Hi Haochen,
on 2023/11/28 15:43, HAO CHEN GUI wrote:
> Hi,
> This patch passes down the equality only flags from
> emit_block_cmp_hints to cmpmem optab so that the target specific expand
> can generate optimized insns for equality only compare. Targets
> (e.g. rs6000) can generate more efficient insn sequence if the block
> compare is equality only.
Add more CCs since this patch changes generic part of code.
>
> Bootstrapped and tested on x86 and powerpc64-linux BE and LE with
> no regressions. Is this OK for trunk?
>
> Thanks
> Gui Haochen
>
> ChangeLog
> Expand: Pass down equality only flag to cmpmem expand
>
> Targets (e.g. rs6000) can generate more efficient insn sequence if the
> block compare is equality only. This patch passes down the equality
> only flags from emit_block_cmp_hints to cmpmem optab so that the target
> specific expand can generate optimized insns for equality only compare.
>
> gcc/
> * expr.cc (expand_cmpstrn_or_cmpmem): Rename to...
Maybe we can still keep this function expand_cmpstrn_or_cmpmem but with
an additional argument like (int equality_only = -1, default as -1 means
the underlying optab expansion doesn't support equality_only, 1 and 0
stands for what you proposed), to avoid to duplicate expand_cmpstrn_or_cmpmem.
> (expand_cmpstrn): ...this.
> (expand_cmpmem): New function. Pass down equality only flag to
> cmpmem expand.
> (emit_block_cmp_via_cmpmem): Add an argument for equality only
> flag and call expand_cmpmem instead of expand_cmpstrn_or_cmpmem.
> (emit_block_cmp_hints): Call emit_block_cmp_via_cmpmem with
> equality only flag.
> * expr.h (expand_cmpstrn, expand_cmpmem): Declare.
> * builtins.cc (expand_builtin_strcmp, expand_builtin_strncmp):
> Call expand_cmpstrn instead of expand_cmpstrn_or_cmpmem.
> * config/i386/i386.md (cmpmemsi): Add the sixth operand for
> equality only flag.
> * config/rs6000/rs6000.md (cmpmemsi): Likewise.
> * config/s390/s390.md (cmpmemsi): Likewise.
> * doc/md.texi (cmpmem): Modify the document and add an operand
> for equality only flag.
>
> patch.diff
> diff --git a/gcc/builtins.cc b/gcc/builtins.cc
> index 5ece0d23eb9..c2dbc25433d 100644
> --- a/gcc/builtins.cc
> +++ b/gcc/builtins.cc
> @@ -4819,7 +4819,7 @@ expand_builtin_strcmp (tree exp, ATTRIBUTE_UNUSED rtx
> target)
> if (len && !TREE_SIDE_EFFECTS (len))
> {
> arg3_rtx = expand_normal (len);
> - result = expand_cmpstrn_or_cmpmem
> + result = expand_cmpstrn
> (cmpstrn_icode, target, arg1_rtx, arg2_rtx, TREE_TYPE (len),
> arg3_rtx, MIN (arg1_align, arg2_align));
> }
> @@ -4929,9 +4929,9 @@ expand_builtin_strncmp (tree exp, ATTRIBUTE_UNUSED rtx
> target,
> rtx arg1_rtx = get_memory_rtx (arg1, len);
> rtx arg2_rtx = get_memory_rtx (arg2, len);
> rtx arg3_rtx = expand_normal (len);
> - result = expand_cmpstrn_or_cmpmem (cmpstrn_icode, target, arg1_rtx,
> - arg2_rtx, TREE_TYPE (len), arg3_rtx,
> - MIN (arg1_align, arg2_align));
> + result = expand_cmpstrn (cmpstrn_icode, target, arg1_rtx, arg2_rtx,
> + TREE_TYPE (len), arg3_rtx,
> + MIN (arg1_align, arg2_align));
>
> tree fndecl = get_callee_fndecl (exp);
> if (result)
> diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md
> index 1b5a794b9e5..775cba5d93d 100644
> --- a/gcc/config/i386/i386.md
> +++ b/gcc/config/i386/i386.md
> @@ -23195,7 +23195,8 @@ (define_expand "cmpmemsi"
> (compare:SI (match_operand:BLK 1 "memory_operand" "")
> (match_operand:BLK 2 "memory_operand" "") ) )
> (use (match_operand 3 "general_operand"))
> - (use (match_operand 4 "immediate_operand"))]
> + (use (match_operand 4 "immediate_operand"))
> + (use (match_operand 5 ""))]
As the other operands are with predicates, maybe i386 folks want to
have "const_0_to_1_operand".
BR,
Kewen
> ""
> {
> if (ix86_expand_cmpstrn_or_cmpmem (operands[0], operands[1],
> diff --git a/gcc/config/rs6000/rs6000.md b/gcc/config/rs6000/rs6000.md
> index 2a1b5ecfaee..e66330f876e 100644
> --- a/gcc/config/rs6000/rs6000.md
> +++ b/gcc/config/rs6000/rs6000.md
> @@ -10097,7 +10097,8 @@ (define_expand "cmpmemsi"
> (compare:SI (match_operand:BLK 1)
> (match_operand:BLK 2)))
> (use (match_operand:SI 3))
> - (use (match_operand:SI 4))])]
> + (use (match_operand:SI 4))
> + (use (match_operand:SI 5))])]
> "TARGET_POPCNTD"
> {
> if (expand_block_compare (operands))
> diff --git a/gcc/config/s390/s390.md b/gcc/config/s390/s390.md
> index 4bdb679daf2..506e79fb035 100644
> --- a/gcc/config/s390/s390.md
> +++ b/gcc/config/s390/s390.md
> @@ -3790,7 +3790,8 @@ (define_expand "cmpmemsi"
> (compare:SI (match_operand:BLK 1 "memory_operand" "")
> (match_operand:BLK 2 "memory_operand" "") ) )
> (use (match_operand:SI 3 "general_operand" ""))
> - (use (match_operand:SI 4 "" ""))]
> + (use (match_operand:SI 4 "" ""))
> + (use (match_operand:SI 5 "" ""))]
> ""
> {
> if (s390_expand_cmpmem (operands[0], operands[1],
> diff --git a/gcc/doc/md.texi b/gcc/doc/md.texi
> index e01cdcbe22c..06955cd7e78 100644
> --- a/gcc/doc/md.texi
> +++ b/gcc/doc/md.texi
> @@ -6992,14 +6992,19 @@ result of the comparison.
>
> @cindex @code{cmpmem@var{m}} instruction pattern
> @item @samp{cmpmem@var{m}}
> -Block compare instruction, with five operands like the operands
> -of @samp{cmpstr@var{m}}. The two memory blocks specified are compared
> -byte by byte in lexicographic order starting at the beginning of each
> -block. Unlike @samp{cmpstr@var{m}} the instruction can prefetch
> -any bytes in the two memory blocks. Also unlike @samp{cmpstr@var{m}}
> -the comparison will not stop if both bytes are zero. The effect of
> -the instruction is to store a value in operand 0 whose sign indicates
> -the result of the comparison.
> +Block compare instruction, with six operands. The first five operands are
> +like the operands of @samp{cmpstr@var{m}}. The last operand indicates
> +whether the comparison is equality or not. Value one means it's an
> +equality only compare and zero means it's a non-equality compare.
> +
> +The two memory blocks specified are compared byte by byte in lexicographic
> +order starting at the beginning of each block. Unlike @samp{cmpstr@var{m}}
> +the instruction can prefetch any bytes in the two memory blocks. Also
> +unlike @samp{cmpstr@var{m}} the comparison will not stop if both bytes are
> +zero. When last operand is zero, the effect of the instruction is to store
> +a value in operand 0 whose sign indicates the result of the comparison.
> +When last operand is one, zero in operand 0 indicates two blocks are equal.
> +All other values in operand 0 indicate two blocks are not equal.
>
> @cindex @code{strlen@var{m}} instruction pattern
> @item @samp{strlen@var{m}}
> diff --git a/gcc/expr.cc b/gcc/expr.cc
> index 6dd9b8f2ce6..3cdc5181bd3 100644
> --- a/gcc/expr.cc
> +++ b/gcc/expr.cc
> @@ -2381,14 +2381,13 @@ emit_block_op_via_libcall (enum built_in_function
> fncode, rtx dst, rtx src,
> return expand_call (call_expr, NULL_RTX, false);
> }
>
> -/* Try to expand cmpstrn or cmpmem operation ICODE with the given operands.
> +/* Try to expand cmpstrn operation ICODE with the given operands.
> ARG3_TYPE is the type of ARG3_RTX. Return the result rtx on success,
> otherwise return null. */
>
> rtx
> -expand_cmpstrn_or_cmpmem (insn_code icode, rtx target, rtx arg1_rtx,
> - rtx arg2_rtx, tree arg3_type, rtx arg3_rtx,
> - HOST_WIDE_INT align)
> +expand_cmpstrn (insn_code icode, rtx target, rtx arg1_rtx, rtx arg2_rtx,
> + tree arg3_type, rtx arg3_rtx, HOST_WIDE_INT align)
> {
> machine_mode insn_mode = insn_data[icode].operand[0].mode;
>
> @@ -2407,6 +2406,34 @@ expand_cmpstrn_or_cmpmem (insn_code icode, rtx target,
> rtx arg1_rtx,
> return NULL_RTX;
> }
>
> +/* Similar as expand_cmpstrn, the last operand indicates whether it is a
> + equality comparison or not. */
> +rtx
> +expand_cmpmem (insn_code icode, rtx target, rtx arg1_rtx, rtx arg2_rtx,
> + tree arg3_type, rtx arg3_rtx, HOST_WIDE_INT align,
> + bool equality_only)
> +{
> + machine_mode insn_mode = insn_data[icode].operand[0].mode;
> +
> + if (target && (!REG_P (target) || HARD_REGISTER_P (target)))
> + target = NULL_RTX;
> +
> + class expand_operand ops[6];
> + create_output_operand (&ops[0], target, insn_mode);
> + create_fixed_operand (&ops[1], arg1_rtx);
> + create_fixed_operand (&ops[2], arg2_rtx);
> + create_convert_operand_from (&ops[3], arg3_rtx, TYPE_MODE (arg3_type),
> + TYPE_UNSIGNED (arg3_type));
> + create_integer_operand (&ops[4], align);
> + if (equality_only)
> + create_integer_operand (&ops[5], 1);
> + else
> + create_integer_operand (&ops[5], 0);
> + if (maybe_expand_insn (icode, 6, ops))
> + return ops[0].value;
> + return NULL_RTX;
> +}
> +
> /* Expand a block compare between X and Y with length LEN using the
> cmpmem optab, placing the result in TARGET. LEN_TYPE is the type
> of the expression that was used to calculate the length. ALIGN
> @@ -2414,7 +2441,7 @@ expand_cmpstrn_or_cmpmem (insn_code icode, rtx target,
> rtx arg1_rtx,
>
> static rtx
> emit_block_cmp_via_cmpmem (rtx x, rtx y, rtx len, tree len_type, rtx target,
> - unsigned align)
> + unsigned align, bool equality_only)
> {
> /* Note: The cmpstrnsi pattern, if it exists, is not suitable for
> implementing memcmp because it will stop if it encounters two
> @@ -2424,7 +2451,8 @@ emit_block_cmp_via_cmpmem (rtx x, rtx y, rtx len, tree
> len_type, rtx target,
> if (icode == CODE_FOR_nothing)
> return NULL_RTX;
>
> - return expand_cmpstrn_or_cmpmem (icode, target, x, y, len_type, len,
> align);
> + return expand_cmpmem (icode, target, x, y, len_type, len, align,
> + equality_only);
> }
>
> /* Emit code to compare a block Y to a block X. This may be done with
> @@ -2469,7 +2497,8 @@ emit_block_cmp_hints (rtx x, rtx y, rtx len, tree
> len_type, rtx target,
> result = compare_by_pieces (x, y, INTVAL (len), target, align,
> y_cfn, y_cfndata);
> else
> - result = emit_block_cmp_via_cmpmem (x, y, len, len_type, target, align);
> + result = emit_block_cmp_via_cmpmem (x, y, len, len_type, target, align,
> + equality_only);
>
> return result;
> }
> diff --git a/gcc/expr.h b/gcc/expr.h
> index 2a172867fdb..64dbbcfcaad 100644
> --- a/gcc/expr.h
> +++ b/gcc/expr.h
> @@ -199,8 +199,9 @@ extern void use_regs (rtx *, int, int);
> extern void use_group_regs (rtx *, rtx);
>
> #ifdef GCC_INSN_CODES_H
> -extern rtx expand_cmpstrn_or_cmpmem (insn_code, rtx, rtx, rtx, tree, rtx,
> - HOST_WIDE_INT);
> +extern rtx expand_cmpstrn (insn_code, rtx, rtx, rtx, tree, rtx,
> HOST_WIDE_INT);
> +extern rtx expand_cmpmem (insn_code, rtx, rtx, rtx, tree, rtx, HOST_WIDE_INT,
> + bool);
> #endif
>
> /* Write zeros through the storage of OBJECT.
