feat(bench): PoC of integration with zkalc (#425)

* feat(bench): PoC of integration with zkalc

* feat(bench): zkalc prepare for adding pairing benches - generic type resulution issue

* feat(bench): add pairing and G2 bench for zkalc and nimble build script

* fix nimble dependencies

* feat(bench-zkalc): polish and output json bench file

* fix(nimble): version gate task-level dependencies

* fix(deps): std/os instead of commandline for 1.6.x and gmp@#head

* fix(deps): .nimble and nimble have different versioning format

* fix(zkalc): workaround generic sandwich bug in Nim 1.6.x nim-lang/Nim#8677 nim-lang/Nim#11225

* fix(zkalc CI): skip zkalc build on 32-bit CI as it needs clang multilib

* fix(zkalc CI): skip Windows as Clang throws fatal error LNK1107

.github/workflows/ci.yml

@@ -262,7 +262,7 @@ jobs:
         run: |
           pacman -S --needed --noconfirm mingw-w64-x86_64-gmp mingw-w64-x86_64-llvm
           nimble refresh --verbose -y
-          nimble install --verbose -y gmp@#head jsony asynctools [email protected]
+          nimble install --verbose -y gmp@#head jsony asynctools [email protected] cliche
 
           cd constantine
           go mod download -modfile=go_test.mod
@@ -272,7 +272,7 @@ jobs:
         shell: bash
         run: |
           nimble refresh --verbose -y
-          nimble install --verbose -y gmp@#head jsony asynctools [email protected]
+          nimble install --verbose -y gmp@#head jsony asynctools [email protected] cliche
 
           cd constantine
           go mod download -modfile=go_test.mod
@@ -337,6 +337,24 @@ jobs:
           cd constantine
           CTT_ASM=0 cargo test -- --nocapture
 
+      - name: Compile Constantine Zkalc benchmark (no assembly)
+        # Skip 32-bit as that would need clang-multilib or -m32
+        # Skip Windows as clang throws fatal error LNK1107
+        if: matrix.target.BACKEND == 'NO_ASM' && matrix.target.cpu != 'i386' && runner.os != 'Windows'
+        shell: bash
+        run: |
+          cd constantine
+          CTT_ASM=0 nimble make_zkalc
+
+      - name: Compile Constantine Zkalc benchmark (with assembly)
+        # Skip 32-bit as that would need clang-multilib or -m32
+        # Skip Windows as clang throws fatal error LNK1107
+        if: matrix.target.BACKEND == 'ASM' && matrix.target.cpu != 'i386' && runner.os != 'Windows'
+        shell: bash
+        run: |
+          cd constantine
+          nimble make_zkalc
+
       - name: Run Constantine in-depth tests (Unix - with GMP, with Assembly)
         if: runner.os != 'Windows' && matrix.target.BACKEND == 'ASM'
         shell: bash

.gitignore

@@ -1,4 +1,4 @@
-# Executables shall be put in an ignored build/ directory
+# Build artifacts shall be put in an ignored build/ directory
 # Ignore dynamic, static libs and libtool archive files
 # -----------------------------------------------------------------------------------------
 build/
@@ -54,4 +54,5 @@ Cargo.lock
 # Perf artifacts
 # -----------------------------------------------------------------------------------------
 perf.data
-perf.data.old
+perf.data.old
+*.bench.json

benchmarks/bench_elliptic_parallel_template.nim

@@ -56,7 +56,7 @@ proc multiAddParallelBench*(EC: typedesc, numInputs: int, iters: int) =
 type BenchMsmContext*[EC] = object
   tp: Threadpool
   numInputs: int
-  coefs: seq[getBigInt(EC.F.Name, kScalarField)]
+  coefs: seq[getBigInt(EC.getName(), kScalarField)]
   points: seq[affine(EC)]
 
 proc createBenchMsmContext*(EC: typedesc, inputSizes: openArray[int]): BenchMsmContext[EC] =
@@ -66,6 +66,7 @@ proc createBenchMsmContext*(EC: typedesc, inputSizes: openArray[int]): BenchMsmC
   const bits = EC.getScalarField().bits()
   type ECaff = affine(EC)
 
+  result.numInputs = maxNumInputs
   result.points = newSeq[ECaff](maxNumInputs)
   result.coefs = newSeq[BigInt[bits]](maxNumInputs)
 
@@ -81,7 +82,7 @@ proc createBenchMsmContext*(EC: typedesc, inputSizes: openArray[int]): BenchMsmC
       var tmp = threadRng.random_unsafe(EC)
       tmp.clearCofactor()
       points[i].affine(tmp)
-      coefs[i] = rng.random_unsafe(BigInt[bits])
+      coefs[i] = threadRng.random_unsafe(BigInt[bits])
 
   let chunks = balancedChunksPrioNumber(0, maxNumInputs, result.tp.numThreads)