build: MSVC portability for runtime and generated code

- psx_cyc.h / psx_cycles.c: _BitScanForward/_BitScanReverse shims for
  __builtin_ctz/__builtin_clz under _MSC_VER
- full_function_emitter.cpp / main_psx.cpp: emit a portable .CRT$XCU
  init-pointer instead of GCC-only __attribute__((constructor)) in
  generated dispatch/game code
- main.cpp: declare cross-language globals extern "C" so MSVC links
  them against their C definitions (no-op for GCC/Clang)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

docs/internal/upstream/msvc-portability-shaneomac.md

@@ -0,0 +1,41 @@
+# MSVC portability from Shaneomac PR #16
+
+## Source and credit
+
+- Source PR: [mstan/psxrecomp#16](https://github.com/mstan/psxrecomp/pull/16)
+- Source commit: [`df4cc5c70ab1e5d0fd2a8f817dd346905d4d0eeb`](https://github.com/mstan/psxrecomp/commit/df4cc5c70ab1e5d0fd2a8f817dd346905d4d0eeb)
+- Author preserved by cherry-pick: Martin Penkava
+  `<mpenkava1337@gmail.com>`.
+- The source commit's Claude Fable 5 co-author trailer is also preserved.
+
+## Adapted scope
+
+The source commit was applied as a focused portability change:
+
+- MSVC bit-scan intrinsics replace unavailable GCC/Clang builtins;
+- generated CPS startup markers use an MSVC `.CRT$XCU` initializer and retain
+  the constructor path for GCC/Clang; and
+- runtime globals defined by C translation units receive explicit C linkage in
+  the C++ entry point.
+
+`recompiler/tests/test_msvc_cps_codegen.py` generates a minimal game dispatch
+and checks that both compiler startup paths are emitted.
+
+## Validation in this branch
+
+- Visual Studio 2022/MSVC 19.41 built `psxrecomp-game`, `psxrecomp-bios`, and
+  `l2_structural_test`; the structural suite passed 44/44.
+- The MSVC-built game recompiler passed
+  `test_msvc_cps_codegen.py` on generated output.
+- A freshly generated BIOS runtime compiled all 63 C/C++ object steps under
+  MSVC, including the new intrinsic and C-linkage paths. The final BIOS-only
+  link remains blocked by the baseline's unrelated missing
+  `psx_game_address_in_text` and `psx_game_text_native_ok` definitions. No
+  change for that pre-existing game-symbol contract is included here.
+
+## Explicitly excluded
+
+No other PR #16 material was imported. In particular, this branch excludes the
+SmackDown-specific batch files, dynamic-code discovery fallback, SPU changes,
+primitive rejection, Vulkan diagnostics/optimization, and CRT shaders. Those
+items require independent review and provenance records.

recompiler/src/full_function_emitter.cpp

@@ -1656,9 +1656,17 @@ void FullFunctionEmitter::emit_dispatch(
         // RECURSION_BUG.md §25 — mark CPS mode at startup for runtime code that
         // must emit the CPS contract (the overlay sljit JIT, overlay_sljit.c).
         out += "\n/* CPS runtime-mode marker (overlay sljit JIT reads g_psx_cps_mode). */\n";
-        out += "__attribute__((constructor)) static void psx_cps_mark_bios(void) {\n";
+        out += "static void psx_cps_mark_bios(void) {\n";
         out += "    extern int g_psx_cps_mode; g_psx_cps_mode = 1;\n";
         out += "}\n";
+        // Run psx_cps_mark_bios before main(). __attribute__((constructor)) is
+        // GCC/Clang-only; MSVC uses a static initializer pointer in .CRT$XCU.
+        out += "#if defined(_MSC_VER)\n";
+        out += "#pragma section(\".CRT$XCU\", read)\n";
+        out += "__declspec(allocate(\".CRT$XCU\")) static void (*psx_cps_mark_bios_ctor)(void) = psx_cps_mark_bios;\n";
+        out += "#else\n";
+        out += "__attribute__((constructor)) static void psx_cps_mark_bios_ctor(void) { psx_cps_mark_bios(); }\n";
+        out += "#endif\n";
     }
 }
 

recompiler/src/main_psx.cpp

@@ -1086,9 +1086,17 @@ int main(int argc, char** argv) {
             // JIT (overlay_sljit.c) emits the CPS contract. Static ctor: no clash
             // with the BIOS dispatch's marker.
             ds << "\n/* CPS runtime-mode marker (overlay sljit JIT reads g_psx_cps_mode). */\n";
-            ds << "__attribute__((constructor)) static void psx_cps_mark_game(void) {\n";
+            ds << "static void psx_cps_mark_game(void) {\n";
             ds << "    extern int g_psx_cps_mode; g_psx_cps_mode = 1;\n";
             ds << "}\n";
+            // Run psx_cps_mark_game before main(). __attribute__((constructor)) is
+            // GCC/Clang-only; MSVC uses a static initializer pointer in .CRT$XCU.
+            ds << "#if defined(_MSC_VER)\n";
+            ds << "#pragma section(\".CRT$XCU\", read)\n";
+            ds << "__declspec(allocate(\".CRT$XCU\")) static void (*psx_cps_mark_game_ctor)(void) = psx_cps_mark_game;\n";
+            ds << "#else\n";
+            ds << "__attribute__((constructor)) static void psx_cps_mark_game_ctor(void) { psx_cps_mark_game(); }\n";
+            ds << "#endif\n";
         }
 
         std::ofstream dispatch_file(dispatch_filename);

recompiler/tests/test_msvc_cps_codegen.py

@@ -0,0 +1,98 @@
+#!/usr/bin/env python3
+"""Verify that CPS game output carries both MSVC and GCC/Clang startup hooks.
+
+Usage: python test_msvc_cps_codegen.py [--recompiler <psxrecomp-game.exe>]
+"""
+
+import argparse
+import os
+import struct
+import subprocess
+import sys
+import tempfile
+
+
+LOAD = 0x80010000
+
+
+def make_exe():
+    header = bytearray(2048)
+    header[0:8] = b"PS-X EXE"
+    struct.pack_into("<I", header, 0x10, LOAD)
+    struct.pack_into("<I", header, 0x18, LOAD)
+    body = struct.pack("<II", 0x03E00008, 0x00000000)  # jr ra; nop
+    struct.pack_into("<I", header, 0x1C, len(body))
+    return bytes(header) + body
+
+
+def generate_dispatch(recompiler):
+    with tempfile.TemporaryDirectory() as tmp:
+        exe = os.path.join(tmp, "tiny.psx")
+        seeds = os.path.join(tmp, "seeds.txt")
+        out = os.path.join(tmp, "out")
+        with open(exe, "wb") as stream:
+            stream.write(make_exe())
+        with open(seeds, "w", encoding="utf-8") as stream:
+            stream.write(f"0x{LOAD:08X}\n")
+        result = subprocess.run(
+            [recompiler, exe, "--seeds", seeds, "--out-dir", out],
+            capture_output=True,
+            text=True,
+        )
+        if result.returncode:
+            raise RuntimeError(result.stderr or result.stdout)
+        dispatches = [
+            os.path.join(out, name)
+            for name in os.listdir(out)
+            if name.endswith("_dispatch.c")
+        ]
+        if len(dispatches) != 1:
+            raise RuntimeError(
+                "expected one generated dispatch, found "
+                f"{[os.path.basename(path) for path in dispatches]}; "
+                f"generator output:\n{result.stdout}"
+            )
+        with open(dispatches[0], encoding="utf-8") as stream:
+            return stream.read()
+
+
+def main():
+    here = os.path.dirname(os.path.abspath(__file__))
+    default = os.path.normpath(
+        os.path.join(here, "..", "build", "psxrecomp-game.exe")
+    )
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--recompiler", default=default)
+    args = parser.parse_args()
+
+    if not os.path.isfile(args.recompiler):
+        print(f"FAIL: recompiler not found: {args.recompiler}", file=sys.stderr)
+        return 2
+
+    try:
+        generated = generate_dispatch(args.recompiler)
+    except (OSError, RuntimeError) as exc:
+        print(f"FAIL: generation failed: {exc}", file=sys.stderr)
+        return 1
+
+    required = (
+        "static void psx_cps_mark_game(void)",
+        "#if defined(_MSC_VER)",
+        '#pragma section(".CRT$XCU", read)',
+        '__declspec(allocate(".CRT$XCU")) static void '
+        "(*psx_cps_mark_game_ctor)(void) = psx_cps_mark_game;",
+        "__attribute__((constructor)) static void psx_cps_mark_game_ctor(void) "
+        "{ psx_cps_mark_game(); }",
+    )
+    missing = [text for text in required if text not in generated]
+    if missing:
+        for text in missing:
+            print(f"FAIL: generated dispatch is missing: {text}", file=sys.stderr)
+        return 1
+
+    print("PASS: generated CPS startup hooks cover MSVC and GCC/Clang.")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

runtime/include/psx_cyc.h

@@ -28,6 +28,9 @@
 #define PSX_CYC_H
 
 #include <stdint.h>
+#if defined(_MSC_VER)
+#include <intrin.h>       /* MSVC intrinsics: _BitScanForward (no __builtin_ctz) */
+#endif
 #include "cpu_state.h"   /* CPUState (guard-safe: cpu_state.h includes us last) */
 
 #ifdef __cplusplus
@@ -49,7 +52,13 @@ static inline void psx_cyc_base(CPUState* cpu) {
 static inline void psx_cyc_deps(CPUState* cpu, uint32_t reg_mask) {
     reg_mask &= 0xFFFFFFFEu;   /* never touch ReadAbsorb[0] */
     while (reg_mask) {
+#if defined(_MSC_VER)
+        unsigned long _psx_ctz_idx;
+        _BitScanForward(&_psx_ctz_idx, reg_mask);
+        unsigned n = (unsigned)_psx_ctz_idx;
+#else
         unsigned n = (unsigned)__builtin_ctz(reg_mask);
+#endif
         cpu->read_absorb[n] = 0u;
         reg_mask &= reg_mask - 1u;
     }

runtime/src/main.cpp

@@ -93,6 +93,21 @@
 
 extern "C" uint64_t gte_get_exec_count(void);
 
+/* Cross-language globals defined in C translation units. Declared extern "C" at
+ * file scope so MSVC gives them C linkage (matching the C definitions); without
+ * this MSVC name-mangles the C++ references and they fail to link. GCC/Clang do
+ * not mangle namespace-scope variables, so this is a no-op there. The existing
+ * block-scope `extern` redeclarations inside functions inherit this C linkage. */
+extern "C" {
+    extern uint64_t psx_cycle_count;
+    extern uint64_t s_frame_count;
+    extern uint32_t g_overlay_region_floor;
+    extern int      g_psx_cps_mode;
+    extern uint64_t g_slice_fired, g_slice_irq_taken, g_dirty_ram_insns_run;
+    extern uint32_t g_slice_exit_pc, g_slice_exit_reason, g_slice_exit_iter;
+    extern uint32_t g_slice_exit_dispatchable, g_slice_exit_dirty, g_slice_exit_in_text, g_slice_exit_want;
+}
+
 /* memory.c */
 extern "C" void     memory_init(const char* bios_path);
 extern "C" void     memory_set_sr_ptr(const uint32_t *p);

runtime/src/psx_cycles.c

@@ -3,6 +3,9 @@
 #include "psx_cycles.h"
 #include "cpu_state.h"
 #include <stdlib.h>
+#if defined(_MSC_VER)
+#include <intrin.h>       /* MSVC intrinsics: _BitScanReverse (no __builtin_clz) */
+#endif
 #include "cdrom.h"
 #include "dma.h"
 #include "interrupts.h"
@@ -451,7 +454,13 @@ static const uint8_t PSX_MULT_TAB24[24] = {
 
 static inline uint32_t psx_clz32(uint32_t v) {
     /* v is never 0 here (callers OR in 0x400). */
+#if defined(_MSC_VER)
+    unsigned long _idx;
+    _BitScanReverse(&_idx, v);   /* index of highest set bit */
+    return (uint32_t)(31u - _idx);
+#else
     return (uint32_t)__builtin_clz(v);
+#endif
 }
 
 uint32_t psx_mult_latency_s(uint32_t rs) {  /* MULT (signed): sign-fold magnitude */