From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: by sourceware.org (Postfix, from userid 7891) id D27E838582A4; Tue, 29 Aug 2023 04:33:25 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org D27E838582A4 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=sourceware.org; s=default; t=1693283605; bh=xZ3iyfnSngi/F8GgKwBf8YeYxWx2gLua5RMDoEDvotU=; h=From:To:Subject:Date:From; b=jZqHjRNmrSsbNY2MZ67A/ZObBW7wpL5DYe9iIdAxsd2aq+ahTvXDknL3KuEN7MPCW mwhebz6/YE948FZs37GalM1RCf9Ecd7RLqS9XSq/hi6bNe4K7bUUMSKUzAg5wD4k5i kqIJfT0ytfmrJYYDB0K6fN1FgAMJTF9ynP67bGlg= Content-Type: text/plain; charset="us-ascii" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit From: Yinyu Cai To: glibc-cvs@sourceware.org Subject: [glibc] LoongArch: Add ifunc support for memrchr{lsx, lasx} X-Act-Checkin: glibc X-Git-Author: dengjianbo X-Git-Refname: refs/heads/master X-Git-Oldrev: 60bcb9acbfcb40d1b613a13b539cb75e500b4ad6 X-Git-Newrev: 55e84dc6edc074f0f87a28bafce3249dfed4b201 Message-Id: <20230829043325.D27E838582A4@sourceware.org> Date: Tue, 29 Aug 2023 04:33:25 +0000 (GMT) List-Id: https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=55e84dc6edc074f0f87a28bafce3249dfed4b201 commit 55e84dc6edc074f0f87a28bafce3249dfed4b201 Author: dengjianbo Date: Mon Aug 28 10:08:37 2023 +0800 LoongArch: Add ifunc support for memrchr{lsx, lasx} According to glibc memrchr microbenchmark, this implementation could reduce the runtime as following: Name Percent of rutime reduced memrchr-lasx 20%-83% memrchr-lsx 20%-64% Diff: --- sysdeps/loongarch/lp64/multiarch/Makefile | 3 + sysdeps/loongarch/lp64/multiarch/ifunc-impl-list.c | 8 ++ sysdeps/loongarch/lp64/multiarch/ifunc-memrchr.h | 40 +++++++ sysdeps/loongarch/lp64/multiarch/memrchr-generic.c | 23 ++++ sysdeps/loongarch/lp64/multiarch/memrchr-lasx.S | 123 +++++++++++++++++++++ sysdeps/loongarch/lp64/multiarch/memrchr-lsx.S | 105 ++++++++++++++++++ sysdeps/loongarch/lp64/multiarch/memrchr.c | 33 ++++++ 7 files changed, 335 insertions(+) diff --git a/sysdeps/loongarch/lp64/multiarch/Makefile b/sysdeps/loongarch/lp64/multiarch/Makefile index 2f4802cfa4..7b87bc9055 100644 --- a/sysdeps/loongarch/lp64/multiarch/Makefile +++ b/sysdeps/loongarch/lp64/multiarch/Makefile @@ -27,5 +27,8 @@ sysdep_routines += \ memchr-aligned \ memchr-lsx \ memchr-lasx \ + memrchr-generic \ + memrchr-lsx \ + memrchr-lasx \ # sysdep_routines endif diff --git a/sysdeps/loongarch/lp64/multiarch/ifunc-impl-list.c b/sysdeps/loongarch/lp64/multiarch/ifunc-impl-list.c index a567b9cf4d..8bd5489ee2 100644 --- a/sysdeps/loongarch/lp64/multiarch/ifunc-impl-list.c +++ b/sysdeps/loongarch/lp64/multiarch/ifunc-impl-list.c @@ -109,5 +109,13 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, #endif IFUNC_IMPL_ADD (array, i, memchr, 1, __memchr_aligned) ) + + IFUNC_IMPL (i, name, memrchr, +#if !defined __loongarch_soft_float + IFUNC_IMPL_ADD (array, i, memrchr, SUPPORT_LASX, __memrchr_lasx) + IFUNC_IMPL_ADD (array, i, memrchr, SUPPORT_LSX, __memrchr_lsx) +#endif + IFUNC_IMPL_ADD (array, i, memrchr, 1, __memrchr_generic) + ) return i; } diff --git a/sysdeps/loongarch/lp64/multiarch/ifunc-memrchr.h b/sysdeps/loongarch/lp64/multiarch/ifunc-memrchr.h new file mode 100644 index 0000000000..8215f9ad94 --- /dev/null +++ b/sysdeps/loongarch/lp64/multiarch/ifunc-memrchr.h @@ -0,0 +1,40 @@ +/* Common definition for memrchr implementation. + All versions must be listed in ifunc-impl-list.c. + Copyright (C) 2023 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 + . */ + +#include +#include + +#if !defined __loongarch_soft_float +extern __typeof (REDIRECT_NAME) OPTIMIZE (lasx) attribute_hidden; +extern __typeof (REDIRECT_NAME) OPTIMIZE (lsx) attribute_hidden; +#endif +extern __typeof (REDIRECT_NAME) OPTIMIZE (generic) attribute_hidden; + +static inline void * +IFUNC_SELECTOR (void) +{ +#if !defined __loongarch_soft_float + if (SUPPORT_LASX) + return OPTIMIZE (lasx); + else if (SUPPORT_LSX) + return OPTIMIZE (lsx); + else +#endif + return OPTIMIZE (generic); +} diff --git a/sysdeps/loongarch/lp64/multiarch/memrchr-generic.c b/sysdeps/loongarch/lp64/multiarch/memrchr-generic.c new file mode 100644 index 0000000000..ced61ebce5 --- /dev/null +++ b/sysdeps/loongarch/lp64/multiarch/memrchr-generic.c @@ -0,0 +1,23 @@ +/* Generic implementation of memrchr. + Copyright (C) 2023 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 + . */ + +#if IS_IN (libc) +# define MEMRCHR __memrchr_generic +#endif + +#include diff --git a/sysdeps/loongarch/lp64/multiarch/memrchr-lasx.S b/sysdeps/loongarch/lp64/multiarch/memrchr-lasx.S new file mode 100644 index 0000000000..5f3e0d06d7 --- /dev/null +++ b/sysdeps/loongarch/lp64/multiarch/memrchr-lasx.S @@ -0,0 +1,123 @@ +/* Optimized memrchr implementation using LoongArch LASX instructions. + Copyright (C) 2023 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 + . */ + +#include +#include +#include + +#if IS_IN (libc) && !defined __loongarch_soft_float + +#ifndef MEMRCHR +# define MEMRCHR __memrchr_lasx +#endif + +LEAF(MEMRCHR, 6) + beqz a2, L(ret0) + addi.d a2, a2, -1 + add.d a3, a0, a2 + andi t1, a3, 0x3f + + bstrins.d a3, zero, 5, 0 + addi.d t1, t1, 1 + xvld xr0, a3, 0 + xvld xr1, a3, 32 + + sub.d t2, zero, t1 + li.d t3, -1 + xvreplgr2vr.b xr2, a1 + andi t4, a0, 0x3f + + srl.d t2, t3, t2 + xvseq.b xr0, xr0, xr2 + xvseq.b xr1, xr1, xr2 + xvmsknz.b xr0, xr0 + + + xvmsknz.b xr1, xr1 + xvpickve.w xr3, xr0, 4 + xvpickve.w xr4, xr1, 4 + vilvl.h vr0, vr3, vr0 + + vilvl.h vr1, vr4, vr1 + vilvl.w vr0, vr1, vr0 + movfr2gr.d t0, fa0 + and t0, t0, t2 + + bltu a2, t1, L(end) + bnez t0, L(found) + bstrins.d a0, zero, 5, 0 +L(loop): + xvld xr0, a3, -64 + + xvld xr1, a3, -32 + addi.d a3, a3, -64 + xvseq.b xr0, xr0, xr2 + xvseq.b xr1, xr1, xr2 + + + beq a0, a3, L(out) + xvmax.bu xr3, xr0, xr1 + xvseteqz.v fcc0, xr3 + bcnez fcc0, L(loop) + + xvmsknz.b xr0, xr0 + xvmsknz.b xr1, xr1 + xvpickve.w xr3, xr0, 4 + xvpickve.w xr4, xr1, 4 + + vilvl.h vr0, vr3, vr0 + vilvl.h vr1, vr4, vr1 + vilvl.w vr0, vr1, vr0 + movfr2gr.d t0, fa0 + +L(found): + addi.d a0, a3, 63 + clz.d t1, t0 + sub.d a0, a0, t1 + jr ra + + +L(out): + xvmsknz.b xr0, xr0 + xvmsknz.b xr1, xr1 + xvpickve.w xr3, xr0, 4 + xvpickve.w xr4, xr1, 4 + + vilvl.h vr0, vr3, vr0 + vilvl.h vr1, vr4, vr1 + vilvl.w vr0, vr1, vr0 + movfr2gr.d t0, fa0 + +L(end): + sll.d t2, t3, t4 + and t0, t0, t2 + addi.d a0, a3, 63 + clz.d t1, t0 + + sub.d a0, a0, t1 + maskeqz a0, a0, t0 + jr ra +L(ret0): + move a0, zero + + + jr ra +END(MEMRCHR) + +libc_hidden_builtin_def (MEMRCHR) +#endif diff --git a/sysdeps/loongarch/lp64/multiarch/memrchr-lsx.S b/sysdeps/loongarch/lp64/multiarch/memrchr-lsx.S new file mode 100644 index 0000000000..39a7c8b076 --- /dev/null +++ b/sysdeps/loongarch/lp64/multiarch/memrchr-lsx.S @@ -0,0 +1,105 @@ +/* Optimized memrchr implementation using LoongArch LSX instructions. + Copyright (C) 2023 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 + . */ + +#include +#include +#include + +#if IS_IN (libc) && !defined __loongarch_soft_float + +# define MEMRCHR __memrchr_lsx + +LEAF(MEMRCHR, 6) + beqz a2, L(ret0) + addi.d a2, a2, -1 + add.d a3, a0, a2 + andi t1, a3, 0x1f + + bstrins.d a3, zero, 4, 0 + addi.d t1, t1, 1 + vld vr0, a3, 0 + vld vr1, a3, 16 + + sub.d t2, zero, t1 + li.d t3, -1 + vreplgr2vr.b vr2, a1 + andi t4, a0, 0x1f + + srl.d t2, t3, t2 + vseq.b vr0, vr0, vr2 + vseq.b vr1, vr1, vr2 + vmsknz.b vr0, vr0 + + + vmsknz.b vr1, vr1 + vilvl.h vr0, vr1, vr0 + movfr2gr.s t0, fa0 + and t0, t0, t2 + + bltu a2, t1, L(end) + bnez t0, L(found) + bstrins.d a0, zero, 4, 0 +L(loop): + vld vr0, a3, -32 + + vld vr1, a3, -16 + addi.d a3, a3, -32 + vseq.b vr0, vr0, vr2 + vseq.b vr1, vr1, vr2 + + beq a0, a3, L(out) + vmax.bu vr3, vr0, vr1 + vseteqz.v fcc0, vr3 + bcnez fcc0, L(loop) + + + vmsknz.b vr0, vr0 + vmsknz.b vr1, vr1 + vilvl.h vr0, vr1, vr0 + movfr2gr.s t0, fa0 + +L(found): + addi.d a0, a3, 31 + clz.w t1, t0 + sub.d a0, a0, t1 + jr ra + +L(out): + vmsknz.b vr0, vr0 + vmsknz.b vr1, vr1 + vilvl.h vr0, vr1, vr0 + movfr2gr.s t0, fa0 + +L(end): + sll.d t2, t3, t4 + and t0, t0, t2 + addi.d a0, a3, 31 + clz.w t1, t0 + + + sub.d a0, a0, t1 + maskeqz a0, a0, t0 + jr ra +L(ret0): + move a0, zero + + jr ra +END(MEMRCHR) + +libc_hidden_builtin_def (MEMRCHR) +#endif diff --git a/sysdeps/loongarch/lp64/multiarch/memrchr.c b/sysdeps/loongarch/lp64/multiarch/memrchr.c new file mode 100644 index 0000000000..8baba9ab7e --- /dev/null +++ b/sysdeps/loongarch/lp64/multiarch/memrchr.c @@ -0,0 +1,33 @@ +/* Multiple versions of memrchr. + All versions must be listed in ifunc-impl-list.c. + Copyright (C) 2023 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 + . */ + +/* Define multiple versions only for the definition in libc. */ +#if IS_IN (libc) +# define memrchr __redirect_memrchr +# include +# undef memrchr + +# define SYMBOL_NAME memrchr +# include "ifunc-memrchr.h" + +libc_ifunc_redirected (__redirect_memrchr, __memrchr, IFUNC_SELECTOR ()); +libc_hidden_def (__memrchr) +weak_alias (__memrchr, memrchr) + +#endif