Merge branch 'main' into feat/vlad/refactor-from-affine

.github/workflows/cpp_cuda.yml

@@ -73,6 +73,8 @@ jobs:
             build_args: -DEXT_FIELD=ON
           - name: stark252
             build_args: -DEXT_FIELD=OFF
+          - name: m31
+            build_args: -DEXT_FIELD=ON
     steps:
     - name: Checkout Repo
       uses: actions/checkout@v4

.github/workflows/rust.yml

@@ -62,8 +62,8 @@ jobs:
       # We need to limit the number of threads to avoid running out of memory on weaker machines
       # ignored tests are polynomial tests. Since they conflict with NTT tests, they are executed separately
       run: |
-        cargo test --workspace --exclude icicle-babybear --exclude icicle-stark252 --release --verbose --features=g2 -- --test-threads=2 --ignored
-        cargo test --workspace --exclude icicle-babybear --exclude icicle-stark252 --release --verbose --features=g2 -- --test-threads=2
+        cargo test --workspace --exclude icicle-babybear --exclude icicle-stark252 --exclude icicle-m31 --release --verbose --features=g2 -- --test-threads=2 --ignored
+        cargo test --workspace --exclude icicle-babybear --exclude icicle-stark252 --exclude icicle-m31 --release --verbose --features=g2 -- --test-threads=2
 
     - name: Run baby bear tests
       working-directory: ./wrappers/rust/icicle-fields/icicle-babybear
@@ -79,6 +79,13 @@ jobs:
         cargo test --release --verbose -- --ignored
         cargo test --release --verbose
 
+    - name: Run m31 tests
+      working-directory: ./wrappers/rust/icicle-fields/icicle-m31
+      if: needs.check-changed-files.outputs.rust == 'true' || needs.check-changed-files.outputs.cpp_cuda == 'true'
+      run: |
+        cargo test --release --verbose -- --ignored
+        cargo test --release --verbose
+
   # build-windows:
   #   name: Build on Windows
   #   runs-on: windows-2022

icicle/cmake/FieldsCommon.cmake

@@ -1,5 +1,5 @@
 function(check_field)
-  set(SUPPORTED_FIELDS babybear;stark252)
+  set(SUPPORTED_FIELDS babybear;stark252;m31)
 
   set(IS_FIELD_SUPPORTED FALSE)
   set(I 1000)

icicle/include/api/babybear.h

@@ -19,36 +19,48 @@
 extern "C" cudaError_t babybear_extension_ntt_cuda(
   const babybear::extension_t* input, int size, ntt::NTTDir dir, ntt::NTTConfig<babybear::scalar_t>& config, babybear::extension_t* output);
 
-extern "C" cudaError_t babybear_create_poseidon2_constants_cuda(
-  int width,
-  int alpha,
-  int internal_rounds,
-  int external_rounds,
-  const babybear::scalar_t* round_constants,
-  const babybear::scalar_t* internal_matrix_diag,
-  poseidon2::MdsType mds_type,
-  poseidon2::DiffusionStrategy diffusion,
-  device_context::DeviceContext& ctx,
-  poseidon2::Poseidon2Constants<babybear::scalar_t>* poseidon_constants);
+extern "C" cudaError_t babybear_initialize_domain(
+  babybear::scalar_t* primitive_root, device_context::DeviceContext& ctx, bool fast_twiddles_mode);
 
-extern "C" cudaError_t babybear_init_poseidon2_constants_cuda(
-  int width,
-  poseidon2::MdsType mds_type,
-  poseidon2::DiffusionStrategy diffusion,
+extern "C" cudaError_t babybear_ntt_cuda(
+  const babybear::scalar_t* input, int size, ntt::NTTDir dir, ntt::NTTConfig<babybear::scalar_t>& config, babybear::scalar_t* output);
+
+extern "C" cudaError_t babybear_release_domain(device_context::DeviceContext& ctx);
+
+extern "C" void babybear_generate_scalars(babybear::scalar_t* scalars, int size);
+
+extern "C" cudaError_t babybear_scalar_convert_montgomery(
+  babybear::scalar_t* d_inout, size_t n, bool is_into, device_context::DeviceContext& ctx);
+
+extern "C" cudaError_t babybear_extension_mul_cuda(
+  babybear::extension_t* vec_a, babybear::extension_t* vec_b, int n, vec_ops::VecOpsConfig& config, babybear::extension_t* result);
+
+extern "C" cudaError_t babybear_extension_add_cuda(
+  babybear::extension_t* vec_a, babybear::extension_t* vec_b, int n, vec_ops::VecOpsConfig& config, babybear::extension_t* result);
+
+extern "C" cudaError_t babybear_extension_accumulate_cuda(
+  babybear::extension_t* vec_a, babybear::extension_t* vec_b, int n, vec_ops::VecOpsConfig& config);
+
+extern "C" cudaError_t babybear_extension_sub_cuda(
+  babybear::extension_t* vec_a, babybear::extension_t* vec_b, int n, vec_ops::VecOpsConfig& config, babybear::extension_t* result);
+
+extern "C" cudaError_t babybear_extension_transpose_matrix_cuda(
+  const babybear::extension_t* input,
+  uint32_t row_size,
+  uint32_t column_size,
+  babybear::extension_t* output,
   device_context::DeviceContext& ctx,
-  poseidon2::Poseidon2Constants<babybear::scalar_t>* poseidon_constants);
+  bool on_device,
+  bool is_async);
 
-extern "C" cudaError_t babybear_poseidon2_hash_cuda(
-  const babybear::scalar_t* input,
-  babybear::scalar_t* output,
-  int number_of_states,
-  int width,
-  const poseidon2::Poseidon2Constants<babybear::scalar_t>& constants,
-  poseidon2::Poseidon2Config& config);
+extern "C" cudaError_t babybear_extension_bit_reverse_cuda(
+  const babybear::extension_t* input, uint64_t n, vec_ops::BitReverseConfig& config, babybear::extension_t* output);
 
-extern "C" cudaError_t babybear_release_poseidon2_constants_cuda(
-  poseidon2::Poseidon2Constants<babybear::scalar_t>* constants,
-  device_context::DeviceContext& ctx);
+
+extern "C" void babybear_extension_generate_scalars(babybear::extension_t* scalars, int size);
+
+extern "C" cudaError_t babybear_extension_scalar_convert_montgomery(
+  babybear::extension_t* d_inout, size_t n, bool is_into, device_context::DeviceContext& ctx);
 
 extern "C" cudaError_t babybear_mul_cuda(
   babybear::scalar_t* vec_a, babybear::scalar_t* vec_b, int n, vec_ops::VecOpsConfig& config, babybear::scalar_t* result);
@@ -72,45 +84,38 @@ extern "C" cudaError_t babybear_transpose_matrix_cuda(
   bool is_async);
 
 extern "C" cudaError_t babybear_bit_reverse_cuda(
-  const babybear::scalar_t* input,
-  uint64_t n,
-  vec_ops::BitReverseConfig& config,
-  babybear::scalar_t* output);
-
-extern "C" void babybear_generate_scalars(babybear::scalar_t* scalars, int size);
-
-extern "C" cudaError_t babybear_scalar_convert_montgomery(
-  babybear::scalar_t* d_inout, size_t n, bool is_into, device_context::DeviceContext& ctx);
-
-extern "C" cudaError_t babybear_initialize_domain(
-  babybear::scalar_t* primitive_root, device_context::DeviceContext& ctx, bool fast_twiddles_mode);
-
-extern "C" cudaError_t babybear_ntt_cuda(
-  const babybear::scalar_t* input, int size, ntt::NTTDir dir, ntt::NTTConfig<babybear::scalar_t>& config, babybear::scalar_t* output);
-
-extern "C" cudaError_t babybear_release_domain(device_context::DeviceContext& ctx);
-
-extern "C" void babybear_extension_generate_scalars(babybear::extension_t* scalars, int size);
+  const babybear::scalar_t* input, uint64_t n, vec_ops::BitReverseConfig& config, babybear::scalar_t* output);
 
-extern "C" cudaError_t babybear_extension_scalar_convert_montgomery(
-  babybear::extension_t* d_inout, size_t n, bool is_into, device_context::DeviceContext& ctx);
 
-extern "C" cudaError_t babybear_extension_mul_cuda(
-  babybear::extension_t* vec_a, babybear::extension_t* vec_b, int n, vec_ops::VecOpsConfig& config, babybear::extension_t* result);
+extern "C" cudaError_t babybear_create_poseidon2_constants_cuda(
+  int width,
+  int alpha,
+  int internal_rounds,
+  int external_rounds,
+  const babybear::scalar_t* round_constants,
+  const babybear::scalar_t* internal_matrix_diag,
+  poseidon2::MdsType mds_type,
+  poseidon2::DiffusionStrategy diffusion,
+  device_context::DeviceContext& ctx,
+  poseidon2::Poseidon2Constants<babybear::scalar_t>* poseidon_constants);
 
-extern "C" cudaError_t babybear_extension_add_cuda(
-  babybear::extension_t* vec_a, babybear::extension_t* vec_b, int n, vec_ops::VecOpsConfig& config, babybear::extension_t* result);
+extern "C" cudaError_t babybear_init_poseidon2_constants_cuda(
+  int width,
+  poseidon2::MdsType mds_type,
+  poseidon2::DiffusionStrategy diffusion,
+  device_context::DeviceContext& ctx,
+  poseidon2::Poseidon2Constants<babybear::scalar_t>* poseidon_constants);
 
-extern "C" cudaError_t babybear_extension_sub_cuda(
-  babybear::extension_t* vec_a, babybear::extension_t* vec_b, int n, vec_ops::VecOpsConfig& config, babybear::extension_t* result);
+extern "C" cudaError_t babybear_poseidon2_hash_cuda(
+  const babybear::scalar_t* input,
+  babybear::scalar_t* output,
+  int number_of_states,
+  int width,
+  const poseidon2::Poseidon2Constants<babybear::scalar_t>& constants,
+  poseidon2::Poseidon2Config& config);
 
-extern "C" cudaError_t babybear_extension_transpose_matrix_cuda(
-  const babybear::extension_t* input,
-  uint32_t row_size,
-  uint32_t column_size,
-  babybear::extension_t* output,
-  device_context::DeviceContext& ctx,
-  bool on_device,
-  bool is_async);
+extern "C" cudaError_t babybear_release_poseidon2_constants_cuda(
+  poseidon2::Poseidon2Constants<babybear::scalar_t>* constants,
+  device_context::DeviceContext& ctx);
 
 #endif

icicle/include/api/bls12_377.h

@@ -16,6 +16,20 @@
 #include "poseidon/poseidon.cuh"
 #include "poseidon/tree/merkle.cuh"
 
+extern "C" bool bls12_377_g2_eq(bls12_377::g2_projective_t* point1, bls12_377::g2_projective_t* point2);
+
+extern "C" void bls12_377_g2_to_affine(bls12_377::g2_projective_t* point, bls12_377::g2_affine_t* point_out);
+
+extern "C" void bls12_377_g2_generate_projective_points(bls12_377::g2_projective_t* points, int size);
+
+extern "C" void bls12_377_g2_generate_affine_points(bls12_377::g2_affine_t* points, int size);
+
+extern "C" cudaError_t bls12_377_g2_affine_convert_montgomery(
+  bls12_377::g2_affine_t* d_inout, size_t n, bool is_into, device_context::DeviceContext& ctx);
+
+extern "C" cudaError_t bls12_377_g2_projective_convert_montgomery(
+  bls12_377::g2_projective_t* d_inout, size_t n, bool is_into, device_context::DeviceContext& ctx);
+
 extern "C" cudaError_t bls12_377_g2_precompute_msm_bases_cuda(
   bls12_377::g2_affine_t* bases,
   int msm_size,
@@ -34,20 +48,6 @@ extern "C" cudaError_t bls12_377_precompute_msm_bases_cuda(
 extern "C" cudaError_t bls12_377_msm_cuda(
   const bls12_377::scalar_t* scalars, const bls12_377::affine_t* points, int msm_size, msm::MSMConfig& config, bls12_377::projective_t* out);
 
-extern "C" bool bls12_377_g2_eq(bls12_377::g2_projective_t* point1, bls12_377::g2_projective_t* point2);
-
-extern "C" void bls12_377_g2_to_affine(bls12_377::g2_projective_t* point, bls12_377::g2_affine_t* point_out);
-
-extern "C" void bls12_377_g2_generate_projective_points(bls12_377::g2_projective_t* points, int size);
-
-extern "C" void bls12_377_g2_generate_affine_points(bls12_377::g2_affine_t* points, int size);
-
-extern "C" cudaError_t bls12_377_g2_affine_convert_montgomery(
-  bls12_377::g2_affine_t* d_inout, size_t n, bool is_into, device_context::DeviceContext& ctx);
-
-extern "C" cudaError_t bls12_377_g2_projective_convert_montgomery(
-  bls12_377::g2_projective_t* d_inout, size_t n, bool is_into, device_context::DeviceContext& ctx);
-
 extern "C" cudaError_t bls12_377_ecntt_cuda(
   const bls12_377::projective_t* input, int size, ntt::NTTDir dir, ntt::NTTConfig<bls12_377::scalar_t>& config, bls12_377::projective_t* output);
 
@@ -65,32 +65,18 @@ extern "C" cudaError_t bls12_377_affine_convert_montgomery(
 extern "C" cudaError_t bls12_377_projective_convert_montgomery(
   bls12_377::projective_t* d_inout, size_t n, bool is_into, device_context::DeviceContext& ctx);
 
-extern "C" cudaError_t bls12_377_create_optimized_poseidon_constants_cuda(
-  int arity,
-  int full_rounds_half,
-  int partial_rounds,
-  const bls12_377::scalar_t* constants,
-  device_context::DeviceContext& ctx,
-  poseidon::PoseidonConstants<bls12_377::scalar_t>* poseidon_constants);
+extern "C" cudaError_t bls12_377_initialize_domain(
+  bls12_377::scalar_t* primitive_root, device_context::DeviceContext& ctx, bool fast_twiddles_mode);
 
-extern "C" cudaError_t bls12_377_init_optimized_poseidon_constants_cuda(
-  int arity, device_context::DeviceContext& ctx, poseidon::PoseidonConstants<bls12_377::scalar_t>* constants);
+extern "C" cudaError_t bls12_377_ntt_cuda(
+  const bls12_377::scalar_t* input, int size, ntt::NTTDir dir, ntt::NTTConfig<bls12_377::scalar_t>& config, bls12_377::scalar_t* output);
 
-extern "C" cudaError_t bls12_377_poseidon_hash_cuda(
-  bls12_377::scalar_t* input,
-  bls12_377::scalar_t* output,
-  int number_of_states,
-  int arity,
-  const poseidon::PoseidonConstants<bls12_377::scalar_t>& constants,
-  poseidon::PoseidonConfig& config);
+extern "C" cudaError_t bls12_377_release_domain(device_context::DeviceContext& ctx);
 
-extern "C" cudaError_t bls12_377_build_poseidon_merkle_tree(
-  const bls12_377::scalar_t* leaves,
-  bls12_377::scalar_t* digests,
-  uint32_t height,
-  int arity,
-  poseidon::PoseidonConstants<bls12_377::scalar_t>& constants,
-  merkle::TreeBuilderConfig& config);
+extern "C" void bls12_377_generate_scalars(bls12_377::scalar_t* scalars, int size);
+
+extern "C" cudaError_t bls12_377_scalar_convert_montgomery(
+  bls12_377::scalar_t* d_inout, size_t n, bool is_into, device_context::DeviceContext& ctx);
 
 extern "C" cudaError_t bls12_377_mul_cuda(
   bls12_377::scalar_t* vec_a, bls12_377::scalar_t* vec_b, int n, vec_ops::VecOpsConfig& config, bls12_377::scalar_t* result);
@@ -114,22 +100,34 @@ extern "C" cudaError_t bls12_377_transpose_matrix_cuda(
   bool is_async);
 
 extern "C" cudaError_t bls12_377_bit_reverse_cuda(
-  const bls12_377::scalar_t* input,
-  uint64_t n,
-  vec_ops::BitReverseConfig& config,
-  bls12_377::scalar_t* output);
+  const bls12_377::scalar_t* input, uint64_t n, vec_ops::BitReverseConfig& config, bls12_377::scalar_t* output);
 
-extern "C" void bls12_377_generate_scalars(bls12_377::scalar_t* scalars, int size);
 
-extern "C" cudaError_t bls12_377_scalar_convert_montgomery(
-  bls12_377::scalar_t* d_inout, size_t n, bool is_into, device_context::DeviceContext& ctx);
+extern "C" cudaError_t bls12_377_create_optimized_poseidon_constants_cuda(
+  int arity,
+  int full_rounds_half,
+  int partial_rounds,
+  const bls12_377::scalar_t* constants,
+  device_context::DeviceContext& ctx,
+  poseidon::PoseidonConstants<bls12_377::scalar_t>* poseidon_constants);
 
-extern "C" cudaError_t bls12_377_initialize_domain(
-  bls12_377::scalar_t* primitive_root, device_context::DeviceContext& ctx, bool fast_twiddles_mode);
+extern "C" cudaError_t bls12_377_init_optimized_poseidon_constants_cuda(
+  int arity, device_context::DeviceContext& ctx, poseidon::PoseidonConstants<bls12_377::scalar_t>* constants);
 
-extern "C" cudaError_t bls12_377_ntt_cuda(
-  const bls12_377::scalar_t* input, int size, ntt::NTTDir dir, ntt::NTTConfig<bls12_377::scalar_t>& config, bls12_377::scalar_t* output);
+extern "C" cudaError_t bls12_377_poseidon_hash_cuda(
+  bls12_377::scalar_t* input,
+  bls12_377::scalar_t* output,
+  int number_of_states,
+  int arity,
+  const poseidon::PoseidonConstants<bls12_377::scalar_t>& constants,
+  poseidon::PoseidonConfig& config);
 
-extern "C" cudaError_t bls12_377_release_domain(device_context::DeviceContext& ctx);
+extern "C" cudaError_t bls12_377_build_poseidon_merkle_tree(
+  const bls12_377::scalar_t* leaves,
+  bls12_377::scalar_t* digests,
+  uint32_t height,
+  int arity,
+  poseidon::PoseidonConstants<bls12_377::scalar_t>& constants,
+  merkle::TreeBuilderConfig& config);
 
 #endif