[v2] RISC-V: Implement autovec copysign.
Checks
Commit Message
Hi,
changes from v1:
- Removed UNSPEC_VNCOPYSIGN
- Adjusted xorsign test expectation.
Regards
Robin
This adds vector copysign, ncopysign and xorsign as well as the
accompanying tests.
gcc/ChangeLog:
* config/riscv/autovec.md (copysign<mode>3): Add expander.
(xorsign<mode>3): Dito.
* config/riscv/riscv-vector-builtins-bases.cc (class vfsgnjn):
New class.
* config/riscv/vector-iterators.md (copysign): Remove ncopysign.
(xorsign): Dito.
(n): Dito.
(x): Dito.
* config/riscv/vector.md (@pred_ncopysign<mode>): Split off.
(@pred_ncopysign<mode>_scalar): Dito.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/rvv/autovec/binop/copysign-run.c: New test.
* gcc.target/riscv/rvv/autovec/binop/copysign-rv64gcv.c: New test.
* gcc.target/riscv/rvv/autovec/binop/copysign-rv32gcv.c: New test.
* gcc.target/riscv/rvv/autovec/binop/copysign-template.h: New test.
* gcc.target/riscv/rvv/autovec/binop/copysign-zvfh-run.c: New test.
---
gcc/config/riscv/autovec.md | 43 +++++++++
.../riscv/riscv-vector-builtins-bases.cc | 18 +++-
gcc/config/riscv/vector-iterators.md | 10 +--
gcc/config/riscv/vector.md | 43 +++++++++
.../riscv/rvv/autovec/binop/copysign-run.c | 89 +++++++++++++++++++
.../rvv/autovec/binop/copysign-rv32gcv.c | 11 +++
.../rvv/autovec/binop/copysign-rv64gcv.c | 11 +++
.../rvv/autovec/binop/copysign-template.h | 78 ++++++++++++++++
.../rvv/autovec/binop/copysign-zvfh-run.c | 83 +++++++++++++++++
9 files changed, 377 insertions(+), 9 deletions(-)
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-run.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-rv32gcv.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-rv64gcv.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-template.h
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-zvfh-run.c
Comments
LGTM.
juzhe.zhong@rivai.ai
From: Robin Dapp
Date: 2023-06-21 20:57
To: juzhe.zhong@rivai.ai; gcc-patches; palmer; kito.cheng; jeffreyalaw
CC: rdapp.gcc
Subject: [PATCH v2] RISC-V: Implement autovec copysign.
Hi,
changes from v1:
- Removed UNSPEC_VNCOPYSIGN
- Adjusted xorsign test expectation.
Regards
Robin
This adds vector copysign, ncopysign and xorsign as well as the
accompanying tests.
gcc/ChangeLog:
* config/riscv/autovec.md (copysign<mode>3): Add expander.
(xorsign<mode>3): Dito.
* config/riscv/riscv-vector-builtins-bases.cc (class vfsgnjn):
New class.
* config/riscv/vector-iterators.md (copysign): Remove ncopysign.
(xorsign): Dito.
(n): Dito.
(x): Dito.
* config/riscv/vector.md (@pred_ncopysign<mode>): Split off.
(@pred_ncopysign<mode>_scalar): Dito.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/rvv/autovec/binop/copysign-run.c: New test.
* gcc.target/riscv/rvv/autovec/binop/copysign-rv64gcv.c: New test.
* gcc.target/riscv/rvv/autovec/binop/copysign-rv32gcv.c: New test.
* gcc.target/riscv/rvv/autovec/binop/copysign-template.h: New test.
* gcc.target/riscv/rvv/autovec/binop/copysign-zvfh-run.c: New test.
---
gcc/config/riscv/autovec.md | 43 +++++++++
.../riscv/riscv-vector-builtins-bases.cc | 18 +++-
gcc/config/riscv/vector-iterators.md | 10 +--
gcc/config/riscv/vector.md | 43 +++++++++
.../riscv/rvv/autovec/binop/copysign-run.c | 89 +++++++++++++++++++
.../rvv/autovec/binop/copysign-rv32gcv.c | 11 +++
.../rvv/autovec/binop/copysign-rv64gcv.c | 11 +++
.../rvv/autovec/binop/copysign-template.h | 78 ++++++++++++++++
.../rvv/autovec/binop/copysign-zvfh-run.c | 83 +++++++++++++++++
9 files changed, 377 insertions(+), 9 deletions(-)
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-run.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-rv32gcv.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-rv64gcv.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-template.h
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-zvfh-run.c
diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md
index f1641d7e1ea..f2e69aaf102 100644
--- a/gcc/config/riscv/autovec.md
+++ b/gcc/config/riscv/autovec.md
@@ -804,3 +804,46 @@ (define_expand "<optab><mode>3"
riscv_vector::RVV_BINOP, operands);
DONE;
})
+
+;; -------------------------------------------------------------------------------
+;; ---- [FP] Sign copying
+;; -------------------------------------------------------------------------------
+;; Includes:
+;; - vfsgnj.vv/vfsgnjn.vv
+;; - vfsgnj.vf/vfsgnjn.vf
+;; -------------------------------------------------------------------------------
+
+;; Leave the pattern like this as to still allow combine to match
+;; a negated copysign (see vector.md) before adding the UNSPEC_VPREDICATE later.
+(define_insn_and_split "copysign<mode>3"
+ [(set (match_operand:VF 0 "register_operand" "=vd, vd, vr, vr")
+ (unspec:VF
+ [(match_operand:VF 1 "register_operand" " vr, vr, vr, vr")
+ (match_operand:VF 2 "register_operand" " vr, vr, vr, vr")] UNSPEC_VCOPYSIGN))]
+ "TARGET_VECTOR && can_create_pseudo_p ()"
+ "#"
+ "&& 1"
+ [(const_int 0)]
+{
+ riscv_vector::emit_vlmax_insn (code_for_pred (UNSPEC_VCOPYSIGN, <MODE>mode),
+ riscv_vector::RVV_BINOP, operands);
+ DONE;
+}
+ [(set_attr "type" "vfsgnj")
+ (set_attr "mode" "<MODE>")])
+
+;; -------------------------------------------------------------------------------
+;; Includes:
+;; - vfsgnjx.vv
+;; - vfsgnjx.vf
+;; -------------------------------------------------------------------------------
+(define_expand "xorsign<mode>3"
+ [(match_operand:VF_AUTO 0 "register_operand")
+ (match_operand:VF_AUTO 1 "register_operand")
+ (match_operand:VF_AUTO 2 "register_operand")]
+ "TARGET_VECTOR"
+{
+ riscv_vector::emit_vlmax_insn (code_for_pred (UNSPEC_VXORSIGN, <MODE>mode),
+ riscv_vector::RVV_BINOP, operands);
+ DONE;
+})
diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc
index c6c53dc13a5..6b8fec47ac9 100644
--- a/gcc/config/riscv/riscv-vector-builtins-bases.cc
+++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc
@@ -1212,7 +1212,7 @@ public:
}
};
-/* Implements vfsqrt7/vfrec7/vfclass/vfsgnj/vfsgnjn/vfsgnjx. */
+/* Implements vfsqrt7/vfrec7/vfclass/vfsgnj/vfsgnjx. */
template<int UNSPEC>
class float_misc : public function_base
{
@@ -1227,6 +1227,20 @@ public:
}
};
+/* Implements vfsgnjn. */
+class vfsgnjn : public function_base
+{
+public:
+ rtx expand (function_expander &e) const override
+ {
+ if (e.op_info->op == OP_TYPE_vf)
+ return e.use_exact_insn (code_for_pred_ncopysign_scalar (e.vector_mode ()));
+ if (e.op_info->op == OP_TYPE_vv)
+ return e.use_exact_insn (code_for_pred_ncopysign (e.vector_mode ()));
+ gcc_unreachable ();
+ }
+};
+
/* Implements vmfeq/vmfne/vmflt/vmfgt/vmfle/vmfge. */
template<rtx_code CODE>
class fcmp : public function_base
@@ -2031,7 +2045,7 @@ static CONSTEXPR const float_misc<UNSPEC_VFREC7> vfrec7_obj;
static CONSTEXPR const binop<SMIN> vfmin_obj;
static CONSTEXPR const binop<SMAX> vfmax_obj;
static CONSTEXPR const float_misc<UNSPEC_VCOPYSIGN> vfsgnj_obj;
-static CONSTEXPR const float_misc<UNSPEC_VNCOPYSIGN> vfsgnjn_obj;
+static CONSTEXPR const vfsgnjn vfsgnjn_obj;
static CONSTEXPR const float_misc<UNSPEC_VXORSIGN> vfsgnjx_obj;
static CONSTEXPR const unop<NEG> vfneg_obj;
static CONSTEXPR const unop<ABS> vfabs_obj;
diff --git a/gcc/config/riscv/vector-iterators.md b/gcc/config/riscv/vector-iterators.md
index 6ca1c54c709..2c0c53c0c28 100644
--- a/gcc/config/riscv/vector-iterators.md
+++ b/gcc/config/riscv/vector-iterators.md
@@ -61,7 +61,6 @@ (define_c_enum "unspec" [
UNSPEC_VFCLASS
UNSPEC_VCOPYSIGN
- UNSPEC_VNCOPYSIGN
UNSPEC_VXORSIGN
UNSPEC_VFCVT
@@ -1480,14 +1479,11 @@ (define_int_attr misc_op [(UNSPEC_VMSBF "sbf") (UNSPEC_VMSIF "sif") (UNSPEC_VMSO
(define_int_attr float_insn_type [(UNSPEC_VFRSQRT7 "vfsqrt") (UNSPEC_VFREC7 "vfrecp")])
-(define_int_iterator VCOPYSIGNS [UNSPEC_VCOPYSIGN UNSPEC_VNCOPYSIGN UNSPEC_VXORSIGN])
+(define_int_iterator VCOPYSIGNS [UNSPEC_VCOPYSIGN UNSPEC_VXORSIGN])
-(define_int_attr copysign [(UNSPEC_VCOPYSIGN "copysign")
- (UNSPEC_VNCOPYSIGN "ncopysign")
- (UNSPEC_VXORSIGN "xorsign")])
+(define_int_attr copysign [(UNSPEC_VCOPYSIGN "copysign") (UNSPEC_VXORSIGN "xorsign")])
-(define_int_attr nx [(UNSPEC_VCOPYSIGN "") (UNSPEC_VNCOPYSIGN "n")
- (UNSPEC_VXORSIGN "x")])
+(define_int_attr nx [(UNSPEC_VCOPYSIGN "") (UNSPEC_VXORSIGN "x")])
(define_int_attr ud [(UNSPEC_VSLIDEUP "up") (UNSPEC_VSLIDEDOWN "down")
(UNSPEC_VSLIDE1UP "1up") (UNSPEC_VSLIDE1DOWN "1down")
diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md
index 884e7435cc2..aa81b9023d8 100644
--- a/gcc/config/riscv/vector.md
+++ b/gcc/config/riscv/vector.md
@@ -5820,6 +5820,27 @@ (define_insn "@pred_<copysign><mode>"
[(set_attr "type" "vfsgnj")
(set_attr "mode" "<MODE>")])
+(define_insn "@pred_ncopysign<mode>"
+ [(set (match_operand:VF 0 "register_operand" "=vd, vd, vr, vr")
+ (if_then_else:VF
+ (unspec:<VM>
+ [(match_operand:<VM> 1 "vector_mask_operand" " vm, vm,Wc1,Wc1")
+ (match_operand 5 "vector_length_operand" " rK, rK, rK, rK")
+ (match_operand 6 "const_int_operand" " i, i, i, i")
+ (match_operand 7 "const_int_operand" " i, i, i, i")
+ (match_operand 8 "const_int_operand" " i, i, i, i")
+ (reg:SI VL_REGNUM)
+ (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE)
+ (neg:VF
+ (unspec:VF
+ [(match_operand:VF 3 "register_operand" " vr, vr, vr, vr")
+ (match_operand:VF 4 "register_operand" " vr, vr, vr, vr")] UNSPEC_VCOPYSIGN))
+ (match_operand:VF 2 "vector_merge_operand" " vu, 0, vu, 0")))]
+ "TARGET_VECTOR"
+ "vfsgnjn.vv\t%0,%3,%4%p1"
+ [(set_attr "type" "vfsgnj")
+ (set_attr "mode" "<MODE>")])
+
(define_insn "@pred_<copysign><mode>_scalar"
[(set (match_operand:VF 0 "register_operand" "=vd, vd, vr, vr")
(if_then_else:VF
@@ -5841,6 +5862,28 @@ (define_insn "@pred_<copysign><mode>_scalar"
[(set_attr "type" "vfsgnj")
(set_attr "mode" "<MODE>")])
+(define_insn "@pred_ncopysign<mode>_scalar"
+ [(set (match_operand:VF 0 "register_operand" "=vd, vd, vr, vr")
+ (if_then_else:VF
+ (unspec:<VM>
+ [(match_operand:<VM> 1 "vector_mask_operand" " vm, vm,Wc1,Wc1")
+ (match_operand 5 "vector_length_operand" " rK, rK, rK, rK")
+ (match_operand 6 "const_int_operand" " i, i, i, i")
+ (match_operand 7 "const_int_operand" " i, i, i, i")
+ (match_operand 8 "const_int_operand" " i, i, i, i")
+ (reg:SI VL_REGNUM)
+ (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE)
+ (neg:VF
+ (unspec:VF
+ [(match_operand:VF 3 "register_operand" " vr, vr, vr, vr")
+ (vec_duplicate:VF
+ (match_operand:<VEL> 4 "register_operand" " f, f, f, f"))] UNSPEC_VCOPYSIGN))
+ (match_operand:VF 2 "vector_merge_operand" " vu, 0, vu, 0")))]
+ "TARGET_VECTOR"
+ "vfsgnjn.vf\t%0,%3,%4%p1"
+ [(set_attr "type" "vfsgnj")
+ (set_attr "mode" "<MODE>")])
+
;; -------------------------------------------------------------------------------
;; ---- Predicated floating-point ternary operations
;; -------------------------------------------------------------------------------
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-run.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-run.c
new file mode 100644
index 00000000000..7a6d429c9a9
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-run.c
@@ -0,0 +1,89 @@
+/* { dg-do run { target { riscv_vector } } } */
+/* { dg-additional-options "-std=c99 -fno-vect-cost-model --param=riscv-autovec-preference=fixed-vlmax -ffast-math" } */
+
+#include "copysign-template.h"
+
+#include <assert.h>
+
+#define SZ 512
+
+#define EPS 1e-6
+
+#define RUN(TYPE,VAL) \
+ TYPE a##TYPE[SZ]; \
+ TYPE b##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ { \
+ a##TYPE[i] = i; \
+ b##TYPE[i] = (i & 1) ? VAL : -VAL; \
+ } \
+ copysign_##TYPE (a##TYPE, a##TYPE, b##TYPE, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (a##TYPE[i] - ((i & 1) ? i : -i)) < EPS); \
+
+#define RUN2(TYPE,VAL) \
+ TYPE a2##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ a2##TYPE[i] = i; \
+ copysigns_##TYPE (a2##TYPE, a2##TYPE, -VAL, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (a2##TYPE[i] + i) < EPS); \
+
+#define RUN3(TYPE,VAL) \
+ TYPE a3##TYPE[SZ]; \
+ TYPE b3##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ { \
+ a3##TYPE[i] = (i & 1) ? -i : i; \
+ b3##TYPE[i] = (i & 1) ? VAL : -VAL; \
+ } \
+ xorsign_##TYPE (a3##TYPE, a3##TYPE, b3##TYPE, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (a3##TYPE[i] + i) < EPS); \
+
+#define RUN4(TYPE,VAL) \
+ TYPE a4##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ a4##TYPE[i] = -i; \
+ xorsigns_##TYPE (a4##TYPE, a4##TYPE, -VAL, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (a4##TYPE[i] - i) < EPS); \
+
+#define RUN5(TYPE,VAL) \
+ TYPE a5##TYPE[SZ]; \
+ TYPE b5##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ { \
+ a5##TYPE[i] = i; \
+ b5##TYPE[i] = (i & 1) ? VAL : -VAL; \
+ } \
+ ncopysign_##TYPE (a5##TYPE, a5##TYPE, b##TYPE, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (-a5##TYPE[i] - ((i & 1) ? i : -i)) < EPS); \
+
+#define RUN6(TYPE,VAL) \
+ TYPE a6##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ a6##TYPE[i] = i; \
+ ncopysigns_##TYPE (a6##TYPE, a6##TYPE, -VAL, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (-a6##TYPE[i] + i) < EPS); \
+
+#define RUN_ALL() \
+ RUN(float, 5) \
+ RUN(double, 6) \
+ RUN2(float, 11) \
+ RUN2(double, 12) \
+ RUN3(float, 16) \
+ RUN3(double, 18) \
+ RUN4(float, 17) \
+ RUN4(double, 19) \
+ RUN5(float, 123) \
+ RUN5(double, 523) \
+ RUN6(float, 777) \
+ RUN6(double, 877) \
+
+int main ()
+{
+ RUN_ALL()
+}
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-rv32gcv.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-rv32gcv.c
new file mode 100644
index 00000000000..db29e37598a
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-rv32gcv.c
@@ -0,0 +1,11 @@
+/* { dg-do compile } */
+/* { dg-additional-options "-std=c99 -O3 -fno-vect-cost-model -march=rv32gcv_zvfh -mabi=ilp32d --param=riscv-autovec-preference=fixed-vlmax -ffast-math" } */
+
+#include "copysign-template.h"
+
+/* { dg-final { scan-assembler-times {\tvfsgnj\.vv} 6 } } */
+/* The vectorizer wraps scalar variants of copysign into vector constants which
+ expand cannot handle currently. Therefore only expect 3 instead of 6
+ vfsgnjx.vv. */
+/* { dg-final { scan-assembler-times {\tvfsgnjx\.vv} 3 } } */
+/* { dg-final { scan-assembler-times {\tvfsgnjn\.vv} 6 } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-rv64gcv.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-rv64gcv.c
new file mode 100644
index 00000000000..1c2504915cc
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-rv64gcv.c
@@ -0,0 +1,11 @@
+/* { dg-do compile } */
+/* { dg-additional-options "-std=c99 -O3 -fno-vect-cost-model -march=rv64gcv_zvfh -mabi=lp64d --param=riscv-autovec-preference=fixed-vlmax -ffast-math" } */
+
+#include "copysign-template.h"
+
+/* { dg-final { scan-assembler-times {\tvfsgnj\.vv} 6 } } */
+/* The vectorizer wraps scalar variants of copysign into vector constants which
+ expand cannot handle currently. Therefore only expect 3 instead of 6
+ vfsgnjx.vv. */
+/* { dg-final { scan-assembler-times {\tvfsgnjx\.vv} 3 } } */
+/* { dg-final { scan-assembler-times {\tvfsgnjn\.vv} 6 } } */
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-template.h b/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-template.h
new file mode 100644
index 00000000000..df2274fd3ef
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-template.h
@@ -0,0 +1,78 @@
+#include <stdint-gcc.h>
+
+#define TEST_TYPE(TYPE, SUFFIX) \
+ __attribute__((noipa)) \
+ void copysign_##TYPE (TYPE *restrict dst, TYPE *restrict a, \
+ TYPE *restrict b, int n) \
+ { \
+ for (int i = 0; i < n; i++) \
+ dst[i] = __builtin_copysign##SUFFIX (a[i], b[i]); \
+ }
+
+#define TEST_TYPE2(TYPE, SUFFIX) \
+ __attribute__((noipa)) \
+ void copysigns_##TYPE (TYPE *restrict dst, TYPE *restrict a, \
+ TYPE b, int n) \
+ { \
+ for (int i = 0; i < n; i++) \
+ dst[i] = __builtin_copysign##SUFFIX (a[i], b); \
+ }
+
+#define TEST_TYPE3(TYPE, SUFFIX) \
+ __attribute__((noipa)) \
+ void xorsign_##TYPE (TYPE *restrict dst, TYPE *restrict a, \
+ TYPE *restrict b, int n) \
+ { \
+ for (int i = 0; i < n; i++) \
+ dst[i] = a[i] * __builtin_copysign##SUFFIX (1.0, b[i]); \
+ }
+
+#define TEST_TYPE4(TYPE, SUFFIX) \
+ __attribute__((noipa)) \
+ void xorsigns_##TYPE (TYPE *restrict dst, TYPE *restrict a, \
+ TYPE b, int n) \
+ { \
+ for (int i = 0; i < n; i++) \
+ dst[i] = a[i] * __builtin_copysign##SUFFIX (1.0, b); \
+ }
+
+#define TEST_TYPE5(TYPE, SUFFIX) \
+ __attribute__((noipa)) \
+ void ncopysign_##TYPE (TYPE *restrict dst, TYPE *restrict a, \
+ TYPE *restrict b, int n) \
+ { \
+ for (int i = 0; i < n; i++) \
+ dst[i] = -__builtin_copysign##SUFFIX (a[i], b[i]); \
+ }
+
+#define TEST_TYPE6(TYPE, SUFFIX) \
+ __attribute__((noipa)) \
+ void ncopysigns_##TYPE (TYPE *restrict dst, TYPE *restrict a, \
+ TYPE b, int n) \
+ { \
+ for (int i = 0; i < n; i++) \
+ dst[i] = -__builtin_copysign##SUFFIX (a[i], b); \
+ }
+
+
+#define TEST_ALL() \
+ TEST_TYPE(_Float16,f16) \
+ TEST_TYPE(float,f) \
+ TEST_TYPE(double,) \
+ TEST_TYPE2(_Float16,f16) \
+ TEST_TYPE2(float,f) \
+ TEST_TYPE2(double,) \
+ TEST_TYPE3(_Float16,f16) \
+ TEST_TYPE3(float,f) \
+ TEST_TYPE3(double,) \
+ TEST_TYPE4(_Float16,f16) \
+ TEST_TYPE4(float,f) \
+ TEST_TYPE4(double,) \
+ TEST_TYPE5(_Float16,f16) \
+ TEST_TYPE5(float,f) \
+ TEST_TYPE5(double,) \
+ TEST_TYPE6(_Float16,f16) \
+ TEST_TYPE6(float,f) \
+ TEST_TYPE6(double,) \
+
+TEST_ALL()
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-zvfh-run.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-zvfh-run.c
new file mode 100644
index 00000000000..7aaac9794f0
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/binop/copysign-zvfh-run.c
@@ -0,0 +1,83 @@
+/* { dg-do run { target { riscv_vector && riscv_zvfh_hw } } } */
+/* { dg-additional-options "-std=c99 -fno-vect-cost-model --param=riscv-autovec-preference=fixed-vlmax -ffast-math" } */
+
+#include "copysign-template.h"
+
+#include <assert.h>
+
+#define SZ 512
+
+#define EPS 1e-6
+
+#define RUN(TYPE,VAL) \
+ TYPE a##TYPE[SZ]; \
+ TYPE b##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ { \
+ a##TYPE[i] = i; \
+ b##TYPE[i] = (i & 1) ? VAL : -VAL; \
+ } \
+ copysign_##TYPE (a##TYPE, a##TYPE, b##TYPE, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (a##TYPE[i] - ((i & 1) ? i : -i)) < EPS); \
+
+#define RUN2(TYPE,VAL) \
+ TYPE a2##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ a2##TYPE[i] = i; \
+ copysigns_##TYPE (a2##TYPE, a2##TYPE, -VAL, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (a2##TYPE[i] + i) < EPS); \
+
+#define RUN3(TYPE,VAL) \
+ TYPE a3##TYPE[SZ]; \
+ TYPE b3##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ { \
+ a3##TYPE[i] = (i & 1) ? -i : i; \
+ b3##TYPE[i] = (i & 1) ? VAL : -VAL; \
+ } \
+ xorsign_##TYPE (a3##TYPE, a3##TYPE, b3##TYPE, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (a3##TYPE[i] + i) < EPS); \
+
+#define RUN4(TYPE,VAL) \
+ TYPE a4##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ a4##TYPE[i] = -i; \
+ xorsigns_##TYPE (a4##TYPE, a4##TYPE, -VAL, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (a4##TYPE[i] - i) < EPS); \
+
+#define RUN5(TYPE,VAL) \
+ TYPE a5##TYPE[SZ]; \
+ TYPE b5##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ { \
+ a5##TYPE[i] = i; \
+ b5##TYPE[i] = (i & 1) ? VAL : -VAL; \
+ } \
+ ncopysign_##TYPE (a5##TYPE, a5##TYPE, b##TYPE, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (-a5##TYPE[i] - ((i & 1) ? i : -i)) < EPS); \
+
+#define RUN6(TYPE,VAL) \
+ TYPE a6##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ a6##TYPE[i] = i; \
+ ncopysigns_##TYPE (a6##TYPE, a6##TYPE, -VAL, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (-a6##TYPE[i] + i) < EPS); \
+
+#define RUN_ALL() \
+ RUN(_Float16, 5) \
+ RUN2(_Float16, 11) \
+ RUN3(_Float16, 16) \
+ RUN4(_Float16, 17) \
+ RUN5(_Float16, 123) \
+ RUN6(_Float16, 777) \
+
+int main ()
+{
+ RUN_ALL()
+}
--
2.40.1
On 6/21/23 08:24, 钟居哲 wrote:
> LGTM.
Likewise. OK for the trunk.
jeff
@@ -804,3 +804,46 @@ (define_expand "<optab><mode>3"
riscv_vector::RVV_BINOP, operands);
DONE;
})
+
+;; -------------------------------------------------------------------------------
+;; ---- [FP] Sign copying
+;; -------------------------------------------------------------------------------
+;; Includes:
+;; - vfsgnj.vv/vfsgnjn.vv
+;; - vfsgnj.vf/vfsgnjn.vf
+;; -------------------------------------------------------------------------------
+
+;; Leave the pattern like this as to still allow combine to match
+;; a negated copysign (see vector.md) before adding the UNSPEC_VPREDICATE later.
+(define_insn_and_split "copysign<mode>3"
+ [(set (match_operand:VF 0 "register_operand" "=vd, vd, vr, vr")
+ (unspec:VF
+ [(match_operand:VF 1 "register_operand" " vr, vr, vr, vr")
+ (match_operand:VF 2 "register_operand" " vr, vr, vr, vr")] UNSPEC_VCOPYSIGN))]
+ "TARGET_VECTOR && can_create_pseudo_p ()"
+ "#"
+ "&& 1"
+ [(const_int 0)]
+{
+ riscv_vector::emit_vlmax_insn (code_for_pred (UNSPEC_VCOPYSIGN, <MODE>mode),
+ riscv_vector::RVV_BINOP, operands);
+ DONE;
+}
+ [(set_attr "type" "vfsgnj")
+ (set_attr "mode" "<MODE>")])
+
+;; -------------------------------------------------------------------------------
+;; Includes:
+;; - vfsgnjx.vv
+;; - vfsgnjx.vf
+;; -------------------------------------------------------------------------------
+(define_expand "xorsign<mode>3"
+ [(match_operand:VF_AUTO 0 "register_operand")
+ (match_operand:VF_AUTO 1 "register_operand")
+ (match_operand:VF_AUTO 2 "register_operand")]
+ "TARGET_VECTOR"
+{
+ riscv_vector::emit_vlmax_insn (code_for_pred (UNSPEC_VXORSIGN, <MODE>mode),
+ riscv_vector::RVV_BINOP, operands);
+ DONE;
+})
@@ -1212,7 +1212,7 @@ public:
}
};
-/* Implements vfsqrt7/vfrec7/vfclass/vfsgnj/vfsgnjn/vfsgnjx. */
+/* Implements vfsqrt7/vfrec7/vfclass/vfsgnj/vfsgnjx. */
template<int UNSPEC>
class float_misc : public function_base
{
@@ -1227,6 +1227,20 @@ public:
}
};
+/* Implements vfsgnjn. */
+class vfsgnjn : public function_base
+{
+public:
+ rtx expand (function_expander &e) const override
+ {
+ if (e.op_info->op == OP_TYPE_vf)
+ return e.use_exact_insn (code_for_pred_ncopysign_scalar (e.vector_mode ()));
+ if (e.op_info->op == OP_TYPE_vv)
+ return e.use_exact_insn (code_for_pred_ncopysign (e.vector_mode ()));
+ gcc_unreachable ();
+ }
+};
+
/* Implements vmfeq/vmfne/vmflt/vmfgt/vmfle/vmfge. */
template<rtx_code CODE>
class fcmp : public function_base
@@ -2031,7 +2045,7 @@ static CONSTEXPR const float_misc<UNSPEC_VFREC7> vfrec7_obj;
static CONSTEXPR const binop<SMIN> vfmin_obj;
static CONSTEXPR const binop<SMAX> vfmax_obj;
static CONSTEXPR const float_misc<UNSPEC_VCOPYSIGN> vfsgnj_obj;
-static CONSTEXPR const float_misc<UNSPEC_VNCOPYSIGN> vfsgnjn_obj;
+static CONSTEXPR const vfsgnjn vfsgnjn_obj;
static CONSTEXPR const float_misc<UNSPEC_VXORSIGN> vfsgnjx_obj;
static CONSTEXPR const unop<NEG> vfneg_obj;
static CONSTEXPR const unop<ABS> vfabs_obj;
@@ -61,7 +61,6 @@ (define_c_enum "unspec" [
UNSPEC_VFCLASS
UNSPEC_VCOPYSIGN
- UNSPEC_VNCOPYSIGN
UNSPEC_VXORSIGN
UNSPEC_VFCVT
@@ -1480,14 +1479,11 @@ (define_int_attr misc_op [(UNSPEC_VMSBF "sbf") (UNSPEC_VMSIF "sif") (UNSPEC_VMSO
(define_int_attr float_insn_type [(UNSPEC_VFRSQRT7 "vfsqrt") (UNSPEC_VFREC7 "vfrecp")])
-(define_int_iterator VCOPYSIGNS [UNSPEC_VCOPYSIGN UNSPEC_VNCOPYSIGN UNSPEC_VXORSIGN])
+(define_int_iterator VCOPYSIGNS [UNSPEC_VCOPYSIGN UNSPEC_VXORSIGN])
-(define_int_attr copysign [(UNSPEC_VCOPYSIGN "copysign")
- (UNSPEC_VNCOPYSIGN "ncopysign")
- (UNSPEC_VXORSIGN "xorsign")])
+(define_int_attr copysign [(UNSPEC_VCOPYSIGN "copysign") (UNSPEC_VXORSIGN "xorsign")])
-(define_int_attr nx [(UNSPEC_VCOPYSIGN "") (UNSPEC_VNCOPYSIGN "n")
- (UNSPEC_VXORSIGN "x")])
+(define_int_attr nx [(UNSPEC_VCOPYSIGN "") (UNSPEC_VXORSIGN "x")])
(define_int_attr ud [(UNSPEC_VSLIDEUP "up") (UNSPEC_VSLIDEDOWN "down")
(UNSPEC_VSLIDE1UP "1up") (UNSPEC_VSLIDE1DOWN "1down")
@@ -5820,6 +5820,27 @@ (define_insn "@pred_<copysign><mode>"
[(set_attr "type" "vfsgnj")
(set_attr "mode" "<MODE>")])
+(define_insn "@pred_ncopysign<mode>"
+ [(set (match_operand:VF 0 "register_operand" "=vd, vd, vr, vr")
+ (if_then_else:VF
+ (unspec:<VM>
+ [(match_operand:<VM> 1 "vector_mask_operand" " vm, vm,Wc1,Wc1")
+ (match_operand 5 "vector_length_operand" " rK, rK, rK, rK")
+ (match_operand 6 "const_int_operand" " i, i, i, i")
+ (match_operand 7 "const_int_operand" " i, i, i, i")
+ (match_operand 8 "const_int_operand" " i, i, i, i")
+ (reg:SI VL_REGNUM)
+ (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE)
+ (neg:VF
+ (unspec:VF
+ [(match_operand:VF 3 "register_operand" " vr, vr, vr, vr")
+ (match_operand:VF 4 "register_operand" " vr, vr, vr, vr")] UNSPEC_VCOPYSIGN))
+ (match_operand:VF 2 "vector_merge_operand" " vu, 0, vu, 0")))]
+ "TARGET_VECTOR"
+ "vfsgnjn.vv\t%0,%3,%4%p1"
+ [(set_attr "type" "vfsgnj")
+ (set_attr "mode" "<MODE>")])
+
(define_insn "@pred_<copysign><mode>_scalar"
[(set (match_operand:VF 0 "register_operand" "=vd, vd, vr, vr")
(if_then_else:VF
@@ -5841,6 +5862,28 @@ (define_insn "@pred_<copysign><mode>_scalar"
[(set_attr "type" "vfsgnj")
(set_attr "mode" "<MODE>")])
+(define_insn "@pred_ncopysign<mode>_scalar"
+ [(set (match_operand:VF 0 "register_operand" "=vd, vd, vr, vr")
+ (if_then_else:VF
+ (unspec:<VM>
+ [(match_operand:<VM> 1 "vector_mask_operand" " vm, vm,Wc1,Wc1")
+ (match_operand 5 "vector_length_operand" " rK, rK, rK, rK")
+ (match_operand 6 "const_int_operand" " i, i, i, i")
+ (match_operand 7 "const_int_operand" " i, i, i, i")
+ (match_operand 8 "const_int_operand" " i, i, i, i")
+ (reg:SI VL_REGNUM)
+ (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE)
+ (neg:VF
+ (unspec:VF
+ [(match_operand:VF 3 "register_operand" " vr, vr, vr, vr")
+ (vec_duplicate:VF
+ (match_operand:<VEL> 4 "register_operand" " f, f, f, f"))] UNSPEC_VCOPYSIGN))
+ (match_operand:VF 2 "vector_merge_operand" " vu, 0, vu, 0")))]
+ "TARGET_VECTOR"
+ "vfsgnjn.vf\t%0,%3,%4%p1"
+ [(set_attr "type" "vfsgnj")
+ (set_attr "mode" "<MODE>")])
+
;; -------------------------------------------------------------------------------
;; ---- Predicated floating-point ternary operations
;; -------------------------------------------------------------------------------
new file mode 100644
@@ -0,0 +1,89 @@
+/* { dg-do run { target { riscv_vector } } } */
+/* { dg-additional-options "-std=c99 -fno-vect-cost-model --param=riscv-autovec-preference=fixed-vlmax -ffast-math" } */
+
+#include "copysign-template.h"
+
+#include <assert.h>
+
+#define SZ 512
+
+#define EPS 1e-6
+
+#define RUN(TYPE,VAL) \
+ TYPE a##TYPE[SZ]; \
+ TYPE b##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ { \
+ a##TYPE[i] = i; \
+ b##TYPE[i] = (i & 1) ? VAL : -VAL; \
+ } \
+ copysign_##TYPE (a##TYPE, a##TYPE, b##TYPE, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (a##TYPE[i] - ((i & 1) ? i : -i)) < EPS); \
+
+#define RUN2(TYPE,VAL) \
+ TYPE a2##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ a2##TYPE[i] = i; \
+ copysigns_##TYPE (a2##TYPE, a2##TYPE, -VAL, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (a2##TYPE[i] + i) < EPS); \
+
+#define RUN3(TYPE,VAL) \
+ TYPE a3##TYPE[SZ]; \
+ TYPE b3##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ { \
+ a3##TYPE[i] = (i & 1) ? -i : i; \
+ b3##TYPE[i] = (i & 1) ? VAL : -VAL; \
+ } \
+ xorsign_##TYPE (a3##TYPE, a3##TYPE, b3##TYPE, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (a3##TYPE[i] + i) < EPS); \
+
+#define RUN4(TYPE,VAL) \
+ TYPE a4##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ a4##TYPE[i] = -i; \
+ xorsigns_##TYPE (a4##TYPE, a4##TYPE, -VAL, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (a4##TYPE[i] - i) < EPS); \
+
+#define RUN5(TYPE,VAL) \
+ TYPE a5##TYPE[SZ]; \
+ TYPE b5##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ { \
+ a5##TYPE[i] = i; \
+ b5##TYPE[i] = (i & 1) ? VAL : -VAL; \
+ } \
+ ncopysign_##TYPE (a5##TYPE, a5##TYPE, b##TYPE, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (-a5##TYPE[i] - ((i & 1) ? i : -i)) < EPS); \
+
+#define RUN6(TYPE,VAL) \
+ TYPE a6##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ a6##TYPE[i] = i; \
+ ncopysigns_##TYPE (a6##TYPE, a6##TYPE, -VAL, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (-a6##TYPE[i] + i) < EPS); \
+
+#define RUN_ALL() \
+ RUN(float, 5) \
+ RUN(double, 6) \
+ RUN2(float, 11) \
+ RUN2(double, 12) \
+ RUN3(float, 16) \
+ RUN3(double, 18) \
+ RUN4(float, 17) \
+ RUN4(double, 19) \
+ RUN5(float, 123) \
+ RUN5(double, 523) \
+ RUN6(float, 777) \
+ RUN6(double, 877) \
+
+int main ()
+{
+ RUN_ALL()
+}
new file mode 100644
@@ -0,0 +1,11 @@
+/* { dg-do compile } */
+/* { dg-additional-options "-std=c99 -O3 -fno-vect-cost-model -march=rv32gcv_zvfh -mabi=ilp32d --param=riscv-autovec-preference=fixed-vlmax -ffast-math" } */
+
+#include "copysign-template.h"
+
+/* { dg-final { scan-assembler-times {\tvfsgnj\.vv} 6 } } */
+/* The vectorizer wraps scalar variants of copysign into vector constants which
+ expand cannot handle currently. Therefore only expect 3 instead of 6
+ vfsgnjx.vv. */
+/* { dg-final { scan-assembler-times {\tvfsgnjx\.vv} 3 } } */
+/* { dg-final { scan-assembler-times {\tvfsgnjn\.vv} 6 } } */
new file mode 100644
@@ -0,0 +1,11 @@
+/* { dg-do compile } */
+/* { dg-additional-options "-std=c99 -O3 -fno-vect-cost-model -march=rv64gcv_zvfh -mabi=lp64d --param=riscv-autovec-preference=fixed-vlmax -ffast-math" } */
+
+#include "copysign-template.h"
+
+/* { dg-final { scan-assembler-times {\tvfsgnj\.vv} 6 } } */
+/* The vectorizer wraps scalar variants of copysign into vector constants which
+ expand cannot handle currently. Therefore only expect 3 instead of 6
+ vfsgnjx.vv. */
+/* { dg-final { scan-assembler-times {\tvfsgnjx\.vv} 3 } } */
+/* { dg-final { scan-assembler-times {\tvfsgnjn\.vv} 6 } } */
new file mode 100644
@@ -0,0 +1,78 @@
+#include <stdint-gcc.h>
+
+#define TEST_TYPE(TYPE, SUFFIX) \
+ __attribute__((noipa)) \
+ void copysign_##TYPE (TYPE *restrict dst, TYPE *restrict a, \
+ TYPE *restrict b, int n) \
+ { \
+ for (int i = 0; i < n; i++) \
+ dst[i] = __builtin_copysign##SUFFIX (a[i], b[i]); \
+ }
+
+#define TEST_TYPE2(TYPE, SUFFIX) \
+ __attribute__((noipa)) \
+ void copysigns_##TYPE (TYPE *restrict dst, TYPE *restrict a, \
+ TYPE b, int n) \
+ { \
+ for (int i = 0; i < n; i++) \
+ dst[i] = __builtin_copysign##SUFFIX (a[i], b); \
+ }
+
+#define TEST_TYPE3(TYPE, SUFFIX) \
+ __attribute__((noipa)) \
+ void xorsign_##TYPE (TYPE *restrict dst, TYPE *restrict a, \
+ TYPE *restrict b, int n) \
+ { \
+ for (int i = 0; i < n; i++) \
+ dst[i] = a[i] * __builtin_copysign##SUFFIX (1.0, b[i]); \
+ }
+
+#define TEST_TYPE4(TYPE, SUFFIX) \
+ __attribute__((noipa)) \
+ void xorsigns_##TYPE (TYPE *restrict dst, TYPE *restrict a, \
+ TYPE b, int n) \
+ { \
+ for (int i = 0; i < n; i++) \
+ dst[i] = a[i] * __builtin_copysign##SUFFIX (1.0, b); \
+ }
+
+#define TEST_TYPE5(TYPE, SUFFIX) \
+ __attribute__((noipa)) \
+ void ncopysign_##TYPE (TYPE *restrict dst, TYPE *restrict a, \
+ TYPE *restrict b, int n) \
+ { \
+ for (int i = 0; i < n; i++) \
+ dst[i] = -__builtin_copysign##SUFFIX (a[i], b[i]); \
+ }
+
+#define TEST_TYPE6(TYPE, SUFFIX) \
+ __attribute__((noipa)) \
+ void ncopysigns_##TYPE (TYPE *restrict dst, TYPE *restrict a, \
+ TYPE b, int n) \
+ { \
+ for (int i = 0; i < n; i++) \
+ dst[i] = -__builtin_copysign##SUFFIX (a[i], b); \
+ }
+
+
+#define TEST_ALL() \
+ TEST_TYPE(_Float16,f16) \
+ TEST_TYPE(float,f) \
+ TEST_TYPE(double,) \
+ TEST_TYPE2(_Float16,f16) \
+ TEST_TYPE2(float,f) \
+ TEST_TYPE2(double,) \
+ TEST_TYPE3(_Float16,f16) \
+ TEST_TYPE3(float,f) \
+ TEST_TYPE3(double,) \
+ TEST_TYPE4(_Float16,f16) \
+ TEST_TYPE4(float,f) \
+ TEST_TYPE4(double,) \
+ TEST_TYPE5(_Float16,f16) \
+ TEST_TYPE5(float,f) \
+ TEST_TYPE5(double,) \
+ TEST_TYPE6(_Float16,f16) \
+ TEST_TYPE6(float,f) \
+ TEST_TYPE6(double,) \
+
+TEST_ALL()
new file mode 100644
@@ -0,0 +1,83 @@
+/* { dg-do run { target { riscv_vector && riscv_zvfh_hw } } } */
+/* { dg-additional-options "-std=c99 -fno-vect-cost-model --param=riscv-autovec-preference=fixed-vlmax -ffast-math" } */
+
+#include "copysign-template.h"
+
+#include <assert.h>
+
+#define SZ 512
+
+#define EPS 1e-6
+
+#define RUN(TYPE,VAL) \
+ TYPE a##TYPE[SZ]; \
+ TYPE b##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ { \
+ a##TYPE[i] = i; \
+ b##TYPE[i] = (i & 1) ? VAL : -VAL; \
+ } \
+ copysign_##TYPE (a##TYPE, a##TYPE, b##TYPE, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (a##TYPE[i] - ((i & 1) ? i : -i)) < EPS); \
+
+#define RUN2(TYPE,VAL) \
+ TYPE a2##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ a2##TYPE[i] = i; \
+ copysigns_##TYPE (a2##TYPE, a2##TYPE, -VAL, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (a2##TYPE[i] + i) < EPS); \
+
+#define RUN3(TYPE,VAL) \
+ TYPE a3##TYPE[SZ]; \
+ TYPE b3##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ { \
+ a3##TYPE[i] = (i & 1) ? -i : i; \
+ b3##TYPE[i] = (i & 1) ? VAL : -VAL; \
+ } \
+ xorsign_##TYPE (a3##TYPE, a3##TYPE, b3##TYPE, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (a3##TYPE[i] + i) < EPS); \
+
+#define RUN4(TYPE,VAL) \
+ TYPE a4##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ a4##TYPE[i] = -i; \
+ xorsigns_##TYPE (a4##TYPE, a4##TYPE, -VAL, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (a4##TYPE[i] - i) < EPS); \
+
+#define RUN5(TYPE,VAL) \
+ TYPE a5##TYPE[SZ]; \
+ TYPE b5##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ { \
+ a5##TYPE[i] = i; \
+ b5##TYPE[i] = (i & 1) ? VAL : -VAL; \
+ } \
+ ncopysign_##TYPE (a5##TYPE, a5##TYPE, b##TYPE, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (-a5##TYPE[i] - ((i & 1) ? i : -i)) < EPS); \
+
+#define RUN6(TYPE,VAL) \
+ TYPE a6##TYPE[SZ]; \
+ for (int i = 0; i < SZ; i++) \
+ a6##TYPE[i] = i; \
+ ncopysigns_##TYPE (a6##TYPE, a6##TYPE, -VAL, SZ); \
+ for (int i = 0; i < SZ; i++) \
+ assert (__builtin_fabs (-a6##TYPE[i] + i) < EPS); \
+
+#define RUN_ALL() \
+ RUN(_Float16, 5) \
+ RUN2(_Float16, 11) \
+ RUN3(_Float16, 16) \
+ RUN4(_Float16, 17) \
+ RUN5(_Float16, 123) \
+ RUN6(_Float16, 777) \
+
+int main ()
+{
+ RUN_ALL()
+}