Don't generate trampoline for deferred static compile failures

Summary:
At some point this was necessary - we didn't *always* call the native entrypoint w/ the first argument. But that time's long gone.

Instead of generate a trampoline at the end of the function let's just go to the de-opt trampoline and have it grab the right function.

Reviewed By: alexmalyshev

Differential Revision: D96419335

fbshipit-source-id: c75e733b8fa9f689fd42c05985f72da60e8d4a08

Author: DinoV

cinderx/Jit/codegen/environ.h

@@ -74,7 +74,6 @@ struct Environ {
     explicit IndirectInfo(void** indirect_ptr) : indirect(indirect_ptr) {}
 
     void** indirect;
-    asmjit::Label trampoline{0};
   };
   UnorderedMap<PyFunctionObject*, IndirectInfo> function_indirections;
 

cinderx/Jit/codegen/gen_asm.cpp

@@ -921,9 +921,7 @@ void* generateFailedDeferredCompileTrampoline() {
 
   annot.add("saveRegisters", &a, annot_cursor);
 
-  // r10 contains the function object from our stub
-  a.mov(x86::rdi, x86::r10);
-  a.mov(x86::rsi, x86::rsp);
+  a.mov(x86::rdi, x86::rsp);
   a.call(reinterpret_cast<uint64_t>(JITRT_FailedDeferredCompileShim));
   a.leave();
   a.ret();
@@ -941,9 +939,7 @@ void* generateFailedDeferredCompileTrampoline() {
 
   annot.add("saveRegisters", &a, annot_cursor);
 
-  // x10 contains the function object from our stub
-  a.mov(a64::x0, a64::x10);
-  a.mov(a64::x1, a64::sp);
+  a.mov(a64::x0, a64::sp);
   a.mov(arch::reg_scratch_br, JITRT_FailedDeferredCompileShim);
   a.blr(arch::reg_scratch_br);
   a.mov(a64::sp, arch::fp);
@@ -2076,17 +2072,6 @@ void NativeGenerator::generateEpilogue(BaseNode* epilogue_cursor) {
       "Epilogue (restore regs; pop native frame; error exit)",
       epilogue_error_cursor);
   env_.addAnnotation("Epilogue", epilogue_cursor);
-  if (env_.function_indirections.size()) {
-    auto jit_helpers = as_->cursor();
-    for (auto& x : env_.function_indirections) {
-      Label trampoline = as_->newLabel();
-      as_->bind(trampoline);
-      as_->mov(x86::r10, reinterpret_cast<uint64_t>(x.first));
-      as_->jmp(reinterpret_cast<uint64_t>(failed_deferred_compile_trampoline_));
-      x.second.trampoline = trampoline;
-    }
-    env_.addAnnotation("JitHelpers", jit_helpers);
-  }
 #elif defined(CINDER_AARCH64)
   bool is_gen = GetFunction()->code->co_flags & kCoFlagsAnyGenerator;
   if (is_gen) {
@@ -2172,18 +2157,6 @@ void NativeGenerator::generateEpilogue(BaseNode* epilogue_cursor) {
       "Epilogue (restore regs; pop native frame; error exit)",
       epilogue_error_cursor);
   env_.addAnnotation("Epilogue", epilogue_cursor);
-  if (env_.function_indirections.size()) {
-    auto jit_helpers = as_->cursor();
-    for (auto& x : env_.function_indirections) {
-      Label trampoline = as_->newLabel();
-      as_->bind(trampoline);
-      as_->mov(a64::x10, reinterpret_cast<uint64_t>(x.first));
-      as_->mov(arch::reg_scratch_br, failed_deferred_compile_trampoline_);
-      as_->br(arch::reg_scratch_br);
-      x.second.trampoline = trampoline;
-    }
-    env_.addAnnotation("JitHelpers", jit_helpers);
-  }
 #else
   CINDER_UNSUPPORTED
 #endif
@@ -2956,10 +2929,7 @@ void NativeGenerator::generateCode(CodeHolder& codeholder) {
   {
     ThreadedCompileSerialize guard;
     for (auto& x : env_.function_indirections) {
-      Label trampoline = x.second.trampoline;
-      *x.second.indirect = reinterpret_cast<void*>(
-          codeholder.labelOffsetFromBase(trampoline) +
-          codeholder.baseAddress());
+      *x.second.indirect = failed_deferred_compile_trampoline_;
     }
   }
 

cinderx/Jit/jit_rt.cpp

@@ -1769,11 +1769,12 @@ PyObject* JITRT_BuildString(
   return _PyUnicode_JoinArray(empty, args, nargs);
 }
 
-JITRT_StaticCallReturn JITRT_FailedDeferredCompileShim(
-    PyFunctionObject* func,
-    PyObject** args) {
+JITRT_StaticCallReturn JITRT_FailedDeferredCompileShim(PyObject** args) {
   void* no_error = reinterpret_cast<void*>(1);
 
+  // The function object is always the first argument in the static calling
+  // convention.
+  PyFunctionObject* func = reinterpret_cast<PyFunctionObject*>(args[0]);
   PyCodeObject* code = (PyCodeObject*)func->func_code;
   int total_args = code->co_argcount;
   if (code->co_flags & CO_VARARGS) {

cinderx/Jit/jit_rt.h

@@ -469,10 +469,11 @@ PyObject* JITRT_UnpackExToTuple(
  * work in this function would have to be done anyway if we were initially
  * making a JIT static -> non-JIT static function anyway, so there is not too
  * much overhead.
+ *
+ * The function object is obtained from args[0] since in the static calling
+ * convention the function is always the first argument.
  */
-JITRT_StaticCallReturn JITRT_FailedDeferredCompileShim(
-    PyFunctionObject* func,
-    PyObject** args);
+JITRT_StaticCallReturn JITRT_FailedDeferredCompileShim(PyObject** args);
 
 JITRT_StaticCallReturn JITRT_CallStaticallyWithPrimitiveSignature(
     PyFunctionObject* func,