RISC-V: Support RVV VFNCVT.F.{X|XU|F}.W rounding mode intrinsic API

This patch would like to support the rounding mode API for the
VFNCVT.F.{X|XU|F}.W as the below samples.

* __riscv_vfncvt_f_x_w_f32m1_rm
* __riscv_vfncvt_f_x_w_f32m1_rm_m
* __riscv_vfncvt_f_xu_w_f32m1_rm
* __riscv_vfncvt_f_xu_w_f32m1_rm_m
* __riscv_vfncvt_f_f_w_f32m1_rm
* __riscv_vfncvt_f_f_w_f32m1_rm_m

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

	* config/riscv/riscv-vector-builtins-bases.cc
	(class vfncvt_f): Add frm_op_type template arg.
	(vfncvt_f_frm_obj): New declaration.
	(BASE): Ditto.
	* config/riscv/riscv-vector-builtins-bases.h: Ditto.
	* config/riscv/riscv-vector-builtins-functions.def
	(vfncvt_f_frm): New intrinsic function def.

gcc/testsuite/ChangeLog:

	* gcc.target/riscv/rvv/base/float-point-ncvt-f.c: New test.

diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc
index acadec2afca759da708b704d2237a22334f12afb..ad04647f9ba500112ebc1207a822453d31e5b4a2 100644 (file)
--- a/gcc/config/riscv/riscv-vector-builtins-bases.cc
+++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc
@@ -1786,9 +1786,15 @@ public:
   }
 };
 
+template<enum frm_op_type FRM_OP = NO_FRM>
 class vfncvt_f : public function_base
 {
 public:
+  bool has_rounding_mode_operand_p () const override
+  {
+    return FRM_OP == HAS_FRM;
+  }
+
   rtx expand (function_expander &e) const override
   {
     if (e.op_info->op == OP_TYPE_f_w)
@@ -2512,7 +2518,8 @@ static CONSTEXPR const vfncvt_x<UNSPEC_UNSIGNED_VFCVT> vfncvt_xu_obj;
 static CONSTEXPR const vfncvt_x<UNSPEC_UNSIGNED_VFCVT, HAS_FRM> vfncvt_xu_frm_obj;
 static CONSTEXPR const vfncvt_rtz_x<FIX> vfncvt_rtz_x_obj;
 static CONSTEXPR const vfncvt_rtz_x<UNSIGNED_FIX> vfncvt_rtz_xu_obj;
-static CONSTEXPR const vfncvt_f vfncvt_f_obj;
+static CONSTEXPR const vfncvt_f<NO_FRM> vfncvt_f_obj;
+static CONSTEXPR const vfncvt_f<HAS_FRM> vfncvt_f_frm_obj;
 static CONSTEXPR const vfncvt_rod_f vfncvt_rod_f_obj;
 static CONSTEXPR const reducop<PLUS> vredsum_obj;
 static CONSTEXPR const reducop<UMAX> vredmaxu_obj;
@@ -2769,6 +2776,7 @@ BASE (vfncvt_xu_frm)
 BASE (vfncvt_rtz_x)
 BASE (vfncvt_rtz_xu)
 BASE (vfncvt_f)
+BASE (vfncvt_f_frm)
 BASE (vfncvt_rod_f)
 BASE (vredsum)
 BASE (vredmaxu)

diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h
index 9bd09a4196035da6c944b7243378bad20b2bc71b..c8c649c4bb0d1fb5e16ae0604f3e849abad4ca5e 100644 (file)
--- a/gcc/config/riscv/riscv-vector-builtins-bases.h
+++ b/gcc/config/riscv/riscv-vector-builtins-bases.h
@@ -226,6 +226,7 @@ extern const function_base *const vfncvt_xu_frm;
 extern const function_base *const vfncvt_rtz_x;
 extern const function_base *const vfncvt_rtz_xu;
 extern const function_base *const vfncvt_f;
+extern const function_base *const vfncvt_f_frm;
 extern const function_base *const vfncvt_rod_f;
 extern const function_base *const vredsum;
 extern const function_base *const vredmaxu;

diff --git a/gcc/config/riscv/riscv-vector-builtins-functions.def b/gcc/config/riscv/riscv-vector-builtins-functions.def
index 1e0e989fc2adffe7688593e744e916461fbba95c..cfbc125dcd8ba45d01fcc86e23134ff0ccfa3873 100644 (file)
--- a/gcc/config/riscv/riscv-vector-builtins-functions.def
+++ b/gcc/config/riscv/riscv-vector-builtins-functions.def
@@ -474,6 +474,9 @@ DEF_RVV_FUNCTION (vfncvt_rod_f, narrow_alu, full_preds, f_to_nf_f_w_ops)
 
 DEF_RVV_FUNCTION (vfncvt_x_frm, narrow_alu_frm, full_preds, f_to_ni_f_w_ops)
 DEF_RVV_FUNCTION (vfncvt_xu_frm, narrow_alu_frm, full_preds, f_to_nu_f_w_ops)
+DEF_RVV_FUNCTION (vfncvt_f_frm, narrow_alu_frm, full_preds, i_to_nf_x_w_ops)
+DEF_RVV_FUNCTION (vfncvt_f_frm, narrow_alu_frm, full_preds, u_to_nf_xu_w_ops)
+DEF_RVV_FUNCTION (vfncvt_f_frm, narrow_alu_frm, full_preds, f_to_nf_f_w_ops)
 
 /* 14. Vector Reduction Operations.  */
 

diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/float-point-ncvt-f.c b/gcc/testsuite/gcc.target/riscv/rvv/base/float-point-ncvt-f.c
new file mode 100644 (file)
index 0000000..d6d4be5
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/base/float-point-ncvt-f.c
@@ -0,0 +1,69 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv -mabi=lp64 -O3 -Wno-psabi" } */
+
+#include "riscv_vector.h"
+
+vfloat32m1_t
+test_riscv_vfncvt_f_x_w_f32m1_rm (vint64m2_t op1, size_t vl) {
+  return __riscv_vfncvt_f_x_w_f32m1_rm (op1, 0, vl);
+}
+
+vfloat32m1_t
+test_vfncvt_f_x_w_f32m1_rm_m (vbool32_t mask, vint64m2_t op1, size_t vl) {
+  return __riscv_vfncvt_f_x_w_f32m1_rm_m (mask, op1, 1, vl);
+}
+
+vfloat32m1_t
+test_riscv_vfncvt_f_xu_w_f32m1_rm (vuint64m2_t op1, size_t vl) {
+  return __riscv_vfncvt_f_xu_w_f32m1_rm (op1, 0, vl);
+}
+
+vfloat32m1_t
+test_vfncvt_f_xu_w_f32m1_rm_m (vbool32_t mask, vuint64m2_t op1, size_t vl) {
+  return __riscv_vfncvt_f_xu_w_f32m1_rm_m (mask, op1, 1, vl);
+}
+
+vfloat32m1_t
+test_riscv_vfncvt_f_f_w_f32m1_rm (vfloat64m2_t op1, size_t vl) {
+  return __riscv_vfncvt_f_f_w_f32m1_rm (op1, 0, vl);
+}
+
+vfloat32m1_t
+test_vfncvt_f_f_w_f32m1_rm_m (vbool32_t mask, vfloat64m2_t op1, size_t vl) {
+  return __riscv_vfncvt_f_f_w_f32m1_rm_m (mask, op1, 1, vl);
+}
+
+vfloat32m1_t
+test_riscv_vfncvt_f_x_w_f32m1 (vint64m2_t op1, size_t vl) {
+  return __riscv_vfncvt_f_x_w_f32m1 (op1, vl);
+}
+
+vfloat32m1_t
+test_vfncvt_f_x_w_f32m1_m (vbool32_t mask, vint64m2_t op1, size_t vl) {
+  return __riscv_vfncvt_f_x_w_f32m1_m (mask, op1, vl);
+}
+
+vfloat32m1_t
+test_riscv_vfncvt_f_xu_w_f32m1 (vuint64m2_t op1, size_t vl) {
+  return __riscv_vfncvt_f_xu_w_f32m1 (op1, vl);
+}
+
+vfloat32m1_t
+test_vfncvt_f_xu_w_f32m1_m (vbool32_t mask, vuint64m2_t op1, size_t vl) {
+  return __riscv_vfncvt_f_xu_w_f32m1_m (mask, op1, vl);
+}
+
+vfloat32m1_t
+test_riscv_vfncvt_f_f_w_f32m1 (vfloat64m2_t op1, size_t vl) {
+  return __riscv_vfncvt_f_f_w_f32m1 (op1, vl);
+}
+
+vfloat32m1_t
+test_vfncvt_f_f_w_f32m1_m (vbool32_t mask, vfloat64m2_t op1, size_t vl) {
+  return __riscv_vfncvt_f_f_w_f32m1_m (mask, op1, vl);
+}
+
+/* { dg-final { scan-assembler-times {vfncvt\.f\.[xuf]+\.w\s+v[0-9]+,\s*v[0-9]+} 12 } } */
+/* { dg-final { scan-assembler-times {frrm\s+[axs][0-9]+} 6 } } */
+/* { dg-final { scan-assembler-times {fsrm\s+[axs][0-9]+} 6 } } */
+/* { dg-final { scan-assembler-times {fsrmi\s+[01234]} 6 } } */