diff --git a/recompiler/src/pgxp_hook_emitter.cpp b/recompiler/src/pgxp_hook_emitter.cpp
index 1455983c..5c54b8f5 100644
--- a/recompiler/src/pgxp_hook_emitter.cpp
+++ b/recompiler/src/pgxp_hook_emitter.cpp
@@ -30,6 +30,13 @@ inline int16_t get_imm16(uint32_t i) { return (int16_t)(i & 0xFFFF); }
  * return early above translate_instruction's main dispatch and are likewise
  * unhooked (they are cull compares, not vertex moves; validation covers). */
 void append_pgxp_hooks(uint32_t instr, std::string& code) {
+    /* A translation ending in a preprocessor directive (the block-cycles
+     * muldiv wrappers end in `#endif`) would swallow an appended hook: the
+     * preprocessor discards tokens after #endif (the "extra tokens at end of
+     * #endif" warning class), so the MULT/DIV/MFLO/MFHI hooks silently never
+     * ran. Give the hook its own line. */
+    if (code.size() >= 6 && code.compare(code.size() - 6, 6, "#endif") == 0)
+        code += "\n    ";
     const uint32_t opcode = (instr >> 26) & 0x3F;
     const uint32_t rs = get_rs(instr);
     const uint32_t rt = get_rt(instr);
diff --git a/runtime/include/overlay_dispatch_preamble.c.inc b/runtime/include/overlay_dispatch_preamble.c.inc
index c2acb8e7..eee589b7 100644
--- a/runtime/include/overlay_dispatch_preamble.c.inc
+++ b/runtime/include/overlay_dispatch_preamble.c.inc
@@ -211,6 +211,23 @@ int psx_ws_cull_vxrange(uint32_t x, uint32_t imm) {
     return (((x + (uint32_t)margin) & 0xFFFFu) <
             (bound + 2u * (uint32_t)margin)) ? 1 : 0;
 }
+/* [[widescreen.cull.widen]] SLT-family helpers. The emitter writes these at
+ * explicitly configured cull sites, INCLUDING in overlay-resident code, so
+ * the DLL must define them or the shard fails to link (undefined reference ->
+ * no native shard -> that mod-patched page runs interpreted forever). Same
+ * contract as the funnel helpers above: self-contained on psx_ws_x_margin(),
+ * identity at margin 0 (4:3), MUST stay byte-for-byte identical to the gpu.c
+ * implementations. */
+int psx_ws_cull_slti_lower(uint32_t sx, uint32_t imm) {
+    int32_t bound = (int32_t)(int16_t)(uint16_t)imm;
+    return ((int32_t)sx < bound - psx_ws_x_margin()) ? 1 : 0;
+}
+int psx_ws_cull_slt_widen(uint32_t rs, uint32_t rt, int bound_is_rt) {
+    const int32_t m = psx_ws_x_margin();
+    const int32_t a = (int32_t)rs, b = (int32_t)rt;
+    return bound_is_rt ? ((a < b + m) ? 1 : 0)
+                       : ((a + m < b) ? 1 : 0);
+}
 void psx_ws_sprite_tag(CPUState *cpu) {
     if (g_cbs.ws_sprite_tag) g_cbs.ws_sprite_tag(cpu);
 }