Message ID | 20200120080401.61443-1-kito.cheng@sifive.com |
---|---|
State | New |
Headers | show |
Series | RISC-V: Disallow regrenme if the TO register never used before for interrupt functions | expand |
On Mon, Jan 20, 2020 at 12:04 AM Kito Cheng <kito.cheng@sifive.com> wrote: > gcc/ChangeLog > > PR target/93304 > * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New. > * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New. > * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined. > > gcc/testsuite/ChangeLog > > PR target/93304 > * gcc.target/riscv/pr93304.c: New test. OK. By the way, you don't need blank lines after the PR target/93304. Jim
Hi Jim: Thanks, fixed and committed, and it's OK to commit to gcc 8/9 next week? On Tue, Jan 21, 2020 at 7:13 AM Jim Wilson <jimw@sifive.com> wrote: > On Mon, Jan 20, 2020 at 12:04 AM Kito Cheng <kito.cheng@sifive.com> wrote: > > gcc/ChangeLog > > > > PR target/93304 > > * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New. > > * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New. > > * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined. > > > > gcc/testsuite/ChangeLog > > > > PR target/93304 > > * gcc.target/riscv/pr93304.c: New test. > > OK. > > By the way, you don't need blank lines after the PR target/93304. > > Jim >
On Mon, Jan 20, 2020 at 6:44 PM Kito Cheng <kito.cheng@sifive.com> wrote:
> Thanks, fixed and committed, and it's OK to commit to gcc 8/9 next week?
Yes, that is OK with me.
Jim
Tested and committed to gcc 9 branch. On Wed, Jan 22, 2020 at 2:35 AM Jim Wilson <jimw@sifive.com> wrote: > On Mon, Jan 20, 2020 at 6:44 PM Kito Cheng <kito.cheng@sifive.com> wrote: > > Thanks, fixed and committed, and it's OK to commit to gcc 8/9 next week? > > Yes, that is OK with me. > > Jim >
diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h index 26b811007a9..8cf9137b5e7 100644 --- a/gcc/config/riscv/riscv-protos.h +++ b/gcc/config/riscv/riscv-protos.h @@ -89,4 +89,6 @@ extern void riscv_init_builtins (void); /* Routines implemented in riscv-common.c. */ extern std::string riscv_arch_str (); +extern bool riscv_hard_regno_rename_ok (unsigned, unsigned); + #endif /* ! GCC_RISCV_PROTOS_H */ diff --git a/gcc/config/riscv/riscv.c b/gcc/config/riscv/riscv.c index 320a70b9c86..573024074e3 100644 --- a/gcc/config/riscv/riscv.c +++ b/gcc/config/riscv/riscv.c @@ -5021,6 +5021,19 @@ riscv_reorg (void) riscv_remove_unneeded_save_restore_calls (); } +/* Return nonzero if register FROM_REGNO can be renamed to register + TO_REGNO. */ + +bool +riscv_hard_regno_rename_ok (unsigned from_regno ATTRIBUTE_UNUSED, + unsigned to_regno) +{ + /* Interrupt functions can only use registers that have already been + saved by the prologue, even if they would normally be + call-clobbered. */ + return !cfun->machine->interrupt_handler_p || df_regs_ever_live_p (to_regno); +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" diff --git a/gcc/config/riscv/riscv.h b/gcc/config/riscv/riscv.h index 0bf3d2f81a8..19438e28fe8 100644 --- a/gcc/config/riscv/riscv.h +++ b/gcc/config/riscv/riscv.h @@ -926,4 +926,6 @@ extern unsigned riscv_stack_boundary; extern void riscv_remove_unneeded_save_restore_calls (void); +#define HARD_REGNO_RENAME_OK(FROM, TO) riscv_hard_regno_rename_ok (FROM, TO) + #endif /* ! GCC_RISCV_H */ diff --git a/gcc/testsuite/gcc.target/riscv/pr93304.c b/gcc/testsuite/gcc.target/riscv/pr93304.c new file mode 100644 index 00000000000..f771e4859a9 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/pr93304.c @@ -0,0 +1,19 @@ +/* Verify the regrename won't rename registers to register which never used + before. */ +/* { dg-do compile } */ +/* { dg-options "-O -frename-registers" } */ + +static unsigned _t = 0; + +void __attribute__ ((interrupt)) +foo (void) +{ + _t++; +} + +/* Register rename will try to use registers from the lower register + regradless of the REG_ALLOC_ORDER. + In theory, t0-t6 should not used in such small program if regrename + not executed incorrectly, because a5-a0 has higher priority in + REG_ALLOC_ORDER. */ +/* { dg-final { scan-assembler-not "t\[0-6\]" } } */