Add experimental use of icicle GPU Acceleration on pse feature

.github/workflows/ci.yml

@@ -2,45 +2,73 @@ name: Tests
 
 on:
   push:
-    branches: ["main", "release-0.3.0"]
-  pull_request:
     branches: ["main"]
+  pull_request:
+    branches: ["main", "develop", "community-edition", "release-*"]
 
 env:
   CARGO_TERM_COLOR: always
 
 jobs:
   build:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-latest-64core-256ram
 
     steps:
       - uses: actions/checkout@v3
       - name: Build
         run: cargo build --verbose
       - name: Run halo2-base tests
+        working-directory: "halo2-base"
         run: |
-          cd halo2-base
-          cargo test -- --test-threads=1
-          cd ..
-      - name: Run halo2-ecc tests MockProver
+          cargo test
+      - name: Run halo2-ecc tests (mock prover)
+        working-directory: "halo2-ecc"
         run: |
-          cd halo2-ecc
-          cargo test -- --test-threads=1 test_fp
-          cargo test -- test_ecc
-          cargo test -- test_secp256k1_ecdsa
-          cargo test -- test_ecdsa
-          cargo test -- test_ec_add
-          cargo test -- test_fixed_base_msm
-          cargo test -- test_msm
-          cargo test -- test_pairing
-          cd ..
-      - name: Run halo2-ecc tests real prover
+          cargo test --lib -- --skip bench
+      - name: Run halo2-ecc tests (real prover)
+        working-directory: "halo2-ecc"
         run: |
-          cd halo2-ecc
-          cargo test --release -- test_fp_assert_eq
+          mv configs/bn254/bench_fixed_msm.t.config configs/bn254/bench_fixed_msm.config
+          mv configs/bn254/bench_msm.t.config configs/bn254/bench_msm.config
+          mv configs/bn254/bench_pairing.t.config configs/bn254/bench_pairing.config
+          mv configs/secp256k1/bench_ecdsa.t.config configs/secp256k1/bench_ecdsa.config
           cargo test --release -- --nocapture bench_secp256k1_ecdsa
-          cargo test --release -- --nocapture bench_ec_add
           cargo test --release -- --nocapture bench_fixed_base_msm
           cargo test --release -- --nocapture bench_msm
           cargo test --release -- --nocapture bench_pairing
-          cd ..
+      - name: Run zkevm tests
+        working-directory: "hashes/zkevm"
+        run: |
+          cargo test packed_multi_keccak_prover::k_14
+          cargo t test_vanilla_keccak_kat_vectors
+
+  lint:
+    name: Lint
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v2
+
+      - name: Install toolchain
+        uses: actions-rs/toolchain@v1
+        with:
+          profile: minimal
+          override: false
+          components: rustfmt, clippy
+
+      - uses: Swatinem/rust-cache@v1
+        with:
+          cache-on-failure: true
+
+      - name: Run fmt
+        run: cargo fmt --all -- --check
+
+      - name: Run clippy
+        run: cargo clippy --all --all-targets -- -D warnings
+
+      - name: Generate Cargo.lock
+        run: cargo generate-lockfile
+
+      - name: Run cargo audit
+        uses: actions-rs/audit-check@v1
+        with:
+          token: ${{ secrets.GITHUB_TOKEN }}

.gitignore

@@ -8,6 +8,10 @@ Cargo.lock
 
 # These are backup files generated by rustfmt
 **/*.rs.bk
+
+# Local IDE configs
+.idea/
+.vscode/
 =======
 /target
 

Cargo.toml

@@ -1,13 +1,10 @@
 [workspace]
-members = [
-    "halo2-base",
-    "halo2-ecc",
-    "hashes/zkevm-keccak",
-]
+members = ["halo2-base", "halo2-ecc", "hashes/zkevm"]
+resolver = "2"
 
 [profile.dev]
 opt-level = 3
-debug = 1 # change to 0 or 2 for more or less debug info
+debug = 2              # change to 0 or 2 for more or less debug info
 overflow-checks = true
 incremental = true
 
@@ -28,7 +25,7 @@ codegen-units = 16
 opt-level = 3
 debug = false
 debug-assertions = false
-lto = "fat" 
+lto = "fat"
 # `codegen-units = 1` can lead to WORSE performance - always bench to find best profile for your machine!
 # codegen-units = 1
 panic = "unwind"
@@ -39,7 +36,6 @@ incremental = false
 inherits = "release"
 debug = true
 
-# patch so snark-verifier uses this crate's halo2-base 
 [patch."https://github.com/axiom-crypto/halo2-lib.git"]
-halo2-base = { path = "./halo2-base" }
-halo2-ecc = { path = "./halo2-ecc" }
+halo2-base = { path = "../halo2-lib/halo2-base" }
+halo2-ecc = { path = "../halo2-lib/halo2-ecc" }

README.md

@@ -1,6 +1,6 @@
 # halo2-lib
 
-This repository aims to provide basic primitives for writing zero-knowledge proof circuits using the [Halo 2](https://zcash.github.io/halo2/) proving stack.
+This repository aims to provide basic primitives for writing zero-knowledge proof circuits using the [Halo 2](https://zcash.github.io/halo2/) proving stack. To discuss or collaborate, join our community on [Telegram](https://t.me/halo2lib).
 
 ## Getting Started
 
@@ -45,6 +45,30 @@ cargo bench --bench inner_product
 
 These benchmarks use the `criterion` crate to run `create_proof` 10 times for statistical analysis. Note the benchmark circuits perform more than a one multiplication / inner product per circuit.
 
+### GPU Acceleration
+
+If you have access to NVIDIA GPUs, you can enable acceleration by building with the feature `halo2-icicle` and setting the following environment variable:
+
+```sh
+export ENABLE_ICICLE_GPU=true
+```
+
+GPU acceleration is provided by [Icicle](https://github.com/ingonyama-zk/icicle)
+
+To go back to running with CPU, the previous environment variable must be **unset** instead of being switched to a value of false:
+
+```sh
+unset ENABLE_ICICLE_GPU
+```
+
+> [!NOTE] 
+> Even with the above environment variable set, for circuits where k <= 8, icicle is only enabled in certain areas where batching MSMs will help; all other places will fallback to using CPU MSM. To change the value of `k` where icicle is enabled, you can set the environment variable `ICICLE_SMALL_CIRCUIT`.
+> 
+> Example: The following will cause icicle single MSM to be used throughout when k > 10 and CPU single MSM with certain locations using icicle batched MSM when k <= 10
+> ```sh
+> export ICICLE_SMALL_CIRCUIT=10
+> ```
+
 ## halo2-ecc
 
 This crate uses `halo2-base` to provide a library of elliptic curve cryptographic primitives. In particular, we support elliptic curves over base fields that are larger than the scalar field used in the proving system (e.g., `F_r` for bn254 when using Halo 2 with a KZG backend).
@@ -130,7 +154,7 @@ The test config file locations are (relative to `halo2-ecc` directory):
 | `test_msm` | `src/bn254/configs/msm_circuit.config` |
 | `test_pairing` | `src/bn254/configs/pairing_circuit.config` |
 
-### Benchmarks
+## Benchmarks
 
 We have tests that are actually benchmarks using the production Halo2 prover.
 As mentioned [above](#Configurable-Circuits), there are different configurations for each circuit that lead to _very_ different proving times. The following benchmarks will take a list of possible configurations and benchmark each one. The results are saved in a file in the `results` directory. We currently supply the configuration lists, which should provide optimal configurations for a given circuit degree `k` (however you can check versus the stdout suggestions to see if they really are optimal!).
@@ -172,7 +196,7 @@ cargo bench --bench fp_mul
 
 This run the same proof generation over 10 runs and collect the average. Each circuit has a fixed configuration chosen for optimal speed. These benchmarks are mostly for use in performance optimization.
 
-## Secp256k1 ECDSA
+### Secp256k1 ECDSA
 
 We provide benchmarks for ECDSA signature verification for the Secp256k1 curve on several different machines. All machines only use CPUs.
 
@@ -215,7 +239,7 @@ The other columns provide information about the [PLONKish arithmetization](https
 
 The r6a has a higher clock speed than the r6g.
 
-## BN254 Pairing
+### BN254 Pairing
 
 We provide benchmarks of the optimal Ate pairing for BN254 on several different machines. All machines only use CPUs.
 
@@ -258,7 +282,7 @@ The other columns provide information about the [PLONKish arithmetization](https
 
 The r6a has a higher clock speed than the r6g. We hypothesize that the Apple Silicon integrated memory leads to the faster performance on the M2 Max.
 
-## BN254 MSM
+### BN254 MSM
 
 We provide benchmarks of multi-scalar multiplication (MSM, multi-exp) with a batch size of `100` for BN254.
 
@@ -275,3 +299,17 @@ cargo test --release --no-default-features --features "halo2-axiom, mimalloc" --
 | 19  | 20         | 3                 | 1         | 32.6s               |
 | 20  | 11         | 2                 | 1         | 41.3s               |
 | 21  | 6          | 1                 | 1         | 51.9s               |
+
+## Projects built with `halo2-lib`
+
+- [Axiom](https://github.com/axiom-crypto/axiom-eth) -- Prove facts about Ethereum on-chain data via aggregate block header, account, and storage proofs.
+- [Proof of Email](https://github.com/zkemail/) -- Prove facts about emails with the same trust assumption as the email domain.
+  - [halo2-regex](https://github.com/zkemail/halo2-regex)
+  - [halo2-zk-email](https://github.com/zkemail/halo2-zk-email)
+  - [halo2-base64](https://github.com/zkemail/halo2-base64)
+  - [halo2-rsa](https://github.com/zkemail/halo2-rsa/tree/feat/new_bigint)
+- [halo2-fri-gadget](https://github.com/maxgillett/halo2-fri-gadget) -- FRI verifier in halo2.
+- [eth-voice-recovery](https://github.com/SoraSuegami/voice_recovery_circuit)
+- [zkevm tx-circuit](https://github.com/scroll-tech/zkevm-circuits/tree/develop/zkevm-circuits/src/tx_circuit)
+- [webauthn-halo2](https://github.com/zkwebauthn/webauthn-halo2) -- Proving and verifying WebAuthn with halo2.
+- [Fixed Point Arithmetic](https://github.com/DCMMC/halo2-scaffold/tree/main/src/gadget) -- Fixed point arithmetic library in halo2.

halo2-base/Cargo.toml

@@ -1,42 +1,48 @@
 [package]
 name = "halo2-base"
-version = "0.3.0"
+version = "0.4.0"
 edition = "2021"
 
 [dependencies]
-itertools = "0.10"
+itertools = "0.11"
 num-bigint = { version = "0.4", features = ["rand"] }
 num-integer = "0.1"
 num-traits = "0.2"
 rand_chacha = "0.3"
 rustc-hash = "1.1"
-ff = "0.12"
-rayon = "1.6.1"
+rayon = "1.8"
 serde = { version = "1.0", features = ["derive"] }
 serde_json = "1.0"
 log = "0.4"
+getset = "0.1.2"
+ark-std = { version = "0.3.0", features = ["print-trace"], optional = true }
 
 # Use Axiom's custom halo2 monorepo for faster proving when feature = "halo2-axiom" is on
-halo2_proofs_axiom = { git = "https://github.com/axiom-crypto/halo2.git", branch = "axiom/dev", package = "halo2_proofs", optional = true }
+halo2_proofs_axiom = { version = "0.4", package = "halo2-axiom", optional = true }
 # Use PSE halo2 and halo2curves for compatibility when feature = "halo2-pse" is on
-halo2_proofs = { git = "https://github.com/privacy-scaling-explorations/halo2.git", tag = "v2023_02_02", optional = true }
+halo2_proofs = { git = "https://github.com/ingonyama-zk/halo2", branch = "axiom-icicle", package = "halo2_proofs", optional = true }
 
+# This is Scroll's audited poseidon circuit. We only use it for the Native Poseidon spec. We do not use the halo2 circuit at all (and it wouldn't even work because the halo2_proofs tag is not compatbile).
+# We forked it to upgrade to ff v0.13 and removed the circuit module
+poseidon-rs = { package = "poseidon-primitives", version = "=0.1.1" }
 # plotting circuit layout
 plotters = { version = "0.3.0", optional = true }
-tabbycat = { version = "0.1", features = ["attributes"], optional = true }
 
 # test-utils
 rand = { version = "0.8", optional = true }
 
 [dev-dependencies]
 ark-std = { version = "0.3.0", features = ["print-trace"] }
 rand = "0.8"
-pprof = { version = "0.11", features = ["criterion", "flamegraph"] }
-criterion = "0.4"
+pprof = { version = "0.13", features = ["criterion", "flamegraph"] }
+criterion = "0.5.1"
 criterion-macro = "0.4"
-rayon = "1.6.1"
 test-case = "3.1.0"
+test-log = "0.2.12"
+env_logger = "0.10.0"
 proptest = "1.1.0"
+# native poseidon for testing
+pse-poseidon = { git = "https://github.com/axiom-crypto/pse-poseidon.git" }
 
 # memory allocation
 [target.'cfg(not(target_env = "msvc"))'.dependencies]
@@ -45,13 +51,16 @@ jemallocator = { version = "0.5", optional = true }
 mimalloc = { version = "0.1", default-features = false, optional = true }
 
 [features]
-default = ["halo2-axiom", "display"]
-dev-graph = ["halo2_proofs?/dev-graph", "halo2_proofs_axiom?/dev-graph", "plotters"]
-halo2-pse = ["halo2_proofs"]
+default = ["halo2-axiom", "display", "test-utils"]
+asm = ["halo2_proofs_axiom?/asm"]
+dev-graph = ["halo2_proofs/dev-graph", "plotters"] # only works with halo2-pse for now
+halo2-pse = ["halo2_proofs/circuit-params"]
+halo2-icicle = ["halo2_proofs/icicle_gpu", "halo2_proofs/circuit-params"]
 halo2-axiom = ["halo2_proofs_axiom"]
+halo2-axiom-icicle = ["halo2_proofs_axiom"]
 display = []
 profile = ["halo2_proofs_axiom?/profile"]
-test-utils = ["dep:rand"]
+test-utils = ["dep:rand", "ark-std"]
 
 [[bench]]
 name = "mul"
@@ -60,3 +69,7 @@ harness = false
 [[bench]]
 name = "inner_product"
 harness = false
+
+[[example]]
+name = "inner_product"
+required-features = ["test-utils"]