add benchmark of f2m_mul loop

Makefile

@@ -9,11 +9,14 @@ v3:
 v2:
 	./deps/v2/solidity/build/solc/solc --strict-assembly --optimize src/v2/benchmark.yul | awk '/Binary representation:/ { getline; print $0 }' | grep . > build/v2-f6m_mul_bench.bin
 	./deps/v2/solidity/build/solc/solc --strict-assembly --optimize src/v2/test.yul | awk '/Binary representation:/ { getline; print $0 }' | grep . > build/v2-test.bin
+	./deps/v2/solidity/build/solc/solc --strict-assembly --optimize src/v2/f2m_mul_benchmark.yul | awk '/Binary representation:/ { getline; print $0 }' | grep . > build/v2-f2m_mul_bench.bin
 v2_no_curve_params:
 	./deps/v2-no-curve-params/solidity/build/solc/solc --strict-assembly --optimize src/v2-no-curve-params/benchmark.yul | awk '/Binary representation:/ { getline; print $0 }' | grep . > build/v2-no-curve-params-f6m_mul_bench.bin
 	./deps/v2-no-curve-params/solidity/build/solc/solc --strict-assembly --optimize src/v2-no-curve-params/test.yul | awk '/Binary representation:/ { getline; print $0 }' | grep . > build/v2-no-curve-params-test.bin
 v3_no_curve_params:
 	./deps/v2-no-curve-params/solidity/build/solc/solc --strict-assembly --optimize src/v3-no-curve-params/benchmark.yul | awk '/Binary representation:/ { getline; print $0 }' | grep . > build/v3-no-curve-params-f6m_mul_bench.bin
+
+
 v1:
 	./deps/v1/solidity/build/solc/solc --strict-assembly --optimize src/v1/benchmark.yul | awk '/Binary representation:/ { getline; print $0 }' | grep . > build/v1-f6m_mul_bench.bin
 

benchmark.sh

@@ -9,3 +9,7 @@ set -e
 
 ./deps/v3/evmone/build/bin/evmone-bench --benchmark_format=json --benchmark_color=false --benchmark_min_time=5 build/v3-f6m_mul_bench.bin 00 74229fc665e6c3f4401905c1a454ea57c8931739d05a074fd60400f19684d680a9e1305c25f13613dcc6cdd6e6e57d0800000000000000000000000000000000 | python3 catch_evmone_bench_err.py
 ./deps/v3-no-curve-params/evmone/build/bin/evmone-bench --benchmark_format=json --benchmark_color=false --benchmark_min_time=5 build/v3-no-curve-params-f6m_mul_bench.bin 00 74229fc665e6c3f4401905c1a454ea57c8931739d05a074fd60400f19684d680a9e1305c25f13613dcc6cdd6e6e57d0800000000000000000000000000000000 | python3 catch_evmone_bench_err.py
+./deps/v3-no-curve-params/evmone/build/bin/evmone-bench --benchmark_format=json --benchmark_color=false --benchmark_min_time=5 build/v3-no-curve-params-f2m_mul_bench.bin 00 aff6e01e71b0709e0f7f6528dd81ea41aed732761d712e6884576a7b04a9490e64be863464e7b928ab4a709d2e2e7b0e00000000000000000000000000000000bb65fcee5ae7b04cef799771e1fd0e597838d8088633a9f99a3e9c62d4d48989b3ef3e3451ad30b40dedb7b6f928ae1400000000000000000000000000000000 | python3 catch_evmone_bench_err.py
+
+# test yul f2m_mul loop
+./deps/v2/evmone/build/bin/evmone-bench --benchmark_format=json --benchmark_color=false --benchmark_min_time=5 build/v2-f2m_mul_bench.bin 00 aff6e01e71b0709e0f7f6528dd81ea41aed732761d712e6884576a7b04a9490e64be863464e7b928ab4a709d2e2e7b0e00000000000000000000000000000000bb65fcee5ae7b04cef799771e1fd0e597838d8088633a9f99a3e9c62d4d48989b3ef3e3451ad30b40dedb7b6f928ae1400000000000000000000000000000000 | python3 catch_evmone_bench_err.py

build/v2-f2m_mul_bench.bin

@@ -0,0 +1 @@
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

build/v3-no-curve-params-f2m_mul_bench.bin

@@ -0,0 +1 @@
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

src/v2/f2m_mul_benchmark.yul

@@ -0,0 +1,83 @@
+{
+    function zero512(dst) {
+        mstore(dst, 0)
+	mstore(add(dst, 32), 0)
+    }
+
+    // r <- x * y
+    function f2m_mul(x, y, r, modulus, inv, mem) {
+
+        let tmp := add(mem, 64)
+        let tmp2 := add(tmp, 64)
+        let zero := add(tmp2, 64)
+	zero512(zero)
+
+        // TODO also cache x0y0 calculation
+
+        // tmp2 = x1y1
+        mulmodmont384(tmp2, add(x, 64), add(y, 64), modulus, inv)
+
+        //r0 = mulNR(tmp2)
+        submod384(r, zero, tmp2, modulus)
+
+        // tmp = x0y0
+        mulmodmont384(tmp, x, y, modulus, inv)
+
+        //r0 = r0 + tmp (x0y0)
+        addmod384(r, r, tmp, modulus)
+
+        // r1 -------------------------------------
+
+        // r1 = ((y0 + y1) * (x0 + x1)) - ((x0 * y0) + (x1 * y1))
+
+        // tmp2 <- tmp (x0 * y0) + tmp2 (x1 * y1)
+        addmod384(tmp2, tmp, tmp2, modulus)
+
+        // tmp = y0 + y1
+        addmod384(tmp, y, add(y, 64), modulus)
+
+        // r1 = x0 + x1
+        addmod384(add(r, 64), x, add(x, 64), modulus)
+
+        // r1 <- r1 (x0 + x1) * tmp (y0 + y1)
+        mulmodmont384(add(r, 64), add(r, 64), tmp, modulus, inv)
+
+        // r1 = r1 [(x0 + x1) * (y0 + y1)] - tmp2 [(x0 * y0) + (x1 * y1)]
+        submod384(add(r, 64), add(r, 64), tmp2, modulus)
+    }
+
+    function test_f2m_mul_loop() {
+            let bls12_mod := msize()
+            mstore(bls12_mod,          0xabaafffffffffeb9ffff53b1feffab1e24f6b0f6a0d23067bf1285f3844b7764)
+            mstore(add(bls12_mod, 32), 0xd7ac4b43b6a71b4b9ae67f39ea11011a00000000000000000000000000000000)
+
+            let inv :=         0x89f3fffcfffcfffd
+            let point1_a := add(bls12_mod, 64)
+            mstore(point1_a,          0x8f2990f3e598f5b1b8f480a3c388306bc023fac151c0104d13ec3aa181599402)
+            mstore(add(point1_a, 32), 0x72d1c8c528a1ce3bcaa280a8e735aa0d00000000000000000000000000000000)
+            mstore(add(point1_a, 64), 0x992d7a27906d4cd530b23a7e8c48c0778f8653fbc3332d63db24339d8bc65d7e)
+            mstore(add(point1_a, 96), 0xe83b6e91c6550f5aceab102e88e9180900000000000000000000000000000000)
+
+            let point1_b := add(point1_a, 128)
+            mstore(point1_b,          0x7299907146816f08c4c6a394e91374ed6ff3618a57358cfb124ee6ab4c560e5c)
+            mstore(add(point1_b, 32), 0xac40700b41e2ee8674680728f0c5a61800000000000000000000000000000000)
+            mstore(add(point1_b, 64), 0x0fd77f62b39eb952a0f8d21cec1f93b1d62dd7923aa86882ddf7dd4d3532b0b7)
+            mstore(add(point1_b, 96), 0xede8f3fc89fa4a79574067e2d9a9d20000000000000000000000000000000000)
+
+            let f2m_result := add(point1_b, 384) // allocate memory past bls12_mod
+
+            let mem := add(f2m_result, 128)
+
+            f2m_mul(point1_a, point1_b, f2m_result, bls12_mod, inv, mem)
+
+            let i := 0
+            for {} lt(i, 1350) {i := add(i, 1)} {
+                f2m_mul(point1_a, f2m_result, point1_b, bls12_mod, inv, mem)
+                f2m_mul(point1_b, f2m_result, point1_a, bls12_mod, inv, mem)
+            }
+
+            return(f2m_result, 128)
+    }
+
+    test_f2m_mul_loop()
+}