triton-windows 3.3.0a0.post11__cp311-cp311-win_amd64.whl → 3.3.0a0.post13__cp311-cp311-win_amd64.whl

triton/__init__.py

@@ -1,18 +1,6 @@
 """isort:skip_file"""
 __version__ = '3.3.0'
 
-# Users may not know how to add cl and CUDA to PATH. Let's do it before loading anything
-import os
-if os.name == "nt":
-    from .windows_utils import find_cuda, find_msvc_winsdk
-    msvc_winsdk_inc_dirs, _ = find_msvc_winsdk()
-    if msvc_winsdk_inc_dirs:
-        cl_path = msvc_winsdk_inc_dirs[0].replace(r"\include", r"\bin\Hostx64\x64")
-        os.environ["PATH"] = cl_path + os.pathsep + os.environ["PATH"]
-    cuda_bin_path, _, _ = find_cuda()
-    if cuda_bin_path:
-        os.environ["PATH"] = cuda_bin_path + os.pathsep + os.environ["PATH"]
-
 # ---------------------------------------
 # Note: import order is significant here.
 

triton/backends/nvidia/compiler.py

@@ -32,9 +32,11 @@ def min_dot_size(target: GPUTarget):
 
 @functools.lru_cache()
 def _path_to_binary(binary: str):
-    binary += sysconfig.get_config_var("EXE")
     paths = [
         os.environ.get(f"TRITON_{binary.upper()}_PATH", ""),
+    ]
+    binary += sysconfig.get_config_var("EXE")
+    paths += [
         os.path.join(os.path.dirname(__file__), "bin", binary),
     ]
     if os.name == "nt":

triton/backends/nvidia/driver.py

@@ -23,14 +23,14 @@ libraries = ['cuda']
 
 @functools.lru_cache()
 def libcuda_dirs():
-    if os.name == "nt":
-        _, _, cuda_lib_dirs = find_cuda()
-        return cuda_lib_dirs
-
     env_libcuda_path = os.getenv("TRITON_LIBCUDA_PATH")
     if env_libcuda_path:
         return [env_libcuda_path]
 
+    if os.name == "nt":
+        _, _, cuda_lib_dirs = find_cuda()
+        return cuda_lib_dirs
+
     libs = subprocess.check_output(["/sbin/ldconfig", "-p"]).decode()
     # each line looks like the following:
     # libcuda.so.1 (libc6,x86-64) => /lib/x86_64-linux-gnu/libcuda.so.1

triton/compiler/compiler.py

@@ -65,10 +65,10 @@ class ASTSource:
         self.attrs = attrs or dict()
         if isinstance(self.signature, str):
             self.signature = {k: v.strip() for k, v in enumerate(self.signature.split(","))}
-        # else:
-        #     for k in self.signature.keys():
-        #         if not isinstance(k, str):
-        #             raise TypeError("Signature keys must be string")
+        else:
+            for k in self.signature.keys():
+                if not isinstance(k, str):
+                    raise TypeError("Signature keys must be string")
 
     def hash(self):
         sorted_sig = [v for k, v in sorted(self.signature.items())]

triton/runtime/build.py

@@ -7,8 +7,31 @@ if os.name == "nt":
     from triton.windows_utils import find_msvc_winsdk, find_python
 
 
+def get_cc():
+    cc = os.environ.get("CC")
+    if cc is None:
+        # Bundled TinyCC
+        cc = os.path.join(sysconfig.get_paths()["platlib"], "triton", "runtime", "tcc", "tcc.exe")
+        if not os.path.exists(cc):
+            cc = None
+    if cc is None:
+        cc = shutil.which("cl")
+    if cc is None:
+        cc = shutil.which("gcc")
+    if cc is None:
+        cc = shutil.which("clang")
+    if cc is None:
+        raise RuntimeError("Failed to find C compiler. Please specify via CC environment variable.")
+    return cc
+
+
+def is_msvc(cc):
+    cc = os.path.basename(cc).lower()
+    return cc == "cl" or cc == "cl.exe"
+
+
 def _cc_cmd(cc, src, out, include_dirs, library_dirs, libraries):
-    if cc.lower().endswith("cl") or cc.lower().endswith("cl.exe"):
+    if is_msvc(cc):
         out_base = os.path.splitext(out)[0]
         cc_cmd = [cc, src, "/nologo", "/O2", "/LD", "/wd4819"]
         cc_cmd += [f"/I{dir}" for dir in include_dirs if dir is not None]
@@ -32,15 +55,7 @@ def _build(name, src, srcdir, library_dirs, include_dirs, libraries):
     suffix = sysconfig.get_config_var('EXT_SUFFIX')
     so = os.path.join(srcdir, '{name}{suffix}'.format(name=name, suffix=suffix))
     # try to avoid setuptools if possible
-    cc = os.environ.get("CC")
-    if cc is None:
-        # TODO: support more things here.
-        cl = shutil.which("cl")
-        gcc = shutil.which("gcc")
-        clang = shutil.which("clang")
-        cc = cl if cl is not None else gcc if gcc is not None else clang
-        if cc is None:
-            raise RuntimeError("Failed to find C compiler. Please specify via CC environment variable.")
+    cc = get_cc()
     # This function was renamed and made public in Python 3.10
     if hasattr(sysconfig, 'get_default_scheme'):
         scheme = sysconfig.get_default_scheme()
@@ -55,7 +70,11 @@ def _build(name, src, srcdir, library_dirs, include_dirs, libraries):
     include_dirs = include_dirs + [srcdir, py_include_dir, *custom_backend_dirs]
     if os.name == "nt":
         library_dirs += find_python()
+    # Link against Python stable ABI
+    # libraries is modified in place
+    if "python3" not in libraries:
         libraries += ["python3"]
+    if is_msvc(cc):
         msvc_winsdk_inc_dirs, msvc_winsdk_lib_dirs = find_msvc_winsdk()
         include_dirs += msvc_winsdk_inc_dirs
         library_dirs += msvc_winsdk_lib_dirs

triton/runtime/tcc/include/_mingw.h

@@ -0,0 +1,170 @@
+/*
+ * _mingw.h
+ *
+ *  This file is for TinyCC and not part of the Mingw32 package.
+ *
+ *  THIS SOFTWARE IS NOT COPYRIGHTED
+ *
+ *  This source code is offered for use in the public domain. You may
+ *  use, modify or distribute it freely.
+ *
+ *  This code is distributed in the hope that it will be useful but
+ *  WITHOUT ANY WARRANTY. ALL WARRANTIES, EXPRESS OR IMPLIED ARE HEREBY
+ *  DISCLAIMED. This includes but is not limited to warranties of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ */
+
+#ifndef __MINGW_H
+#define __MINGW_H
+
+/* some winapi files define these before including _mingw.h --> */
+#undef __cdecl
+#undef _X86_
+#undef WIN32
+/* <-- */
+
+#include <stddef.h>
+#include <stdarg.h>
+
+#define __int8 char
+#define __int16 short
+#define __int32 int
+#define __int64 long long
+#define _HAVE_INT64
+
+#define __cdecl
+#define __declspec(x) __attribute__((x))
+#define __unaligned __attribute__((packed))
+#define __fastcall __attribute__((fastcall))
+
+#define __MSVCRT__ 1
+#undef _MSVCRT_
+#define __MINGW_IMPORT extern __declspec(dllimport)
+#define __MINGW_ATTRIB_NORETURN
+#define __MINGW_ATTRIB_CONST
+#define __MINGW_ATTRIB_DEPRECATED
+#define __MINGW_ATTRIB_MALLOC
+#define __MINGW_ATTRIB_PURE
+#define __MINGW_ATTRIB_NONNULL(arg)
+#define __MINGW_NOTHROW
+#define __GNUC_VA_LIST
+
+#define _CRTIMP extern
+#define __CRT_INLINE extern __inline__
+
+#define _CRT_ALIGN(x) __attribute__((aligned(x)))
+#define DECLSPEC_ALIGN(x) __attribute__((aligned(x)))
+#define _CRT_PACKING 8
+#define __CRT_UNALIGNED
+#define _CONST_RETURN
+
+#ifndef _TRUNCATE
+#define _TRUNCATE ((size_t)-1)
+#endif
+
+#define __CRT_STRINGIZE(_Value) #_Value
+#define _CRT_STRINGIZE(_Value) __CRT_STRINGIZE(_Value)
+#define __CRT_WIDE(_String) L ## _String
+#define _CRT_WIDE(_String) __CRT_WIDE(_String)
+
+#ifdef _WIN64
+#define __stdcall
+#define _AMD64_ 1
+#define __x86_64 1
+#define _M_X64 100 /* Visual Studio */
+#define _M_AMD64 100 /* Visual Studio */
+#define USE_MINGW_SETJMP_TWO_ARGS
+#define mingw_getsp tinyc_getbp
+#define __TRY__
+#else
+#define __stdcall __attribute__((__stdcall__))
+#define _X86_ 1
+#define _M_IX86 300 /* Visual Studio */
+#define WIN32 1
+#define _USE_32BIT_TIME_T
+#ifdef __arm__
+#define __TRY__
+#else
+#define __TRY__ void __try__(void**), *_sehrec[6]; __try__(_sehrec);
+#endif
+#endif
+
+/* in stddef.h */
+#define _SIZE_T_DEFINED
+#define _SSIZE_T_DEFINED
+#define _PTRDIFF_T_DEFINED
+#define _WCHAR_T_DEFINED
+#define _UINTPTR_T_DEFINED
+#define _INTPTR_T_DEFINED
+#define _INTEGRAL_MAX_BITS 64
+
+#ifndef _TIME32_T_DEFINED
+#define _TIME32_T_DEFINED
+typedef long __time32_t;
+#endif
+
+#ifndef _TIME64_T_DEFINED
+#define _TIME64_T_DEFINED
+typedef long long __time64_t;
+#endif
+
+#ifndef _TIME_T_DEFINED
+#define _TIME_T_DEFINED
+#ifdef _USE_32BIT_TIME_T
+typedef __time32_t time_t;
+#else
+typedef __time64_t time_t;
+#endif
+#endif
+
+#ifndef _WCTYPE_T_DEFINED
+#define _WCTYPE_T_DEFINED
+typedef wchar_t wctype_t;
+#endif
+
+#ifndef _WINT_T
+#define _WINT_T
+typedef __WINT_TYPE__ wint_t;
+#endif
+
+typedef int errno_t;
+#define _ERRCODE_DEFINED
+
+typedef struct threadlocaleinfostruct *pthreadlocinfo;
+typedef struct threadmbcinfostruct *pthreadmbcinfo;
+typedef struct localeinfo_struct _locale_tstruct,*_locale_t;
+
+/* for winapi */
+#define _ANONYMOUS_UNION
+#define _ANONYMOUS_STRUCT
+#define DECLSPEC_NORETURN
+#define DECLARE_STDCALL_P(type) __stdcall type
+#define NOSERVICE 1
+#define NOMCX 1
+#define NOIME 1
+#define __INTRIN_H_
+#ifndef DUMMYUNIONNAME
+#  define DUMMYUNIONNAME
+#  define DUMMYUNIONNAME1
+#  define DUMMYUNIONNAME2
+#  define DUMMYUNIONNAME3
+#  define DUMMYUNIONNAME4
+#  define DUMMYUNIONNAME5
+#endif
+#ifndef DUMMYSTRUCTNAME
+#  define DUMMYSTRUCTNAME
+#endif
+#ifndef WINVER
+# define WINVER 0x0502
+#endif
+#ifndef _WIN32_WINNT
+# define _WIN32_WINNT 0x502
+#endif
+
+#define __C89_NAMELESS
+#define __MINGW_EXTENSION
+#define WINAPI_FAMILY_PARTITION(X) 1
+#define MINGW_HAS_SECURE_API
+
+#endif /* __MINGW_H */

triton/runtime/tcc/include/assert.h

@@ -0,0 +1,57 @@
+/**
+ * This file has no copyright assigned and is placed in the Public Domain.
+ * This file is part of the w64 mingw-runtime package.
+ * No warranty is given; refer to the file DISCLAIMER within this package.
+ */
+#ifndef __ASSERT_H_
+#define __ASSERT_H_
+
+#include <_mingw.h>
+#ifdef __cplusplus
+#include <stdlib.h>
+#endif
+
+#ifdef NDEBUG
+#ifndef assert
+#define assert(_Expression) ((void)0)
+#endif
+#else
+
+#ifndef _CRT_TERMINATE_DEFINED
+#define _CRT_TERMINATE_DEFINED
+  void __cdecl __MINGW_NOTHROW exit(int _Code) __MINGW_ATTRIB_NORETURN;
+ _CRTIMP void __cdecl __MINGW_NOTHROW _exit(int _Code) __MINGW_ATTRIB_NORETURN;
+#if !defined __NO_ISOCEXT /* extern stub in static libmingwex.a */
+/* C99 function name */
+void __cdecl _Exit(int) __MINGW_ATTRIB_NORETURN;
+__CRT_INLINE __MINGW_ATTRIB_NORETURN void __cdecl _Exit(int status)
+{  _exit(status); }
+#endif
+
+#pragma push_macro("abort")
+#undef abort
+  void __cdecl __declspec(noreturn) abort(void);
+#pragma pop_macro("abort")
+
+#endif
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+extern void __cdecl _wassert(const wchar_t *_Message,const wchar_t *_File,unsigned _Line);
+extern void __cdecl _assert(const char *, const char *, unsigned);
+
+#ifdef __cplusplus
+}
+#endif
+
+#ifndef assert
+//#define assert(_Expression) (void)((!!(_Expression)) || (_wassert(_CRT_WIDE(#_Expression),_CRT_WIDE(__FILE__),__LINE__),0))
+#define assert(e) ((e) ? (void)0 : _assert(#e, __FILE__, __LINE__))
+#endif
+
+#endif
+
+#endif