Releases: Chia-Network/bladebit
v2.0.0-beta1
Bladebit v2.0.0-beta1
- Fixed regression with the
-n
parameter. - Fixed issue which caused
-b 128
to serialize plots incorrectly in Phase 3. - Enabled 128 buckets (via
-b 128
).
Full Changelog: v2.0.0-alpha2...v2.0.0-beta1
v2.0.0-alpha2
Bladebit v2.0.0-alpha2
Bladebit2 alpha 2 adds optimizations to Phase 1 and reduces the required RAM to hold all of temp2 in-cache by roughly half.
64 buckets option (-b 64
) has also been added which requires rougly 11 GiB and will perform better in general.
All of temp2's data can be kept in RAM by specifying a cache of 99G and enabling alternating I/O via -a --cache 99G
(assuming 64 buckets used and a standard block size of 4096).
Plotting with all temp2 in-RAM on a 128G machine
diskplot -f <farmer_public_key> -c <pool_contract_address> diskplot -a --cache 99G -t1 <temp1_path> <output_path>
Use bladebit diskplot -h
for more info on plotting and more information on each of the diskplot
options, such as per-phase thread control.
Known Issues
- In-RAM plotting (
memplot
command) is currently disabled. -n
parameters is not supported yet and results in memory leaks and other undefined behavior.- Setting buckets to
128
serializes plot files incorrectly, resulting in invalid plots (currently disabled). - Larger than usual file system block sizes may crash on Phase 1.
- Only 1 instance can be used given the same
--temp1/2
directories because the file names are the same. For the time being different subdirectories can be used for multi-instance, if needed, which should not be the case.
Full Changelog: v2.0.0-alpha1...v2.0.0-alpha2
v2.0.0-alpha1
Bladebit v2.0.0-alpha1
Initial alpha release of Bladebit 2 with support for disk plotting
Use bladebit --help
and bladebit help diskplot
for more information and usage for disk plotting.
For a basic usage instructions, see the (in-progress) Bladebit 2 User Guide.
Known issues
- In-RAM plotting (
ramplot
command) is currently disabled. -n
parameters is not supported yet and results in memory leaks and other undefined behavior.- Setting buckets to
128
serializes plot files incorrectly, resulting in invalid plots. - Setting buckets to
1024
crashes on Phase 3. - Larger than usual file system block sizes may crash on Phase 1.
memtest
andiotest
do not interleave memory across NUMA nodes.
Full Changelog: v1.2.4...v2.0.0-alpha1
v1.2.4
What's Changed
- Fixes issues where some platforms showed the wrong version number.
Full Changelog: v1.2.3...v1.2.4
v1.2.3
v1.2.2
v1.2.0
v1.1.1
Fixed bug where --no-numa
would incorrectly disable cpu affinity instead (thanks to @malventano for figuring this out).
v1.1.0
Version 1.1.0
Changes:
- Output buffering has been disabled (thanks @mocksoul)
- Phase 1 sorting has been optimized with performance boost on average of 20 seconds+
- Allow displaying memo with the
--show-memo
switch. - Specify a memo for debugging with the
--memo
switch. - Disable automatic CPU affinity with the
--no-cpu-affinity
switch. Seebladebit --help
for more info on what this might be used for.