triton-lang/triton

Pattern Findings

3109 matches across 18 categories. Click a row to expand file-level details.

Decorative Section Separators365 hits · 954 pts

Severity	File	Line	Snippet
MEDIUM	python/test/unit/cuda/test_libdevice_cuda.py	12	# -----------------------
MEDIUM	python/test/unit/cuda/test_libdevice_cuda.py	14	# -----------------------
MEDIUM	python/test/unit/language/test_standard.py	8	# ---------------
MEDIUM	python/test/unit/language/test_standard.py	10	# ---------------
MEDIUM	python/test/unit/language/test_standard.py	23	# ---------------
MEDIUM	python/test/unit/language/test_standard.py	25	# ---------------
MEDIUM	python/test/unit/language/test_standard.py	62	# ---------------
MEDIUM	python/test/unit/language/test_standard.py	64	# ---------------
MEDIUM	python/test/unit/language/test_core.py	369	# ---------------
MEDIUM	python/test/unit/language/test_core.py	371	# ---------------
MEDIUM	python/test/unit/language/test_core.py	515	# ---------------
MEDIUM	python/test/unit/language/test_core.py	556	# ---------------
MEDIUM	python/test/unit/language/test_core.py	558	# ---------------
MEDIUM	python/test/unit/language/test_core.py	583	# ---------------
MEDIUM	python/test/unit/language/test_core.py	585	# ---------------
MEDIUM	python/test/unit/language/test_core.py	615	# ----------
MEDIUM	python/test/unit/language/test_core.py	617	# ----------
MEDIUM	python/test/unit/language/test_core.py	664	# ------------------
MEDIUM	python/test/unit/language/test_core.py	666	# ------------------
MEDIUM	python/test/unit/language/test_core.py	681	# ----------------
MEDIUM	python/test/unit/language/test_core.py	683	# ----------------
MEDIUM	python/test/unit/language/test_core.py	792	# ----------------------------
MEDIUM	python/test/unit/language/test_core.py	794	# ----------------------------
MEDIUM	python/test/unit/language/test_core.py	808	# ---------------
MEDIUM	python/test/unit/language/test_core.py	810	# ---------------
MEDIUM	python/test/unit/language/test_core.py	902	# ---------------
MEDIUM	python/test/unit/language/test_core.py	904	# ---------------
MEDIUM	python/test/unit/language/test_core.py	935	# ----------------
MEDIUM	python/test/unit/language/test_core.py	937	# ----------------
MEDIUM	python/test/unit/language/test_core.py	1006	# -------------
MEDIUM	python/test/unit/language/test_core.py	1008	# -------------
MEDIUM	python/test/unit/language/test_core.py	1067	# ----------------
MEDIUM	python/test/unit/language/test_core.py	1069	# ----------------
MEDIUM	python/test/unit/language/test_core.py	1114	# ----------------
MEDIUM	python/test/unit/language/test_core.py	1116	# ----------------
MEDIUM	python/test/unit/language/test_core.py	1148	# ----------------
MEDIUM	python/test/unit/language/test_core.py	1150	# ----------------
MEDIUM	python/test/unit/language/test_core.py	1175	# ----------------
MEDIUM	python/test/unit/language/test_core.py	1177	# ----------------
MEDIUM	python/test/unit/language/test_core.py	1358	# ---------------
MEDIUM	python/test/unit/language/test_core.py	1360	# ---------------
MEDIUM	python/test/unit/language/test_core.py	1870	# ---------------
MEDIUM	python/test/unit/language/test_core.py	1872	# ---------------
MEDIUM	python/test/unit/language/test_core.py	2316	# ---------------
MEDIUM	python/test/unit/language/test_core.py	2318	# ---------------
MEDIUM	python/test/unit/language/test_core.py	2871	# ---------------
MEDIUM	python/test/unit/language/test_core.py	2873	# ---------------
MEDIUM	python/test/unit/language/test_core.py	2921	# ------------------------
MEDIUM	python/test/unit/language/test_core.py	2923	# ------------------------
MEDIUM	python/test/unit/language/test_core.py	3118	# ---------------
MEDIUM	python/test/unit/language/test_core.py	3120	# ---------------
MEDIUM	python/test/unit/language/test_core.py	3235	# ---------------
MEDIUM	python/test/unit/language/test_core.py	3237	# ---------------
MEDIUM	python/test/unit/language/test_core.py	4605	# ---------------
MEDIUM	python/test/unit/language/test_core.py	4607	# ---------------
MEDIUM	python/test/unit/language/test_core.py	4782	# --------------------
MEDIUM	python/test/unit/language/test_core.py	4784	# --------------------
MEDIUM	python/test/unit/language/test_core.py	5054	# -------------
MEDIUM	python/test/unit/language/test_core.py	5056	# -------------
MEDIUM	python/test/unit/language/test_core.py	4301	# ---------------
305 more matches not shown…

Hyper-Verbose Identifiers1065 hits · 891 pts

Severity	File	Line	Snippet
LOW	setup.py	550	def get_triton_version_suffix():
LOW	python/build_helpers.py	126	def _download_file_with_urllib(url: str, path: str, label: str):
LOW	python/build_helpers.py	380	def _get_thirdparty_package_cmake_vars(package: Package, helper_args: BuildHelperArgs):
LOW	python/build_helpers.py	414	def get_thirdparty_cmake_vars(packages: list[str], helper_args: BuildHelperArgs):
LOW	python/build_helpers.py	434	def write_thirdparty_cmake_vars(output: str, packages: list[str], helper_args: BuildHelperArgs):
LOW	python/build_helpers.py	482	def download_and_copy_dependencies(helper_args: BuildHelperArgs):
LOW	python/test/conftest.py	45	def fresh_knobs_including_libraries():
LOW	python/test/unit/test_debug.py	24	def _run_device_assert_barrier(device):
LOW	python/test/unit/test_debug.py	38	def _run_expect_zero_device_assert(device):
LOW	python/test/unit/test_debug.py	54	def test_expect_zero_device_assert(device):
LOW	python/test/unit/test_debug.py	76	def test_device_assert_barrier(device):
LOW	python/test/unit/test_debug.py	151	def test_sanitize_int_add_overflow(x, y, x_dtype, y_dtype, debug, should_overflow, device):
LOW	python/test/unit/test_debug.py	167	def test_sanitize_int_mul_overflow(x, y, x_dtype, y_dtype, debug, should_overflow, device):
LOW	python/test/unit/test_debug.py	182	def test_sanitize_int_sub_overflow(x, y, x_dtype, y_dtype, debug, should_overflow, device):
LOW	python/test/unit/test_perf_warning.py	12	def enable_diagnostics_context(value):
LOW	python/test/unit/test_perf_warning.py	98	def test_remark_vectorization(capfd, fresh_triton_cache):
LOW	python/test/unit/test_perf_warning.py	162	def test_remark_swp_op_before_operands(capfd, fresh_triton_cache):
LOW	python/test/unit/tools/test_aot.py	321	def compile_aot_kernel_no_specialization(dir, kernel_path, dtype, BM, BN, BK, target=None):
LOW	python/test/unit/tools/test_aot.py	366	def generate_matmul_test_data(dir, M, N, K):
LOW	python/test/unit/tools/test_aot.py	392	def test_compile_link_matmul_no_specialization():
LOW	python/test/unit/tools/test_aot.py	459	def test_launcher_has_no_available_kernel():
LOW	python/test/unit/tools/test_aot.py	497	def test_compile_link_autotune_matmul():
LOW	python/test/unit/tools/test_linear_layout.py	36	def test_operator_mul_identity():
LOW	python/test/unit/tools/test_linear_layout.py	42	def test_operator_mul_disjoint_dims():
LOW	python/test/unit/tools/test_linear_layout.py	75	def test_get_matrix_view_identity():
LOW	python/test/unit/tools/test_linear_layout.py	83	def test_get_matrix_view_strided():
LOW	python/test/unit/tools/test_linear_layout.py	92	def test_get_matrix_view_from_bases():
LOW	python/test/unit/tools/test_slice_kernel.py	45	def test_slice_kernel_basic_module_slicing(tmp_path):
LOW	python/test/unit/tools/test_slice_kernel.py	646	def test_slice_kernel_treats_assign_targets_as_locals(tmp_path):
LOW	python/test/unit/tools/test_slice_kernel.py	121	def test_slice_kernel_does_not_treat_site_packages_as_stdlib(tmp_path, monkeypatch):
LOW	python/test/unit/tools/test_slice_kernel.py	167	def test_slice_kernel_supports_injected_decorator_matchers(tmp_path):
LOW	python/test/unit/tools/test_slice_kernel.py	263	def test_slice_kernel_translate_to_gluon_keeps_tensor_method_rewrites(tmp_path):
LOW	python/test/unit/tools/test_slice_kernel.py	299	def test_slice_kernel_translate_to_gluon_inlines_descriptor_adapter(tmp_path):
LOW	python/test/unit/tools/test_slice_kernel.py	346	def test_slice_kernel_binds_local_imports(tmp_path):
LOW	python/test/unit/tools/test_slice_kernel.py	376	def test_slice_kernel_function_import(tmp_path):
LOW	python/test/unit/tools/test_slice_kernel.py	411	def test_slice_kernel_function_import_module_value(tmp_path):
LOW	python/test/unit/tools/test_slice_kernel.py	442	def test_slice_kernel_function_relative_import(tmp_path):
LOW	python/test/unit/tools/test_slice_kernel.py	488	def test_slice_kernel_function_from_import_value(tmp_path):
LOW	python/test/unit/tools/test_slice_kernel.py	519	def test_slice_kernel_function_absolute_import(tmp_path):
LOW	python/test/unit/tools/test_slice_kernel.py	565	def test_slice_kernel_function_module_relative_import(tmp_path):
LOW	python/test/unit/tools/test_slice_kernel.py	607	def test_slice_kernel_function_module_relative_import_leaf(tmp_path):
LOW	python/test/unit/tools/test_slice_kernel.py	671	def test_slice_kernel_treats_annassign_targets_as_locals(tmp_path):
LOW	python/test/unit/tools/test_slice_kernel.py	695	def test_slice_kernel_treats_assign_and_annassign_targets_as_locals(tmp_path):
LOW	python/test/unit/tools/test_slice_kernel.py	733	def test_slice_kernel_translate_to_gluon_avoids_double_descriptor_wrap(tmp_path):
LOW	python/test/unit/tools/test_slice_kernel.py	765	def test_translate_to_gluon_explicit_expand_dims_rewrites_layout(tmp_path):
LOW	python/test/unit/tools/test_slice_kernel.py	799	def test_translate_to_gluon_member_fn_expand_dims_rewrites_layout(tmp_path):
LOW	python/test/unit/tools/test_slice_kernel.py	833	def test_slice_kernel_public_imports():
LOW	python/test/unit/tools/test_irsource.py	10	def test_mlir_attribute_parsing(tmp_path: pathlib.Path) -> None:
LOW	python/test/unit/tools/test_disasm.py	25	def test_extract_handles_large_instruction_offsets(monkeypatch):
LOW	python/test/unit/tools/test_triton_to_gluon.py	100	def test_triton_to_gluon_dot_minimal(tmp_path):
LOW	python/test/unit/tools/test_triton_to_gluon.py	222	def test_triton_to_gluon_dot_scaled(
LOW	python/test/unit/tools/test_triton_to_gluon.py	272	def dot_transposed_operand_tile_kernel(
LOW	python/test/unit/tools/test_triton_to_gluon.py	308	def test_triton_to_gluon_dot_transposed_operands(lhs_transposed, rhs_transposed, tmp_path):
LOW	python/test/unit/tools/test_triton_to_gluon.py	435	def _skip_unless_descriptor_target():
LOW	python/test/unit/tools/test_triton_to_gluon.py	442	def test_triton_to_gluon_descriptor_roundtrip(tmp_path):
LOW	python/test/unit/tools/test_triton_to_gluon.py	466	def test_triton_to_gluon_descriptor_load_roundtrip(tmp_path):
LOW	python/test/unit/tools/test_triton_to_gluon.py	488	def make_tensor_descriptor_copy_kernel(x_ptr, y_ptr, M, N, BLOCK_M: tl.constexpr, BLOCK_N: tl.constexpr):
LOW	python/test/unit/tools/test_triton_to_gluon.py	505	def test_triton_to_gluon_make_tensor_descriptor(tmp_path, with_allocator):
LOW	python/test/unit/tools/test_triton_to_gluon.py	539	def test_triton_reshape_trans(tmp_path, TRANS_KIND):
LOW	python/test/unit/tools/test_triton_to_gluon.py	720	def gather_scatter_roundtrip_kernel(out_ptr, in_ptr, idx_ptr, X: tl.constexpr, Y: tl.constexpr, BLOCK_X: tl.constexpr,
1005 more matches not shown…

Over-Commented Block851 hits · 822 pts

Severity	File	Line	Snippet
LOW	unittest/Dialect/TritonGPU/SwizzleTest.cpp	1	#include "triton/Dialect/Triton/IR/Dialect.h"
LOW	unittest/Dialect/TritonGPU/SwizzleTest.cpp	241	int maskedLaneIdx;
LOW	unittest/Dialect/TritonGPU/SwizzleTest.cpp	261	} else {
LOW	…test/Dialect/TritonGPU/LinearLayoutConversionsTest.cpp	3501	// to load the data. If the layouts differ, MMA would compute wrong addresses.
LOW	bin/triton-llvm-opt.cpp	1	/// Trimmed down clone of llvm opt to be able to test triton custom llvm ir
LOW	bin/triton-tensor-layout.cpp	1	#include "RegisterTritonDialects.h"
LOW	bin/triton-tensor-layout.cpp	21	//
LOW	bin/RegisterTritonDialects.h	1	#pragma once
LOW	bin/RegisterTritonDialects.h	21	#include "TritonAMDGPUTransforms/Passes.h"
LOW	python/test/unit/cuda/test_tma_store_gemm.py	1	# Copyright (c) 2023 NVIDIA Corporation & Affiliates. All rights reserved.
LOW	python/test/unit/language/test_line_info.py	261	def kernel_basic(src, N, BLOCK_SIZE: tl.constexpr):
LOW	python/test/unit/language/test_line_info.py	281	# CHECK: } loc(#loc)
LOW	python/test/unit/language/test_line_info.py	481	kernel_info = kernel.warmup(torch.float32, torch.float32, grid=(1, ))
LOW	python/test/unit/language/test_line_info.py	501	# CHECK: %offsets_1 = arith.addi %offsets_0, %offsets : tensor<16xi32> loc(#loc16)
LOW	python/test/gluon/test_frontend.py	221	@gluon.jit
LOW	python/test/gluon/test_frontend.py	521	# CHECK-NEXT: [[OUTS:%.]]:3 = ttg.warp_specialize([[A]], [[B]], [[C]], [[A]], [[B]], [[C]]) {{.}}requestedRegis
LOW	python/test/gluon/test_frontend.py	581	# CHECK-DAG: [[BLOCKED_NW4:#.*]] = #ttg.blocked<{sizePerThread = [1], threadsPerWarp = [32], warpsPerCTA = [4], orde
LOW	python/test/gluon/test_frontend.py	1901	def test_auto_layout_broadcast():
LOW	python/triton/experimental/gsan/src/GSan.h	1	#pragma once
LOW	…on/experimental/gluon/language/amd/gfx1250/_layouts.py	141	# tile (i.e. the tile's shape is already expressed in memory order):
LOW	…on/experimental/gluon/language/amd/gfx1250/_layouts.py	161	# M=3: w2 w3
LOW	python/triton/tools/disasm.py	1	# MIT License
LOW	…triton/tools/triton_to_gluon_translator/ordered_set.py	1	"""
LOW	python/examples/gluon/02-conv-dgrad.py	61	#
LOW	python/examples/gluon/02-conv-fprop.py	61
LOW	python/examples/gluon/02-conv-fprop.py	481	output = torch.empty((N, out_h, out_w, Co), device=input_tensor.device, dtype=TORCH_GEMM_DTYPE)
LOW	python/examples/gluon/02-conv-wgrad.py	61	# grad_W[Co, RSCi] = grad_out[M, Co]^T @ im2col(input)[M, RSCi]
LOW	python/triton_kernels/triton_kernels/matmul.py	61	out_handles: torch.Tensor
LOW	python/triton_kernels/triton_kernels/matmul.py	201	k: torch.empty(v[0], device=allocation.device, dtype=v[1])
LOW	…_kernels/tensor_details/layout_details/hopper_value.py	121	init_shape = data.shape
LOW	…riton_kernels/tensor_details/layout_details/strided.py	1	from dataclasses import dataclass
LOW	…n_kernels/tensor_details/layout_details/torch_utils.py	1	import torch
LOW	…n_kernels/tensor_details/layout_details/torch_utils.py	21	# if not is_fp4:
LOW	python/tutorials/02-fused-softmax.py	61
LOW	python/tutorials/02-fused-softmax.py	181
LOW	python/tutorials/08-grouped-gemm.py	1	"""
LOW	python/tutorials/04-low-memory-dropout.py	21	# The dropout operator was first introduced in [SRIVASTAVA2014]_ as a way to improve the performance
LOW	python/tutorials/04-low-memory-dropout.py	81	output = dropout(x, x_keep=x_keep, p=p)
LOW	python/tutorials/04-low-memory-dropout.py	161
LOW	python/tutorials/03-matrix-multiplication.py	21	# -----------
LOW	python/tutorials/03-matrix-multiplication.py	41	# for k in range(0, K, BLOCK_SIZE_K):
LOW	python/tutorials/03-matrix-multiplication.py	61	# For a row-major 2D tensor :code:`X`, the memory location of :code:`X[i, j]` is given
LOW	python/tutorials/03-matrix-multiplication.py	81	# offs_k = tl.arange(0, BLOCK_SIZE_K)
LOW	python/tutorials/03-matrix-multiplication.py	101	#
LOW	python/tutorials/03-matrix-multiplication.py	121	# # Number of programs ids along the N axis
LOW	python/tutorials/03-matrix-multiplication.py	281	# We will advance this pointer as we move in the K direction
LOW	python/tutorials/05-layer-norm.py	21	# It takes a vector :math:`x` as input and produces a vector :math:`y` of the same shape as output.
LOW	python/tutorials/05-layer-norm.py	101	#
LOW	python/tutorials/10-block-scaled-matmul.py	21	python 10-block-scaled-matmul.py --format mxfp8 --K_range 8192 16384 --K_step 2048 --bench
LOW	python/tutorials/10-block-scaled-matmul.py	41	# where scale_a and scale_b are the blocked scale factors for the A and B matrices.
LOW	python/tutorials/10-block-scaled-matmul.py	61	# In order to conform with Triton's language semantics for dot_scaled, the scale factors
LOW	python/tutorials/10-block-scaled-matmul.py	81	#
LOW	python/tutorials/10-block-scaled-matmul.py	101	#
LOW	python/tutorials/gluon/05-wgmma.py	41	# Let's illustrate WGMMA with a trivial kernel launched with grid size (1, ).
LOW	python/tutorials/gluon/05-wgmma.py	61	#
LOW	python/tutorials/gluon/05-wgmma.py	81	#
LOW	python/tutorials/gluon/05-wgmma.py	101	#
LOW	python/tutorials/gluon/05-wgmma.py	181	# and all this does is immediately wait for 0 outstanding operations. In
LOW	python/tutorials/gluon/05-wgmma.py	261	# False 32 8.48
LOW	python/tutorials/gluon/05-wgmma.py	421	# We should try to pick the largest `n` for the WGMMA layout. Based on the
791 more matches not shown…

Unused Imports499 hits · 388 pts

Severity	File	Line
LOW	python/test/unit/runtime/test_build.py	1
LOW	python/test/unit/language/test_annotations.py	1
LOW	python/test/gsan/test_allocator.py	1
LOW	python/test/gsan/test_gsan_failures.py	1
LOW	python/test/gsan/test_symmetric_memory.py	1
LOW	python/test/gsan/test_gsan.py	1
LOW	python/triton/knobs.py	1
LOW	python/triton/__init__.py	8
LOW	python/triton/__init__.py	8
LOW	python/triton/__init__.py	8
LOW	python/triton/__init__.py	8
LOW	python/triton/__init__.py	8
LOW	python/triton/__init__.py	8
LOW	python/triton/__init__.py	8
LOW	python/triton/__init__.py	8
LOW	python/triton/__init__.py	8
LOW	python/triton/__init__.py	8
LOW	python/triton/__init__.py	20
LOW	python/triton/__init__.py	21
LOW	python/triton/__init__.py	21
LOW	python/triton/__init__.py	22
LOW	python/triton/__init__.py	22
LOW	python/triton/__init__.py	23
LOW	python/triton/__init__.py	24
LOW	python/triton/__init__.py	27
LOW	python/triton/__init__.py	28
LOW	python/triton/_utils.py	1
LOW	python/triton/experimental/gsan/__init__.py	1
LOW	python/triton/experimental/gsan/__init__.py	1
LOW	python/triton/experimental/gsan/_testing_utils.py	1
LOW	python/triton/experimental/gsan/_allocator.py	1
LOW	python/triton/experimental/gsan/symmetric_memory.py	1
LOW	python/triton/experimental/gsan/_testing.py	1
LOW	python/triton/experimental/gsan/_stream_sync.py	1
LOW	python/triton/experimental/gsan/_utils.py	1
LOW	python/triton/experimental/gluon/__init__.py	1
LOW	python/triton/experimental/gluon/__init__.py	1
LOW	python/triton/experimental/gluon/__init__.py	1
LOW	python/triton/experimental/gluon/__init__.py	2
LOW	python/triton/experimental/gluon/__init__.py	2
LOW	python/triton/experimental/gluon/__init__.py	3
LOW	python/triton/experimental/gluon/__init__.py	4
LOW	python/triton/experimental/gluon/_runtime.py	1
LOW	python/triton/experimental/gluon/_runtime.py	4
LOW	python/triton/experimental/gluon/language/__init__.py	1
LOW	python/triton/experimental/gluon/language/__init__.py	1
LOW	python/triton/experimental/gluon/language/__init__.py	1
LOW	python/triton/experimental/gluon/language/__init__.py	1
LOW	python/triton/experimental/gluon/language/__init__.py	1
LOW	python/triton/experimental/gluon/language/__init__.py	1
LOW	python/triton/experimental/gluon/language/__init__.py	1
LOW	python/triton/experimental/gluon/language/__init__.py	1
LOW	python/triton/experimental/gluon/language/__init__.py	1
LOW	python/triton/experimental/gluon/language/__init__.py	1
LOW	python/triton/experimental/gluon/language/__init__.py	1
LOW	python/triton/experimental/gluon/language/__init__.py	1
LOW	python/triton/experimental/gluon/language/__init__.py	1
LOW	python/triton/experimental/gluon/language/__init__.py	1
LOW	python/triton/experimental/gluon/language/__init__.py	1
LOW	python/triton/experimental/gluon/language/__init__.py	1
439 more matches not shown…

Hallucination Indicators15 hits · 145 pts

Severity	File	Line	Snippet
CRITICAL	python/test/unit/language/test_core.py	4001	if not is_interpreter() and triton.runtime.driver.active.utils.get_device_properties(
CRITICAL	python/test/unit/language/test_matmul.py	104	shared_mem_avail = triton.runtime.driver.active.utils.get_device_properties(0)["max_shared_mem"]
CRITICAL	python/test/gluon/test_frontend.py	786	%0 = tt.call @triton.experimental.gluon.language.nvidia.ampere.mbarrier.allocate_mbarrier__cNone_cTrue() : () -> !tt
CRITICAL	python/test/gluon/test_frontend.py	791	tt.func private @triton.experimental.gluon.language.nvidia.ampere.mbarrier.allocate_mbarrier__cNone_cTrue() -> !ttg.me
CRITICAL	python/test/gluon/test_frontend.py	3487	%13 = tt.call @triton.experimental.gluon.language._standard.zeros__Tc1T_cfp16_cAL() : () -> tensor<1xf16, #gluon.aut
CRITICAL	python/test/gluon/test_frontend.py	3515	tt.func private @triton.experimental.gluon.language._standard.zeros__Tc1T_cfp16_cAL() -> tensor<1xf16, #gluon.auto_enc
CRITICAL	python/test/gluon/test_frontend.py	3546	%1 = tt.call @triton.experimental.gluon.language._standard.zeros__Tc1T_cbf16_cAL() : () -> tensor<1xbf16, #gluon.aut
CRITICAL	python/test/gluon/test_frontend.py	3564	tt.func private @triton.experimental.gluon.language._standard.zeros__Tc1T_cbf16_cAL() -> tensor<1xbf16, #gluon.auto_en
CRITICAL	python/triton/language/core.py	3450	triton.runtime.driver.active.utils.set_printf_fifo_size(size_bytes)
CRITICAL	third_party/proton/proton/hooks/instrumentation.py	157	max_shared_mem = triton.runtime.driver.active.utils.get_device_properties(device)["max_shared_mem"]
CRITICAL	third_party/proton/proton/hooks/instrumentation.py	182	arch = triton.runtime.driver.active.utils.get_device_properties(device)["arch"].split(":")[0]
CRITICAL	third_party/amd/backend/driver.py	242	return triton.runtime.driver.active.utils.build_signature_metadata(kernel_signature)
CRITICAL	third_party/nvidia/backend/driver.py	182	return triton.runtime.driver.active.utils.build_signature_metadata(kernel_signature)
CRITICAL	third_party/nvidia/backend/driver.py	238	cu_tensor_map = triton.runtime.driver.active.utils.fill_tma_descriptor_im2col(
CRITICAL	third_party/nvidia/backend/driver.py	253	cu_tensor_map = triton.runtime.driver.active.utils.fill_tma_descriptor_tiled(

Deep Nesting151 hits · 128 pts

Severity	File	Line
LOW	setup.py	129
LOW	python/build_helpers.py	313
LOW	python/test/unit/tools/test_triton_to_gluon.py	520
LOW	python/test/unit/runtime/test_specialize.py	41
LOW	python/test/unit/language/test_conversions.py	15
LOW	python/test/unit/language/test_conversions.py	276
LOW	python/test/unit/language/test_core.py	382
LOW	python/test/unit/language/test_core.py	2557
LOW	python/test/unit/language/test_core.py	2735
LOW	python/test/unit/language/test_core.py	3240
LOW	python/test/unit/language/test_core.py	3404
LOW	python/test/unit/language/test_core.py	4216
LOW	python/test/unit/language/test_core.py	4441
LOW	python/test/unit/language/test_core.py	4612
LOW	python/test/unit/language/test_core.py	5064
LOW	python/test/unit/language/test_core.py	5537
LOW	python/test/unit/language/test_core.py	2561
LOW	python/test/unit/language/test_core.py	5067
LOW	python/test/unit/language/test_core.py	5540
LOW	python/test/unit/language/test_pipeliner.py	215
LOW	python/test/unit/language/test_subprocess.py	37
LOW	python/test/unit/language/test_matmul.py	489
LOW	python/test/unit/language/test_matmul.py	628
LOW	python/test/unit/language/test_matmul.py	684
LOW	python/test/unit/language/test_line_info.py	138
LOW	python/test/unit/language/test_line_info.py	183
LOW	python/test/unit/language/test_annotations.py	62
LOW	python/test/unit/language/print_helper.py	107
LOW	python/test/unit/language/test_tensor_descriptor.py	1555
LOW	python/test/unit/language/test_tensor_descriptor.py	1564
LOW	python/test/gsan/test_gsan_failures.py	94
LOW	python/test/gsan/test_symmetric_memory.py	369
LOW	python/test/gsan/test_gsan.py	421
LOW	python/test/gluon/test_lowerings.py	14
LOW	python/test/gluon/test_lowerings.py	699
LOW	python/test/gluon/test_frontend.py	4236
LOW	python/test/gluon/test_frontend.py	4239
LOW	python/test/gluon/test_core.py	1193
LOW	python/test/gluon/test_core.py	2164
LOW	python/test/gluon/test_core.py	2821
LOW	python/test/gluon/test_core.py	2933
LOW	python/test/gluon/test_core.py	1273
LOW	python/test/gluon/test_core.py	2171
LOW	python/test/gluon/test_consan.py	967
LOW	python/test/gluon/test_consan.py	2751
LOW	python/test/gluon/test_consan.py	995
LOW	python/test/gluon/test_consan.py	2762
LOW	python/test/gluon/test_fpsan.py	417
LOW	python/test/gluon/test_fpsan.py	715
LOW	python/test/gluon/test_fpsan.py	769
LOW	python/test/gluon/test_fpsan.py	921
LOW	python/test/regression/test_functional_regressions.py	137
LOW	python/test/regression/test_functional_regressions.py	140
LOW	python/triton/knobs.py	297
LOW	python/triton/testing.py	44
LOW	python/triton/testing.py	99
LOW	python/triton/testing.py	166
LOW	python/triton/testing.py	308
LOW	python/triton/testing.py	623
LOW	python/triton/testing.py	576
91 more matches not shown…

Verbosity Indicators41 hits · 79 pts

Severity	File	Line	Snippet
LOW	python/triton/runtime/jit.py	202	# The purpose of this function is to visit everything in `arguments`
LOW	python/tutorials/gluon/13-conv-im2col.py	149	# Step 1: Extract patches for each output position
LOW	python/tutorials/gluon/13-conv-im2col.py	156	# Step 2: Stack patches into im2col matrix A (M=4, K=4)
LOW	python/tutorials/gluon/13-conv-im2col.py	167	# Step 3: Reshape filter into weight matrix W (Co=1, K=4)
LOW	python/tutorials/gluon/13-conv-im2col.py	171	# Step 4: Output = A @ W^T
LOW	third_party/amd/lib/TritonAMDGPUToLLVM/TargetInfo.cpp	443	// Step 1: Right shift for 8 lanes.
LOW	third_party/amd/lib/TritonAMDGPUToLLVM/TargetInfo.cpp	446	// Step 2: Right shift for 4 lanes.
LOW	third_party/amd/lib/TritonAMDGPUToLLVM/TargetInfo.cpp	449	// Step 3: Right shift for 2 lanes.
LOW	third_party/amd/lib/TritonAMDGPUToLLVM/TargetInfo.cpp	452	// Step 4: Right shift for 1 lane.
LOW	third_party/amd/lib/TritonAMDGPUToLLVM/TargetInfo.cpp	455	// Step 5: Broadcast lane 15 of each row to all the lanes of its next row.
LOW	third_party/amd/lib/TritonAMDGPUToLLVM/TargetInfo.cpp	458	// Step 6: Broadcast lane 31 to lane 32-63.
LOW	third_party/amd/lib/TritonAMDGPUToLLVM/TargetInfo.cpp	463	// Step 7: Read the reduction result from lane 63 and broadcast with
LOW	third_party/amd/lib/TritonAMDGPUToLLVM/Utility.cpp	1038	// Step 1: extract EM bits for elements 0,2,4,6 and 1,3,5,7 respectively.
LOW	third_party/amd/lib/TritonAMDGPUToLLVM/Utility.cpp	1045	// Step 2: convert fp4 to fp8 using LUT
LOW	third_party/amd/lib/TritonAMDGPUToLLVM/Utility.cpp	1055	// Step 3: extract sign bits
LOW	third_party/amd/lib/TritonAMDGPUToLLVM/Utility.cpp	1059	// Step 4: assemble 4 packed fp8 values w/ sign
LOW	third_party/amd/lib/TritonAMDGPUToLLVM/Utility.cpp	1063	// Step 5: convert fp8 to fp32
LOW	third_party/amd/lib/TritonAMDGPUToLLVM/Utility.cpp	1145	// Step 1: extract EM bits for elements 0,2,4,6 and 1,3,5,7 respectively.
LOW	third_party/amd/lib/TritonAMDGPUToLLVM/Utility.cpp	1153	// Step 2: extract S bit for elements 0,2,4,6 and 1,3,5,7
LOW	third_party/amd/lib/TritonAMDGPUToLLVM/Utility.cpp	1161	// Step 3: Upcast elements 0,2,4,6 to 4 16-bit elements
LOW	third_party/amd/lib/TritonAMDGPUToLLVM/Utility.cpp	1190	// Step 4: Upcast elements 1,3,5,7 to 4 16-bit elements
LOW	third_party/amd/lib/TritonAMDGPUToLLVM/Utility.cpp	1220	// Step 5: Reorder 16-bit elements to be 0,1,2,3,4,5,6,7
LOW	…_party/amd/lib/TritonAMDGPUToLLVM/DotOpToLLVM/MFMA.cpp	485	// Step 1: construct each kBase-element vector by
LOW	…_party/amd/lib/TritonAMDGPUToLLVM/DotOpToLLVM/MFMA.cpp	497	// Step 2: process rawElems based on element type
LOW	…_party/amd/lib/TritonAMDGPUToLLVM/DotOpToLLVM/MFMA.cpp	517	// Step 3: Insert the processed vals into the ValueTable
LOW	…/lib/Transforms/WarpSpecialization/WSCodePartition.cpp	1229	// Step 1: collect all communications between producers and consumers.
LOW	…/lib/Transforms/WarpSpecialization/WSCodePartition.cpp	1240	// Step 2: group channels
LOW	…/lib/Transforms/WarpSpecialization/WSCodePartition.cpp	1249	// Step 3: reorder producer ops and the backward slices of the producer ops.
LOW	…/lib/Transforms/WarpSpecialization/WSCodePartition.cpp	1252	// Step 4: find top-level ops that contain a channel, also create new ForOps
LOW	…/lib/Transforms/WarpSpecialization/WSCodePartition.cpp	1272	// Step 5: Create buffers. An array of buffers for each channel.
LOW	…/lib/Transforms/WarpSpecialization/WSCodePartition.cpp	1280	// Step 6: Lower the loads. Also add local copy ops for non-load
LOW	…/lib/Transforms/WarpSpecialization/WSCodePartition.cpp	1290	// Step 7: Create tokens. A set of tokens for each group of channels for
LOW	…/lib/Transforms/WarpSpecialization/WSCodePartition.cpp	1301	// Step 8: add async communication ops (ProducerAcquire etc). Also lower
LOW	…/hopper/lib/Transforms/WarpSpecialization/WSBuffer.cpp	304	// Step 1: Append accumCnts as forOp arguments.
LOW	…/hopper/lib/Transforms/WarpSpecialization/WSBuffer.cpp	308	// Step 2: Add accumCnts to yieldOp.
LOW	…/hopper/lib/Transforms/WarpSpecialization/WSBuffer.cpp	317	// Step 3: Create loop arguments for the new ForOp.
LOW	…/hopper/lib/Transforms/WarpSpecialization/WSBuffer.cpp	326	// Step 4: Create newForOp and take the region of the original forOp.
LOW	…/hopper/lib/Transforms/WarpSpecialization/WSBuffer.cpp	332	// Step 5: Replace forOp with newForOp.
LOW	third_party/nvidia/lib/NVGPUToLLVM/NVGPUToLLVMPass.cpp	604	// TODO: pred will have a long liverange, we need to check if this is a
LOW	lib/Dialect/TritonGPU/IR/LinearLayoutConversions.cpp	1065	// Step 3: Along the "register" dim, remove any all-zero bases.
LOW	…t/TritonInstrument/Transforms/ConcurrencySanitizer.cpp	673	// Op is writing to the buffer, we need to check if anything else

Excessive Try-Catch Wrapping45 hits · 51 pts

Severity	File	Line	Snippet
LOW	setup.py	528	except Exception:
LOW	setup.py	536	except Exception:
MEDIUM	setup.py	524	def get_git_commit_hash(length=8):
MEDIUM	setup.py	532	def get_git_branch():
LOW	python/test/unit/runtime/test_cache.py	821	except Exception as e:
LOW	python/test/unit/language/test_core.py	5044	except Exception as e:
LOW	python/test/unit/language/test_decorator.py	27	except Exception as e:
LOW	python/test/backend/test_mir_stage.py	11	except Exception:
MEDIUM	python/test/backend/test_mir_stage.py	8	def is_hip():
LOW	python/triton/_internal_testing.py	262	except Exception as e:
LOW	python/triton/experimental/gsan/symmetric_memory.py	152	except Exception:
LOW	python/triton/experimental/gsan/symmetric_memory.py	252	except Exception:
LOW	python/triton/experimental/gsan/symmetric_memory.py	444	except Exception:
MEDIUM	python/triton/experimental/gsan/symmetric_memory.py	249	def __del__(self) -> None:
LOW	python/triton/experimental/gsan/_utils.py	109	except Exception:
LOW	python/triton/runtime/cache.py	81	except Exception:
LOW	python/triton/runtime/cache.py	225	except Exception:
LOW	python/triton/runtime/_async_compile.py	22	except Exception:
LOW	python/triton/runtime/interpreter.py	1406	except Exception as e:
LOW	python/triton/runtime/interpreter.py	1447	except Exception:
LOW	python/triton/runtime/interpreter.py	1540	except Exception as e:
LOW	python/triton/runtime/autotuner.py	159	except Exception as e:
LOW	python/triton/compiler/compiler.py	308	except Exception as e:
LOW	python/triton/compiler/code_generator.py	1375	except Exception as e:
LOW	python/triton/compiler/code_generator.py	1426	except Exception as e:
LOW	python/triton/compiler/code_generator.py	1597	except Exception as e:
LOW	python/triton/compiler/code_generator.py	1632	except Exception as e:
LOW	python/examples/gluon/02-conv-dgrad.py	980	except Exception:
MEDIUM	python/examples/gluon/02-conv-dgrad.py	937	def _benchmark_dgrad_config(
LOW	python/examples/gluon/02-conv-wgrad.py	735	except Exception:
MEDIUM	python/examples/gluon/02-conv-wgrad.py	692	def _benchmark_wgrad_config(
LOW	third_party/proton/test/test_instrumentation.py	938	except Exception as e:
MEDIUM	third_party/proton/test/test_instrumentation.py	935	def run_kernel():
LOW	third_party/proton/test/test_api.py	98	except Exception as e:
LOW	third_party/proton/test/test_api.py	120	except Exception as e:
LOW	third_party/proton/test/test_api.py	133	except Exception as e:
LOW	third_party/proton/test/test_api.py	287	except Exception as e:
LOW	third_party/proton/test/test_api.py	404	except Exception as e:
LOW	third_party/proton/test/test_api.py	414	except Exception as e:
LOW	…ton/test/reproducers/cupti_graph_replay_heap_growth.py	275	except Exception:
LOW	third_party/proton/proton/proton.py	46	except Exception as e:
MEDIUM	third_party/proton/proton/proton.py	47	print("An error occurred while executing the script:")
LOW	third_party/amd/python/test/test_gluon_gfx1250.py	1994	except Exception as e:
LOW	third_party/amd/backend/driver.py	45	except Exception:
LOW	third_party/nvidia/backend/driver.py	58	except Exception:

Cross-File Repetition10 hits · 50 pts

Severity	File	Snippet
HIGH	python/test/unit/tools/test_slice_kernel.py	import math def common_util() -> int: return math.prod([11, 33])
HIGH	python/test/unit/tools/test_slice_kernel.py	import math def common_util() -> int: return math.prod([11, 33])
HIGH	python/test/unit/tools/test_slice_kernel.py	import math def common_util() -> int: return math.prod([11, 33])
HIGH	python/test/unit/tools/test_slice_kernel.py	import math def common_util() -> int: return math.prod([11, 33])
HIGH	python/test/unit/tools/test_slice_kernel.py	import math def common_util() -> int: return math.prod([11, 33]) def helper() -> none: common_util() def kernel() -> non
HIGH	python/test/unit/tools/test_slice_kernel.py	import math def common_util() -> int: return math.prod([11, 33]) def helper() -> none: common_util() def kernel() -> non
HIGH	python/test/unit/tools/test_slice_kernel.py	import math def common_util() -> int: return math.prod([11, 33]) def helper() -> none: common_util() def kernel() -> non
HIGH	…iton/experimental/gluon/language/amd/rdna4/__init__.py	computes matrix-multiplication of a * b + acc using amd wmma instruction. args: a (tensor): the operand a to be multipli
HIGH	…iton/experimental/gluon/language/amd/rdna3/__init__.py	computes matrix-multiplication of a * b + acc using amd wmma instruction. args: a (tensor): the operand a to be multipli
HIGH	…on/experimental/gluon/language/amd/gfx1250/__init__.py	computes matrix-multiplication of a * b + acc using amd wmma instruction. args: a (tensor): the operand a to be multipli

Self-Referential Comments15 hits · 43 pts

Severity	File	Line	Snippet
MEDIUM	python/build_helpers.py	356	# Create a stable symlink that doesn't include revision
MEDIUM	python/test/gluon/test_core.py	2679	# Create a test matrix with known values
MEDIUM	python/test/gluon/test_core.py	3227	# Create a test matrix [N, M]
MEDIUM	python/test/gluon/test_core.py	3388	# Create a test tensor [N, M, P]
MEDIUM	python/triton/experimental/gluon/language/_semantic.py	618	# Create the warp specialize op.
MEDIUM	python/triton/experimental/gluon/language/_layouts.py	617	# Create a idendity mapping based on shape_per_cta + order
MEDIUM	python/triton/runtime/jit.py	400	# Create the function argument list and the dict entries for the return statement
MEDIUM	python/triton/runtime/jit.py	1138	# Create a bound function to support constexpr_function methods
MEDIUM	python/triton/language/core.py	1659	# Define the wrapped Triton value type.
MEDIUM	python/tutorials/02-fused-softmax.py	173	# Create a number of persistent programs.
MEDIUM	python/tutorials/01-vector-add.py	46	# Create a mask to guard memory operations against out-of-bounds accesses.
MEDIUM	docs/conf.py	6	# This file is execfile()d with the current directory set to its
MEDIUM	third_party/proton/proton/hooks/instrumentation.py	39	# Create the buffer
MEDIUM	third_party/proton/proton/hooks/instrumentation.py	111	# Create the appropriate mode instance
MEDIUM	third_party/amd/python/test/test_gluon_gfx1250.py	2212	# Create a flattened view into the nD reference to unify the indexing logic over all dimensions

AI Slop Vocabulary15 hits · 40 pts

Severity	File	Line	Snippet
MEDIUM	…riton/tools/triton_to_gluon_translator/slice_kernel.py	112	# HACK: An AST rewrite would be more robust, but this works for now.
MEDIUM	python/triton/runtime/cache.py	115	# use temp dir to be robust against program interruptions
LOW	python/triton/runtime/interpreter.py	1444	# e.g., dynamically generated functions, we cannot rewrite it so just return the original function
MEDIUM	python/triton_kernels/bench/bench_dense_matmul.py	3	# This is not meant to be a comprehensive benchmark of triton_kernels.
MEDIUM	python/triton_kernels/bench/bench_dense_matmul.py	135	# even for the fp16 benchmark path. Match that when reproducing the OpenAI harness.
MEDIUM	python/tutorials/09-persistent-matmul.py	411	# Enable warp specialization to leverage async warp scheduling in the GPU.
MEDIUM	python/tutorials/10-block-scaled-matmul.py	390	# With 256 elements we better utilize the L2 and don't require the TMA
MEDIUM	python/tutorials/gluon/11-tcgen05-mma-scaled.py	654	# elements. With 256 bytes along the inner dimension, we better utilize the
MEDIUM	python/tutorials/gluon/08-warp-specialization.py	369	# specialization is especially powerful when used to build intricate schedules
MEDIUM	python/tutorials/gluon/02-layouts.py	868	# Linear layouts are extremely poweful, and can be used in conjunction with
MEDIUM	python/tutorials/gluon/07-persistence.py	841	# work in conjunction with the GPU, getting the best of both worlds. This is
LOW	third_party/amd/python/test/test_gluon_gfx1250.py	2776	# In practice, since we know that phase is initialized to 0, we can just set count = NUM_WARPS * WARP_SIZE and call
LOW	third_party/amd/python/test/test_gluon_gfx1250.py	2837	# In practice, since we know that phase is initialized to 0, we can just set count = NUM_WARPS and call directly ttg
MEDIUM	…party/amd/lib/TritonAMDGPUTransforms/ScheduleLoops.cpp	210	/// which is not coalesced. To better utilize global memory bandwidth,
MEDIUM	…party/amd/lib/TritonAMDGPUToLLVM/LoadStoreOpToLLVM.cpp	2248	// In the case of unpaired f16 elements utilize dpp instructions to

Overly Generic Function Names17 hits · 18 pts

Severity	File	Line	Snippet
LOW	python/test/unit/tools/test_slice_kernel.py	636	def helper() -> None:
LOW	python/test/unit/tools/test_slice_kernel.py	652	def helper() -> int:
LOW	python/test/unit/tools/test_slice_kernel.py	130	def helper() -> int:
LOW	python/test/unit/tools/test_slice_kernel.py	157	def helper() -> int:
LOW	python/test/unit/tools/test_slice_kernel.py	382	def helper() -> None:
LOW	python/test/unit/tools/test_slice_kernel.py	397	def helper() -> None:
LOW	python/test/unit/tools/test_slice_kernel.py	417	def helper():
LOW	python/test/unit/tools/test_slice_kernel.py	432	def helper():
LOW	python/test/unit/tools/test_slice_kernel.py	455	def helper() -> None:
LOW	python/test/unit/tools/test_slice_kernel.py	474	def helper() -> None:
LOW	python/test/unit/tools/test_slice_kernel.py	494	def helper():
LOW	python/test/unit/tools/test_slice_kernel.py	509	def helper():
LOW	python/test/unit/tools/test_slice_kernel.py	536	def helper() -> None:
LOW	python/test/unit/tools/test_slice_kernel.py	555	def helper() -> None:
LOW	python/test/unit/tools/test_slice_kernel.py	578	def helper() -> None:
LOW	python/test/unit/tools/test_slice_kernel.py	597	def helper() -> None:
LOW	python/test/unit/tools/test_slice_kernel.py	620	def helper() -> None:

Redundant / Tautological Comments10 hits · 13 pts

Severity	File	Line	Snippet
LOW	python/build_helpers.py	96	# Set timeout to 300 seconds to prevent the request from hanging forever.
LOW	python/triton/compiler/code_generator.py	133	# Check if the given syntax node has an "early" return
LOW	utils/generate-test-checks.py	177	# Check if an existing variable exists for this name.
LOW	utils/generate-test-checks.py	320	# Check if this is an attribute definition and process it
LOW	third_party/proton/proton/hooks/hook.py	93	# Check if any other sessions rely on this hook
LOW	third_party/amd/python/test/address_sanitizer_helper.py	23	#Set access to go out of bounds for ASAN test
LOW	third_party/amd/backend/driver.py	330	# Check if cooperative groups are supported on the device.
LOW	.github/workflows/integration-tests-nvidia.yml	41	# Check if files exist before proceeding
LOW	.github/workflows/integration-tests-amd.yml	66	# Check if files exist before proceeding
LOW	.github/workflows/build-macos.yml	37	# Check if files exist before proceeding

Cross-Language Confusion2 hits · 10 pts

Severity	File	Line	Snippet
HIGH	python/test/unit/tools/test_aot.py	147	while (fscanf(file, "%hd,", &buffer[index]) != EOF && index < size) {
HIGH	python/triton/experimental/gsan/symmetric_memory.py	396	stack.push(listener)

Slop Phrases4 hits · 8 pts

Severity	File	Line	Snippet
LOW	python/test/unit/language/test_core.py	6446	# reuse the temp variable and make sure to check that it isn't creating incorrect IR.
LOW	python/tutorials/01-vector-add.py	74	# - Don't forget to pass meta-parameters as keywords arguments.
MEDIUM	python/tutorials/gluon/14-multicta.py	359	# It's worth noting that the pattern changes a bit once the TMA has to wait on the tcgen05_mma
MEDIUM	python/tutorials/gluon/02-layouts.py	839	# require reordering registers within a thread (which is free), you can use

Docstring Block Structure1 hit · 5 pts

Severity	File	Line	Snippet
HIGH	…/triton/experimental/gluon/language/amd/gfx1250/tdm.py	161	Update selected fields of a TDM descriptor; return a new descriptor SSA value. Each parameter is independently opti

Dead Code2 hits · 4 pts

Severity	File	Line	Snippet
MEDIUM	python/test/unit/language/test_compile_errors.py	171
MEDIUM	python/triton_kernels/triton_kernels/tensor.py	98

Example Usage Blocks1 hit · 2 pts

Severity	File	Line	Snippet
LOW	bin/triton-tensor-layout.cpp	23	// Example usage:

Score History

Severity Breakdown

Pattern Findings