Use all best practice to squeeze a few more TPS out of Prover (#38)

Cargo.toml

@@ -42,3 +42,12 @@ tonic-prost-build = "0.14.1"
 
 #[dev.dependencies]
 criterion = { version = "0.7.0", features = ["async_tokio"] }
+
+[profile.release]
+lto = "fat"
+codegen-units = 1
+opt-level = 3
+debug = false
+panic = "abort"
+strip = true
+incremental = false

Makefile

@@ -0,0 +1,15 @@
+NATIVE_FLAGS = -C target-cpu=native
+
+.PHONY: build
+build:
+	cargo build --release $(ARGS)
+
+RAYON_NUM_THREADS ?= 64
+PROOF_SERVICE_COUNT ?= 64
+PROOF_COUNT ?= 64
+
+.PHONY: bench
+bench:
+	RAYON_NUM_THREADS=$(RAYON_NUM_THREADS) PROOF_SERVICE_COUNT=$(PROOF_SERVICE_COUNT) PROOF_COUNT=$(PROOF_COUNT) RUSTFLAGS="$(NATIVE_FLAGS)" cargo bench $(ARGS)
+
+# Example: make bench RAYON_NUM_THREADS=128 PROOF_SERVICE_COUNT=32 PROOF_COUNT=8 ARGS="--bench prover_bench"

prover_cli/Cargo.toml

@@ -24,7 +24,3 @@ opentelemetry-otlp = { version = "0.30.0", features = [
     "tls-roots",
 ] }
 tracing-opentelemetry = "0.31.0"
-
-[profile.release]
-codegen-units = 1    # Do not split into multiple codegen units, improve performance, may compile slower
-lto = "fat"          # Enables Link Time Optimization, enabling more aggressive optimizations across the entire codebase