[PATCH] soft-fp: Add fixhf[uns][di|si] and float[uns][di|si]hf

[PATCH] soft-fp: Add fixhf[uns][di|si] and float[uns][di|si]hf

[Kito Cheng]

Add more IEEE half conversion routines:

- Convert 32-bit/64-bit integer to IEEE half.
- Convert IEEE half to 32-bit/64-bit integer.

They are required by RISC-V _Float16 support, tested with RISC-V GCC.
---
 soft-fp/fixhfdi.c     | 45 +++++++++++++++++++++++++++++++++++++++++++
 soft-fp/fixhfsi.c     | 45 +++++++++++++++++++++++++++++++++++++++++++
 soft-fp/fixunshfdi.c  | 45 +++++++++++++++++++++++++++++++++++++++++++
 soft-fp/fixunshfsi.c  | 45 +++++++++++++++++++++++++++++++++++++++++++
 soft-fp/floatdihf.c   | 45 +++++++++++++++++++++++++++++++++++++++++++
 soft-fp/floatsihf.c   | 45 +++++++++++++++++++++++++++++++++++++++++++
 soft-fp/floatundihf.c | 45 +++++++++++++++++++++++++++++++++++++++++++
 soft-fp/floatunsihf.c | 45 +++++++++++++++++++++++++++++++++++++++++++
 8 files changed, 360 insertions(+)
 create mode 100644 soft-fp/fixhfdi.c
 create mode 100644 soft-fp/fixhfsi.c
 create mode 100644 soft-fp/fixunshfdi.c
 create mode 100644 soft-fp/fixunshfsi.c
 create mode 100644 soft-fp/floatdihf.c
 create mode 100644 soft-fp/floatsihf.c
 create mode 100644 soft-fp/floatundihf.c
 create mode 100644 soft-fp/floatunsihf.c

diff --git a/soft-fp/fixhfdi.c b/soft-fp/fixhfdi.c
new file mode 100644
index 0000000000..630a1e3c5f
--- /dev/null
+++ b/soft-fp/fixhfdi.c
@@ -0,0 +1,45 @@
+/* Software floating-point emulation.
+   Convert a IEEE half to 64bit signed integer
+   Copyright (C) 2022 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   In addition to the permissions in the GNU Lesser General Public
+   License, the Free Software Foundation gives you unlimited
+   permission to link the compiled version of this file into
+   combinations with other programs, and to distribute those
+   combinations without any restriction coming from the use of this
+   file.  (The Lesser General Public License restrictions do apply in
+   other respects; for example, they cover modification of the file,
+   and distribution when not linked into a combine executable.)
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include "soft-fp.h"
+#include "half.h"
+
+DItype
+__fixhfdi (HFtype a)
+{
+  FP_DECL_EX;
+  FP_DECL_H (A);
+  UDItype r;
+
+  FP_INIT_EXCEPTIONS;
+  FP_UNPACK_RAW_H (A, a);
+  FP_TO_INT_H (r, A, DI_BITS, 1);
+  FP_HANDLE_EXCEPTIONS;
+
+  return r;
+}
diff --git a/soft-fp/fixhfsi.c b/soft-fp/fixhfsi.c
new file mode 100644
index 0000000000..4699a75533
--- /dev/null
+++ b/soft-fp/fixhfsi.c
@@ -0,0 +1,45 @@
+/* Software floating-point emulation.
+   Convert a IEEE half to 32bit signed integer
+   Copyright (C) 2022 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   In addition to the permissions in the GNU Lesser General Public
+   License, the Free Software Foundation gives you unlimited
+   permission to link the compiled version of this file into
+   combinations with other programs, and to distribute those
+   combinations without any restriction coming from the use of this
+   file.  (The Lesser General Public License restrictions do apply in
+   other respects; for example, they cover modification of the file,
+   and distribution when not linked into a combine executable.)
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include "soft-fp.h"
+#include "half.h"
+
+SItype
+__fixhfsi (HFtype a)
+{
+  FP_DECL_EX;
+  FP_DECL_H (A);
+  USItype r;
+
+  FP_INIT_EXCEPTIONS;
+  FP_UNPACK_RAW_H (A, a);
+  FP_TO_INT_H (r, A, SI_BITS, 1);
+  FP_HANDLE_EXCEPTIONS;
+
+  return r;
+}
diff --git a/soft-fp/fixunshfdi.c b/soft-fp/fixunshfdi.c
new file mode 100644
index 0000000000..3ca9c1aee4
--- /dev/null
+++ b/soft-fp/fixunshfdi.c
@@ -0,0 +1,45 @@
+/* Software floating-point emulation.
+   Convert IEEE half to 64bit unsigned integer
+   Copyright (C) 2022 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   In addition to the permissions in the GNU Lesser General Public
+   License, the Free Software Foundation gives you unlimited
+   permission to link the compiled version of this file into
+   combinations with other programs, and to distribute those
+   combinations without any restriction coming from the use of this
+   file.  (The Lesser General Public License restrictions do apply in
+   other respects; for example, they cover modification of the file,
+   and distribution when not linked into a combine executable.)
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include "soft-fp.h"
+#include "half.h"
+
+UDItype
+__fixunshfdi (HFtype a)
+{
+  FP_DECL_EX;
+  FP_DECL_H (A);
+  UDItype r;
+
+  FP_INIT_EXCEPTIONS;
+  FP_UNPACK_RAW_H (A, a);
+  FP_TO_INT_H (r, A, DI_BITS, 0);
+  FP_HANDLE_EXCEPTIONS;
+
+  return r;
+}
diff --git a/soft-fp/fixunshfsi.c b/soft-fp/fixunshfsi.c
new file mode 100644
index 0000000000..66dd6996ba
--- /dev/null
+++ b/soft-fp/fixunshfsi.c
@@ -0,0 +1,45 @@
+/* Software floating-point emulation.
+   Convert IEEE half to 32bit unsigned integer
+   Copyright (C) 2022 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   In addition to the permissions in the GNU Lesser General Public
+   License, the Free Software Foundation gives you unlimited
+   permission to link the compiled version of this file into
+   combinations with other programs, and to distribute those
+   combinations without any restriction coming from the use of this
+   file.  (The Lesser General Public License restrictions do apply in
+   other respects; for example, they cover modification of the file,
+   and distribution when not linked into a combine executable.)
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include "soft-fp.h"
+#include "half.h"
+
+USItype
+__fixunshfsi (HFtype a)
+{
+  FP_DECL_EX;
+  FP_DECL_H (A);
+  USItype r;
+
+  FP_INIT_EXCEPTIONS;
+  FP_UNPACK_RAW_H (A, a);
+  FP_TO_INT_H (r, A, SI_BITS, 0);
+  FP_HANDLE_EXCEPTIONS;
+
+  return r;
+}
diff --git a/soft-fp/floatdihf.c b/soft-fp/floatdihf.c
new file mode 100644
index 0000000000..284a516cb9
--- /dev/null
+++ b/soft-fp/floatdihf.c
@@ -0,0 +1,45 @@
+/* Software floating-point emulation.
+   Convert a 64bit signed integer to IEEE half
+   Copyright (C) 2022 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   In addition to the permissions in the GNU Lesser General Public
+   License, the Free Software Foundation gives you unlimited
+   permission to link the compiled version of this file into
+   combinations with other programs, and to distribute those
+   combinations without any restriction coming from the use of this
+   file.  (The Lesser General Public License restrictions do apply in
+   other respects; for example, they cover modification of the file,
+   and distribution when not linked into a combine executable.)
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include "soft-fp.h"
+#include "half.h"
+
+HFtype
+__floatdihf (DItype i)
+{
+  FP_DECL_EX;
+  FP_DECL_H (A);
+  HFtype a;
+
+  FP_INIT_ROUNDMODE;
+  FP_FROM_INT_H (A, i, DI_BITS, UDItype);
+  FP_PACK_RAW_H (a, A);
+  FP_HANDLE_EXCEPTIONS;
+
+  return a;
+}
diff --git a/soft-fp/floatsihf.c b/soft-fp/floatsihf.c
new file mode 100644
index 0000000000..2caa11d99c
--- /dev/null
+++ b/soft-fp/floatsihf.c
@@ -0,0 +1,45 @@
+/* Software floating-point emulation.
+   Convert a 32bit signed integer to IEEE half
+   Copyright (C) 2022 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   In addition to the permissions in the GNU Lesser General Public
+   License, the Free Software Foundation gives you unlimited
+   permission to link the compiled version of this file into
+   combinations with other programs, and to distribute those
+   combinations without any restriction coming from the use of this
+   file.  (The Lesser General Public License restrictions do apply in
+   other respects; for example, they cover modification of the file,
+   and distribution when not linked into a combine executable.)
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include "soft-fp.h"
+#include "half.h"
+
+HFtype
+__floatsihf (SItype i)
+{
+  FP_DECL_EX;
+  FP_DECL_H (A);
+  HFtype a;
+
+  FP_INIT_ROUNDMODE;
+  FP_FROM_INT_H (A, i, SI_BITS, USItype);
+  FP_PACK_RAW_H (a, A);
+  FP_HANDLE_EXCEPTIONS;
+
+  return a;
+}
diff --git a/soft-fp/floatundihf.c b/soft-fp/floatundihf.c
new file mode 100644
index 0000000000..f325a8099e
--- /dev/null
+++ b/soft-fp/floatundihf.c
@@ -0,0 +1,45 @@
+/* Software floating-point emulation.
+   Convert a 64bit unsigned integer to IEEE half.
+   Copyright (C) 2022 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   In addition to the permissions in the GNU Lesser General Public
+   License, the Free Software Foundation gives you unlimited
+   permission to link the compiled version of this file into
+   combinations with other programs, and to distribute those
+   combinations without any restriction coming from the use of this
+   file.  (The Lesser General Public License restrictions do apply in
+   other respects; for example, they cover modification of the file,
+   and distribution when not linked into a combine executable.)
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include "soft-fp.h"
+#include "half.h"
+
+HFtype
+__floatundihf (UDItype i)
+{
+  FP_DECL_EX;
+  FP_DECL_H (A);
+  HFtype a;
+
+  FP_INIT_ROUNDMODE;
+  FP_FROM_INT_H (A, i, DI_BITS, UDItype);
+  FP_PACK_RAW_H (a, A);
+  FP_HANDLE_EXCEPTIONS;
+
+  return a;
+}
diff --git a/soft-fp/floatunsihf.c b/soft-fp/floatunsihf.c
new file mode 100644
index 0000000000..8d690a038f
--- /dev/null
+++ b/soft-fp/floatunsihf.c
@@ -0,0 +1,45 @@
+/* Software floating-point emulation.
+   Convert a 32bit unsigned integer to IEEE half.
+   Copyright (C) 2022 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   In addition to the permissions in the GNU Lesser General Public
+   License, the Free Software Foundation gives you unlimited
+   permission to link the compiled version of this file into
+   combinations with other programs, and to distribute those
+   combinations without any restriction coming from the use of this
+   file.  (The Lesser General Public License restrictions do apply in
+   other respects; for example, they cover modification of the file,
+   and distribution when not linked into a combine executable.)
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include "soft-fp.h"
+#include "half.h"
+
+HFtype
+__floatunsihf (USItype i)
+{
+  FP_DECL_EX;
+  FP_DECL_H (A);
+  HFtype a;
+
+  FP_INIT_ROUNDMODE;
+  FP_FROM_INT_H (A, i, SI_BITS, USItype);
+  FP_PACK_RAW_H (a, A);
+  FP_HANDLE_EXCEPTIONS;
+
+  return a;
+}
-- 
2.34.0

Re: [PATCH] soft-fp: Add fixhf[uns][di|si] and float[uns][di|si]hf

[Joseph Myers]

On Tue, 5 Jul 2022, Kito Cheng wrote:

> Add more IEEE half conversion routines:
> 
> - Convert 32-bit/64-bit integer to IEEE half.
> - Convert IEEE half to 32-bit/64-bit integer.
> 
> They are required by RISC-V _Float16 support, tested with RISC-V GCC.

OK.

-- 
Joseph S. Myers
joseph@codesourcery.com

Re: [PATCH] soft-fp: Add fixhf[uns][di|si] and float[uns][di|si]hf

[Kito Cheng]

Hi Joseph:

I didn't have commit access for glibc, could you help me to commit that? thanks!

On Wed, Jul 20, 2022 at 4:59 AM Joseph Myers <joseph@codesourcery.com> wrote:
>
> On Tue, 5 Jul 2022, Kito Cheng wrote:
>
> > Add more IEEE half conversion routines:
> >
> > - Convert 32-bit/64-bit integer to IEEE half.
> > - Convert IEEE half to 32-bit/64-bit integer.
> >
> > They are required by RISC-V _Float16 support, tested with RISC-V GCC.
>
> OK.
>
> --
> Joseph S. Myers
> joseph@codesourcery.com

Re: [PATCH] soft-fp: Add fixhf[uns][di|si] and float[uns][di|si]hf

[Adhemerval Zanella Netto]

I will take care of it.

On 21/07/22 05:55, Kito Cheng via Libc-alpha wrote:
> Hi Joseph:
> 
> I didn't have commit access for glibc, could you help me to commit that? thanks!
> 
> On Wed, Jul 20, 2022 at 4:59 AM Joseph Myers <joseph@codesourcery.com> wrote:
>>
>> On Tue, 5 Jul 2022, Kito Cheng wrote:
>>
>>> Add more IEEE half conversion routines:
>>>
>>> - Convert 32-bit/64-bit integer to IEEE half.
>>> - Convert IEEE half to 32-bit/64-bit integer.
>>>
>>> They are required by RISC-V _Float16 support, tested with RISC-V GCC.
>>
>> OK.
>>
>> --
>> Joseph S. Myers
>> joseph@codesourcery.com

Re: [PATCH] soft-fp: Add fixhf[uns][di|si] and float[uns][di|si]hf

[Kito Cheng]

Hi Adhemerval:

Could you commit this, 2.36 has been branch out, so this should be OK
for master now ?

Thanks!


On Thu, Jul 21, 2022 at 7:34 PM Adhemerval Zanella Netto via
Libc-alpha <libc-alpha@sourceware.org> wrote:
>
> I will take care of it.
>
> On 21/07/22 05:55, Kito Cheng via Libc-alpha wrote:
> > Hi Joseph:
> >
> > I didn't have commit access for glibc, could you help me to commit that? thanks!
> >
> > On Wed, Jul 20, 2022 at 4:59 AM Joseph Myers <joseph@codesourcery.com> wrote:
> >>
> >> On Tue, 5 Jul 2022, Kito Cheng wrote:
> >>
> >>> Add more IEEE half conversion routines:
> >>>
> >>> - Convert 32-bit/64-bit integer to IEEE half.
> >>> - Convert IEEE half to 32-bit/64-bit integer.
> >>>
> >>> They are required by RISC-V _Float16 support, tested with RISC-V GCC.
> >>
> >> OK.
> >>
> >> --
> >> Joseph S. Myers
> >> joseph@codesourcery.com

Re: [PATCH] soft-fp: Add fixhf[uns][di|si] and float[uns][di|si]hf

[Adhemerval Zanella Netto]

On 08/08/22 04:29, Kito Cheng wrote:
> Hi Adhemerval:
> 
> Could you commit this, 2.36 has been branch out, so this should be OK
> for master now ?
> 
> Thanks!

I will do it, thanks for remind me.

> 
> 
> On Thu, Jul 21, 2022 at 7:34 PM Adhemerval Zanella Netto via
> Libc-alpha <libc-alpha@sourceware.org> wrote:
>>
>> I will take care of it.
>>
>> On 21/07/22 05:55, Kito Cheng via Libc-alpha wrote:
>>> Hi Joseph:
>>>
>>> I didn't have commit access for glibc, could you help me to commit that? thanks!
>>>
>>> On Wed, Jul 20, 2022 at 4:59 AM Joseph Myers <joseph@codesourcery.com> wrote:
>>>>
>>>> On Tue, 5 Jul 2022, Kito Cheng wrote:
>>>>
>>>>> Add more IEEE half conversion routines:
>>>>>
>>>>> - Convert 32-bit/64-bit integer to IEEE half.
>>>>> - Convert IEEE half to 32-bit/64-bit integer.
>>>>>
>>>>> They are required by RISC-V _Float16 support, tested with RISC-V GCC.
>>>>
>>>> OK.
>>>>
>>>> --
>>>> Joseph S. Myers
>>>> joseph@codesourcery.com