diff options
| author | Vladimir N. Makarov <vmakarov@redhat.com> | 2025-02-25 15:01:15 -0500 |
|---|---|---|
| committer | Vladimir N. Makarov <vmakarov@redhat.com> | 2025-02-25 15:04:50 -0500 |
| commit | 2341f675edadd6370147d2bc55ca7761a7ecfaa1 (patch) | |
| tree | 4b96fc958ea2f747355e8a314a69cd103d80a397 /gcc/lra-int.h | |
| parent | 0bb431d0a77cf8dc790b9c61539b3eb6ab1710f0 (diff) | |
| download | gcc-2341f675edadd6370147d2bc55ca7761a7ecfaa1.zip gcc-2341f675edadd6370147d2bc55ca7761a7ecfaa1.tar.gz gcc-2341f675edadd6370147d2bc55ca7761a7ecfaa1.tar.bz2 | |
[PR115458][LRA]: Run split sub-pass more times
In this PR case LRA needs to provide too many hard regs for insn
reloads, where some reload pseudos require 8 aligned regs for
themselves. As the last attempt, LRA tries to split live ranges of
hard regs for insn reload pseudos. It is a very rare case. An
inheritance pseudo involving a reload pseudo of the insn can be
spilled in the assignment sub-pass run right after splitting and we need
to run split sub-pass for the inheritance pseudo now.
gcc/ChangeLog:
PR target/115458
* lra-int.h (LRA_MAX_FAILED_SPLITS): Define and check its value.
(lra_split_hard_reg_for): Change prototype.
* lra.cc (lra): Try to split hard reg range several times after a
failure.
* lra-assigns.cc (lra_split_hard_reg_for): Add an arg, a flag of
giving up. Report asm error and nullify the asm insn depending on
the arg value.
gcc/testsuite/ChangeLog:
PR target/115458
* g++.target/riscv/pr115458.C: New.
Diffstat (limited to 'gcc/lra-int.h')
| -rw-r--r-- | gcc/lra-int.h | 14 |
1 files changed, 13 insertions, 1 deletions
diff --git a/gcc/lra-int.h b/gcc/lra-int.h index 1f89e06..ad42f48 100644 --- a/gcc/lra-int.h +++ b/gcc/lra-int.h @@ -252,6 +252,18 @@ typedef class lra_insn_recog_data *lra_insn_recog_data_t; for preventing LRA cycling in a bug case. */ #define LRA_MAX_ASSIGNMENT_ITERATION_NUMBER 30 +/* Maximum allowed number of tries to split hard reg live ranges after failure + in assignment of reload pseudos. Theoretical bound for the value is the + number of the insn reload pseudos plus the number of inheritance pseudos + generated from the reload pseudos. This bound can be achieved when all the + reload pseudos and the inheritance pseudos require hard reg splitting for + their assignment. This is extremely unlikely event. */ +#define LRA_MAX_FAILED_SPLITS 10 + +#if LRA_MAX_FAILED_SPLITS >= LRA_MAX_ASSIGNMENT_ITERATION_NUMBER +#error wrong LRA_MAX_FAILED_SPLITS value +#endif + /* The maximal number of inheritance/split passes in LRA. It should be more 1 in order to perform caller saves transformations and much less MAX_CONSTRAINT_ITERATION_NUMBER to prevent LRA to do as many @@ -392,7 +404,7 @@ extern int lra_assignment_iter; extern int lra_assignment_iter_after_spill; extern void lra_setup_reg_renumber (int, int, bool); extern bool lra_assign (bool &); -extern bool lra_split_hard_reg_for (void); +extern bool lra_split_hard_reg_for (bool fail_p); /* lra-coalesce.cc: */ |