Re: [PATCH v1] RISC-V: Support RVV VFWMSAC rounding mode intrinsic API

["juzhe.zhong@rivai.ai" <juzhe.zhong@rivai.ai>]

[-- Attachment #1: Type: text/plain, Size: 6307 bytes --]


LGTM


juzhe.zhong@rivai.ai
 
From: pan2.li
Date: 2023-08-14 11:29
To: gcc-patches
CC: juzhe.zhong; pan2.li; yanzhang.wang; kito.cheng
Subject: [PATCH v1] RISC-V: Support RVV VFWMSAC rounding mode intrinsic API
From: Pan Li <pan2.li@intel.com>
 
This patch would like to support the rounding mode API for the
VFWMSAC as the below samples.
 
* __riscv_vfwmsac_vv_f64m2_rm
* __riscv_vfwmsac_vv_f64m2_rm_m
* __riscv_vfwmsac_vf_f64m2_rm
* __riscv_vfwmsac_vf_f64m2_rm_m
 
Signed-off-by: Pan Li <pan2.li@intel.com>
 
gcc/ChangeLog:
 
* config/riscv/riscv-vector-builtins-bases.cc
(class vfwmsac_frm): New class for frm.
(vfwmsac_frm_obj): New declaration.
(BASE): Ditto.
* config/riscv/riscv-vector-builtins-bases.h: Ditto.
* config/riscv/riscv-vector-builtins-functions.def
(vfwmsac_frm): New intrinsic function definition.
 
gcc/testsuite/ChangeLog:
 
* gcc.target/riscv/rvv/base/float-point-wmsac.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-wmsac.c        | 47 +++++++++++++++++++
4 files changed, 75 insertions(+)
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/base/float-point-wmsac.c
 
diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc
index 4a7f2b8e3e9..5a5da903cb2 100644
--- a/gcc/config/riscv/riscv-vector-builtins-bases.cc
+++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc
@@ -585,6 +585,29 @@ public:
   }
};
+/* Implements below instructions for frm
+   - vfwmsac
+*/
+class vfwmsac_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_widen_ternop_insn (
+ code_for_pred_widen_mul_scalar (MINUS, e.vector_mode ()));
+    if (e.op_info->op == OP_TYPE_vv)
+      return e.use_widen_ternop_insn (
+ code_for_pred_widen_mul (MINUS, e.vector_mode ()));
+
+    gcc_unreachable ();
+  }
+};
+
/* Implements vrsub.  */
class vrsub : public function_base
{
@@ -2365,6 +2388,7 @@ static CONSTEXPR const vfwmacc_frm vfwmacc_frm_obj;
static CONSTEXPR const vfwnmacc vfwnmacc_obj;
static CONSTEXPR const vfwnmacc_frm vfwnmacc_frm_obj;
static CONSTEXPR const vfwmsac vfwmsac_obj;
+static CONSTEXPR const vfwmsac_frm vfwmsac_frm_obj;
static CONSTEXPR const vfwnmsac vfwnmsac_obj;
static CONSTEXPR const unop<SQRT> vfsqrt_obj;
static CONSTEXPR const float_misc<UNSPEC_VFRSQRT7> vfrsqrt7_obj;
@@ -2610,6 +2634,7 @@ BASE (vfwmacc_frm)
BASE (vfwnmacc)
BASE (vfwnmacc_frm)
BASE (vfwmsac)
+BASE (vfwmsac_frm)
BASE (vfwnmsac)
BASE (vfsqrt)
BASE (vfrsqrt7)
diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h
index 27c7deb4ec2..09356dd7ac8 100644
--- a/gcc/config/riscv/riscv-vector-builtins-bases.h
+++ b/gcc/config/riscv/riscv-vector-builtins-bases.h
@@ -180,6 +180,7 @@ extern const function_base *const vfwmacc_frm;
extern const function_base *const vfwnmacc;
extern const function_base *const vfwnmacc_frm;
extern const function_base *const vfwmsac;
+extern const function_base *const vfwmsac_frm;
extern const function_base *const vfwnmsac;
extern const function_base *const vfsqrt;
extern const function_base *const vfrsqrt7;
diff --git a/gcc/config/riscv/riscv-vector-builtins-functions.def b/gcc/config/riscv/riscv-vector-builtins-functions.def
index 481c3b899f2..e2a79607d04 100644
--- a/gcc/config/riscv/riscv-vector-builtins-functions.def
+++ b/gcc/config/riscv/riscv-vector-builtins-functions.def
@@ -380,6 +380,8 @@ DEF_RVV_FUNCTION (vfwmacc_frm, alu_frm, full_preds, f_wwvv_ops)
DEF_RVV_FUNCTION (vfwmacc_frm, alu_frm, full_preds, f_wwfv_ops)
DEF_RVV_FUNCTION (vfwnmacc_frm, alu_frm, full_preds, f_wwvv_ops)
DEF_RVV_FUNCTION (vfwnmacc_frm, alu_frm, full_preds, f_wwfv_ops)
+DEF_RVV_FUNCTION (vfwmsac_frm, alu_frm, full_preds, f_wwvv_ops)
+DEF_RVV_FUNCTION (vfwmsac_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)
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/float-point-wmsac.c b/gcc/testsuite/gcc.target/riscv/rvv/base/float-point-wmsac.c
new file mode 100644
index 00000000000..886a0b13695
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/base/float-point-wmsac.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;
+
+vfloat64m2_t
+test_vfwmsac_vv_f32m1_rm (vfloat64m2_t vd, vfloat32m1_t op1, vfloat32m1_t op2,
+   size_t vl) {
+  return __riscv_vfwmsac_vv_f64m2_rm (vd, op1, op2, 0, vl);
+}
+
+vfloat64m2_t
+test_vfwmsac_vv_f32m1_rm_m (vbool32_t mask, vfloat64m2_t vd, vfloat32m1_t op1,
+     vfloat32m1_t op2, size_t vl) {
+  return __riscv_vfwmsac_vv_f64m2_rm_m (mask, vd, op1, op2, 1, vl);
+}
+
+vfloat64m2_t
+test_vfwmsac_vf_f32m1_rm (vfloat64m2_t vd, float32_t op1, vfloat32m1_t op2,
+   size_t vl) {
+  return __riscv_vfwmsac_vf_f64m2_rm (vd, op1, op2, 2, vl);
+}
+
+vfloat64m2_t
+test_vfwmsac_vf_f32m1_rm_m (vbool32_t mask, vfloat64m2_t vd, float32_t op1,
+     vfloat32m1_t op2, size_t vl) {
+  return __riscv_vfwmsac_vf_f64m2_rm_m (mask, vd, op1, op2, 3, vl);
+}
+
+vfloat64m2_t
+test_vfwmsac_vv_f32m1 (vfloat64m2_t vd, vfloat32m1_t op1, vfloat32m1_t op2,
+        size_t vl) {
+  return __riscv_vfwmsac_vv_f64m2 (vd, op1, op2, vl);
+}
+
+vfloat64m2_t
+test_vfwmsac_vv_f32m1_m (vbool32_t mask, vfloat64m2_t vd, vfloat32m1_t op1,
+ vfloat32m1_t op2, size_t vl) {
+  return __riscv_vfwmsac_vv_f64m2_m (mask, vd, op1, op2, vl);
+}
+
+/* { dg-final { scan-assembler-times {vfwmsac\.[vw][vf]\s+v[0-9]+,\s*[fav]+[0-9]+,\s*[fav]+[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