[software] Format the data generation scripts

pulp-platform · Jan 8, 2024 · 01a6581 · 01a6581
1 parent 01d9fd2
commit 01a6581
Show file tree

Hide file tree

Showing 10 changed files with 114 additions and 95 deletions.
diff --git a/Makefile b/Makefile
@@ -181,6 +181,7 @@ toolchain/riscv-opcodes/*:
 
 format:
 	$(ROOT_DIR)/scripts/run_clang_format.py --clang-format-executable=$(LLVM_INSTALL_DIR)/bin/clang-format -i -r $(ROOT_DIR)
+	find ./software/runtime/data -name '*.py' -exec autopep8 --in-place --aggressive {} +
 
 clean: clean_test
 	rm -rf $(INSTALL_DIR)
diff --git a/software/runtime/data/data_chest_f16.py b/software/runtime/data/data_chest_f16.py
@@ -116,7 +116,6 @@ def main():
     vector_pilot_rx = np.concatenate(vector_pilot_rx, axis=0)
     vector_pilot_tx = np.concatenate(vector_pilot_tx, axis=0)
     vector_Hest = np.concatenate(vector_Hest, axis=0)
-    print(vector_Hest)
 
     kwargs = {'name': 'data_chest_f16',
               'pilot_rx': vector_pilot_rx,

diff --git a/software/runtime/data/data_cholesky_f16.py b/software/runtime/data/data_cholesky_f16.py
@@ -10,15 +10,15 @@
 import numpy as np
 import argparse
 import pathlib
-from scipy.linalg import solve_triangular
 from mako.template import Template
 
 
 ##################
 # compute_result #
 ##################
 
-def gen_data_header_file(outdir: pathlib.Path.cwd(), tpl: pathlib.Path.cwd(), **kwargs):
+def gen_data_header_file(outdir: pathlib.Path.cwd(),
+                         tpl: pathlib.Path.cwd(), **kwargs):
 
     file = outdir / f"{kwargs['name']}.h"
 

diff --git a/software/runtime/data/data_cholesky_q16.py b/software/runtime/data/data_cholesky_q16.py
@@ -10,7 +10,6 @@
 import numpy as np
 import argparse
 import pathlib
-from scipy.linalg import solve_triangular
 from mako.template import Template
 
 
@@ -19,7 +18,8 @@
 ##################
 
 
-def gen_data_header_file(outdir: pathlib.Path.cwd(), tpl: pathlib.Path.cwd(), **kwargs):
+def gen_data_header_file(outdir: pathlib.Path.cwd(),
+                         tpl: pathlib.Path.cwd(), **kwargs):
 
     file = outdir / f"{kwargs['name']}.h"
 
@@ -81,8 +81,10 @@ def main():
     vector_L = []
     for k in range(n_samples):
         # Create hermitian matrix
-        H = (np.random.randint(-2**(15), 2**(15) - 1, n_matrix * n_matrix, dtype=np.int16) + 1.j * \
-            np.random.randint(-2**(15), 2**(15) - 1, n_matrix * n_matrix, dtype=np.int16)).reshape(n_matrix, n_matrix)
+        H = np.random.randint(-2**(15), 2**(15) - 1, n_matrix * n_matrix, dtype=np.int16) \
+            + 1.j * np.random.randint(-2**(15), 2**(15) - 1, n_matrix *
+                                      n_matrix, dtype=np.int16)
+        H = H.reshape(n_matrix, n_matrix)
         # Matrix to be inverted
         H_h = (np.asmatrix(H).H)
         # H_H = np.asmatrix(H).H

diff --git a/software/runtime/data/data_cholesky_q32.py b/software/runtime/data/data_cholesky_q32.py
@@ -18,7 +18,8 @@
 # compute_result #
 ##################
 
-def gen_data_header_file(outdir: pathlib.Path.cwd(), tpl: pathlib.Path.cwd(), **kwargs):
+def gen_data_header_file(outdir: pathlib.Path.cwd(),
+                         tpl: pathlib.Path.cwd(), **kwargs):
 
     file = outdir / f"{kwargs['name']}.h"
 
@@ -68,24 +69,30 @@ def main():
     n_matrix = args.dimension
 
     # Create hermitian matrix
-    H = np.random.randint(-2**(15), 2**(15) - 1, n_matrix * n_matrix, dtype=np.int32)
-    y = np.random.randint(-2**(15), 2**(15) - 1, n_matrix, dtype=np.int32)
+    L = np.random.randint(-2**(15), 2**(15) - 1,
+                          size=(n_matrix, n_matrix), dtype=np.int32)
+    L = np.tril(L).astype(np.int32)
+    G = np.dot(np.asmatrix(L), np.asmatrix(L).transpose())
 
-    H = H.reshape(n_matrix, n_matrix)
-    G = H * np.asmatrix(H).T
-    # Cholesky decomposition
-    L = np.linalg.cholesky(G)
+    y = np.random.randint(-2**(15), 2**(15) - 1, n_matrix, dtype=np.int32)
 
     # Linear system solution
     y = solve_triangular(L, y, lower=True)
-    x = solve_triangular(np.asmatrix(L).H, y)
+    # x = solve_triangular(np.asmatrix(L).T, y)
 
     # Reshape
-    G = np.reshape(np.asarray(G), (n_matrix * n_matrix), order='C').astype(np.int32)
-    L = np.reshape(np.asarray(L), (n_matrix * n_matrix), order='C').astype(np.int32)
+    G = np.reshape(
+        np.asarray(G),
+        (n_matrix * n_matrix),
+        order='C').astype(
+        np.int32)
+    L = np.reshape(
+        np.asarray(L),
+        (n_matrix * n_matrix),
+        order='C').astype(
+        np.int32)
     y = np.reshape(np.asarray(y), (n_matrix), order='C').astype(np.int32)
 
-
     kwargs = {'name': 'data_cholesky_q32',
               'G': G,
               'L': L,
@@ -94,5 +101,6 @@ def main():
 
     gen_data_header_file(args.outdir, args.tpl, **kwargs)
 
+
 if __name__ == "__main__":
     main()
diff --git a/software/runtime/data/data_cmatmul_f16.py b/software/runtime/data/data_cmatmul_f16.py
@@ -96,7 +96,6 @@ def main():
     A = np.reshape(A, (matrix_M * matrix_N), order='C')
     B = np.reshape(B, (matrix_N * matrix_P), order='C')
     C = np.reshape(C, (matrix_M * matrix_P), order='C')
-    print(A)
 
     A = np.column_stack((A.real, A.imag)).astype(np.float16).flatten()
     B = np.column_stack((B.real, B.imag)).astype(np.float16).flatten()

diff --git a/software/runtime/data/data_mimo_mmse_f16.py b/software/runtime/data/data_mimo_mmse_f16.py
@@ -18,7 +18,8 @@
 # compute_result #
 ##################
 
-def gen_data_header_file(outdir: pathlib.Path.cwd(), tpl: pathlib.Path.cwd(), **kwargs):
+def gen_data_header_file(outdir: pathlib.Path.cwd(),
+                         tpl: pathlib.Path.cwd(), **kwargs):
 
     file = outdir / f"{kwargs['name']}.h"
 
@@ -52,9 +53,9 @@ def gen_input_data(N_rx, N_tx):
     x = solve_triangular(np.asmatrix(L).H, y2)
 
     sigma = sigma + 0j
-    H = np.reshape(np.asarray(H), (N_tx*N_rx), order='C')
-    G = np.reshape(np.asarray(G), (N_tx*N_tx), order='C')
-    L = np.reshape(np.asarray(L), (N_tx*N_tx), order='C')
+    H = np.reshape(np.asarray(H), (N_tx * N_rx), order='C')
+    G = np.reshape(np.asarray(G), (N_tx * N_tx), order='C')
+    L = np.reshape(np.asarray(L), (N_tx * N_tx), order='C')
     sigma = np.column_stack((sigma.real, sigma.imag)
                             ).astype(np.float16).flatten()
     H = np.column_stack((H.real, H.imag)).astype(np.float16).flatten()
@@ -123,20 +124,20 @@ def main():
     N_rx = args.receivers
     itr = args.iterations
 
-    sigma = np.zeros([itr, 2*N_tx])
-    H_RI = np.zeros([itr, 2*N_tx*N_rx])
-    G_RI = np.zeros([itr, 2*N_tx*N_tx])
-    y_RI = np.zeros([itr, 2*N_rx])
-    x_RI = np.zeros([itr, 2*N_tx])
+    sigma = np.zeros([itr, 2 * N_tx])
+    H_RI = np.zeros([itr, 2 * N_tx * N_rx])
+    G_RI = np.zeros([itr, 2 * N_tx * N_tx])
+    y_RI = np.zeros([itr, 2 * N_rx])
+    x_RI = np.zeros([itr, 2 * N_tx])
     for k in range(itr):
-        sigma[k, :], H_RI[k, :], G_RI[k, :], y_RI[k,
-                                                  :], x_RI[k, :] = gen_input_data(N_rx, N_tx)
-
-    sigma = np.reshape(sigma, (2*N_tx*itr)).astype(np.float16)
-    H_RI = np.reshape(H_RI, (2*N_rx*N_tx*itr)).astype(np.float16)
-    G_RI = np.reshape(G_RI, (2*N_tx*N_tx*itr)).astype(np.float16)
-    y_RI = np.reshape(y_RI, (2*N_rx*itr)).astype(np.float16)
-    x_RI = np.reshape(x_RI, (2*N_tx*itr)).astype(np.float16)
+        sigma[k, :], H_RI[k, :], G_RI[k, :], \
+            y_RI[k, :], x_RI[k, :] = gen_input_data(N_rx, N_tx)
+
+    sigma = np.reshape(sigma, (2 * N_tx * itr)).astype(np.float16)
+    H_RI = np.reshape(H_RI, (2 * N_rx * N_tx * itr)).astype(np.float16)
+    G_RI = np.reshape(G_RI, (2 * N_tx * N_tx * itr)).astype(np.float16)
+    y_RI = np.reshape(y_RI, (2 * N_rx * itr)).astype(np.float16)
+    x_RI = np.reshape(x_RI, (2 * N_tx * itr)).astype(np.float16)
 
     kwargs = {'name': 'data_mimo_mmse_f16',
               'H': H_RI,

diff --git a/software/runtime/data/data_mimo_mmse_f32.py b/software/runtime/data/data_mimo_mmse_f32.py
@@ -18,7 +18,8 @@
 # compute_result #
 ##################
 
-def gen_data_header_file(outdir: pathlib.Path.cwd(), tpl: pathlib.Path.cwd(), **kwargs):
+def gen_data_header_file(outdir: pathlib.Path.cwd(),
+                         tpl: pathlib.Path.cwd(), **kwargs):
 
     file = outdir / f"{kwargs['name']}.h"
 
@@ -52,9 +53,9 @@ def gen_input_data(N_rx, N_tx):
     y2 = solve_triangular(L, y1, lower=True)
     x = solve_triangular(np.asmatrix(L).H, y2)
 
-    H = np.reshape(np.asarray(H), (N_tx*N_rx), order='C')
-    G = np.reshape(np.asarray(G), (N_tx*N_tx), order='C')
-    L = np.reshape(np.asarray(L), (N_tx*N_tx), order='C')
+    H = np.reshape(np.asarray(H), (N_tx * N_rx), order='C')
+    G = np.reshape(np.asarray(G), (N_tx * N_tx), order='C')
+    L = np.reshape(np.asarray(L), (N_tx * N_tx), order='C')
     H = np.column_stack((H.real, H.imag)).astype(np.float32).flatten()
     G = np.column_stack((G.real, G.imag)).astype(np.float32).flatten()
     L = np.column_stack((L.real, L.imag)).astype(np.float32).flatten()
@@ -122,19 +123,19 @@ def main():
     itr = args.iterations
 
     sigma = np.zeros([itr, N_tx])
-    H_RI = np.zeros([itr, 2*N_tx*N_rx])
-    G_RI = np.zeros([itr, 2*N_tx*N_tx])
-    y_RI = np.zeros([itr, 2*N_rx])
-    x_RI = np.zeros([itr, 2*N_tx])
+    H_RI = np.zeros([itr, 2 * N_tx * N_rx])
+    G_RI = np.zeros([itr, 2 * N_tx * N_tx])
+    y_RI = np.zeros([itr, 2 * N_rx])
+    x_RI = np.zeros([itr, 2 * N_tx])
     for k in range(itr):
-        sigma[k, :], H_RI[k, :], G_RI[k, :], y_RI[k,
-                                                  :], x_RI[k, :] = gen_input_data(N_rx, N_tx)
-
-    sigma = np.reshape(sigma, (N_tx*itr))
-    H_RI = np.reshape(H_RI, (2*N_rx*N_tx*itr))
-    G_RI = np.reshape(G_RI, (2*N_tx*N_tx*itr))
-    y_RI = np.reshape(y_RI, (2*N_rx*itr))
-    x_RI = np.reshape(x_RI, (2*N_tx*itr))
+        sigma[k, :], H_RI[k, :], G_RI[k, :], \
+            y_RI[k, :], x_RI[k, :] = gen_input_data(N_rx, N_tx)
+
+    sigma = np.reshape(sigma, (N_tx * itr))
+    H_RI = np.reshape(H_RI, (2 * N_rx * N_tx * itr))
+    G_RI = np.reshape(G_RI, (2 * N_tx * N_tx * itr))
+    y_RI = np.reshape(y_RI, (2 * N_rx * itr))
+    x_RI = np.reshape(x_RI, (2 * N_tx * itr))
 
     kwargs = {'name': 'data_mimo_mmse_f32',
               'H': H_RI,

diff --git a/software/runtime/data/data_mimo_mmse_q16.py b/software/runtime/data/data_mimo_mmse_q16.py
@@ -18,7 +18,8 @@
 # compute_result #
 ##################
 
-def gen_data_header_file(outdir: pathlib.Path.cwd(), tpl: pathlib.Path.cwd(), **kwargs):
+def gen_data_header_file(outdir: pathlib.Path.cwd(),
+                         tpl: pathlib.Path.cwd(), **kwargs):
 
     file = outdir / f"{kwargs['name']}.h"
 
@@ -32,12 +33,13 @@ def gen_data_header_file(outdir: pathlib.Path.cwd(), tpl: pathlib.Path.cwd(), **
 def gen_input_data(N_rx, N_tx):
 
     # Create channel matrix
-    H = np.random.randint(-2**(15), 2**(15) - 1, N_rx * N_tx, dtype=np.int16) + 1.j * \
-        np.random.randint(-2**(15), 2**(15) - 1, N_rx * N_tx, dtype=np.int16)
+    H = np.random.randint(-2**(15), 2**(15) - 1, N_rx * N_tx, dtype=np.int16) + \
+        1.j * np.random.randint(-2**(15), 2**(15) - 1,
+                                N_rx * N_tx, dtype=np.int16)
     H = H.reshape(N_rx, N_tx)
     # Create input vector
-    y = np.random.randint(-2**(15), 2**(15) - 1, N_rx, dtype=np.int16) + 1.j * \
-        np.random.randint(-2**(15), 2**(15) - 1, N_rx, dtype=np.int16)
+    y = np.random.randint(-2**(15), 2**(15) - 1, N_rx, dtype=np.int16) + \
+        1.j * np.random.randint(-2**(15), 2**(15) - 1, N_rx, dtype=np.int16)
     # Generate noise variance
     sigma = np.random.randint(-2**(15), 2**(15) - 1, N_tx, dtype=np.int16)
 
@@ -50,17 +52,19 @@ def gen_input_data(N_rx, N_tx):
     y1 = np.transpose(np.dot(H_h, y))
 
     # Cholesky decomposition
-    #L = np.linalg.cholesky(G)
+    # L = np.linalg.cholesky(G)
     L = G
     # Linear system solution
     y2 = solve_triangular(L, y1, lower=True)
     x = solve_triangular(np.asmatrix(L).H, y2)
 
     sigma = sigma + 0j
-    H = np.reshape(np.asarray(H), (N_rx*N_tx), order='C')
-    G = np.reshape(np.asarray(G), (N_tx*N_tx), order='C')
-    L = np.reshape(np.asarray(L), (N_tx*N_tx), order='C')
-    sigma = np.column_stack((sigma.real, sigma.imag)).astype(np.int16).flatten()
+    H = np.reshape(np.asarray(H), (N_rx * N_tx), order='C')
+    G = np.reshape(np.asarray(G), (N_tx * N_tx), order='C')
+    L = np.reshape(np.asarray(L), (N_tx * N_tx), order='C')
+    sigma = np.column_stack(
+        (sigma.real, sigma.imag)).astype(
+        np.int16).flatten()
     H = np.column_stack((H.real, H.imag)).astype(np.int16).flatten()
     G = np.column_stack((G.real, G.imag)).astype(np.int16).flatten()
     L = np.column_stack((L.real, L.imag)).astype(np.int16).flatten()
@@ -126,23 +130,23 @@ def main():
     N_rx = args.receivers
     itr = args.iterations
 
-    sigma = np.zeros([itr, 2*N_tx], dtype=np.int16)
-    H_RI = np.zeros([itr, 2*N_tx*N_rx], dtype=np.int16)
-    G_RI = np.zeros([itr, 2*N_tx*N_tx], dtype=np.int16)
-    y_RI = np.zeros([itr, 2*N_rx], dtype=np.int16)
-    x_RI = np.zeros([itr, 2*N_tx], dtype=np.int16)
+    sigma = np.zeros([itr, 2 * N_tx], dtype=np.int16)
+    H_RI = np.zeros([itr, 2 * N_tx * N_rx], dtype=np.int16)
+    G_RI = np.zeros([itr, 2 * N_tx * N_tx], dtype=np.int16)
+    y_RI = np.zeros([itr, 2 * N_rx], dtype=np.int16)
+    x_RI = np.zeros([itr, 2 * N_tx], dtype=np.int16)
     for k in range(itr):
-        [   sigma[k, :],
+        [sigma[k, :],
             H_RI[k, :],
             G_RI[k, :],
             y_RI[k, :],
-            x_RI[k, :] ] = gen_input_data(N_rx, N_tx)
+            x_RI[k, :]] = gen_input_data(N_rx, N_tx)
 
-    sigma = np.reshape(sigma, (2*N_tx*itr)).astype(np.int16)
-    H_RI = np.reshape(H_RI, (2*N_rx*N_tx*itr)).astype(np.int16)
-    G_RI = np.reshape(G_RI, (2*N_tx*N_tx*itr)).astype(np.int16)
-    y_RI = np.reshape(y_RI, (2*N_rx*itr)).astype(np.int16)
-    x_RI = np.reshape(x_RI, (2*N_tx*itr)).astype(np.int16)
+    sigma = np.reshape(sigma, (2 * N_tx * itr)).astype(np.int16)
+    H_RI = np.reshape(H_RI, (2 * N_rx * N_tx * itr)).astype(np.int16)
+    G_RI = np.reshape(G_RI, (2 * N_tx * N_tx * itr)).astype(np.int16)
+    y_RI = np.reshape(y_RI, (2 * N_rx * itr)).astype(np.int16)
+    x_RI = np.reshape(x_RI, (2 * N_tx * itr)).astype(np.int16)
 
     kwargs = {'name': 'data_mimo_mmse_q16',
               'H': H_RI,