On Mon, Jun 30, 2025 at 2:06 PM Hongtao Liu <crazy...@gmail.com> wrote: > > On Mon, Jun 30, 2025 at 11:16 AM H.J. Lu <hjl.to...@gmail.com> wrote: > > > > On Mon, Jun 30, 2025 at 10:37 AM Hongtao Liu <crazy...@gmail.com> wrote: > > > > > > On Sat, Jun 28, 2025 at 8:30 PM H.J. Lu <hjl.to...@gmail.com> wrote: > > > > > > > > Update functions with no_callee_saved_registers/preserve_none attribute > > > > to preserve frame pointer since caller may use it to save the current > > > > stack: > > > > > > > > pushq %rbp > > > > movq %rsp, %rbp > > > > ... > > > > call function > > > > ... > > > > leave > > > > ret > > > > > > > > If callee changes frame pointer without restoring it, caller will fail > > > > to restore its stack after callee returns. > > > Do we know why the caller failed to restore rbp? Are there any > > > assumptions in the middle-end that frame pointers must be callee saved > > > registers(even if it's marked as caller-saved)? > > > > Since RBP is changed by callee, "leave", which does > > > > mov %rbp, %rsp > > pop %rbp > > > > no longer works as RBP is wrong. > Reading the testcase, it looks like the issue is related to the > mixusaged of omit-frame-pointer in the callee but > no-omit-frame-pointer in the caller side relies on the fact that rbp > must be a callee-saved register?
Correct. RBP must be a callee-saved register. The bug is only fixed on master. I'd like to backport it to release branches after a few days. > Since ira will take HARD_FRAME_POINTER_REGNUM as a fixed register in > the caller(when no-omit-frame-pointer) but callee takes it as a > scratch register (since omit-frame-pointer && no_callee_saved_register > attribute) > > > > > > > > > > > > > /* The current function is a function specified with the > > > "no_callee_saved_registers" attribute. */ > > > TYPE_NO_CALLEE_SAVED_REGISTERS, > > > - /* The current function is a function specified with the "noreturn" > > > - attribute. */ > > > - TYPE_NO_CALLEE_SAVED_REGISTERS_EXCEPT_BP, > > > > > > Comments of "noreturn" part should be merged into that of > > > TYPE_NO_CALLEE_SAVED_REGISTERS. > > > > Will update. > > > > > +callee-saved registers. That is, all registers, except for stack and > > > +frame pointers, can be used as scratch registers. For example, this > > > > > > The patch only excludes frame pointers but the document mentions both > > > stack and frame pointers? > > > > Stack pointer is implied. My patch makes it clear. > It's a fixed register, I c. > > > > > > > > > > > -- > > > BR, > > > Hongtao > > > > > > > > -- > > H.J. > > > > -- > BR, > Hongtao -- H.J.
