Add MSVC support to setup.py for Windows builds

- Introduced platform-specific compiler and linker flag handling.
- Adjusted `nvcc` path resolution for Windows.
- Ensured compatibility with MSVC by adding appropriate flags and conditionals.

setup.py

@@ -45,18 +45,41 @@
     # Supported NVIDIA GPU architectures.
     SUPPORTED_ARCHS = {"8.0", "8.6", "8.9", "9.0", "12.0"}
 
-    # Compiler flags.
-    CXX_FLAGS = ["-g", "-O3", "-fopenmp", "-lgomp", "-std=c++17", "-DENABLE_BF16"]
-    NVCC_FLAGS = [
-        "-O3",
-        "-std=c++17",
-        "-U__CUDA_NO_HALF_OPERATORS__",
-        "-U__CUDA_NO_HALF_CONVERSIONS__",
-        "--use_fast_math",
-        "--threads=8",
-        "-Xptxas=-v",
-        "-diag-suppress=174",
-    ]
+    # Compiler flags (platform-specific)
+    is_msvc = os.name == "nt"
+    if is_msvc:
+        # MSVC-compatible flags
+        CXX_FLAGS = [
+            "/O2", "/std:c++17", "/openmp", "/EHsc", "/DNOMINMAX",
+            "/D_ENABLE_EXTENDED_ALIGNED_STORAGE", "/MP", "/permissive-", "/Zc:__cplusplus",
+            "/D_WIN32", "/DUSE_CUDA",
+        ]
+        NVCC_FLAGS = [
+            "-O3",
+            "-std=c++17",
+            "-U__CUDA_NO_HALF_OPERATORS__",
+            "-U__CUDA_NO_HALF_CONVERSIONS__",
+            "--use_fast_math",
+            "--threads=8",
+            "-Xptxas=-v",
+            "-diag-suppress=174",
+            # Host compiler flags for MSVC
+            "-Xcompiler", "/DWIN32",
+            "-Xcompiler", "/DUSE_CUDA",
+        ]
+    else:
+        # GCC/Clang flags (Linux/macOS)
+        CXX_FLAGS = ["-g", "-O3", "-fopenmp", "-lgomp", "-std=c++17", "-DENABLE_BF16"]
+        NVCC_FLAGS = [
+            "-O3",
+            "-std=c++17",
+            "-U__CUDA_NO_HALF_OPERATORS__",
+            "-U__CUDA_NO_HALF_CONVERSIONS__",
+            "--use_fast_math",
+            "--threads=8",
+            "-Xptxas=-v",
+            "-diag-suppress=174",
+        ]
 
     # Append flags from env if provided
     cxx_append = os.getenv("CXX_APPEND_FLAGS", "").strip()
@@ -66,9 +89,10 @@
     if nvcc_append:
         NVCC_FLAGS += nvcc_append.split()
 
-    ABI = 1 if torch._C._GLIBCXX_USE_CXX11_ABI else 0
-    CXX_FLAGS += [f"-D_GLIBCXX_USE_CXX11_ABI={ABI}"]
-    NVCC_FLAGS += [f"-D_GLIBCXX_USE_CXX11_ABI={ABI}"]
+    ABI = 1 if not is_msvc and getattr(torch._C, "_GLIBCXX_USE_CXX11_ABI", False) else 0
+    if not is_msvc:
+        CXX_FLAGS += [f"-D_GLIBCXX_USE_CXX11_ABI={ABI}"]
+        NVCC_FLAGS += [f"-D_GLIBCXX_USE_CXX11_ABI={ABI}"]
 
     if CUDA_HOME is None:
         raise RuntimeError(
@@ -79,8 +103,9 @@ def get_nvcc_cuda_version(cuda_dir: str) -> Version:
 
         Adapted from https://github.com/NVIDIA/apex/blob/8b7a1ff183741dd8f9b87e7bafd04cfde99cea28/setup.py
         """
-        nvcc_output = subprocess.check_output([cuda_dir + "/bin/nvcc", "-V"],
-                                              universal_newlines=True)
+        nvcc_name = "nvcc.exe" if os.name == "nt" else "nvcc"
+        nvcc_path = os.path.join(cuda_dir, "bin", nvcc_name)
+        nvcc_output = subprocess.check_output([nvcc_path, "-V"], universal_newlines=True)
         output = nvcc_output.split()
         release_idx = output.index("release") + 1
         nvcc_cuda_version = parse(output[release_idx].split(",")[0])
@@ -202,7 +227,7 @@ def get_nvcc_cuda_version(cuda_dir: str) -> Version:
                     "csrc/qattn/qk_int_sv_f8_cuda_sm90.cu",
                 ],
                 extra_compile_args={"cxx": CXX_FLAGS, "nvcc": NVCC_FLAGS},
-                extra_link_args=['-lcuda'],
+                extra_link_args=([] if is_msvc else ['-lcuda']),
             )
         )