RISC-V: Support RVV VFWADD rounding mode intrinsic API

This patch would like to support the rounding mode API for the VFWADD
VFSUB and VFRSUB as below samples.

* __riscv_vfwadd_vv_f64m2_rm
* __riscv_vfwadd_vv_f64m2_rm_m
* __riscv_vfwadd_vf_f64m2_rm
* __riscv_vfwadd_vf_f64m2_rm_m
* __riscv_vfwadd_wv_f64m2_rm
* __riscv_vfwadd_wv_f64m2_rm_m
* __riscv_vfwadd_wf_f64m2_rm
* __riscv_vfwadd_wf_f64m2_rm_m

Signed-off-by: Pan Li <pan2.li@intel.com>

gcc/ChangeLog:

	* config/riscv/riscv-vector-builtins-bases.cc
	(class widen_binop_frm): New class for binop frm.
	(BASE): Add vfwadd_frm.
	* config/riscv/riscv-vector-builtins-bases.h: New declaration.
	* config/riscv/riscv-vector-builtins-functions.def
	(vfwadd_frm): New function definition.
	* config/riscv/riscv-vector-builtins-shapes.cc
	(BASE_NAME_MAX_LEN): New macro.
	(struct alu_frm_def): Leverage new base class.
	(struct build_frm_base): New build base for frm.
	(struct widen_alu_frm_def): New struct for widen alu frm.
	(SHAPE): Add widen_alu_frm shape.
	* config/riscv/riscv-vector-builtins-shapes.h: New declaration.
	* config/riscv/vector.md (frm_mode): Add vfwalu type.

gcc/testsuite/ChangeLog:

	* gcc.target/riscv/rvv/base/float-point-widening-add.c: New test.

gcc/config/riscv/riscv-vector-builtins-bases.cc

@@ -315,6 +315,41 @@ public:
   }
 };
 
+/* Implements below instructions for frm
+   - vfwadd
+*/
+template<rtx_code CODE>
+class widen_binop_frm : public function_base
+{
+public:
+  bool has_rounding_mode_operand_p () const override { return true; }
+
+  rtx expand (function_expander &e) const override
+  {
+    switch (e.op_info->op)
+      {
+      case OP_TYPE_vv:
+	return e.use_exact_insn (
+	  code_for_pred_dual_widen (CODE, e.vector_mode ()));
+      case OP_TYPE_vf:
+	return e.use_exact_insn (
+	  code_for_pred_dual_widen_scalar (CODE, e.vector_mode ()));
+      case OP_TYPE_wv:
+	if (CODE == PLUS)
+	  return e.use_exact_insn (
+	    code_for_pred_single_widen_add (e.vector_mode ()));
+	else
+	  return e.use_exact_insn (
+	    code_for_pred_single_widen_sub (e.vector_mode ()));
+      case OP_TYPE_wf:
+	return e.use_exact_insn (
+	  code_for_pred_single_widen_scalar (CODE, e.vector_mode ()));
+      default:
+	gcc_unreachable ();
+      }
+  }
+};
+
 /* Implements vrsub.  */
 class vrsub : public function_base
 {
@@ -2063,6 +2098,7 @@ static CONSTEXPR const binop_frm<MINUS> vfsub_frm_obj;
 static CONSTEXPR const reverse_binop<MINUS> vfrsub_obj;
 static CONSTEXPR const reverse_binop_frm<MINUS> vfrsub_frm_obj;
 static CONSTEXPR const widen_binop<PLUS> vfwadd_obj;
+static CONSTEXPR const widen_binop_frm<PLUS> vfwadd_frm_obj;
 static CONSTEXPR const widen_binop<MINUS> vfwsub_obj;
 static CONSTEXPR const binop<MULT> vfmul_obj;
 static CONSTEXPR const binop<DIV> vfdiv_obj;
@@ -2292,6 +2328,7 @@ BASE (vfsub_frm)
 BASE (vfrsub)
 BASE (vfrsub_frm)
 BASE (vfwadd)
+BASE (vfwadd_frm)
 BASE (vfwsub)
 BASE (vfmul)
 BASE (vfdiv)

gcc/config/riscv/riscv-vector-builtins-bases.h

@@ -148,6 +148,7 @@ extern const function_base *const vfsub_frm;
 extern const function_base *const vfrsub;
 extern const function_base *const vfrsub_frm;
 extern const function_base *const vfwadd;
+extern const function_base *const vfwadd_frm;
 extern const function_base *const vfwsub;
 extern const function_base *const vfmul;
 extern const function_base *const vfmul;

gcc/config/riscv/riscv-vector-builtins-functions.def

@@ -304,6 +304,10 @@ DEF_RVV_FUNCTION (vfwadd, widen_alu, full_preds, f_wwv_ops)
 DEF_RVV_FUNCTION (vfwadd, widen_alu, full_preds, f_wwf_ops)
 DEF_RVV_FUNCTION (vfwsub, widen_alu, full_preds, f_wwv_ops)
 DEF_RVV_FUNCTION (vfwsub, widen_alu, full_preds, f_wwf_ops)
+DEF_RVV_FUNCTION (vfwadd_frm, widen_alu_frm, full_preds, f_wvv_ops)
+DEF_RVV_FUNCTION (vfwadd_frm, widen_alu_frm, full_preds, f_wvf_ops)
+DEF_RVV_FUNCTION (vfwadd_frm, widen_alu_frm, full_preds, f_wwv_ops)
+DEF_RVV_FUNCTION (vfwadd_frm, widen_alu_frm, full_preds, f_wwf_ops)
 
 // 13.4. Vector Single-Width Floating-Point Multiply/Divide Instructions
 DEF_RVV_FUNCTION (vfmul, alu, full_preds, f_vvv_ops)

gcc/config/riscv/riscv-vector-builtins-shapes.cc

@@ -75,6 +75,8 @@ build_all (function_builder &b, const function_group_info &group)
   static CONSTEXPR const DEF##_def VAR##_obj; \
   namespace shapes { const function_shape *const VAR = &VAR##_obj; }
 
+#define BASE_NAME_MAX_LEN 16
+
 /* Base class for for build.  */
 struct build_base : public function_shape
 {
@@ -226,8 +228,8 @@ struct alu_def : public build_base
   }
 };
 
-/* alu_frm_def class.  */
-struct alu_frm_def : public build_base
+/* The base class for frm build.  */
+struct build_frm_base : public build_base
 {
   /* Normalize vf<op>_frm to vf<op>.  */
   static void normalize_base_name (char *to, const char *from, int limit)
@@ -241,10 +243,29 @@ struct alu_frm_def : public build_base
     to[limit - 1] = '\0';
   }
 
+  bool check (function_checker &c) const override
+  {
+    gcc_assert (c.any_type_float_p ());
+
+    /* Check whether rounding mode argument is a valid immediate.  */
+    if (c.base->has_rounding_mode_operand_p ())
+      {
+	unsigned int frm_num = c.arg_num () - 2;
+
+	return c.require_immediate (frm_num, FRM_STATIC_MIN, FRM_STATIC_MAX);
+      }
+
+    return true;
+  }
+};
+
+/* alu_frm_def class.  */
+struct alu_frm_def : public build_frm_base
+{
   char *get_name (function_builder &b, const function_instance &instance,
 		  bool overloaded_p) const override
   {
-    char base_name[16] = {};
+    char base_name[BASE_NAME_MAX_LEN] = {};
 
     /* Return nullptr if it can not be overloaded.  */
     if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred))
@@ -275,20 +296,40 @@ struct alu_frm_def : public build_base
 
     return b.finish_name ();
   }
+};
 
-  bool check (function_checker &c) const override
+/* widen_alu_frm_def class.  */
+struct widen_alu_frm_def : public build_frm_base
+{
+  char *get_name (function_builder &b, const function_instance &instance,
+		  bool overloaded_p) const override
   {
-    gcc_assert (c.any_type_float_p ());
+    char base_name[BASE_NAME_MAX_LEN] = {};
 
-    /* Check whether rounding mode argument is a valid immediate.  */
-    if (c.base->has_rounding_mode_operand_p ())
-      {
-	unsigned int frm_num = c.arg_num () - 2;
+    normalize_base_name (base_name, instance.base_name, sizeof (base_name));
 
-	return c.require_immediate (frm_num, FRM_STATIC_MIN, FRM_STATIC_MAX);
-      }
+    b.append_base_name (base_name);
 
-    return true;
+    /* vop<sew> --> vop<sew>_<op>.  */
+    b.append_name (operand_suffixes[instance.op_info->op]);
+
+    /* vop<sew>_<op> --> vop<sew>_<op>_<type>.  */
+    if (!overloaded_p)
+      b.append_name (type_suffixes[instance.type.index].vector);
+
+    /* According to rvv-intrinsic-doc, it does not add "_rm" suffix
+       for vop_rm C++ overloaded API.  */
+    if (!overloaded_p)
+      b.append_name ("_rm");
+
+    /* According to rvv-intrinsic-doc, it does not add "_m" suffix
+       for vop_m C++ overloaded API.  */
+    if (overloaded_p && instance.pred == PRED_TYPE_m)
+      return b.finish_name ();
+
+    b.append_name (predication_suffixes[instance.pred]);
+
+    return b.finish_name ();
   }
 };
 
@@ -811,6 +852,7 @@ SHAPE(indexed_loadstore, indexed_loadstore)
 SHAPE(alu, alu)
 SHAPE(alu_frm, alu_frm)
 SHAPE(widen_alu, widen_alu)
+SHAPE(widen_alu_frm, widen_alu_frm)
 SHAPE(no_mask_policy, no_mask_policy)
 SHAPE(return_mask, return_mask)
 SHAPE(narrow_alu, narrow_alu)

gcc/config/riscv/riscv-vector-builtins-shapes.h

@@ -31,6 +31,7 @@ extern const function_shape *const indexed_loadstore;
 extern const function_shape *const alu;
 extern const function_shape *const alu_frm;
 extern const function_shape *const widen_alu;
+extern const function_shape *const widen_alu_frm;
 extern const function_shape *const no_mask_policy;
 extern const function_shape *const return_mask;
 extern const function_shape *const narrow_alu;

gcc/config/riscv/vector.md

@@ -866,7 +866,7 @@
 
 ;; Defines rounding mode of an floating-point operation.
 (define_attr "frm_mode" "rne,rtz,rdn,rup,rmm,dyn,dyn_exit,dyn_call,none"
-  (cond [(eq_attr "type" "vfalu")
+  (cond [(eq_attr "type" "vfalu,vfwalu")
          (cond
 	   [(match_test "INTVAL (operands[9]) == riscv_vector::FRM_RNE")
 	    (const_string "rne")

gcc/testsuite/gcc.target/riscv/rvv/base/float-point-widening-add.c

@@ -0,0 +1,66 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv -mabi=lp64 -O3 -Wno-psabi" } */
+
+#include "riscv_vector.h"
+
+typedef float float32_t;
+
+vfloat64m2_t
+test_vfwadd_vv_f32m1_rm (vfloat32m1_t op1, vfloat32m1_t op2, size_t vl) {
+  return __riscv_vfwadd_vv_f64m2_rm (op1, op2, 0, vl);
+}
+
+vfloat64m2_t
+test_vfwadd_vv_f32m1_rm_m (vbool32_t mask, vfloat32m1_t op1, vfloat32m1_t op2,
+			   size_t vl) {
+  return __riscv_vfwadd_vv_f64m2_rm_m (mask, op1, op2, 1, vl);
+}
+
+vfloat64m2_t
+test_vfwadd_vf_f32m1_rm (vfloat32m1_t op1, float32_t op2, size_t vl) {
+  return __riscv_vfwadd_vf_f64m2_rm (op1, op2, 2, vl);
+}
+
+vfloat64m2_t
+test_vfwadd_vf_f32m1_rm_m (vbool32_t mask, vfloat32m1_t op1, float32_t op2,
+			   size_t vl) {
+  return __riscv_vfwadd_vf_f64m2_rm_m (mask, op1, op2, 3, vl);
+}
+
+vfloat64m2_t
+test_vfwadd_wv_f32m1_rm (vfloat64m2_t op1, vfloat32m1_t op2, size_t vl) {
+  return __riscv_vfwadd_wv_f64m2_rm (op1, op2, 0, vl);
+}
+
+vfloat64m2_t
+test_vfwadd_wv_f32m1_rm_m (vbool32_t mask, vfloat64m2_t op1, vfloat32m1_t op2,
+			   size_t vl) {
+  return __riscv_vfwadd_wv_f64m2_rm_m (mask, op1, op2, 1, vl);
+}
+
+vfloat64m2_t
+test_vfwadd_wf_f32m1_rm (vfloat64m2_t op1, float32_t op2, size_t vl) {
+  return __riscv_vfwadd_wf_f64m2_rm (op1, op2, 2, vl);
+}
+
+vfloat64m2_t
+test_vfwadd_wf_f32m1_rm_m (vbool32_t mask, vfloat64m2_t op1, float32_t op2,
+			   size_t vl) {
+  return __riscv_vfwadd_wf_f64m2_rm_m (mask, op1, op2, 3, vl);
+}
+
+vfloat64m2_t
+test_vfwadd_vv_f32m1 (vfloat32m1_t op1, vfloat32m1_t op2, size_t vl) {
+  return __riscv_vfwadd_vv_f64m2 (op1, op2, vl);
+}
+
+vfloat64m2_t
+test_vfwadd_vv_f32m1_m (vbool32_t mask, vfloat32m1_t op1, vfloat32m1_t op2,
+			   size_t vl) {
+  return __riscv_vfwadd_vv_f64m2_m (mask, op1, op2, vl);
+}
+
+/* { dg-final { scan-assembler-times {vfwadd\.[vw][vf]\s+v[0-9]+,\s*v[0-9]+,\s*[fav]+[0-9]+} 10 } } */
+/* { dg-final { scan-assembler-times {frrm\s+[axs][0-9]+} 8 } } */
+/* { dg-final { scan-assembler-times {fsrm\s+[axs][0-9]+} 8 } } */
+/* { dg-final { scan-assembler-times {fsrmi\s+[01234]} 8 } } */