i386: Add peephole2 patterns to improve subtract with borrow with memory destination [PR79173]

This patch adds subborrow<mode> alternative so that it can have memory
destination and adds various peephole2s which help to match it.

2023-06-15  Jakub Jelinek  <jakub@redhat.com>

	PR middle-end/79173
	* config/i386/i386.md (subborrow<mode>): Add alternative with
	memory destination and add for it define_peephole2
	TARGET_READ_MODIFY_WRITE/-Os patterns to prefer using memory
	destination in these patterns.

1 files changed, 151 insertions, 3 deletions

diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md
index 23be04f..75eda25 100644
--- a/gcc/config/i386/i386.md
+++ b/gcc/config/i386/i386.md
@@ -8381,13 +8381,13 @@
   [(set (reg:CCC FLAGS_REG)
 	(compare:CCC
 	  (zero_extend:<DWI>
-	    (match_operand:SWI48 1 "nonimmediate_operand" "0"))
+	    (match_operand:SWI48 1 "nonimmediate_operand" "0,0"))
 	  (plus:<DWI>
 	    (match_operator:<DWI> 4 "ix86_carry_flag_operator"
 	      [(match_operand 3 "flags_reg_operand") (const_int 0)])
 	    (zero_extend:<DWI>
-	      (match_operand:SWI48 2 "nonimmediate_operand" "rm")))))
-   (set (match_operand:SWI48 0 "register_operand" "=r")
+	      (match_operand:SWI48 2 "nonimmediate_operand" "r,rm")))))
+   (set (match_operand:SWI48 0 "nonimmediate_operand" "=rm,r")
 	(minus:SWI48 (minus:SWI48
 		       (match_dup 1)
 		       (match_operator:SWI48 5 "ix86_carry_flag_operator"
@@ -8400,6 +8400,154 @@
    (set_attr "pent_pair" "pu")
    (set_attr "mode" "<MODE>")])
 
+(define_peephole2
+  [(set (match_operand:SWI48 0 "general_reg_operand")
+	(match_operand:SWI48 1 "memory_operand"))
+   (parallel [(set (reg:CCC FLAGS_REG)
+		   (compare:CCC
+		     (zero_extend:<DWI> (match_dup 0))
+		     (plus:<DWI>
+		       (match_operator:<DWI> 4 "ix86_carry_flag_operator"
+			 [(match_operand 3 "flags_reg_operand") (const_int 0)])
+		       (zero_extend:<DWI>
+			 (match_operand:SWI48 2 "memory_operand")))))
+	      (set (match_dup 0)
+		   (minus:SWI48
+		     (minus:SWI48
+		       (match_dup 0)
+		       (match_operator:SWI48 5 "ix86_carry_flag_operator"
+			 [(match_dup 3) (const_int 0)]))
+		     (match_dup 2)))])
+   (set (match_dup 1) (match_dup 0))]
+  "(TARGET_READ_MODIFY_WRITE || optimize_insn_for_size_p ())
+   && peep2_reg_dead_p (3, operands[0])
+   && !reg_overlap_mentioned_p (operands[0], operands[1])
+   && !reg_overlap_mentioned_p (operands[0], operands[2])"
+  [(set (match_dup 0) (match_dup 2))
+   (parallel [(set (reg:CCC FLAGS_REG)
+		   (compare:CCC
+		     (zero_extend:<DWI> (match_dup 1))
+		     (plus:<DWI> (match_op_dup 4
+				   [(match_dup 3) (const_int 0)])
+				 (zero_extend:<DWI> (match_dup 0)))))
+	      (set (match_dup 1)
+		   (minus:SWI48 (minus:SWI48 (match_dup 1)
+					     (match_op_dup 5
+					       [(match_dup 3) (const_int 0)]))
+				(match_dup 0)))])])
+
+(define_peephole2
+  [(set (match_operand:SWI48 6 "general_reg_operand")
+	(match_operand:SWI48 7 "memory_operand"))
+   (set (match_operand:SWI48 8 "general_reg_operand")
+	(match_operand:SWI48 9 "memory_operand"))
+   (parallel [(set (reg:CCC FLAGS_REG)
+		   (compare:CCC
+		     (zero_extend:<DWI>
+		       (match_operand:SWI48 0 "general_reg_operand"))
+		     (plus:<DWI>
+		       (match_operator:<DWI> 4 "ix86_carry_flag_operator"
+			 [(match_operand 3 "flags_reg_operand") (const_int 0)])
+		       (zero_extend:<DWI>
+			 (match_operand:SWI48 2 "general_reg_operand")))))
+	      (set (match_dup 0)
+		   (minus:SWI48
+		     (minus:SWI48
+		       (match_dup 0)
+		       (match_operator:SWI48 5 "ix86_carry_flag_operator"
+			 [(match_dup 3) (const_int 0)]))
+		     (match_dup 2)))])
+   (set (match_operand:SWI48 1 "memory_operand") (match_dup 0))]
+  "(TARGET_READ_MODIFY_WRITE || optimize_insn_for_size_p ())
+   && peep2_reg_dead_p (4, operands[0])
+   && peep2_reg_dead_p (3, operands[2])
+   && !reg_overlap_mentioned_p (operands[0], operands[1])
+   && !reg_overlap_mentioned_p (operands[2], operands[1])
+   && !reg_overlap_mentioned_p (operands[6], operands[9])
+   && (rtx_equal_p (operands[6], operands[0])
+       ? (rtx_equal_p (operands[7], operands[1])
+	  && rtx_equal_p (operands[8], operands[2]))
+       : (rtx_equal_p (operands[8], operands[0])
+	  && rtx_equal_p (operands[9], operands[1])
+	  && rtx_equal_p (operands[6], operands[2])))"
+  [(set (match_dup 0) (match_dup 9))
+   (parallel [(set (reg:CCC FLAGS_REG)
+		   (compare:CCC
+		     (zero_extend:<DWI> (match_dup 1))
+		     (plus:<DWI> (match_op_dup 4
+				   [(match_dup 3) (const_int 0)])
+				 (zero_extend:<DWI> (match_dup 0)))))
+	      (set (match_dup 1)
+		   (minus:SWI48 (minus:SWI48 (match_dup 1)
+					     (match_op_dup 5
+					       [(match_dup 3) (const_int 0)]))
+				(match_dup 0)))])]
+{
+  if (!rtx_equal_p (operands[6], operands[0]))
+    operands[9] = operands[7];
+})
+
+(define_peephole2
+  [(set (match_operand:SWI48 6 "general_reg_operand")
+	(match_operand:SWI48 7 "memory_operand"))
+   (set (match_operand:SWI48 8 "general_reg_operand")
+	(match_operand:SWI48 9 "memory_operand"))
+   (parallel [(set (reg:CCC FLAGS_REG)
+		   (compare:CCC
+		     (zero_extend:<DWI>
+		       (match_operand:SWI48 0 "general_reg_operand"))
+		     (plus:<DWI>
+		       (match_operator:<DWI> 4 "ix86_carry_flag_operator"
+			 [(match_operand 3 "flags_reg_operand") (const_int 0)])
+		       (zero_extend:<DWI>
+			 (match_operand:SWI48 2 "general_reg_operand")))))
+	      (set (match_dup 0)
+		   (minus:SWI48
+		     (minus:SWI48
+		       (match_dup 0)
+		       (match_operator:SWI48 5 "ix86_carry_flag_operator"
+			 [(match_dup 3) (const_int 0)]))
+		     (match_dup 2)))])
+   (set (match_operand:QI 10 "general_reg_operand")
+	(ltu:QI (reg:CCC FLAGS_REG) (const_int 0)))
+   (set (match_operand:SWI48 11 "general_reg_operand")
+	(zero_extend:SWI48 (match_dup 10)))
+   (set (match_operand:SWI48 1 "memory_operand") (match_dup 0))]
+  "(TARGET_READ_MODIFY_WRITE || optimize_insn_for_size_p ())
+   && peep2_reg_dead_p (6, operands[0])
+   && peep2_reg_dead_p (3, operands[2])
+   && !reg_overlap_mentioned_p (operands[0], operands[1])
+   && !reg_overlap_mentioned_p (operands[2], operands[1])
+   && !reg_overlap_mentioned_p (operands[6], operands[9])
+   && !reg_overlap_mentioned_p (operands[0], operands[10])
+   && !reg_overlap_mentioned_p (operands[10], operands[1])
+   && !reg_overlap_mentioned_p (operands[0], operands[11])
+   && !reg_overlap_mentioned_p (operands[11], operands[1])
+   && (rtx_equal_p (operands[6], operands[0])
+       ? (rtx_equal_p (operands[7], operands[1])
+	  && rtx_equal_p (operands[8], operands[2]))
+       : (rtx_equal_p (operands[8], operands[0])
+	  && rtx_equal_p (operands[9], operands[1])
+	  && rtx_equal_p (operands[6], operands[2])))"
+  [(set (match_dup 0) (match_dup 9))
+   (parallel [(set (reg:CCC FLAGS_REG)
+		   (compare:CCC
+		     (zero_extend:<DWI> (match_dup 1))
+		     (plus:<DWI> (match_op_dup 4
+				   [(match_dup 3) (const_int 0)])
+				 (zero_extend:<DWI> (match_dup 0)))))
+	      (set (match_dup 1)
+		   (minus:SWI48 (minus:SWI48 (match_dup 1)
+					     (match_op_dup 5
+					       [(match_dup 3) (const_int 0)]))
+				(match_dup 0)))])
+   (set (match_dup 10) (ltu:QI (reg:CCC FLAGS_REG) (const_int 0)))
+   (set (match_dup 11) (zero_extend:SWI48 (match_dup 10)))]
+{
+  if (!rtx_equal_p (operands[6], operands[0]))
+    operands[9] = operands[7];
+})
+
 (define_expand "subborrow<mode>_0"
   [(parallel
      [(set (reg:CC FLAGS_REG)