validation errors

[KlausT]

Looks like the 64bit build with CUDA 7.5 is broken. At least under Windows.

[ZeroASIC]

CUDA 7.5 gives a memcpy error on Ubuntu 16.04 64bit. I tracked down a bug for a different project that suggests it might be fixed in CUDA 8.0.

Here's the error I get: /usr/include/string.h: In function ‘void* mempcpyinline(void, const void_, size_t)’: /usr/include/string.h:652:42: error: ‘memcpy’ was not declared in this scope return (char *) memcpy (dest, src, n) + __n;

Here's the issue for the other project https://github.com/BVLC/caffe/issues/4046

[KlausT]

I think you get this error because the gcc version is not compatible with your CUDA version.

[ZeroASIC]

i just got it to compile by adding -D_FORCE_INLINES to line 394 of the makefile: NVCC_GENCODE = -D_FORCE_INLINES -gencode=arch=compute_30,code=\"sm_30,compute_30\" -gencode=arch=compute_35,code=\"sm_35,compute_35\" -gencode=arch=compute_50,code=\"sm_50\" -gencode=arch=compute_52,code=\"sm_52,compute_52\" all: cpuminer-config.h

[KlausT]

Looks like CUDA 7.5 is generally broken. CUDA 6.5 works without errors, but is not as fast for Pascal cards. Since Nvidia has released CUDA 8 now, I will test this.

[KlausT]

ok, I have tested CUDA 8 now. Works great, no validation errors. Using CUDA 8 is recommended now, at least for Maxwell and Pascal cards.

[ceozero]

ccminer-cryptonight is not the cudart32_80.dll... where is the cudart32_80.dll? thank you.

[KlausT]

the dll is build into the exe It should work without it