Replace FEATURE_EH_FUNCLETS in JIT with runtime switch (#99191)

* Replace FEATURE_EH_FUNCLETS/FEATURE_EH_CALLFINALLY_THUNKS in JIT with runtime switch

* Cache Native AOT ABI check to see if TP improves

---------

Co-authored-by: Bruce Forstall <[email protected]>

Author: filipnavara

docs/design/coreclr/botr/clr-abi.md

@@ -177,11 +177,13 @@ This section describes the conventions the JIT needs to follow when generating c
 
 ## Funclets
 
-For all platforms except Windows/x86, all managed EH handlers (finally, fault, filter, filter-handler, and catch) are extracted into their own 'funclets'. To the OS they are treated just like first class functions (separate PDATA and XDATA (`RUNTIME_FUNCTION` entry), etc.). The CLR currently treats them just like part of the parent function in many ways. The main function and all funclets must be allocated in a single code allocation (see hot cold splitting). They 'share' GC info. Only the main function prolog can be hot patched.
+For all platforms except Windows/x86 on CoreCLR, all managed EH handlers (finally, fault, filter, filter-handler, and catch) are extracted into their own 'funclets'. To the OS they are treated just like first class functions (separate PDATA and XDATA (`RUNTIME_FUNCTION` entry), etc.). The CLR currently treats them just like part of the parent function in many ways. The main function and all funclets must be allocated in a single code allocation (see hot cold splitting). They 'share' GC info. Only the main function prolog can be hot patched.
 
 The only way to enter a handler funclet is via a call. In the case of an exception, the call is from the VM's EH subsystem as part of exception dispatch/unwind. In the non-exceptional case, this is called local unwind or a non-local exit. In C# this is accomplished by simply falling-through/out of a try body or an explicit goto. In IL this is always accomplished via a LEAVE opcode, within a try body, targeting an IL offset outside the try body. In such cases the call is from the JITed code of the parent function.
 
-For Windows/x86, all handlers are generated within the method body, typically in lexical order. A nested try/catch is generated completely within the EH region in which it is nested. These handlers are essentially "in-line funclets", but they do not look like normal functions: they do not have a normal prolog or epilog, although they do have special entry/exit and register conventions. Also, nested handlers are not un-nested as for funclets: the code for a nested handler is generated within the handler in which it is nested.
+For Windows/x86 on CoreCLR, all handlers are generated within the method body, typically in lexical order. A nested try/catch is generated completely within the EH region in which it is nested. These handlers are essentially "in-line funclets", but they do not look like normal functions: they do not have a normal prolog or epilog, although they do have special entry/exit and register conventions. Also, nested handlers are not un-nested as for funclets: the code for a nested handler is generated within the handler in which it is nested.
+
+For Windows/x86 on NativeAOT and Linux/x86, funclets are used just like on other platforms.
 
 ## Cloned finallys
 

src/coreclr/clrdefinitions.cmake

@@ -288,8 +288,4 @@ function(set_target_definitions_to_custom_os_and_arch)
   if (TARGETDETAILS_ARCH STREQUAL "armel")
     target_compile_definitions(${TARGETDETAILS_TARGET} PRIVATE ARM_SOFTFP)
   endif()
-
-  if (NOT (TARGETDETAILS_ARCH STREQUAL "x86") OR (TARGETDETAILS_OS MATCHES "^unix") OR (TARGETDETAILS_OS MATCHES "win_aot"))
-    target_compile_definitions(${TARGETDETAILS_TARGET} PRIVATE FEATURE_EH_FUNCLETS)
-  endif (NOT (TARGETDETAILS_ARCH STREQUAL "x86") OR (TARGETDETAILS_OS MATCHES "^unix") OR (TARGETDETAILS_OS MATCHES "win_aot"))
 endfunction()

src/coreclr/crosscomponents.cmake

@@ -25,13 +25,6 @@ if (CLR_CMAKE_HOST_OS STREQUAL CLR_CMAKE_TARGET_OS OR CLR_CMAKE_TARGET_IOS OR CL
                 DESTINATIONS .
                 COMPONENT crosscomponents
             )
-            if (CLR_CMAKE_TARGET_ARCH_I386)
-                install_clr (TARGETS
-                    clrjit_win_aot_${ARCH_TARGET_NAME}_${ARCH_HOST_NAME}
-                    DESTINATIONS .
-                    COMPONENT crosscomponents
-                )
-            endif()
         endif()
     endif()
 endif()

src/coreclr/inc/clrnt.h

@@ -201,7 +201,6 @@ typedef struct _DISPATCHER_CONTEXT {
 #define RUNTIME_FUNCTION__BeginAddress(prf)             (prf)->BeginAddress
 #define RUNTIME_FUNCTION__SetBeginAddress(prf,addr)     ((prf)->BeginAddress = (addr))
 
-#ifdef FEATURE_EH_FUNCLETS
 #include "win64unwind.h"
 #include "daccess.h"
 
@@ -235,7 +234,6 @@ RtlVirtualUnwind (
     __inout_opt PT_KNONVOLATILE_CONTEXT_POINTERS ContextPointers
     );
 #endif // HOST_X86
-#endif // FEATURE_EH_FUNCLETS
 
 #endif // TARGET_X86
 

src/coreclr/inc/gcinfo.h

@@ -25,7 +25,7 @@ const unsigned        OFFSET_MASK  = 0x3;  // mask to access the low 2 bits
 //
 const unsigned  byref_OFFSET_FLAG  = 0x1;  // the offset is an interior ptr
 const unsigned pinned_OFFSET_FLAG  = 0x2;  // the offset is a pinned ptr
-#if defined(TARGET_X86) && !defined(FEATURE_EH_FUNCLETS)
+#if defined(TARGET_X86)
 // JIT32_ENCODER has additional restriction on x86 without funclets: 
 // - for untracked locals the flags allowed are "pinned" and "byref"
 // - for tracked locals the flags allowed are "this" and "byref"

src/coreclr/jit/CMakeLists.txt

@@ -23,8 +23,6 @@ function(create_standalone_jit)
 
   if(TARGETDETAILS_OS STREQUAL "unix_osx" OR TARGETDETAILS_OS STREQUAL "unix_anyos")
     set(JIT_ARCH_LINK_LIBRARIES gcinfo_unix_${TARGETDETAILS_ARCH})
-  elseif(TARGETDETAILS_OS STREQUAL "win_aot")
-    set(JIT_ARCH_LINK_LIBRARIES gcinfo_win_${TARGETDETAILS_ARCH})
   else()
     set(JIT_ARCH_LINK_LIBRARIES gcinfo_${TARGETDETAILS_OS}_${TARGETDETAILS_ARCH})
   endif()
@@ -658,7 +656,6 @@ else()
   create_standalone_jit(TARGET clrjit_universal_arm_${ARCH_HOST_NAME} OS universal ARCH arm DESTINATIONS .)
   target_compile_definitions(clrjit_universal_arm_${ARCH_HOST_NAME} PRIVATE ARM_SOFTFP CONFIGURABLE_ARM_ABI)
   create_standalone_jit(TARGET clrjit_win_x86_${ARCH_HOST_NAME} OS win ARCH x86 DESTINATIONS .)
-  create_standalone_jit(TARGET clrjit_win_aot_x86_${ARCH_HOST_NAME} OS win_aot ARCH x86 DESTINATIONS .)
 endif (CLR_CMAKE_TARGET_ARCH_RISCV64)
 
 if (CLR_CMAKE_TARGET_ARCH_I386 AND CLR_CMAKE_TARGET_UNIX)

src/coreclr/jit/block.cpp

@@ -1812,9 +1812,7 @@ bool BasicBlock::hasEHBoundaryIn() const
     bool returnVal = (bbCatchTyp != BBCT_NONE);
     if (!returnVal)
     {
-#if FEATURE_EH_FUNCLETS
         assert(!HasFlag(BBF_FUNCLET_BEG));
-#endif // FEATURE_EH_FUNCLETS
     }
     return returnVal;
 }
@@ -1833,16 +1831,7 @@ bool BasicBlock::hasEHBoundaryIn() const
 //
 bool BasicBlock::hasEHBoundaryOut() const
 {
-    bool returnVal = KindIs(BBJ_EHFILTERRET, BBJ_EHFINALLYRET, BBJ_EHFAULTRET);
-
-#if FEATURE_EH_FUNCLETS
-    if (bbKind == BBJ_EHCATCHRET)
-    {
-        returnVal = true;
-    }
-#endif // FEATURE_EH_FUNCLETS
-
-    return returnVal;
+    return KindIs(BBJ_EHFILTERRET, BBJ_EHFINALLYRET, BBJ_EHFAULTRET, BBJ_EHCATCHRET);
 }
 
 //------------------------------------------------------------------------

src/coreclr/jit/block.h

@@ -66,7 +66,7 @@ enum BBKinds : BYTE
     BBJ_EHFINALLYRET,// block ends with 'endfinally' (for finally)
     BBJ_EHFAULTRET,  // block ends with 'endfinally' (IL alias for 'endfault') (for fault)
     BBJ_EHFILTERRET, // block ends with 'endfilter'
-    BBJ_EHCATCHRET,  // block ends with a leave out of a catch (only #if defined(FEATURE_EH_FUNCLETS))
+    BBJ_EHCATCHRET,  // block ends with a leave out of a catch
     BBJ_THROW,       // block ends with 'throw'
     BBJ_RETURN,      // block ends with 'ret'
     BBJ_ALWAYS,      // block always jumps to the target

src/coreclr/jit/codegen.h

@@ -559,8 +559,6 @@ class CodeGen final : public CodeGenInterface
     void genFnProlog();
     void genFnEpilog(BasicBlock* block);
 
-#if defined(FEATURE_EH_FUNCLETS)
-
     void genReserveFuncletProlog(BasicBlock* block);
     void genReserveFuncletEpilog(BasicBlock* block);
     void genFuncletProlog(BasicBlock* block);
@@ -643,16 +641,6 @@ class CodeGen final : public CodeGenInterface
 
     void genUpdateCurrentFunclet(BasicBlock* block);
 
-#else // !FEATURE_EH_FUNCLETS
-
-    // This is a no-op when there are no funclets!
-    void genUpdateCurrentFunclet(BasicBlock* block)
-    {
-        return;
-    }
-
-#endif // !FEATURE_EH_FUNCLETS
-
     void genGeneratePrologsAndEpilogs();
 
 #if defined(DEBUG)
@@ -747,9 +735,7 @@ class CodeGen final : public CodeGenInterface
     void siOpenScopesForNonTrackedVars(const BasicBlock* block, unsigned int lastBlockILEndOffset);
 
 protected:
-#if defined(FEATURE_EH_FUNCLETS)
     bool siInFuncletRegion; // Have we seen the start of the funclet region?
-#endif                      // FEATURE_EH_FUNCLETS
 
     IL_OFFSET siLastEndOffs; // IL offset of the (exclusive) end of the last block processed
 
@@ -1294,11 +1280,10 @@ class CodeGen final : public CodeGenInterface
     void genCodeForBfiz(GenTreeOp* tree);
 #endif // TARGET_ARM64
 
-#if defined(FEATURE_EH_FUNCLETS)
     void genEHCatchRet(BasicBlock* block);
-#else  // !FEATURE_EH_FUNCLETS
+#if defined(FEATURE_EH_WINDOWS_X86)
     void genEHFinallyOrFilterRet(BasicBlock* block);
-#endif // !FEATURE_EH_FUNCLETS
+#endif // FEATURE_EH_WINDOWS_X86
 
     void genMultiRegStoreToSIMDLocal(GenTreeLclVar* lclNode);
     void genMultiRegStoreToLocal(GenTreeLclVar* lclNode);