pow-miner-gpu
pow-miner-gpu copied to clipboard
After update openCL for 2x Boost the CL kernel contains error and too much warnings
It seems that newer CL kernel contains syntax errors for example like that: [ OpenCL: create context for platform #0 device #0 ATI RV770, max work group size is 256 ] [ 1][t 4][2022-01-05 01:16:56.9380746][opencl.cpp:127] [ OpenCL: ERROR ] "C:\Users\ADMINI~1\AppData\Local\Temp\OCL3E63.tmp.cl", line 67: warning: storage-class specifiers not allowed for routines static vec8u sha256_transform(vec16u data, vec8u state) ^
"C:\Users\ADMINI~1\AppData\Local\Temp\OCL3E63.tmp.cl", line 119: warning: storage-class specifiers not allowed for routines static vec8u sha256_transform_2(vec16u data, vec8u state) ^
"C:\Users\ADMINI~1\AppData\Local\Temp\OCL3E63.tmp.cl", line 399: warning:
nested comment is not allowed
//*/
^
"C:\Users\ADMINI~1\AppData\Local\Temp\OCL3E63.tmp.cl", line 413: warning: goto statement may cause irreducible control flow goto next_try; ^
"C:\Users\ADMINI~1\AppData\Local\Temp\OCL3E63.tmp.cl", line 419: error: function "atomic_add" declared implicitly if (atomic_add(&result->found, 1) == 0) ^
1 error detected in the compilation of "C:\Users\ADMINI~1\AppData\Local\Temp\OCL3E63.tmp.cl".
Internal error: clc compiler invocation failed.
So now the CL kernel is not working under some drivers and GPU platforms For example if we try to comment if (atomic_add(&result->found, 1) == 0) вот здесь ошибка
to
// if (atomic_add(&result->found, 1) == 0) result->found++;
Then the kernel starts working but hashrate is extremely low.