gaspersic/triton
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Improve

Browse Source
This commit is contained in:
Jokeren
2022-10-23 20:32:25 -07:00
parent db3aa1d1fb
commit baab18e1d1
2 changed files with 12 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files

29
python/triton/compiler.py
View File

@@ -959,7 +959,7 @@ def generate_launcher(identifier, constants, signature):
"int64_t": "L", "int64_t": "L",
}[ty] }[ty]
format = "iiiiiKKOOO" + ''.join([format_of(_extracted_type(ty)) for ty in signature.values()]) format = "iiiiiKKOOO" + ''.join([format_of(_extracted_type(ty)) for ty in signature.values()]) + "O"
# generate glue code # generate glue code
src = f""" src = f"""
@@ -1022,34 +1022,22 @@ static PyObject* launch(PyObject* self, PyObject* args) {{
PyObject *launch_enter_hook = NULL; PyObject *launch_enter_hook = NULL;
PyObject *launch_exit_hook = NULL; PyObject *launch_exit_hook = NULL;
PyObject *compiled_kernel = NULL; PyObject *compiled_kernel = NULL;
PyObject *hook_ret = NULL; PyObject *constants = NULL;
{' '.join([f"{_extracted_type(ty)} _arg{i}; " for i, ty in signature.items()])} {' '.join([f"{_extracted_type(ty)} _arg{i}; " for i, ty in signature.items()])}
if(!PyArg_ParseTuple(args, \"{format}\", &gridX, &gridY, &gridZ, &num_warps, &shared_memory, &_stream, &_function, &launch_enter_hook, &launch_exit_hook, &compiled_kernel, {', '.join(f"&_arg{i}" for i, ty in signature.items())})) {{ if(!PyArg_ParseTuple(args, \"{format}\", &gridX, &gridY, &gridZ, &num_warps, &shared_memory, &_stream, &_function, &launch_enter_hook, &launch_exit_hook, &compiled_kernel, {', '.join(f"&_arg{i}" for i, ty in signature.items())}, &constants)) {{
return NULL; return NULL;
}} }}
if (launch_enter_hook != Py_None) {{ if (launch_enter_hook != Py_None) {{
PyObject *new_args = PyTuple_Pack(1, compiled_kernel); PyObject_CallObject(launch_enter_hook, args);
hook_ret = PyObject_CallObject(launch_enter_hook, new_args);
Py_DECREF(new_args);
}} }}
_launch(gridX, gridY, gridZ, num_warps, shared_memory, (CUstream)_stream, (CUfunction)_function, {', '.join(f"getPointer(_arg{i},{i})" if ty[0]=="*" else f"_arg{i}"for i, ty in signature.items())}); _launch(gridX, gridY, gridZ, num_warps, shared_memory, (CUstream)_stream, (CUfunction)_function, {', '.join(f"getPointer(_arg{i},{i})" if ty[0]=="*" else f"_arg{i}"for i, ty in signature.items())});
if (launch_exit_hook != Py_None) {{ if (launch_exit_hook != Py_None) {{
PyObject *new_args = NULL; PyObject_CallObject(launch_exit_hook, args);
if (hook_ret) {{
new_args = PyTuple_Pack(2, compiled_kernel, hook_ret);
}} else {{
new_args = PyTuple_Pack(1, compiled_kernel);
}}
hook_ret = PyObject_CallObject(launch_exit_hook, new_args);
Py_DECREF(new_args);
}} }}
if (hook_ret) {{
Py_DECREF(hook_ret);
}}
if(PyErr_Occurred()) {{ if(PyErr_Occurred()) {{
return NULL; return NULL;
}} }}
@@ -1214,7 +1202,7 @@ def make_fn_cache_key(fn_hash, signature, configs, constants, num_warps, num_sta
return key return key
def compile(fn, signature: str, device: int = -1, constants=dict(), num_warps: int = 4, def compile(fn, signature: str, device: int = -1, constants=dict(), constexpr=tuple(), num_warps: int = 4,
num_stages: int = 3, extern_libs=None, configs=None, cc=0, warm_cache_only=False): num_stages: int = 3, extern_libs=None, configs=None, cc=0, warm_cache_only=False):
# we get the kernel, i.e. the first function generated in the module # we get the kernel, i.e. the first function generated in the module
assert len(configs) == 1 assert len(configs) == 1
@@ -1250,7 +1238,7 @@ def compile(fn, signature: str, device: int = -1, constants=dict(), num_warps: i
not fn_cache_manager.has_file(llir_name): not fn_cache_manager.has_file(llir_name):
asm, shared, kernel_name = _compile(fn, signature, device, constants, configs[0], num_warps, num_stages, asm, shared, kernel_name = _compile(fn, signature, device, constants, configs[0], num_warps, num_stages,
extern_libs, "cubin", cc) extern_libs, "cubin", cc)
metadata = {"name": kernel_name, "shared": shared, "num_warps": num_warps, "num_stages": num_stages} metadata = {"name": kernel_name, "shared": shared, "num_warps": num_warps, "num_stages": num_stages, "constexpr": constexpr}
fn_cache_manager.put(asm["cubin"], cubin_name) fn_cache_manager.put(asm["cubin"], cubin_name)
fn_cache_manager.put(asm["ptx"], ptx_name, binary=False) fn_cache_manager.put(asm["ptx"], ptx_name, binary=False)
fn_cache_manager.put(asm["ttir"], ttir_name, binary=False) fn_cache_manager.put(asm["ttir"], ttir_name, binary=False)
@@ -1282,6 +1270,7 @@ class CompiledKernel:
self.shared = metadata["shared"] self.shared = metadata["shared"]
self.num_warps = metadata["num_warps"] self.num_warps = metadata["num_warps"]
self.num_stages = metadata["num_stages"] self.num_stages = metadata["num_stages"]
self.constexpr = metadata["constexpr"]
# initialize asm dict # initialize asm dict
self.asm = dict() self.asm = dict()
with open(os.path.join(cache_dir, f"{fn_name}.cubin"), "rb") as f: with open(os.path.join(cache_dir, f"{fn_name}.cubin"), "rb") as f:
@@ -1305,7 +1294,7 @@ class CompiledKernel:
if stream is None: if stream is None:
stream = torch.cuda.current_stream().cuda_stream stream = torch.cuda.current_stream().cuda_stream
self.c_wrapper(grid[0], grid[1], grid[2], self.num_warps, self.shared, stream, self.cu_function, self.c_wrapper(grid[0], grid[1], grid[2], self.num_warps, self.shared, stream, self.cu_function,
CompiledKernel.launch_enter_hook, CompiledKernel.launch_exit_hook, self, *args) CompiledKernel.launch_enter_hook, CompiledKernel.launch_exit_hook, self, *args, self.constexpr)
return runner return runner
def get_sass(self, fun=None): def get_sass(self, fun=None):

6
python/triton/runtime/jit.py
View File

@@ -254,7 +254,7 @@ def {self.fn.__name__}({', '.join(self.arg_names)}, grid, num_warps=4, num_stage
try: try:
bin = cache[device][key] bin = cache[device][key]
if not warmup: if not warmup:
bin.c_wrapper(grid_0, grid_1, grid_2, bin.num_warps, bin.shared, stream, bin.cu_function, triton.compiler.CompiledKernel.launch_enter_hook, triton.compiler.CompiledKernel.launch_exit_hook, bin, {args}) bin.c_wrapper(grid_0, grid_1, grid_2, bin.num_warps, bin.shared, stream, bin.cu_function, triton.compiler.CompiledKernel.launch_enter_hook, triton.compiler.CompiledKernel.launch_exit_hook, bin, {args}, constexpr_key)
return bin return bin
# kernel not cached -- compile # kernel not cached -- compile
except KeyError: except KeyError:
@@ -272,9 +272,9 @@ def {self.fn.__name__}({', '.join(self.arg_names)}, grid, num_warps=4, num_stage
if callable(arg): if callable(arg):
raise TypeError(f"Callable constexpr at index {i} is not supported") raise TypeError(f"Callable constexpr at index {i} is not supported")
if not self._call_hook(key, signature, device, constants, num_warps, num_stages, extern_libs, configs): if not self._call_hook(key, signature, device, constants, num_warps, num_stages, extern_libs, configs):
bin = triton.compile(self, signature, device, constants, num_warps, num_stages, extern_libs=extern_libs, configs=configs) bin = triton.compile(self, signature, device, constants, constexpr_key, num_warps, num_stages, extern_libs=extern_libs, configs=configs)
if not warmup: if not warmup:
bin.c_wrapper(grid_0, grid_1, grid_2, bin.num_warps, bin.shared, stream, bin.cu_function, triton.compiler.CompiledKernel.launch_enter_hook, triton.compiler.CompiledKernel.launch_exit_hook, bin, *args) bin.c_wrapper(grid_0, grid_1, grid_2, bin.num_warps, bin.shared, stream, bin.cu_function, triton.compiler.CompiledKernel.launch_enter_hook, triton.compiler.CompiledKernel.launch_exit_hook, bin, *args, constexpr_key)
self.cache[device][key] = bin self.cache[device][key] = bin
return bin return bin
return None return None