From patchwork Tue Sep 25 14:56:09 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Jason A. Donenfeld" X-Patchwork-Id: 147508 Delivered-To: patch@linaro.org Received: by 2002:a2e:8595:0:0:0:0:0 with SMTP id b21-v6csp832768lji; Tue, 25 Sep 2018 07:59:25 -0700 (PDT) X-Google-Smtp-Source: ACcGV61eoHkFUXX0VzpjAjv/PoOdu+hd9F/rdW3sFIgZNC+4ny03+TR7pM2cvGBmjo625aBCgeaH X-Received: by 2002:a17:902:bd4a:: with SMTP id b10-v6mr1618785plx.209.1537887565324; Tue, 25 Sep 2018 07:59:25 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1537887565; cv=none; d=google.com; s=arc-20160816; b=YtJw3jBXRNhKzEWn/RxWQpwFcFQqWuoOniV4/YX9ysTzCOdpWZPDFXXY7lqS9N3OuC xK+ko+z34OX6icaqPr0BSmURhyFMuVwWsQnnSX1wEXg9owV4JykYlJTQ59pjy9Z3BaP3 5Rr0g/ylY+kboGZsmkuTl6D+UySBCXuLNNvQzC5IhiTVIvDPkQ0bmF50BMlX5NnAtUN1 nyQgBEeXb/1LdtQpf58ZFS4wHKme3K1GbtQh9PQizR3VR/K59udeJybmTr9m6MUAYbQ5 1AJvdhYLlbxX0TI6FLRBDhcFHxlrN1gbLbpkn3vCzJP7FsSZ/hIkAijGMF5zVY0zsrh4 +eWg== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=list-id:precedence:sender:content-transfer-encoding:mime-version :references:in-reply-to:message-id:date:subject:cc:to:from :dkim-signature; bh=euYPe28/SduDQLV/gSW6C2NaXOOX85VH3JM5KwCltoc=; b=0kyV1MQPITQp38tiVX6MlPtWT9L32AcJ4dkp1kMzyNYeuLiv8t9x3WhoKiuAZPpncM 8C02FDVgX5jQzLchRodpZ16J3ghF5uQCTZJ2gkHDA23wZFOaoFHjxNph/Dd50SsaEwUL MMVTonkuyKtnQQw9ZdcbltNmPrJFkHk1LnguosUPXzjUj/mUTpCxMwXtodDkbYR0gw0U UWuvh1HjdpZ7isoPPG8SacoPDpl9nLQOtmA/IPIj9h9dCVAnPu9lHBCkO5k7DikQVGFZ 5T0YY0tMScqoTYimjxW6AD66j+JxKABtH4i13anqOU5IjPefHecSKHQUFtxDTpMzHJml F/GA== ARC-Authentication-Results: i=1; mx.google.com; dkim=pass header.i=@zx2c4.com header.s=mail header.b="AU5t/mBp"; spf=pass (google.com: best guess record for domain of linux-crypto-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) smtp.mailfrom=linux-crypto-owner@vger.kernel.org; dmarc=pass (p=NONE sp=NONE dis=NONE) header.from=zx2c4.com Return-Path: Received: from vger.kernel.org (vger.kernel.org. [209.132.180.67]) by mx.google.com with ESMTP id cf13-v6si2754759plb.334.2018.09.25.07.59.24; Tue, 25 Sep 2018 07:59:25 -0700 (PDT) Received-SPF: pass (google.com: best guess record for domain of linux-crypto-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67; Authentication-Results: mx.google.com; dkim=pass header.i=@zx2c4.com header.s=mail header.b="AU5t/mBp"; spf=pass (google.com: best guess record for domain of linux-crypto-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) smtp.mailfrom=linux-crypto-owner@vger.kernel.org; dmarc=pass (p=NONE sp=NONE dis=NONE) header.from=zx2c4.com Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1729272AbeIYVHQ (ORCPT + 2 others); Tue, 25 Sep 2018 17:07:16 -0400 Received: from frisell.zx2c4.com ([192.95.5.64]:38531 "EHLO frisell.zx2c4.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1729654AbeIYVFF (ORCPT ); Tue, 25 Sep 2018 17:05:05 -0400 Received: by frisell.zx2c4.com (ZX2C4 Mail Server) with ESMTP id f0204edd; Tue, 25 Sep 2018 14:38:31 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha1; c=relaxed; d=zx2c4.com; h=from:to:cc :subject:date:message-id:in-reply-to:references:mime-version :content-transfer-encoding; s=mail; bh=Q9NUTEV+hrOCv8K1XwjQbiiU8 R4=; b=AU5t/mBpgeCuhag6lLceXPk1hbH0gQNUyFLSK/ryrXg/KcMEfPLZyVBTs 0jC0OmzShqQNpLNLVUWktAEcJXHJlbHlwwW55ABSO/h0rJiBtIM+2i+rCixO9RA4 SFnCQfgGDvXvZUjW4/U2O0LdNatBZjNYmdxB+cXt2fk5xd5xUqVXboTki3ZfAQI7 xo+vv6NfR3cRiGMZSLOBCtinLxlh0vd2NC5nCCrKHj+R3ntvn6IzJtFDckfKfubg z39wLJCQHn5g36yjXrFK1j/WUin/9Lc3UtN7EsYHK3d0WogEV4jBSB9qY73FN5Lv y6Ev8E8W6SfSNusmibT82Zb/SSiRw== Received: by frisell.zx2c4.com (ZX2C4 Mail Server) with ESMTPSA id 9b6a1e57 (TLSv1.2:ECDHE-RSA-AES256-GCM-SHA384:256:NO); Tue, 25 Sep 2018 14:38:30 +0000 (UTC) From: "Jason A. Donenfeld" To: linux-kernel@vger.kernel.org, netdev@vger.kernel.org, linux-crypto@vger.kernel.org, davem@davemloft.net, gregkh@linuxfoundation.org Cc: "Jason A. Donenfeld" , Samuel Neves , Andy Lutomirski , Jean-Philippe Aumasson , Andy Polyakov , Thomas Gleixner , Ingo Molnar , x86@kernel.org Subject: [PATCH net-next v6 10/23] zinc: Poly1305 x86_64 implementation Date: Tue, 25 Sep 2018 16:56:09 +0200 Message-Id: <20180925145622.29959-11-Jason@zx2c4.com> In-Reply-To: <20180925145622.29959-1-Jason@zx2c4.com> References: <20180925145622.29959-1-Jason@zx2c4.com> MIME-Version: 1.0 Sender: linux-crypto-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-crypto@vger.kernel.org This provides AVX, AVX-2, and AVX-512F implementations for Poly1305. The AVX-512F implementation is disabled on Skylake, due to throttling. These come from Andy Polyakov's implementation, with the following modifications from Samuel Neves: - Some cosmetic changes, like renaming labels to .Lname, constants, and other Linux conventions. - CPU feature checking is done in C by the glue code, so that has been removed from the assembly. - poly1305_blocks_avx512 jumped to the middle of the poly1305_blocks_avx2 for the final blocks. To appease objtool, the relevant tail avx2 code was duplicated for the avx512 function. - The original uses %rbp as a scratch register. However, the kernel expects %rbp to be a valid frame pointer at any given time in order to do proper unwinding. Thus we need to alter the code in order to preserve it. The most straightforward manner in which this was accomplished was by replacing $d3, formerly %r10, by %rdi, and replacing %rbp by %r10. Because %rdi, a pointer to the context structure, does not change and is not used by poly1305_iteration, it is safe to use it here, and the overhead of saving and restoring it should be minimal. - The original hardcodes returns as .byte 0xf3,0xc3, aka "rep ret". We replace this by "ret". "rep ret" was meant to help with AMD K8 chips, cf. http://repzret.org/p/repzret. It makes no sense to continue to use this kludge for code that won't even run on ancient AMD chips. While this is CRYPTOGAMS code, the originating code for this happens to be the same as OpenSSL's commit 4dfe4310c31c4483705991d9a798ce9be1ed1c68 Cycle counts on a Core i7 6700HQ using the AVX-2 codepath: size old new ---- ---- ---- 0 70 68 16 92 90 32 134 104 48 172 120 64 218 136 80 254 158 96 298 174 112 342 192 128 388 212 144 428 228 160 466 246 176 510 264 192 550 282 208 594 302 224 628 316 240 676 334 256 716 354 272 764 374 288 802 352 304 420 366 320 428 360 336 484 378 352 426 384 368 478 400 384 488 394 400 542 408 416 486 416 432 534 430 448 544 422 464 600 438 480 540 448 496 594 464 512 602 456 528 656 476 544 600 480 560 650 494 576 664 490 592 714 508 608 656 514 624 708 532 640 716 524 656 770 536 672 716 548 688 770 562 704 774 552 720 826 568 736 768 574 752 822 592 768 830 584 784 884 602 800 828 610 816 884 628 832 888 618 848 942 632 864 884 644 880 936 660 896 948 652 912 1000 664 928 942 676 944 994 690 960 1002 680 976 1054 694 992 1002 706 1008 1052 720 Cycle counts on a Xeon Gold 5120 using the AVX-512 codepath: size old new ---- ---- ---- 0 74 70 16 96 92 32 136 106 48 184 124 64 218 138 80 260 160 96 300 176 112 342 194 128 384 212 144 420 226 160 464 248 176 504 264 192 544 282 208 582 300 224 624 318 240 662 338 256 708 358 272 748 372 288 788 358 304 422 370 320 432 364 336 486 380 352 434 390 368 480 408 384 490 398 400 542 412 416 492 426 432 538 436 448 546 432 464 600 448 480 548 456 496 594 476 512 606 470 528 656 480 544 606 498 560 652 512 576 662 508 592 716 522 608 664 538 624 710 552 640 720 516 656 772 526 672 722 544 688 768 556 704 778 556 720 832 568 736 780 584 752 826 600 768 836 560 784 888 572 800 838 588 816 884 604 832 894 598 848 946 612 864 896 628 880 942 644 896 952 608 912 1004 616 928 954 634 944 1000 646 960 1008 646 976 1062 658 992 1012 674 1008 1058 690 Signed-off-by: Jason A. Donenfeld Signed-off-by: Samuel Neves Co-developed-by: Samuel Neves Based-on-code-from: Andy Polyakov Cc: Andy Lutomirski Cc: Greg KH Cc: Jean-Philippe Aumasson Cc: Andy Polyakov Cc: Thomas Gleixner Cc: Ingo Molnar Cc: x86@kernel.org --- lib/zinc/Makefile | 1 + lib/zinc/poly1305/poly1305-x86_64-glue.h | 157 ++ lib/zinc/poly1305/poly1305-x86_64.S | 2792 ++++++++++++++++++++++ lib/zinc/poly1305/poly1305.c | 4 +- 4 files changed, 2953 insertions(+), 1 deletion(-) create mode 100644 lib/zinc/poly1305/poly1305-x86_64-glue.h create mode 100644 lib/zinc/poly1305/poly1305-x86_64.S -- 2.19.0 diff --git a/lib/zinc/Makefile b/lib/zinc/Makefile index 6fc9626c55fa..a8943d960b6a 100644 --- a/lib/zinc/Makefile +++ b/lib/zinc/Makefile @@ -11,4 +11,5 @@ AFLAGS_chacha20-mips.o += -O2 # This is required to fill the branch delay slots obj-$(CONFIG_ZINC_CHACHA20) += zinc_chacha20.o zinc_poly1305-y := poly1305/poly1305.o +zinc_poly1305-$(CONFIG_ZINC_ARCH_X86_64) += poly1305/poly1305-x86_64.o obj-$(CONFIG_ZINC_POLY1305) += zinc_poly1305.o diff --git a/lib/zinc/poly1305/poly1305-x86_64-glue.h b/lib/zinc/poly1305/poly1305-x86_64-glue.h new file mode 100644 index 000000000000..3ca1bdd02c6b --- /dev/null +++ b/lib/zinc/poly1305/poly1305-x86_64-glue.h @@ -0,0 +1,157 @@ +/* SPDX-License-Identifier: GPL-2.0 OR MIT */ +/* + * Copyright (C) 2015-2018 Jason A. Donenfeld . All Rights Reserved. + */ + +#include +#include +#include + +asmlinkage void poly1305_init_x86_64(void *ctx, + const u8 key[POLY1305_KEY_SIZE]); +asmlinkage void poly1305_blocks_x86_64(void *ctx, const u8 *inp, + const size_t len, const u32 padbit); +asmlinkage void poly1305_emit_x86_64(void *ctx, u8 mac[POLY1305_MAC_SIZE], + const u32 nonce[4]); +#ifdef CONFIG_AS_AVX +asmlinkage void poly1305_emit_avx(void *ctx, u8 mac[POLY1305_MAC_SIZE], + const u32 nonce[4]); +asmlinkage void poly1305_blocks_avx(void *ctx, const u8 *inp, const size_t len, + const u32 padbit); +#endif +#ifdef CONFIG_AS_AVX2 +asmlinkage void poly1305_blocks_avx2(void *ctx, const u8 *inp, const size_t len, + const u32 padbit); +#endif +#ifdef CONFIG_AS_AVX512 +asmlinkage void poly1305_blocks_avx512(void *ctx, const u8 *inp, + const size_t len, const u32 padbit); +#endif + +static bool poly1305_use_avx __ro_after_init; +static bool poly1305_use_avx2 __ro_after_init; +static bool poly1305_use_avx512 __ro_after_init; + +static void __init poly1305_fpu_init(void) +{ + poly1305_use_avx = + boot_cpu_has(X86_FEATURE_AVX) && + cpu_has_xfeatures(XFEATURE_MASK_SSE | XFEATURE_MASK_YMM, NULL); + poly1305_use_avx2 = + boot_cpu_has(X86_FEATURE_AVX) && + boot_cpu_has(X86_FEATURE_AVX2) && + cpu_has_xfeatures(XFEATURE_MASK_SSE | XFEATURE_MASK_YMM, NULL); + poly1305_use_avx512 = + boot_cpu_has(X86_FEATURE_AVX) && + boot_cpu_has(X86_FEATURE_AVX2) && + boot_cpu_has(X86_FEATURE_AVX512F) && + cpu_has_xfeatures(XFEATURE_MASK_SSE | XFEATURE_MASK_YMM | + XFEATURE_MASK_AVX512, NULL) && + /* Skylake downclocks unacceptably much when using zmm. */ + boot_cpu_data.x86_model != INTEL_FAM6_SKYLAKE_X; +} + +static inline bool poly1305_init_arch(void *ctx, + const u8 key[POLY1305_KEY_SIZE]) +{ + poly1305_init_x86_64(ctx, key); + return true; +} + +struct poly1305_arch_internal { + union { + struct { + u32 h[5]; + u32 is_base2_26; + }; + u64 hs[3]; + }; + u64 r[2]; + u64 pad; + struct { u32 r2, r1, r4, r3; } rn[9]; +}; + +static void convert_to_base2_64(void *ctx) +{ + struct poly1305_arch_internal *state = ctx; + u32 cy; + + if (!state->is_base2_26) + return; + + cy = state->h[0] >> 26; state->h[0] &= 0x3ffffff; state->h[1] += cy; + cy = state->h[1] >> 26; state->h[1] &= 0x3ffffff; state->h[2] += cy; + cy = state->h[2] >> 26; state->h[2] &= 0x3ffffff; state->h[3] += cy; + cy = state->h[3] >> 26; state->h[3] &= 0x3ffffff; state->h[4] += cy; + state->hs[0] = ((u64)state->h[2] << 52) | ((u64)state->h[1] << 26) | state->h[0]; + state->hs[1] = ((u64)state->h[4] << 40) | ((u64)state->h[3] << 14) | (state->h[2] >> 12); + state->hs[2] = state->h[4] >> 24; +#define ULT(a, b) ((a ^ ((a ^ b) | ((a - b) ^ b))) >> (sizeof(a) * 8 - 1)) + cy = (state->hs[2] >> 2) + (state->hs[2] & ~3ULL); + state->hs[2] &= 3; + state->hs[0] += cy; + state->hs[1] += (cy = ULT(state->hs[0], cy)); + state->hs[2] += ULT(state->hs[1], cy); +#undef ULT + state->is_base2_26 = 0; +} + +static inline bool poly1305_blocks_arch(void *ctx, const u8 *inp, + const size_t len, const u32 padbit, + simd_context_t *simd_context) +{ + struct poly1305_arch_internal *state = ctx; + + if (!poly1305_use_avx || + (len < (POLY1305_BLOCK_SIZE * 18) && !state->is_base2_26) || + !simd_use(simd_context)) + goto scalar; + +#ifdef CONFIG_AS_AVX512 + if (poly1305_use_avx512) { + poly1305_blocks_avx512(ctx, inp, len, padbit); + return true; + } +#endif + +#ifdef CONFIG_AS_AVX2 + if (poly1305_use_avx2) { + poly1305_blocks_avx2(ctx, inp, len, padbit); + return true; + } +#endif + +#ifdef CONFIG_AS_AVX + if (poly1305_use_avx) { + poly1305_blocks_avx(ctx, inp, len, padbit); + return true; + } +#endif + +scalar: + convert_to_base2_64(ctx); + poly1305_blocks_x86_64(ctx, inp, len, padbit); + return true; +} + +static inline bool poly1305_emit_arch(void *ctx, u8 mac[POLY1305_MAC_SIZE], + const u32 nonce[4], + simd_context_t *simd_context) +{ + struct poly1305_arch_internal *state = ctx; + + if (!poly1305_use_avx || !state->is_base2_26 ||!simd_use(simd_context)) + goto scalar; + +#ifdef CONFIG_AS_AVX + if (poly1305_use_avx || poly1305_use_avx2 || poly1305_use_avx512) { + poly1305_emit_avx(ctx, mac, nonce); + return true; + } +#endif + +scalar: + convert_to_base2_64(ctx); + poly1305_emit_x86_64(ctx, mac, nonce); + return true; +} diff --git a/lib/zinc/poly1305/poly1305-x86_64.S b/lib/zinc/poly1305/poly1305-x86_64.S new file mode 100644 index 000000000000..3c3f2b4d880b --- /dev/null +++ b/lib/zinc/poly1305/poly1305-x86_64.S @@ -0,0 +1,2792 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2017 Samuel Neves . All Rights Reserved. + * Copyright (C) 2015-2018 Jason A. Donenfeld . All Rights Reserved. + * Copyright (C) 2006-2017 CRYPTOGAMS by . All Rights Reserved. + * + * This is based in part on Andy Polyakov's implementation from CRYPTOGAMS. + */ + +#include + +.section .rodata.cst192.Lconst, "aM", @progbits, 192 +.align 64 +.Lconst: +.long 0x0ffffff,0,0x0ffffff,0,0x0ffffff,0,0x0ffffff,0 +.long 16777216,0,16777216,0,16777216,0,16777216,0 +.long 0x3ffffff,0,0x3ffffff,0,0x3ffffff,0,0x3ffffff,0 +.long 2,2,2,3,2,0,2,1 +.long 0,0,0,1, 0,2,0,3, 0,4,0,5, 0,6,0,7 + +.text + +.align 32 +ENTRY(poly1305_init_x86_64) + xorq %rax,%rax + movq %rax,0(%rdi) + movq %rax,8(%rdi) + movq %rax,16(%rdi) + + cmpq $0,%rsi + je .Lno_key + + movq $0x0ffffffc0fffffff,%rax + movq $0x0ffffffc0ffffffc,%rcx + andq 0(%rsi),%rax + andq 8(%rsi),%rcx + movq %rax,24(%rdi) + movq %rcx,32(%rdi) + movl $1,%eax +.Lno_key: + ret +ENDPROC(poly1305_init_x86_64) + +.align 32 +ENTRY(poly1305_blocks_x86_64) +.Lblocks: + shrq $4,%rdx + jz .Lno_data + + pushq %rbx + pushq %r12 + pushq %r13 + pushq %r14 + pushq %r15 + pushq %rdi + +.Lblocks_body: + + movq %rdx,%r15 + + movq 24(%rdi),%r11 + movq 32(%rdi),%r13 + + movq 0(%rdi),%r14 + movq 8(%rdi),%rbx + movq 16(%rdi),%r10 + + movq %r13,%r12 + shrq $2,%r13 + movq %r12,%rax + addq %r12,%r13 + jmp .Loop + +.align 32 +.Loop: + + addq 0(%rsi),%r14 + adcq 8(%rsi),%rbx + leaq 16(%rsi),%rsi + adcq %rcx,%r10 + mulq %r14 + movq %rax,%r9 + movq %r11,%rax + movq %rdx,%rdi + + mulq %r14 + movq %rax,%r14 + movq %r11,%rax + movq %rdx,%r8 + + mulq %rbx + addq %rax,%r9 + movq %r13,%rax + adcq %rdx,%rdi + + mulq %rbx + movq %r10,%rbx + addq %rax,%r14 + adcq %rdx,%r8 + + imulq %r13,%rbx + addq %rbx,%r9 + movq %r8,%rbx + adcq $0,%rdi + + imulq %r11,%r10 + addq %r9,%rbx + movq $-4,%rax + adcq %r10,%rdi + + andq %rdi,%rax + movq %rdi,%r10 + shrq $2,%rdi + andq $3,%r10 + addq %rdi,%rax + addq %rax,%r14 + adcq $0,%rbx + adcq $0,%r10 + + movq %r12,%rax + decq %r15 + jnz .Loop + + movq 0(%rsp),%rdi + + movq %r14,0(%rdi) + movq %rbx,8(%rdi) + movq %r10,16(%rdi) + + movq 8(%rsp),%r15 + movq 16(%rsp),%r14 + movq 24(%rsp),%r13 + movq 32(%rsp),%r12 + movq 40(%rsp),%rbx + leaq 48(%rsp),%rsp +.Lno_data: +.Lblocks_epilogue: + ret +ENDPROC(poly1305_blocks_x86_64) + +.align 32 +ENTRY(poly1305_emit_x86_64) +.Lemit: + movq 0(%rdi),%r8 + movq 8(%rdi),%r9 + movq 16(%rdi),%r10 + + movq %r8,%rax + addq $5,%r8 + movq %r9,%rcx + adcq $0,%r9 + adcq $0,%r10 + shrq $2,%r10 + cmovnzq %r8,%rax + cmovnzq %r9,%rcx + + addq 0(%rdx),%rax + adcq 8(%rdx),%rcx + movq %rax,0(%rsi) + movq %rcx,8(%rsi) + + ret +ENDPROC(poly1305_emit_x86_64) + +.macro __poly1305_block + mulq %r14 + movq %rax,%r9 + movq %r11,%rax + movq %rdx,%rdi + + mulq %r14 + movq %rax,%r14 + movq %r11,%rax + movq %rdx,%r8 + + mulq %rbx + addq %rax,%r9 + movq %r13,%rax + adcq %rdx,%rdi + + mulq %rbx + movq %r10,%rbx + addq %rax,%r14 + adcq %rdx,%r8 + + imulq %r13,%rbx + addq %rbx,%r9 + movq %r8,%rbx + adcq $0,%rdi + + imulq %r11,%r10 + addq %r9,%rbx + movq $-4,%rax + adcq %r10,%rdi + + andq %rdi,%rax + movq %rdi,%r10 + shrq $2,%rdi + andq $3,%r10 + addq %rdi,%rax + addq %rax,%r14 + adcq $0,%rbx + adcq $0,%r10 +.endm + +.macro __poly1305_init_avx + movq %r11,%r14 + movq %r12,%rbx + xorq %r10,%r10 + + leaq 48+64(%rdi),%rdi + + movq %r12,%rax + movq %rdi,0(%rsp) + __poly1305_block + movq 0(%rsp),%rdi + + movl $0x3ffffff,%eax + movl $0x3ffffff,%edx + movq %r14,%r8 + andl %r14d,%eax + movq %r11,%r9 + andl %r11d,%edx + movl %eax,-64(%rdi) + shrq $26,%r8 + movl %edx,-60(%rdi) + shrq $26,%r9 + + movl $0x3ffffff,%eax + movl $0x3ffffff,%edx + andl %r8d,%eax + andl %r9d,%edx + movl %eax,-48(%rdi) + leal (%rax,%rax,4),%eax + movl %edx,-44(%rdi) + leal (%rdx,%rdx,4),%edx + movl %eax,-32(%rdi) + shrq $26,%r8 + movl %edx,-28(%rdi) + shrq $26,%r9 + + movq %rbx,%rax + movq %r12,%rdx + shlq $12,%rax + shlq $12,%rdx + orq %r8,%rax + orq %r9,%rdx + andl $0x3ffffff,%eax + andl $0x3ffffff,%edx + movl %eax,-16(%rdi) + leal (%rax,%rax,4),%eax + movl %edx,-12(%rdi) + leal (%rdx,%rdx,4),%edx + movl %eax,0(%rdi) + movq %rbx,%r8 + movl %edx,4(%rdi) + movq %r12,%r9 + + movl $0x3ffffff,%eax + movl $0x3ffffff,%edx + shrq $14,%r8 + shrq $14,%r9 + andl %r8d,%eax + andl %r9d,%edx + movl %eax,16(%rdi) + leal (%rax,%rax,4),%eax + movl %edx,20(%rdi) + leal (%rdx,%rdx,4),%edx + movl %eax,32(%rdi) + shrq $26,%r8 + movl %edx,36(%rdi) + shrq $26,%r9 + + movq %r10,%rax + shlq $24,%rax + orq %rax,%r8 + movl %r8d,48(%rdi) + leaq (%r8,%r8,4),%r8 + movl %r9d,52(%rdi) + leaq (%r9,%r9,4),%r9 + movl %r8d,64(%rdi) + movl %r9d,68(%rdi) + + movq %r12,%rax + movq %rdi,0(%rsp) + __poly1305_block + movq 0(%rsp),%rdi + + movl $0x3ffffff,%eax + movq %r14,%r8 + andl %r14d,%eax + shrq $26,%r8 + movl %eax,-52(%rdi) + + movl $0x3ffffff,%edx + andl %r8d,%edx + movl %edx,-36(%rdi) + leal (%rdx,%rdx,4),%edx + shrq $26,%r8 + movl %edx,-20(%rdi) + + movq %rbx,%rax + shlq $12,%rax + orq %r8,%rax + andl $0x3ffffff,%eax + movl %eax,-4(%rdi) + leal (%rax,%rax,4),%eax + movq %rbx,%r8 + movl %eax,12(%rdi) + + movl $0x3ffffff,%edx + shrq $14,%r8 + andl %r8d,%edx + movl %edx,28(%rdi) + leal (%rdx,%rdx,4),%edx + shrq $26,%r8 + movl %edx,44(%rdi) + + movq %r10,%rax + shlq $24,%rax + orq %rax,%r8 + movl %r8d,60(%rdi) + leaq (%r8,%r8,4),%r8 + movl %r8d,76(%rdi) + + movq %r12,%rax + movq %rdi,0(%rsp) + __poly1305_block + movq 0(%rsp),%rdi + + movl $0x3ffffff,%eax + movq %r14,%r8 + andl %r14d,%eax + shrq $26,%r8 + movl %eax,-56(%rdi) + + movl $0x3ffffff,%edx + andl %r8d,%edx + movl %edx,-40(%rdi) + leal (%rdx,%rdx,4),%edx + shrq $26,%r8 + movl %edx,-24(%rdi) + + movq %rbx,%rax + shlq $12,%rax + orq %r8,%rax + andl $0x3ffffff,%eax + movl %eax,-8(%rdi) + leal (%rax,%rax,4),%eax + movq %rbx,%r8 + movl %eax,8(%rdi) + + movl $0x3ffffff,%edx + shrq $14,%r8 + andl %r8d,%edx + movl %edx,24(%rdi) + leal (%rdx,%rdx,4),%edx + shrq $26,%r8 + movl %edx,40(%rdi) + + movq %r10,%rax + shlq $24,%rax + orq %rax,%r8 + movl %r8d,56(%rdi) + leaq (%r8,%r8,4),%r8 + movl %r8d,72(%rdi) + + leaq -48-64(%rdi),%rdi +.endm + +#ifdef CONFIG_AS_AVX +.align 32 +ENTRY(poly1305_blocks_avx) + + movl 20(%rdi),%r8d + cmpq $128,%rdx + jae .Lblocks_avx + testl %r8d,%r8d + jz .Lblocks + +.Lblocks_avx: + andq $-16,%rdx + jz .Lno_data_avx + + vzeroupper + + testl %r8d,%r8d + jz .Lbase2_64_avx + + testq $31,%rdx + jz .Leven_avx + + pushq %rbx + pushq %r12 + pushq %r13 + pushq %r14 + pushq %r15 + pushq %rdi + +.Lblocks_avx_body: + + movq %rdx,%r15 + + movq 0(%rdi),%r8 + movq 8(%rdi),%r9 + movl 16(%rdi),%r10d + + movq 24(%rdi),%r11 + movq 32(%rdi),%r13 + + + movl %r8d,%r14d + andq $-2147483648,%r8 + movq %r9,%r12 + movl %r9d,%ebx + andq $-2147483648,%r9 + + shrq $6,%r8 + shlq $52,%r12 + addq %r8,%r14 + shrq $12,%rbx + shrq $18,%r9 + addq %r12,%r14 + adcq %r9,%rbx + + movq %r10,%r8 + shlq $40,%r8 + shrq $24,%r10 + addq %r8,%rbx + adcq $0,%r10 + + movq $-4,%r9 + movq %r10,%r8 + andq %r10,%r9 + shrq $2,%r8 + andq $3,%r10 + addq %r9,%r8 + addq %r8,%r14 + adcq $0,%rbx + adcq $0,%r10 + + movq %r13,%r12 + movq %r13,%rax + shrq $2,%r13 + addq %r12,%r13 + + addq 0(%rsi),%r14 + adcq 8(%rsi),%rbx + leaq 16(%rsi),%rsi + adcq %rcx,%r10 + + movq %rdi,0(%rsp) + __poly1305_block + movq 0(%rsp),%rdi + + testq %rcx,%rcx + jz .Lstore_base2_64_avx + + + movq %r14,%rax + movq %r14,%rdx + shrq $52,%r14 + movq %rbx,%r11 + movq %rbx,%r12 + shrq $26,%rdx + andq $0x3ffffff,%rax + shlq $12,%r11 + andq $0x3ffffff,%rdx + shrq $14,%rbx + orq %r11,%r14 + shlq $24,%r10 + andq $0x3ffffff,%r14 + shrq $40,%r12 + andq $0x3ffffff,%rbx + orq %r12,%r10 + + subq $16,%r15 + jz .Lstore_base2_26_avx + + vmovd %eax,%xmm0 + vmovd %edx,%xmm1 + vmovd %r14d,%xmm2 + vmovd %ebx,%xmm3 + vmovd %r10d,%xmm4 + jmp .Lproceed_avx + +.align 32 +.Lstore_base2_64_avx: + movq %r14,0(%rdi) + movq %rbx,8(%rdi) + movq %r10,16(%rdi) + jmp .Ldone_avx + +.align 16 +.Lstore_base2_26_avx: + movl %eax,0(%rdi) + movl %edx,4(%rdi) + movl %r14d,8(%rdi) + movl %ebx,12(%rdi) + movl %r10d,16(%rdi) +.align 16 +.Ldone_avx: + movq 8(%rsp),%r15 + movq 16(%rsp),%r14 + movq 24(%rsp),%r13 + movq 32(%rsp),%r12 + movq 40(%rsp),%rbx + leaq 48(%rsp),%rsp + +.Lno_data_avx: +.Lblocks_avx_epilogue: + ret + +.align 32 +.Lbase2_64_avx: + + pushq %rbx + pushq %r12 + pushq %r13 + pushq %r14 + pushq %r15 + pushq %rdi + +.Lbase2_64_avx_body: + + movq %rdx,%r15 + + movq 24(%rdi),%r11 + movq 32(%rdi),%r13 + + movq 0(%rdi),%r14 + movq 8(%rdi),%rbx + movl 16(%rdi),%r10d + + movq %r13,%r12 + movq %r13,%rax + shrq $2,%r13 + addq %r12,%r13 + + testq $31,%rdx + jz .Linit_avx + + addq 0(%rsi),%r14 + adcq 8(%rsi),%rbx + leaq 16(%rsi),%rsi + adcq %rcx,%r10 + subq $16,%r15 + + movq %rdi,0(%rsp) + __poly1305_block + movq 0(%rsp),%rdi + +.Linit_avx: + + movq %r14,%rax + movq %r14,%rdx + shrq $52,%r14 + movq %rbx,%r8 + movq %rbx,%r9 + shrq $26,%rdx + andq $0x3ffffff,%rax + shlq $12,%r8 + andq $0x3ffffff,%rdx + shrq $14,%rbx + orq %r8,%r14 + shlq $24,%r10 + andq $0x3ffffff,%r14 + shrq $40,%r9 + andq $0x3ffffff,%rbx + orq %r9,%r10 + + vmovd %eax,%xmm0 + vmovd %edx,%xmm1 + vmovd %r14d,%xmm2 + vmovd %ebx,%xmm3 + vmovd %r10d,%xmm4 + movl $1,20(%rdi) + + __poly1305_init_avx + +.Lproceed_avx: + movq %r15,%rdx + + movq 8(%rsp),%r15 + movq 16(%rsp),%r14 + movq 24(%rsp),%r13 + movq 32(%rsp),%r12 + movq 40(%rsp),%rbx + leaq 48(%rsp),%rax + leaq 48(%rsp),%rsp + +.Lbase2_64_avx_epilogue: + jmp .Ldo_avx + + +.align 32 +.Leven_avx: + vmovd 0(%rdi),%xmm0 + vmovd 4(%rdi),%xmm1 + vmovd 8(%rdi),%xmm2 + vmovd 12(%rdi),%xmm3 + vmovd 16(%rdi),%xmm4 + +.Ldo_avx: + leaq 8(%rsp),%r10 + andq $-32,%rsp + subq $8,%rsp + leaq -88(%rsp),%r11 + subq $0x178,%rsp + subq $64,%rdx + leaq -32(%rsi),%rax + cmovcq %rax,%rsi + + vmovdqu 48(%rdi),%xmm14 + leaq 112(%rdi),%rdi + leaq .Lconst(%rip),%rcx + + vmovdqu 32(%rsi),%xmm5 + vmovdqu 48(%rsi),%xmm6 + vmovdqa 64(%rcx),%xmm15 + + vpsrldq $6,%xmm5,%xmm7 + vpsrldq $6,%xmm6,%xmm8 + vpunpckhqdq %xmm6,%xmm5,%xmm9 + vpunpcklqdq %xmm6,%xmm5,%xmm5 + vpunpcklqdq %xmm8,%xmm7,%xmm8 + + vpsrlq $40,%xmm9,%xmm9 + vpsrlq $26,%xmm5,%xmm6 + vpand %xmm15,%xmm5,%xmm5 + vpsrlq $4,%xmm8,%xmm7 + vpand %xmm15,%xmm6,%xmm6 + vpsrlq $30,%xmm8,%xmm8 + vpand %xmm15,%xmm7,%xmm7 + vpand %xmm15,%xmm8,%xmm8 + vpor 32(%rcx),%xmm9,%xmm9 + + jbe .Lskip_loop_avx + + + vmovdqu -48(%rdi),%xmm11 + vmovdqu -32(%rdi),%xmm12 + vpshufd $0xEE,%xmm14,%xmm13 + vpshufd $0x44,%xmm14,%xmm10 + vmovdqa %xmm13,-144(%r11) + vmovdqa %xmm10,0(%rsp) + vpshufd $0xEE,%xmm11,%xmm14 + vmovdqu -16(%rdi),%xmm10 + vpshufd $0x44,%xmm11,%xmm11 + vmovdqa %xmm14,-128(%r11) + vmovdqa %xmm11,16(%rsp) + vpshufd $0xEE,%xmm12,%xmm13 + vmovdqu 0(%rdi),%xmm11 + vpshufd $0x44,%xmm12,%xmm12 + vmovdqa %xmm13,-112(%r11) + vmovdqa %xmm12,32(%rsp) + vpshufd $0xEE,%xmm10,%xmm14 + vmovdqu 16(%rdi),%xmm12 + vpshufd $0x44,%xmm10,%xmm10 + vmovdqa %xmm14,-96(%r11) + vmovdqa %xmm10,48(%rsp) + vpshufd $0xEE,%xmm11,%xmm13 + vmovdqu 32(%rdi),%xmm10 + vpshufd $0x44,%xmm11,%xmm11 + vmovdqa %xmm13,-80(%r11) + vmovdqa %xmm11,64(%rsp) + vpshufd $0xEE,%xmm12,%xmm14 + vmovdqu 48(%rdi),%xmm11 + vpshufd $0x44,%xmm12,%xmm12 + vmovdqa %xmm14,-64(%r11) + vmovdqa %xmm12,80(%rsp) + vpshufd $0xEE,%xmm10,%xmm13 + vmovdqu 64(%rdi),%xmm12 + vpshufd $0x44,%xmm10,%xmm10 + vmovdqa %xmm13,-48(%r11) + vmovdqa %xmm10,96(%rsp) + vpshufd $0xEE,%xmm11,%xmm14 + vpshufd $0x44,%xmm11,%xmm11 + vmovdqa %xmm14,-32(%r11) + vmovdqa %xmm11,112(%rsp) + vpshufd $0xEE,%xmm12,%xmm13 + vmovdqa 0(%rsp),%xmm14 + vpshufd $0x44,%xmm12,%xmm12 + vmovdqa %xmm13,-16(%r11) + vmovdqa %xmm12,128(%rsp) + + jmp .Loop_avx + +.align 32 +.Loop_avx: + + vpmuludq %xmm5,%xmm14,%xmm10 + vpmuludq %xmm6,%xmm14,%xmm11 + vmovdqa %xmm2,32(%r11) + vpmuludq %xmm7,%xmm14,%xmm12 + vmovdqa 16(%rsp),%xmm2 + vpmuludq %xmm8,%xmm14,%xmm13 + vpmuludq %xmm9,%xmm14,%xmm14 + + vmovdqa %xmm0,0(%r11) + vpmuludq 32(%rsp),%xmm9,%xmm0 + vmovdqa %xmm1,16(%r11) + vpmuludq %xmm8,%xmm2,%xmm1 + vpaddq %xmm0,%xmm10,%xmm10 + vpaddq %xmm1,%xmm14,%xmm14 + vmovdqa %xmm3,48(%r11) + vpmuludq %xmm7,%xmm2,%xmm0 + vpmuludq %xmm6,%xmm2,%xmm1 + vpaddq %xmm0,%xmm13,%xmm13 + vmovdqa 48(%rsp),%xmm3 + vpaddq %xmm1,%xmm12,%xmm12 + vmovdqa %xmm4,64(%r11) + vpmuludq %xmm5,%xmm2,%xmm2 + vpmuludq %xmm7,%xmm3,%xmm0 + vpaddq %xmm2,%xmm11,%xmm11 + + vmovdqa 64(%rsp),%xmm4 + vpaddq %xmm0,%xmm14,%xmm14 + vpmuludq %xmm6,%xmm3,%xmm1 + vpmuludq %xmm5,%xmm3,%xmm3 + vpaddq %xmm1,%xmm13,%xmm13 + vmovdqa 80(%rsp),%xmm2 + vpaddq %xmm3,%xmm12,%xmm12 + vpmuludq %xmm9,%xmm4,%xmm0 + vpmuludq %xmm8,%xmm4,%xmm4 + vpaddq %xmm0,%xmm11,%xmm11 + vmovdqa 96(%rsp),%xmm3 + vpaddq %xmm4,%xmm10,%xmm10 + + vmovdqa 128(%rsp),%xmm4 + vpmuludq %xmm6,%xmm2,%xmm1 + vpmuludq %xmm5,%xmm2,%xmm2 + vpaddq %xmm1,%xmm14,%xmm14 + vpaddq %xmm2,%xmm13,%xmm13 + vpmuludq %xmm9,%xmm3,%xmm0 + vpmuludq %xmm8,%xmm3,%xmm1 + vpaddq %xmm0,%xmm12,%xmm12 + vmovdqu 0(%rsi),%xmm0 + vpaddq %xmm1,%xmm11,%xmm11 + vpmuludq %xmm7,%xmm3,%xmm3 + vpmuludq %xmm7,%xmm4,%xmm7 + vpaddq %xmm3,%xmm10,%xmm10 + + vmovdqu 16(%rsi),%xmm1 + vpaddq %xmm7,%xmm11,%xmm11 + vpmuludq %xmm8,%xmm4,%xmm8 + vpmuludq %xmm9,%xmm4,%xmm9 + vpsrldq $6,%xmm0,%xmm2 + vpaddq %xmm8,%xmm12,%xmm12 + vpaddq %xmm9,%xmm13,%xmm13 + vpsrldq $6,%xmm1,%xmm3 + vpmuludq 112(%rsp),%xmm5,%xmm9 + vpmuludq %xmm6,%xmm4,%xmm5 + vpunpckhqdq %xmm1,%xmm0,%xmm4 + vpaddq %xmm9,%xmm14,%xmm14 + vmovdqa -144(%r11),%xmm9 + vpaddq %xmm5,%xmm10,%xmm10 + + vpunpcklqdq %xmm1,%xmm0,%xmm0 + vpunpcklqdq %xmm3,%xmm2,%xmm3 + + + vpsrldq $5,%xmm4,%xmm4 + vpsrlq $26,%xmm0,%xmm1 + vpand %xmm15,%xmm0,%xmm0 + vpsrlq $4,%xmm3,%xmm2 + vpand %xmm15,%xmm1,%xmm1 + vpand 0(%rcx),%xmm4,%xmm4 + vpsrlq $30,%xmm3,%xmm3 + vpand %xmm15,%xmm2,%xmm2 + vpand %xmm15,%xmm3,%xmm3 + vpor 32(%rcx),%xmm4,%xmm4 + + vpaddq 0(%r11),%xmm0,%xmm0 + vpaddq 16(%r11),%xmm1,%xmm1 + vpaddq 32(%r11),%xmm2,%xmm2 + vpaddq 48(%r11),%xmm3,%xmm3 + vpaddq 64(%r11),%xmm4,%xmm4 + + leaq 32(%rsi),%rax + leaq 64(%rsi),%rsi + subq $64,%rdx + cmovcq %rax,%rsi + + vpmuludq %xmm0,%xmm9,%xmm5 + vpmuludq %xmm1,%xmm9,%xmm6 + vpaddq %xmm5,%xmm10,%xmm10 + vpaddq %xmm6,%xmm11,%xmm11 + vmovdqa -128(%r11),%xmm7 + vpmuludq %xmm2,%xmm9,%xmm5 + vpmuludq %xmm3,%xmm9,%xmm6 + vpaddq %xmm5,%xmm12,%xmm12 + vpaddq %xmm6,%xmm13,%xmm13 + vpmuludq %xmm4,%xmm9,%xmm9 + vpmuludq -112(%r11),%xmm4,%xmm5 + vpaddq %xmm9,%xmm14,%xmm14 + + vpaddq %xmm5,%xmm10,%xmm10 + vpmuludq %xmm2,%xmm7,%xmm6 + vpmuludq %xmm3,%xmm7,%xmm5 + vpaddq %xmm6,%xmm13,%xmm13 + vmovdqa -96(%r11),%xmm8 + vpaddq %xmm5,%xmm14,%xmm14 + vpmuludq %xmm1,%xmm7,%xmm6 + vpmuludq %xmm0,%xmm7,%xmm7 + vpaddq %xmm6,%xmm12,%xmm12 + vpaddq %xmm7,%xmm11,%xmm11 + + vmovdqa -80(%r11),%xmm9 + vpmuludq %xmm2,%xmm8,%xmm5 + vpmuludq %xmm1,%xmm8,%xmm6 + vpaddq %xmm5,%xmm14,%xmm14 + vpaddq %xmm6,%xmm13,%xmm13 + vmovdqa -64(%r11),%xmm7 + vpmuludq %xmm0,%xmm8,%xmm8 + vpmuludq %xmm4,%xmm9,%xmm5 + vpaddq %xmm8,%xmm12,%xmm12 + vpaddq %xmm5,%xmm11,%xmm11 + vmovdqa -48(%r11),%xmm8 + vpmuludq %xmm3,%xmm9,%xmm9 + vpmuludq %xmm1,%xmm7,%xmm6 + vpaddq %xmm9,%xmm10,%xmm10 + + vmovdqa -16(%r11),%xmm9 + vpaddq %xmm6,%xmm14,%xmm14 + vpmuludq %xmm0,%xmm7,%xmm7 + vpmuludq %xmm4,%xmm8,%xmm5 + vpaddq %xmm7,%xmm13,%xmm13 + vpaddq %xmm5,%xmm12,%xmm12 + vmovdqu 32(%rsi),%xmm5 + vpmuludq %xmm3,%xmm8,%xmm7 + vpmuludq %xmm2,%xmm8,%xmm8 + vpaddq %xmm7,%xmm11,%xmm11 + vmovdqu 48(%rsi),%xmm6 + vpaddq %xmm8,%xmm10,%xmm10 + + vpmuludq %xmm2,%xmm9,%xmm2 + vpmuludq %xmm3,%xmm9,%xmm3 + vpsrldq $6,%xmm5,%xmm7 + vpaddq %xmm2,%xmm11,%xmm11 + vpmuludq %xmm4,%xmm9,%xmm4 + vpsrldq $6,%xmm6,%xmm8 + vpaddq %xmm3,%xmm12,%xmm2 + vpaddq %xmm4,%xmm13,%xmm3 + vpmuludq -32(%r11),%xmm0,%xmm4 + vpmuludq %xmm1,%xmm9,%xmm0 + vpunpckhqdq %xmm6,%xmm5,%xmm9 + vpaddq %xmm4,%xmm14,%xmm4 + vpaddq %xmm0,%xmm10,%xmm0 + + vpunpcklqdq %xmm6,%xmm5,%xmm5 + vpunpcklqdq %xmm8,%xmm7,%xmm8 + + + vpsrldq $5,%xmm9,%xmm9 + vpsrlq $26,%xmm5,%xmm6 + vmovdqa 0(%rsp),%xmm14 + vpand %xmm15,%xmm5,%xmm5 + vpsrlq $4,%xmm8,%xmm7 + vpand %xmm15,%xmm6,%xmm6 + vpand 0(%rcx),%xmm9,%xmm9 + vpsrlq $30,%xmm8,%xmm8 + vpand %xmm15,%xmm7,%xmm7 + vpand %xmm15,%xmm8,%xmm8 + vpor 32(%rcx),%xmm9,%xmm9 + + vpsrlq $26,%xmm3,%xmm13 + vpand %xmm15,%xmm3,%xmm3 + vpaddq %xmm13,%xmm4,%xmm4 + + vpsrlq $26,%xmm0,%xmm10 + vpand %xmm15,%xmm0,%xmm0 + vpaddq %xmm10,%xmm11,%xmm1 + + vpsrlq $26,%xmm4,%xmm10 + vpand %xmm15,%xmm4,%xmm4 + + vpsrlq $26,%xmm1,%xmm11 + vpand %xmm15,%xmm1,%xmm1 + vpaddq %xmm11,%xmm2,%xmm2 + + vpaddq %xmm10,%xmm0,%xmm0 + vpsllq $2,%xmm10,%xmm10 + vpaddq %xmm10,%xmm0,%xmm0 + + vpsrlq $26,%xmm2,%xmm12 + vpand %xmm15,%xmm2,%xmm2 + vpaddq %xmm12,%xmm3,%xmm3 + + vpsrlq $26,%xmm0,%xmm10 + vpand %xmm15,%xmm0,%xmm0 + vpaddq %xmm10,%xmm1,%xmm1 + + vpsrlq $26,%xmm3,%xmm13 + vpand %xmm15,%xmm3,%xmm3 + vpaddq %xmm13,%xmm4,%xmm4 + + ja .Loop_avx + +.Lskip_loop_avx: + vpshufd $0x10,%xmm14,%xmm14 + addq $32,%rdx + jnz .Long_tail_avx + + vpaddq %xmm2,%xmm7,%xmm7 + vpaddq %xmm0,%xmm5,%xmm5 + vpaddq %xmm1,%xmm6,%xmm6 + vpaddq %xmm3,%xmm8,%xmm8 + vpaddq %xmm4,%xmm9,%xmm9 + +.Long_tail_avx: + vmovdqa %xmm2,32(%r11) + vmovdqa %xmm0,0(%r11) + vmovdqa %xmm1,16(%r11) + vmovdqa %xmm3,48(%r11) + vmovdqa %xmm4,64(%r11) + + vpmuludq %xmm7,%xmm14,%xmm12 + vpmuludq %xmm5,%xmm14,%xmm10 + vpshufd $0x10,-48(%rdi),%xmm2 + vpmuludq %xmm6,%xmm14,%xmm11 + vpmuludq %xmm8,%xmm14,%xmm13 + vpmuludq %xmm9,%xmm14,%xmm14 + + vpmuludq %xmm8,%xmm2,%xmm0 + vpaddq %xmm0,%xmm14,%xmm14 + vpshufd $0x10,-32(%rdi),%xmm3 + vpmuludq %xmm7,%xmm2,%xmm1 + vpaddq %xmm1,%xmm13,%xmm13 + vpshufd $0x10,-16(%rdi),%xmm4 + vpmuludq %xmm6,%xmm2,%xmm0 + vpaddq %xmm0,%xmm12,%xmm12 + vpmuludq %xmm5,%xmm2,%xmm2 + vpaddq %xmm2,%xmm11,%xmm11 + vpmuludq %xmm9,%xmm3,%xmm3 + vpaddq %xmm3,%xmm10,%xmm10 + + vpshufd $0x10,0(%rdi),%xmm2 + vpmuludq %xmm7,%xmm4,%xmm1 + vpaddq %xmm1,%xmm14,%xmm14 + vpmuludq %xmm6,%xmm4,%xmm0 + vpaddq %xmm0,%xmm13,%xmm13 + vpshufd $0x10,16(%rdi),%xmm3 + vpmuludq %xmm5,%xmm4,%xmm4 + vpaddq %xmm4,%xmm12,%xmm12 + vpmuludq %xmm9,%xmm2,%xmm1 + vpaddq %xmm1,%xmm11,%xmm11 + vpshufd $0x10,32(%rdi),%xmm4 + vpmuludq %xmm8,%xmm2,%xmm2 + vpaddq %xmm2,%xmm10,%xmm10 + + vpmuludq %xmm6,%xmm3,%xmm0 + vpaddq %xmm0,%xmm14,%xmm14 + vpmuludq %xmm5,%xmm3,%xmm3 + vpaddq %xmm3,%xmm13,%xmm13 + vpshufd $0x10,48(%rdi),%xmm2 + vpmuludq %xmm9,%xmm4,%xmm1 + vpaddq %xmm1,%xmm12,%xmm12 + vpshufd $0x10,64(%rdi),%xmm3 + vpmuludq %xmm8,%xmm4,%xmm0 + vpaddq %xmm0,%xmm11,%xmm11 + vpmuludq %xmm7,%xmm4,%xmm4 + vpaddq %xmm4,%xmm10,%xmm10 + + vpmuludq %xmm5,%xmm2,%xmm2 + vpaddq %xmm2,%xmm14,%xmm14 + vpmuludq %xmm9,%xmm3,%xmm1 + vpaddq %xmm1,%xmm13,%xmm13 + vpmuludq %xmm8,%xmm3,%xmm0 + vpaddq %xmm0,%xmm12,%xmm12 + vpmuludq %xmm7,%xmm3,%xmm1 + vpaddq %xmm1,%xmm11,%xmm11 + vpmuludq %xmm6,%xmm3,%xmm3 + vpaddq %xmm3,%xmm10,%xmm10 + + jz .Lshort_tail_avx + + vmovdqu 0(%rsi),%xmm0 + vmovdqu 16(%rsi),%xmm1 + + vpsrldq $6,%xmm0,%xmm2 + vpsrldq $6,%xmm1,%xmm3 + vpunpckhqdq %xmm1,%xmm0,%xmm4 + vpunpcklqdq %xmm1,%xmm0,%xmm0 + vpunpcklqdq %xmm3,%xmm2,%xmm3 + + vpsrlq $40,%xmm4,%xmm4 + vpsrlq $26,%xmm0,%xmm1 + vpand %xmm15,%xmm0,%xmm0 + vpsrlq $4,%xmm3,%xmm2 + vpand %xmm15,%xmm1,%xmm1 + vpsrlq $30,%xmm3,%xmm3 + vpand %xmm15,%xmm2,%xmm2 + vpand %xmm15,%xmm3,%xmm3 + vpor 32(%rcx),%xmm4,%xmm4 + + vpshufd $0x32,-64(%rdi),%xmm9 + vpaddq 0(%r11),%xmm0,%xmm0 + vpaddq 16(%r11),%xmm1,%xmm1 + vpaddq 32(%r11),%xmm2,%xmm2 + vpaddq 48(%r11),%xmm3,%xmm3 + vpaddq 64(%r11),%xmm4,%xmm4 + + vpmuludq %xmm0,%xmm9,%xmm5 + vpaddq %xmm5,%xmm10,%xmm10 + vpmuludq %xmm1,%xmm9,%xmm6 + vpaddq %xmm6,%xmm11,%xmm11 + vpmuludq %xmm2,%xmm9,%xmm5 + vpaddq %xmm5,%xmm12,%xmm12 + vpshufd $0x32,-48(%rdi),%xmm7 + vpmuludq %xmm3,%xmm9,%xmm6 + vpaddq %xmm6,%xmm13,%xmm13 + vpmuludq %xmm4,%xmm9,%xmm9 + vpaddq %xmm9,%xmm14,%xmm14 + + vpmuludq %xmm3,%xmm7,%xmm5 + vpaddq %xmm5,%xmm14,%xmm14 + vpshufd $0x32,-32(%rdi),%xmm8 + vpmuludq %xmm2,%xmm7,%xmm6 + vpaddq %xmm6,%xmm13,%xmm13 + vpshufd $0x32,-16(%rdi),%xmm9 + vpmuludq %xmm1,%xmm7,%xmm5 + vpaddq %xmm5,%xmm12,%xmm12 + vpmuludq %xmm0,%xmm7,%xmm7 + vpaddq %xmm7,%xmm11,%xmm11 + vpmuludq %xmm4,%xmm8,%xmm8 + vpaddq %xmm8,%xmm10,%xmm10 + + vpshufd $0x32,0(%rdi),%xmm7 + vpmuludq %xmm2,%xmm9,%xmm6 + vpaddq %xmm6,%xmm14,%xmm14 + vpmuludq %xmm1,%xmm9,%xmm5 + vpaddq %xmm5,%xmm13,%xmm13 + vpshufd $0x32,16(%rdi),%xmm8 + vpmuludq %xmm0,%xmm9,%xmm9 + vpaddq %xmm9,%xmm12,%xmm12 + vpmuludq %xmm4,%xmm7,%xmm6 + vpaddq %xmm6,%xmm11,%xmm11 + vpshufd $0x32,32(%rdi),%xmm9 + vpmuludq %xmm3,%xmm7,%xmm7 + vpaddq %xmm7,%xmm10,%xmm10 + + vpmuludq %xmm1,%xmm8,%xmm5 + vpaddq %xmm5,%xmm14,%xmm14 + vpmuludq %xmm0,%xmm8,%xmm8 + vpaddq %xmm8,%xmm13,%xmm13 + vpshufd $0x32,48(%rdi),%xmm7 + vpmuludq %xmm4,%xmm9,%xmm6 + vpaddq %xmm6,%xmm12,%xmm12 + vpshufd $0x32,64(%rdi),%xmm8 + vpmuludq %xmm3,%xmm9,%xmm5 + vpaddq %xmm5,%xmm11,%xmm11 + vpmuludq %xmm2,%xmm9,%xmm9 + vpaddq %xmm9,%xmm10,%xmm10 + + vpmuludq %xmm0,%xmm7,%xmm7 + vpaddq %xmm7,%xmm14,%xmm14 + vpmuludq %xmm4,%xmm8,%xmm6 + vpaddq %xmm6,%xmm13,%xmm13 + vpmuludq %xmm3,%xmm8,%xmm5 + vpaddq %xmm5,%xmm12,%xmm12 + vpmuludq %xmm2,%xmm8,%xmm6 + vpaddq %xmm6,%xmm11,%xmm11 + vpmuludq %xmm1,%xmm8,%xmm8 + vpaddq %xmm8,%xmm10,%xmm10 + +.Lshort_tail_avx: + + vpsrldq $8,%xmm14,%xmm9 + vpsrldq $8,%xmm13,%xmm8 + vpsrldq $8,%xmm11,%xmm6 + vpsrldq $8,%xmm10,%xmm5 + vpsrldq $8,%xmm12,%xmm7 + vpaddq %xmm8,%xmm13,%xmm13 + vpaddq %xmm9,%xmm14,%xmm14 + vpaddq %xmm5,%xmm10,%xmm10 + vpaddq %xmm6,%xmm11,%xmm11 + vpaddq %xmm7,%xmm12,%xmm12 + + vpsrlq $26,%xmm13,%xmm3 + vpand %xmm15,%xmm13,%xmm13 + vpaddq %xmm3,%xmm14,%xmm14 + + vpsrlq $26,%xmm10,%xmm0 + vpand %xmm15,%xmm10,%xmm10 + vpaddq %xmm0,%xmm11,%xmm11 + + vpsrlq $26,%xmm14,%xmm4 + vpand %xmm15,%xmm14,%xmm14 + + vpsrlq $26,%xmm11,%xmm1 + vpand %xmm15,%xmm11,%xmm11 + vpaddq %xmm1,%xmm12,%xmm12 + + vpaddq %xmm4,%xmm10,%xmm10 + vpsllq $2,%xmm4,%xmm4 + vpaddq %xmm4,%xmm10,%xmm10 + + vpsrlq $26,%xmm12,%xmm2 + vpand %xmm15,%xmm12,%xmm12 + vpaddq %xmm2,%xmm13,%xmm13 + + vpsrlq $26,%xmm10,%xmm0 + vpand %xmm15,%xmm10,%xmm10 + vpaddq %xmm0,%xmm11,%xmm11 + + vpsrlq $26,%xmm13,%xmm3 + vpand %xmm15,%xmm13,%xmm13 + vpaddq %xmm3,%xmm14,%xmm14 + + vmovd %xmm10,-112(%rdi) + vmovd %xmm11,-108(%rdi) + vmovd %xmm12,-104(%rdi) + vmovd %xmm13,-100(%rdi) + vmovd %xmm14,-96(%rdi) + leaq -8(%r10),%rsp + + vzeroupper + ret +ENDPROC(poly1305_blocks_avx) + +.align 32 +ENTRY(poly1305_emit_avx) + cmpl $0,20(%rdi) + je .Lemit + + movl 0(%rdi),%eax + movl 4(%rdi),%ecx + movl 8(%rdi),%r8d + movl 12(%rdi),%r11d + movl 16(%rdi),%r10d + + shlq $26,%rcx + movq %r8,%r9 + shlq $52,%r8 + addq %rcx,%rax + shrq $12,%r9 + addq %rax,%r8 + adcq $0,%r9 + + shlq $14,%r11 + movq %r10,%rax + shrq $24,%r10 + addq %r11,%r9 + shlq $40,%rax + addq %rax,%r9 + adcq $0,%r10 + + movq %r10,%rax + movq %r10,%rcx + andq $3,%r10 + shrq $2,%rax + andq $-4,%rcx + addq %rcx,%rax + addq %rax,%r8 + adcq $0,%r9 + adcq $0,%r10 + + movq %r8,%rax + addq $5,%r8 + movq %r9,%rcx + adcq $0,%r9 + adcq $0,%r10 + shrq $2,%r10 + cmovnzq %r8,%rax + cmovnzq %r9,%rcx + + addq 0(%rdx),%rax + adcq 8(%rdx),%rcx + movq %rax,0(%rsi) + movq %rcx,8(%rsi) + + ret +ENDPROC(poly1305_emit_avx) +#endif /* CONFIG_AS_AVX */ + +#ifdef CONFIG_AS_AVX2 +.align 32 +ENTRY(poly1305_blocks_avx2) + + movl 20(%rdi),%r8d + cmpq $128,%rdx + jae .Lblocks_avx2 + testl %r8d,%r8d + jz .Lblocks + +.Lblocks_avx2: + andq $-16,%rdx + jz .Lno_data_avx2 + + vzeroupper + + testl %r8d,%r8d + jz .Lbase2_64_avx2 + + testq $63,%rdx + jz .Leven_avx2 + + pushq %rbx + pushq %r12 + pushq %r13 + pushq %r14 + pushq %r15 + pushq %rdi + +.Lblocks_avx2_body: + + movq %rdx,%r15 + + movq 0(%rdi),%r8 + movq 8(%rdi),%r9 + movl 16(%rdi),%r10d + + movq 24(%rdi),%r11 + movq 32(%rdi),%r13 + + + movl %r8d,%r14d + andq $-2147483648,%r8 + movq %r9,%r12 + movl %r9d,%ebx + andq $-2147483648,%r9 + + shrq $6,%r8 + shlq $52,%r12 + addq %r8,%r14 + shrq $12,%rbx + shrq $18,%r9 + addq %r12,%r14 + adcq %r9,%rbx + + movq %r10,%r8 + shlq $40,%r8 + shrq $24,%r10 + addq %r8,%rbx + adcq $0,%r10 + + movq $-4,%r9 + movq %r10,%r8 + andq %r10,%r9 + shrq $2,%r8 + andq $3,%r10 + addq %r9,%r8 + addq %r8,%r14 + adcq $0,%rbx + adcq $0,%r10 + + movq %r13,%r12 + movq %r13,%rax + shrq $2,%r13 + addq %r12,%r13 + +.Lbase2_26_pre_avx2: + addq 0(%rsi),%r14 + adcq 8(%rsi),%rbx + leaq 16(%rsi),%rsi + adcq %rcx,%r10 + subq $16,%r15 + + movq %rdi,0(%rsp) + __poly1305_block + movq 0(%rsp),%rdi + movq %r12,%rax + + testq $63,%r15 + jnz .Lbase2_26_pre_avx2 + + testq %rcx,%rcx + jz .Lstore_base2_64_avx2 + + + movq %r14,%rax + movq %r14,%rdx + shrq $52,%r14 + movq %rbx,%r11 + movq %rbx,%r12 + shrq $26,%rdx + andq $0x3ffffff,%rax + shlq $12,%r11 + andq $0x3ffffff,%rdx + shrq $14,%rbx + orq %r11,%r14 + shlq $24,%r10 + andq $0x3ffffff,%r14 + shrq $40,%r12 + andq $0x3ffffff,%rbx + orq %r12,%r10 + + testq %r15,%r15 + jz .Lstore_base2_26_avx2 + + vmovd %eax,%xmm0 + vmovd %edx,%xmm1 + vmovd %r14d,%xmm2 + vmovd %ebx,%xmm3 + vmovd %r10d,%xmm4 + jmp .Lproceed_avx2 + +.align 32 +.Lstore_base2_64_avx2: + movq %r14,0(%rdi) + movq %rbx,8(%rdi) + movq %r10,16(%rdi) + jmp .Ldone_avx2 + +.align 16 +.Lstore_base2_26_avx2: + movl %eax,0(%rdi) + movl %edx,4(%rdi) + movl %r14d,8(%rdi) + movl %ebx,12(%rdi) + movl %r10d,16(%rdi) +.align 16 +.Ldone_avx2: + movq 8(%rsp),%r15 + movq 16(%rsp),%r14 + movq 24(%rsp),%r13 + movq 32(%rsp),%r12 + movq 40(%rsp),%rbx + leaq 48(%rsp),%rsp + +.Lno_data_avx2: +.Lblocks_avx2_epilogue: + ret + + +.align 32 +.Lbase2_64_avx2: + + + pushq %rbx + pushq %r12 + pushq %r13 + pushq %r14 + pushq %r15 + pushq %rdi + +.Lbase2_64_avx2_body: + + movq %rdx,%r15 + + movq 24(%rdi),%r11 + movq 32(%rdi),%r13 + + movq 0(%rdi),%r14 + movq 8(%rdi),%rbx + movl 16(%rdi),%r10d + + movq %r13,%r12 + movq %r13,%rax + shrq $2,%r13 + addq %r12,%r13 + + testq $63,%rdx + jz .Linit_avx2 + +.Lbase2_64_pre_avx2: + addq 0(%rsi),%r14 + adcq 8(%rsi),%rbx + leaq 16(%rsi),%rsi + adcq %rcx,%r10 + subq $16,%r15 + + movq %rdi,0(%rsp) + __poly1305_block + movq 0(%rsp),%rdi + movq %r12,%rax + + testq $63,%r15 + jnz .Lbase2_64_pre_avx2 + +.Linit_avx2: + + movq %r14,%rax + movq %r14,%rdx + shrq $52,%r14 + movq %rbx,%r8 + movq %rbx,%r9 + shrq $26,%rdx + andq $0x3ffffff,%rax + shlq $12,%r8 + andq $0x3ffffff,%rdx + shrq $14,%rbx + orq %r8,%r14 + shlq $24,%r10 + andq $0x3ffffff,%r14 + shrq $40,%r9 + andq $0x3ffffff,%rbx + orq %r9,%r10 + + vmovd %eax,%xmm0 + vmovd %edx,%xmm1 + vmovd %r14d,%xmm2 + vmovd %ebx,%xmm3 + vmovd %r10d,%xmm4 + movl $1,20(%rdi) + + __poly1305_init_avx + +.Lproceed_avx2: + movq %r15,%rdx + + movq 8(%rsp),%r15 + movq 16(%rsp),%r14 + movq 24(%rsp),%r13 + movq 32(%rsp),%r12 + movq 40(%rsp),%rbx + leaq 48(%rsp),%rax + leaq 48(%rsp),%rsp + +.Lbase2_64_avx2_epilogue: + jmp .Ldo_avx2 + + +.align 32 +.Leven_avx2: + + vmovd 0(%rdi),%xmm0 + vmovd 4(%rdi),%xmm1 + vmovd 8(%rdi),%xmm2 + vmovd 12(%rdi),%xmm3 + vmovd 16(%rdi),%xmm4 + +.Ldo_avx2: + leaq 8(%rsp),%r10 + subq $0x128,%rsp + leaq .Lconst(%rip),%rcx + leaq 48+64(%rdi),%rdi + vmovdqa 96(%rcx),%ymm7 + + + vmovdqu -64(%rdi),%xmm9 + andq $-512,%rsp + vmovdqu -48(%rdi),%xmm10 + vmovdqu -32(%rdi),%xmm6 + vmovdqu -16(%rdi),%xmm11 + vmovdqu 0(%rdi),%xmm12 + vmovdqu 16(%rdi),%xmm13 + leaq 144(%rsp),%rax + vmovdqu 32(%rdi),%xmm14 + vpermd %ymm9,%ymm7,%ymm9 + vmovdqu 48(%rdi),%xmm15 + vpermd %ymm10,%ymm7,%ymm10 + vmovdqu 64(%rdi),%xmm5 + vpermd %ymm6,%ymm7,%ymm6 + vmovdqa %ymm9,0(%rsp) + vpermd %ymm11,%ymm7,%ymm11 + vmovdqa %ymm10,32-144(%rax) + vpermd %ymm12,%ymm7,%ymm12 + vmovdqa %ymm6,64-144(%rax) + vpermd %ymm13,%ymm7,%ymm13 + vmovdqa %ymm11,96-144(%rax) + vpermd %ymm14,%ymm7,%ymm14 + vmovdqa %ymm12,128-144(%rax) + vpermd %ymm15,%ymm7,%ymm15 + vmovdqa %ymm13,160-144(%rax) + vpermd %ymm5,%ymm7,%ymm5 + vmovdqa %ymm14,192-144(%rax) + vmovdqa %ymm15,224-144(%rax) + vmovdqa %ymm5,256-144(%rax) + vmovdqa 64(%rcx),%ymm5 + + + + vmovdqu 0(%rsi),%xmm7 + vmovdqu 16(%rsi),%xmm8 + vinserti128 $1,32(%rsi),%ymm7,%ymm7 + vinserti128 $1,48(%rsi),%ymm8,%ymm8 + leaq 64(%rsi),%rsi + + vpsrldq $6,%ymm7,%ymm9 + vpsrldq $6,%ymm8,%ymm10 + vpunpckhqdq %ymm8,%ymm7,%ymm6 + vpunpcklqdq %ymm10,%ymm9,%ymm9 + vpunpcklqdq %ymm8,%ymm7,%ymm7 + + vpsrlq $30,%ymm9,%ymm10 + vpsrlq $4,%ymm9,%ymm9 + vpsrlq $26,%ymm7,%ymm8 + vpsrlq $40,%ymm6,%ymm6 + vpand %ymm5,%ymm9,%ymm9 + vpand %ymm5,%ymm7,%ymm7 + vpand %ymm5,%ymm8,%ymm8 + vpand %ymm5,%ymm10,%ymm10 + vpor 32(%rcx),%ymm6,%ymm6 + + vpaddq %ymm2,%ymm9,%ymm2 + subq $64,%rdx + jz .Ltail_avx2 + jmp .Loop_avx2 + +.align 32 +.Loop_avx2: + + vpaddq %ymm0,%ymm7,%ymm0 + vmovdqa 0(%rsp),%ymm7 + vpaddq %ymm1,%ymm8,%ymm1 + vmovdqa 32(%rsp),%ymm8 + vpaddq %ymm3,%ymm10,%ymm3 + vmovdqa 96(%rsp),%ymm9 + vpaddq %ymm4,%ymm6,%ymm4 + vmovdqa 48(%rax),%ymm10 + vmovdqa 112(%rax),%ymm5 + + vpmuludq %ymm2,%ymm7,%ymm13 + vpmuludq %ymm2,%ymm8,%ymm14 + vpmuludq %ymm2,%ymm9,%ymm15 + vpmuludq %ymm2,%ymm10,%ymm11 + vpmuludq %ymm2,%ymm5,%ymm12 + + vpmuludq %ymm0,%ymm8,%ymm6 + vpmuludq %ymm1,%ymm8,%ymm2 + vpaddq %ymm6,%ymm12,%ymm12 + vpaddq %ymm2,%ymm13,%ymm13 + vpmuludq %ymm3,%ymm8,%ymm6 + vpmuludq 64(%rsp),%ymm4,%ymm2 + vpaddq %ymm6,%ymm15,%ymm15 + vpaddq %ymm2,%ymm11,%ymm11 + vmovdqa -16(%rax),%ymm8 + + vpmuludq %ymm0,%ymm7,%ymm6 + vpmuludq %ymm1,%ymm7,%ymm2 + vpaddq %ymm6,%ymm11,%ymm11 + vpaddq %ymm2,%ymm12,%ymm12 + vpmuludq %ymm3,%ymm7,%ymm6 + vpmuludq %ymm4,%ymm7,%ymm2 + vmovdqu 0(%rsi),%xmm7 + vpaddq %ymm6,%ymm14,%ymm14 + vpaddq %ymm2,%ymm15,%ymm15 + vinserti128 $1,32(%rsi),%ymm7,%ymm7 + + vpmuludq %ymm3,%ymm8,%ymm6 + vpmuludq %ymm4,%ymm8,%ymm2 + vmovdqu 16(%rsi),%xmm8 + vpaddq %ymm6,%ymm11,%ymm11 + vpaddq %ymm2,%ymm12,%ymm12 + vmovdqa 16(%rax),%ymm2 + vpmuludq %ymm1,%ymm9,%ymm6 + vpmuludq %ymm0,%ymm9,%ymm9 + vpaddq %ymm6,%ymm14,%ymm14 + vpaddq %ymm9,%ymm13,%ymm13 + vinserti128 $1,48(%rsi),%ymm8,%ymm8 + leaq 64(%rsi),%rsi + + vpmuludq %ymm1,%ymm2,%ymm6 + vpmuludq %ymm0,%ymm2,%ymm2 + vpsrldq $6,%ymm7,%ymm9 + vpaddq %ymm6,%ymm15,%ymm15 + vpaddq %ymm2,%ymm14,%ymm14 + vpmuludq %ymm3,%ymm10,%ymm6 + vpmuludq %ymm4,%ymm10,%ymm2 + vpsrldq $6,%ymm8,%ymm10 + vpaddq %ymm6,%ymm12,%ymm12 + vpaddq %ymm2,%ymm13,%ymm13 + vpunpckhqdq %ymm8,%ymm7,%ymm6 + + vpmuludq %ymm3,%ymm5,%ymm3 + vpmuludq %ymm4,%ymm5,%ymm4 + vpunpcklqdq %ymm8,%ymm7,%ymm7 + vpaddq %ymm3,%ymm13,%ymm2 + vpaddq %ymm4,%ymm14,%ymm3 + vpunpcklqdq %ymm10,%ymm9,%ymm10 + vpmuludq 80(%rax),%ymm0,%ymm4 + vpmuludq %ymm1,%ymm5,%ymm0 + vmovdqa 64(%rcx),%ymm5 + vpaddq %ymm4,%ymm15,%ymm4 + vpaddq %ymm0,%ymm11,%ymm0 + + vpsrlq $26,%ymm3,%ymm14 + vpand %ymm5,%ymm3,%ymm3 + vpaddq %ymm14,%ymm4,%ymm4 + + vpsrlq $26,%ymm0,%ymm11 + vpand %ymm5,%ymm0,%ymm0 + vpaddq %ymm11,%ymm12,%ymm1 + + vpsrlq $26,%ymm4,%ymm15 + vpand %ymm5,%ymm4,%ymm4 + + vpsrlq $4,%ymm10,%ymm9 + + vpsrlq $26,%ymm1,%ymm12 + vpand %ymm5,%ymm1,%ymm1 + vpaddq %ymm12,%ymm2,%ymm2 + + vpaddq %ymm15,%ymm0,%ymm0 + vpsllq $2,%ymm15,%ymm15 + vpaddq %ymm15,%ymm0,%ymm0 + + vpand %ymm5,%ymm9,%ymm9 + vpsrlq $26,%ymm7,%ymm8 + + vpsrlq $26,%ymm2,%ymm13 + vpand %ymm5,%ymm2,%ymm2 + vpaddq %ymm13,%ymm3,%ymm3 + + vpaddq %ymm9,%ymm2,%ymm2 + vpsrlq $30,%ymm10,%ymm10 + + vpsrlq $26,%ymm0,%ymm11 + vpand %ymm5,%ymm0,%ymm0 + vpaddq %ymm11,%ymm1,%ymm1 + + vpsrlq $40,%ymm6,%ymm6 + + vpsrlq $26,%ymm3,%ymm14 + vpand %ymm5,%ymm3,%ymm3 + vpaddq %ymm14,%ymm4,%ymm4 + + vpand %ymm5,%ymm7,%ymm7 + vpand %ymm5,%ymm8,%ymm8 + vpand %ymm5,%ymm10,%ymm10 + vpor 32(%rcx),%ymm6,%ymm6 + + subq $64,%rdx + jnz .Loop_avx2 + +.byte 0x66,0x90 +.Ltail_avx2: + + vpaddq %ymm0,%ymm7,%ymm0 + vmovdqu 4(%rsp),%ymm7 + vpaddq %ymm1,%ymm8,%ymm1 + vmovdqu 36(%rsp),%ymm8 + vpaddq %ymm3,%ymm10,%ymm3 + vmovdqu 100(%rsp),%ymm9 + vpaddq %ymm4,%ymm6,%ymm4 + vmovdqu 52(%rax),%ymm10 + vmovdqu 116(%rax),%ymm5 + + vpmuludq %ymm2,%ymm7,%ymm13 + vpmuludq %ymm2,%ymm8,%ymm14 + vpmuludq %ymm2,%ymm9,%ymm15 + vpmuludq %ymm2,%ymm10,%ymm11 + vpmuludq %ymm2,%ymm5,%ymm12 + + vpmuludq %ymm0,%ymm8,%ymm6 + vpmuludq %ymm1,%ymm8,%ymm2 + vpaddq %ymm6,%ymm12,%ymm12 + vpaddq %ymm2,%ymm13,%ymm13 + vpmuludq %ymm3,%ymm8,%ymm6 + vpmuludq 68(%rsp),%ymm4,%ymm2 + vpaddq %ymm6,%ymm15,%ymm15 + vpaddq %ymm2,%ymm11,%ymm11 + + vpmuludq %ymm0,%ymm7,%ymm6 + vpmuludq %ymm1,%ymm7,%ymm2 + vpaddq %ymm6,%ymm11,%ymm11 + vmovdqu -12(%rax),%ymm8 + vpaddq %ymm2,%ymm12,%ymm12 + vpmuludq %ymm3,%ymm7,%ymm6 + vpmuludq %ymm4,%ymm7,%ymm2 + vpaddq %ymm6,%ymm14,%ymm14 + vpaddq %ymm2,%ymm15,%ymm15 + + vpmuludq %ymm3,%ymm8,%ymm6 + vpmuludq %ymm4,%ymm8,%ymm2 + vpaddq %ymm6,%ymm11,%ymm11 + vpaddq %ymm2,%ymm12,%ymm12 + vmovdqu 20(%rax),%ymm2 + vpmuludq %ymm1,%ymm9,%ymm6 + vpmuludq %ymm0,%ymm9,%ymm9 + vpaddq %ymm6,%ymm14,%ymm14 + vpaddq %ymm9,%ymm13,%ymm13 + + vpmuludq %ymm1,%ymm2,%ymm6 + vpmuludq %ymm0,%ymm2,%ymm2 + vpaddq %ymm6,%ymm15,%ymm15 + vpaddq %ymm2,%ymm14,%ymm14 + vpmuludq %ymm3,%ymm10,%ymm6 + vpmuludq %ymm4,%ymm10,%ymm2 + vpaddq %ymm6,%ymm12,%ymm12 + vpaddq %ymm2,%ymm13,%ymm13 + + vpmuludq %ymm3,%ymm5,%ymm3 + vpmuludq %ymm4,%ymm5,%ymm4 + vpaddq %ymm3,%ymm13,%ymm2 + vpaddq %ymm4,%ymm14,%ymm3 + vpmuludq 84(%rax),%ymm0,%ymm4 + vpmuludq %ymm1,%ymm5,%ymm0 + vmovdqa 64(%rcx),%ymm5 + vpaddq %ymm4,%ymm15,%ymm4 + vpaddq %ymm0,%ymm11,%ymm0 + + vpsrldq $8,%ymm12,%ymm8 + vpsrldq $8,%ymm2,%ymm9 + vpsrldq $8,%ymm3,%ymm10 + vpsrldq $8,%ymm4,%ymm6 + vpsrldq $8,%ymm0,%ymm7 + vpaddq %ymm8,%ymm12,%ymm12 + vpaddq %ymm9,%ymm2,%ymm2 + vpaddq %ymm10,%ymm3,%ymm3 + vpaddq %ymm6,%ymm4,%ymm4 + vpaddq %ymm7,%ymm0,%ymm0 + + vpermq $0x2,%ymm3,%ymm10 + vpermq $0x2,%ymm4,%ymm6 + vpermq $0x2,%ymm0,%ymm7 + vpermq $0x2,%ymm12,%ymm8 + vpermq $0x2,%ymm2,%ymm9 + vpaddq %ymm10,%ymm3,%ymm3 + vpaddq %ymm6,%ymm4,%ymm4 + vpaddq %ymm7,%ymm0,%ymm0 + vpaddq %ymm8,%ymm12,%ymm12 + vpaddq %ymm9,%ymm2,%ymm2 + + vpsrlq $26,%ymm3,%ymm14 + vpand %ymm5,%ymm3,%ymm3 + vpaddq %ymm14,%ymm4,%ymm4 + + vpsrlq $26,%ymm0,%ymm11 + vpand %ymm5,%ymm0,%ymm0 + vpaddq %ymm11,%ymm12,%ymm1 + + vpsrlq $26,%ymm4,%ymm15 + vpand %ymm5,%ymm4,%ymm4 + + vpsrlq $26,%ymm1,%ymm12 + vpand %ymm5,%ymm1,%ymm1 + vpaddq %ymm12,%ymm2,%ymm2 + + vpaddq %ymm15,%ymm0,%ymm0 + vpsllq $2,%ymm15,%ymm15 + vpaddq %ymm15,%ymm0,%ymm0 + + vpsrlq $26,%ymm2,%ymm13 + vpand %ymm5,%ymm2,%ymm2 + vpaddq %ymm13,%ymm3,%ymm3 + + vpsrlq $26,%ymm0,%ymm11 + vpand %ymm5,%ymm0,%ymm0 + vpaddq %ymm11,%ymm1,%ymm1 + + vpsrlq $26,%ymm3,%ymm14 + vpand %ymm5,%ymm3,%ymm3 + vpaddq %ymm14,%ymm4,%ymm4 + + vmovd %xmm0,-112(%rdi) + vmovd %xmm1,-108(%rdi) + vmovd %xmm2,-104(%rdi) + vmovd %xmm3,-100(%rdi) + vmovd %xmm4,-96(%rdi) + leaq -8(%r10),%rsp + + vzeroupper + ret + +ENDPROC(poly1305_blocks_avx2) +#endif /* CONFIG_AS_AVX2 */ + +#ifdef CONFIG_AS_AVX512 +.align 32 +ENTRY(poly1305_blocks_avx512) + + movl 20(%rdi),%r8d + cmpq $128,%rdx + jae .Lblocks_avx2_512 + testl %r8d,%r8d + jz .Lblocks + +.Lblocks_avx2_512: + andq $-16,%rdx + jz .Lno_data_avx2_512 + + vzeroupper + + testl %r8d,%r8d + jz .Lbase2_64_avx2_512 + + testq $63,%rdx + jz .Leven_avx2_512 + + pushq %rbx + pushq %r12 + pushq %r13 + pushq %r14 + pushq %r15 + pushq %rdi + +.Lblocks_avx2_body_512: + + movq %rdx,%r15 + + movq 0(%rdi),%r8 + movq 8(%rdi),%r9 + movl 16(%rdi),%r10d + + movq 24(%rdi),%r11 + movq 32(%rdi),%r13 + + + movl %r8d,%r14d + andq $-2147483648,%r8 + movq %r9,%r12 + movl %r9d,%ebx + andq $-2147483648,%r9 + + shrq $6,%r8 + shlq $52,%r12 + addq %r8,%r14 + shrq $12,%rbx + shrq $18,%r9 + addq %r12,%r14 + adcq %r9,%rbx + + movq %r10,%r8 + shlq $40,%r8 + shrq $24,%r10 + addq %r8,%rbx + adcq $0,%r10 + + movq $-4,%r9 + movq %r10,%r8 + andq %r10,%r9 + shrq $2,%r8 + andq $3,%r10 + addq %r9,%r8 + addq %r8,%r14 + adcq $0,%rbx + adcq $0,%r10 + + movq %r13,%r12 + movq %r13,%rax + shrq $2,%r13 + addq %r12,%r13 + +.Lbase2_26_pre_avx2_512: + addq 0(%rsi),%r14 + adcq 8(%rsi),%rbx + leaq 16(%rsi),%rsi + adcq %rcx,%r10 + subq $16,%r15 + + movq %rdi,0(%rsp) + __poly1305_block + movq 0(%rsp),%rdi + movq %r12,%rax + + testq $63,%r15 + jnz .Lbase2_26_pre_avx2_512 + + testq %rcx,%rcx + jz .Lstore_base2_64_avx2_512 + + + movq %r14,%rax + movq %r14,%rdx + shrq $52,%r14 + movq %rbx,%r11 + movq %rbx,%r12 + shrq $26,%rdx + andq $0x3ffffff,%rax + shlq $12,%r11 + andq $0x3ffffff,%rdx + shrq $14,%rbx + orq %r11,%r14 + shlq $24,%r10 + andq $0x3ffffff,%r14 + shrq $40,%r12 + andq $0x3ffffff,%rbx + orq %r12,%r10 + + testq %r15,%r15 + jz .Lstore_base2_26_avx2_512 + + vmovd %eax,%xmm0 + vmovd %edx,%xmm1 + vmovd %r14d,%xmm2 + vmovd %ebx,%xmm3 + vmovd %r10d,%xmm4 + jmp .Lproceed_avx2_512 + +.align 32 +.Lstore_base2_64_avx2_512: + movq %r14,0(%rdi) + movq %rbx,8(%rdi) + movq %r10,16(%rdi) + jmp .Ldone_avx2_512 + +.align 16 +.Lstore_base2_26_avx2_512: + movl %eax,0(%rdi) + movl %edx,4(%rdi) + movl %r14d,8(%rdi) + movl %ebx,12(%rdi) + movl %r10d,16(%rdi) +.align 16 +.Ldone_avx2_512: + movq 8(%rsp),%r15 + movq 16(%rsp),%r14 + movq 24(%rsp),%r13 + movq 32(%rsp),%r12 + movq 40(%rsp),%rbx + leaq 48(%rsp),%rsp + +.Lno_data_avx2_512: +.Lblocks_avx2_epilogue_512: + ret + + +.align 32 +.Lbase2_64_avx2_512: + + pushq %rbx + pushq %r12 + pushq %r13 + pushq %r14 + pushq %r15 + pushq %rdi + +.Lbase2_64_avx2_body_512: + + movq %rdx,%r15 + + movq 24(%rdi),%r11 + movq 32(%rdi),%r13 + + movq 0(%rdi),%r14 + movq 8(%rdi),%rbx + movl 16(%rdi),%r10d + + movq %r13,%r12 + movq %r13,%rax + shrq $2,%r13 + addq %r12,%r13 + + testq $63,%rdx + jz .Linit_avx2_512 + +.Lbase2_64_pre_avx2_512: + addq 0(%rsi),%r14 + adcq 8(%rsi),%rbx + leaq 16(%rsi),%rsi + adcq %rcx,%r10 + subq $16,%r15 + + movq %rdi,0(%rsp) + __poly1305_block + movq 0(%rsp),%rdi + movq %r12,%rax + + testq $63,%r15 + jnz .Lbase2_64_pre_avx2_512 + +.Linit_avx2_512: + + movq %r14,%rax + movq %r14,%rdx + shrq $52,%r14 + movq %rbx,%r8 + movq %rbx,%r9 + shrq $26,%rdx + andq $0x3ffffff,%rax + shlq $12,%r8 + andq $0x3ffffff,%rdx + shrq $14,%rbx + orq %r8,%r14 + shlq $24,%r10 + andq $0x3ffffff,%r14 + shrq $40,%r9 + andq $0x3ffffff,%rbx + orq %r9,%r10 + + vmovd %eax,%xmm0 + vmovd %edx,%xmm1 + vmovd %r14d,%xmm2 + vmovd %ebx,%xmm3 + vmovd %r10d,%xmm4 + movl $1,20(%rdi) + + __poly1305_init_avx + +.Lproceed_avx2_512: + movq %r15,%rdx + + movq 8(%rsp),%r15 + movq 16(%rsp),%r14 + movq 24(%rsp),%r13 + movq 32(%rsp),%r12 + movq 40(%rsp),%rbx + leaq 48(%rsp),%rax + leaq 48(%rsp),%rsp + +.Lbase2_64_avx2_epilogue_512: + jmp .Ldo_avx2_512 + + +.align 32 +.Leven_avx2_512: + + vmovd 0(%rdi),%xmm0 + vmovd 4(%rdi),%xmm1 + vmovd 8(%rdi),%xmm2 + vmovd 12(%rdi),%xmm3 + vmovd 16(%rdi),%xmm4 + +.Ldo_avx2_512: + cmpq $512,%rdx + jae .Lblocks_avx512 +.Lskip_avx512: + leaq 8(%rsp),%r10 + + subq $0x128,%rsp + leaq .Lconst(%rip),%rcx + leaq 48+64(%rdi),%rdi + vmovdqa 96(%rcx),%ymm7 + + + vmovdqu -64(%rdi),%xmm9 + andq $-512,%rsp + vmovdqu -48(%rdi),%xmm10 + vmovdqu -32(%rdi),%xmm6 + vmovdqu -16(%rdi),%xmm11 + vmovdqu 0(%rdi),%xmm12 + vmovdqu 16(%rdi),%xmm13 + leaq 144(%rsp),%rax + vmovdqu 32(%rdi),%xmm14 + vpermd %ymm9,%ymm7,%ymm9 + vmovdqu 48(%rdi),%xmm15 + vpermd %ymm10,%ymm7,%ymm10 + vmovdqu 64(%rdi),%xmm5 + vpermd %ymm6,%ymm7,%ymm6 + vmovdqa %ymm9,0(%rsp) + vpermd %ymm11,%ymm7,%ymm11 + vmovdqa %ymm10,32-144(%rax) + vpermd %ymm12,%ymm7,%ymm12 + vmovdqa %ymm6,64-144(%rax) + vpermd %ymm13,%ymm7,%ymm13 + vmovdqa %ymm11,96-144(%rax) + vpermd %ymm14,%ymm7,%ymm14 + vmovdqa %ymm12,128-144(%rax) + vpermd %ymm15,%ymm7,%ymm15 + vmovdqa %ymm13,160-144(%rax) + vpermd %ymm5,%ymm7,%ymm5 + vmovdqa %ymm14,192-144(%rax) + vmovdqa %ymm15,224-144(%rax) + vmovdqa %ymm5,256-144(%rax) + vmovdqa 64(%rcx),%ymm5 + + + + vmovdqu 0(%rsi),%xmm7 + vmovdqu 16(%rsi),%xmm8 + vinserti128 $1,32(%rsi),%ymm7,%ymm7 + vinserti128 $1,48(%rsi),%ymm8,%ymm8 + leaq 64(%rsi),%rsi + + vpsrldq $6,%ymm7,%ymm9 + vpsrldq $6,%ymm8,%ymm10 + vpunpckhqdq %ymm8,%ymm7,%ymm6 + vpunpcklqdq %ymm10,%ymm9,%ymm9 + vpunpcklqdq %ymm8,%ymm7,%ymm7 + + vpsrlq $30,%ymm9,%ymm10 + vpsrlq $4,%ymm9,%ymm9 + vpsrlq $26,%ymm7,%ymm8 + vpsrlq $40,%ymm6,%ymm6 + vpand %ymm5,%ymm9,%ymm9 + vpand %ymm5,%ymm7,%ymm7 + vpand %ymm5,%ymm8,%ymm8 + vpand %ymm5,%ymm10,%ymm10 + vpor 32(%rcx),%ymm6,%ymm6 + + vpaddq %ymm2,%ymm9,%ymm2 + subq $64,%rdx + jz .Ltail_avx2_512 + jmp .Loop_avx2_512 + +.align 32 +.Loop_avx2_512: + + vpaddq %ymm0,%ymm7,%ymm0 + vmovdqa 0(%rsp),%ymm7 + vpaddq %ymm1,%ymm8,%ymm1 + vmovdqa 32(%rsp),%ymm8 + vpaddq %ymm3,%ymm10,%ymm3 + vmovdqa 96(%rsp),%ymm9 + vpaddq %ymm4,%ymm6,%ymm4 + vmovdqa 48(%rax),%ymm10 + vmovdqa 112(%rax),%ymm5 + + vpmuludq %ymm2,%ymm7,%ymm13 + vpmuludq %ymm2,%ymm8,%ymm14 + vpmuludq %ymm2,%ymm9,%ymm15 + vpmuludq %ymm2,%ymm10,%ymm11 + vpmuludq %ymm2,%ymm5,%ymm12 + + vpmuludq %ymm0,%ymm8,%ymm6 + vpmuludq %ymm1,%ymm8,%ymm2 + vpaddq %ymm6,%ymm12,%ymm12 + vpaddq %ymm2,%ymm13,%ymm13 + vpmuludq %ymm3,%ymm8,%ymm6 + vpmuludq 64(%rsp),%ymm4,%ymm2 + vpaddq %ymm6,%ymm15,%ymm15 + vpaddq %ymm2,%ymm11,%ymm11 + vmovdqa -16(%rax),%ymm8 + + vpmuludq %ymm0,%ymm7,%ymm6 + vpmuludq %ymm1,%ymm7,%ymm2 + vpaddq %ymm6,%ymm11,%ymm11 + vpaddq %ymm2,%ymm12,%ymm12 + vpmuludq %ymm3,%ymm7,%ymm6 + vpmuludq %ymm4,%ymm7,%ymm2 + vmovdqu 0(%rsi),%xmm7 + vpaddq %ymm6,%ymm14,%ymm14 + vpaddq %ymm2,%ymm15,%ymm15 + vinserti128 $1,32(%rsi),%ymm7,%ymm7 + + vpmuludq %ymm3,%ymm8,%ymm6 + vpmuludq %ymm4,%ymm8,%ymm2 + vmovdqu 16(%rsi),%xmm8 + vpaddq %ymm6,%ymm11,%ymm11 + vpaddq %ymm2,%ymm12,%ymm12 + vmovdqa 16(%rax),%ymm2 + vpmuludq %ymm1,%ymm9,%ymm6 + vpmuludq %ymm0,%ymm9,%ymm9 + vpaddq %ymm6,%ymm14,%ymm14 + vpaddq %ymm9,%ymm13,%ymm13 + vinserti128 $1,48(%rsi),%ymm8,%ymm8 + leaq 64(%rsi),%rsi + + vpmuludq %ymm1,%ymm2,%ymm6 + vpmuludq %ymm0,%ymm2,%ymm2 + vpsrldq $6,%ymm7,%ymm9 + vpaddq %ymm6,%ymm15,%ymm15 + vpaddq %ymm2,%ymm14,%ymm14 + vpmuludq %ymm3,%ymm10,%ymm6 + vpmuludq %ymm4,%ymm10,%ymm2 + vpsrldq $6,%ymm8,%ymm10 + vpaddq %ymm6,%ymm12,%ymm12 + vpaddq %ymm2,%ymm13,%ymm13 + vpunpckhqdq %ymm8,%ymm7,%ymm6 + + vpmuludq %ymm3,%ymm5,%ymm3 + vpmuludq %ymm4,%ymm5,%ymm4 + vpunpcklqdq %ymm8,%ymm7,%ymm7 + vpaddq %ymm3,%ymm13,%ymm2 + vpaddq %ymm4,%ymm14,%ymm3 + vpunpcklqdq %ymm10,%ymm9,%ymm10 + vpmuludq 80(%rax),%ymm0,%ymm4 + vpmuludq %ymm1,%ymm5,%ymm0 + vmovdqa 64(%rcx),%ymm5 + vpaddq %ymm4,%ymm15,%ymm4 + vpaddq %ymm0,%ymm11,%ymm0 + + vpsrlq $26,%ymm3,%ymm14 + vpand %ymm5,%ymm3,%ymm3 + vpaddq %ymm14,%ymm4,%ymm4 + + vpsrlq $26,%ymm0,%ymm11 + vpand %ymm5,%ymm0,%ymm0 + vpaddq %ymm11,%ymm12,%ymm1 + + vpsrlq $26,%ymm4,%ymm15 + vpand %ymm5,%ymm4,%ymm4 + + vpsrlq $4,%ymm10,%ymm9 + + vpsrlq $26,%ymm1,%ymm12 + vpand %ymm5,%ymm1,%ymm1 + vpaddq %ymm12,%ymm2,%ymm2 + + vpaddq %ymm15,%ymm0,%ymm0 + vpsllq $2,%ymm15,%ymm15 + vpaddq %ymm15,%ymm0,%ymm0 + + vpand %ymm5,%ymm9,%ymm9 + vpsrlq $26,%ymm7,%ymm8 + + vpsrlq $26,%ymm2,%ymm13 + vpand %ymm5,%ymm2,%ymm2 + vpaddq %ymm13,%ymm3,%ymm3 + + vpaddq %ymm9,%ymm2,%ymm2 + vpsrlq $30,%ymm10,%ymm10 + + vpsrlq $26,%ymm0,%ymm11 + vpand %ymm5,%ymm0,%ymm0 + vpaddq %ymm11,%ymm1,%ymm1 + + vpsrlq $40,%ymm6,%ymm6 + + vpsrlq $26,%ymm3,%ymm14 + vpand %ymm5,%ymm3,%ymm3 + vpaddq %ymm14,%ymm4,%ymm4 + + vpand %ymm5,%ymm7,%ymm7 + vpand %ymm5,%ymm8,%ymm8 + vpand %ymm5,%ymm10,%ymm10 + vpor 32(%rcx),%ymm6,%ymm6 + + subq $64,%rdx + jnz .Loop_avx2_512 + +.byte 0x66,0x90 +.Ltail_avx2_512: + + vpaddq %ymm0,%ymm7,%ymm0 + vmovdqu 4(%rsp),%ymm7 + vpaddq %ymm1,%ymm8,%ymm1 + vmovdqu 36(%rsp),%ymm8 + vpaddq %ymm3,%ymm10,%ymm3 + vmovdqu 100(%rsp),%ymm9 + vpaddq %ymm4,%ymm6,%ymm4 + vmovdqu 52(%rax),%ymm10 + vmovdqu 116(%rax),%ymm5 + + vpmuludq %ymm2,%ymm7,%ymm13 + vpmuludq %ymm2,%ymm8,%ymm14 + vpmuludq %ymm2,%ymm9,%ymm15 + vpmuludq %ymm2,%ymm10,%ymm11 + vpmuludq %ymm2,%ymm5,%ymm12 + + vpmuludq %ymm0,%ymm8,%ymm6 + vpmuludq %ymm1,%ymm8,%ymm2 + vpaddq %ymm6,%ymm12,%ymm12 + vpaddq %ymm2,%ymm13,%ymm13 + vpmuludq %ymm3,%ymm8,%ymm6 + vpmuludq 68(%rsp),%ymm4,%ymm2 + vpaddq %ymm6,%ymm15,%ymm15 + vpaddq %ymm2,%ymm11,%ymm11 + + vpmuludq %ymm0,%ymm7,%ymm6 + vpmuludq %ymm1,%ymm7,%ymm2 + vpaddq %ymm6,%ymm11,%ymm11 + vmovdqu -12(%rax),%ymm8 + vpaddq %ymm2,%ymm12,%ymm12 + vpmuludq %ymm3,%ymm7,%ymm6 + vpmuludq %ymm4,%ymm7,%ymm2 + vpaddq %ymm6,%ymm14,%ymm14 + vpaddq %ymm2,%ymm15,%ymm15 + + vpmuludq %ymm3,%ymm8,%ymm6 + vpmuludq %ymm4,%ymm8,%ymm2 + vpaddq %ymm6,%ymm11,%ymm11 + vpaddq %ymm2,%ymm12,%ymm12 + vmovdqu 20(%rax),%ymm2 + vpmuludq %ymm1,%ymm9,%ymm6 + vpmuludq %ymm0,%ymm9,%ymm9 + vpaddq %ymm6,%ymm14,%ymm14 + vpaddq %ymm9,%ymm13,%ymm13 + + vpmuludq %ymm1,%ymm2,%ymm6 + vpmuludq %ymm0,%ymm2,%ymm2 + vpaddq %ymm6,%ymm15,%ymm15 + vpaddq %ymm2,%ymm14,%ymm14 + vpmuludq %ymm3,%ymm10,%ymm6 + vpmuludq %ymm4,%ymm10,%ymm2 + vpaddq %ymm6,%ymm12,%ymm12 + vpaddq %ymm2,%ymm13,%ymm13 + + vpmuludq %ymm3,%ymm5,%ymm3 + vpmuludq %ymm4,%ymm5,%ymm4 + vpaddq %ymm3,%ymm13,%ymm2 + vpaddq %ymm4,%ymm14,%ymm3 + vpmuludq 84(%rax),%ymm0,%ymm4 + vpmuludq %ymm1,%ymm5,%ymm0 + vmovdqa 64(%rcx),%ymm5 + vpaddq %ymm4,%ymm15,%ymm4 + vpaddq %ymm0,%ymm11,%ymm0 + + vpsrldq $8,%ymm12,%ymm8 + vpsrldq $8,%ymm2,%ymm9 + vpsrldq $8,%ymm3,%ymm10 + vpsrldq $8,%ymm4,%ymm6 + vpsrldq $8,%ymm0,%ymm7 + vpaddq %ymm8,%ymm12,%ymm12 + vpaddq %ymm9,%ymm2,%ymm2 + vpaddq %ymm10,%ymm3,%ymm3 + vpaddq %ymm6,%ymm4,%ymm4 + vpaddq %ymm7,%ymm0,%ymm0 + + vpermq $0x2,%ymm3,%ymm10 + vpermq $0x2,%ymm4,%ymm6 + vpermq $0x2,%ymm0,%ymm7 + vpermq $0x2,%ymm12,%ymm8 + vpermq $0x2,%ymm2,%ymm9 + vpaddq %ymm10,%ymm3,%ymm3 + vpaddq %ymm6,%ymm4,%ymm4 + vpaddq %ymm7,%ymm0,%ymm0 + vpaddq %ymm8,%ymm12,%ymm12 + vpaddq %ymm9,%ymm2,%ymm2 + + vpsrlq $26,%ymm3,%ymm14 + vpand %ymm5,%ymm3,%ymm3 + vpaddq %ymm14,%ymm4,%ymm4 + + vpsrlq $26,%ymm0,%ymm11 + vpand %ymm5,%ymm0,%ymm0 + vpaddq %ymm11,%ymm12,%ymm1 + + vpsrlq $26,%ymm4,%ymm15 + vpand %ymm5,%ymm4,%ymm4 + + vpsrlq $26,%ymm1,%ymm12 + vpand %ymm5,%ymm1,%ymm1 + vpaddq %ymm12,%ymm2,%ymm2 + + vpaddq %ymm15,%ymm0,%ymm0 + vpsllq $2,%ymm15,%ymm15 + vpaddq %ymm15,%ymm0,%ymm0 + + vpsrlq $26,%ymm2,%ymm13 + vpand %ymm5,%ymm2,%ymm2 + vpaddq %ymm13,%ymm3,%ymm3 + + vpsrlq $26,%ymm0,%ymm11 + vpand %ymm5,%ymm0,%ymm0 + vpaddq %ymm11,%ymm1,%ymm1 + + vpsrlq $26,%ymm3,%ymm14 + vpand %ymm5,%ymm3,%ymm3 + vpaddq %ymm14,%ymm4,%ymm4 + + vmovd %xmm0,-112(%rdi) + vmovd %xmm1,-108(%rdi) + vmovd %xmm2,-104(%rdi) + vmovd %xmm3,-100(%rdi) + vmovd %xmm4,-96(%rdi) + leaq -8(%r10),%rsp + + vzeroupper + ret + +.Lblocks_avx512: + + movl $15,%eax + kmovw %eax,%k2 + leaq 8(%rsp),%r10 + + subq $0x128,%rsp + leaq .Lconst(%rip),%rcx + leaq 48+64(%rdi),%rdi + vmovdqa 96(%rcx),%ymm9 + + vmovdqu32 -64(%rdi),%zmm16{%k2}{z} + andq $-512,%rsp + vmovdqu32 -48(%rdi),%zmm17{%k2}{z} + movq $0x20,%rax + vmovdqu32 -32(%rdi),%zmm21{%k2}{z} + vmovdqu32 -16(%rdi),%zmm18{%k2}{z} + vmovdqu32 0(%rdi),%zmm22{%k2}{z} + vmovdqu32 16(%rdi),%zmm19{%k2}{z} + vmovdqu32 32(%rdi),%zmm23{%k2}{z} + vmovdqu32 48(%rdi),%zmm20{%k2}{z} + vmovdqu32 64(%rdi),%zmm24{%k2}{z} + vpermd %zmm16,%zmm9,%zmm16 + vpbroadcastq 64(%rcx),%zmm5 + vpermd %zmm17,%zmm9,%zmm17 + vpermd %zmm21,%zmm9,%zmm21 + vpermd %zmm18,%zmm9,%zmm18 + vmovdqa64 %zmm16,0(%rsp){%k2} + vpsrlq $32,%zmm16,%zmm7 + vpermd %zmm22,%zmm9,%zmm22 + vmovdqu64 %zmm17,0(%rsp,%rax,1){%k2} + vpsrlq $32,%zmm17,%zmm8 + vpermd %zmm19,%zmm9,%zmm19 + vmovdqa64 %zmm21,64(%rsp){%k2} + vpermd %zmm23,%zmm9,%zmm23 + vpermd %zmm20,%zmm9,%zmm20 + vmovdqu64 %zmm18,64(%rsp,%rax,1){%k2} + vpermd %zmm24,%zmm9,%zmm24 + vmovdqa64 %zmm22,128(%rsp){%k2} + vmovdqu64 %zmm19,128(%rsp,%rax,1){%k2} + vmovdqa64 %zmm23,192(%rsp){%k2} + vmovdqu64 %zmm20,192(%rsp,%rax,1){%k2} + vmovdqa64 %zmm24,256(%rsp){%k2} + + vpmuludq %zmm7,%zmm16,%zmm11 + vpmuludq %zmm7,%zmm17,%zmm12 + vpmuludq %zmm7,%zmm18,%zmm13 + vpmuludq %zmm7,%zmm19,%zmm14 + vpmuludq %zmm7,%zmm20,%zmm15 + vpsrlq $32,%zmm18,%zmm9 + + vpmuludq %zmm8,%zmm24,%zmm25 + vpmuludq %zmm8,%zmm16,%zmm26 + vpmuludq %zmm8,%zmm17,%zmm27 + vpmuludq %zmm8,%zmm18,%zmm28 + vpmuludq %zmm8,%zmm19,%zmm29 + vpsrlq $32,%zmm19,%zmm10 + vpaddq %zmm25,%zmm11,%zmm11 + vpaddq %zmm26,%zmm12,%zmm12 + vpaddq %zmm27,%zmm13,%zmm13 + vpaddq %zmm28,%zmm14,%zmm14 + vpaddq %zmm29,%zmm15,%zmm15 + + vpmuludq %zmm9,%zmm23,%zmm25 + vpmuludq %zmm9,%zmm24,%zmm26 + vpmuludq %zmm9,%zmm17,%zmm28 + vpmuludq %zmm9,%zmm18,%zmm29 + vpmuludq %zmm9,%zmm16,%zmm27 + vpsrlq $32,%zmm20,%zmm6 + vpaddq %zmm25,%zmm11,%zmm11 + vpaddq %zmm26,%zmm12,%zmm12 + vpaddq %zmm28,%zmm14,%zmm14 + vpaddq %zmm29,%zmm15,%zmm15 + vpaddq %zmm27,%zmm13,%zmm13 + + vpmuludq %zmm10,%zmm22,%zmm25 + vpmuludq %zmm10,%zmm16,%zmm28 + vpmuludq %zmm10,%zmm17,%zmm29 + vpmuludq %zmm10,%zmm23,%zmm26 + vpmuludq %zmm10,%zmm24,%zmm27 + vpaddq %zmm25,%zmm11,%zmm11 + vpaddq %zmm28,%zmm14,%zmm14 + vpaddq %zmm29,%zmm15,%zmm15 + vpaddq %zmm26,%zmm12,%zmm12 + vpaddq %zmm27,%zmm13,%zmm13 + + vpmuludq %zmm6,%zmm24,%zmm28 + vpmuludq %zmm6,%zmm16,%zmm29 + vpmuludq %zmm6,%zmm21,%zmm25 + vpmuludq %zmm6,%zmm22,%zmm26 + vpmuludq %zmm6,%zmm23,%zmm27 + vpaddq %zmm28,%zmm14,%zmm14 + vpaddq %zmm29,%zmm15,%zmm15 + vpaddq %zmm25,%zmm11,%zmm11 + vpaddq %zmm26,%zmm12,%zmm12 + vpaddq %zmm27,%zmm13,%zmm13 + + vmovdqu64 0(%rsi),%zmm10 + vmovdqu64 64(%rsi),%zmm6 + leaq 128(%rsi),%rsi + + vpsrlq $26,%zmm14,%zmm28 + vpandq %zmm5,%zmm14,%zmm14 + vpaddq %zmm28,%zmm15,%zmm15 + + vpsrlq $26,%zmm11,%zmm25 + vpandq %zmm5,%zmm11,%zmm11 + vpaddq %zmm25,%zmm12,%zmm12 + + vpsrlq $26,%zmm15,%zmm29 + vpandq %zmm5,%zmm15,%zmm15 + + vpsrlq $26,%zmm12,%zmm26 + vpandq %zmm5,%zmm12,%zmm12 + vpaddq %zmm26,%zmm13,%zmm13 + + vpaddq %zmm29,%zmm11,%zmm11 + vpsllq $2,%zmm29,%zmm29 + vpaddq %zmm29,%zmm11,%zmm11 + + vpsrlq $26,%zmm13,%zmm27 + vpandq %zmm5,%zmm13,%zmm13 + vpaddq %zmm27,%zmm14,%zmm14 + + vpsrlq $26,%zmm11,%zmm25 + vpandq %zmm5,%zmm11,%zmm11 + vpaddq %zmm25,%zmm12,%zmm12 + + vpsrlq $26,%zmm14,%zmm28 + vpandq %zmm5,%zmm14,%zmm14 + vpaddq %zmm28,%zmm15,%zmm15 + + vpunpcklqdq %zmm6,%zmm10,%zmm7 + vpunpckhqdq %zmm6,%zmm10,%zmm6 + + vmovdqa32 128(%rcx),%zmm25 + movl $0x7777,%eax + kmovw %eax,%k1 + + vpermd %zmm16,%zmm25,%zmm16 + vpermd %zmm17,%zmm25,%zmm17 + vpermd %zmm18,%zmm25,%zmm18 + vpermd %zmm19,%zmm25,%zmm19 + vpermd %zmm20,%zmm25,%zmm20 + + vpermd %zmm11,%zmm25,%zmm16{%k1} + vpermd %zmm12,%zmm25,%zmm17{%k1} + vpermd %zmm13,%zmm25,%zmm18{%k1} + vpermd %zmm14,%zmm25,%zmm19{%k1} + vpermd %zmm15,%zmm25,%zmm20{%k1} + + vpslld $2,%zmm17,%zmm21 + vpslld $2,%zmm18,%zmm22 + vpslld $2,%zmm19,%zmm23 + vpslld $2,%zmm20,%zmm24 + vpaddd %zmm17,%zmm21,%zmm21 + vpaddd %zmm18,%zmm22,%zmm22 + vpaddd %zmm19,%zmm23,%zmm23 + vpaddd %zmm20,%zmm24,%zmm24 + + vpbroadcastq 32(%rcx),%zmm30 + + vpsrlq $52,%zmm7,%zmm9 + vpsllq $12,%zmm6,%zmm10 + vporq %zmm10,%zmm9,%zmm9 + vpsrlq $26,%zmm7,%zmm8 + vpsrlq $14,%zmm6,%zmm10 + vpsrlq $40,%zmm6,%zmm6 + vpandq %zmm5,%zmm9,%zmm9 + vpandq %zmm5,%zmm7,%zmm7 + + vpaddq %zmm2,%zmm9,%zmm2 + subq $192,%rdx + jbe .Ltail_avx512 + jmp .Loop_avx512 + +.align 32 +.Loop_avx512: + + vpmuludq %zmm2,%zmm17,%zmm14 + vpaddq %zmm0,%zmm7,%zmm0 + vpmuludq %zmm2,%zmm18,%zmm15 + vpandq %zmm5,%zmm8,%zmm8 + vpmuludq %zmm2,%zmm23,%zmm11 + vpandq %zmm5,%zmm10,%zmm10 + vpmuludq %zmm2,%zmm24,%zmm12 + vporq %zmm30,%zmm6,%zmm6 + vpmuludq %zmm2,%zmm16,%zmm13 + vpaddq %zmm1,%zmm8,%zmm1 + vpaddq %zmm3,%zmm10,%zmm3 + vpaddq %zmm4,%zmm6,%zmm4 + + vmovdqu64 0(%rsi),%zmm10 + vmovdqu64 64(%rsi),%zmm6 + leaq 128(%rsi),%rsi + vpmuludq %zmm0,%zmm19,%zmm28 + vpmuludq %zmm0,%zmm20,%zmm29 + vpmuludq %zmm0,%zmm16,%zmm25 + vpmuludq %zmm0,%zmm17,%zmm26 + vpaddq %zmm28,%zmm14,%zmm14 + vpaddq %zmm29,%zmm15,%zmm15 + vpaddq %zmm25,%zmm11,%zmm11 + vpaddq %zmm26,%zmm12,%zmm12 + + vpmuludq %zmm1,%zmm18,%zmm28 + vpmuludq %zmm1,%zmm19,%zmm29 + vpmuludq %zmm1,%zmm24,%zmm25 + vpmuludq %zmm0,%zmm18,%zmm27 + vpaddq %zmm28,%zmm14,%zmm14 + vpaddq %zmm29,%zmm15,%zmm15 + vpaddq %zmm25,%zmm11,%zmm11 + vpaddq %zmm27,%zmm13,%zmm13 + + vpunpcklqdq %zmm6,%zmm10,%zmm7 + vpunpckhqdq %zmm6,%zmm10,%zmm6 + + vpmuludq %zmm3,%zmm16,%zmm28 + vpmuludq %zmm3,%zmm17,%zmm29 + vpmuludq %zmm1,%zmm16,%zmm26 + vpmuludq %zmm1,%zmm17,%zmm27 + vpaddq %zmm28,%zmm14,%zmm14 + vpaddq %zmm29,%zmm15,%zmm15 + vpaddq %zmm26,%zmm12,%zmm12 + vpaddq %zmm27,%zmm13,%zmm13 + + vpmuludq %zmm4,%zmm24,%zmm28 + vpmuludq %zmm4,%zmm16,%zmm29 + vpmuludq %zmm3,%zmm22,%zmm25 + vpmuludq %zmm3,%zmm23,%zmm26 + vpaddq %zmm28,%zmm14,%zmm14 + vpmuludq %zmm3,%zmm24,%zmm27 + vpaddq %zmm29,%zmm15,%zmm15 + vpaddq %zmm25,%zmm11,%zmm11 + vpaddq %zmm26,%zmm12,%zmm12 + vpaddq %zmm27,%zmm13,%zmm13 + + vpmuludq %zmm4,%zmm21,%zmm25 + vpmuludq %zmm4,%zmm22,%zmm26 + vpmuludq %zmm4,%zmm23,%zmm27 + vpaddq %zmm25,%zmm11,%zmm0 + vpaddq %zmm26,%zmm12,%zmm1 + vpaddq %zmm27,%zmm13,%zmm2 + + vpsrlq $52,%zmm7,%zmm9 + vpsllq $12,%zmm6,%zmm10 + + vpsrlq $26,%zmm14,%zmm3 + vpandq %zmm5,%zmm14,%zmm14 + vpaddq %zmm3,%zmm15,%zmm4 + + vporq %zmm10,%zmm9,%zmm9 + + vpsrlq $26,%zmm0,%zmm11 + vpandq %zmm5,%zmm0,%zmm0 + vpaddq %zmm11,%zmm1,%zmm1 + + vpandq %zmm5,%zmm9,%zmm9 + + vpsrlq $26,%zmm4,%zmm15 + vpandq %zmm5,%zmm4,%zmm4 + + vpsrlq $26,%zmm1,%zmm12 + vpandq %zmm5,%zmm1,%zmm1 + vpaddq %zmm12,%zmm2,%zmm2 + + vpaddq %zmm15,%zmm0,%zmm0 + vpsllq $2,%zmm15,%zmm15 + vpaddq %zmm15,%zmm0,%zmm0 + + vpaddq %zmm9,%zmm2,%zmm2 + vpsrlq $26,%zmm7,%zmm8 + + vpsrlq $26,%zmm2,%zmm13 + vpandq %zmm5,%zmm2,%zmm2 + vpaddq %zmm13,%zmm14,%zmm3 + + vpsrlq $14,%zmm6,%zmm10 + + vpsrlq $26,%zmm0,%zmm11 + vpandq %zmm5,%zmm0,%zmm0 + vpaddq %zmm11,%zmm1,%zmm1 + + vpsrlq $40,%zmm6,%zmm6 + + vpsrlq $26,%zmm3,%zmm14 + vpandq %zmm5,%zmm3,%zmm3 + vpaddq %zmm14,%zmm4,%zmm4 + + vpandq %zmm5,%zmm7,%zmm7 + + subq $128,%rdx + ja .Loop_avx512 + +.Ltail_avx512: + + vpsrlq $32,%zmm16,%zmm16 + vpsrlq $32,%zmm17,%zmm17 + vpsrlq $32,%zmm18,%zmm18 + vpsrlq $32,%zmm23,%zmm23 + vpsrlq $32,%zmm24,%zmm24 + vpsrlq $32,%zmm19,%zmm19 + vpsrlq $32,%zmm20,%zmm20 + vpsrlq $32,%zmm21,%zmm21 + vpsrlq $32,%zmm22,%zmm22 + + leaq (%rsi,%rdx,1),%rsi + + vpaddq %zmm0,%zmm7,%zmm0 + + vpmuludq %zmm2,%zmm17,%zmm14 + vpmuludq %zmm2,%zmm18,%zmm15 + vpmuludq %zmm2,%zmm23,%zmm11 + vpandq %zmm5,%zmm8,%zmm8 + vpmuludq %zmm2,%zmm24,%zmm12 + vpandq %zmm5,%zmm10,%zmm10 + vpmuludq %zmm2,%zmm16,%zmm13 + vporq %zmm30,%zmm6,%zmm6 + vpaddq %zmm1,%zmm8,%zmm1 + vpaddq %zmm3,%zmm10,%zmm3 + vpaddq %zmm4,%zmm6,%zmm4 + + vmovdqu 0(%rsi),%xmm7 + vpmuludq %zmm0,%zmm19,%zmm28 + vpmuludq %zmm0,%zmm20,%zmm29 + vpmuludq %zmm0,%zmm16,%zmm25 + vpmuludq %zmm0,%zmm17,%zmm26 + vpaddq %zmm28,%zmm14,%zmm14 + vpaddq %zmm29,%zmm15,%zmm15 + vpaddq %zmm25,%zmm11,%zmm11 + vpaddq %zmm26,%zmm12,%zmm12 + + vmovdqu 16(%rsi),%xmm8 + vpmuludq %zmm1,%zmm18,%zmm28 + vpmuludq %zmm1,%zmm19,%zmm29 + vpmuludq %zmm1,%zmm24,%zmm25 + vpmuludq %zmm0,%zmm18,%zmm27 + vpaddq %zmm28,%zmm14,%zmm14 + vpaddq %zmm29,%zmm15,%zmm15 + vpaddq %zmm25,%zmm11,%zmm11 + vpaddq %zmm27,%zmm13,%zmm13 + + vinserti128 $1,32(%rsi),%ymm7,%ymm7 + vpmuludq %zmm3,%zmm16,%zmm28 + vpmuludq %zmm3,%zmm17,%zmm29 + vpmuludq %zmm1,%zmm16,%zmm26 + vpmuludq %zmm1,%zmm17,%zmm27 + vpaddq %zmm28,%zmm14,%zmm14 + vpaddq %zmm29,%zmm15,%zmm15 + vpaddq %zmm26,%zmm12,%zmm12 + vpaddq %zmm27,%zmm13,%zmm13 + + vinserti128 $1,48(%rsi),%ymm8,%ymm8 + vpmuludq %zmm4,%zmm24,%zmm28 + vpmuludq %zmm4,%zmm16,%zmm29 + vpmuludq %zmm3,%zmm22,%zmm25 + vpmuludq %zmm3,%zmm23,%zmm26 + vpmuludq %zmm3,%zmm24,%zmm27 + vpaddq %zmm28,%zmm14,%zmm3 + vpaddq %zmm29,%zmm15,%zmm15 + vpaddq %zmm25,%zmm11,%zmm11 + vpaddq %zmm26,%zmm12,%zmm12 + vpaddq %zmm27,%zmm13,%zmm13 + + vpmuludq %zmm4,%zmm21,%zmm25 + vpmuludq %zmm4,%zmm22,%zmm26 + vpmuludq %zmm4,%zmm23,%zmm27 + vpaddq %zmm25,%zmm11,%zmm0 + vpaddq %zmm26,%zmm12,%zmm1 + vpaddq %zmm27,%zmm13,%zmm2 + + movl $1,%eax + vpermq $0xb1,%zmm3,%zmm14 + vpermq $0xb1,%zmm15,%zmm4 + vpermq $0xb1,%zmm0,%zmm11 + vpermq $0xb1,%zmm1,%zmm12 + vpermq $0xb1,%zmm2,%zmm13 + vpaddq %zmm14,%zmm3,%zmm3 + vpaddq %zmm15,%zmm4,%zmm4 + vpaddq %zmm11,%zmm0,%zmm0 + vpaddq %zmm12,%zmm1,%zmm1 + vpaddq %zmm13,%zmm2,%zmm2 + + kmovw %eax,%k3 + vpermq $0x2,%zmm3,%zmm14 + vpermq $0x2,%zmm4,%zmm15 + vpermq $0x2,%zmm0,%zmm11 + vpermq $0x2,%zmm1,%zmm12 + vpermq $0x2,%zmm2,%zmm13 + vpaddq %zmm14,%zmm3,%zmm3 + vpaddq %zmm15,%zmm4,%zmm4 + vpaddq %zmm11,%zmm0,%zmm0 + vpaddq %zmm12,%zmm1,%zmm1 + vpaddq %zmm13,%zmm2,%zmm2 + + vextracti64x4 $0x1,%zmm3,%ymm14 + vextracti64x4 $0x1,%zmm4,%ymm15 + vextracti64x4 $0x1,%zmm0,%ymm11 + vextracti64x4 $0x1,%zmm1,%ymm12 + vextracti64x4 $0x1,%zmm2,%ymm13 + vpaddq %zmm14,%zmm3,%zmm3{%k3}{z} + vpaddq %zmm15,%zmm4,%zmm4{%k3}{z} + vpaddq %zmm11,%zmm0,%zmm0{%k3}{z} + vpaddq %zmm12,%zmm1,%zmm1{%k3}{z} + vpaddq %zmm13,%zmm2,%zmm2{%k3}{z} + + vpsrlq $26,%ymm3,%ymm14 + vpand %ymm5,%ymm3,%ymm3 + vpsrldq $6,%ymm7,%ymm9 + vpsrldq $6,%ymm8,%ymm10 + vpunpckhqdq %ymm8,%ymm7,%ymm6 + vpaddq %ymm14,%ymm4,%ymm4 + + vpsrlq $26,%ymm0,%ymm11 + vpand %ymm5,%ymm0,%ymm0 + vpunpcklqdq %ymm10,%ymm9,%ymm9 + vpunpcklqdq %ymm8,%ymm7,%ymm7 + vpaddq %ymm11,%ymm1,%ymm1 + + vpsrlq $26,%ymm4,%ymm15 + vpand %ymm5,%ymm4,%ymm4 + + vpsrlq $26,%ymm1,%ymm12 + vpand %ymm5,%ymm1,%ymm1 + vpsrlq $30,%ymm9,%ymm10 + vpsrlq $4,%ymm9,%ymm9 + vpaddq %ymm12,%ymm2,%ymm2 + + vpaddq %ymm15,%ymm0,%ymm0 + vpsllq $2,%ymm15,%ymm15 + vpsrlq $26,%ymm7,%ymm8 + vpsrlq $40,%ymm6,%ymm6 + vpaddq %ymm15,%ymm0,%ymm0 + + vpsrlq $26,%ymm2,%ymm13 + vpand %ymm5,%ymm2,%ymm2 + vpand %ymm5,%ymm9,%ymm9 + vpand %ymm5,%ymm7,%ymm7 + vpaddq %ymm13,%ymm3,%ymm3 + + vpsrlq $26,%ymm0,%ymm11 + vpand %ymm5,%ymm0,%ymm0 + vpaddq %ymm2,%ymm9,%ymm2 + vpand %ymm5,%ymm8,%ymm8 + vpaddq %ymm11,%ymm1,%ymm1 + + vpsrlq $26,%ymm3,%ymm14 + vpand %ymm5,%ymm3,%ymm3 + vpand %ymm5,%ymm10,%ymm10 + vpor 32(%rcx),%ymm6,%ymm6 + vpaddq %ymm14,%ymm4,%ymm4 + + leaq 144(%rsp),%rax + addq $64,%rdx + jnz .Ltail_avx2_512 + + vpsubq %ymm9,%ymm2,%ymm2 + vmovd %xmm0,-112(%rdi) + vmovd %xmm1,-108(%rdi) + vmovd %xmm2,-104(%rdi) + vmovd %xmm3,-100(%rdi) + vmovd %xmm4,-96(%rdi) + vzeroall + leaq -8(%r10),%rsp + + ret + +ENDPROC(poly1305_blocks_avx512) +#endif /* CONFIG_AS_AVX512 */ diff --git a/lib/zinc/poly1305/poly1305.c b/lib/zinc/poly1305/poly1305.c index 538abc359d1d..2ae1b3cb66cd 100644 --- a/lib/zinc/poly1305/poly1305.c +++ b/lib/zinc/poly1305/poly1305.c @@ -15,7 +15,9 @@ #include #include -#ifndef HAVE_POLY1305_ARCH_IMPLEMENTATION +#if defined(CONFIG_ZINC_ARCH_X86_64) +#include "poly1305-x86_64-glue.h" +#else static inline bool poly1305_init_arch(void *ctx, const u8 key[POLY1305_KEY_SIZE]) {