i386: Optimization for APX NDD is always zero-uppered for ADD

gcc/ChangeLog:

	PR target/113729
	* config/i386/i386.md (*addqi_1_zext<mode><nf_name>): New
	define.
	(*addhi_1_zext<mode><nf_name>): Ditto.

gcc/testsuite/ChangeLog:

	* gcc.target/i386/pr113729.c: New test.

2 files changed, 107 insertions, 0 deletions

diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md
index 1a6188f..5f23727 100644
--- a/gcc/config/i386/i386.md
+++ b/gcc/config/i386/i386.md
@@ -6572,6 +6572,86 @@
    (set_attr "has_nf" "1")
    (set_attr "mode" "<MODE>")])
 
+;; For APX instruction with an NDD, the destination GPR  will get the
+;; instruction’s result in bits [OSIZE-1:0] and, if OSIZE < 64b, have
+;; its upper bits [63:OSIZE] zeroed.
+
+(define_insn "*addqi_1_zext<mode><nf_name>"
+  [(set (match_operand:SWI248x 0 "register_operand" "=r,r")
+	(zero_extend:SWI248x
+	  (plus:QI (match_operand:QI 1 "nonimmediate_operand" "%rm,r")
+		   (match_operand:QI 2 "general_operand" "rn,m"))))]
+  "TARGET_APX_NDD && <nf_condition>
+   && !(MEM_P (operands[1]) && MEM_P (operands[2]))"
+{
+  switch (get_attr_type (insn))
+    {
+    case TYPE_INCDEC:
+      if (operands[2] == const1_rtx)
+	return "<nf_prefix>inc{b}\t{%1, %b0|%b0, %1}";
+      else
+	{
+	  gcc_assert (operands[2] == constm1_rtx);
+	  return "<nf_prefix>dec{b}\t{%1, %b0|%b0, %1}";
+	}
+
+    default:
+      if (x86_maybe_negate_const_int (&operands[2], QImode))
+	return "<nf_prefix>sub{b}\t{%2, %1, %b0|%b0, %1, %2}";
+      return "<nf_prefix>add{b}\t{%2, %1, %b0|%b0, %1, %2}";
+    }
+}
+  [(set (attr "type")
+     (cond [(match_operand:QI 2 "incdec_operand")
+	      (const_string "incdec")
+	   ]
+	   (const_string "alu")))
+   (set (attr "length_immediate")
+      (if_then_else
+	(and (eq_attr "type" "alu") (match_operand 2 "const128_operand"))
+	(const_string "1")
+	(const_string "*")))
+   (set_attr "has_nf" "1")
+   (set_attr "mode" "QI")])
+
+(define_insn "*addhi_1_zext<mode><nf_name>"
+  [(set (match_operand:SWI48x 0 "register_operand" "=r,r")
+	(zero_extend:SWI48x
+	  (plus:HI (match_operand:HI 1 "nonimmediate_operand" "%rm,r")
+		   (match_operand:HI 2 "general_operand" "rn,m"))))]
+  "TARGET_APX_NDD && <nf_condition>
+   && !(MEM_P (operands[1]) && MEM_P (operands[2]))"
+{
+  switch (get_attr_type (insn))
+    {
+    case TYPE_INCDEC:
+      if (operands[2] == const1_rtx)
+	return "<nf_prefix>inc{w}\t{%1, %w0|%w0, %1}";
+      else
+	{
+	  gcc_assert (operands[2] == constm1_rtx);
+	  return "<nf_prefix>dec{w}\t{%1, %w0|%w0, %1}";
+	}
+
+    default:
+      if (x86_maybe_negate_const_int (&operands[2], HImode))
+	return "<nf_prefix>sub{w}\t{%2, %1, %w0|%w0, %1, %2}";
+      return "<nf_prefix>add{w}\t{%2, %1, %w0|%w0, %1, %2}";
+    }
+}
+  [(set (attr "type")
+     (cond [(match_operand:QI 2 "incdec_operand")
+	      (const_string "incdec")
+	   ]
+	   (const_string "alu")))
+   (set (attr "length_immediate")
+      (if_then_else
+	(and (eq_attr "type" "alu") (match_operand 2 "const128_operand"))
+	(const_string "1")
+	(const_string "*")))
+   (set_attr "has_nf" "1")
+   (set_attr "mode" "HI")])
+
 ;; It may seem that nonimmediate operand is proper one for operand 1.
 ;; The addsi_1 pattern allows nonimmediate operand at that place and
 ;; we take care in ix86_binary_operator_ok to not allow two memory
diff --git a/gcc/testsuite/gcc.target/i386/pr113729.c b/gcc/testsuite/gcc.target/i386/pr113729.c
new file mode 100644
index 0000000..34518a5
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr113729.c
@@ -0,0 +1,27 @@
+/* { dg-do compile { target { ! ia32 } } } */
+/* { dg-options "-mapx-features=ndd -march=x86-64 -O2" } */
+/* { dg-final { scan-assembler-not "movz"} } */
+
+#include <stdint.h>
+
+#define F(TYPE1, TYPE2, OP_NAME, OP)                \
+TYPE1                                               \
+__attribute__ ((noipa))                             \
+f_##OP_NAME##_##TYPE2##_##TYPE1 (unsigned TYPE2 b)  \
+{                                                   \
+  return (unsigned TYPE2) (200 OP b);               \
+}                                                   \
+TYPE1                                               \
+__attribute__ ((noipa))                             \
+f1_##OP_NAME##_##TYPE2##_##TYPE1                    \
+(unsigned TYPE2 a, unsigned TYPE2 b)                \
+{                                                   \
+  return (unsigned TYPE2) (a OP b);                 \
+}					
+
+/* addqi_1_zext<mode> */
+F (short, char, add, +)
+F (int, char, add, +)
+F (int64_t, char, add, +)
+F (int, short, add, +)
+F (int64_t, short, add, +)