mirror of
https://github.com/albertobsd/keyhunt.git
synced 2025-03-17 21:31:49 +01:00
I get mixed result about the speed performance: - bP Table creation is 4 or 5 times faster - address and rmd160 get the same speed. WHY? - xpoint get less speed. WTF? - bsgs not implemented yet. I need to move every mode to his own funtion to see if the xpoint speed get the double speed as spected. If nothing of this works, i will only keep the changes about the bP table creation (the only good change at the momment)
20 lines
1.1 KiB
Makefile
20 lines
1.1 KiB
Makefile
default:
|
|
gcc -O3 -c bloom/bloom.c -o bloom.o
|
|
g++ -O3 -c sha256/sha256.c -o sha256.o
|
|
gcc -O3 -c base58/base58.c -o base58.o
|
|
gcc -O3 -c rmd160/rmd160.c -o rmd160.o
|
|
gcc -O3 -c sha3/sha3.c -o sha3.o
|
|
gcc -O3 -c xxhash/xxhash.c -o xxhash.o
|
|
g++ -O3 -c util.c -o util.o
|
|
g++ -m64 -mssse3 -Wno-unused-result -Wno-write-strings -O2 -c secp256k1/Int.cpp -o Int.o
|
|
g++ -m64 -mssse3 -Wno-unused-result -Wno-write-strings -O2 -c secp256k1/Point.cpp -o Point.o
|
|
g++ -m64 -mssse3 -Wno-unused-result -Wno-write-strings -O2 -c secp256k1/SECP256K1.cpp -o SECP256K1.o
|
|
g++ -m64 -mssse3 -Wno-unused-result -Wno-write-strings -O2 -c secp256k1/IntMod.cpp -o IntMod.o
|
|
g++ -m64 -mssse3 -Wno-unused-result -Wno-write-strings -O2 -c secp256k1/Random.cpp -o Random.o
|
|
g++ -m64 -mssse3 -Wno-unused-result -Wno-write-strings -O2 -c secp256k1/IntGroup.cpp -o IntGroup.o
|
|
g++ -o keyhunt keyhunt.c base58.o rmd160.o sha256.o bloom.o xxhash.o util.o Int.o Point.o SECP256K1.o IntMod.o Random.o IntGroup.o -lgmp -lm -lpthread
|
|
#gcc -O3 hexcharstoraw.c -o hexcharstoraw util.o -lm
|
|
g++ -o bPfile bPfile.c util.o -lgmp -lm
|
|
clean:
|
|
rm -r *.o
|