Human readable changelog:
- More drastic improvements to ModMiner clock adjustment algorithm. Downclocking should be more conservative, and BFGMiner will even raise the clock speed when it's doing well.
- ModMiners are now kept busier (less idle time) and nonces are checked against the previous work to catch races. This may result in a higher utility (effective hashrate).
- Tested and fixed problems managing throttling BitFORCE Singles. An 880 MH/s Single that was throttling to 440 MH/s before (and throwing lots of garbled errors) is now working cleanly at 590 MH/s - a 34% improvement in hashrate!
- A bunch of annoying crashes were fixed:
- With high-share devices when network connectivity is lost.
- GPU mining not functional when any OpenCL platform fails.
- Occasional deadlock/hang during exit.
- New --debuglog option to include debug information in log file (see README) but not in main console.
- The scrypt code was updated - now ALL results from a GPU are checked to see if they're valid and if not, instead of being submitted where they are guaranteed to be rejected, they register as hardware errors. This will make tuning your scrypt parameters easier as HW error count will rise if you've set the parameters too high.
- Updated miner.php from Kano
Full changelog
- Bugfix: Define my_cancellable_getch in miner.h
- Escape " and \ when writing json config file
- miner.php allow a custom page section to select all fields with '*' - e.g. to create a STATS section on a custom page
- miner.php optional single rig totals (on by default)
- Bugfix: Initialize submitting mutex
- Bugfix: bitforce: Allocate enough space for FTDI description pointers
- Queue one request for each staged request removed, keeping the staged request count optimal at all times.
- Bugfix: Avoid cancelling threads while locks are held
- Set recognizable names on threads for debugging
- Bugfix: Don't keep making new get_work threads if all pools are dead
- Enable configuring submission thread limit with --submit-threads option
- Bugfix: Limit active submission threads to 0x40 so we don't overflow
- Bugfix: Properly handle switching to pools that aren't on the latest block, and warn if a pool actively switches to an old block
- Log more details of reasons in stale_work debug messages
- Failsafe against stale-on-arrival work: disable the pool
- Bugfix: Debug message should show "Work stale due to work restart" when it's not a share
- windows-build: Remove APP SDK section since it is no longer needed
- modminer: HACK: Let last_work handle the end of the work, and start the next one immediately
- Bugfix: modminer: Remove erroneous "else" statement, to fix hashrate reporting
- README: Document user group required for FPGAs on Gentoo and Ubuntu
- BFGMiner-specific README adjustments
- Bugfix: opencl: Ignore error getting device ids from platforms unless they are explicitly chosen
- New --debuglog option to include debug info in stderr logfile even if not in the console
- Bumped down debhelper compatibility reqs so that this will build on Lucid.
- Updated to match packaging changes.
- Switched to native packages so we don't have to muck around creating fake upstream tarballs, and can easily generate minor versions for upload to Launchpad.
- Removed accidentally included debugging line.
- Minor version bump again because of launchpad. Will sort this out for next release.
- Updated to patch bitforce module issue on Debian/Ubuntu.
- Added local quilt config dir to ignore.
- modminer: Check nonce against previous work, in case of race
- Bugfix: Enable --kernel-path option if ModMiner or Ztex is enabled (even if no OpenCL)
- Bugfix: Escape backslashes and double-quotes in strings that rightfully may have them, when writing JSON config file
- Clean object (.o) and dependency (.d) files out of source tree
- Bugfix: bitforce: Don't count hashes that never happened due to throttling
- Bugfix: Deal with serial_open timeout maximum (25.5s)
- fpgautils: Linux only supports uint8_t decisecond values for timeouts, so use uint8_t for timeout value; this gets smart compilers to throw warnings when overflowed in some cases
- bitforce: Reduce serial timeout to 25 seconds (was 30) and increase job long timeout to 25 seconds (was 15) to handle throttling gracefully
- modminer: Add debug info to API extra device stats
- modminer: Raise clock speed when there's only good nonces for a while
- modminer: Only print clock speed adjustments when they actually change
- modminer: Increase tolerance for bad nonces to 2%
- modminer: Reset bad-nonce ratio measurement when the clock speed changes
- Bugfix: bitforce: Include the correct device id in "garbled response" warning
- ADL: Add attribution and disclaimer to interfaces
- Cleaned out refs to AMD SDKs.
- Updated README about debian packaging, changelog with minor version bump to work around Launchpad reqs.
- Updated changelog with Ubuntu release specific version, needed to build for multiple releases. Also stripped out ADL SDK stuff in the build rules.
- Initial work to adjust debian packaging from cgminer. Should build correctly now with pbuilder/pdebuild, and include docs.
- Adapt miner code to free ADL structures
- Import free ADL interfaces
- Include scrypt.h in Makefile.
- Fix windows bitforce build.
- Convert the serial autodetect functions to use int instead of char to enumerate devices.
- Deuglify windows autodetect code for BFL.
- There is no point zeroing temperature in BFL if we fail to get a response, and we should register it as a HW error, suggesting throttling.
- Update SCRYPT README with information about HW errors.
- Use the scrypt CPU code to confirm results from OCL code, and mark failures as HW errors, making it easier to tune scrypt parameters.
- We may as well leave one curl still available per pool instead of reaping the last one.
- Display reaped debug message outside mutex lock to avoid recursive locking.
- api.c update API start message and include port number
- miner.php ignore arg when readonly
- miner.php allow pool inputs: delete, addpool, poolpriority
- bitforce: Reopen on communication error
- Bugfix: Calculate hw err percent for the affected FPGA only
- make-release: Adapt to new autogen by using NOCONFIGURE var