diff --git a/contrib/c-integer-semantics.py b/contrib/c-integer-semantics.py index 1cc1a142d9b76725b9b4a700052c877fa2861fbe..38b5b673e872c31e244fa5ecfab3419e2cbe08cb 100644 --- a/contrib/c-integer-semantics.py +++ b/contrib/c-integer-semantics.py @@ -139,7 +139,7 @@ def main(): if cresult != presult: print(a, b, cresult, presult) - #print(int_mod(552, -918), 552 % -918) + # print(int_mod(552, -918), 552 % -918) print(cmod(23, -11), 23 % -11) diff --git a/examples/python/ispc-stream-harness.py b/examples/python/ispc-stream-harness.py index 32da802dfaad464c0137db790d19d64cb5557d76..94bb5d9032b58256385e42219e5ebd6566034f77 100644 --- a/examples/python/ispc-stream-harness.py +++ b/examples/python/ispc-stream-harness.py @@ -88,16 +88,16 @@ def main(): [("tasksys.cpp", tasksys_source)], cxx_options=["-g", "-fopenmp", "-DISPC_USE_OMP"], ispc_options=([ - #"-g", "--no-omit-frame-pointer", + # "-g", "--no-omit-frame-pointer", "--target=avx2-i32x8", "--opt=force-aligned-memory", "--opt=disable-loop-unroll", - #"--opt=fast-math", - #"--opt=disable-fma", + # "--opt=fast-math", + # "--opt=disable-fma", ] + (["--addressing=64"] if INDEX_DTYPE == np.int64 else []) ), - #ispc_bin="/home/andreask/pack/ispc-v1.9.0-linux/ispc", + # ispc_bin="/home/andreask/pack/ispc-v1.9.0-linux/ispc", quiet=False, ) diff --git a/loopy/check.py b/loopy/check.py index 3f65ad7ab4a62363c130efefb6e0ab2ea2ac755b..9c6e4f9516685ef43730dd25068950ebf0790333 100644 --- a/loopy/check.py +++ b/loopy/check.py @@ -1847,9 +1847,9 @@ def check_implemented_domains(kernel, implemented_domains, code=None): pt = diff_set.sample_point() assert not pt.is_void() - #pt_set = isl.Set.from_point(pt) - #lines.append("point implemented: %s" % (pt_set <= insn_impl_domain)) - #lines.append("point desired: %s" % (pt_set <= desired_domain)) + # pt_set = isl.Set.from_point(pt) + # lines.append("point implemented: %s" % (pt_set <= insn_impl_domain)) + # lines.append("point desired: %s" % (pt_set <= desired_domain)) iname_to_dim = pt.get_space().get_var_dict() point_axes = [] diff --git a/loopy/kernel/creation.py b/loopy/kernel/creation.py index 0f7e783a3381eb2828e7ebc308e6f467842b6b13..10aab71f1ae2cc6f51a637ae4cff762e9850459c 100644 --- a/loopy/kernel/creation.py +++ b/loopy/kernel/creation.py @@ -902,7 +902,7 @@ def parse_instructions(instructions, defines): insn_options_stack.append(options) - #add to the if_stack + # add to the if_stack if_options = options.copy() if_options["insn_predicates"] = options["predicates"] if_predicates_stack.append(if_options) @@ -953,7 +953,7 @@ def parse_instructions(instructions, defines): | additional_preds ) if_options["predicates"] = additional_preds - #hold on to this for comparison / stack popping later + # hold on to this for comparison / stack popping later if_options["insn_predicates"] = options["predicates"] insn_options_stack.append(options) @@ -967,7 +967,7 @@ def parse_instructions(instructions, defines): if insn == "end": obj = insn_options_stack.pop() - #if this object is the end of an if statement + # if this object is the end of an if statement if obj["predicates"] == if_predicates_stack[-1]["insn_predicates"] and\ if_predicates_stack[-1]["insn_predicates"] and\ obj["within_inames"] == if_predicates_stack[-1]["within_inames"]: diff --git a/loopy/library/reduction.py b/loopy/library/reduction.py index c4374eee1e7d1e65851af6099614760241d1c236..648b7a197a0d23e4a4742939983928b47de32528 100644 --- a/loopy/library/reduction.py +++ b/loopy/library/reduction.py @@ -179,10 +179,10 @@ def get_le_neutral(dtype): if dtype.numpy_dtype.kind == "f": # OpenCL 1.2, section 6.12.2 if dtype.numpy_dtype.itemsize == 4: - #float + # float return var("INFINITY") elif dtype.numpy_dtype.itemsize == 8: - #double + # double return var("HUGE_VAL") elif dtype.numpy_dtype.kind == "i": @@ -210,10 +210,10 @@ def get_ge_neutral(dtype): if dtype.numpy_dtype.kind == "f": # OpenCL 1.2, section 6.12.2 if dtype.numpy_dtype.itemsize == 4: - #float + # float return -var("INFINITY") elif dtype.numpy_dtype.itemsize == 8: - #double + # double return -var("HUGE_VAL") elif dtype.numpy_dtype.kind == "i": # OpenCL 1.1, section 6.11.3 diff --git a/loopy/preprocess.py b/loopy/preprocess.py index f6f42be59ecd68863888ec5e60517524bfed4767..c787cedbfe01d3283f4b1230efdb91f7cd77c06a 100644 --- a/loopy/preprocess.py +++ b/loopy/preprocess.py @@ -476,7 +476,7 @@ def check_atomic_loads(kernel): accessed_atomic_vars = (insn.dependency_names() & atomicity_candidates)\ - {insn.assignee_var_names()[0]} if not accessed_atomic_vars <= atomic_accesses: - #if we're missing some + # if we're missing some missed = accessed_atomic_vars - atomic_accesses for x in missed: if {x} & atomicity_candidates: diff --git a/loopy/schedule/__init__.py b/loopy/schedule/__init__.py index a2c82f1253a6e8393e8f35f29427b5fcb86fa849..136f0e6014adf1ccce94265a2c8d71cfc37a5441 100644 --- a/loopy/schedule/__init__.py +++ b/loopy/schedule/__init__.py @@ -908,8 +908,8 @@ def _generate_loop_schedules_internal( if debug.debug_length == len(debug.longest_rejected_schedule): print("WHY IS THIS A DEAD-END SCHEDULE?") - #if len(schedule) == 2: - #from pudb import set_trace; set_trace() + # if len(schedule) == 2: + # from pudb import set_trace; set_trace() # }}} diff --git a/loopy/target/c/c_execution.py b/loopy/target/c/c_execution.py index b1685cad1781db6e11f597edf7a3029f74461775..ebf2fd611225bc1415cca3a82ba68b140dc91f58 100644 --- a/loopy/target/c/c_execution.py +++ b/loopy/target/c/c_execution.py @@ -152,7 +152,7 @@ class CExecutionWrapperGenerator(ExecutionWrapperGeneratorBase): gen("{} = {}.strides".format(strify(expected_strides), arg.name)) - #check strides + # check strides if not skip_arg_checks: strides_check_expr = self.get_strides_check_expr( [strify(s) for s in sym_shape], diff --git a/loopy/target/execution.py b/loopy/target/execution.py index dcf162ae7ec3553b94dcdd3808226ca601fb016d..fe0c7442b36951b77ad7d5ad618bc9b4d335eebc 100644 --- a/loopy/target/execution.py +++ b/loopy/target/execution.py @@ -695,7 +695,7 @@ class ExecutionWrapperGeneratorBase(ABC): self.generate_value_arg_check(gen, program[entrypoint], kai) args = self.generate_arg_setup(gen, program[entrypoint], kai) - #FIXME: should we make this as a dict as well. + # FIXME: should we make this as a dict as well. host_program_name = codegen_result.host_programs[entrypoint].name self.generate_invocation(gen, program[entrypoint], kai, diff --git a/loopy/target/opencl.py b/loopy/target/opencl.py index a2ac18f1f2f078e619f0086442b7686a5f401465..38426323003c3ca1119df073ad0c2c091de0e785 100644 --- a/loopy/target/opencl.py +++ b/loopy/target/opencl.py @@ -257,7 +257,7 @@ class OpenCLCallable(ScalarCallable): for id in arg_id_to_dtype: if not -1 <= id <= 1: - #FIXME: Do we need to raise here?: + # FIXME: Do we need to raise here?: # The pattern we generally follow is that if we don't find # a function, then we just return None raise LoopyError("%s can take only two arguments." % name) diff --git a/loopy/target/pyopencl_execution.py b/loopy/target/pyopencl_execution.py index b65bdc66ee832827a2c4fe3cf7dd0c1bc54d715a..02dbcfc899138549fe3650d69e3ab07363ddc9ef 100644 --- a/loopy/target/pyopencl_execution.py +++ b/loopy/target/pyopencl_execution.py @@ -313,7 +313,7 @@ class PyOpenCLExecutor(ExecutorBase): dev_code = codegen_result.device_code() if t_unit[self.entrypoint].options.write_code: - #FIXME: redirect to "translation unit" level option as well. + # FIXME: redirect to "translation unit" level option as well. output = dev_code if self.t_unit[self.entrypoint].options.allow_terminal_colors: output = get_highlighted_code(output) @@ -327,13 +327,13 @@ class PyOpenCLExecutor(ExecutorBase): outf.write(output) if t_unit[self.entrypoint].options.edit_code: - #FIXME: redirect to "translation unit" level option as well. + # FIXME: redirect to "translation unit" level option as well. from pytools import invoke_editor dev_code = invoke_editor(dev_code, "code.cl") import pyopencl as cl - #FIXME: redirect to "translation unit" level option as well. + # FIXME: redirect to "translation unit" level option as well. cl_program = ( cl.Program(self.context, dev_code) .build(options=t_unit[self.entrypoint].options.build_options)) diff --git a/loopy/transform/data.py b/loopy/transform/data.py index cb5fe829d2a5b847d5b17e7dcf122e2b62ca0e3f..87cb7befb1b9e4ab1a1c9b62f168b5d1874cfa38 100644 --- a/loopy/transform/data.py +++ b/loopy/transform/data.py @@ -654,7 +654,7 @@ def set_argument_order(kernel, arg_names): :arg arg_names: A list (or comma-separated string) or argument names. All arguments must be in this list. """ - #FIXME: @inducer -- shoulld this only affect the root kernel, or should it + # FIXME: @inducer -- shoulld this only affect the root kernel, or should it # take a within? if isinstance(arg_names, str): diff --git a/proto-tests/test_fem_assembly.py b/proto-tests/test_fem_assembly.py index 2bee025f11c5d2e4149a9a6738f2c88a2ebb67ab..0f4e95b4715465a30596f3761ee88cc9e541d5b3 100644 --- a/proto-tests/test_fem_assembly.py +++ b/proto-tests/test_fem_assembly.py @@ -123,7 +123,7 @@ def test_laplacian_stiffness(ctx_factory): loop_priority=loop_prio) kernel_gen = lp.check_kernels(kernel_gen, dict(Nc=Nc)) - #print lp.preprocess_kernel(var_knl) + # print lp.preprocess_kernel(var_knl) lp.auto_test_vs_ref(seq_knl, ctx, kernel_gen, op_count=0, op_label="GFlops", diff --git a/proto-tests/test_sem.py b/proto-tests/test_sem.py index 56a0d5e2569ed9722f09c4536d1ad60958badf67..5155a001165305333b0634a220de79be839bcc76 100644 --- a/proto-tests/test_sem.py +++ b/proto-tests/test_sem.py @@ -56,11 +56,11 @@ def test_laplacian(ctx_factory): ], name="semlap", assumptions="K>=1") - #print lp.preprocess_kernel(knl, cse_ok=True) - #1/0 + # print lp.preprocess_kernel(knl, cse_ok=True) + # 1/0 # - #print knl - #1/0 + # print knl + # 1/0 knl = lp.realize_cse(knl, "urf", np.float32, ["o1"]) knl = lp.realize_cse(knl, "usf", np.float32, ["o2"]) knl = lp.realize_cse(knl, "utf", np.float32, ["o3"]) @@ -76,7 +76,7 @@ def test_laplacian(ctx_factory): if 0: pass #seq_knl = lp.add_prefetch(knl, "G", ["gi", "m", "j", "k"], "G[gi,e,m,j,k]", default_tag="l.auto") # noqa - #seq_knl = lp.add_prefetch(seq_knl, "D", ["m", "j"], default_tag="l.auto") + # seq_knl = lp.add_prefetch(seq_knl, "D", ["m", "j"], default_tag="l.auto") #seq_knl = lp.add_prefetch(seq_knl, "u", ["i", "j", "k"], "u[*,i,j,k]", default_tag="l.auto") # noqa else: seq_knl = knl @@ -87,11 +87,11 @@ def test_laplacian(ctx_factory): knl = lp.add_prefetch(knl, "D", ["m", "j"], default_tag="l.auto") #knl = lp.add_prefetch(knl, "u", ["i", "j", "k"], "u[*,i,j,k]", default_tag="l.auto") # noqa - #knl = lp.split_iname(knl, "e_inner", 4, inner_tag="ilp") + # knl = lp.split_iname(knl, "e_inner", 4, inner_tag="ilp") - #print seq_knl - #print lp.preprocess_kernel(knl) - #1/0 + # print seq_knl + # print lp.preprocess_kernel(knl) + # 1/0 knl = lp.tag_inames(knl, dict(i="l.0", j="l.1")) @@ -164,11 +164,11 @@ def test_laplacian_lmem(ctx_factory): default_tag="l.auto") knl = lp.add_prefetch(knl, "u", ["i", "j", "k"], "u[*,i,j,k]", default_tag="l.auto") - #knl = lp.split_iname(knl, "e_inner", 4, inner_tag="ilp") + # knl = lp.split_iname(knl, "e_inner", 4, inner_tag="ilp") - #print seq_knl - #print lp.preprocess_kernel(knl) - #1/0 + # print seq_knl + # print lp.preprocess_kernel(knl) + # 1/0 knl = lp.tag_inames(knl, dict(i="l.0", j="l.1")) @@ -240,8 +240,8 @@ def test_laplacian_lmem_ilp(ctx_factory): knl = lp.add_prefetch(knl, "D", ["m", "j"], default_tag="l.auto") - #print seq_knl - #1/0 + # print seq_knl + # 1/0 knl = lp.tag_inames(knl, dict(i="l.0", j="l.1")) @@ -447,7 +447,7 @@ def test_advect_dealias(ctx_factory): knl = lp.tag_inames(knl, dict(i="l.0", j="l.1")) print(knl) - #1/0 + # 1/0 kernel_gen = lp.generate_loop_schedules(knl) kernel_gen = lp.check_kernels(kernel_gen, dict(K=1000), kill_level_min=5) @@ -508,7 +508,7 @@ def test_interp_diff(ctx_factory): knl = lp.tag_inames(knl, dict(i="l.0", j="l.1")) print(knl) - #1/0 + # 1/0 kernel_gen = lp.generate_loop_schedules(knl) kernel_gen = lp.check_kernels(kernel_gen, dict(K=1000), kill_level_min=5) diff --git a/proto-tests/test_sem_tim.py b/proto-tests/test_sem_tim.py index a4af60a958e8a3f7ce4d02d2896ebfeb0229b9aa..4bf448fc1d2890e0811ad1b8f777ebf935b9d843 100644 --- a/proto-tests/test_sem_tim.py +++ b/proto-tests/test_sem_tim.py @@ -56,11 +56,11 @@ def test_laplacian(ctx_factory): ], name="semlap", assumptions="K>=1") - #print(lp.preprocess_kernel(knl, cse_ok=True)) - #1/0 + # print(lp.preprocess_kernel(knl, cse_ok=True)) + # 1/0 # - #print(knl) - #1/0 + # print(knl) + # 1/0 knl = lp.realize_cse(knl, "urf", np.float32, ["o1"]) knl = lp.realize_cse(knl, "usf", np.float32, ["o2"]) knl = lp.realize_cse(knl, "utf", np.float32, ["o3"]) @@ -76,7 +76,7 @@ def test_laplacian(ctx_factory): if 0: pass #seq_knl = lp.add_prefetch(knl, "G", ["gi", "m", "j", "k"], "G[gi,e,m,j,k]", default_tag="l.auto") # noqa - #seq_knl = lp.add_prefetch(seq_knl, "D", ["m", "j"], default_tag="l.auto") + # seq_knl = lp.add_prefetch(seq_knl, "D", ["m", "j"], default_tag="l.auto") #seq_knl = lp.add_prefetch(seq_knl, "u", ["i", "j", "k"], "u[*,i,j,k]", default_tag="l.auto") # noqa else: seq_knl = knl @@ -89,11 +89,11 @@ def test_laplacian(ctx_factory): default_tag="l.auto") #knl = lp.add_prefetch(knl, "u", ["i", "j", "k"], "u[*,i,j,k]", default_tag="l.auto") # noqa - #knl = lp.split_iname(knl, "e_inner", 4, inner_tag="ilp") + # knl = lp.split_iname(knl, "e_inner", 4, inner_tag="ilp") - #print(seq_knl) - #print(lp.preprocess_kernel(knl)) - #1/0 + # print(seq_knl) + # print(lp.preprocess_kernel(knl)) + # 1/0 knl = lp.tag_inames(knl, dict(i="l.0", j="l.1")) @@ -108,7 +108,7 @@ def test_laplacian(ctx_factory): parameters={"K": K}, print_seq_code=True) -#TW: start here +# TW: start here def test_laplacian_lmem(ctx_factory): dtype = np.float32 ctx = ctx_factory() @@ -170,15 +170,15 @@ def test_laplacian_lmem(ctx_factory): #knl = lp.add_prefetch(knl, "G", [2,3,4], default_tag="l.auto") # axis/argument indices on G # noqa #knl = lp.add_prefetch(knl, "G", ["i", "j", "m", "k"], default_tag="l.auto") # axis/argument indices on G # noqa - #print(knl) - #1/0 + # print(knl) + # 1/0 - #knl = lp.split_iname(knl, "e_inner", 4, inner_tag="ilp") + # knl = lp.split_iname(knl, "e_inner", 4, inner_tag="ilp") # knl = lp.join_dimensions(knl, ["i", "j"], "i_and_j") - #print(seq_knl) - #print(lp.preprocess_kernel(knl)) - #1/0 + # print(seq_knl) + # print(lp.preprocess_kernel(knl)) + # 1/0 # TW: turned this off since it generated: # ValueError: cannot tag 'i_and_j'--not known @@ -248,8 +248,8 @@ def test_laplacian_lmem_ilp(ctx_factory): knl = lp.add_prefetch(knl, "G", ["m", "i", "j", "k", "e_inner_inner"], default_tag="l.auto") # noqa knl = lp.add_prefetch(knl, "D", ["m", "j"], default_tag="l.auto") - #print(seq_knl) - #1/0 + # print(seq_knl) + # 1/0 knl = lp.tag_inames(knl, dict(i="l.0", j="l.1")) @@ -455,7 +455,7 @@ def test_advect_dealias(ctx_factory): knl = lp.tag_inames(knl, dict(i="l.0", j="l.1")) print(knl) - #1/0 + # 1/0 kernel_gen = lp.generate_loop_schedules(knl) kernel_gen = lp.check_kernels(kernel_gen, dict(K=1000), kill_level_min=5) @@ -516,7 +516,7 @@ def test_interp_diff(ctx_factory): knl = lp.tag_inames(knl, dict(i="l.0", j="l.1")) print(knl) - #1/0 + # 1/0 kernel_gen = lp.generate_loop_schedules(knl) kernel_gen = lp.check_kernels(kernel_gen, dict(K=1000), kill_level_min=5) diff --git a/pyproject.toml b/pyproject.toml index a3d19c2cfde6b148415451abb07b5e2c62c4a6bb..5f1403f34f61811e802a6156742049fcb172fb5b 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -32,7 +32,6 @@ extend-ignore = [ # FIXME "NPY002", # numpy rng "C408", # unnecssary dict() -> literal - "E265", # block comment should start with "F841", # local variable unused ] diff --git a/test/test_apps.py b/test/test_apps.py index 6b6b65bf4616f0cb234fbe6e314a2f1058b9c3cf..cbe3e76e8e04a886c6a40f0e995c8cb06095cff6 100644 --- a/test/test_apps.py +++ b/test/test_apps.py @@ -86,7 +86,7 @@ def test_convolution(ctx_factory): ref_knl = knl def variant_0(knl): - #knl = lp.split_iname(knl, "im_x", 16, inner_tag="l.0") + # knl = lp.split_iname(knl, "im_x", 16, inner_tag="l.0") knl = lp.prioritize_loops(knl, "iimg,im_x,im_y,ifeat,f_x,f_y") return knl @@ -108,8 +108,8 @@ def test_convolution(ctx_factory): return knl for variant in [ - #variant_0, - #variant_1, + # variant_0, + # variant_1, variant_2 ]: lp.auto_test_vs_ref(ref_knl, ctx, variant(knl), @@ -157,7 +157,7 @@ def test_convolution_with_nonzero_base(ctx_factory): f_w = 3 def variant_0(knl): - #knl = lp.split_iname(knl, "im_x", 16, inner_tag="l.0") + # knl = lp.split_iname(knl, "im_x", 16, inner_tag="l.0") knl = lp.prioritize_loops(knl, "iimg,im_x,im_y,ifeat,f_x,f_y") return knl @@ -357,7 +357,7 @@ def test_stencil(ctx_factory): return knl for variant in [ - #variant_1, + # variant_1, variant_2, ]: lp.auto_test_vs_ref(ref_knl, ctx, variant(knl), @@ -513,7 +513,7 @@ def test_fd_demo(): "result[i+1,j+1] = u[i + 1, j + 1]**2 + -1 + (-4)*u[i + 1, j + 1] \ + u[i + 1 + 1, j + 1] + u[i + 1 + -1, j + 1] \ + u[i + 1, j + 1 + 1] + u[i + 1, j + 1 + -1]") - #assumptions="n mod 16=0") + # assumptions="n mod 16=0") knl = lp.split_iname(knl, "i", 16, outer_tag="g.1", inner_tag="l.1") knl = lp.split_iname(knl, @@ -523,8 +523,8 @@ def test_fd_demo(): fetch_bounding_box=True, default_tag="l.auto") - #n = 1000 - #u = cl.clrandom.rand(queue, (n+2, n+2), dtype=np.float32) + # n = 1000 + # u = cl.clrandom.rand(queue, (n+2, n+2), dtype=np.float32) knl = lp.set_options(knl, write_code=True) knl = lp.add_and_infer_dtypes(knl, dict(u=np.float32)) @@ -604,7 +604,7 @@ def test_poisson_fem(ctx_factory): )) for variant in [ - #variant_1, + # variant_1, variant_2 ]: knl = variant(knl) diff --git a/test/test_dg.py b/test/test_dg.py index 318b4c3698f1209893b9b71daff4d2e053d5a351..125e5b9a62c8cc973a6fc3ae65a56f3128dff721 100644 --- a/test/test_dg.py +++ b/test/test_dg.py @@ -36,7 +36,7 @@ from loopy.version import LOOPY_USE_LANGUAGE_VERSION_2018_2 # noqa def test_dg_volume(ctx_factory): - #logging.basicConfig(level=logging.DEBUG) + # logging.basicConfig(level=logging.DEBUG) dtype = np.float32 dtype4 = cl.array.vec.float4 @@ -168,7 +168,7 @@ def test_dg_volume(ctx_factory): for variant in variants: lp.auto_test_vs_ref( seq_knl, ctx, variant(knl), parameters=parameters_dict, - #codegen_kwargs=dict(with_annotation=True) + # codegen_kwargs=dict(with_annotation=True) ) diff --git a/test/test_diff.py b/test/test_diff.py index 2498170736216fb4b7f7fa7d4e352dd49637c583..f06618b9f14ca1f9cfc9eb5d33e8157d08bd1f65 100644 --- a/test/test_diff.py +++ b/test/test_diff.py @@ -63,8 +63,8 @@ def test_diff(ctx_factory): knl = lp.fix_parameters(knl, n=50) from loopy.transform.diff import diff_kernel - #FIXME Is this the correct interface. Does it make sense to take the entire - #translation unit? + # FIXME Is this the correct interface. Does it make sense to take the entire + # translation unit? dknl, diff_map = diff_kernel(knl["diff"], "z", "x") dknl = knl.with_kernel(dknl) dknl = lp.remove_unused_arguments(dknl) diff --git a/test/test_domain.py b/test/test_domain.py index 8f29295066189f1c0ccf8ea11a910e98c2fa05de..9f0bb4158cac1be6798e13f67ed482980e3f1c64 100644 --- a/test/test_domain.py +++ b/test/test_domain.py @@ -301,8 +301,8 @@ def test_equality_constraints(ctx_factory): knl = lp.add_inames_to_insn(knl, "j_inner, j_outer", "id:set_b") - #print(knl) - #print(knl.domains[0].detect_equalities()) + # print(knl) + # print(knl.domains[0].detect_equalities()) lp.auto_test_vs_ref(seq_knl, ctx, knl, parameters=dict(n=n), print_ref_code=True) diff --git a/test/test_expression.py b/test/test_expression.py index 9af21cab814bcc855f862a1cfe60c8bcbf55e186..16b1e1bd37ba3041b20581db4ca0e67df06fab9c 100644 --- a/test/test_expression.py +++ b/test/test_expression.py @@ -291,7 +291,7 @@ def test_fuzz_expression_code_gen(ctx_factory, expr_type, random_seed, target_cl var_name = "expr%d" % i # print(expr) - #assert_parse_roundtrip(expr) + # assert_parse_roundtrip(expr) if expr_type in ["int", "int_nonneg"]: result_type_iinfo = np.iinfo(np.int32) @@ -451,8 +451,8 @@ def test_indexof_vec(ctx_factory): knl = lp.set_options(knl, write_code=True) (evt, (out,)) = knl(queue) - #out = out.get() - #assert np.array_equal(out.ravel(order="C"), np.arange(25)) + # out = out.get() + # assert np.array_equal(out.ravel(order="C"), np.arange(25)) def test_is_expression_equal(): diff --git a/test/test_linalg.py b/test/test_linalg.py index d31185ae0330479b2275359226ba0dfb0f314c62..ce43f19dad101e95b37ba638c833641c31fd597a 100644 --- a/test/test_linalg.py +++ b/test/test_linalg.py @@ -113,7 +113,7 @@ def test_axpy(ctx_factory): outer_tag="unr", inner_tag="l.0") return knl - #for variant in [ variant_gpu]: + # for variant in [ variant_gpu]: for variant in [variant_cpu, variant_gpu]: lp.auto_test_vs_ref(seq_knl, ctx, variant(knl), op_count=[np.dtype(dtype).itemsize*n*3/1e9], @@ -264,8 +264,8 @@ def test_funny_shape_matrix_mul(ctx_factory): outer_tag="g.1", inner_tag="l.0") knl = lp.split_iname(knl, "k", 32) - #knl = lp.add_prefetch(knl, "a", ["k_inner", "i_inner"], default_tag="l.auto") - #knl = lp.add_prefetch(knl, "b", ["j_inner", "k_inner"], default_tag="l.auto") + # knl = lp.add_prefetch(knl, "a", ["k_inner", "i_inner"], default_tag="l.auto") + # knl = lp.add_prefetch(knl, "b", ["j_inner", "k_inner"], default_tag="l.auto") knl = lp.extract_subst(knl, "a_acc", "a[i1,i2]", parameters="i1, i2") knl = lp.extract_subst(knl, "b_acc", "b[i1,i2]", parameters="i1, i2") knl = lp.precompute(knl, "a_acc", "k_inner,i_inner", @@ -285,7 +285,7 @@ def test_rank_one(ctx_factory): ctx = ctx_factory() order = "F" - #n = int(get_suitable_size(ctx)**(2.7/2)) + # n = int(get_suitable_size(ctx)**(2.7/2)) n = 16**3 knl = lp.make_kernel( @@ -442,7 +442,7 @@ def test_intel_matrix_mul(ctx_factory): knl = lp.split_iname(knl, "j", j_reg*j_chunks, outer_tag="g.1") knl = lp.split_iname(knl, "j_inner", j_reg, outer_tag="l.1", inner_tag="ilp") knl = lp.split_iname(knl, "k", 16) - #knl = lp.split_iname(knl, "k_inner", 8, outer_tag="unr") + # knl = lp.split_iname(knl, "k_inner", 8, outer_tag="unr") knl = lp.add_prefetch(knl, "a", ["i_inner_inner", "k_inner", "i_inner_outer"], fetch_outer_inames="i_outer, j_outer, k_outer", @@ -452,12 +452,12 @@ def test_intel_matrix_mul(ctx_factory): default_tag="l.auto") # FIXME: Grouped prefetch - #knl = lp.add_prefetch(knl, "a", ["k_inner", ("i_inner_inner", "i_inner_outer")], + # knl = lp.add_prefetch(knl, "a", ["k_inner", ("i_inner_inner", "i_inner_outer")], # default_tag="l.auto") - #knl = lp.add_prefetch(knl, "b", + # knl = lp.add_prefetch(knl, "b", # ["k_inner", ("j_inner_inner", "j_inner_outer"),], default_tag="l.auto") - #hints=["k_outer", "k_inner_outer", "k_inner_inner"] + # hints=["k_outer", "k_inner_outer", "k_inner_inner"] lp.auto_test_vs_ref(seq_knl, ctx, knl, op_count=[2*n**3/1e9], op_label=["GFlops"], @@ -506,9 +506,9 @@ def test_magma_fermi_matrix_mul(ctx_factory): knl = lp.split_iname(knl, "k", 16) knl = lp.split_iname(knl, "k_inner", 8, outer_tag="unr") # FIXME - #knl = lp.add_prefetch(knl, "a", ["k_inner", "i_inner_inner", "i_inner_outer"], + # knl = lp.add_prefetch(knl, "a", ["k_inner", "i_inner_inner", "i_inner_outer"], # default_tag="l.auto") - #knl = lp.add_prefetch(knl, "b", + # knl = lp.add_prefetch(knl, "b", # ["k_inner", ("j_inner_inner", "j_inner_outer"),], default_tag="l.auto") lp.auto_test_vs_ref(seq_knl, ctx, knl, diff --git a/test/test_loopy.py b/test/test_loopy.py index 5637c118746da4d1bbfb54edc693864fa7e047ce..00e29b91e738593b204c67d37b60d470c9dc5b8c 100644 --- a/test/test_loopy.py +++ b/test/test_loopy.py @@ -79,7 +79,7 @@ def test_globals_decl_once_with_multi_subprogram(ctx_factory): def test_complicated_subst(ctx_factory): - #ctx = ctx_factory() + # ctx = ctx_factory() knl = lp.make_kernel( "{[i]: 0<=i