gIOTA

Client library for the IOTA reference implementation (IRI).

This branch supports FPGA based hardware POW accelerator

This library is still in flux and there maybe breaking changes.

Consider to use a dependency tool to use vendoring, e.g. godep, glide or govendor.

Refer to godoc for details.

Install

You will need C compiler for linux to compile PoW routine in C.

    $ go get -u github.com/iotaledger/giota

You will need C compiler and OpenCL environemnt(hardware and software) to compile PoW routine for GPU and need to add opencl tag when you build.

	$ go build -tags=gpu

Examples

import "github.com/iotaledger/giota"

//Trits
tritsFrom:=[]int8{1,-1,1,0,1,1,0,-1,0}
trits,err:=giota.ToTrits(tritsFrom)

//Trytes
trytes:=trits.Trytes()
trytesFrom:="ABCDEAAC9ACB9PO..."
trytes2,err:=giota.ToTrytes(trytesFrom)

//Hash
hash:=trytes.Hash()

//API
api := giota.NewAPI("http://localhost:14265", nil)
resp, err := api.FindTransactions([]Trytes{"DEXRPL...SJRU"})

///Address
index:=0
security:=2
adr,err:=giota.NewAddress(trytes,index,security) //without checksum.
adrWithChecksum := adr.WithChecksum() //adrWithChecksum is trytes type.

//transaction
tx,err:=giota.NewTransaction(trytes)
mwm := 14
if tx.HasValidNonce(mwm){...}
trytes2:=tx.trytes()

//create signature
key := giota.NewKey(seed, index, security)
norm := bundleHash.Normalize()
sign := giota.Sign(norm[:27], key[:6561/3])

//validate signature
if giota.ValidateSig(adr, []giota.Trytes{sign}, bundleHash) {...}

//send
trs := []giota.Transfer{
	giota.Transfer{
		Address: "KTXF...QTIWOWTY",
		Value:   20,
		Tag: "MOUDAMEPO",
	},
}
_, pow := giota.GetBestPoW()
bdl, err = giota.Send(api, seed, security, trs, mwm, pow)


// promote transaction
trs := []giota.Transfer{
	giota.Transfer{
		Address: "999...9999999",
		Value:   0,
		Tag: "PROMOTESPAM",
	},
}
tail := []giota.Trytes("NLN...TY99999")
_, pow := giota.GetBestPoW()
bdl, err = giota.Promote(api, tail, giota.Depth, trs, mwm, pow)

PoW(Proof of Work) Benchmarking

You can benchmark PoWs(by C,Go,SSE) by

    $ go test -v -run Pow

or if you want to add OpenCL PoW,

    $ go test -tags=gpu -v -run Pow

then it outputs like:

	$ go test -tags=gpu -v -run Pow
=== RUN   TestPowC
--- PASS: TestPowC (15.93s)
	pow_c_test.go:50: 1550 kH/sec on C PoW
=== RUN   TestPowCL
--- PASS: TestPowCL (17.45s)
	pow_cl_test.go:49: 332 kH/sec on GPU PoW
=== RUN   TestPowGo
--- PASS: TestPowGo (21.21s)
	pow_go_test.go:50: 1164 kH/sec on Go PoW
=== RUN   TestPowSSE
--- PASS: TestPowSSE (13.41s)
	pow_sse_test.go:52: 2292 kH/sec on SSE PoW

Note that in travis CI the result is:

=== RUN   TestPowSSE
--- PASS: TestPowSSE (2.73s)
	pow_sse_test.go:52: 12902 kH/sec on SSE PoW
=== RUN   TestPowSSE1
--- PASS: TestPowSSE1 (16.19s)
	pow_sse_test.go:59: 1900 kH/sec on SSE PoW
=== RUN   TestPowSSE32
--- PASS: TestPowSSE32 (1.36s)
	pow_sse_test.go:67: 16117 kH/sec on SSE PoW
=== RUN   TestPowSSE64
--- PASS: TestPowSSE64 (0.73s)
	pow_sse_test.go:75: 20226 kH/sec on SSE PoW

It gets over 20MH/s for 64 threads using SSE2.

Now IOTA uses Min Weight Magnitude = 15, which means 3^15≒14M Hashes are needed to finish PoW in average. So it takes just 14/20 < 0.7sec for 1 tx to do PoW.

TODO

• Multisig
• More tests :(

---

Released under the MIT License.