Refactor benchmarks

latticefold/Cargo.toml

@@ -21,6 +21,13 @@ thiserror = { workspace = true }
 num-bigint = { workspace = true }
 rayon = { version = "1.10.0", optional = true }
 
+
+[build-dependencies]
+serde = { version = "1.0.216", features = ["derive"] }
+toml = "0.8.19"
+quote = "1.0.37"
+proc-macro2 = "1.0.92"
+
 [features]
 default = [ "std" ]
 std = [ "ark-std/std", "cyclotomic-rings/std", "rand/std" ]
@@ -40,6 +47,12 @@ dhat-heap = []
 [profile.release]
 debug = 1
 
+[profile.bench]
+lto = false
+codegen-units = 16
+opt-level = 3
+rpath = true
+
 [lints.clippy]
 single_match = "warn"
 single_match_else = "warn"
@@ -60,6 +73,7 @@ implicit_clone = "warn"
 criterion = "0.5.1"
 dhat = "0.3.2"
 humansize = "2.1.3"
+lazy_static = "1.5.0"
 
 [[bench]]
 name = "ajtai"
@@ -69,10 +83,12 @@ harness = false
 name = "linearization"
 harness = false
 
+
 [[bench]]
 name = "decomposition"
 harness = false
 
+
 [[bench]]
 name = "folding"
 harness = false

latticefold/benches/README.md

@@ -0,0 +1,96 @@
+# Benchmarking with Environment Variables
+
+This benchmark suite allows users to selectively enable or disable certain benchmarks using environment variables. By leveraging environment variables, users can tailor the benchmark runs to specific configurations.
+
+---
+
+## Environment Variables
+
+The following environment variables can be set to **enable or filter benchmarks**:
+
+### Boolean Flags (Enable/Disable Filtering)
+- `GOLDILOCKS`
+- `STARK`
+- `BABYBEAR`
+- `FROG`
+
+- `PROVER`
+- `VERIFIER`
+- `AJTAI`
+
+- `LINEARIZATION`
+- `DECOMPOSITION`
+- `FOLDING`
+- `E2E`
+
+> **Default Behavior**:  
+If none of the flags for a group are set, all flags are enabled by default.
+
+### Optional Variables (Numeric Values)
+The following variables control specific parameters for benchmarks. If not set, they will be ignored:
+- `DURATION` (default: `50.0` seconds as floating point number) – Duration for benchmarks
+- `WARMUP` (default: `1.0` seconds as floating point number) – Warmup time
+
+---
+
+## How It Works
+- If none of `LINEARIZATION`, `DECOMPOSITION`, `FOLDING`, and `E2E` is set, all benchmarks would be run.
+- If some of them is set, only set benchmarks would be run.
+- Similarly, if none of `PROVER`, `VERIFIER` and `AJTAI` is set, all benchmarks would be run, otherwise,
+  only set benchmarks would be run.
+- `GOLDILOCKS`, `STARK`, `BABYBEAR` and `FROG` are used similarly.
+
+---
+
+## Examples
+
+### **Bash**
+
+1. **Run only `LINEARIZATION` benchmarks:**
+   ```bash
+   LINEARIZATION=1 cargo bench
+   ```
+
+2. **Run only `PROVER` `DECOMPOSITION` benchmarks:**
+   ```bash
+   PROVER=1 DECOMPOSITION=1 cargo bench
+   ```
+
+3. **Run benchmarks matching specific parameters for GOLDILOCKS ring:**
+   ```bash
+   KAPPA=10 GOLDILOCKS=1 L=8 GOLDILOCKS=1 cargo bench
+   ```
+
+---
+
+### **PowerShell**
+1. **Run only `LINEARIZATION` benchmarks:**
+   ```powershell
+   $env:LINEARIZATION=1; cargo bench
+   ```
+
+2. **Run only `PROVER` `DECOMPOSITION` benchmarks:**
+   ```powershell
+   $env:PROVER=1; $env:DECOMPOSITION=1; cargo bench
+   ```
+
+3. **Run benchmarks matching specific parameters for GOLDILOCKS ring:**
+   ```powershell
+   $env:KAPPA=10; $env:GOLDILOCKS=1; $env:L=8; $env:GOLDILOCKS=1; cargo bench
+   ```
+
+---
+
+## Numeric Parameters
+
+Numeric parameters can be used to filter benchmarks, if they are set they only benchmarks with matching parameter would be run
+
+- X_LEN 
+- KAPPA 
+- W 
+- WIT_LEN 
+- B 
+- L 
+- B_SMALL 
+- K
+