correct issue with rx5xx with most recent drivers

correct issue with most recent drivers

correct issue with most recent drivers

improve miner stability

improve miner stability in case of bad pool connection which could have prevent to receive properly the messages from the pool.

performance improvement

Improved hashrate on both mtp.cl and mtp_vega.cl

mtp.cl for RX5xx and RX4xx

1.69MH @1250MH/z on RX580
do not use this kernel for vega

mtp_vega
1.685MH @1250MHz on RX580
3.1MH/s @1350MHz on Vega 56

sgminer with mtp algo

fix problem with duplicate
clean kernel area
=> mtp for 580
=> mtp_vega for vega or radeon vii
=> mtp_nvidia for nvidia cards

sgminer with mtp

• fix network difficulty report

• fix share difficulty report

• fix share report, shares are not reported as untracked anymore

• fix HW report (solution not validated on cpu)

• small speed bump for vega in kernel mtp_nvidia4.cl (around 1.8MH/s on vega 56)

sgminer with mtp

• improve miner stability
• minor changes in kernel
• best kernel for 580 => --kernel mtp (1.5-1.6Mh/s; 1.1-1.2 on vega 56)
• best kernel for vega => --kernel mtp_nvidia (~1.5-1.6Mh/s on vega 56)

sgminer with mtp

• more stable sgminer core and less memory leak

• new kernels
  --kernel mtp
  --kernel mtp_vega (experimental kernel)
  --kernel mtp_nvidia (works best with nvidia cards)

• new option in command line (mandatory) --worksize ( --worksize 64 seems to works best for amd)

sgminer with mtp

kernel optimization. 50% (and more) on standard kernel, a little more on new kernel mtp_vega

program tends to crash a bit due to some issues in json (work in progress).

--kernel mtp for 580 and older model
--kernel mtp_vega for vega (this kernel seems also a bit faster than standard on 580)

sgminer with mtp

tested on 580

sgminer with mtp

correct problem with maximum buffer size allocation