From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-pl1-x630.google.com (mail-pl1-x630.google.com [IPv6:2607:f8b0:4864:20::630]) by sourceware.org (Postfix) with ESMTPS id 4078F3854579 for ; Fri, 18 Nov 2022 21:23:04 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.1 sourceware.org 4078F3854579 Authentication-Results: sourceware.org; dmarc=pass (p=none dis=none) header.from=gmail.com Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=gmail.com Received: by mail-pl1-x630.google.com with SMTP id k7so5643616pll.6 for ; Fri, 18 Nov 2022 13:23:04 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20210112; h=content-transfer-encoding:mime-version:references:in-reply-to :message-id:date:subject:cc:to:from:from:to:cc:subject:date :message-id:reply-to; bh=VT4d3v0fW5XOceNgq52N7edH9fycQ3BeNQ2foMhv0b0=; b=anr8WUSj5OrCDSkSRmZKwFvCa744EBoqpQZEgoQP0x83ZJNBYuPw0JNLPfC6jO432t 6Uz5xBBeRV+3spfTIjJ4KjMcBKOa4bEV3PgEPhnzo5ijUsFRbLVn4CPkZDwJ1tRNJd6C 1VnCsTNkRZ/3F+vJPi4cgt4uU4UkjR+qy76UlQ1x7YL1UB1FoRDd6rB8ULTLaW4jFw+c u/LKhk1q70l7MI0yO7lPtyHUdRv0kc1eGLV885KNMEkM5mH7J/8D7KP5SPbXml6oJvfu Zi130miIcbjwPWkPsWSyaPu18QjjzjyNfqjL1CkVAYQcBiWkuzBjjasaNfn5AYBggKTb 4PZQ== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=content-transfer-encoding:mime-version:references:in-reply-to :message-id:date:subject:cc:to:from:x-gm-message-state:from:to:cc :subject:date:message-id:reply-to; bh=VT4d3v0fW5XOceNgq52N7edH9fycQ3BeNQ2foMhv0b0=; b=t9Ueoy8clA/Z6npzE2Z0RHHYCPwXaoI+Wh+LTgzuEVvJE23evTfpaH3TU/UF0AGuVU WV39MqFM9LGC1rLmHmeFftHCXMCWc6xZSsfZd7ctpCZkNUaP9gjdIaJonRBcVDU20Uu4 LzJAjqtooiUzvN+ec+HcALIk3tV/3rc1V7ih6KWNNIHxxoEmRKKZ3lCW73hB+PRIyLRn gTiUZFb43CAwe7dLifxr84aMYo3xRacp3KTUwju4zY+3GxdQ3gB9riGyWjajp7KjvDAr fbk6MscbWYZJyzJbEsWqsLrbZFKIvTUSqBJsbExUE7ocheEw2E8ArsnjJzlqqpDl+H7e UX8g== X-Gm-Message-State: ANoB5pkjcjKYvgQjRgSBLE5D/tO3twtPsbDG5r9flcGH+v1Kqn1tEw6+ cU9ojJy2sWs3Xr/v4hFpcV+iIrCCWjU= X-Google-Smtp-Source: AA0mqf7W2Vq65OwKjDbFKL4AeaiiprIec6z78a53k8fCkkWBsoXbZaKk28H3AzxvhGkJbnIJPp/VoQ== X-Received: by 2002:a17:902:bcc7:b0:188:5c52:83e1 with SMTP id o7-20020a170902bcc700b001885c5283e1mr1244376pls.128.1668806582690; Fri, 18 Nov 2022 13:23:02 -0800 (PST) Received: from noahgold-desk.. ([192.55.60.47]) by smtp.gmail.com with ESMTPSA id m1-20020a170902db0100b00188fce6e8a5sm2489232plx.290.2022.11.18.13.23.01 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Fri, 18 Nov 2022 13:23:02 -0800 (PST) From: Noah Goldstein To: libc-alpha@sourceware.org Cc: goldstein.w.n@gmail.com, hjl.tools@gmail.com, andrey.kolesov@intel.com, carlos@systemhalted.org Subject: [PATCH v4 3/3] x86/fpu: Factor out shared avx2/avx512 code in svml_{s|d}_wrapper_impl.h Date: Fri, 18 Nov 2022 13:22:55 -0800 Message-Id: <20221118212255.1305681-3-goldstein.w.n@gmail.com> X-Mailer: git-send-email 2.34.1 In-Reply-To: <20221118212255.1305681-1-goldstein.w.n@gmail.com> References: <20221118190835.1033248-1-goldstein.w.n@gmail.com> <20221118212255.1305681-1-goldstein.w.n@gmail.com> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Spam-Status: No, score=-12.2 required=5.0 tests=BAYES_00,DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,DKIM_VALID_EF,FREEMAIL_FROM,GIT_PATCH_0,KAM_SHORT,RCVD_IN_DNSWL_NONE,SPF_HELO_NONE,SPF_PASS,TXREP autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on server2.sourceware.org List-Id: Code is exactly the same for the two so better to only maintain one version. All math and mathvec tests pass on x86. --- sysdeps/x86_64/fpu/svml_d_wrapper_impl.h | 172 +------------------- sysdeps/x86_64/fpu/svml_s_wrapper_impl.h | 172 +------------------- sysdeps/x86_64/fpu/svml_sd_wrapper_impl.h | 190 ++++++++++++++++++++++ 3 files changed, 192 insertions(+), 342 deletions(-) create mode 100644 sysdeps/x86_64/fpu/svml_sd_wrapper_impl.h diff --git a/sysdeps/x86_64/fpu/svml_d_wrapper_impl.h b/sysdeps/x86_64/fpu/svml_d_wrapper_impl.h index 78c30c56cb..52407da8ed 100644 --- a/sysdeps/x86_64/fpu/svml_d_wrapper_impl.h +++ b/sysdeps/x86_64/fpu/svml_d_wrapper_impl.h @@ -82,174 +82,4 @@ ret .endm -/* AVX/AVX2 ISA version as wrapper to SSE ISA version. */ -.macro WRAPPER_IMPL_AVX callee - pushq %rbp - cfi_adjust_cfa_offset (8) - cfi_rel_offset (%rbp, 0) - movq %rsp, %rbp - cfi_def_cfa_register (%rbp) - andq $-32, %rsp - subq $32, %rsp - vmovaps %ymm0, (%rsp) - vzeroupper - call HIDDEN_JUMPTARGET(\callee) - vmovaps %xmm0, (%rsp) - vmovaps 16(%rsp), %xmm0 - call HIDDEN_JUMPTARGET(\callee) - /* combine xmm0 (return of second call) with result of first - call (saved on stack). Might be worth exploring logic that - uses `vpblend` and reads in ymm1 using -16(rsp). */ - vmovaps (%rsp), %xmm1 - vinsertf128 $1, %xmm0, %ymm1, %ymm0 - movq %rbp, %rsp - cfi_def_cfa_register (%rsp) - popq %rbp - cfi_adjust_cfa_offset (-8) - cfi_restore (%rbp) - ret -.endm - -/* 2 argument AVX/AVX2 ISA version as wrapper to SSE ISA version. */ -.macro WRAPPER_IMPL_AVX_ff callee - pushq %rbp - cfi_adjust_cfa_offset (8) - cfi_rel_offset (%rbp, 0) - movq %rsp, %rbp - cfi_def_cfa_register (%rbp) - andq $-32, %rsp - subq $64, %rsp - vmovaps %ymm0, (%rsp) - vmovaps %ymm1, 32(%rsp) - vzeroupper - call HIDDEN_JUMPTARGET(\callee) - vmovaps 48(%rsp), %xmm1 - vmovaps %xmm0, (%rsp) - vmovaps 16(%rsp), %xmm0 - call HIDDEN_JUMPTARGET(\callee) - /* combine xmm0 (return of second call) with result of first - call (saved on stack). Might be worth exploring logic that - uses `vpblend` and reads in ymm1 using -16(rsp). */ - vmovaps (%rsp), %xmm1 - vinsertf128 $1, %xmm0, %ymm1, %ymm0 - movq %rbp, %rsp - cfi_def_cfa_register (%rsp) - popq %rbp - cfi_adjust_cfa_offset (-8) - cfi_restore (%rbp) - ret -.endm - -/* 3 argument AVX/AVX2 ISA version as wrapper to SSE ISA version. */ -.macro WRAPPER_IMPL_AVX_fFF callee - pushq %rbp - cfi_adjust_cfa_offset (8) - cfi_rel_offset (%rbp, 0) - movq %rsp, %rbp - andq $-32, %rsp - subq $32, %rsp - vmovaps %ymm0, (%rsp) - pushq %rbx - pushq %r14 - movq %rdi, %rbx - movq %rsi, %r14 - vzeroupper - call HIDDEN_JUMPTARGET(\callee) - vmovaps 32(%rsp), %xmm0 - leaq 16(%rbx), %rdi - leaq 16(%r14), %rsi - call HIDDEN_JUMPTARGET(\callee) - popq %r14 - popq %rbx - movq %rbp, %rsp - cfi_def_cfa_register (%rsp) - popq %rbp - cfi_adjust_cfa_offset (-8) - cfi_restore (%rbp) - ret -.endm - -/* AVX512 ISA version as wrapper to AVX2 ISA version. */ -.macro WRAPPER_IMPL_AVX512 callee - pushq %rbp - cfi_adjust_cfa_offset (8) - cfi_rel_offset (%rbp, 0) - movq %rsp, %rbp - cfi_def_cfa_register (%rbp) - andq $-64, %rsp - subq $64, %rsp - vmovups %zmm0, (%rsp) - call HIDDEN_JUMPTARGET(\callee) - vmovupd %ymm0, (%rsp) - vmovupd 32(%rsp), %ymm0 - call HIDDEN_JUMPTARGET(\callee) - /* combine ymm0 (return of second call) with result of first - call (saved on stack). */ - vmovaps (%rsp), %ymm1 - vinserti64x4 $0x1, %ymm0, %zmm1, %zmm0 - movq %rbp, %rsp - cfi_def_cfa_register (%rsp) - popq %rbp - cfi_adjust_cfa_offset (-8) - cfi_restore (%rbp) - ret -.endm - -/* 2 argument AVX512 ISA version as wrapper to AVX2 ISA version. */ -.macro WRAPPER_IMPL_AVX512_ff callee - pushq %rbp - cfi_adjust_cfa_offset (8) - cfi_rel_offset (%rbp, 0) - movq %rsp, %rbp - cfi_def_cfa_register (%rbp) - andq $-64, %rsp - addq $-128, %rsp - vmovups %zmm0, (%rsp) - vmovups %zmm1, 64(%rsp) - /* ymm0 and ymm1 are already set. */ - call HIDDEN_JUMPTARGET(\callee) - vmovups 96(%rsp), %ymm1 - vmovaps %ymm0, (%rsp) - vmovups 32(%rsp), %ymm0 - call HIDDEN_JUMPTARGET(\callee) - /* combine ymm0 (return of second call) with result of first - call (saved on stack). */ - vmovaps (%rsp), %ymm1 - vinserti64x4 $0x1, %ymm0, %zmm1, %zmm0 - movq %rbp, %rsp - cfi_def_cfa_register (%rsp) - popq %rbp - cfi_adjust_cfa_offset (-8) - cfi_restore (%rbp) - ret -.endm - -/* 3 argument AVX512 ISA version as wrapper to AVX2 ISA version. */ -.macro WRAPPER_IMPL_AVX512_fFF callee - pushq %rbp - cfi_adjust_cfa_offset (8) - cfi_rel_offset (%rbp, 0) - movq %rsp, %rbp - cfi_def_cfa_register (%rbp) - andq $-64, %rsp - subq $64, %rsp - vmovaps %zmm0, (%rsp) - pushq %rbx - pushq %r14 - movq %rdi, %rbx - movq %rsi, %r14 - /* ymm0 is already set. */ - call HIDDEN_JUMPTARGET(\callee) - vmovaps 48(%rsp), %ymm0 - leaq 32(%rbx), %rdi - leaq 32(%r14), %rsi - call HIDDEN_JUMPTARGET(\callee) - popq %r14 - popq %rbx - movq %rbp, %rsp - cfi_def_cfa_register (%rsp) - popq %rbp - cfi_adjust_cfa_offset (-8) - cfi_restore (%rbp) - ret -.endm +#include "svml_sd_wrapper_impl.h" diff --git a/sysdeps/x86_64/fpu/svml_s_wrapper_impl.h b/sysdeps/x86_64/fpu/svml_s_wrapper_impl.h index 43f2b91f32..d9266563ef 100644 --- a/sysdeps/x86_64/fpu/svml_s_wrapper_impl.h +++ b/sysdeps/x86_64/fpu/svml_s_wrapper_impl.h @@ -118,174 +118,4 @@ ret .endm -/* AVX/AVX2 ISA version as wrapper to SSE ISA version. */ -.macro WRAPPER_IMPL_AVX callee - pushq %rbp - cfi_adjust_cfa_offset (8) - cfi_rel_offset (%rbp, 0) - movq %rsp, %rbp - cfi_def_cfa_register (%rbp) - andq $-32, %rsp - subq $32, %rsp - vmovaps %ymm0, (%rsp) - vzeroupper - call HIDDEN_JUMPTARGET(\callee) - vmovaps %xmm0, (%rsp) - vmovaps 16(%rsp), %xmm0 - call HIDDEN_JUMPTARGET(\callee) - /* combine xmm0 (return of second call) with result of first - call (saved on stack). Might be worth exploring logic that - uses `vpblend` and reads in ymm1 using -16(rsp). */ - vmovaps (%rsp), %xmm1 - vinsertf128 $1, %xmm0, %ymm1, %ymm0 - movq %rbp, %rsp - cfi_def_cfa_register (%rsp) - popq %rbp - cfi_adjust_cfa_offset (-8) - cfi_restore (%rbp) - ret -.endm - -/* 2 argument AVX/AVX2 ISA version as wrapper to SSE ISA version. */ -.macro WRAPPER_IMPL_AVX_ff callee - pushq %rbp - cfi_adjust_cfa_offset (8) - cfi_rel_offset (%rbp, 0) - movq %rsp, %rbp - cfi_def_cfa_register (%rbp) - andq $-32, %rsp - subq $64, %rsp - vmovaps %ymm0, (%rsp) - vmovaps %ymm1, 32(%rsp) - vzeroupper - call HIDDEN_JUMPTARGET(\callee) - vmovaps 48(%rsp), %xmm1 - vmovaps %xmm0, (%rsp) - vmovaps 16(%rsp), %xmm0 - call HIDDEN_JUMPTARGET(\callee) - /* combine xmm0 (return of second call) with result of first - call (saved on stack). Might be worth exploring logic that - uses `vpblend` and reads in ymm1 using -16(rsp). */ - vmovaps (%rsp), %xmm1 - vinsertf128 $1, %xmm0, %ymm1, %ymm0 - movq %rbp, %rsp - cfi_def_cfa_register (%rsp) - popq %rbp - cfi_adjust_cfa_offset (-8) - cfi_restore (%rbp) - ret -.endm - -/* 3 argument AVX/AVX2 ISA version as wrapper to SSE ISA version. */ -.macro WRAPPER_IMPL_AVX_fFF callee - pushq %rbp - cfi_adjust_cfa_offset (8) - cfi_rel_offset (%rbp, 0) - movq %rsp, %rbp - andq $-32, %rsp - subq $32, %rsp - vmovaps %ymm0, (%rsp) - pushq %rbx - pushq %r14 - movq %rdi, %rbx - movq %rsi, %r14 - vzeroupper - call HIDDEN_JUMPTARGET(\callee) - vmovaps 32(%rsp), %xmm0 - leaq 16(%rbx), %rdi - leaq 16(%r14), %rsi - call HIDDEN_JUMPTARGET(\callee) - popq %r14 - popq %rbx - movq %rbp, %rsp - cfi_def_cfa_register (%rsp) - popq %rbp - cfi_adjust_cfa_offset (-8) - cfi_restore (%rbp) - ret -.endm - -/* AVX512 ISA version as wrapper to AVX2 ISA version. */ -.macro WRAPPER_IMPL_AVX512 callee - pushq %rbp - cfi_adjust_cfa_offset (8) - cfi_rel_offset (%rbp, 0) - movq %rsp, %rbp - cfi_def_cfa_register (%rbp) - andq $-64, %rsp - subq $64, %rsp - vmovups %zmm0, (%rsp) - call HIDDEN_JUMPTARGET(\callee) - vmovupd %ymm0, (%rsp) - vmovupd 32(%rsp), %ymm0 - call HIDDEN_JUMPTARGET(\callee) - /* combine ymm0 (return of second call) with result of first - call (saved on stack). */ - vmovaps (%rsp), %ymm1 - vinserti64x4 $0x1, %ymm0, %zmm1, %zmm0 - movq %rbp, %rsp - cfi_def_cfa_register (%rsp) - popq %rbp - cfi_adjust_cfa_offset (-8) - cfi_restore (%rbp) - ret -.endm - -/* 2 argument AVX512 ISA version as wrapper to AVX2 ISA version. */ -.macro WRAPPER_IMPL_AVX512_ff callee - pushq %rbp - cfi_adjust_cfa_offset (8) - cfi_rel_offset (%rbp, 0) - movq %rsp, %rbp - cfi_def_cfa_register (%rbp) - andq $-64, %rsp - addq $-128, %rsp - vmovups %zmm0, (%rsp) - vmovups %zmm1, 64(%rsp) - /* ymm0 and ymm1 are already set. */ - call HIDDEN_JUMPTARGET(\callee) - vmovups 96(%rsp), %ymm1 - vmovaps %ymm0, (%rsp) - vmovups 32(%rsp), %ymm0 - call HIDDEN_JUMPTARGET(\callee) - /* combine ymm0 (return of second call) with result of first - call (saved on stack). */ - vmovaps (%rsp), %ymm1 - vinserti64x4 $0x1, %ymm0, %zmm1, %zmm0 - movq %rbp, %rsp - cfi_def_cfa_register (%rsp) - popq %rbp - cfi_adjust_cfa_offset (-8) - cfi_restore (%rbp) - ret -.endm - -/* 3 argument AVX512 ISA version as wrapper to AVX2 ISA version. */ -.macro WRAPPER_IMPL_AVX512_fFF callee - pushq %rbp - cfi_adjust_cfa_offset (8) - cfi_rel_offset (%rbp, 0) - movq %rsp, %rbp - cfi_def_cfa_register (%rbp) - andq $-64, %rsp - subq $64, %rsp - vmovaps %zmm0, (%rsp) - pushq %rbx - pushq %r14 - movq %rdi, %rbx - movq %rsi, %r14 - /* ymm0 is already set. */ - call HIDDEN_JUMPTARGET(\callee) - vmovaps 48(%rsp), %ymm0 - leaq 32(%rbx), %rdi - leaq 32(%r14), %rsi - call HIDDEN_JUMPTARGET(\callee) - popq %r14 - popq %rbx - movq %rbp, %rsp - cfi_def_cfa_register (%rsp) - popq %rbp - cfi_adjust_cfa_offset (-8) - cfi_restore (%rbp) - ret -.endm +#include "svml_sd_wrapper_impl.h" diff --git a/sysdeps/x86_64/fpu/svml_sd_wrapper_impl.h b/sysdeps/x86_64/fpu/svml_sd_wrapper_impl.h new file mode 100644 index 0000000000..bd934ad578 --- /dev/null +++ b/sysdeps/x86_64/fpu/svml_sd_wrapper_impl.h @@ -0,0 +1,190 @@ +/* Common float/double wrapper implementations of vector math + functions. + 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 + . */ + +/* AVX/AVX2 ISA version as wrapper to SSE ISA version. */ +.macro WRAPPER_IMPL_AVX callee + pushq %rbp + cfi_adjust_cfa_offset (8) + cfi_rel_offset (%rbp, 0) + movq %rsp, %rbp + cfi_def_cfa_register (%rbp) + andq $-32, %rsp + subq $32, %rsp + vmovaps %ymm0, (%rsp) + vzeroupper + call HIDDEN_JUMPTARGET(\callee) + vmovaps %xmm0, (%rsp) + vmovaps 16(%rsp), %xmm0 + call HIDDEN_JUMPTARGET(\callee) + /* combine xmm0 (return of second call) with result of first + call (saved on stack). Might be worth exploring logic that + uses `vpblend` and reads in ymm1 using -16(rsp). */ + vmovaps (%rsp), %xmm1 + vinsertf128 $1, %xmm0, %ymm1, %ymm0 + movq %rbp, %rsp + cfi_def_cfa_register (%rsp) + popq %rbp + cfi_adjust_cfa_offset (-8) + cfi_restore (%rbp) + ret +.endm + +/* 2 argument AVX/AVX2 ISA version as wrapper to SSE ISA version. */ +.macro WRAPPER_IMPL_AVX_ff callee + pushq %rbp + cfi_adjust_cfa_offset (8) + cfi_rel_offset (%rbp, 0) + movq %rsp, %rbp + cfi_def_cfa_register (%rbp) + andq $-32, %rsp + subq $64, %rsp + vmovaps %ymm0, (%rsp) + vmovaps %ymm1, 32(%rsp) + vzeroupper + call HIDDEN_JUMPTARGET(\callee) + vmovaps 48(%rsp), %xmm1 + vmovaps %xmm0, (%rsp) + vmovaps 16(%rsp), %xmm0 + call HIDDEN_JUMPTARGET(\callee) + /* combine xmm0 (return of second call) with result of first + call (saved on stack). Might be worth exploring logic that + uses `vpblend` and reads in ymm1 using -16(rsp). */ + vmovaps (%rsp), %xmm1 + vinsertf128 $1, %xmm0, %ymm1, %ymm0 + movq %rbp, %rsp + cfi_def_cfa_register (%rsp) + popq %rbp + cfi_adjust_cfa_offset (-8) + cfi_restore (%rbp) + ret +.endm + +/* 3 argument AVX/AVX2 ISA version as wrapper to SSE ISA version. */ +.macro WRAPPER_IMPL_AVX_fFF callee + pushq %rbp + cfi_adjust_cfa_offset (8) + cfi_rel_offset (%rbp, 0) + movq %rsp, %rbp + andq $-32, %rsp + subq $32, %rsp + vmovaps %ymm0, (%rsp) + pushq %rbx + pushq %r14 + movq %rdi, %rbx + movq %rsi, %r14 + vzeroupper + call HIDDEN_JUMPTARGET(\callee) + vmovaps 32(%rsp), %xmm0 + leaq 16(%rbx), %rdi + leaq 16(%r14), %rsi + call HIDDEN_JUMPTARGET(\callee) + popq %r14 + popq %rbx + movq %rbp, %rsp + cfi_def_cfa_register (%rsp) + popq %rbp + cfi_adjust_cfa_offset (-8) + cfi_restore (%rbp) + ret +.endm + +/* AVX512 ISA version as wrapper to AVX2 ISA version. */ +.macro WRAPPER_IMPL_AVX512 callee + pushq %rbp + cfi_adjust_cfa_offset (8) + cfi_rel_offset (%rbp, 0) + movq %rsp, %rbp + cfi_def_cfa_register (%rbp) + andq $-64, %rsp + subq $64, %rsp + vmovups %zmm0, (%rsp) + call HIDDEN_JUMPTARGET(\callee) + vmovupd %ymm0, (%rsp) + vmovupd 32(%rsp), %ymm0 + call HIDDEN_JUMPTARGET(\callee) + /* combine ymm0 (return of second call) with result of first + call (saved on stack). */ + vmovaps (%rsp), %ymm1 + vinserti64x4 $0x1, %ymm0, %zmm1, %zmm0 + movq %rbp, %rsp + cfi_def_cfa_register (%rsp) + popq %rbp + cfi_adjust_cfa_offset (-8) + cfi_restore (%rbp) + ret +.endm + +/* 2 argument AVX512 ISA version as wrapper to AVX2 ISA version. */ +.macro WRAPPER_IMPL_AVX512_ff callee + pushq %rbp + cfi_adjust_cfa_offset (8) + cfi_rel_offset (%rbp, 0) + movq %rsp, %rbp + cfi_def_cfa_register (%rbp) + andq $-64, %rsp + addq $-128, %rsp + vmovups %zmm0, (%rsp) + vmovups %zmm1, 64(%rsp) + /* ymm0 and ymm1 are already set. */ + call HIDDEN_JUMPTARGET(\callee) + vmovups 96(%rsp), %ymm1 + vmovaps %ymm0, (%rsp) + vmovups 32(%rsp), %ymm0 + call HIDDEN_JUMPTARGET(\callee) + /* combine ymm0 (return of second call) with result of first + call (saved on stack). */ + vmovaps (%rsp), %ymm1 + vinserti64x4 $0x1, %ymm0, %zmm1, %zmm0 + movq %rbp, %rsp + cfi_def_cfa_register (%rsp) + popq %rbp + cfi_adjust_cfa_offset (-8) + cfi_restore (%rbp) + ret +.endm + +/* 3 argument AVX512 ISA version as wrapper to AVX2 ISA version. */ +.macro WRAPPER_IMPL_AVX512_fFF callee + pushq %rbp + cfi_adjust_cfa_offset (8) + cfi_rel_offset (%rbp, 0) + movq %rsp, %rbp + cfi_def_cfa_register (%rbp) + andq $-64, %rsp + subq $64, %rsp + vmovaps %zmm0, (%rsp) + pushq %rbx + pushq %r14 + movq %rdi, %rbx + movq %rsi, %r14 + /* ymm0 is already set. */ + call HIDDEN_JUMPTARGET(\callee) + vmovaps 48(%rsp), %ymm0 + leaq 32(%rbx), %rdi + leaq 32(%r14), %rsi + call HIDDEN_JUMPTARGET(\callee) + popq %r14 + popq %rbx + movq %rbp, %rsp + cfi_def_cfa_register (%rsp) + popq %rbp + cfi_adjust_cfa_offset (-8) + cfi_restore (%rbp) + ret +.endm -- 2.34.1