Split benchmark for unqry pointwise operators

Author: tengqm

benchmark/performance_utils.py

@@ -697,6 +697,36 @@ def get_tflops(self, op, *args, **kwargs):
         return total_flops
 
 
+class UnaryPointwiseBenchmark(Benchmark):
+    """
+    Base class for benchmarking unary pointwise operations.
+    """
+
+    DEFAULT_METRICS = DEFAULT_METRICS[:] + ["tflops"]
+
+    def set_more_shapes(self):
+        special_shapes_2d = [(1024, 2**i) for i in range(0, 20, 4)]
+        sp_shapes_3d = [(64, 64, 2**i) for i in range(0, 15, 4)]
+        return special_shapes_2d + sp_shapes_3d
+
+    def get_input_iter(self, cur_dtype) -> Generator:
+        for shape in self.shapes:
+            inp = generate_tensor_input(shape, cur_dtype, self.device)
+            yield inp,
+
+    def get_tflops(self, op, *args, **kwargs):
+        shape = list(args[0].shape)
+        return torch.tensor(shape).prod().item()
+
+
+class UnaryPointwiseOutBenchmark(UnaryPointwiseBenchmark):
+    def get_input_iter(self, cur_dtype) -> Generator:
+        for shape in self.shapes:
+            inp = generate_tensor_input(shape, cur_dtype, self.device)
+            out = torch.empty_like(inp)
+            yield inp, {"out": out}
+
+
 def generate_tensor_input(shape, dtype, device):
     if dtype in FLOAT_DTYPES:
         return torch.randn(shape, dtype=dtype, device=device)

benchmark/test_abs.py

@@ -0,0 +1,21 @@
+import pytest
+import torch
+
+from . import attri_util as attrs
+from . import performance_utils as base
+
+
+@pytest.mark.abs
+def test_abs():
+    bench = base.UnaryPointwiseBenchmark(
+        op_name="abs", torch_op=torch.abs, dtypes=attrs.FLOAT_DTYPES
+    )
+    bench.run()
+
+
+@pytest.mark.abs_
+def test_abs_inplace():
+    bench = base.UnaryPointwiseBenchmark(
+        op_name="abs_", torch_op=torch.abs_, dtypes=attrs.FLOAT_DTYPES, is_inplace=True
+    )
+    bench.run()

benchmark/test_absolute.py

@@ -0,0 +1,13 @@
+import pytest
+import torch
+
+from . import attri_util as attrs
+from . import performance_utils as base
+
+
+@pytest.mark.absolute
+def test_absolute():
+    bench = base.UnaryPointwiseBenchmark(
+        op_name="absolute", torch_op=torch.absolute, dtypes=attrs.FLOAT_DTYPES
+    )
+    bench.run()

benchmark/test_acos.py

@@ -0,0 +1,13 @@
+import pytest
+import torch
+
+from . import attri_util as attrs
+from . import performance_utils as base
+
+
+@pytest.mark.acos
+def test_acos():
+    bench = base.UnaryPointwiseBenchmark(
+        op_name="acos", torch_op=torch.acos, dtypes=attrs.FLOAT_DTYPES
+    )
+    bench.run()

benchmark/test_alias_copy.py

@@ -0,0 +1,25 @@
+import pytest
+import torch
+
+from . import attri_util as attrs
+from . import performance_utils as base
+
+
+@pytest.mark.alias_copy
+def test_alias_copy():
+    bench = base.UnaryPointwiseBenchmark(
+        op_name="alias_copy",
+        torch_op=torch.ops.aten.alias_copy,
+        dtypes=attrs.FLOAT_DTYPES,
+    )
+    bench.run()
+
+
+@pytest.mark.alias_copy_out
+def test_alias_copy_out():
+    bench = base.UnaryPointwiseOutBenchmark(
+        op_name="alias_copy_out",
+        torch_op=torch.ops.aten.alias_copy,
+        dtypes=attrs.FLOAT_DTYPES,
+    )
+    bench.run()

benchmark/test_angle.py

@@ -0,0 +1,18 @@
+import pytest
+import torch
+
+from . import attri_util as attrs
+from . import performance_utils as base
+
+
+@pytest.mark.angle
+def test_angle():
+    bench = base.UnaryPointwiseBenchmark(
+        op_name="angle",
+        torch_op=torch.angle,
+        dtypes=attrs.COMPLEX_DTYPES
+        + [torch.float32]
+        + attrs.INT_DTYPES
+        + attrs.BOOL_DTYPES,
+    )
+    bench.run()

benchmark/test_apply_repetition_penalties.py

@@ -0,0 +1,68 @@
+import pytest
+import torch
+
+import flag_gems
+
+from . import attri_util as attrs
+from . import performance_utils as base
+
+
+class RepetitionPenaltyBenchmark(base.Benchmark):
+    def __init__(self, op_name, torch_op, dtypes):
+        super().__init__(op_name, torch_op, dtypes)
+        self.gems_op = None
+
+    def set_shapes(self, shape_file_path=None):
+        self.shapes = [
+            (1, 1024),
+            (1, 4096),
+            (1, 8192),
+            (8, 4096),
+            (16, 4096),
+            (32, 1024),
+            (8, 8192),
+            (64, 32000),
+        ]
+
+    def get_input_iter(self, dtype):
+        for shape in self.shapes:
+            num_seqs, vocab_size = shape
+            yield (
+                torch.randn(shape, dtype=dtype, device=self.device),
+                torch.randint(0, 2, shape, dtype=torch.bool, device=self.device),
+                torch.randint(0, 2, shape, dtype=torch.bool, device=self.device),
+                torch.empty(num_seqs, dtype=dtype, device=self.device).uniform_(
+                    1.0, 2.0
+                ),
+            )
+
+    def set_gems(self, gems_op):
+        self.gems_op = gems_op
+
+
+UNSUPPORTED_VENDORS = {
+    "metax",
+    "kunlunxin",
+    "iluvatar",
+    "mthreads",
+    "hygon",
+    "cambricon",
+}
+
+
+@pytest.mark.skipif(base.SkipVersion("vllm", "<0.4"), reason="vLLM <0.4 not supported")
+@pytest.mark.skipif(not torch.cuda.is_available(), reason="CUDA required")
+@pytest.mark.skipif(
+    flag_gems.vendor_name in UNSUPPORTED_VENDORS, reason="Vendor not supported"
+)
+@pytest.mark.apply_repetition_penalties
+def test_apply_repetition_penalties():
+    vllm_ops = pytest.importorskip("vllm._custom_ops")
+
+    bench = RepetitionPenaltyBenchmark(
+        op_name="apply_repetition_penalties",
+        torch_op=vllm_ops.apply_repetition_penalties,
+        dtypes=attrs.FLOAT_DTYPES,
+    )
+    bench.set_gems(flag_gems.apply_repetition_penalties)
+    bench.run()

benchmark/test_arcsinh.py

@@ -0,0 +1,34 @@
+import pytest
+import torch
+
+from . import attri_util as attrs
+from . import performance_utils as base
+
+
+@pytest.mark.arcsinh
+def test_arcsinh():
+    bench = base.UnaryPointwiseBenchmark(
+        op_name="arcsinh", torch_op=torch.arcsinh, dtypes=attrs.FLOAT_DTYPES
+    )
+    bench.run()
+
+
+@pytest.mark.arcsinh_
+def test_arcsinh_inplace():
+    bench = base.UnaryPointwiseBenchmark(
+        op_name="arcsinh_",
+        torch_op=lambda a: a.arcsinh_(),
+        dtypes=attrs.FLOAT_DTYPES,
+        is_inplace=True,
+    )
+    bench.run()
+
+
+@pytest.mark.arcsinh_out
+def test_arcsinh_out():
+    bench = base.UnaryPointwiseOutBenchmark(
+        op_name="arcsinh_out",
+        torch_op=torch.arcsinh,
+        dtypes=attrs.FLOAT_DTYPES,
+    )
+    bench.run()

benchmark/test_arctanh.py

@@ -0,0 +1,15 @@
+import pytest
+
+from . import attri_util as attrs
+from . import performance_utils as base
+
+
+@pytest.mark.arctanh_
+def test_arctanh_inplace():
+    bench = base.UnaryPointwiseBenchmark(
+        op_name="arctanh_",
+        torch_op=lambda a: a.arctanh_(),
+        dtypes=attrs.FLOAT_DTYPES,
+        is_inplace=True,
+    )
+    bench.run()

benchmark/test_asinh.py

@@ -0,0 +1,24 @@
+import pytest
+import torch
+
+from . import attri_util as attrs
+from . import performance_utils as base
+
+
+@pytest.mark.asinh
+def test_asinh():
+    bench = base.UnaryPointwiseBenchmark(
+        op_name="asinh", torch_op=torch.asinh, dtypes=attrs.FLOAT_DTYPES
+    )
+    bench.run()
+
+
+@pytest.mark.asinh_
+def test_asinh_inplace():
+    bench = base.UnaryPointwiseBenchmark(
+        op_name="asinh_",
+        torch_op=lambda a: a.asinh_(),
+        dtypes=attrs.FLOAT_DTYPES,
+        is_inplace=True,
+    )
+    bench.run()