README
¶
DERO AstroBWT CPU Mining Proof-of-Work
AstroBWT Testnet
[AstroBWT Mainnet HardFork on block 4550555, March 7,2020. ~0200-GMT.](DERO HardFork on block 4550555, March 7,2020. ~0200-GMT.)
Table of Contents
AstroBWT BUILDING
go get -u github.com/deroproject/astrobwt/miner
SAMPLE OUTPUT
DERO AstroBWT Miner v0.01 alpha
CPU: Intel(R) Xeon(R) CPU E3-1270 v6 @ 3.80GHz PhysicalThreads:1
Threads Total Time Total Iterations Time/PoW Hash Rate/Sec
1 3.272996982s 100 32.729969ms 30.6
2 3.572288466s 200 17.861442ms 56.0
3 4.013980986s 300 13.379936ms 74.7
4 4.704899609s 400 11.762249ms 85.0
5 5.784798143s 500 11.569596ms 86.4
6 6.629462384s 600 11.049103ms 90.5
7 8.351780961s 700 11.931115ms 83.8
8 10.49473002s 800 13.118412ms 76.2
AstroBWT Pseudo CODE
1. Calulate SHA3-256 of input data
2. Expand data using Salsa20 cipher 69371 bytes
3. Calculate BWT of step 2
4. Calculate SHA3-256 of BWT data
5. Expand data using Salsa20 cipher 69371 + random number based on step 4
6. Calculate BWT of data from step 5
7. Calculate SHA3-256 of BWT data from step 6
Explaining AstroBWT
AstroBWT is not a product of the current cryptosphere approach. It has roots in Information Theory and the Compression Domains.
- AstroBWT is based on mathematical proofs and research, unlike many other CPU mining ALGOs.
- All current mining algorithms are static. To explain more simply, all current cryptocurrency mining algorithms follow data dependent branches, loops or conditions. Dero also built out a RandomX implementation in Golang that has been discarded for use but open sourced to the public RandomX in Golang.
- AstroBWT, as the name implies, has BWT at it's core. BWT has been in research for the last 3 decades and numerous optimization attempts of GPU/FPGA have taken place. For more information on that topic, please refer to literature in the Information Theory and Compression Domains. However, all known implementations to this date, could not deliver an improvement even twice that of CPUs.
- All major providers (such as INTEL, NVIDIA etc) have already provided optimized implementations of BWT. Since, BWT has been used quite often in the general information theory and compression domains, it has been a subject of intensive studies. https://software.intel.com/en-us/ipp-dev-reference-burrows-wheeler-transform
- In the coming months or years that AstroBWT is optimized or shown to have a significant performance boost on FPGAs, ASICs, or GPUs there would be a benefit for everyone. Such an advancement could even trigger the next revolution in Bioinformatics, Signal Processing, DNA Sequencing or other numerous domains where BWT is used. So, AstroBWT will not only serve as an ASIC/FPGA and GPU resisitant algorithm but it will aslo succeed in helping scientific research if it is optimized for these things.
Read more about research of BWT FPGA implementation here and local copy of this research pdf.
AstroBWT Hash Rates
1. Submit and compare hashes with other devices(On Github).:
2. Submit and compare hashes with other devices((https://benchmark.dero.network/).
| SNo. | CPU | Miner | Hash Rates | Threads | OS |
|---|---|---|---|---|---|
| 1 | AMD Ryzen Threadripper 1950X | Official | 560 H/s | - | Linux |
| 2 | Ryzen 7 2700x DDR4 2666 "Stock settings" | XMRig-5.9.0 | 610 H/s | - | Windows 10 |
| 3 | Qualcomm SDM845 Snapdragon 845 (10 nm) CPU Octa-core (4x2.8 GHz Kryo 385 Gold & 4x1.7 GHz Kryo 385 Silver) | XMRigCC | 220 H/s | - | Ubuntu |
| 4 | Ryzen 7 3800X 4.4Ghz | XMRig | 970 H/s | - | Linux |
Documentation
¶
Overview ¶
Package suffixarray implements substring search in logarithmic time using an in-memory suffix array.
Example use:
// create index for some data index := suffixarray.New(data) // lookup byte slice s offsets1 := index.Lookup(s, -1) // the list of all indices where s occurs in data offsets2 := index.Lookup(s, 3) // the list of at most 3 indices where s occurs in data
Index ¶
- Constants
- Variables
- func BWT(input []byte) ([]byte, int)
- func BWT_0alloc(input []byte, sa []int32, bwt []byte) int
- func BigEndian_Uint64(b []byte) uint64
- func FromSuffixArray(s []byte, sa []int, es byte) ([]byte, error)
- func InverseTransform(t []byte, es byte) []byte
- func POW(inputdata []byte) (outputhash [32]byte)
- func POW_0alloc(inputdata []byte) (outputhash [32]byte)
- func POW_optimized_v1(inputdata []byte, max_limit int) (outputhash [32]byte, success bool)
- func POW_optimized_v2(inputdata []byte, max_limit int, data *Data) (outputhash [32]byte, success bool)
- func SuffixArray(s []byte) []int
- func Transform(s []byte, es byte) ([]byte, error)
- type Data
- type Index
Constants ¶
const ALLOCATION_SIZE = MAX_LENGTH
const COUNTING_SORT_BITS uint64 = 10
const COUNTING_SORT_SIZE uint64 = 1 << COUNTING_SORT_BITS
const MAX_LENGTH int = 1024*1024 + stage1_length + 1024
Variables ¶
var ErrInvalidSuffixArray = errors.New("bwt: invalid suffix array")
ErrInvalidSuffixArray means length of sa is not equal to 1+len(s)
Functions ¶
func BWT_0alloc ¶
input byte sa should be len(input) + 1 result len len(input) + 1
func BigEndian_Uint64 ¶
func FromSuffixArray ¶
FromSuffixArray compute BWT from sa
func InverseTransform ¶
InverseTransform reverses the bwt to original byte slice. Not optimized yet.
func POW_0alloc ¶
func POW_optimized_v1 ¶
func POW_optimized_v2 ¶
Types ¶
type Index ¶
type Index struct {
// contains filtered or unexported fields
}
Index implements a suffix array for fast substring search.
func (*Index) Bytes ¶
Bytes returns the data over which the index was created. It must not be modified.
func (*Index) FindAllIndex ¶
FindAllIndex returns a sorted list of non-overlapping matches of the regular expression r, where a match is a pair of indices specifying the matched slice of x.Bytes(). If n < 0, all matches are returned in successive order. Otherwise, at most n matches are returned and they may not be successive. The result is nil if there are no matches, or if n == 0.
func (*Index) Lookup ¶
Lookup returns an unsorted list of at most n indices where the byte string s occurs in the indexed data. If n < 0, all occurrences are returned. The result is nil if s is empty, s is not found, or n == 0. Lookup time is O(log(N)*len(s) + len(result)) where N is the size of the indexed data.
Source Files