LGTM
juzhe.zh...@rivai.ai
From: pan2.li
Date: 2023-08-12 12:48
To: gcc-patches
CC: juzhe.zhong; jeffreyalaw; pan2.li; yanzhang.wang; kito.cheng
Subject: [PATCH v1] RISC-V: Support RVV VFNMSUB rounding mode intrinsic API
From: Pan Li <pan2...@intel.com>
This patch would like to support the rounding mode API for the
VFNMSUB as the below samples.
* __riscv_vfnmsub_vv_f32m1_rm
* __riscv_vfnmsub_vv_f32m1_rm_m
* __riscv_vfnmsub_vf_f32m1_rm
* __riscv_vfnmsub_vf_f32m1_rm_m
Signed-off-by: Pan Li <pan2...@intel.com>
gcc/ChangeLog:
* config/riscv/riscv-vector-builtins-bases.cc
(class vfnmsub_frm): New class for vfnmsub frm.
(vfnmsub_frm): New declaration.
(BASE): Ditto.
* config/riscv/riscv-vector-builtins-bases.h: Ditto.
* config/riscv/riscv-vector-builtins-functions.def
(vfnmsub_frm): New function declaration.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/rvv/base/float-point-nmsub.c: New test.
---
.../riscv/riscv-vector-builtins-bases.cc | 25 ++++++++++
.../riscv/riscv-vector-builtins-bases.h | 1 +
.../riscv/riscv-vector-builtins-functions.def | 2 +
.../riscv/rvv/base/float-point-nmsub.c | 47 +++++++++++++++++++
4 files changed, 75 insertions(+)
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/base/float-point-nmsub.c
diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc
b/gcc/config/riscv/riscv-vector-builtins-bases.cc
index 381bc72c784..e14e9aa7809 100644
--- a/gcc/config/riscv/riscv-vector-builtins-bases.cc
+++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc
@@ -516,6 +516,29 @@ public:
}
};
+/* Implements below instructions for frm
+ - vfnmsub
+*/
+class vfnmsub_frm : public function_base
+{
+public:
+ bool has_rounding_mode_operand_p () const override { return true; }
+
+ bool has_merge_operand_p () const override { return false; }
+
+ rtx expand (function_expander &e) const override
+ {
+ if (e.op_info->op == OP_TYPE_vf)
+ return e.use_ternop_insn (
+ false, code_for_pred_mul_neg_scalar (PLUS, e.vector_mode ()));
+ if (e.op_info->op == OP_TYPE_vv)
+ return e.use_ternop_insn (
+ false, code_for_pred_mul_neg (PLUS, e.vector_mode ()));
+
+ gcc_unreachable ();
+ }
+};
+
/* Implements vrsub. */
class vrsub : public function_base
{
@@ -2282,6 +2305,7 @@ static CONSTEXPR const vfnmsac_frm vfnmsac_frm_obj;
static CONSTEXPR const vfmadd vfmadd_obj;
static CONSTEXPR const vfmadd_frm vfmadd_frm_obj;
static CONSTEXPR const vfnmsub vfnmsub_obj;
+static CONSTEXPR const vfnmsub_frm vfnmsub_frm_obj;
static CONSTEXPR const vfnmacc vfnmacc_obj;
static CONSTEXPR const vfnmacc_frm vfnmacc_frm_obj;
static CONSTEXPR const vfmsac vfmsac_obj;
@@ -2524,6 +2548,7 @@ BASE (vfnmsac_frm)
BASE (vfmadd)
BASE (vfmadd_frm)
BASE (vfnmsub)
+BASE (vfnmsub_frm)
BASE (vfnmacc)
BASE (vfnmacc_frm)
BASE (vfmsac)
diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h
b/gcc/config/riscv/riscv-vector-builtins-bases.h
index 99cfbfd78c8..e60cebab4ae 100644
--- a/gcc/config/riscv/riscv-vector-builtins-bases.h
+++ b/gcc/config/riscv/riscv-vector-builtins-bases.h
@@ -166,6 +166,7 @@ extern const function_base *const vfnmsac_frm;
extern const function_base *const vfmadd;
extern const function_base *const vfmadd_frm;
extern const function_base *const vfnmsub;
+extern const function_base *const vfnmsub_frm;
extern const function_base *const vfnmacc;
extern const function_base *const vfnmacc_frm;
extern const function_base *const vfmsac;
diff --git a/gcc/config/riscv/riscv-vector-builtins-functions.def
b/gcc/config/riscv/riscv-vector-builtins-functions.def
index 75235ec01d3..d75b281eebe 100644
--- a/gcc/config/riscv/riscv-vector-builtins-functions.def
+++ b/gcc/config/riscv/riscv-vector-builtins-functions.def
@@ -363,6 +363,8 @@ DEF_RVV_FUNCTION (vfnmadd_frm, alu_frm, full_preds,
f_vvvv_ops)
DEF_RVV_FUNCTION (vfnmadd_frm, alu_frm, full_preds, f_vvfv_ops)
DEF_RVV_FUNCTION (vfmsub_frm, alu_frm, full_preds, f_vvvv_ops)
DEF_RVV_FUNCTION (vfmsub_frm, alu_frm, full_preds, f_vvfv_ops)
+DEF_RVV_FUNCTION (vfnmsub_frm, alu_frm, full_preds, f_vvvv_ops)
+DEF_RVV_FUNCTION (vfnmsub_frm, alu_frm, full_preds, f_vvfv_ops)
// 13.7. Vector Widening Floating-Point Fused Multiply-Add Instructions
DEF_RVV_FUNCTION (vfwmacc, alu, full_preds, f_wwvv_ops)
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/float-point-nmsub.c
b/gcc/testsuite/gcc.target/riscv/rvv/base/float-point-nmsub.c
new file mode 100644
index 00000000000..1b3e939b1e1
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/base/float-point-nmsub.c
@@ -0,0 +1,47 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv -mabi=lp64 -O3 -Wno-psabi" } */
+
+#include "riscv_vector.h"
+
+typedef float float32_t;
+
+vfloat32m1_t
+test_riscv_vfnmsub_vv_f32m1_rm (vfloat32m1_t vd, vfloat32m1_t op1,
+ vfloat32m1_t op2, size_t vl) {
+ return __riscv_vfnmsub_vv_f32m1_rm (vd, op1, op2, 0, vl);
+}
+
+vfloat32m1_t
+test_vfnmsub_vv_f32m1_rm_m (vbool32_t mask, vfloat32m1_t vd, vfloat32m1_t op1,
+ vfloat32m1_t op2, size_t vl) {
+ return __riscv_vfnmsub_vv_f32m1_rm_m (mask, vd, op1, op2, 1, vl);
+}
+
+vfloat32m1_t
+test_vfnmsub_vf_f32m1_rm (vfloat32m1_t vd, float32_t op1, vfloat32m1_t op2,
+ size_t vl) {
+ return __riscv_vfnmsub_vf_f32m1_rm (vd, op1, op2, 2, vl);
+}
+
+vfloat32m1_t
+test_vfnmsub_vf_f32m1_rm_m (vfloat32m1_t vd, vbool32_t mask, float32_t op1,
+ vfloat32m1_t op2, size_t vl) {
+ return __riscv_vfnmsub_vf_f32m1_rm_m (mask, vd, op1, op2, 3, vl);
+}
+
+vfloat32m1_t
+test_riscv_vfnmsub_vv_f32m1 (vfloat32m1_t vd, vfloat32m1_t op1,
+ vfloat32m1_t op2, size_t vl) {
+ return __riscv_vfnmsub_vv_f32m1 (vd, op1, op2, vl);
+}
+
+vfloat32m1_t
+test_vfnmsub_vv_f32m1_m (vbool32_t mask, vfloat32m1_t vd, vfloat32m1_t op1,
+ vfloat32m1_t op2, size_t vl) {
+ return __riscv_vfnmsub_vv_f32m1_m (mask, vd, op1, op2, vl);
+}
+
+/* { dg-final { scan-assembler-times
{vfnmsub\.v[vf]\s+v[0-9]+,\s*[fav]+[0-9]+,\s*v[0-9]+} 6 } } */
+/* { dg-final { scan-assembler-times {frrm\s+[axs][0-9]+} 4 } } */
+/* { dg-final { scan-assembler-times {fsrm\s+[axs][0-9]+} 4 } } */
+/* { dg-final { scan-assembler-times {fsrmi\s+[01234]} 4 } } */
--
2.34.1
