so close

Cargo.toml

@@ -29,7 +29,7 @@ cuda-mobile = []
 
 [dependencies]
 semolina = "~0.1.3"
-sppark = { git = "https://github.com/lurk-lab/sppark.git", branch = "preallocated-msm" }
+sppark = { git = "https://github.com/lurk-lab/sppark.git", branch = "pushing-the-limit" }
 pasta_curves = { git = "https://github.com/lurk-lab/pasta_curves", branch = "dev", version = ">=0.3.1, <=0.5", features = ["repr-c"] }
 paste = "1.0.14"
 

cuda/pallas.cu

@@ -15,10 +15,17 @@ typedef bucket_t::affine_t affine_t;
 typedef vesta_t scalar_t;
 
 #include <msm/pippenger.cuh>
+#include <spmvm/spmvm.cuh>
 
 #ifndef __CUDA_ARCH__
 
-extern "C" void drop_msm_context_pallas(msm_context_t<affine_t::mem_t> &ref) {
+extern "C" RustError spmvm_pallas(scalar_t *scalars, size_t nscalars)
+{
+    return double_scalars<scalar_t>(scalars, nscalars);
+}
+
+extern "C" void drop_msm_context_pallas(msm_context_t<affine_t::mem_t> &ref)
+{
     CUDA_OK(cudaFree(ref.d_points));
 }
 
@@ -35,7 +42,7 @@ extern "C" RustError cuda_pippenger_pallas(point_t *out, const affine_t points[]
 }
 
 extern "C" RustError cuda_pippenger_pallas_with(point_t *out, msm_context_t<affine_t::mem_t> *msm_context, size_t npoints,
-                                                       const scalar_t scalars[])
+                                                const scalar_t scalars[])
 {
     return mult_pippenger_with<bucket_t, point_t, affine_t, scalar_t>(out, msm_context, npoints, scalars);
 }

examples/spmvm.rs

@@ -0,0 +1,41 @@
+// Copyright Supranational LLC
+// Licensed under the Apache License, Version 2.0, see LICENSE for details.
+// SPDX-License-Identifier: Apache-2.0
+
+use std::time::Instant;
+
+use pasta_curves::group::ff::PrimeField;
+use pasta_msm::spmvm::double_pallas;
+
+pub fn generate_scalars<F: PrimeField>(
+    len: usize,
+) -> Vec<F> {
+    let mut rng = rand::thread_rng();
+    let scalars = (0..len)
+        .map(|_| F::random(&mut rng))
+        .collect::<Vec<_>>();
+
+    scalars
+}
+
+/// cargo run --release --example spmvm
+fn main() {
+    let npow: usize = std::env::var("NPOW")
+        .unwrap_or("23".to_string())
+        .parse()
+        .unwrap();
+    let n = 1usize << npow;
+
+    let mut scalars = generate_scalars(n);
+
+    let start = Instant::now();
+    let double_scalars = scalars.iter().map(|x| x + x).collect::<Vec<_>>();
+    println!("cpu took: {:?}", start.elapsed());
+
+    let start = Instant::now();
+    double_pallas(&mut scalars);
+    println!("gpu took: {:?}", start.elapsed());
+
+    assert_eq!(double_scalars, scalars);
+    println!("success!");
+}

src/lib.rs

@@ -2,6 +2,8 @@
 // Licensed under the Apache License, Version 2.0, see LICENSE for details.
 // SPDX-License-Identifier: Apache-2.0
 
+pub mod spmvm;
+
 extern crate semolina;
 
 #[cfg(feature = "cuda")]

src/spmvm.rs

@@ -0,0 +1,21 @@
+use pasta_curves::pallas;
+
+pub fn sparse_matrix_mul_pallas(scalars: &mut [pallas::Scalar]) {
+    extern "C" {
+        fn spmvm_pallas(
+            scalars: *mut pallas::Scalar,
+            nscalars: usize,
+        ) -> sppark::Error;
+    }
+
+    let nscalars = scalars.len();
+    let err = unsafe {
+        spmvm_pallas(
+            scalars.as_mut_ptr(),
+            nscalars,
+        )
+    };
+    if err.code != 0 {
+        panic!("{}", String::from(err));
+    }
+}