[RFA,PR,target/115478] Accept ADD, IOR or XOR when combining objects with no bits in common

  So the change to prefer ADD over IOR for combining two objects with no 
bits in common is (IMHO) generally good.  It has some minor fallout.

In particular the aarch64 port (and I suspect others) have patterns that 
recognize IOR, but not PLUS or XOR for these cases and thus tests which 
expected to optimize with IOR are no longer optimizing.

Roger suggested using a code iterator for this purpose.  Richard S. 
suggested a new match operator to cover those cases.

I really like the match operator idea, but as Richard S. notes in the PR 
it would require either not validating the "no bits in common", which 
dramatically reduces the utility IMHO or we'd need some work to allow 
consistent results without polluting the nonzero bits cache.

So this patch goes back to Roger's idea of just using a match iterator 
in the aarch64 backend (and presumably anywhere else we see this popping 
up).

Bootstrapped and regression tested on aarch64-linux-gnu where it fixes 
bitint-args.c (as expected).

OK for the trunk?

Jeff
PR target/115478
gcc/
	* config/aarch64/iterators.md (any_or_plus): New code iterator.
	* config/aarch64/aarch64.md (extr<mode>5_insn): Use any_or_plus.
	(extr<mode>5_insn_alt, extrsi5_insn_uxtw): Likewise.
	(extrsi5_insn_uxtw_alt, extrsi5_insn_di): Likewise.

gcc/testsuite/
	* gcc.target/aarch64/bitint-args.c: Update expected output.
  

Jeff Law <jeffreyalaw@gmail.com> writes:
> So the change to prefer ADD over IOR for combining two objects with no 
> bits in common is (IMHO) generally good.  It has some minor fallout.
>
> In particular the aarch64 port (and I suspect others) have patterns that 
> recognize IOR, but not PLUS or XOR for these cases and thus tests which 
> expected to optimize with IOR are no longer optimizing.
>
> Roger suggested using a code iterator for this purpose.  Richard S. 
> suggested a new match operator to cover those cases.
>
> I really like the match operator idea, but as Richard S. notes in the PR 
> it would require either not validating the "no bits in common", which 
> dramatically reduces the utility IMHO or we'd need some work to allow 
> consistent results without polluting the nonzero bits cache.
>
> So this patch goes back to Roger's idea of just using a match iterator 
> in the aarch64 backend (and presumably anywhere else we see this popping 
> up).
>
> Bootstrapped and regression tested on aarch64-linux-gnu where it fixes 
> bitint-args.c (as expected).
>
> OK for the trunk?
>
> Jeff
>
> 	PR target/115478
> gcc/
> 	* config/aarch64/iterators.md (any_or_plus): New code iterator.
> 	* config/aarch64/aarch64.md (extr<mode>5_insn): Use any_or_plus.
> 	(extr<mode>5_insn_alt, extrsi5_insn_uxtw): Likewise.
> 	(extrsi5_insn_uxtw_alt, extrsi5_insn_di): Likewise.
>
> gcc/testsuite/
> 	* gcc.target/aarch64/bitint-args.c: Update expected output.

OK, thanks!

(For the record, I agree that the match_operator thing requires too many
changes for stage 4.)

Richard

> diff --git a/gcc/config/aarch64/aarch64.md b/gcc/config/aarch64/aarch64.md
> index 071058dbeb3..cfe730f3732 100644
> --- a/gcc/config/aarch64/aarch64.md
> +++ b/gcc/config/aarch64/aarch64.md
> @@ -6194,10 +6194,11 @@
>  
>  (define_insn "*extr<mode>5_insn"
>    [(set (match_operand:GPI 0 "register_operand" "=r")
> -	(ior:GPI (ashift:GPI (match_operand:GPI 1 "register_operand" "r")
> -			     (match_operand 3 "const_int_operand" "n"))
> -		 (lshiftrt:GPI (match_operand:GPI 2 "register_operand" "r")
> -			       (match_operand 4 "const_int_operand" "n"))))]
> +	(any_or_plus:GPI
> +	  (ashift:GPI (match_operand:GPI 1 "register_operand" "r")
> +		      (match_operand 3 "const_int_operand" "n"))
> +	  (lshiftrt:GPI (match_operand:GPI 2 "register_operand" "r")
> +		      (match_operand 4 "const_int_operand" "n"))))]
>    "UINTVAL (operands[3]) < GET_MODE_BITSIZE (<MODE>mode) &&
>     (UINTVAL (operands[3]) + UINTVAL (operands[4]) == GET_MODE_BITSIZE (<MODE>mode))"
>    "extr\\t%<w>0, %<w>1, %<w>2, %4"
> @@ -6208,10 +6209,11 @@
>  ;; so we have to match both orderings.
>  (define_insn "*extr<mode>5_insn_alt"
>    [(set (match_operand:GPI 0 "register_operand" "=r")
> -	(ior:GPI  (lshiftrt:GPI (match_operand:GPI 2 "register_operand" "r")
> -			        (match_operand 4 "const_int_operand" "n"))
> -		  (ashift:GPI (match_operand:GPI 1 "register_operand" "r")
> -			      (match_operand 3 "const_int_operand" "n"))))]
> +	(any_or_plus:GPI
> +	  (lshiftrt:GPI (match_operand:GPI 2 "register_operand" "r")
> +		        (match_operand 4 "const_int_operand" "n"))
> +	  (ashift:GPI (match_operand:GPI 1 "register_operand" "r")
> +		      (match_operand 3 "const_int_operand" "n"))))]
>    "UINTVAL (operands[3]) < GET_MODE_BITSIZE (<MODE>mode)
>     && (UINTVAL (operands[3]) + UINTVAL (operands[4])
>         == GET_MODE_BITSIZE (<MODE>mode))"
> @@ -6223,10 +6225,11 @@
>  (define_insn "*extrsi5_insn_uxtw"
>    [(set (match_operand:DI 0 "register_operand" "=r")
>  	(zero_extend:DI
> -	 (ior:SI (ashift:SI (match_operand:SI 1 "register_operand" "r")
> -			    (match_operand 3 "const_int_operand" "n"))
> -		 (lshiftrt:SI (match_operand:SI 2 "register_operand" "r")
> -			      (match_operand 4 "const_int_operand" "n")))))]
> +	 (any_or_plus:SI
> +	   (ashift:SI (match_operand:SI 1 "register_operand" "r")
> +		      (match_operand 3 "const_int_operand" "n"))
> +	   (lshiftrt:SI (match_operand:SI 2 "register_operand" "r")
> +		      (match_operand 4 "const_int_operand" "n")))))]
>    "UINTVAL (operands[3]) < 32 &&
>     (UINTVAL (operands[3]) + UINTVAL (operands[4]) == 32)"
>    "extr\\t%w0, %w1, %w2, %4"
> @@ -6236,10 +6239,11 @@
>  (define_insn "*extrsi5_insn_uxtw_alt"
>    [(set (match_operand:DI 0 "register_operand" "=r")
>  	(zero_extend:DI
> -	 (ior:SI (lshiftrt:SI (match_operand:SI 2 "register_operand" "r")
> -			       (match_operand 4 "const_int_operand" "n"))
> -		 (ashift:SI (match_operand:SI 1 "register_operand" "r")
> -			    (match_operand 3 "const_int_operand" "n")))))]
> +	 (any_or_plus:SI
> +	   (lshiftrt:SI (match_operand:SI 2 "register_operand" "r")
> +			(match_operand 4 "const_int_operand" "n"))
> +	   (ashift:SI (match_operand:SI 1 "register_operand" "r")
> +		      (match_operand 3 "const_int_operand" "n")))))]
>    "UINTVAL (operands[3]) < 32 &&
>     (UINTVAL (operands[3]) + UINTVAL (operands[4]) == 32)"
>    "extr\\t%w0, %w1, %w2, %4"
> @@ -6248,13 +6252,13 @@
>  
>  (define_insn "*extrsi5_insn_di"
>    [(set (match_operand:SI 0 "register_operand" "=r")
> -	(ior:SI (ashift:SI (match_operand:SI 1 "register_operand" "r")
> -			   (match_operand 3 "const_int_operand" "n"))
> -		(match_operator:SI 6 "subreg_lowpart_operator"
> -		  [(zero_extract:DI
> -		     (match_operand:DI 2 "register_operand" "r")
> -		     (match_operand 5 "const_int_operand" "n")
> -		     (match_operand 4 "const_int_operand" "n"))])))]
> +	(any_or_plus:SI (ashift:SI (match_operand:SI 1 "register_operand" "r")
> +				    (match_operand 3 "const_int_operand" "n"))
> +			 (match_operator:SI 6 "subreg_lowpart_operator"
> +			  [(zero_extract:DI
> +			     (match_operand:DI 2 "register_operand" "r")
> +			     (match_operand 5 "const_int_operand" "n")
> +			     (match_operand 4 "const_int_operand" "n"))])))]
>    "UINTVAL (operands[3]) < 32
>     && UINTVAL (operands[3]) + UINTVAL (operands[4]) == 32
>     && INTVAL (operands[3]) == INTVAL (operands[5])"
> diff --git a/gcc/config/aarch64/iterators.md b/gcc/config/aarch64/iterators.md
> index 9fbd7493988..5bfd6e7d362 100644
> --- a/gcc/config/aarch64/iterators.md
> +++ b/gcc/config/aarch64/iterators.md
> @@ -2657,6 +2657,10 @@
>  ;; Code iterator for logical operations
>  (define_code_iterator LOGICAL [and ior xor])
>  
> +;; Code iterator for operations that are equivalent when the
> +;; two input operands are known have disjoint bits set.
> +(define_code_iterator any_or_plus [plus ior xor])
> +
>  ;; LOGICAL with plus, for when | gets converted to +.
>  (define_code_iterator LOGICAL_OR_PLUS [and ior xor plus])
>  
> diff --git a/gcc/testsuite/gcc.target/aarch64/bitint-args.c b/gcc/testsuite/gcc.target/aarch64/bitint-args.c
> index e7e1099c303..06b47246fa5 100644
> --- a/gcc/testsuite/gcc.target/aarch64/bitint-args.c
> +++ b/gcc/testsuite/gcc.target/aarch64/bitint-args.c
> @@ -69,7 +69,7 @@ CHECK_ARG(65)
>  ** f65:
>  **	extr	(x[0-9]+), x3, x2, 1
>  **	and	(x[0-9]+), x2, 1
> -**	orr	(x[0-9]+), \2, \1, lsl 1
> +**	add	(x[0-9]+), \2, \1, lsl 1
>  **	asr	(x[0-9]+), \1, 63
>  **	stp	\3, \4, \[x0\]
>  **	ret
> @@ -80,7 +80,7 @@ CHECK_ARG(127)
>  ** f127:
>  **	extr	(x[0-9]+), x3, x2, 63
>  **	and	(x[0-9]+), x2, 9223372036854775807
> -**	orr	(x[0-9]+), \2, \1, lsl 63
> +**	add	(x[0-9]+), \2, \1, lsl 63
>  **	asr	(x[0-9]+), \1, 1
>  **	stp	\3, \4, \[x0\]
>  **	ret