Title: BFI_INT with ATI7970 on Linux Post by: flok on October 01, 2012, 11:40:38 AM Hi,
I'm trying to mine using: - phoenix 1.7.5 - linux with kernel 3.4.0 - fglrx 1:12-6+point-1 - ATI7970 I'm trying to run it with BFI_INT but it complains "Failed to apply BFI_INT patch to kernel". Any idea how to fix this? (without it, it seems that -k poclbm AGGRESSION=13 is the most efficient: 595 MHash/s, but with loads of "Warning: work queue empty, miner is idle" unfortunately). Title: Re: BFI_INT with ATI7970 on Linux Post by: DiabloD3 on October 03, 2012, 11:52:10 PM Hi, I'm trying to mine using: - phoenix 1.7.5 - linux with kernel 3.4.0 - fglrx 1:12-6+point-1 - ATI7970 I'm trying to run it with BFI_INT but it complains "Failed to apply BFI_INT patch to kernel". Any idea how to fix this? (without it, it seems that -k poclbm AGGRESSION=13 is the most efficient: 595 MHash/s, but with loads of "Warning: work queue empty, miner is idle" unfortunately). GCN does not have that instruction, AMD's OpenCL compiler does the proper thing. Title: Re: BFI_INT with ATI7970 on Linux Post by: flok on October 04, 2012, 08:10:35 AM Hi, I'm trying to mine using: - phoenix 1.7.5 - linux with kernel 3.4.0 - fglrx 1:12-6+point-1 - ATI7970 I'm trying to run it with BFI_INT but it complains "Failed to apply BFI_INT patch to kernel". Any idea how to fix this? (without it, it seems that -k poclbm AGGRESSION=13 is the most efficient: 595 MHash/s, but with loads of "Warning: work queue empty, miner is idle" unfortunately). GCN does not have that instruction, AMD's OpenCL compiler does the proper thing. Can you elaborate on that? I did not install a package with gcn in the name, only pyopencl and fglrx. Title: Re: BFI_INT with ATI7970 on Linux Post by: DiabloD3 on October 04, 2012, 08:40:53 AM Hi, I'm trying to mine using: - phoenix 1.7.5 - linux with kernel 3.4.0 - fglrx 1:12-6+point-1 - ATI7970 I'm trying to run it with BFI_INT but it complains "Failed to apply BFI_INT patch to kernel". Any idea how to fix this? (without it, it seems that -k poclbm AGGRESSION=13 is the most efficient: 595 MHash/s, but with loads of "Warning: work queue empty, miner is idle" unfortunately). GCN does not have that instruction, AMD's OpenCL compiler does the proper thing. Can you elaborate on that? I did not install a package with gcn in the name, only pyopencl and fglrx. GCN is your hardware, 79xx. |