Hi!
While DI <-> BF conversions can be handled (and are) through
DI <-> XF <-> BF and for narrower integral modes even sometimes
through DF or SF, because XFmode has 64-bit mantissa and so all
the DImode values are exactly representable in XFmode.
That is not the case for TImode, and while e.g. the HF -> TI
conversions are IMHO useless in libgcc, because HFmode has
-65504.0f16, 65504.0f16 range, all the integers will be already
representable in SImode (or even HImode for unsigned) and so
I think HF -> DI -> TI conversions are faster and valid,
BFmode has roughly the same range as SFmode and so we absolutely need
the TI -> BF conversions to avoid double rounding.
As for BF -> TI conversions, they can be either also implemented
in libgcc (as done in the patch), or they could be implemented
as BF -> SF -> TI conversions with the same code generation used
elsewhere, just doing the 16-bit left shift of the bits - I think
we don't need to handle sNaNs during the BF -> SF part because
SF -> TI (which is already a libcall too) will handle that too.
Thoughts on this? I guess my preference would be the BF -> SF -> TI
path because we won't need to waste
32: 0000000000015e10 321 FUNC GLOBAL DEFAULT 13 __fixbfti@@GCC_13.0.0
89: 0000000000015f60 299 FUNC GLOBAL DEFAULT 13
__fixunsbfti@@GCC_13.0.0
If so, I'd need to cut the fix parts of the patch below and
do something in the middle-end.
2022-11-16 Jakub Jelinek <ja...@redhat.com>
PR target/107703
* soft-fp/fixbfti.c: New file.
* soft-fp/fixunsbfti.c: New file.
* soft-fp/floattibf.c: New file.
* soft-fp/floatuntibf.c: New file.
* config/i386/libgcc-glibc.ver: Export __fix{,uns}bfti
and __float{,un}tibf @ GCC_13.0.0.
* config/i386/64/t-softfp (softfp_extras): Add
fixbfti, fixunsbfti, floattibf and floatuntibf.
(CFLAGS-fixbfti.c, CFLAGS-fixunsbfti.c, CFLAGS-floattibf.c,
CFLAGS-floatunstibf.c): Add -msse2.
--- libgcc/soft-fp/fixbfti.c.jj 2022-11-15 19:07:41.747901840 +0100
+++ libgcc/soft-fp/fixbfti.c 2022-11-15 19:13:45.990927498 +0100
@@ -0,0 +1,45 @@
+/* Software floating-point emulation.
+ Convert bfloat16 to 128bit signed integer
+ Copyright (C) 2007-2022 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
+ <https://www.gnu.org/licenses/>. */
+
+#include "soft-fp.h"
+#include "brain.h"
+
+TItype
+__fixbfti (BFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_B (A);
+ UTItype r;
+
+ FP_INIT_EXCEPTIONS;
+ FP_UNPACK_RAW_B (A, a);
+ FP_TO_INT_B (r, A, TI_BITS, 1);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
--- libgcc/soft-fp/fixunsbfti.c.jj 2022-11-15 19:08:24.400319352 +0100
+++ libgcc/soft-fp/fixunsbfti.c 2022-11-15 19:13:50.200870008 +0100
@@ -0,0 +1,45 @@
+/* Software floating-point emulation.
+ Convert bfloat16 to 128bit unsigned integer
+ Copyright (C) 2007-2022 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
+ <https://www.gnu.org/licenses/>. */
+
+#include "soft-fp.h"
+#include "brain.h"
+
+UTItype
+__fixunsbfti (BFtype a)
+{
+ FP_DECL_EX;
+ FP_DECL_B (A);
+ UTItype r;
+
+ FP_INIT_EXCEPTIONS;
+ FP_UNPACK_RAW_B (A, a);
+ FP_TO_INT_B (r, A, TI_BITS, 0);
+ FP_HANDLE_EXCEPTIONS;
+
+ return r;
+}
--- libgcc/soft-fp/floattibf.c.jj 2022-11-15 19:10:13.147834226 +0100
+++ libgcc/soft-fp/floattibf.c 2022-11-15 19:13:40.939996482 +0100
@@ -0,0 +1,45 @@
+/* Software floating-point emulation.
+ Convert a 128bit signed integer to bfloat16
+ Copyright (C) 2007-2022 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
+ <https://www.gnu.org/licenses/>. */
+
+#include "soft-fp.h"
+#include "brain.h"
+
+BFtype
+__floattibf (TItype i)
+{
+ FP_DECL_EX;
+ FP_DECL_B (A);
+ BFtype a;
+
+ FP_INIT_ROUNDMODE;
+ FP_FROM_INT_B (A, i, TI_BITS, UTItype);
+ FP_PACK_RAW_B (a, A);
+ FP_HANDLE_EXCEPTIONS;
+
+ return a;
+}
--- libgcc/soft-fp/floatuntibf.c.jj 2022-11-15 19:10:53.363285018 +0100
+++ libgcc/soft-fp/floatuntibf.c 2022-11-15 19:13:31.399126765 +0100
@@ -0,0 +1,45 @@
+/* Software floating-point emulation.
+ Convert a 128bit unsigned integer to bfloat16
+ Copyright (C) 2007-2022 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
+ <https://www.gnu.org/licenses/>. */
+
+#include "soft-fp.h"
+#include "brain.h"
+
+BFtype
+__floatuntibf (UTItype i)
+{
+ FP_DECL_EX;
+ FP_DECL_B (A);
+ BFtype a;
+
+ FP_INIT_ROUNDMODE;
+ FP_FROM_INT_B (A, i, TI_BITS, UTItype);
+ FP_PACK_RAW_B (a, A);
+ FP_HANDLE_EXCEPTIONS;
+
+ return a;
+}
--- libgcc/config/i386/libgcc-glibc.ver.jj 2022-10-14 09:35:56.269989297
+0200
+++ libgcc/config/i386/libgcc-glibc.ver 2022-11-15 19:26:37.125389987 +0100
@@ -218,6 +218,10 @@ GCC_12.0.0 {
%inherit GCC_13.0.0 GCC_12.0.0
GCC_13.0.0 {
__extendbfsf2
+ __fixbfti
+ __fixunsbfti
+ __floattibf
+ __floatuntibf
__truncdfbf2
__truncsfbf2
__trunctfbf2
--- libgcc/config/i386/64/t-softfp.jj 2021-12-30 15:12:44.111138056 +0100
+++ libgcc/config/i386/64/t-softfp 2022-11-15 19:26:07.184799287 +0100
@@ -1,6 +1,11 @@
-softfp_extras := fixhfti fixunshfti floattihf floatuntihf
+softfp_extras := fixhfti fixunshfti floattihf floatuntihf \
+ fixbfti fixunsbfti floattibf floatuntibf
CFLAGS-fixhfti.c += -msse2
CFLAGS-fixunshfti.c += -msse2
CFLAGS-floattihf.c += -msse2
CFLAGS-floatunstihf.c += -msse2
+CFLAGS-fixbfti.c += -msse2
+CFLAGS-fixunsbfti.c += -msse2
+CFLAGS-floattibf.c += -msse2
+CFLAGS-floatunstibf.c += -msse2
Jakub
