Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[CPU] Use static compilation for kernels. #29

Merged
merged 1 commit into from
Jun 20, 2024
Merged

[CPU] Use static compilation for kernels. #29

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions python/triton/runtime/build.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -48,6 +48,8 @@ def _build(name, src, srcdir, library_dirs, include_dirs, libraries):
# CPU backend uses C++ (driver.cpp). Some old version compilers need a specific C++17 flag.
if src.endswith(".cpp") or src.endswith(".cc"):
cc_cmd += ["-std=c++17", "-fopenmp"]
if src.endswith(".s"):
cc_cmd += ["-gdwarf-5"]
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think we can follow it up later on a different platform and compiler (e.g., Mac/clang), but this makes the problem much simpler: just compiling .s. Looks good.

ret = subprocess.check_call(cc_cmd)
if ret == 0:
return so
Expand Down
16 changes: 4 additions & 12 deletions third_party/cpu/backend/compiler.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -42,7 +42,7 @@ def supports_target(target: GPUTarget):

def __init__(self, target: tuple) -> None:
super().__init__(target)
self.binary_ext = "bc"
self.binary_ext = "asm"

def parse_options(self, opts) -> Any:
args = {k: opts[k] for k in CPUOptions.__dataclass_fields__.keys() if k in opts}
Expand Down Expand Up @@ -138,22 +138,14 @@ def make_llir(src, metadata, options):
return ret

@staticmethod
def make_bc(src, metadata, options):
if os.environ.get("TRITON_CPU_ASM_DUMP", "0") == "1":
from triton.runtime.cache import get_cache_manager

asm = llvm.translate_to_host_asm(src, options.enable_fp_fusion)
fn_cache_manager = get_cache_manager(metadata['hash'])
fn_cache_manager.put(asm, f"{metadata['name']}.asm")

ret = llvm.translate_to_bc(src)
return ret
def make_asm(src, metadata, options):
return llvm.translate_to_host_asm(src, options.enable_fp_fusion)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Did you measure the compilation time? It tends to be very long in some cases. For sure, we will address it later.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It should be the same as we had for JIT. It can be quite long when we produce huge basic blocks (like tens or even hundreds of thousands of operations) because some optimizations can be O(basic_block_size^2). We should generate better code for those cases and use smaller block sizes for the CPU.


def add_stages(self, stages, options):
stages["ttir"] = lambda src, metadata: self.make_ttir(src, metadata, options)
stages["ttcir"] = lambda src, metadata: self.make_ttcir(src, metadata, options)
stages["llir"] = lambda src, metadata: self.make_llir(src, metadata, options)
stages["bc"] = lambda src, metadata: self.make_bc(src, metadata, options)
stages["asm"] = lambda src, metadata: self.make_asm(src, metadata, options)

@functools.lru_cache()
def hash(self):
Expand Down
224 changes: 0 additions & 224 deletions third_party/cpu/backend/driver.cpp
View file
Open in desktop

This file was deleted.

94 changes: 23 additions & 71 deletions third_party/cpu/backend/driver.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,74 +8,9 @@
from triton.backends.compiler import GPUTarget

dirname = os.getenv("TRITON_SYS_PATH", default="/usr/local")
llvm_root = os.getenv("LLVM_PATH", default="~/.triton/llvm")
llvm_root = os.path.expanduser(llvm_root)
llvm_dirs = os.listdir(llvm_root)
if len(llvm_dirs) == 1:
llvm_root = os.path.join(llvm_root, llvm_dirs[0])
include_dir = [
os.path.join(dirname, "include"),
os.path.join(llvm_root, "include"),
]
library_dir = [os.path.join(dirname, "lib"), os.path.join(llvm_root, "lib")]
libraries = [
"LLVMOrcJIT",
"LLVMPasses",
"LLVMX86CodeGen",
"LLVMX86AsmParser",
"LLVMX86Desc",
"LLVMX86Info",
"LLVMGlobalISel",
"LLVMSelectionDAG",
"LLVMHipStdPar",
"LLVMCoroutines",
"LLVMipo",
"LLVMFrontendOpenMP",
"LLVMInstrumentation",
"LLVMAsmPrinter",
"LLVMCodeGen",
"LLVMObjCARCOpts",
"LLVMLinker",
"LLVMVectorize",
"LLVMScalarOpts",
"LLVMInstCombine",
"LLVMFrontendOffloading",
"LLVMExecutionEngine",
"LLVMAggressiveInstCombine",
"LLVMTransformUtils",
"LLVMTarget",
"LLVMRuntimeDyld",
"LLVMJITLink",
"LLVMIRPrinter",
"LLVMBitWriter",
"LLVMAnalysis",
"LLVMProfileData",
"LLVMSymbolize",
"LLVMDebugInfoDWARF",
"LLVMObject",
"LLVMTextAPI",
"LLVMMCParser",
"LLVMMCDisassembler",
"LLVMMC",
"LLVMIRReader",
"LLVMCFGuard",
"LLVMBitReader",
"LLVMAsmParser",
"LLVMCore",
"LLVMBinaryFormat",
"LLVMOrcTargetProcess",
"LLVMTargetParser",
"LLVMRemarks",
"LLVMOrcShared",
"LLVMOption",
"LLVMDebugInfoCodeView",
"LLVMCodeGenTypes",
"LLVMBitstreamReader",
"LLVMSupport",
"LLVMDemangle",
"stdc++",
"z",
]
include_dir = [os.path.join(dirname, "include")]
library_dir = [os.path.join(dirname, "lib")]
libraries = ["stdc++"]


def compile_module_from_src(src, name):
Expand Down Expand Up @@ -110,9 +45,26 @@ def __new__(cls):
return cls.instance

def __init__(self):
dirname = os.path.dirname(os.path.realpath(__file__))
mod = compile_module_from_src(Path(os.path.join(dirname, "driver.cpp")).read_text(), "cpu_utils")
self.load_binary = mod.load_binary
pass

def load_binary(self, name, src, shared_mem, device):
# src actually holds asm text, compile to a shared library.
key = hashlib.md5(src).hexdigest()
cache = get_cache_manager(key)
cache_path = cache.get_file(f"{name}.so")
if cache_path is None:
with tempfile.TemporaryDirectory() as tmpdir:
asm_path = os.path.join(tmpdir, "kernel.s")
Path(asm_path).write_bytes(src)
Path("kernel.s").write_bytes(src)
so = _build(name, asm_path, tmpdir, library_dir, include_dir, ["gcc", "m"])
with open(so, "rb") as f:
cache_path = cache.put(f.read(), f"{name}.so", binary=True)
import ctypes
lib = ctypes.cdll.LoadLibrary(cache_path)
fn_ptr = getattr(lib, name)
fn_ptr_as_void_p = ctypes.cast(fn_ptr, ctypes.c_void_p).value
return (fn_ptr, fn_ptr_as_void_p, 0, 0)

def get_device_properties(self, *args):
return {"max_shared_mem": 0}
Expand Down
Loading