RISC-V: Support RVV VFSQRT rounding mode intrinsic API

This patch would like to support the rounding mode API for the
VFSQRT as the below samples.

* __riscv_vfsqrt_v_f32m1_rm
* __riscv_vfsqrt_v_f32m1_rm_m

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

	* config/riscv/riscv-vector-builtins-bases.cc
	(class unop_frm): New class for frm.
	(vfsqrt_frm_obj): New declaration.
	(BASE): Ditto.
	* config/riscv/riscv-vector-builtins-bases.h: Ditto.
	* config/riscv/riscv-vector-builtins-functions.def
	(vfsqrt_frm): New intrinsic function definition.

gcc/testsuite/ChangeLog:

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

diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc
index b458560a0407de325b1d32d65d3d8dd41b051607..2074dac0f167a04a44d3bee101c9b99b44b6506d 100644 (file)
--- a/gcc/config/riscv/riscv-vector-builtins-bases.cc
+++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc
@@ -631,6 +631,21 @@ public:
   }
 };
 
+/* Implements below instructions for frm
+   - vfsqrt
+*/
+template<rtx_code CODE>
+class unop_frm : public function_base
+{
+public:
+  bool has_rounding_mode_operand_p () const override { return true; }
+
+  rtx expand (function_expander &e) const override
+  {
+    return e.use_exact_insn (code_for_pred (CODE, e.vector_mode ()));
+  }
+};
+
 /* Implements vrsub.  */
 class vrsub : public function_base
 {
@@ -2415,6 +2430,7 @@ static CONSTEXPR const vfwmsac_frm vfwmsac_frm_obj;
 static CONSTEXPR const vfwnmsac vfwnmsac_obj;
 static CONSTEXPR const vfwnmsac_frm vfwnmsac_frm_obj;
 static CONSTEXPR const unop<SQRT> vfsqrt_obj;
+static CONSTEXPR const unop_frm<SQRT> vfsqrt_frm_obj;
 static CONSTEXPR const float_misc<UNSPEC_VFRSQRT7> vfrsqrt7_obj;
 static CONSTEXPR const float_misc<UNSPEC_VFREC7> vfrec7_obj;
 static CONSTEXPR const binop<SMIN> vfmin_obj;
@@ -2662,6 +2678,7 @@ BASE (vfwmsac_frm)
 BASE (vfwnmsac)
 BASE (vfwnmsac_frm)
 BASE (vfsqrt)
+BASE (vfsqrt_frm)
 BASE (vfrsqrt7)
 BASE (vfrec7)
 BASE (vfmin)

diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h
index 85e8b9a37691d36344df249fa1591dffc9e8b59f..5c91381bd4cdf35ff7f29e630cf9e4a92ca70c6d 100644 (file)
--- a/gcc/config/riscv/riscv-vector-builtins-bases.h
+++ b/gcc/config/riscv/riscv-vector-builtins-bases.h
@@ -184,6 +184,7 @@ extern const function_base *const vfwmsac_frm;
 extern const function_base *const vfwnmsac;
 extern const function_base *const vfwnmsac_frm;
 extern const function_base *const vfsqrt;
+extern const function_base *const vfsqrt_frm;
 extern const function_base *const vfrsqrt7;
 extern const function_base *const vfrec7;
 extern const function_base *const vfmin;

diff --git a/gcc/config/riscv/riscv-vector-builtins-functions.def b/gcc/config/riscv/riscv-vector-builtins-functions.def
index 7e2a4ab2969600668a74c4ec8fa74f3d0d29088f..a821aca6a4bc818fc3a8af9b4a904abd18f1bed4 100644 (file)
--- a/gcc/config/riscv/riscv-vector-builtins-functions.def
+++ b/gcc/config/riscv/riscv-vector-builtins-functions.def
@@ -388,6 +388,8 @@ DEF_RVV_FUNCTION (vfwnmsac_frm, alu_frm, full_preds, f_wwfv_ops)
 // 13.8. Vector Floating-Point Square-Root Instruction
 DEF_RVV_FUNCTION (vfsqrt, alu, full_preds, f_v_ops)
 
+DEF_RVV_FUNCTION (vfsqrt_frm, alu_frm, full_preds, f_v_ops)
+
 // 13.9. Vector Floating-Point Reciprocal Square-Root Estimate Instruction
 DEF_RVV_FUNCTION (vfrsqrt7, alu, full_preds, f_v_ops)
 

diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/float-point-sqrt.c b/gcc/testsuite/gcc.target/riscv/rvv/base/float-point-sqrt.c
new file mode 100644 (file)
index 0000000..afd1fb2
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/base/float-point-sqrt.c
@@ -0,0 +1,31 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv -mabi=lp64 -O3 -Wno-psabi" } */
+
+#include "riscv_vector.h"
+
+typedef float float32_t;
+
+vfloat32m1_t
+test_riscv_vfsqrt_vv_f32m1_rm (vfloat32m1_t op1, size_t vl) {
+  return __riscv_vfsqrt_v_f32m1_rm (op1, 0, vl);
+}
+
+vfloat32m1_t
+test_vfsqrt_vv_f32m1_rm_m (vbool32_t mask, vfloat32m1_t op1, size_t vl) {
+  return __riscv_vfsqrt_v_f32m1_rm_m (mask, op1, 1, vl);
+}
+
+vfloat32m1_t
+test_riscv_vfsqrt_vv_f32m1 (vfloat32m1_t op1, size_t vl) {
+  return __riscv_vfsqrt_v_f32m1 (op1, vl);
+}
+
+vfloat32m1_t
+test_vfsqrt_vv_f32m1_m (vbool32_t mask, vfloat32m1_t op1, size_t vl) {
+  return __riscv_vfsqrt_v_f32m1_m (mask, op1, vl);
+}
+
+/* { dg-final { scan-assembler-times {vfsqrt\.v\s+v[0-9]+,\s*v[0-9]+} 4 } } */
+/* { dg-final { scan-assembler-times {frrm\s+[axs][0-9]+} 2 } } */
+/* { dg-final { scan-assembler-times {fsrm\s+[axs][0-9]+} 2 } } */
+/* { dg-final { scan-assembler-times {fsrmi\s+[01234]} 2 } } */