-
Notifications
You must be signed in to change notification settings - Fork 189
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Os x port #30
Comments
I could help you if you showed me the errors you see. |
well.. successful build but sa-solver fails now. ./sa-solver -v |
see my changes on https://github.com/justvanbloom/silentarmy/tree/mac |
Weird. Your system's OpenCL compiler requires definining prototypes? Trying adding them and see what happens... |
ok i will try. but i think its another problem also. global variables must have a constant address space qualifier. but when i add __constant nothing changes. have you a hint? |
Sorry I don't know. Maybe your OpenCL compiler is confused by the syntax? Try removing the whole "__attribute((reqd_work_group_size(64, 1, 1)))" |
ok, great steps forward now make runs with only a few warnings now. have inlcuded ocldump on my fork perhaps this helps
so now when i ran
i get Solving default all-zero 140-byte header |
Hi! I got it working on OSX (hackintosh), at least the solver runs but i am not sure if i have correct results. And i am still seeing the "pipe closed by peer" issue. To get rid of the constant address space error, I removed the define for the KERNEL_ROUND(N) completely and just wrote the 7 method declarations by hand instead. ./sa-solver --list ./sa-solver --nonces 10 --use 1 Solving default all-zero 140-byte header |
Wohoo! Jo let us work on this. |
Or post the funcs you mean here... |
Well it seems that the CPU works, but the GPU (tested on a GTX970 and ATI Radeon 6750) does not produce correct solutions at the moment. |
in the first moment i thougt ok, but then also saw Share above target failures. |
When testing, should cpu and gpu devices produce the same result? So at the moment is seems cpu is okay, gpu is false. |
jvb-MacBook-Pro:silentarmy oliverfolz$ ./sa-solver --nonce 1 -v -v --use 1 2>&1 | grep Soln: jvb-MacBook-Pro:silentarmy oliverfolz$ ./sa-solver --nonce 1 -v -v --use 0 2>&1 | grep Soln: working? |
From one tester: Devices on platform "Apple": Total 4.6 sol/s [dev1 5.2] 9 shares It is very strange, like a gtx 970 should provide 50 sol/s |
no, using the gpu i get a different Solution: hackpro:silentarmy matthiasposch$ ./sa-solver --nonce 1 -v -v --use 1 2>&1 | grep Soln: ./sa-solver --nonce 20 --use 1 (GTX 970) so if 2 instances were running on the cpu, 40 Sol/s could be possible. But still my version does not calculate correctly. i´ll have to look into it. It seems that your version provides correct results on the gpu? |
Jep. Mine is fine. |
Just clone this repo and do tests plz. https://github.com/justvanbloom/silentarmy/tree/beta1mac?files=1 |
I see you guys are making progress, nice :) As long as you produce the same solutions as |
Thx! Yes. Verified and working. Mined pover 1000 sol/s on diffrent pools so far. All accepted shares. But i do not stop. I have the feeleing there is more room. |
No i am still having the same issue that my gpu is not delivering correct results. |
Maybe i can port the opencl kernel to metal. |
I am posting my -v -v -v -v output, probably one of you can spot why the results are different (the numbers are a myth to me for now :). Should this output be exactly the same on gpu and cpu? CPU: GPU: |
Cpu in mac version (latest v5 silentarmy) is broken. Gpu works fine. |
What type of system do you use to mine- a real mac ? When i find time i will test the kernel under linux. |
macbook pro, imac and hackintosh. |
Using linux everything works fine and i get 45 Sols/s on my gtx 970 It seems that cpp (clang) does not like the ## concatenation within a #define. If you make sure that cpp-6 (the brew version) of the preprocessor is used during build, you could revert to the original input.cl kernel. see https://github.com/mposch/silentarmy/blob/master/Makefile |
I see @justvanbloom said maybe i can port the opencl kernel to metal. I want to know is it feasible? thx |
@justvanbloom How u solve Mac "make" err issue ? And How to find path of LIBOPENCL? My whole Mac has no libOpenCL.so |
Dear,
I'll do the os x port.
Changed compiler structs and opencl headers but get a bunch of errors. Any clue?
The text was updated successfully, but these errors were encountered: