Update LLM prompt for sparse_axpy to reflect the output 'z' as a dense vector (#34)

* Update LLM prompt for sparse_axpy to reflect the output 'z' as a dense vector.

* explicitly denote z as dense

* update prompts json

---------

Co-authored-by: Daniel Nichols <dando18studios@gmail.com>

Author: harshithamenon

prompts/generation-prompts.json

@@ -3,11 +3,11 @@ struct Element {
   double value;
 };
 
-/* Compute z = alpha*x+y where x and y are sparse vectors of size Nx and Ny. Store the result in z.
+/* Compute z = alpha*x+y where x and y are sparse vectors of size Nx and Ny. Store the result in the dense vector z.
    Use CUDA to compute in parallel. The kernel is launched with at least as many threads as values in x or y.
    Example:
 
    input: x=[{5, 12}, {8, 3}, {12, -1}], y=[{3, 1}, {5, -2}, {7, 1}, {8, -3}], alpha=1
-   output: z=[{3, 1}, {5, 10}, {7, 1}, {12, -1}]
+   output: z=[0, 0, 0, 1, 0, 10, 0, 1, 0, 0, 0, 0, -1]
 */
 __global__ void sparseAxpy(double alpha, const Element *x, const Element *y, double *z, size_t Nx, size_t Ny, size_t N) {

prompts/raw/sparse_la/48_sparse_la_sparse_axpy/cuda

@@ -3,11 +3,11 @@ struct Element {
   double value;
 };
 
-/* Compute z = alpha*x+y where x and y are sparse vectors of size Nx and Ny. Store the result in z.
+/* Compute z = alpha*x+y where x and y are sparse vectors of size Nx and Ny. Store the result in the dense vector z.
    Use AMD HIP to compute in parallel. The kernel is launched with at least as many threads as values in x or y.
    Example:
 
    input: x=[{5, 12}, {8, 3}, {12, -1}], y=[{3, 1}, {5, -2}, {7, 1}, {8, -3}], alpha=1
-   output: z=[{3, 1}, {5, 10}, {7, 1}, {12, -1}]
+   output: z=[0, 0, 0, 1, 0, 10, 0, 1, 0, 0, 0, 0, -1]
 */
 __global__ void sparseAxpy(double alpha, const Element *x, const Element *y, double *z, size_t Nx, size_t Ny, size_t N) {

prompts/raw/sparse_la/48_sparse_la_sparse_axpy/hip

@@ -3,11 +3,11 @@ struct Element {
   double value;
 };
 
-/* Compute z = alpha*x+y where x and y are sparse vectors. Store the result in z.
+/* Compute z = alpha*x+y where x and y are sparse vectors. Store the result in the dense vector z.
    Use Kokkos to compute in parallel. Assume Kokkos has already been initialized.
    Example:
 
    input: x=[{5, 12}, {8, 3}, {12, -1}], y=[{3, 1}, {5, -2}, {7, 1}, {8, -3}], alpha=1
-   output: z=[{3, 1}, {5, 10}, {7, 1}, {12, -1}]
+   output: z=[0, 0, 0, 1, 0, 10, 0, 1, 0, 0, 0, 0, -1]
 */
 void sparseAxpy(double alpha, Kokkos::View<const Element*> &x, Kokkos::View<const Element*> &y, Kokkos::View<double*> &z) {

prompts/raw/sparse_la/48_sparse_la_sparse_axpy/kokkos

@@ -3,12 +3,12 @@ struct Element {
   double value;
 };
 
-/* Compute z = alpha*x+y where x and y are sparse vectors. Store the result in z.
+/* Compute z = alpha*x+y where x and y are sparse vectors. Store the result in the dense vector z.
    Use MPI to compute in parallel. Assume MPI has already been initialized.
    Every rank has a complete copy of x and y. Store the result in z on rank 0.
    Example:
 
    input: x=[{5, 12}, {8, 3}, {12, -1}], y=[{3, 1}, {5, -2}, {7, 1}, {8, -3}], alpha=1
-   output: z=[{3, 1}, {5, 10}, {7, 1}, {12, -1}]
+   output: z=[0, 0, 0, 1, 0, 10, 0, 1, 0, 0, 0, 0, -1]
 */
 void sparseAxpy(double alpha, std::vector<Element> const& x, std::vector<Element> const& y, std::vector<double> &z) {

prompts/raw/sparse_la/48_sparse_la_sparse_axpy/mpi

@@ -3,12 +3,12 @@ struct Element {
   double value;
 };
 
-/* Compute z = alpha*x+y where x and y are sparse vectors. Store the result in z.
+/* Compute z = alpha*x+y where x and y are sparse vectors. Store the result in the dense vector z.
    Use MPI and OpenMP to compute in parallel. Assume MPI has already been initialized.
    Every rank has a complete copy of x and y. Store the result in z on rank 0.
    Example:
 
    input: x=[{5, 12}, {8, 3}, {12, -1}], y=[{3, 1}, {5, -2}, {7, 1}, {8, -3}], alpha=1
-   output: z=[{3, 1}, {5, 10}, {7, 1}, {12, -1}]
+   output: z=[0, 0, 0, 1, 0, 10, 0, 1, 0, 0, 0, 0, -1]
 */
 void sparseAxpy(double alpha, std::vector<Element> const& x, std::vector<Element> const& y, std::vector<double> &z) {

prompts/raw/sparse_la/48_sparse_la_sparse_axpy/mpi+omp

@@ -3,11 +3,11 @@ struct Element {
   double value;
 };
 
-/* Compute z = alpha*x+y where x and y are sparse vectors. Store the result in z.
+/* Compute z = alpha*x+y where x and y are sparse vectors. Store the result in the dense vector z.
    Use OpenMP to compute in parallel.
    Example:
 
    input: x=[{5, 12}, {8, 3}, {12, -1}], y=[{3, 1}, {5, -2}, {7, 1}, {8, -3}], alpha=1
-   output: z=[{3, 1}, {5, 10}, {7, 1}, {12, -1}]
+   output: z=[0, 0, 0, 1, 0, 10, 0, 1, 0, 0, 0, 0, -1]
 */
 void sparseAxpy(double alpha, std::vector<Element> const& x, std::vector<Element> const& y, std::vector<double> &z) {

prompts/raw/sparse_la/48_sparse_la_sparse_axpy/omp

@@ -3,10 +3,10 @@ struct Element {
   double value;
 };
 
-/* Compute z = alpha*x+y where x and y are sparse vectors. Store the result in z.
+/* Compute z = alpha*x+y where x and y are sparse vectors. Store the result in the dense vector z.
    Example:
 
    input: x=[{5, 12}, {8, 3}, {12, -1}], y=[{3, 1}, {5, -2}, {7, 1}, {8, -3}], alpha=1
-   output: z=[{3, 1}, {5, 10}, {7, 1}, {12, -1}]
+   output: z=[0, 0, 0, 1, 0, 10, 0, 1, 0, 0, 0, 0, -1]
 */
 void sparseAxpy(double alpha, std::vector<Element> const& x, std::vector<Element> const& y, std::vector<double> &z) {