Matrix to PIM-Matrix conversion

pim-os/src/bin/samsung_matrix_vector_multiply.rs

@@ -26,17 +26,10 @@ const X16_COLUMNS: usize = COLUMNS / 16;
 pub extern "C" fn main() {
     pim::state::set_kernel(&samsung_matrix_vector_mul::KERNEL);
 
-    let matrix = Box::new(pim::continuous_array::Matrix::<ROWS, X16_COLUMNS>(
-        SMatrix::from_fn(|r, c| {
-            if c > 0 {
-                return F16x16::zero();
-            }
+    let mut matrix = SMatrix::<_, ROWS, COLUMNS>::zeros();
+    matrix.fill_lower_triangle(F16x1::one(), 0);
 
-            let mut entry = F16x16::zero();
-            entry.0.iter_mut().take(r).for_each(|val| *val = F16x1::one());
-            
-            entry
-        })));
+    let pim_matrix = Box::new(pim::continuous_array::Matrix::from(matrix));
 
     let input_vector = SVector::<_, X16_COLUMNS>::from_element(F16x16::one());
     let interleaved_input_vector = Box::new(interleaved_array::Vector::from(input_vector));
@@ -53,7 +46,7 @@ pub extern "C" fn main() {
         pim::state::set_bank_mode(BankMode::PimAllBank);
 
         samsung_matrix_vector_mul::execute(
-            matrix.as_ref(),
+            pim_matrix.as_ref(),
             interleaved_input_vector.as_ref(),
             output_partial_sum_vector.as_mut(),
             dummy.as_ref(),

pim-os/src/lib.rs

@@ -1,3 +1,4 @@
+#![feature(iter_array_chunks)]
 #![no_std]
 
 use core::sync::atomic::{compiler_fence, Ordering};

pim-os/src/pim/continuous_array.rs

@@ -1,4 +1,4 @@
-use super::vector::F16x16;
+use super::vector::{F16x1, F16x16};
 use core::fmt::Display;
 use nalgebra::SMatrix;
 
@@ -11,3 +11,18 @@ impl<const R: usize, const C: usize> Display for Matrix<R, C> {
         self.0.fmt(f)
     }
 }
+
+impl<const R: usize, const C: usize, const X16C: usize> From<SMatrix<F16x1, R, C>>
+    for Matrix<R, X16C>
+{
+    fn from(matrix: SMatrix<F16x1, R, C>) -> Self {
+        Self(SMatrix::from_row_iterator(
+            matrix
+                .transpose()
+                .iter()
+                .map(|e| *e)
+                .array_chunks::<16>()
+                .map(|chunk| F16x16(chunk)),
+        ))
+    }
+}