./Crypto/ -I./Cuda/ -I./Main/ -I./ -lm
-D_GPU_
Core.o: In function `core':
Core.c:(.text+0x211): undefined reference to `cpu_Core'
collect2: ld returned 1 exit status
make: *** [truecrack] Error 1
Original comment by marcandr...@gmail.com
on 23 May 2012 at 7:13
The problem is that nvcc don't pass the right parameters to the preprocessor.
If you want to compile the sources in gpu mode, the nvcc should pass to the
preprocessor the label _GPU_. The purpose of this label is compile only the
pieces of code optimized for GPU and not for CPU. If the label _GPU_ is set,
Core.c uses the function `cuda_Core()`; instead the function `cpu_Core()` is
called.
I test with different version of cuda, different version of the kernel (from
2.6.x to the most recent version) and it worked well. So i propose to retry the
installation of nvidia cuda and use an older version of gcc (some version of
cuda don't support the recent gcc versions).
Anyway in a future version of truecrack, i will distinguish the CPU and GPU
version as different programs to avoid this kind of problem with preprocessor.
Original comment by luck87
on 19 Jun 2012 at 11:07
[deleted comment]
i recently forgot my overcomplicated truecrypt password and i was wondering if
there is a way to run this with windows 7?
Original comment by doriangray216
on 22 Jul 2012 at 11:28
cat Cuda/CudaXts.cu >> Cuda.cu
/usr/local/cuda/bin/nvcc -c Cuda.cu -I./Common/ -I./Crypto/ -I./Cuda/ -I./Main/
-I./ -lm -D_GPU_ --opencc-options -OPT:Olimit=0 -o Cuda.o
./Cuda.cu(189): Warning: Cannot tell what pointer points to, assuming global
memory space
./Cuda.cu(196): Warning: Cannot tell what pointer points to, assuming global
memory space
./Cuda.cu(189): Warning: Cannot tell what pointer points to, assuming global
memory space
./Cuda.cu(196): Warning: Cannot tell what pointer points to, assuming global
memory space
./Cuda.cu(1164): Warning: Cannot tell what pointer points to, assuming global
memory space
/usr/local/cuda/bin/nvcc -o truecrack Cuda.o Volumes.o Crc.o Endian.o Core.o
Charset.o Utils.o Main.o -I./Common/ -I./Crypto/ -I./Cuda/ -I./Main/ -I./ -lm
-D_GPU_
Core.o: In function `core':
Core.c:(.text+0x211): undefined reference to `cpu_Core'
collect2: ld returned 1 exit status
make: *** [truecrack] Error 1
This is solved in the following manner....
make clean
make GPU=false
.... Let it finish
mv truecrack truecrack-CPU
make clean
make GPU=true
.... Let it finish
mv truecrack truecrack-GPU
No core issues.... the real core issue is due to parts of the code has been
compiled for different execution models and should have been recompiled as the
command line switches have changed.
Hope this helps
Tom
Original comment by tomsmith...@gmail.com
on 10 Oct 2012 at 3:18
I notice a solution to the second part of the problem with the error status,
but I am not sure of what the solution to the first part about the warning is,
could you explain for me please. I compiled in both modes and tested on a
volume and the CPU mode returned a hit on the password but the GPU mode
returned a not defined result, I noticed as above the only difference in both
compilations is the warning about the pointers, any help please.
Original comment by yehshuam...@gmail.com
on 23 Nov 2012 at 8:35
Original issue reported on code.google.com by
marcandr...@gmail.com
on 23 May 2012 at 7:12