Lab_B_Systolic_array

Introduce how to implement systolic array with HLS.

Introduction

The Systolic array is used in matrix multiplication. This architecture can reduce the number of times the hardware reads data.

In this Lab, the matrix multiplication of A[16][16]*B[16][16] will be implemented, and Store the multiplied result in C[16][16].

Matrix multiplication corresponds to computation using the systolic array schema as follows:

Code & Pragma

In for loop, we can use "pragma HLS LOOP_TRIPCOUNT" to help calculate the time required for the for loop to complete the operation.

readA:
    for (int loc = 0, i = 0, j = 0; loc < a_row * a_col; loc++, j++) {
#pragma HLS LOOP_TRIPCOUNT min = c_size* c_size max = c_size * c_size
        if (j == a_col) {
            i++;
            j = 0;
        }
        localA[i][j] = a[loc];
    }

"HLS ARRAY_PARTITION variable = localA dim = 1 complete":The matrix localA will be cut into multiple arrays in the row direction

"HLS ARRAY_PARTITION variable = localB dim = 2 complete":The matrix localB will be cut into multiple arrays in the col direction

"HLS ARRAY_PARTITION variable = localC dim = 0 complete":Assume that each element of matrix C is 8bit, the matrix localC will be put in a register which is 16*16*8 bit

#pragma HLS ARRAY_PARTITION variable = localA dim = 1 complete

#pragma HLS ARRAY_PARTITION variable = localB dim = 2 complete

#pragma HLS ARRAY_PARTITION variable = localC dim = 0 complete

"HLS UNROLL" allows the operations in the for loop to be processed in parallel.

systolic1:
    for (int k = 0; k < a_col; k++) {
#pragma HLS LOOP_TRIPCOUNT min = c_size max = c_size
    systolic2:
        for (int i = 0; i < MAX_SIZE; i++) {
#pragma HLS UNROLL
        systolic3:
            for (int j = 0; j < MAX_SIZE; j++) {
#pragma HLS UNROLL
                // Get previous sum
                int last = (k == 0) ? 0 : localC[i][j];

                // Update current sum
                // Handle boundary conditions
                int a_val = (i < a_row && k < a_col) ? localA[i][k] : 0;
                int b_val = (k < b_row && j < b_col) ? localB[k][j] : 0;
                int result = last + a_val * b_val;

                // Write back results
                localC[i][j] = result;
            }
        }
    }

Build flow

You can refer to the file . The only difference is 3.2.2., just give "binary_container_1.xclbin" this argument.

2021.2-Workbook-Lab3_new.pdf

Result