From patchwork Fri Feb 25 08:44:09 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Hongyu Wang X-Patchwork-Id: 1597543 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@bilbo.ozlabs.org Authentication-Results: bilbo.ozlabs.org; dkim=pass (1024-bit key; unprotected) header.d=gcc.gnu.org header.i=@gcc.gnu.org header.a=rsa-sha256 header.s=default header.b=fwCd8l+X; dkim-atps=neutral Authentication-Results: ozlabs.org; spf=pass (sender SPF authorized) smtp.mailfrom=gcc.gnu.org (client-ip=8.43.85.97; helo=sourceware.org; envelope-from=gcc-patches-bounces+incoming=patchwork.ozlabs.org@gcc.gnu.org; receiver=) Received: from sourceware.org (ip-8-43-85-97.sourceware.org [8.43.85.97]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange X25519 server-signature RSA-PSS (4096 bits) server-digest SHA256) (No client certificate requested) by bilbo.ozlabs.org (Postfix) with ESMTPS id 4K4jyT0H1gz9sG1 for ; Fri, 25 Feb 2022 19:45:03 +1100 (AEDT) Received: from server2.sourceware.org (localhost [IPv6:::1]) by sourceware.org (Postfix) with ESMTP id 11954385803E for ; Fri, 25 Feb 2022 08:45:01 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 11954385803E DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org; s=default; t=1645778701; bh=bf1wSOqzfkhNkPLhfZiU5SC6NwWbXWbqf8X0ADnO1Q0=; h=To:Subject:Date:List-Id:List-Unsubscribe:List-Archive:List-Post: List-Help:List-Subscribe:From:Reply-To:Cc:From; b=fwCd8l+Xq4ljVPmA9qM/cYFWS+jE4DtmFiZUY2RXZAe+5MbuORwC1Onyp0kXSODti NV6y+3+LR6R3LMux4aZB4nvDc9znO1JJQun33L21PGnowjDvv6P7zYtxsTxJosTH6T PEIh3CuV5xXp2nBl6YIigARhqjbwMJ6YvNVEvjIQ= X-Original-To: gcc-patches@gcc.gnu.org Delivered-To: gcc-patches@gcc.gnu.org Received: from mga02.intel.com (mga02.intel.com [134.134.136.20]) by sourceware.org (Postfix) with ESMTPS id 0FFFB3858402 for ; Fri, 25 Feb 2022 08:44:12 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.1 sourceware.org 0FFFB3858402 X-IronPort-AV: E=McAfee;i="6200,9189,10268"; a="239857862" X-IronPort-AV: E=Sophos;i="5.90,136,1643702400"; d="scan'208";a="239857862" Received: from orsmga004.jf.intel.com ([10.7.209.38]) by orsmga101.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 25 Feb 2022 00:44:11 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.90,136,1643702400"; d="scan'208";a="640045210" Received: from scymds01.sc.intel.com ([10.148.94.138]) by orsmga004.jf.intel.com with ESMTP; 25 Feb 2022 00:44:11 -0800 Received: from shliclel320.sh.intel.com (shliclel320.sh.intel.com [10.239.236.50]) by scymds01.sc.intel.com with ESMTP id 21P8iALh016660; Fri, 25 Feb 2022 00:44:10 -0800 To: hongtao.liu@intel.com Subject: [PATCH] AVX512F: Add helper enumeration for ternary logic intrinsics. Date: Fri, 25 Feb 2022 16:44:09 +0800 Message-Id: <20220225084409.58620-1-hongyu.wang@intel.com> X-Mailer: git-send-email 2.18.1 X-Spam-Status: No, score=-10.3 required=5.0 tests=BAYES_00, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, FREEMAIL_ENVFROM_END_DIGIT, FREEMAIL_FORGED_FROMDOMAIN, FREEMAIL_FROM, GIT_PATCH_0, HEADER_FROM_DIFFERENT_DOMAINS, KAM_SHORT, RCVD_IN_MSPIKE_H3, RCVD_IN_MSPIKE_WL, SPF_HELO_NONE, SPF_SOFTFAIL, TXREP, T_SCC_BODY_TEXT_LINE autolearn=ham autolearn_force=no version=3.4.4 X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on server2.sourceware.org X-BeenThere: gcc-patches@gcc.gnu.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Gcc-patches mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-Patchwork-Original-From: Hongyu Wang via Gcc-patches From: Hongyu Wang Reply-To: Hongyu Wang Cc: gcc-patches@gcc.gnu.org Errors-To: gcc-patches-bounces+incoming=patchwork.ozlabs.org@gcc.gnu.org Sender: "Gcc-patches" Hi, This patch intends to sync with llvm change in https://reviews.llvm.org/D120307 to add enumeration and truncate imm to unsigned char, so users could use ~ on immediates. Bootstraped/regtested on x86_64-pc-linux-gnu{-m32,}. Ok for master? gcc/ChangeLog: * config/i386/avx512fintrin.h (_MM_TERNLOG_ENUM): New enum. (_mm512_ternarylogic_epi64): Truncate imm to unsigned char to avoid error when using ~enum as parameter. (_mm512_mask_ternarylogic_epi64): Likewise. (_mm512_maskz_ternarylogic_epi64): Likewise. (_mm512_ternarylogic_epi32): Likewise. (_mm512_mask_ternarylogic_epi32): Likewise. (_mm512_maskz_ternarylogic_epi32): Likewise. * config/i386/avx512vlintrin.h (_mm256_ternarylogic_epi64): Adjust imm param type to unsigned char. (_mm256_mask_ternarylogic_epi64): Likewise. (_mm256_maskz_ternarylogic_epi64): Likewise. (_mm256_ternarylogic_epi32): Likewise. (_mm256_mask_ternarylogic_epi32): Likewise. (_mm256_maskz_ternarylogic_epi32): Likewise. (_mm_ternarylogic_epi64): Likewise. (_mm_mask_ternarylogic_epi64): Likewise. (_mm_maskz_ternarylogic_epi64): Likewise. (_mm_ternarylogic_epi32): Likewise. (_mm_mask_ternarylogic_epi32): Likewise. (_mm_maskz_ternarylogic_epi32): Likewise. gcc/testsuite/ChangeLog: * gcc.target/i386/avx512f-vpternlogd-1.c: Use new enum. * gcc.target/i386/avx512f-vpternlogq-1.c: Likewise. * gcc.target/i386/avx512vl-vpternlogd-1.c: Likewise. * gcc.target/i386/avx512vl-vpternlogq-1.c: Likewise. * gcc.target/i386/testimm-10.c: Remove imm check for vpternlog insns since the imm has been truncated in intrinsic. --- gcc/config/i386/avx512fintrin.h | 132 ++++++--- gcc/config/i386/avx512vlintrin.h | 278 +++++++++++------- .../gcc.target/i386/avx512f-vpternlogd-1.c | 7 +- .../gcc.target/i386/avx512f-vpternlogq-1.c | 7 +- .../gcc.target/i386/avx512vl-vpternlogd-1.c | 13 +- .../gcc.target/i386/avx512vl-vpternlogq-1.c | 14 +- gcc/testsuite/gcc.target/i386/testimm-10.c | 7 - 7 files changed, 285 insertions(+), 173 deletions(-) diff --git a/gcc/config/i386/avx512fintrin.h b/gcc/config/i386/avx512fintrin.h index bc10c823c76..29511fd2831 100644 --- a/gcc/config/i386/avx512fintrin.h +++ b/gcc/config/i386/avx512fintrin.h @@ -1639,16 +1639,27 @@ _mm_maskz_sub_round_ss (__mmask8 __U, __m128 __A, __m128 __B, #endif +/* Constant helper to represent the ternary logic operations among + vector A, B and C. */ +typedef enum +{ + _MM_TERNLOG_A = 0xF0, + _MM_TERNLOG_B = 0xCC, + _MM_TERNLOG_C = 0xAA +} _MM_TERNLOG_ENUM; + #ifdef __OPTIMIZE__ extern __inline __m512i __attribute__ ((__gnu_inline__, __always_inline__, __artificial__)) _mm512_ternarylogic_epi64 (__m512i __A, __m512i __B, __m512i __C, const int __imm) { - return (__m512i) __builtin_ia32_pternlogq512_mask ((__v8di) __A, - (__v8di) __B, - (__v8di) __C, __imm, - (__mmask8) -1); + return (__m512i) + __builtin_ia32_pternlogq512_mask ((__v8di) __A, + (__v8di) __B, + (__v8di) __C, + (unsigned char) __imm, + (__mmask8) -1); } extern __inline __m512i @@ -1656,10 +1667,12 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__)) _mm512_mask_ternarylogic_epi64 (__m512i __A, __mmask8 __U, __m512i __B, __m512i __C, const int __imm) { - return (__m512i) __builtin_ia32_pternlogq512_mask ((__v8di) __A, - (__v8di) __B, - (__v8di) __C, __imm, - (__mmask8) __U); + return (__m512i) + __builtin_ia32_pternlogq512_mask ((__v8di) __A, + (__v8di) __B, + (__v8di) __C, + (unsigned char) __imm, + (__mmask8) __U); } extern __inline __m512i @@ -1667,10 +1680,12 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__)) _mm512_maskz_ternarylogic_epi64 (__mmask8 __U, __m512i __A, __m512i __B, __m512i __C, const int __imm) { - return (__m512i) __builtin_ia32_pternlogq512_maskz ((__v8di) __A, - (__v8di) __B, - (__v8di) __C, - __imm, (__mmask8) __U); + return (__m512i) + __builtin_ia32_pternlogq512_maskz ((__v8di) __A, + (__v8di) __B, + (__v8di) __C, + (unsigned char) __imm, + (__mmask8) __U); } extern __inline __m512i @@ -1678,10 +1693,12 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__)) _mm512_ternarylogic_epi32 (__m512i __A, __m512i __B, __m512i __C, const int __imm) { - return (__m512i) __builtin_ia32_pternlogd512_mask ((__v16si) __A, - (__v16si) __B, - (__v16si) __C, - __imm, (__mmask16) -1); + return (__m512i) + __builtin_ia32_pternlogd512_mask ((__v16si) __A, + (__v16si) __B, + (__v16si) __C, + (unsigned char) __imm, + (__mmask16) -1); } extern __inline __m512i @@ -1689,10 +1706,12 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__)) _mm512_mask_ternarylogic_epi32 (__m512i __A, __mmask16 __U, __m512i __B, __m512i __C, const int __imm) { - return (__m512i) __builtin_ia32_pternlogd512_mask ((__v16si) __A, - (__v16si) __B, - (__v16si) __C, - __imm, (__mmask16) __U); + return (__m512i) + __builtin_ia32_pternlogd512_mask ((__v16si) __A, + (__v16si) __B, + (__v16si) __C, + (unsigned char) __imm, + (__mmask16) __U); } extern __inline __m512i @@ -1700,33 +1719,56 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__)) _mm512_maskz_ternarylogic_epi32 (__mmask16 __U, __m512i __A, __m512i __B, __m512i __C, const int __imm) { - return (__m512i) __builtin_ia32_pternlogd512_maskz ((__v16si) __A, - (__v16si) __B, - (__v16si) __C, - __imm, (__mmask16) __U); + return (__m512i) + __builtin_ia32_pternlogd512_maskz ((__v16si) __A, + (__v16si) __B, + (__v16si) __C, + (unsigned char) __imm, + (__mmask16) __U); } #else -#define _mm512_ternarylogic_epi64(A, B, C, I) \ - ((__m512i) __builtin_ia32_pternlogq512_mask ((__v8di)(__m512i)(A), \ - (__v8di)(__m512i)(B), (__v8di)(__m512i)(C), (int)(I), (__mmask8)-1)) -#define _mm512_mask_ternarylogic_epi64(A, U, B, C, I) \ - ((__m512i) __builtin_ia32_pternlogq512_mask ((__v8di)(__m512i)(A), \ - (__v8di)(__m512i)(B), (__v8di)(__m512i)(C), (int)(I), (__mmask8)(U))) -#define _mm512_maskz_ternarylogic_epi64(U, A, B, C, I) \ - ((__m512i) __builtin_ia32_pternlogq512_maskz ((__v8di)(__m512i)(A), \ - (__v8di)(__m512i)(B), (__v8di)(__m512i)(C), (int)(I), (__mmask8)(U))) -#define _mm512_ternarylogic_epi32(A, B, C, I) \ - ((__m512i) __builtin_ia32_pternlogd512_mask ((__v16si)(__m512i)(A), \ - (__v16si)(__m512i)(B), (__v16si)(__m512i)(C), (int)(I), \ - (__mmask16)-1)) -#define _mm512_mask_ternarylogic_epi32(A, U, B, C, I) \ - ((__m512i) __builtin_ia32_pternlogd512_mask ((__v16si)(__m512i)(A), \ - (__v16si)(__m512i)(B), (__v16si)(__m512i)(C), (int)(I), \ - (__mmask16)(U))) -#define _mm512_maskz_ternarylogic_epi32(U, A, B, C, I) \ - ((__m512i) __builtin_ia32_pternlogd512_maskz ((__v16si)(__m512i)(A), \ - (__v16si)(__m512i)(B), (__v16si)(__m512i)(C), (int)(I), \ - (__mmask16)(U))) +#define _mm512_ternarylogic_epi64(A, B, C, I) \ + ((__m512i) \ + __builtin_ia32_pternlogq512_mask ((__v8di) (__m512i) (A), \ + (__v8di) (__m512i) (B), \ + (__v8di) (__m512i) (C), \ + (unsigned char) (I), \ + (__mmask8) -1)) +#define _mm512_mask_ternarylogic_epi64(A, U, B, C, I) \ + ((__m512i) \ + __builtin_ia32_pternlogq512_mask ((__v8di) (__m512i) (A), \ + (__v8di) (__m512i) (B), \ + (__v8di) (__m512i) (C), \ + (unsigned char)(I), \ + (__mmask8) (U))) +#define _mm512_maskz_ternarylogic_epi64(U, A, B, C, I) \ + ((__m512i) \ + __builtin_ia32_pternlogq512_maskz ((__v8di) (__m512i) (A), \ + (__v8di) (__m512i) (B), \ + (__v8di) (__m512i) (C), \ + (unsigned char) (I), \ + (__mmask8) (U))) +#define _mm512_ternarylogic_epi32(A, B, C, I) \ + ((__m512i) \ + __builtin_ia32_pternlogd512_mask ((__v16si) (__m512i) (A), \ + (__v16si) (__m512i) (B), \ + (__v16si) (__m512i) (C), \ + (unsigned char) (I), \ + (__mmask16) -1)) +#define _mm512_mask_ternarylogic_epi32(A, U, B, C, I) \ + ((__m512i) \ + __builtin_ia32_pternlogd512_mask ((__v16si) (__m512i) (A), \ + (__v16si) (__m512i) (B), \ + (__v16si) (__m512i) (C), \ + (unsigned char) (I), \ + (__mmask16) (U))) +#define _mm512_maskz_ternarylogic_epi32(U, A, B, C, I) \ + ((__m512i) \ + __builtin_ia32_pternlogd512_maskz ((__v16si) (__m512i) (A), \ + (__v16si) (__m512i) (B), \ + (__v16si) (__m512i) (C), \ + (unsigned char) (I), \ + (__mmask16) (U))) #endif extern __inline __m512d diff --git a/gcc/config/i386/avx512vlintrin.h b/gcc/config/i386/avx512vlintrin.h index bbced242506..26b286eae6b 100644 --- a/gcc/config/i386/avx512vlintrin.h +++ b/gcc/config/i386/avx512vlintrin.h @@ -10575,10 +10575,12 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__)) _mm256_ternarylogic_epi64 (__m256i __A, __m256i __B, __m256i __C, const int __imm) { - return (__m256i) __builtin_ia32_pternlogq256_mask ((__v4di) __A, - (__v4di) __B, - (__v4di) __C, __imm, - (__mmask8) -1); + return (__m256i) + __builtin_ia32_pternlogq256_mask ((__v4di) __A, + (__v4di) __B, + (__v4di) __C, + (unsigned char) __imm, + (__mmask8) -1); } extern __inline __m256i @@ -10587,10 +10589,12 @@ _mm256_mask_ternarylogic_epi64 (__m256i __A, __mmask8 __U, __m256i __B, __m256i __C, const int __imm) { - return (__m256i) __builtin_ia32_pternlogq256_mask ((__v4di) __A, - (__v4di) __B, - (__v4di) __C, __imm, - (__mmask8) __U); + return (__m256i) + __builtin_ia32_pternlogq256_mask ((__v4di) __A, + (__v4di) __B, + (__v4di) __C, + (unsigned char) __imm, + (__mmask8) __U); } extern __inline __m256i @@ -10599,11 +10603,12 @@ _mm256_maskz_ternarylogic_epi64 (__mmask8 __U, __m256i __A, __m256i __B, __m256i __C, const int __imm) { - return (__m256i) __builtin_ia32_pternlogq256_maskz ((__v4di) __A, - (__v4di) __B, - (__v4di) __C, - __imm, - (__mmask8) __U); + return (__m256i) + __builtin_ia32_pternlogq256_maskz ((__v4di) __A, + (__v4di) __B, + (__v4di) __C, + (unsigned char) __imm, + (__mmask8) __U); } extern __inline __m256i @@ -10611,10 +10616,12 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__)) _mm256_ternarylogic_epi32 (__m256i __A, __m256i __B, __m256i __C, const int __imm) { - return (__m256i) __builtin_ia32_pternlogd256_mask ((__v8si) __A, - (__v8si) __B, - (__v8si) __C, __imm, - (__mmask8) -1); + return (__m256i) + __builtin_ia32_pternlogd256_mask ((__v8si) __A, + (__v8si) __B, + (__v8si) __C, + (unsigned char) __imm, + (__mmask8) -1); } extern __inline __m256i @@ -10623,10 +10630,12 @@ _mm256_mask_ternarylogic_epi32 (__m256i __A, __mmask8 __U, __m256i __B, __m256i __C, const int __imm) { - return (__m256i) __builtin_ia32_pternlogd256_mask ((__v8si) __A, - (__v8si) __B, - (__v8si) __C, __imm, - (__mmask8) __U); + return (__m256i) + __builtin_ia32_pternlogd256_mask ((__v8si) __A, + (__v8si) __B, + (__v8si) __C, + (unsigned char) __imm, + (__mmask8) __U); } extern __inline __m256i @@ -10635,11 +10644,12 @@ _mm256_maskz_ternarylogic_epi32 (__mmask8 __U, __m256i __A, __m256i __B, __m256i __C, const int __imm) { - return (__m256i) __builtin_ia32_pternlogd256_maskz ((__v8si) __A, - (__v8si) __B, - (__v8si) __C, - __imm, - (__mmask8) __U); + return (__m256i) + __builtin_ia32_pternlogd256_maskz ((__v8si) __A, + (__v8si) __B, + (__v8si) __C, + (unsigned char) __imm, + (__mmask8) __U); } extern __inline __m128i @@ -10647,33 +10657,40 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__)) _mm_ternarylogic_epi64 (__m128i __A, __m128i __B, __m128i __C, const int __imm) { - return (__m128i) __builtin_ia32_pternlogq128_mask ((__v2di) __A, - (__v2di) __B, - (__v2di) __C, __imm, - (__mmask8) -1); + return (__m128i) + __builtin_ia32_pternlogq128_mask ((__v2di) __A, + (__v2di) __B, + (__v2di) __C, + (unsigned char) __imm, + (__mmask8) -1); } extern __inline __m128i __attribute__ ((__gnu_inline__, __always_inline__, __artificial__)) _mm_mask_ternarylogic_epi64 (__m128i __A, __mmask8 __U, - __m128i __B, __m128i __C, const int __imm) + __m128i __B, __m128i __C, + const int __imm) { - return (__m128i) __builtin_ia32_pternlogq128_mask ((__v2di) __A, - (__v2di) __B, - (__v2di) __C, __imm, - (__mmask8) __U); + return (__m128i) + __builtin_ia32_pternlogq128_mask ((__v2di) __A, + (__v2di) __B, + (__v2di) __C, + (unsigned char) __imm, + (__mmask8) __U); } extern __inline __m128i __attribute__ ((__gnu_inline__, __always_inline__, __artificial__)) _mm_maskz_ternarylogic_epi64 (__mmask8 __U, __m128i __A, - __m128i __B, __m128i __C, const int __imm) + __m128i __B, __m128i __C, + const int __imm) { - return (__m128i) __builtin_ia32_pternlogq128_maskz ((__v2di) __A, - (__v2di) __B, - (__v2di) __C, - __imm, - (__mmask8) __U); + return (__m128i) + __builtin_ia32_pternlogq128_maskz ((__v2di) __A, + (__v2di) __B, + (__v2di) __C, + (unsigned char) __imm, + (__mmask8) __U); } extern __inline __m128i @@ -10681,33 +10698,40 @@ __attribute__ ((__gnu_inline__, __always_inline__, __artificial__)) _mm_ternarylogic_epi32 (__m128i __A, __m128i __B, __m128i __C, const int __imm) { - return (__m128i) __builtin_ia32_pternlogd128_mask ((__v4si) __A, - (__v4si) __B, - (__v4si) __C, __imm, - (__mmask8) -1); + return (__m128i) + __builtin_ia32_pternlogd128_mask ((__v4si) __A, + (__v4si) __B, + (__v4si) __C, + (unsigned char) __imm, + (__mmask8) -1); } extern __inline __m128i __attribute__ ((__gnu_inline__, __always_inline__, __artificial__)) _mm_mask_ternarylogic_epi32 (__m128i __A, __mmask8 __U, - __m128i __B, __m128i __C, const int __imm) + __m128i __B, __m128i __C, + const int __imm) { - return (__m128i) __builtin_ia32_pternlogd128_mask ((__v4si) __A, - (__v4si) __B, - (__v4si) __C, __imm, - (__mmask8) __U); + return (__m128i) + __builtin_ia32_pternlogd128_mask ((__v4si) __A, + (__v4si) __B, + (__v4si) __C, + (unsigned char) __imm, + (__mmask8) __U); } extern __inline __m128i __attribute__ ((__gnu_inline__, __always_inline__, __artificial__)) _mm_maskz_ternarylogic_epi32 (__mmask8 __U, __m128i __A, - __m128i __B, __m128i __C, const int __imm) + __m128i __B, __m128i __C, + const int __imm) { - return (__m128i) __builtin_ia32_pternlogd128_maskz ((__v4si) __A, - (__v4si) __B, - (__v4si) __C, - __imm, - (__mmask8) __U); + return (__m128i) + __builtin_ia32_pternlogd128_maskz ((__v4si) __A, + (__v4si) __B, + (__v4si) __C, + (unsigned char) __imm, + (__mmask8) __U); } extern __inline __m256 @@ -12910,53 +12934,101 @@ _mm256_permutex_pd (__m256d __X, const int __M) (__v2di)(__m128i)_mm_setzero_si128 (),\ (__mmask8)(U))) -#define _mm256_ternarylogic_epi64(A, B, C, I) \ - ((__m256i) __builtin_ia32_pternlogq256_mask ((__v4di)(__m256i)(A), \ - (__v4di)(__m256i)(B), (__v4di)(__m256i)(C), (int)(I), (__mmask8)-1)) - -#define _mm256_mask_ternarylogic_epi64(A, U, B, C, I) \ - ((__m256i) __builtin_ia32_pternlogq256_mask ((__v4di)(__m256i)(A), \ - (__v4di)(__m256i)(B), (__v4di)(__m256i)(C), (int)(I), (__mmask8)(U))) - -#define _mm256_maskz_ternarylogic_epi64(U, A, B, C, I) \ - ((__m256i) __builtin_ia32_pternlogq256_maskz ((__v4di)(__m256i)(A), \ - (__v4di)(__m256i)(B), (__v4di)(__m256i)(C), (int)(I), (__mmask8)(U))) - -#define _mm256_ternarylogic_epi32(A, B, C, I) \ - ((__m256i) __builtin_ia32_pternlogd256_mask ((__v8si)(__m256i)(A), \ - (__v8si)(__m256i)(B), (__v8si)(__m256i)(C), (int)(I), (__mmask8)-1)) - -#define _mm256_mask_ternarylogic_epi32(A, U, B, C, I) \ - ((__m256i) __builtin_ia32_pternlogd256_mask ((__v8si)(__m256i)(A), \ - (__v8si)(__m256i)(B), (__v8si)(__m256i)(C), (int)(I), (__mmask8)(U))) - -#define _mm256_maskz_ternarylogic_epi32(U, A, B, C, I) \ - ((__m256i) __builtin_ia32_pternlogd256_maskz ((__v8si)(__m256i)(A), \ - (__v8si)(__m256i)(B), (__v8si)(__m256i)(C), (int)(I), (__mmask8)(U))) - -#define _mm_ternarylogic_epi64(A, B, C, I) \ - ((__m128i) __builtin_ia32_pternlogq128_mask ((__v2di)(__m128i)(A), \ - (__v2di)(__m128i)(B), (__v2di)(__m128i)(C), (int)(I), (__mmask8)-1)) - -#define _mm_mask_ternarylogic_epi64(A, U, B, C, I) \ - ((__m128i) __builtin_ia32_pternlogq128_mask ((__v2di)(__m128i)(A), \ - (__v2di)(__m128i)(B), (__v2di)(__m128i)(C), (int)(I), (__mmask8)(U))) - -#define _mm_maskz_ternarylogic_epi64(U, A, B, C, I) \ - ((__m128i) __builtin_ia32_pternlogq128_maskz ((__v2di)(__m128i)(A), \ - (__v2di)(__m128i)(B), (__v2di)(__m128i)(C), (int)(I), (__mmask8)(U))) - -#define _mm_ternarylogic_epi32(A, B, C, I) \ - ((__m128i) __builtin_ia32_pternlogd128_mask ((__v4si)(__m128i)(A), \ - (__v4si)(__m128i)(B), (__v4si)(__m128i)(C), (int)(I), (__mmask8)-1)) - -#define _mm_mask_ternarylogic_epi32(A, U, B, C, I) \ - ((__m128i) __builtin_ia32_pternlogd128_mask ((__v4si)(__m128i)(A), \ - (__v4si)(__m128i)(B), (__v4si)(__m128i)(C), (int)(I), (__mmask8)(U))) - -#define _mm_maskz_ternarylogic_epi32(U, A, B, C, I) \ - ((__m128i) __builtin_ia32_pternlogd128_maskz ((__v4si)(__m128i)(A), \ - (__v4si)(__m128i)(B), (__v4si)(__m128i)(C), (int)(I), (__mmask8)(U))) +#define _mm256_ternarylogic_epi64(A, B, C, I) \ + ((__m256i) \ + __builtin_ia32_pternlogq256_mask ((__v4di) (__m256i) (A), \ + (__v4di) (__m256i) (B), \ + (__v4di) (__m256i) (C), \ + (unsigned char) (I), \ + (__mmask8) -1)) + +#define _mm256_mask_ternarylogic_epi64(A, U, B, C, I) \ + ((__m256i) \ + __builtin_ia32_pternlogq256_mask ((__v4di) (__m256i) (A), \ + (__v4di) (__m256i) (B), \ + (__v4di) (__m256i) (C), \ + (unsigned char) (I), \ + (__mmask8) (U))) + +#define _mm256_maskz_ternarylogic_epi64(U, A, B, C, I) \ + ((__m256i) \ + __builtin_ia32_pternlogq256_maskz ((__v4di) (__m256i) (A), \ + (__v4di) (__m256i) (B), \ + (__v4di) (__m256i) (C), \ + (unsigned char) (I), \ + (__mmask8) (U))) + +#define _mm256_ternarylogic_epi32(A, B, C, I) \ + ((__m256i) \ + __builtin_ia32_pternlogd256_mask ((__v8si) (__m256i) (A), \ + (__v8si) (__m256i) (B), \ + (__v8si) (__m256i) (C), \ + (unsigned char) (I), \ + (__mmask8) -1)) + +#define _mm256_mask_ternarylogic_epi32(A, U, B, C, I) \ + ((__m256i) \ + __builtin_ia32_pternlogd256_mask ((__v8si) (__m256i) (A), \ + (__v8si) (__m256i) (B), \ + (__v8si) (__m256i) (C), \ + (unsigned char) (I), \ + (__mmask8) (U))) + +#define _mm256_maskz_ternarylogic_epi32(U, A, B, C, I) \ + ((__m256i) \ + __builtin_ia32_pternlogd256_maskz ((__v8si) (__m256i) (A), \ + (__v8si) (__m256i) (B), \ + (__v8si) (__m256i) (C), \ + (unsigned char) (I), \ + (__mmask8) (U))) + +#define _mm_ternarylogic_epi64(A, B, C, I) \ + ((__m128i) \ + __builtin_ia32_pternlogq128_mask ((__v2di) (__m128i) (A), \ + (__v2di) (__m128i) (B), \ + (__v2di) (__m128i) (C), \ + (unsigned char) (I), \ + (__mmask8) -1)) + +#define _mm_mask_ternarylogic_epi64(A, U, B, C, I) \ + ((__m128i) \ + __builtin_ia32_pternlogq128_mask ((__v2di) (__m128i) (A), \ + (__v2di) (__m128i) (B), \ + (__v2di) (__m128i) (C), \ + (unsigned char) (I), \ + (__mmask8) (U))) + +#define _mm_maskz_ternarylogic_epi64(U, A, B, C, I) \ + ((__m128i) \ + __builtin_ia32_pternlogq128_maskz ((__v2di) (__m128i) (A), \ + (__v2di) (__m128i) (B), \ + (__v2di) (__m128i) (C), \ + (unsigned char) (I), \ + (__mmask8) (U))) + +#define _mm_ternarylogic_epi32(A, B, C, I) \ + ((__m128i) \ + __builtin_ia32_pternlogd128_mask ((__v4si) (__m128i) (A), \ + (__v4si) (__m128i) (B), \ + (__v4si) (__m128i) (C), \ + (unsigned char) (I), \ + (__mmask8) -1)) + +#define _mm_mask_ternarylogic_epi32(A, U, B, C, I) \ + ((__m128i) \ + __builtin_ia32_pternlogd128_mask ((__v4si) (__m128i) (A), \ + (__v4si) (__m128i) (B), \ + (__v4si) (__m128i) (C), \ + (unsigned char) (I), \ + (__mmask8) (U))) + +#define _mm_maskz_ternarylogic_epi32(U, A, B, C, I) \ + ((__m128i) \ + __builtin_ia32_pternlogd128_maskz ((__v4si) (__m128i) (A), \ + (__v4si) (__m128i) (B), \ + (__v4si) (__m128i) (C), \ + (unsigned char) (I), \ + (__mmask8) (U))) #define _mm256_roundscale_ps(A, B) \ ((__m256) __builtin_ia32_rndscaleps_256_mask ((__v8sf)(__m256)(A), \ diff --git a/gcc/testsuite/gcc.target/i386/avx512f-vpternlogd-1.c b/gcc/testsuite/gcc.target/i386/avx512f-vpternlogd-1.c index b14af235ea3..a88153a85db 100644 --- a/gcc/testsuite/gcc.target/i386/avx512f-vpternlogd-1.c +++ b/gcc/testsuite/gcc.target/i386/avx512f-vpternlogd-1.c @@ -12,7 +12,8 @@ volatile __mmask16 m; void extern avx512f_test (void) { - x = _mm512_ternarylogic_epi32 (x, y, z, 0xF0); - x = _mm512_mask_ternarylogic_epi32 (x, m, y, z, 0xF0); - x = _mm512_maskz_ternarylogic_epi32 (m, x, y, z, 0xF0); + x = _mm512_ternarylogic_epi32 (x, y, z, _MM_TERNLOG_A); + x = _mm512_mask_ternarylogic_epi32 (x, m, y, z, _MM_TERNLOG_B); + x = _mm512_maskz_ternarylogic_epi32 (m, x, y, z, + _MM_TERNLOG_A | _MM_TERNLOG_C); } diff --git a/gcc/testsuite/gcc.target/i386/avx512f-vpternlogq-1.c b/gcc/testsuite/gcc.target/i386/avx512f-vpternlogq-1.c index e7aa7a7d3ed..ef302464765 100644 --- a/gcc/testsuite/gcc.target/i386/avx512f-vpternlogq-1.c +++ b/gcc/testsuite/gcc.target/i386/avx512f-vpternlogq-1.c @@ -12,7 +12,8 @@ volatile __mmask8 m; void extern avx512f_test (void) { - x = _mm512_ternarylogic_epi64 (x, y, z, 0xF0); - x = _mm512_mask_ternarylogic_epi64 (x, m, y, z, 0xF0); - x = _mm512_maskz_ternarylogic_epi64 (m, x, y, z, 0xF0); + x = _mm512_ternarylogic_epi64 (x, y, z, _MM_TERNLOG_A); + x = _mm512_mask_ternarylogic_epi64 (x, m, y, z, _MM_TERNLOG_B); + x = _mm512_maskz_ternarylogic_epi64 (m, x, y, z, + _MM_TERNLOG_A | _MM_TERNLOG_C); } diff --git a/gcc/testsuite/gcc.target/i386/avx512vl-vpternlogd-1.c b/gcc/testsuite/gcc.target/i386/avx512vl-vpternlogd-1.c index 590e1ef3213..045a266664c 100644 --- a/gcc/testsuite/gcc.target/i386/avx512vl-vpternlogd-1.c +++ b/gcc/testsuite/gcc.target/i386/avx512vl-vpternlogd-1.c @@ -16,11 +16,12 @@ volatile __mmask8 m; void extern avx512vl_test (void) { - y = _mm256_ternarylogic_epi32 (y, y2, y3, 0xF0); - y = _mm256_mask_ternarylogic_epi32 (y, m, y2, y3, 0xF0); - y = _mm256_maskz_ternarylogic_epi32 (m, y, y2, y3, 0xF0); + y = _mm256_ternarylogic_epi32 (y, y2, y3, _MM_TERNLOG_A); + y = _mm256_mask_ternarylogic_epi32 (y, m, y2, y3, _MM_TERNLOG_B); + y = _mm256_maskz_ternarylogic_epi32 (m, y, y2, y3, _MM_TERNLOG_C); - x = _mm_ternarylogic_epi32 (x, x2, x3, 0xF0); - x = _mm_mask_ternarylogic_epi32 (x, m, x2, x3, 0xF0); - x = _mm_maskz_ternarylogic_epi32 (m, x, x2, x3, 0xF0); + x = _mm_ternarylogic_epi32 (x, x2, x3, _MM_TERNLOG_A); + x = _mm_mask_ternarylogic_epi32 (x, m, x2, x3, ~_MM_TERNLOG_B); + x = _mm_maskz_ternarylogic_epi32 (m, x, x2, x3, + _MM_TERNLOG_A | _MM_TERNLOG_C); } diff --git a/gcc/testsuite/gcc.target/i386/avx512vl-vpternlogq-1.c b/gcc/testsuite/gcc.target/i386/avx512vl-vpternlogq-1.c index 3e08a71e703..3a6707c8f65 100644 --- a/gcc/testsuite/gcc.target/i386/avx512vl-vpternlogq-1.c +++ b/gcc/testsuite/gcc.target/i386/avx512vl-vpternlogq-1.c @@ -16,11 +16,13 @@ volatile __mmask8 m; void extern avx512vl_test (void) { - y = _mm256_ternarylogic_epi64 (y, y2, y3, 0xF0); - y = _mm256_mask_ternarylogic_epi64 (y, m, y2, y3, 0xF0); - y = _mm256_maskz_ternarylogic_epi64 (m, y, y2, y3, 0xF0); + y = _mm256_ternarylogic_epi64 (y, y2, y3, _MM_TERNLOG_A); + y = _mm256_mask_ternarylogic_epi64 (y, m, y2, y3, ~_MM_TERNLOG_B); + y = _mm256_maskz_ternarylogic_epi64 (m, y, y2, y3, + _MM_TERNLOG_A | _MM_TERNLOG_C); - x = _mm_ternarylogic_epi64 (x, x2, x3, 0xF0); - x = _mm_mask_ternarylogic_epi64 (x, m, x2, x3, 0xF0); - x = _mm_maskz_ternarylogic_epi64 (m, x, x2, x3, 0xF0); + x = _mm_ternarylogic_epi64 (x, x2, x3, _MM_TERNLOG_A); + x = _mm_mask_ternarylogic_epi64 (x, m, x2, x3, ~_MM_TERNLOG_B); + x = _mm_maskz_ternarylogic_epi64 (m, x, x2, x3, + _MM_TERNLOG_A | _MM_TERNLOG_C); } diff --git a/gcc/testsuite/gcc.target/i386/testimm-10.c b/gcc/testsuite/gcc.target/i386/testimm-10.c index d0e9b42f2fe..426295a439e 100644 --- a/gcc/testsuite/gcc.target/i386/testimm-10.c +++ b/gcc/testsuite/gcc.target/i386/testimm-10.c @@ -22,13 +22,6 @@ test8bit (void) m512i = _mm512_mask_permutex_epi64 (m512i, mmask8, m512i, 256); /* { dg-error "the last argument must be an 8-bit immediate" } */ m512i = _mm512_maskz_permutex_epi64 (mmask8, m512i, 256); /* { dg-error "the last argument must be an 8-bit immediate" } */ - m512i = _mm512_ternarylogic_epi64 (m512i, m512i, m512i, 256); /* { dg-error "the last argument must be an 8-bit immediate" } */ - m512i = _mm512_mask_ternarylogic_epi64 (m512i, mmask8, m512i, m512i, 256); /* { dg-error "the last argument must be an 8-bit immediate" } */ - m512i = _mm512_maskz_ternarylogic_epi64 (mmask8, m512i, m512i, m512i, 256); /* { dg-error "the last argument must be an 8-bit immediate" } */ - m512i = _mm512_ternarylogic_epi32 (m512i, m512i, m512i, 256); /* { dg-error "the last argument must be an 8-bit immediate" } */ - m512i = _mm512_mask_ternarylogic_epi32 (m512i, mmask16, m512i, m512i, 256); /* { dg-error "the last argument must be an 8-bit immediate" } */ - m512i = _mm512_maskz_ternarylogic_epi32 (mmask16, m512i, m512i, m512i, 256); /* { dg-error "the last argument must be an 8-bit immediate" } */ - m512i = _mm512_shuffle_epi32 (m512i, 256); /* { dg-error "the last argument must be an 8-bit immediate" } */ m512i = _mm512_mask_shuffle_epi32 (m512i, mmask16, m512i, 256); /* { dg-error "the last argument must be an 8-bit immediate" } */ m512i = _mm512_maskz_shuffle_epi32 (mmask16, m512i, 256); /* { dg-error "the last argument must be an 8-bit immediate" } */