On 29/04/23 5:03 am, Jeff Law wrote:
> 
> 
> On 4/28/23 16:42, Hans-Peter Nilsson wrote:
>> On Sat, 22 Apr 2023, Ajit Agarwal via Gcc-patches wrote:
>>
>>> Hello All:
>>>
>>> This new version of patch 4 use improve ree pass for rs6000 target using 
>>> defined ABI interfaces.
>>> Bootstrapped and regtested on power64-linux-gnu.
>>>
>>> Thanks & Regards
>>> Ajit
>>>
>>>
>>>     ree: Improve ree pass for rs6000 target using defined abi interfaces
>>>
>>>          For rs6000 target we see redundant zero and sign
>>>          extension and done to improve ree pass to eliminate
>>>          such redundant zero and sign extension using defines
>>>          ABI interfaces.
>>>
>>>          2023-04-22  Ajit Kumar Agarwal  <aagar...@linux.ibm.com>
>>>
>>> gcc/ChangeLog:
>>>
>>>          * ree.cc (combline_reaching_defs): Add zero_extend
>>>          using defined abi interfaces.
>>>          (add_removable_extension): use of defined abi interfaces
>>>          for no reaching defs.
>>>          (abi_extension_candidate_return_reg_p): New defined ABI function.
>>>          (abi_extension_candidate_p): New defined ABI function.
>>>          (abi_extension_candidate_argno_p): New defined ABI function.
>>>          (abi_handle_regs_without_defs_p): New defined ABI function.
>>>
>>> gcc/testsuite/ChangeLog:
>>>
>>>          * g++.target/powerpc/zext-elim-3.C
>>> ---
>>>   gcc/ree.cc                                    | 176 +++++++++++++++---
>>>   .../g++.target/powerpc/zext-elim-3.C          |  16 ++
>>>   2 files changed, 162 insertions(+), 30 deletions(-)
>>>   create mode 100644 gcc/testsuite/g++.target/powerpc/zext-elim-3.C
>>>
>>> diff --git a/gcc/ree.cc b/gcc/ree.cc
>>> index 413aec7c8eb..0de96b1ece1 100644
>>> --- a/gcc/ree.cc
>>> +++ b/gcc/ree.cc
>>> @@ -473,7 +473,8 @@ get_defs (rtx_insn *insn, rtx reg, vec<rtx_insn *> 
>>> *dest)
>>>       break;
>>>       }
>>>   -  gcc_assert (use != NULL);
>>> +  if (use == NULL)
>>> +    return NULL;
>>>       ref_chain = DF_REF_CHAIN (use);
>>>   @@ -514,7 +515,8 @@ get_uses (rtx_insn *insn, rtx reg)
>>>       if (REGNO (DF_REF_REG (def)) == REGNO (reg))
>>>         break;
>>>   -  gcc_assert (def != NULL);
>>> +  if (def == NULL)
>>> +    return NULL;
>>>       ref_chain = DF_REF_CHAIN (def);
>>>   @@ -750,6 +752,103 @@ get_extended_src_reg (rtx src)
>>>     return src;
>>>   }
>>>   +/* Return TRUE if the candidate insn is zero extend and regno is
>>> +   an return  registers.  */
>>> +
>>> +static bool
>>> +abi_extension_candidate_return_reg_p (rtx_insn *insn, int regno)
>>> +{
>>> +  rtx set = single_set (insn);
>>> +
>>> +  if (GET_CODE (SET_SRC (set)) !=  ZERO_EXTEND)
>>> +    return false;
>>> +
>>> +  if (FUNCTION_VALUE_REGNO_P (regno))
>>> +    return true;
>>> +
>>> +  return false;
>>> +}
>>> +
>>> +/* Return TRUE if reg source operand of zero_extend is argument registers
>>> +   and not return registers and source and destination operand are same
>>> +   and mode of source and destination operand are not same.  */
>>> +
>>> +static bool
>>> +abi_extension_candidate_p (rtx_insn *insn)
>>> +{
>>> +  rtx set = single_set (insn);
>>> +
>>> +  if (GET_CODE (SET_SRC (set)) !=  ZERO_EXTEND)
>>> +    return false;
>>> +
>>> +  machine_mode ext_dst_mode = GET_MODE (SET_DEST (set));
>>> +  rtx orig_src = XEXP (SET_SRC (set),0);
>>> +
>>> +  bool copy_needed
>>> +    = (REGNO (SET_DEST (set)) != REGNO (XEXP (SET_SRC (set), 0)));
>>> +
>>> +  if (!copy_needed && ext_dst_mode != GET_MODE (orig_src)
>>> +      && FUNCTION_ARG_REGNO_P (REGNO (orig_src))
>>> +      && !abi_extension_candidate_return_reg_p (insn, REGNO (orig_src)))
>>> +    return true;
>>> +
>>> +  return false;
>>> +}
>>> +
>>> +/* Return TRUE if the candidate insn is zero extend and regno is
>>> +   an argument registers.  */
>>> +
>>> +static bool
>>> +abi_extension_candidate_argno_p (rtx_code code, int regno)
>>> +{
>>> +  if (code !=  ZERO_EXTEND)
>>> +    return false;
>>> +
>>> +  if (FUNCTION_ARG_REGNO_P (regno))
>>> +    return true;
>>> +
>>> +  return false;
>>> +}
>>
>> I don't see anything in those functions that checks if
>> ZERO_EXTEND is actually a feature of the ABI, e.g. as opposed to
>> no extension or SIGN_EXTEND.  Do I miss something?
> I don't think you missed anything.  That was one of the points I was making 
> last week.  Somewhere, somehow we need to describe what the ABI mandates and 
> guarantees.
> 
> So while what Ajit has done is a step forward, at some point the actual 
> details of the ABI need to be described in a way that can be checked and 
> consumed by REE.


The ABI we need for ree pass are the argument registers and return registers. 
Based on that I have described interfaces that we need. Other than that we dont 
any other ABI hooks. I have used FUNCTION_VALUE_REGNO_P and 
FuNCTION_ARG_REGNO_P abi hooks.

Thanks & Regards
Ajit
> 
> Jeff

Reply via email to