Test of different Java based SHA-256 hash implementations

hash_comp.groovy

@Grab(group='com.google.guava', module='guava', version='19.0')
@Grab(group='commons-codec', module='commons-codec', version='1.13')
@Grab(group='bouncycastle', module='bcprov-jdk15', version='140')

import groovy.transform.CompileStatic
import java.security.MessageDigest
import java.nio.charset.StandardCharsets
import com.google.common.hash.Hashing
import org.apache.commons.codec.digest.DigestUtils
import org.bouncycastle.crypto.digests.SHA256Digest

//test string for all hashing implementations
String test = "lwkjt23uy45pojsdf;lnwo45y23po5i;lknwe;lknasdflnqw3uo5"

String hashJava(String str){
	MessageDigest digest = MessageDigest.getInstance("SHA-256");
	byte[] encodedhash = digest.digest(str.getBytes(StandardCharsets.UTF_8));
	return bytesToHex(encodedhash)
}

@CompileStatic
String bytesToHex(byte[] hash) {
    StringBuffer hexString = new StringBuffer();
    for (int i = 0; i < hash.length; i++) {
    String hex = Integer.toHexString(0xff & hash[i]);
    if(hex.length() == 1) hexString.append('0');
        hexString.append(hex);
    }
    return hexString.toString();
}

@CompileStatic
String groovyHash(String str){
	str.digest('SHA-256')
}

@CompileStatic
String guavaHash(String str){
	Hashing.sha256().hashString(str, StandardCharsets.UTF_8).toString();
}

@CompileStatic
String apacheHash(String str){
	DigestUtils.sha256Hex(str)
}

@CompileStatic
String bouncyHash(String str){
	SHA256Digest digest = new SHA256Digest();
    byte[] keyByteArray = str.getBytes(StandardCharsets.UTF_8)
    byte[] output = new byte[digest.getDigestSize()];
    digest.update(keyByteArray, 0, keyByteArray.length);
    digest.doFinal(output, 0);
    bytesToHex(output)
}

long iterations = 100_0000
println "Hashing ${iterations} iterations of SHA-256"
//warm up, throw away 
for (int x=0; x<1000; x++) hashJava(test)
long start = System.currentTimeMillis()
for (int x=0; x<iterations; x++) hashJava(test)
long end = System.currentTimeMillis()
println ("time java: ${end-start}\t\t${iterations*1000/(end-start)} hashes/sec")

for (int x=0; x<1000; x++) groovyHash(test)
start = System.currentTimeMillis()
for (int x=0; x<iterations; x++) groovyHash(test)
end = System.currentTimeMillis()
println ("time groovy: ${end-start}\t${iterations*1000/(end-start)} hashes/sec")

for (int x=0; x<1000; x++) apacheHash(test)
start = System.currentTimeMillis()
for (int x=0; x<iterations; x++) apacheHash(test)
end = System.currentTimeMillis()
println ("time apache: ${end-start}\t\t${iterations*1000/(end-start)} hashes/sec")

for (int x=0; x<1000; x++) guavaHash(test)
start = System.currentTimeMillis()
for (int x=0; x<iterations; x++) guavaHash(test)
end = System.currentTimeMillis()
println ("time guava: ${end-start}\t\t${iterations*1000/(end-start)} hashes/sec")

for (int x=0; x<1000; x++) bouncyHash(test)
start = System.currentTimeMillis()
for (int x=0; x<iterations; x++) bouncyHash(test)
end = System.currentTimeMillis()
println ("time bouncy: ${end-start}\t\t${iterations*1000/(end-start)} hashes/sec")

I thought the same thing and moved them around. I ended up getting the same results -- and consistently the same results. I have also heard that Guava wraps standard, but there too I get better results. If you can find where they get similar results, that would be helpful.

I think bytesToHex function makes the performance for MessageDigest and commons-codec different. If make both hashJava and apacheHash return byte[] to avoid transforming to hex string. Then they can get similar results.

I did, same performance difference even in different order with different warms ups. Something different is happening even if same algo implementation.

On Wed, Nov 6, 2019 at 9:54 AM Marcel Hellkamp ***@***.***> wrote: Apache commons-codec does not implement its own MessageDigest algorithms, but uses the standard java implementation. Apache beeing more than twice as fast as java would not make any sense. Whatever your benchmark is measuring, it is not SHA256 performance. Your warm-up phase is probably way too short. Switching around the order in which the tests are run will significantly change the results. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <https://gist.github.com/a60d61a2cc3afba1e8813b338ecd1501?email_source=notifications&email_token=AAA6NSC5GRN6AIB6JXZJ6YLQSMAETA5CNFSM4JBEZVS2YY3PNVWWK3TUL52HS4DFVNDWS43UINXW23LFNZ2KUY3PNVWWK3TUL5UWJTQAF3XVO#gistcomment-3075927>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAA6NSEBED3UZUEREQIQB7LQSMAETANCNFSM4JBEZVSQ> .

-- Sent from my phone

Took advice to remove the String conversion portions here: https://gist.github.com/scoroberts/77cc2d5c28fa2faeb1626d8eb39ec8c8