[AArch64][SVE2] Shift-Right Accumulate combine patterns

This patch adds combining support for SVE2's shift-right accumulate
instructions.

2019-09-27  Yuliang Wang  <yuliang.wang@arm.com>

gcc/
	* config/aarch64/aarch64-sve2.md (aarch64_sve2_sra<mode>):
	New combine pattern.

gcc/testsuite/
	* gcc.target/aarch64/sve2/shracc_1.c: New test.

From-SVN: r276174

gcc/ChangeLog

+2019-09-27  Yuliang Wang  <yuliang.wang@arm.com>
+
+	* config/aarch64/aarch64-sve2.md (aarch64_sve2_sra<mode>):
+	New combine pattern.
+
 2019-09-26  Max Filippov  <jcmvbkbc@gmail.com>
 
 	* config/xtensa/xtensa.c (hwloop_optimize): Insert zero overhead

gcc/config/aarch64/aarch64-sve2.md

@@ -123,3 +123,22 @@
   }
 )
 
+;; Unpredicated signed / unsigned shift-right accumulate.
+(define_insn_and_rewrite "*aarch64_sve2_sra<mode>"
+  [(set (match_operand:SVE_I 0 "register_operand" "=w")
+	(plus:SVE_I
+	  (unspec:SVE_I
+	    [(match_operand 4)
+	     (SHIFTRT:SVE_I
+	       (match_operand:SVE_I 2 "register_operand" "w")
+	       (match_operand:SVE_I 3 "aarch64_simd_rshift_imm" "Dr"))]
+	    UNSPEC_PRED_X)
+	 (match_operand:SVE_I 1 "register_operand" "0")))]
+  "TARGET_SVE2"
+  "<sra_op>sra\t%0.<Vetype>, %2.<Vetype>, #%3"
+  "&& !CONSTANT_P (operands[4])"
+  {
+    operands[4] = CONSTM1_RTX (<VPRED>mode);
+  }
+)
+

gcc/testsuite/ChangeLog

+2019-09-27  Yuliang Wang  <yuliang.wang@arm.com>
+
+	* gcc.target/aarch64/sve2/shracc_1.c: New test.
+
 2019-09-26  Eric Botcazou  <ebotcazou@adacore.com>
 
 	* gcc.dg/cpp/ucs.c: Add test for new warning and adjust.

gcc/testsuite/gcc.target/aarch64/sve2/shracc_1.c

+/* { dg-do compile } */
+/* { dg-options "-O2 -ftree-vectorize -fdump-tree-vect-details --save-temps" } */
+
+#include <stdint.h>
+
+#define SHRACC(TYPE,SHIFT)			\
+void __attribute__ ((noinline, noclone))	\
+f_##TYPE##_##SHIFT				\
+  (TYPE *restrict a, TYPE *restrict b, int n)	\
+{						\
+  for (int i = 0; i < n; i++)			\
+    a[i] += b[i] >> (SHIFT);			\
+}
+
+SHRACC (int8_t, 5);
+SHRACC (int16_t, 14);
+SHRACC (int32_t, 19);
+SHRACC (int64_t, 27);
+
+SHRACC (uint8_t, 2);
+SHRACC (uint16_t, 6);
+SHRACC (uint32_t, 24);
+SHRACC (uint64_t, 53);
+
+/* { dg-final { scan-tree-dump-times "vectorized 1 loops in function" 8 "vect" } } */
+
+/* { dg-final { scan-assembler-not {\tasr\t} } } */
+/* { dg-final { scan-assembler-not {\tlsr\t} } } */
+/* { dg-final { scan-assembler-not {\tadd\t} } } */
+
+/* { dg-final { scan-assembler-times {\tssra\tz[0-9]+\.b, z[0-9]+\.b, #5\n} 1 } } */
+/* { dg-final { scan-assembler-times {\tssra\tz[0-9]+\.h, z[0-9]+\.h, #14\n} 1 } } */
+/* { dg-final { scan-assembler-times {\tssra\tz[0-9]+\.s, z[0-9]+\.s, #19\n} 1 } } */
+/* { dg-final { scan-assembler-times {\tssra\tz[0-9]+\.d, z[0-9]+\.d, #27\n} 1 } } */
+
+/* { dg-final { scan-assembler-times {\tusra\tz[0-9]+\.b, z[0-9]+\.b, #2\n} 1 } } */
+/* { dg-final { scan-assembler-times {\tusra\tz[0-9]+\.h, z[0-9]+\.h, #6\n} 1 } } */
+/* { dg-final { scan-assembler-times {\tusra\tz[0-9]+\.s, z[0-9]+\.s, #24\n} 1 } } */
+/* { dg-final { scan-assembler-times {\tusra\tz[0-9]+\.d, z[0-9]+\.d, #53\n} 1 } } */