Author

Topic: cudaMiner patch for x86_64 linux (Read 1095 times)

newbie
Activity: 41
Merit: 0
April 20, 2013, 04:34:45 PM
#6
Little patch for compiling & running cudaminer on native 64bit linux.

mk.junkyard.one.pl/cudaminer-2013.04.17-64bit.patch.gz

To apply to cudaMiner source download the file & issue command:
Code:
zcat cudaminer-2013.04.17-64bit.patch.gz | patch -p1

I only tested the salsa kernel, and i'm not sure if titan kernel will work.

thank you for the patch. Works well on RHEL6.4 clone with CUDA 5 on GT430 .
running "as-is" 21Khash/s
newbie
Activity: 41
Merit: 0
April 20, 2013, 08:11:12 AM
#5
@dbabo: the patch is for 2013.04.17, not 2013.04.14 version.
ehhh shoot. me is stupid $)
thank you.
But after patching the kernel i got it to compile very late last night.
Thank you Misiolap!
newbie
Activity: 14
Merit: 0
April 20, 2013, 04:19:15 AM
#4
@dbabo: the patch is for 2013.04.17, not 2013.04.14 version.
newbie
Activity: 21
Merit: 0
April 19, 2013, 11:29:50 PM
#3
Be careful with using -O3; Some commonly in use versions of GCC drop about 20% efficiency with one of the vector optimizations added by -O3

Check it against -O2, and if you get a performance hit from using -O3, read the GCC manpage and manually disable the added O3 optimizations one by one until you find what's hitting you.
newbie
Activity: 41
Merit: 0
April 19, 2013, 10:23:32 PM
#2
Little patch for compiling & running cudaminer on native 64bit linux.

mk.junkyard.one.pl/cudaminer-2013.04.17-64bit.patch.gz

To apply to cudaMiner source download the file & issue command:
Code:
zcat cudaminer-2013.04.17-64bit.patch.gz | patch -p1

I only tested the salsa kernel, and i'm not sure if titan kernel will work.

not sure. This is what i have:
Code:
unzip cudaminer-src-2013.04.14.zip && cp cudaminer-2013.04.17-64bit.patch.gz ./cudaminer-src-2013.04.14 && cd cudaminer-src-2013.04.14

pwd
/home/az/LTC/cudaminer-2013-04-14/cudaminer-src-2013.04.14
[az@server cudaminer-src-2013.04.14]$ zcat cudaminer-2013.04.17-64bit.patch.gz | patch -p1
patching file configure.sh
patching file salsa_kernel.cu
Hunk #1 FAILED at 50.
Hunk #2 succeeded at 146 (offset -3 lines).
Hunk #3 succeeded at 170 (offset -3 lines).
Hunk #4 succeeded at 187 (offset -3 lines).
Hunk #5 succeeded at 200 (offset -3 lines).
Hunk #6 succeeded at 208 (offset -3 lines).
Hunk #7 succeeded at 225 (offset -3 lines).
Hunk #8 succeeded at 234 (offset -3 lines).
Hunk #9 FAILED at 245.
Hunk #10 FAILED at 269.
Hunk #11 FAILED at 280.
Hunk #12 succeeded at 283 (offset -7 lines).
Hunk #13 succeeded at 295 (offset -7 lines).
Hunk #14 succeeded at 314 (offset -7 lines).
Hunk #15 succeeded at 332 (offset -7 lines).
Hunk #16 succeeded at 364 (offset -7 lines).
Hunk #17 succeeded at 384 (offset -7 lines).
Hunk #18 succeeded at 392 (offset -7 lines).
Hunk #19 FAILED at 410.
Hunk #20 FAILED at 440.
Hunk #21 FAILED at 448.
Hunk #22 FAILED at 464.
Hunk #23 FAILED at 474.
Hunk #24 succeeded at 472 (offset -15 lines).
Hunk #25 FAILED at 722.
Hunk #26 FAILED at 738.
Hunk #27 FAILED at 1031.
Hunk #28 FAILED at 1047.
13 out of 28 hunks FAILED -- saving rejects to file salsa_kernel.cu.rej
patching file titan_kernel.cu
Hunk #2 FAILED at 35.
Hunk #3 succeeded at 81 with fuzz 2 (offset -3 lines).
Hunk #4 FAILED at 137.
Hunk #5 succeeded at 183 with fuzz 2 (offset -36 lines).
Hunk #6 FAILED at 274.
3 out of 6 hunks FAILED -- saving rejects to file titan_kernel.cu.rej
patching file util.cpp


and after :
Code:
chmod +x configure autotools.sh configure.sh  
[az@server cudaminer-src-2013.04.14]$ ./autotools.sh
[az@server cudaminer-src-2013.04.14]$ ./configure.sh

make

usr/local/cuda/bin/nvcc -O3 -arch=compute_10 --maxrregcount=124 --ptxas-options=-v -I./compat/jansson -o salsa_kernel.o -c salsa_kernel.cu
salsa_kernel.cu(269): error: no instance of overloaded function "tex2D" matches the argument list
            argument types are: (texture, unsigned int, unsigned int)

salsa_kernel.cu(278): error: no instance of overloaded function "tex2D" matches the argument list
            argument types are: (texture, unsigned int, unsigned int)
          detected during instantiation of "void scrypt_core_kernel_specialB_tex(uint32_t *) [with WARPS_PER_BLOCK=1]"
(1138): here


Additional info:
Code:
echo $LD_LIBRARY_PATH
/usr/local/cuda/lib64:/usr/local/cuda/lib:/usr/local/cuda/livb64:/usr/local/cuda/livb64:
[az@server cudaminer-src-2013.04.14]$ more /etc/ld.so.conf.d/nvidia.conf
/usr/lib64/nvidia
/usr/lib64/vdpau
/usr/lib/nvidia
/usr/lib/vdpau
[az@server cudaminer-src-2013.04.14]$ more /etc/ld.so.conf.d/cuda.conf
/usr/local/cuda/lib64
/usr/local/cuda/lib

rpm -qa | grep 686
glibc-2.12-1.107.el6.i686
nss-softokn-freebl-3.12.9-11.el6.i686
glibc-devel-2.12-1.107.el6.i686


newbie
Activity: 14
Merit: 0
April 18, 2013, 06:22:37 PM
#1
Little patch for compiling & running cudaminer on native 64bit linux.

mk.junkyard.one.pl/cudaminer-2013.04.17-64bit.patch.gz

To apply to cudaMiner source download the file & issue command:
Code:
zcat cudaminer-2013.04.17-64bit.patch.gz | patch -p1

I only tested the salsa kernel, and i'm not sure if titan kernel will work.
Jump to: