diff --git a/brainpy/_src/math/event/tests/event_csrmv_taichi_VS_event_csrmv.py b/brainpy/_src/math/event/tests/event_csrmv_taichi_VS_event_csrmv.py
index 39a813264..47c60b2d6 100644
--- a/brainpy/_src/math/event/tests/event_csrmv_taichi_VS_event_csrmv.py
+++ b/brainpy/_src/math/event/tests/event_csrmv_taichi_VS_event_csrmv.py
@@ -278,10 +278,10 @@ def test_event_ell_gpu(s, p, values_type, events_type, transpose):
               for _transpose in transpose:
                 taichi_aot_time_1, taichi_aot_time_2, taichi_aot_time_3, taichi_aot_time_4, taichi_aot_time_5,\
                     brainpy_time_1, brainpy_time_2, brainpy_time_3, brainpy_time_4, brainpy_time_5, speedup = test_event_ell_cpu(_s, _p, _values_type, _events_type, _transpose)
-              # append to dataframe
-              df.loc[df.shape[0]] = [_s, _p, 'cpu', _values_type, _events_type, _transpose,
-                                    taichi_aot_time_1, taichi_aot_time_2, taichi_aot_time_3, taichi_aot_time_4, taichi_aot_time_5,
-                                    brainpy_time_1, brainpy_time_2, brainpy_time_3, brainpy_time_4, brainpy_time_5, speedup]
+                # append to dataframe
+                df.loc[df.shape[0]] = [_s, _p, 'cpu', _values_type, _events_type, _transpose,
+                                      taichi_aot_time_1, taichi_aot_time_2, taichi_aot_time_3, taichi_aot_time_4, taichi_aot_time_5,
+                                      brainpy_time_1, brainpy_time_2, brainpy_time_3, brainpy_time_4, brainpy_time_5, speedup]
   df.to_csv(f'{PATH}/event_csrmv_cpu.csv', index=False)
 
 if (bm.get_platform() == 'gpu'):
@@ -292,10 +292,10 @@ def test_event_ell_gpu(s, p, values_type, events_type, transpose):
               for _transpose in transpose:
                 taichi_aot_time_1, taichi_aot_time_2, taichi_aot_time_3, taichi_aot_time_4, taichi_aot_time_5,\
                     brainpy_time_1, brainpy_time_2, brainpy_time_3, brainpy_time_4, brainpy_time_5, speedup = test_event_ell_gpu(_s, _p, _values_type, _events_type, _transpose)
-              # append to dataframe
-              df.loc[df.shape[0]] = [_s, _p, 'gpu', _values_type, _events_type, transpose,
-                                    taichi_aot_time_1, taichi_aot_time_2, taichi_aot_time_3, taichi_aot_time_4, taichi_aot_time_5,
-                                    brainpy_time_1, brainpy_time_2, brainpy_time_3, brainpy_time_4, brainpy_time_5, speedup]
+                # append to dataframe
+                df.loc[df.shape[0]] = [_s, _p, 'gpu', _values_type, _events_type, _transpose,
+                                      taichi_aot_time_1, taichi_aot_time_2, taichi_aot_time_3, taichi_aot_time_4, taichi_aot_time_5,
+                                      brainpy_time_1, brainpy_time_2, brainpy_time_3, brainpy_time_4, brainpy_time_5, speedup]
   df.to_csv(f'{PATH}/event_csrmv_gpu.csv', index=False)
 
 # if (bm.get_platform() == 'gpu'):
diff --git a/brainpy/_src/math/sparse/_csr_mv_taichi.py b/brainpy/_src/math/sparse/_csr_mv_taichi.py
index 6f739e7bd..d3a6e10b6 100644
--- a/brainpy/_src/math/sparse/_csr_mv_taichi.py
+++ b/brainpy/_src/math/sparse/_csr_mv_taichi.py
@@ -74,9 +74,18 @@ def _sparse_csr_matvec_transpose_homo_gpu(values: ti.types.ndarray(ndim=1),
                                      vector: ti.types.ndarray(ndim=1),
                                      out: ti.types.ndarray(ndim=1)):
   value = values[0]
-  for row_i in range(row_ptr.shape[0] - 1):
-    for j in range(row_ptr[row_i], row_ptr[row_i + 1]):
+  total_rows = row_ptr.shape[0] - 1
+  for i in range(total_rows * 32):
+    row_i = i >> 5
+    index = i & 31
+    j = row_ptr[row_i] + index
+    end_index = row_ptr[row_i + 1]
+    while j < end_index:
       out[col_indices[j]] += value * vector[row_i]
+      j += 32
+  # for row_i in range(row_ptr.shape[0] - 1):
+  #   for j in range(row_ptr[row_i], row_ptr[row_i + 1]):
+  #     out[col_indices[j]] += value * vector[row_i]
 
 
 @ti.kernel
@@ -86,11 +95,21 @@ def _sparse_csr_matvec_homo_gpu(values: ti.types.ndarray(ndim=1),
                            vector: ti.types.ndarray(ndim=1),
                            out: ti.types.ndarray(ndim=1)):
   value = values[0]
-  for row_i in range(row_ptr.shape[0] - 1):
+  total_rows = row_ptr.shape[0] - 1
+  for i in range(total_rows * 32):
+    row_i = i >> 5
+    index = i & 31
     r = 0.
-    for j in range(row_ptr[row_i], row_ptr[row_i + 1]):
+    j = row_ptr[row_i] + index
+    end_index = row_ptr[row_i + 1]
+    while j < end_index:
       r += value * vector[col_indices[j]]
-    out[row_i] = r
+      j += 32
+  # for row_i in range(row_ptr.shape[0] - 1):
+  #   r = 0.
+  #   for j in range(row_ptr[row_i], row_ptr[row_i + 1]):
+  #     r += value * vector[col_indices[j]]
+  #   out[row_i] = r
 
 # heter
 
@@ -100,9 +119,18 @@ def _sparse_csr_matvec_transpose_heter_gpu(values: ti.types.ndarray(ndim=1),
                                      row_ptr: ti.types.ndarray(ndim=1),
                                      vector: ti.types.ndarray(ndim=1),
                                      out: ti.types.ndarray(ndim=1)):
-  for row_i in range(row_ptr.shape[0] - 1):
-    for j in range(row_ptr[row_i], row_ptr[row_i + 1]):
+  total_rows = row_ptr.shape[0] - 1
+  for i in range(total_rows * 32):
+    row_i = i >> 5
+    index = i & 31
+    j = row_ptr[row_i] + index
+    end_index = row_ptr[row_i + 1]
+    while j < end_index:
       out[col_indices[j]] += values[j] * vector[row_i]
+      j += 32
+  # for row_i in range(row_ptr.shape[0] - 1):
+  #   for j in range(row_ptr[row_i], row_ptr[row_i + 1]):
+  #     out[col_indices[j]] += values[j] * vector[row_i]
 
 
 @ti.kernel
@@ -111,11 +139,21 @@ def _sparse_csr_matvec_heter_gpu(values: ti.types.ndarray(ndim=1),
                            row_ptr: ti.types.ndarray(ndim=1),
                            vector: ti.types.ndarray(ndim=1),
                            out: ti.types.ndarray(ndim=1)):
-  for row_i in range(row_ptr.shape[0] - 1):
+  total_rows = row_ptr.shape[0] - 1
+  for i in range(total_rows * 32):
+    row_i = i >> 5
+    index = i & 31
     r = 0.
-    for j in range(row_ptr[row_i], row_ptr[row_i + 1]):
+    j = row_ptr[row_i] + index
+    end_index = row_ptr[row_i + 1]
+    while j < end_index:
       r += values[j] * vector[col_indices[j]]
-    out[row_i] = r
+      j += 32
+  # for row_i in range(row_ptr.shape[0] - 1):
+  #   r = 0.
+  #   for j in range(row_ptr[row_i], row_ptr[row_i + 1]):
+  #     r += values[j] * vector[col_indices[j]]
+  #   out[row_i] = r
 
 
 def _sparse_csr_matvec_jvp_values(val_dot, values, col_indices, row_ptr, vector, *, outs, transpose, shape):
diff --git a/brainpy/_src/math/sparse/tests/csrmv_taichi_VS_csrmv.py b/brainpy/_src/math/sparse/tests/csrmv_taichi_VS_csrmv.py
index 9583743fb..606715e91 100644
--- a/brainpy/_src/math/sparse/tests/csrmv_taichi_VS_csrmv.py
+++ b/brainpy/_src/math/sparse/tests/csrmv_taichi_VS_csrmv.py
@@ -14,7 +14,7 @@
 
 bm.set_platform('gpu')
 
-s = [1000, 2500, 5000, 10000, 25000, 50000]
+s = [1000, 5000, 10000, 15000, 20000, 25000, 30000]
 p = [0.1, 0.2, 0.3, 0.4, 0.5]
 values_type = ['homo', 'heter']
 events_type = ['float']
@@ -266,11 +266,11 @@ def test_event_ell_gpu(s, p, values_type, events_type, transpose):
               for _transpose in transpose:
                 taichi_aot_time_1, taichi_aot_time_2, taichi_aot_time_3, taichi_aot_time_4, taichi_aot_time_5,\
                     brainpy_time_1, brainpy_time_2, brainpy_time_3, brainpy_time_4, brainpy_time_5, speedup = test_event_ell_cpu(_s, _p, _values_type, _events_type, _transpose)
-              # append to dataframe
-              df.loc[df.shape[0]] = [_s, _p, 'cpu', _values_type, _events_type, _transpose,
-                                    taichi_aot_time_1, taichi_aot_time_2, taichi_aot_time_3, taichi_aot_time_4, taichi_aot_time_5,
-                                    brainpy_time_1, brainpy_time_2, brainpy_time_3, brainpy_time_4, brainpy_time_5, speedup]
-  df.to_csv(f'{PATH}/event_csrmv_cpu.csv', index=False)
+                # append to dataframe
+                df.loc[df.shape[0]] = [_s, _p, 'cpu', _values_type, _events_type, _transpose,
+                                      taichi_aot_time_1, taichi_aot_time_2, taichi_aot_time_3, taichi_aot_time_4, taichi_aot_time_5,
+                                      brainpy_time_1, brainpy_time_2, brainpy_time_3, brainpy_time_4, brainpy_time_5, speedup]
+  df.to_csv(f'{PATH}/csrmv_cpu.csv', index=False)
 
 if (bm.get_platform() == 'gpu'):
   for _s in s:
@@ -280,11 +280,11 @@ def test_event_ell_gpu(s, p, values_type, events_type, transpose):
               for _transpose in transpose:
                 taichi_aot_time_1, taichi_aot_time_2, taichi_aot_time_3, taichi_aot_time_4, taichi_aot_time_5,\
                     brainpy_time_1, brainpy_time_2, brainpy_time_3, brainpy_time_4, brainpy_time_5, speedup = test_event_ell_gpu(_s, _p, _values_type, _events_type, _transpose)
-              # append to dataframe
-              df.loc[df.shape[0]] = [_s, _p, 'gpu', _values_type, _events_type, transpose,
-                                    taichi_aot_time_1, taichi_aot_time_2, taichi_aot_time_3, taichi_aot_time_4, taichi_aot_time_5,
-                                    brainpy_time_1, brainpy_time_2, brainpy_time_3, brainpy_time_4, brainpy_time_5, speedup]
-  df.to_csv(f'{PATH}/event_csrmv_gpu.csv', index=False)
+                # append to dataframe
+                df.loc[df.shape[0]] = [_s, _p, 'gpu', _values_type, _events_type, _transpose,
+                                      taichi_aot_time_1, taichi_aot_time_2, taichi_aot_time_3, taichi_aot_time_4, taichi_aot_time_5,
+                                      brainpy_time_1, brainpy_time_2, brainpy_time_3, brainpy_time_4, brainpy_time_5, speedup]
+  df.to_csv(f'{PATH}/csrmv_gpu.csv', index=False)
 
 # if (bm.get_platform() == 'gpu'):
 #   for _s in s: