circuitpython/py/nlrx64.c

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

133 lines
4.9 KiB
C
Raw Permalink Normal View History

/*
* This file is part of the MicroPython project, http://micropython.org/
*
* The MIT License (MIT)
*
* Copyright (c) 2013-2017 Damien P. George
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
* THE SOFTWARE.
*/
#include "py/mpstate.h"
2018-07-12 14:13:51 -04:00
#if defined(MICROPY_NLR_X64) && MICROPY_NLR_X64
#undef nlr_push
// x86-64 callee-save registers are:
// rbx, rbp, rsp, r12, r13, r14, r15
Revert "py/nlr: Factor out common NLR code to generic functions." This reverts commit 6a3a742a6c9caaa2be0fd0aac7a5df4ac816081c. The above commit has number of faults starting from the motivation down to the actual implementation. 1. Faulty implementation. The original code contained functions like: NORETURN void nlr_jump(void *val) { nlr_buf_t **top_ptr = &MP_STATE_THREAD(nlr_top); nlr_buf_t *top = *top_ptr; ... __asm volatile ( "mov %0, %%edx \n" // %edx points to nlr_buf "mov 28(%%edx), %%esi \n" // load saved %esi "mov 24(%%edx), %%edi \n" // load saved %edi "mov 20(%%edx), %%ebx \n" // load saved %ebx "mov 16(%%edx), %%esp \n" // load saved %esp "mov 12(%%edx), %%ebp \n" // load saved %ebp "mov 8(%%edx), %%eax \n" // load saved %eip "mov %%eax, (%%esp) \n" // store saved %eip to stack "xor %%eax, %%eax \n" // clear return register "inc %%al \n" // increase to make 1, non-local return "ret \n" // return : // output operands : "r"(top) // input operands : // clobbered registers ); } Which clearly stated that C-level variable should be a parameter of the assembly, whcih then moved it into correct register. Whereas now it's: NORETURN void nlr_jump_tail(nlr_buf_t *top) { (void)top; __asm volatile ( "mov 28(%edx), %esi \n" // load saved %esi "mov 24(%edx), %edi \n" // load saved %edi "mov 20(%edx), %ebx \n" // load saved %ebx "mov 16(%edx), %esp \n" // load saved %esp "mov 12(%edx), %ebp \n" // load saved %ebp "mov 8(%edx), %eax \n" // load saved %eip "mov %eax, (%esp) \n" // store saved %eip to stack "xor %eax, %eax \n" // clear return register "inc %al \n" // increase to make 1, non-local return "ret \n" // return ); for (;;); // needed to silence compiler warning } Which just tries to perform operations on a completely random register (edx in this case). The outcome is the expected: saving the pure random luck of the compiler putting the right value in the random register above, there's a crash. 2. Non-critical assessment. The original commit message says "There is a small overhead introduced (typically 1 machine instruction)". That machine instruction is a call if a compiler doesn't perform tail optimization (happens regularly), and it's 1 instruction only with the broken code shown above, fixing it requires adding more. With inefficiencies already presented in the NLR code, the overhead becomes "considerable" (several times more than 1%), not "small". The commit message also says "This eliminates duplicated code.". An obvious way to eliminate duplication would be to factor out common code to macros, not introduce overhead and breakage like above. 3. Faulty motivation. All this started with a report of warnings/errors happening for a niche compiler. It could have been solved in one the direct ways: a) fixing it just for affected compiler(s); b) rewriting it in proper assembly (like it was before BTW); c) by not doing anything at all, MICROPY_NLR_SETJMP exists exactly to address minor-impact cases like thar (where a) or b) are not applicable). Instead, a backwards "solution" was put forward, leading to all the issues above. The best action thus appears to be revert and rework, not trying to work around what went haywire in the first place.
2017-12-26 11:39:51 -05:00
__attribute__((used)) unsigned int nlr_push_tail(nlr_buf_t *nlr);
#if !MICROPY_NLR_OS_WINDOWS
#if defined(__clang__) || (defined(__GNUC__) && __GNUC__ >= 8)
#define USE_NAKED 1
#else
// On older gcc the equivalent here is to force omit-frame-pointer
__attribute__((optimize("omit-frame-pointer")))
#endif
#endif
#if !defined(USE_NAKED)
#define USE_NAKED 0
#endif
#if USE_NAKED
// nlr_push prelude should never push frame pointer register ebp onto the stack
__attribute__((naked))
#endif
unsigned int nlr_push(nlr_buf_t *nlr) {
#if !USE_NAKED
(void)nlr;
#endif
#if MICROPY_NLR_OS_WINDOWS
__asm volatile (
2021-03-15 09:57:36 -04:00
"movq (%rsp), %rax \n" // load return %rip
"movq %rax, 16(%rcx) \n" // store %rip into nlr_buf
"movq %rbp, 24(%rcx) \n" // store %rbp into nlr_buf
"movq %rsp, 32(%rcx) \n" // store %rsp into nlr_buf
"movq %rbx, 40(%rcx) \n" // store %rbx into nlr_buf
"movq %r12, 48(%rcx) \n" // store %r12 into nlr_buf
"movq %r13, 56(%rcx) \n" // store %r13 into nlr_buf
"movq %r14, 64(%rcx) \n" // store %r14 into nlr_buf
"movq %r15, 72(%rcx) \n" // store %r15 into nlr_buf
"movq %rdi, 80(%rcx) \n" // store %rdr into nlr_buf
"movq %rsi, 88(%rcx) \n" // store %rsi into nlr_buf
"jmp nlr_push_tail \n" // do the rest in C
);
#else
__asm volatile (
2021-03-15 09:57:36 -04:00
"movq (%rsp), %rax \n" // load return %rip
"movq %rax, 16(%rdi) \n" // store %rip into nlr_buf
"movq %rbp, 24(%rdi) \n" // store %rbp into nlr_buf
"movq %rsp, 32(%rdi) \n" // store %rsp into nlr_buf
"movq %rbx, 40(%rdi) \n" // store %rbx into nlr_buf
"movq %r12, 48(%rdi) \n" // store %r12 into nlr_buf
"movq %r13, 56(%rdi) \n" // store %r13 into nlr_buf
"movq %r14, 64(%rdi) \n" // store %r14 into nlr_buf
"movq %r15, 72(%rdi) \n" // store %r15 into nlr_buf
#if defined(__APPLE__) && defined(__MACH__)
2021-03-15 09:57:36 -04:00
"jmp _nlr_push_tail \n" // do the rest in C
#else
"jmp nlr_push_tail \n" // do the rest in C
#endif
);
#endif
#if !USE_NAKED
return 0; // needed to silence compiler warning
#endif
}
Revert "py/nlr: Factor out common NLR code to generic functions." This reverts commit 6a3a742a6c9caaa2be0fd0aac7a5df4ac816081c. The above commit has number of faults starting from the motivation down to the actual implementation. 1. Faulty implementation. The original code contained functions like: NORETURN void nlr_jump(void *val) { nlr_buf_t **top_ptr = &MP_STATE_THREAD(nlr_top); nlr_buf_t *top = *top_ptr; ... __asm volatile ( "mov %0, %%edx \n" // %edx points to nlr_buf "mov 28(%%edx), %%esi \n" // load saved %esi "mov 24(%%edx), %%edi \n" // load saved %edi "mov 20(%%edx), %%ebx \n" // load saved %ebx "mov 16(%%edx), %%esp \n" // load saved %esp "mov 12(%%edx), %%ebp \n" // load saved %ebp "mov 8(%%edx), %%eax \n" // load saved %eip "mov %%eax, (%%esp) \n" // store saved %eip to stack "xor %%eax, %%eax \n" // clear return register "inc %%al \n" // increase to make 1, non-local return "ret \n" // return : // output operands : "r"(top) // input operands : // clobbered registers ); } Which clearly stated that C-level variable should be a parameter of the assembly, whcih then moved it into correct register. Whereas now it's: NORETURN void nlr_jump_tail(nlr_buf_t *top) { (void)top; __asm volatile ( "mov 28(%edx), %esi \n" // load saved %esi "mov 24(%edx), %edi \n" // load saved %edi "mov 20(%edx), %ebx \n" // load saved %ebx "mov 16(%edx), %esp \n" // load saved %esp "mov 12(%edx), %ebp \n" // load saved %ebp "mov 8(%edx), %eax \n" // load saved %eip "mov %eax, (%esp) \n" // store saved %eip to stack "xor %eax, %eax \n" // clear return register "inc %al \n" // increase to make 1, non-local return "ret \n" // return ); for (;;); // needed to silence compiler warning } Which just tries to perform operations on a completely random register (edx in this case). The outcome is the expected: saving the pure random luck of the compiler putting the right value in the random register above, there's a crash. 2. Non-critical assessment. The original commit message says "There is a small overhead introduced (typically 1 machine instruction)". That machine instruction is a call if a compiler doesn't perform tail optimization (happens regularly), and it's 1 instruction only with the broken code shown above, fixing it requires adding more. With inefficiencies already presented in the NLR code, the overhead becomes "considerable" (several times more than 1%), not "small". The commit message also says "This eliminates duplicated code.". An obvious way to eliminate duplication would be to factor out common code to macros, not introduce overhead and breakage like above. 3. Faulty motivation. All this started with a report of warnings/errors happening for a niche compiler. It could have been solved in one the direct ways: a) fixing it just for affected compiler(s); b) rewriting it in proper assembly (like it was before BTW); c) by not doing anything at all, MICROPY_NLR_SETJMP exists exactly to address minor-impact cases like thar (where a) or b) are not applicable). Instead, a backwards "solution" was put forward, leading to all the issues above. The best action thus appears to be revert and rework, not trying to work around what went haywire in the first place.
2017-12-26 11:39:51 -05:00
NORETURN void nlr_jump(void *val) {
MP_NLR_JUMP_HEAD(val, top)
__asm volatile (
2021-03-15 09:57:36 -04:00
"movq %0, %%rcx \n" // %rcx points to nlr_buf
#if MICROPY_NLR_OS_WINDOWS
"movq 88(%%rcx), %%rsi \n" // load saved %rsi
2021-02-17 18:28:09 -05:00
"movq 80(%%rcx), %%rdi \n" // load saved %rdi
2021-03-15 09:57:36 -04:00
#endif
"movq 72(%%rcx), %%r15 \n" // load saved %r15
"movq 64(%%rcx), %%r14 \n" // load saved %r14
"movq 56(%%rcx), %%r13 \n" // load saved %r13
"movq 48(%%rcx), %%r12 \n" // load saved %r12
"movq 40(%%rcx), %%rbx \n" // load saved %rbx
"movq 32(%%rcx), %%rsp \n" // load saved %rsp
"movq 24(%%rcx), %%rbp \n" // load saved %rbp
"movq 16(%%rcx), %%rax \n" // load saved %rip
"movq %%rax, (%%rsp) \n" // store saved %rip to stack
"xorq %%rax, %%rax \n" // clear return register
"inc %%al \n" // increase to make 1, non-local return
"ret \n" // return
: // output operands
: "r" (top) // input operands
: // clobbered registers
);
MP_UNREACHABLE
}
#endif // MICROPY_NLR_X64