mirrors/cpython
1
0
Fork 0
mirror of https://github.com/python/cpython.git synced 2025-07-23 03:05:38 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 15

GH-135904: Improve the JIT's performance on macOS (GH-136528)

Browse source
This commit is contained in:
Brandt Bucher 2025-07-14 10:14:20 -07:00 • committed by GitHub
parent a68ddea3bf
commit 3d8c38f6db
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
6 changed files with 73 additions and 61 deletions
Download patch file Download diff file Expand all files Collapse all files

11
Python/jit.c
View file

@ -431,8 +431,10 @@ void patch_aarch64_trampoline(unsigned char *location, int ordinal, jit_state *s
#if defined(__aarch64__) || defined(_M_ARM64)
#define TRAMPOLINE_SIZE 16
#define DATA_ALIGN 8
#else
#define TRAMPOLINE_SIZE 0
#define DATA_ALIGN 1
#endif
// Generate and patch AArch64 trampolines. The symbols to jump to are stored
@ -522,8 +524,9 @@ _PyJIT_Compile(_PyExecutorObject *executor, const _PyUOpInstruction trace[], siz
// Round up to the nearest page:
size_t page_size = get_page_size();
assert((page_size & (page_size - 1)) == 0);
size_t padding = page_size - ((code_size + state.trampolines.size + data_size) & (page_size - 1));
size_t total_size = code_size + state.trampolines.size + data_size + padding;
size_t code_padding = DATA_ALIGN - ((code_size + state.trampolines.size) & (DATA_ALIGN - 1));
size_t padding = page_size - ((code_size + state.trampolines.size + code_padding + data_size) & (page_size - 1));
size_t total_size = code_size + state.trampolines.size + code_padding + data_size + padding;
unsigned char *memory = jit_alloc(total_size);
if (memory == NULL) {
return -1;
@ -545,7 +548,7 @@ _PyJIT_Compile(_PyExecutorObject *executor, const _PyUOpInstruction trace[], siz
// Loop again to emit the code:
unsigned char *code = memory;
state.trampolines.mem = memory + code_size;
unsigned char *data = memory + code_size + state.trampolines.size;
unsigned char *data = memory + code_size + state.trampolines.size + code_padding;
// Compile the shim, which handles converting between the native
// calling convention and the calling convention used by jitted code
// (which may be different for efficiency reasons).
@ -567,7 +570,7 @@ _PyJIT_Compile(_PyExecutorObject *executor, const _PyUOpInstruction trace[], siz
code += group->code_size;
data += group->data_size;
assert(code == memory + code_size);
assert(data == memory + code_size + state.trampolines.size + data_size);
assert(data == memory + code_size + state.trampolines.size + code_padding + data_size);
#ifdef MAP_JIT
pthread_jit_write_protect_np(1);
#endif