gh-149800: Split perf trampoline assembly into per-architecture files (#149894)

Author: stratakis

Makefile.pre.in

@@ -3111,9 +3111,18 @@ config.status:	$(srcdir)/configure
 
 .PRECIOUS: config.status $(BUILDPYTHON) Makefile Makefile.pre
 
-Python/asm_trampoline.o: $(srcdir)/Python/asm_trampoline.S
+Python/asm_trampoline_x86_64.o: $(srcdir)/Python/asm_trampoline_x86_64.S
 	$(CC) -c $(PY_CORE_CFLAGS) -o $@ $<
 
+Python/asm_trampoline_aarch64.o: $(srcdir)/Python/asm_trampoline_aarch64.S
+	$(CC) -c $(PY_CORE_CFLAGS) -o $@ $<
+
+Python/asm_trampoline_riscv64.o: $(srcdir)/Python/asm_trampoline_riscv64.S
+	$(CC) -c $(PY_CORE_CFLAGS) -o $@ $<
+
+Python/asm_trampoline_universal2.o: Python/asm_trampoline_aarch64.o Python/asm_trampoline_x86_64.o
+	lipo -create -output $@ Python/asm_trampoline_aarch64.o Python/asm_trampoline_x86_64.o
+
 Python/emscripten_trampoline_inner.wasm: $(srcdir)/Python/emscripten_trampoline_inner.c
 	# emcc has a path that ends with emsdk/upstream/emscripten/emcc, we're looking for emsdk/upstream/bin/clang.
 	$$(dirname $$(dirname $(CC)))/bin/clang -o $@ $< -mgc -O2 -Wl,--no-entry -Wl,--import-table -Wl,--import-memory -target wasm32-unknown-unknown -nostdlib

Python/asm_trampoline_aarch64.h Python/asm_trampoline_aarch64.SPython/asm_trampoline_aarch64.h renamed to Python/asm_trampoline_aarch64.S

@@ -1,6 +1,3 @@
-#ifndef ASM_TRAMPOLINE_AARCH_64_H_
-#define ASM_TRAMPOLINE_AARCH_64_H_
-
 /*
  * References:
  *  - https://developer.arm.com/documentation/101028/0012/5--Feature-test-macros
@@ -38,6 +35,31 @@
   #define GNU_PROPERTY_AARCH64_GCS 0
 #endif
 
+    .text
+#if defined(__aarch64__) && defined(__AARCH64EL__) && !defined(__ILP32__)
+#if defined(__APPLE__)
+    .globl	__Py_trampoline_func_start
+__Py_trampoline_func_start:
+#else
+    .globl	_Py_trampoline_func_start
+_Py_trampoline_func_start:
+#endif
+    SIGN_LR
+    stp     x29, x30, [sp, -16]!
+    mov     x29, sp
+    blr     x3
+    ldp     x29, x30, [sp], 16
+    VERIFY_LR
+    ret
+#if defined(__APPLE__)
+    .globl	__Py_trampoline_func_end
+__Py_trampoline_func_end:
+#else
+    .globl	_Py_trampoline_func_end
+_Py_trampoline_func_end:
+    .section        .note.GNU-stack,"",@progbits
+#endif
+
 /* Add the BTI, PAC and GCS support to GNU Notes section */
 #if GNU_PROPERTY_AARCH64_BTI != 0 || GNU_PROPERTY_AARCH64_POINTER_AUTH != 0 || GNU_PROPERTY_AARCH64_GCS != 0
     .pushsection .note.gnu.property, "a"; /* Start a new allocatable section */
@@ -52,5 +74,4 @@
     .long 0; /* padding for 8 byte alignment */
     .popsection; /* end the section */
 #endif
-
-#endif
+#endif // __aarch64__ && __AARCH64EL__ && !__ILP32__

Python/asm_trampoline_riscv64.S

@@ -0,0 +1,12 @@
+    .text
+    .globl	_Py_trampoline_func_start
+_Py_trampoline_func_start:
+    addi    sp,sp,-16
+    sd      ra,8(sp)
+    jalr    a3
+    ld      ra,8(sp)
+    addi    sp,sp,16
+    jr      ra
+    .globl	_Py_trampoline_func_end
+_Py_trampoline_func_end:
+    .section        .note.GNU-stack,"",@progbits

Python/asm_trampoline.S Python/asm_trampoline_x86_64.SPython/asm_trampoline.S renamed to Python/asm_trampoline_x86_64.S

@@ -1,24 +1,12 @@
-#include "asm_trampoline_aarch64.h"
-
     .text
+#ifdef __x86_64__
 #if defined(__APPLE__)
     .globl	__Py_trampoline_func_start
-#else
-    .globl	_Py_trampoline_func_start
-#endif
-# The following assembly is equivalent to:
-# PyObject *
-# trampoline(PyThreadState *ts, _PyInterpreterFrame *f,
-#            int throwflag, py_evaluator evaluator)
-# {
-#     return evaluator(ts, f, throwflag);
-# }
-#if defined(__APPLE__)
 __Py_trampoline_func_start:
 #else
+    .globl	_Py_trampoline_func_start
 _Py_trampoline_func_start:
 #endif
-#ifdef __x86_64__
 #if defined(__CET__) && (__CET__ & 1)
     endbr64
 #endif
@@ -27,26 +15,6 @@ _Py_trampoline_func_start:
     call   *%rcx
     pop    %rbp
     ret
-#endif // __x86_64__
-#if defined(__aarch64__) && defined(__AARCH64EL__) && !defined(__ILP32__)
-    // ARM64 little endian, 64bit ABI
-    // generate with aarch64-linux-gnu-gcc 12.1
-    SIGN_LR
-    stp     x29, x30, [sp, -16]!
-    mov     x29, sp
-    blr     x3
-    ldp     x29, x30, [sp], 16
-    VERIFY_LR
-    ret
-#endif
-#ifdef __riscv
-    addi    sp,sp,-16
-    sd      ra,8(sp)
-    jalr    a3
-    ld      ra,8(sp)
-    addi    sp,sp,16
-    jr      ra
-#endif
 #if defined(__APPLE__)
     .globl	__Py_trampoline_func_end
 __Py_trampoline_func_end:
@@ -56,7 +24,7 @@ _Py_trampoline_func_end:
     .section        .note.GNU-stack,"",@progbits
 #endif
 # Note for indicating the assembly code supports CET
-#if defined(__x86_64__) && defined(__CET__) && (__CET__ & 1)
+#if defined(__CET__) && (__CET__ & 1)
     .section    .note.gnu.property,"a"
     .align 8
     .long    1f - 0f
@@ -73,4 +41,5 @@ _Py_trampoline_func_end:
 3:
     .align 8
 4:
+#endif
 #endif // __x86_64__

configure

@@ -3866,20 +3866,37 @@ AC_MSG_RESULT([$SHLIBS])
 dnl perf trampoline is Linux and macOS specific and requires an arch-specific
 dnl trampoline in assembly.
 AC_MSG_CHECKING([perf trampoline])
+PERF_TRAMPOLINE_OBJ=""
 AS_CASE([$PLATFORM_TRIPLET],
-  [x86_64-linux-gnu], [perf_trampoline=yes],
-  [aarch64-linux-gnu], [perf_trampoline=yes],
+  [x86_64-linux-gnu], [perf_trampoline=yes
+                        PERF_TRAMPOLINE_OBJ=Python/asm_trampoline_x86_64.o],
+  [aarch64-linux-gnu], [perf_trampoline=yes
+                         PERF_TRAMPOLINE_OBJ=Python/asm_trampoline_aarch64.o],
   [darwin], [AS_CASE([$MACOSX_DEPLOYMENT_TARGET],
                 [[10.[0-9]|10.1[0-1]]], [perf_trampoline=no],
-                [perf_trampoline=yes]
+                [perf_trampoline=yes
+                 if test "${enable_universalsdk}" && test "$UNIVERSAL_ARCHS" = "universal2"; then
+                   PERF_TRAMPOLINE_OBJ=Python/asm_trampoline_universal2.o
+                 else
+                   case "$host_cpu" in
+                     x86_64)
+                       PERF_TRAMPOLINE_OBJ=Python/asm_trampoline_x86_64.o
+                       ;;
+                     aarch64|arm64)
+                       PERF_TRAMPOLINE_OBJ=Python/asm_trampoline_aarch64.o
+                       ;;
+                     *)
+                       perf_trampoline=no
+                       ;;
+                   esac
+                 fi]
             )],
   [perf_trampoline=no]
 )
 AC_MSG_RESULT([$perf_trampoline])
 
 AS_VAR_IF([perf_trampoline], [yes], [
   AC_DEFINE([PY_HAVE_PERF_TRAMPOLINE], [1], [Define to 1 if you have the perf trampoline.])
-  PERF_TRAMPOLINE_OBJ=Python/asm_trampoline.o
 ])
 AC_SUBST([PERF_TRAMPOLINE_OBJ])