* [PATCH 0/9] LoongArch: Use builtins with GCC >= 13
@ 2022-11-18 11:29 Xiaolin Tang
2022-11-22 16:20 ` Xi Ruoyao via Libc-alpha
0 siblings, 1 reply; 9+ messages in thread
From: Xiaolin Tang @ 2022-11-18 11:29 UTC (permalink / raw)
To: adhemerval.zanella, libc-alpha, caiyinyu, xry111
Cc: xuchenghua, chenglulu, Xiaolin Tang
Patch 0001-0007:
GCC 13, LoongArch support builtins for {,l,ll}rint, logb,
use these built-ins instead of generic function implementation.
Patch 0008:
GCC 13, LoongArch support ___builtin_scalbn{,f} with -fno-math-errno,
but only "libm" can use -fno-math-errno in GLIBC, and scalbn is in libc
instead of libm because __printf_fp calls it. So inline assembler of
the scalbn functions is a better choice for LoongArch.
Patch 0009: Add LoongArch support for scalb[f].
Discuss with Xi Ruoyao<xry111@xry111.site>
Link:
https://sourceware.org/pipermail/libc-alpha/2022-November/143425.html
Xi Ruoyao (1):
LoongArch: Use __builtin_rint{,f} with GCC >= 13
Xiaolin Tang (8):
Use GCC builtins for lrint functions if desired.
LoongArch: Use __builtin_lrint{,f} with GCC >= 13
Use GCC builtins for llrint functions if desired.
LoongArch: Use __builtin_llrint{,f} with GCC >= 13
Use GCC builtins for logb functions if desired.
LoongArch: Use __builtin_logb{,f} with GCC >= 13
LoongArch: Add support for scalbn[f]
LoongArch: Add support for scalb[f]
sysdeps/generic/math-use-builtins-llrint.h | 4 ++
sysdeps/generic/math-use-builtins-logb.h | 4 ++
sysdeps/generic/math-use-builtins-lrint.h | 4 ++
sysdeps/generic/math-use-builtins.h | 3 +
sysdeps/ieee754/dbl-64/s_llrint.c | 18 ++++--
sysdeps/ieee754/dbl-64/s_logb.c | 5 ++
sysdeps/ieee754/dbl-64/s_lrint.c | 18 ++++--
sysdeps/ieee754/float128/float128_private.h | 6 ++
sysdeps/ieee754/flt-32/s_llrintf.c | 18 ++++--
sysdeps/ieee754/flt-32/s_logbf.c | 5 ++
sysdeps/ieee754/flt-32/s_lrintf.c | 18 ++++--
sysdeps/ieee754/ldbl-128/s_llrintl.c | 17 ++++--
sysdeps/ieee754/ldbl-128/s_logbl.c | 6 ++
sysdeps/ieee754/ldbl-128/s_lrintl.c | 17 ++++--
sysdeps/loongarch/fpu/e_scalb.c | 60 +++++++++++++++++++
sysdeps/loongarch/fpu/e_scalbf.c | 60 +++++++++++++++++++
.../loongarch/fpu/math-use-builtins-llrint.h | 10 ++++
.../loongarch/fpu/math-use-builtins-logb.h | 10 ++++
.../loongarch/fpu/math-use-builtins-lrint.h | 10 ++++
.../loongarch/fpu/math-use-builtins-rint.h | 9 +++
sysdeps/loongarch/fpu/s_scalbn.c | 29 +++++++++
sysdeps/loongarch/fpu/s_scalbnf.c | 29 +++++++++
22 files changed, 326 insertions(+), 34 deletions(-)
create mode 100644 sysdeps/generic/math-use-builtins-llrint.h
create mode 100644 sysdeps/generic/math-use-builtins-logb.h
create mode 100644 sysdeps/generic/math-use-builtins-lrint.h
create mode 100644 sysdeps/loongarch/fpu/e_scalb.c
create mode 100644 sysdeps/loongarch/fpu/e_scalbf.c
create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-llrint.h
create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-logb.h
create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-lrint.h
create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-rint.h
create mode 100644 sysdeps/loongarch/fpu/s_scalbn.c
create mode 100644 sysdeps/loongarch/fpu/s_scalbnf.c
--
2.20.1
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH 0/9] LoongArch: Use builtins with GCC >= 13
2022-11-18 11:29 [PATCH 0/9] LoongArch: Use builtins with GCC >= 13 Xiaolin Tang
@ 2022-11-22 16:20 ` Xi Ruoyao via Libc-alpha
2022-11-23 2:14 ` tangxiaolin
2022-11-23 2:55 ` tangxiaolin
0 siblings, 2 replies; 9+ messages in thread
From: Xi Ruoyao via Libc-alpha @ 2022-11-22 16:20 UTC (permalink / raw)
To: Xiaolin Tang, adhemerval.zanella, libc-alpha, caiyinyu
Cc: xuchenghua, chenglulu
Hi Xiaolin,
I guess you need to resent the series because {1,8,9}/9 lacks the proper
In-Reply-To: header, causing them missing from the patchwork interface
(https://patchwork.sourceware.org/project/glibc/list/?series=14494).
The reviewers will use the patchwork interface to track the patches.
On Fri, 2022-11-18 at 19:29 +0800, Xiaolin Tang wrote:
> Patch 0001-0007:
> GCC 13, LoongArch support builtins for {,l,ll}rint, logb,
> use these built-ins instead of generic function implementation.
>
> Patch 0008:
> GCC 13, LoongArch support ___builtin_scalbn{,f} with -fno-math-errno,
> but only "libm" can use -fno-math-errno in GLIBC, and scalbn is in
> libc
> instead of libm because __printf_fp calls it. So inline assembler of
> the scalbn functions is a better choice for LoongArch.
>
> Patch 0009: Add LoongArch support for scalb[f].
>
> Discuss with Xi Ruoyao<xry111@xry111.site>
> Link:
> https://sourceware.org/pipermail/libc-alpha/2022-November/143425.html
>
> Xi Ruoyao (1):
> LoongArch: Use __builtin_rint{,f} with GCC >= 13
>
> Xiaolin Tang (8):
> Use GCC builtins for lrint functions if desired.
> LoongArch: Use __builtin_lrint{,f} with GCC >= 13
> Use GCC builtins for llrint functions if desired.
> LoongArch: Use __builtin_llrint{,f} with GCC >= 13
> Use GCC builtins for logb functions if desired.
> LoongArch: Use __builtin_logb{,f} with GCC >= 13
> LoongArch: Add support for scalbn[f]
> LoongArch: Add support for scalb[f]
>
> sysdeps/generic/math-use-builtins-llrint.h | 4 ++
> sysdeps/generic/math-use-builtins-logb.h | 4 ++
> sysdeps/generic/math-use-builtins-lrint.h | 4 ++
> sysdeps/generic/math-use-builtins.h | 3 +
> sysdeps/ieee754/dbl-64/s_llrint.c | 18 ++++--
> sysdeps/ieee754/dbl-64/s_logb.c | 5 ++
> sysdeps/ieee754/dbl-64/s_lrint.c | 18 ++++--
> sysdeps/ieee754/float128/float128_private.h | 6 ++
> sysdeps/ieee754/flt-32/s_llrintf.c | 18 ++++--
> sysdeps/ieee754/flt-32/s_logbf.c | 5 ++
> sysdeps/ieee754/flt-32/s_lrintf.c | 18 ++++--
> sysdeps/ieee754/ldbl-128/s_llrintl.c | 17 ++++--
> sysdeps/ieee754/ldbl-128/s_logbl.c | 6 ++
> sysdeps/ieee754/ldbl-128/s_lrintl.c | 17 ++++--
> sysdeps/loongarch/fpu/e_scalb.c | 60
> +++++++++++++++++++
> sysdeps/loongarch/fpu/e_scalbf.c | 60
> +++++++++++++++++++
> .../loongarch/fpu/math-use-builtins-llrint.h | 10 ++++
> .../loongarch/fpu/math-use-builtins-logb.h | 10 ++++
> .../loongarch/fpu/math-use-builtins-lrint.h | 10 ++++
> .../loongarch/fpu/math-use-builtins-rint.h | 9 +++
> sysdeps/loongarch/fpu/s_scalbn.c | 29 +++++++++
> sysdeps/loongarch/fpu/s_scalbnf.c | 29 +++++++++
> 22 files changed, 326 insertions(+), 34 deletions(-)
> create mode 100644 sysdeps/generic/math-use-builtins-llrint.h
> create mode 100644 sysdeps/generic/math-use-builtins-logb.h
> create mode 100644 sysdeps/generic/math-use-builtins-lrint.h
> create mode 100644 sysdeps/loongarch/fpu/e_scalb.c
> create mode 100644 sysdeps/loongarch/fpu/e_scalbf.c
> create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-llrint.h
> create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-logb.h
> create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-lrint.h
> create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-rint.h
> create mode 100644 sysdeps/loongarch/fpu/s_scalbn.c
> create mode 100644 sysdeps/loongarch/fpu/s_scalbnf.c
>
--
Xi Ruoyao <xry111@xry111.site>
School of Aerospace Science and Technology, Xidian University
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH 0/9] LoongArch: Use builtins with GCC >= 13
2022-11-22 16:20 ` Xi Ruoyao via Libc-alpha
@ 2022-11-23 2:14 ` tangxiaolin
2022-11-23 2:55 ` tangxiaolin
1 sibling, 0 replies; 9+ messages in thread
From: tangxiaolin @ 2022-11-23 2:14 UTC (permalink / raw)
To: Xi Ruoyao, adhemerval.zanella, libc-alpha, caiyinyu; +Cc: xuchenghua, chenglulu
OK, thank you for pointing this out. Please delete them, and I will
resend this series.
在 2022/11/23 上午12:20, Xi Ruoyao 写道:
> Hi Xiaolin,
>
> I guess you need to resent the series because {1,8,9}/9 lacks the proper
> In-Reply-To: header, causing them missing from the patchwork interface
> (https://patchwork.sourceware.org/project/glibc/list/?series=14494).
> The reviewers will use the patchwork interface to track the patches.
>
> On Fri, 2022-11-18 at 19:29 +0800, Xiaolin Tang wrote:
>> Patch 0001-0007:
>> GCC 13, LoongArch support builtins for {,l,ll}rint, logb,
>> use these built-ins instead of generic function implementation.
>>
>> Patch 0008:
>> GCC 13, LoongArch support ___builtin_scalbn{,f} with -fno-math-errno,
>> but only "libm" can use -fno-math-errno in GLIBC, and scalbn is in
>> libc
>> instead of libm because __printf_fp calls it. So inline assembler of
>> the scalbn functions is a better choice for LoongArch.
>>
>> Patch 0009: Add LoongArch support for scalb[f].
>>
>> Discuss with Xi Ruoyao<xry111@xry111.site>
>> Link:
>> https://sourceware.org/pipermail/libc-alpha/2022-November/143425.html
>>
>> Xi Ruoyao (1):
>> LoongArch: Use __builtin_rint{,f} with GCC >= 13
>>
>> Xiaolin Tang (8):
>> Use GCC builtins for lrint functions if desired.
>> LoongArch: Use __builtin_lrint{,f} with GCC >= 13
>> Use GCC builtins for llrint functions if desired.
>> LoongArch: Use __builtin_llrint{,f} with GCC >= 13
>> Use GCC builtins for logb functions if desired.
>> LoongArch: Use __builtin_logb{,f} with GCC >= 13
>> LoongArch: Add support for scalbn[f]
>> LoongArch: Add support for scalb[f]
>>
>> sysdeps/generic/math-use-builtins-llrint.h | 4 ++
>> sysdeps/generic/math-use-builtins-logb.h | 4 ++
>> sysdeps/generic/math-use-builtins-lrint.h | 4 ++
>> sysdeps/generic/math-use-builtins.h | 3 +
>> sysdeps/ieee754/dbl-64/s_llrint.c | 18 ++++--
>> sysdeps/ieee754/dbl-64/s_logb.c | 5 ++
>> sysdeps/ieee754/dbl-64/s_lrint.c | 18 ++++--
>> sysdeps/ieee754/float128/float128_private.h | 6 ++
>> sysdeps/ieee754/flt-32/s_llrintf.c | 18 ++++--
>> sysdeps/ieee754/flt-32/s_logbf.c | 5 ++
>> sysdeps/ieee754/flt-32/s_lrintf.c | 18 ++++--
>> sysdeps/ieee754/ldbl-128/s_llrintl.c | 17 ++++--
>> sysdeps/ieee754/ldbl-128/s_logbl.c | 6 ++
>> sysdeps/ieee754/ldbl-128/s_lrintl.c | 17 ++++--
>> sysdeps/loongarch/fpu/e_scalb.c | 60
>> +++++++++++++++++++
>> sysdeps/loongarch/fpu/e_scalbf.c | 60
>> +++++++++++++++++++
>> .../loongarch/fpu/math-use-builtins-llrint.h | 10 ++++
>> .../loongarch/fpu/math-use-builtins-logb.h | 10 ++++
>> .../loongarch/fpu/math-use-builtins-lrint.h | 10 ++++
>> .../loongarch/fpu/math-use-builtins-rint.h | 9 +++
>> sysdeps/loongarch/fpu/s_scalbn.c | 29 +++++++++
>> sysdeps/loongarch/fpu/s_scalbnf.c | 29 +++++++++
>> 22 files changed, 326 insertions(+), 34 deletions(-)
>> create mode 100644 sysdeps/generic/math-use-builtins-llrint.h
>> create mode 100644 sysdeps/generic/math-use-builtins-logb.h
>> create mode 100644 sysdeps/generic/math-use-builtins-lrint.h
>> create mode 100644 sysdeps/loongarch/fpu/e_scalb.c
>> create mode 100644 sysdeps/loongarch/fpu/e_scalbf.c
>> create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-llrint.h
>> create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-logb.h
>> create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-lrint.h
>> create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-rint.h
>> create mode 100644 sysdeps/loongarch/fpu/s_scalbn.c
>> create mode 100644 sysdeps/loongarch/fpu/s_scalbnf.c
>>
^ permalink raw reply [flat|nested] 9+ messages in thread
* [PATCH 0/9] LoongArch: Use builtins with GCC >= 13
@ 2022-11-23 2:38 Xiaolin Tang
2022-11-23 2:38 ` [PATCH 1/9] LoongArch: Use __builtin_rint{,f} " Xiaolin Tang
` (2 more replies)
0 siblings, 3 replies; 9+ messages in thread
From: Xiaolin Tang @ 2022-11-23 2:38 UTC (permalink / raw)
To: adhemerval.zanella, libc-alpha, caiyinyu, xry111
Cc: xuchenghua, chenglulu, Xiaolin Tang
Patch 0001-0007:
GCC 13, LoongArch support builtins for {,l,ll}rint, logb,
use these built-ins instead of generic function implementation.
Patch 0008:
GCC 13, LoongArch support ___builtin_scalbn{,f} with -fno-math-errno,
but only "libm" can use -fno-math-errno in GLIBC, and scalbn is in libc
instead of libm because __printf_fp calls it. So inline assembler of
the scalbn functions is a better choice for LoongArch.
Patch 0009: Add LoongArch support for scalb[f].
Discuss with Xi Ruoyao<xry111@xry111.site>
Link:
https://sourceware.org/pipermail/libc-alpha/2022-November/143425.html
Xi Ruoyao (1):
LoongArch: Use __builtin_rint{,f} with GCC >= 13
Xiaolin Tang (8):
Use GCC builtins for lrint functions if desired.
LoongArch: Use __builtin_lrint{,f} with GCC >= 13
Use GCC builtins for llrint functions if desired.
LoongArch: Use __builtin_llrint{,f} with GCC >= 13
Use GCC builtins for logb functions if desired.
LoongArch: Use __builtin_logb{,f} with GCC >= 13
LoongArch: Add support for scalbn[f]
LoongArch: Add support for scalb[f]
sysdeps/generic/math-use-builtins-llrint.h | 4 ++
sysdeps/generic/math-use-builtins-logb.h | 4 ++
sysdeps/generic/math-use-builtins-lrint.h | 4 ++
sysdeps/generic/math-use-builtins.h | 3 +
sysdeps/ieee754/dbl-64/s_llrint.c | 18 ++++--
sysdeps/ieee754/dbl-64/s_logb.c | 5 ++
sysdeps/ieee754/dbl-64/s_lrint.c | 18 ++++--
sysdeps/ieee754/float128/float128_private.h | 6 ++
sysdeps/ieee754/flt-32/s_llrintf.c | 18 ++++--
sysdeps/ieee754/flt-32/s_logbf.c | 5 ++
sysdeps/ieee754/flt-32/s_lrintf.c | 18 ++++--
sysdeps/ieee754/ldbl-128/s_llrintl.c | 17 ++++--
sysdeps/ieee754/ldbl-128/s_logbl.c | 6 ++
sysdeps/ieee754/ldbl-128/s_lrintl.c | 17 ++++--
sysdeps/loongarch/fpu/e_scalb.c | 60 +++++++++++++++++++
sysdeps/loongarch/fpu/e_scalbf.c | 60 +++++++++++++++++++
.../loongarch/fpu/math-use-builtins-llrint.h | 10 ++++
.../loongarch/fpu/math-use-builtins-logb.h | 10 ++++
.../loongarch/fpu/math-use-builtins-lrint.h | 10 ++++
.../loongarch/fpu/math-use-builtins-rint.h | 9 +++
sysdeps/loongarch/fpu/s_scalbn.c | 29 +++++++++
sysdeps/loongarch/fpu/s_scalbnf.c | 29 +++++++++
22 files changed, 326 insertions(+), 34 deletions(-)
create mode 100644 sysdeps/generic/math-use-builtins-llrint.h
create mode 100644 sysdeps/generic/math-use-builtins-logb.h
create mode 100644 sysdeps/generic/math-use-builtins-lrint.h
create mode 100644 sysdeps/loongarch/fpu/e_scalb.c
create mode 100644 sysdeps/loongarch/fpu/e_scalbf.c
create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-llrint.h
create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-logb.h
create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-lrint.h
create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-rint.h
create mode 100644 sysdeps/loongarch/fpu/s_scalbn.c
create mode 100644 sysdeps/loongarch/fpu/s_scalbnf.c
--
2.20.1
^ permalink raw reply [flat|nested] 9+ messages in thread
* [PATCH 1/9] LoongArch: Use __builtin_rint{,f} with GCC >= 13
2022-11-23 2:38 [PATCH 0/9] LoongArch: Use builtins with GCC >= 13 Xiaolin Tang
@ 2022-11-23 2:38 ` Xiaolin Tang
2022-11-23 2:38 ` [PATCH 8/9] LoongArch: Add support for scalbn[f] Xiaolin Tang
2022-11-23 2:38 ` [PATCH 9/9] LoongArch: Add support for scalb[f] Xiaolin Tang
2 siblings, 0 replies; 9+ messages in thread
From: Xiaolin Tang @ 2022-11-23 2:38 UTC (permalink / raw)
To: adhemerval.zanella, libc-alpha, caiyinyu, xry111; +Cc: xuchenghua, chenglulu
From: Xi Ruoyao <xry111@xry111.site>
GCC 13 compiles these built-ins to frint.{d,s} instruction.
Link: https://gcc.gnu.org/r13-3919
---
sysdeps/loongarch/fpu/math-use-builtins-rint.h | 9 +++++++++
1 file changed, 9 insertions(+)
create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-rint.h
diff --git a/sysdeps/loongarch/fpu/math-use-builtins-rint.h b/sysdeps/loongarch/fpu/math-use-builtins-rint.h
new file mode 100644
index 0000000000..cd91482f5f
--- /dev/null
+++ b/sysdeps/loongarch/fpu/math-use-builtins-rint.h
@@ -0,0 +1,9 @@
+#if __GNUC_PREREQ (13, 0)
+# define USE_RINT_BUILTIN 1
+# define USE_RINTF_BUILTIN 1
+#else
+# define USE_RINT_BUILTIN 0
+# define USE_RINTF_BUILTIN 0
+#endif
+#define USE_RINTL_BUILTIN 0
+#define USE_RINTF128_BUILTIN 0
--
2.20.1
^ permalink raw reply related [flat|nested] 9+ messages in thread
* [PATCH 8/9] LoongArch: Add support for scalbn[f]
2022-11-23 2:38 [PATCH 0/9] LoongArch: Use builtins with GCC >= 13 Xiaolin Tang
2022-11-23 2:38 ` [PATCH 1/9] LoongArch: Use __builtin_rint{,f} " Xiaolin Tang
@ 2022-11-23 2:38 ` Xiaolin Tang
2022-11-23 2:38 ` [PATCH 9/9] LoongArch: Add support for scalb[f] Xiaolin Tang
2 siblings, 0 replies; 9+ messages in thread
From: Xiaolin Tang @ 2022-11-23 2:38 UTC (permalink / raw)
To: adhemerval.zanella, libc-alpha, caiyinyu, xry111
Cc: xuchenghua, chenglulu, Xiaolin Tang
Add inline assembler for the scalbn functions. Passes GLIBC regression.
GCC 13, LoongArch support ___builtin_scalbn{,f} with -fno-math-errno,
but only "libm" can use -fno-math-errno in GLIBC, and scalbn is in libc
instead of libm because __printf_fp calls it.
---
sysdeps/loongarch/fpu/s_scalbn.c | 29 +++++++++++++++++++++++++++++
sysdeps/loongarch/fpu/s_scalbnf.c | 29 +++++++++++++++++++++++++++++
2 files changed, 58 insertions(+)
create mode 100644 sysdeps/loongarch/fpu/s_scalbn.c
create mode 100644 sysdeps/loongarch/fpu/s_scalbnf.c
diff --git a/sysdeps/loongarch/fpu/s_scalbn.c b/sysdeps/loongarch/fpu/s_scalbn.c
new file mode 100644
index 0000000000..c03e81a31e
--- /dev/null
+++ b/sysdeps/loongarch/fpu/s_scalbn.c
@@ -0,0 +1,29 @@
+/* scalbn(). LoongArch version.
+ 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.
+
+ 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/>. */
+
+#define NO_MATH_REDIRECT
+#include <math.h>
+
+double
+__scalbn (double x, int fn)
+{
+ double tmp;
+ asm volatile ("movgr2fr.d \t%0, %1" : "=f" (tmp) : "r" (fn));
+ asm volatile ("fscaleb.d \t%0, %1, %2" : "=f" (x) : "f" (x), "f" (tmp));
+ return x;
+}
diff --git a/sysdeps/loongarch/fpu/s_scalbnf.c b/sysdeps/loongarch/fpu/s_scalbnf.c
new file mode 100644
index 0000000000..15e6428044
--- /dev/null
+++ b/sysdeps/loongarch/fpu/s_scalbnf.c
@@ -0,0 +1,29 @@
+/* scalbnf(). LoongArch version.
+ 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.
+
+ 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/>. */
+
+#define NO_MATH_REDIRECT
+#include <math.h>
+
+float
+__scalbnf (float x, int fn)
+{
+ float tmp;
+ asm volatile ("movgr2fr.w \t%0, %1" : "=f" (tmp) : "r" (fn));
+ asm volatile ("fscaleb.s \t%0, %1, %2" : "=f" (x) : "f" (x), "f" (tmp));
+ return x;
+}
--
2.20.1
^ permalink raw reply related [flat|nested] 9+ messages in thread
* [PATCH 9/9] LoongArch: Add support for scalb[f]
2022-11-23 2:38 [PATCH 0/9] LoongArch: Use builtins with GCC >= 13 Xiaolin Tang
2022-11-23 2:38 ` [PATCH 1/9] LoongArch: Use __builtin_rint{,f} " Xiaolin Tang
2022-11-23 2:38 ` [PATCH 8/9] LoongArch: Add support for scalbn[f] Xiaolin Tang
@ 2022-11-23 2:38 ` Xiaolin Tang
2 siblings, 0 replies; 9+ messages in thread
From: Xiaolin Tang @ 2022-11-23 2:38 UTC (permalink / raw)
To: adhemerval.zanella, libc-alpha, caiyinyu, xry111
Cc: xuchenghua, chenglulu, Xiaolin Tang
Add inline assembler for the scalb functions. Passes GLIBC regression.
---
sysdeps/loongarch/fpu/e_scalb.c | 60 ++++++++++++++++++++++++++++++++
sysdeps/loongarch/fpu/e_scalbf.c | 60 ++++++++++++++++++++++++++++++++
2 files changed, 120 insertions(+)
create mode 100644 sysdeps/loongarch/fpu/e_scalb.c
create mode 100644 sysdeps/loongarch/fpu/e_scalbf.c
diff --git a/sysdeps/loongarch/fpu/e_scalb.c b/sysdeps/loongarch/fpu/e_scalb.c
new file mode 100644
index 0000000000..9fdf21fa3a
--- /dev/null
+++ b/sysdeps/loongarch/fpu/e_scalb.c
@@ -0,0 +1,60 @@
+/* scalb(). LoongArch version.
+ 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.
+
+ 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/>. */
+
+#define NO_MATH_REDIRECT
+#include <math.h>
+#include <libm-alias-finite.h>
+#include <fpu_control.h>
+#include <float.h>
+
+double
+__ieee754_scalb (double x, double fn)
+{
+ int x_cond;
+ int fn_cond;
+ asm volatile ("fclass.d \t%0, %1" : "=f" (x_cond) : "f" (x));
+ asm volatile ("fclass.d \t%0, %1" : "=f" (fn_cond) : "f" (fn));
+
+ if (__glibc_unlikely(( x_cond | fn_cond) & _FCLASS_NAN))
+ return x * fn;
+ else if (__glibc_unlikely(fn_cond & _FCLASS_INF))
+ {
+ if (!(fn_cond & _FCLASS_MINF))
+ return x * fn;
+ else
+ return x / -fn;
+ }
+ else if (__glibc_likely(-DBL_MAX < fn && fn < DBL_MAX))
+ {
+ double rint_fn, tmp;
+
+ /* rint_fn = rint(fn) */
+ asm volatile ("frint.d \t%0, %1" : "=f" (rint_fn) : "f" (fn));
+
+ if (rint_fn != fn )
+ return (x - x) / (x - x);
+
+ asm volatile ("ftintrz.l.d \t%0, %1" : "=f" (tmp) : "f" (rint_fn));
+ asm volatile ("fscaleb.d \t%0, %1, %2" : "=f" (x) : "f" (x), "f" (tmp));
+ }
+ else
+ asm volatile ("fscaleb.d \t%0, %1, %2" : "=f" (x) : "f" (x), "f" (fn));
+
+ return x;
+}
+libm_alias_finite (__ieee754_scalb, __scalb)
diff --git a/sysdeps/loongarch/fpu/e_scalbf.c b/sysdeps/loongarch/fpu/e_scalbf.c
new file mode 100644
index 0000000000..fae2e92e20
--- /dev/null
+++ b/sysdeps/loongarch/fpu/e_scalbf.c
@@ -0,0 +1,60 @@
+/* scalbf(). LoongArch version.
+ 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.
+
+ 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/>. */
+
+#define NO_MATH_REDIRECT
+#include <math.h>
+#include <libm-alias-finite.h>
+#include <fpu_control.h>
+#include <float.h>
+
+float
+__ieee754_scalbf (float x, float fn)
+{
+ int x_cond;
+ int fn_cond;
+ asm volatile ("fclass.s \t%0, %1" : "=f" (x_cond) : "f" (x));
+ asm volatile ("fclass.s \t%0, %1" : "=f" (fn_cond) : "f" (fn));
+
+ if (__glibc_unlikely(( x_cond | fn_cond) & _FCLASS_NAN))
+ return x * fn;
+ else if (__glibc_unlikely(fn_cond & _FCLASS_INF))
+ {
+ if (!(fn_cond & _FCLASS_MINF))
+ return x * fn;
+ else
+ return x / -fn;
+ }
+ else if (__glibc_likely(-FLT_MAX < fn && fn < FLT_MAX))
+ {
+ float rintf_fn, tmp;
+
+ /* rintf_fn = rintf(fn) */
+ asm volatile ("frint.s \t%0, %1" : "=f" (rintf_fn) : "f" (fn));
+
+ if (rintf_fn != fn )
+ return (x - x) / (x - x);
+
+ asm volatile ("ftintrz.w.s \t%0, %1" : "=f" (tmp) : "f" (rintf_fn));
+ asm volatile ("fscaleb.s \t%0, %1, %2" : "=f" (x) : "f" (x), "f" (tmp));
+ }
+ else
+ asm volatile ("fscaleb.s \t%0, %1, %2" : "=f" (x) : "f" (x), "f" (fn));
+
+ return x;
+}
+libm_alias_finite (__ieee754_scalb, __scalb)
--
2.20.1
^ permalink raw reply related [flat|nested] 9+ messages in thread
* Re: [PATCH 0/9] LoongArch: Use builtins with GCC >= 13
2022-11-22 16:20 ` Xi Ruoyao via Libc-alpha
2022-11-23 2:14 ` tangxiaolin
@ 2022-11-23 2:55 ` tangxiaolin
2022-11-23 3:18 ` Xi Ruoyao via Libc-alpha
1 sibling, 1 reply; 9+ messages in thread
From: tangxiaolin @ 2022-11-23 2:55 UTC (permalink / raw)
To: Xi Ruoyao, adhemerval.zanella, libc-alpha, caiyinyu; +Cc: xuchenghua, chenglulu
Oops, I just resend {1,8,9}/9 with cover-patch, and still haven't solved
this problem.
I will resend this series again with v2 version, which is added ilogb
function.
在 2022/11/23 上午12:20, Xi Ruoyao 写道:
> Hi Xiaolin,
>
> I guess you need to resent the series because {1,8,9}/9 lacks the proper
> In-Reply-To: header, causing them missing from the patchwork interface
> (https://patchwork.sourceware.org/project/glibc/list/?series=14494).
> The reviewers will use the patchwork interface to track the patches.
>
> On Fri, 2022-11-18 at 19:29 +0800, Xiaolin Tang wrote:
>> Patch 0001-0007:
>> GCC 13, LoongArch support builtins for {,l,ll}rint, logb,
>> use these built-ins instead of generic function implementation.
>>
>> Patch 0008:
>> GCC 13, LoongArch support ___builtin_scalbn{,f} with -fno-math-errno,
>> but only "libm" can use -fno-math-errno in GLIBC, and scalbn is in
>> libc
>> instead of libm because __printf_fp calls it. So inline assembler of
>> the scalbn functions is a better choice for LoongArch.
>>
>> Patch 0009: Add LoongArch support for scalb[f].
>>
>> Discuss with Xi Ruoyao<xry111@xry111.site>
>> Link:
>> https://sourceware.org/pipermail/libc-alpha/2022-November/143425.html
>>
>> Xi Ruoyao (1):
>> LoongArch: Use __builtin_rint{,f} with GCC >= 13
>>
>> Xiaolin Tang (8):
>> Use GCC builtins for lrint functions if desired.
>> LoongArch: Use __builtin_lrint{,f} with GCC >= 13
>> Use GCC builtins for llrint functions if desired.
>> LoongArch: Use __builtin_llrint{,f} with GCC >= 13
>> Use GCC builtins for logb functions if desired.
>> LoongArch: Use __builtin_logb{,f} with GCC >= 13
>> LoongArch: Add support for scalbn[f]
>> LoongArch: Add support for scalb[f]
>>
>> sysdeps/generic/math-use-builtins-llrint.h | 4 ++
>> sysdeps/generic/math-use-builtins-logb.h | 4 ++
>> sysdeps/generic/math-use-builtins-lrint.h | 4 ++
>> sysdeps/generic/math-use-builtins.h | 3 +
>> sysdeps/ieee754/dbl-64/s_llrint.c | 18 ++++--
>> sysdeps/ieee754/dbl-64/s_logb.c | 5 ++
>> sysdeps/ieee754/dbl-64/s_lrint.c | 18 ++++--
>> sysdeps/ieee754/float128/float128_private.h | 6 ++
>> sysdeps/ieee754/flt-32/s_llrintf.c | 18 ++++--
>> sysdeps/ieee754/flt-32/s_logbf.c | 5 ++
>> sysdeps/ieee754/flt-32/s_lrintf.c | 18 ++++--
>> sysdeps/ieee754/ldbl-128/s_llrintl.c | 17 ++++--
>> sysdeps/ieee754/ldbl-128/s_logbl.c | 6 ++
>> sysdeps/ieee754/ldbl-128/s_lrintl.c | 17 ++++--
>> sysdeps/loongarch/fpu/e_scalb.c | 60
>> +++++++++++++++++++
>> sysdeps/loongarch/fpu/e_scalbf.c | 60
>> +++++++++++++++++++
>> .../loongarch/fpu/math-use-builtins-llrint.h | 10 ++++
>> .../loongarch/fpu/math-use-builtins-logb.h | 10 ++++
>> .../loongarch/fpu/math-use-builtins-lrint.h | 10 ++++
>> .../loongarch/fpu/math-use-builtins-rint.h | 9 +++
>> sysdeps/loongarch/fpu/s_scalbn.c | 29 +++++++++
>> sysdeps/loongarch/fpu/s_scalbnf.c | 29 +++++++++
>> 22 files changed, 326 insertions(+), 34 deletions(-)
>> create mode 100644 sysdeps/generic/math-use-builtins-llrint.h
>> create mode 100644 sysdeps/generic/math-use-builtins-logb.h
>> create mode 100644 sysdeps/generic/math-use-builtins-lrint.h
>> create mode 100644 sysdeps/loongarch/fpu/e_scalb.c
>> create mode 100644 sysdeps/loongarch/fpu/e_scalbf.c
>> create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-llrint.h
>> create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-logb.h
>> create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-lrint.h
>> create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-rint.h
>> create mode 100644 sysdeps/loongarch/fpu/s_scalbn.c
>> create mode 100644 sysdeps/loongarch/fpu/s_scalbnf.c
>>
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH 0/9] LoongArch: Use builtins with GCC >= 13
2022-11-23 2:55 ` tangxiaolin
@ 2022-11-23 3:18 ` Xi Ruoyao via Libc-alpha
0 siblings, 0 replies; 9+ messages in thread
From: Xi Ruoyao via Libc-alpha @ 2022-11-23 3:18 UTC (permalink / raw)
To: tangxiaolin, adhemerval.zanella, libc-alpha, caiyinyu
Cc: xuchenghua, chenglulu
On Wed, 2022-11-23 at 10:55 +0800, tangxiaolin wrote:
> Oops, I just resend {1,8,9}/9 with cover-patch, and still haven't solved
> this problem.
>
> I will resend this series again with v2 version, which is added ilogb
> function.
Try "git send-email". I had used to send long series manually with a
email client, but at last I found "git send-email" really makes our life
easier :).
--
Xi Ruoyao <xry111@xry111.site>
School of Aerospace Science and Technology, Xidian University
^ permalink raw reply [flat|nested] 9+ messages in thread
end of thread, other threads:[~2022-11-23 3:18 UTC | newest]
Thread overview: 9+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-11-23 2:38 [PATCH 0/9] LoongArch: Use builtins with GCC >= 13 Xiaolin Tang
2022-11-23 2:38 ` [PATCH 1/9] LoongArch: Use __builtin_rint{,f} " Xiaolin Tang
2022-11-23 2:38 ` [PATCH 8/9] LoongArch: Add support for scalbn[f] Xiaolin Tang
2022-11-23 2:38 ` [PATCH 9/9] LoongArch: Add support for scalb[f] Xiaolin Tang
-- strict thread matches above, loose matches on Subject: below --
2022-11-18 11:29 [PATCH 0/9] LoongArch: Use builtins with GCC >= 13 Xiaolin Tang
2022-11-22 16:20 ` Xi Ruoyao via Libc-alpha
2022-11-23 2:14 ` tangxiaolin
2022-11-23 2:55 ` tangxiaolin
2022-11-23 3:18 ` Xi Ruoyao via Libc-alpha
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).