fix test code

test/test_correctness.py

@@ -9,24 +9,27 @@
 
 @pytest.fixture(params=[0, 1, 2, 3])
 def compile_lamm(request):
-    completed = subprocess.run(
+    subprocess.run(
+        args=LAMMCommand.clean(),
+        capture_output=True,
+        cwd=LAMM_PROJECT_DIR,
+        check=False,
+    )
+    return subprocess.run(
         args=LAMMCommand.compile(opt_level=request.param),
         capture_output=True,
-        timeout=60,
         cwd=LAMM_PROJECT_DIR,
-        shell=True,
         check=False,
     )
-    # assert completed.returncode == 0, completed.stderr
-    return request.param
 
-@pytest.mark.parametrize("dtype", ["f32", "q4_0", "q4_1", "q5_0"])
+@pytest.mark.parametrize("dtype", ["f32", "q4_0", "q4_1", "q8_0"])
 def test_matmul_correctness(compile_lamm, dtype):
+    assert compile_lamm.returncode == 0, compile_lamm.stderr.decode("utf-8")
     completed = subprocess.run(
         args=LAMMCommand.run_benchmark(ggml_type=dtype),
         capture_output=True,
         timeout=60,
         cwd=LAMM_PROJECT_DIR,
         check=False,
     )
-    assert completed.returncode == 0, completed.stdout.decode("utf-8")
+    assert completed.returncode == 0, completed.stdout.decode("utf-8")

test/utils.py

@@ -5,21 +5,17 @@
 class LAMMCommand:
 
     @staticmethod
-    def compile(opt_level: int=3, debug: bool=False, clean: bool=True) -> list[str]:
-
-        clean_cmd = ['make', 'clean', '&&']
-
+    def clean() -> list[str]:
+        return ['make', 'clean']
+
+    @staticmethod
+    def compile(opt_level: int=3, debug: bool=False) -> list[str]:
         assert opt_level >= 0 and opt_level <= 3, f'Optimization level only support 0-3, but got {opt_level}'
         make_options = [f'LAMM_OPT_LEVEL={opt_level}']
         if debug:
             make_options.append('LAMM_DEBUG=1')
-
         compile_cmd = ['make', 'benchmark'] + make_options
-
-        if clean:
-            return clean_cmd + compile_cmd
-        else:
-            return compile_cmd        
+        return compile_cmd        
 
     @staticmethod
     def run_benchmark(ggml_type: str, n_threads: int=1, n_iters: int=1) -> list[str]: