Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from lists1p.gnu.org (lists1p.gnu.org [209.51.188.17]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.lore.kernel.org (Postfix) with ESMTPS id EC5D4C54FD2 for ; Thu, 30 Jul 2026 03:10:49 +0000 (UTC) Received: from localhost ([::1] helo=lists1p.gnu.org) by lists1p.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1wpH6Y-0002es-LM; Wed, 29 Jul 2026 23:06:42 -0400 Received: from eggs.gnu.org ([2001:470:142:3::10]) by lists1p.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1wpH6X-0002SO-0g for qemu-devel@nongnu.org; Wed, 29 Jul 2026 23:06:41 -0400 Received: from rev.ng ([94.130.142.21]) by eggs.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1wpH6U-0002vo-Tb for qemu-devel@nongnu.org; Wed, 29 Jul 2026 23:06:40 -0400 DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=rev.ng; s=dkim; h=Content-Transfer-Encoding:MIME-Version:References:In-Reply-To: Message-ID:Date:Subject:Cc:To:From:Sender:Reply-To:Content-Type:Content-ID: Content-Description:Resent-Date:Resent-From:Resent-Sender:Resent-To:Resent-Cc :Resent-Message-ID:List-Id:List-Help:List-Unsubscribe:List-Subscribe: List-Post:List-Owner:List-Archive:List-Unsubscribe:List-Unsubscribe-Post: List-Help; bh=LRMzawlJ3Kw59JwVhx4oKCorGEMhoXcYEID8YEaEEqs=; b=IiiIKoamNBqq1Jy KIwh6ayfWOPaEQWZ/QEgQB8DCQKuB/leIQRazrT0jg7KZky8kvWKWOxWtcdOFcE6unUARYJ8hUnE8 7rJGn23tvLOQbkRYGW+jcc6BhyPq0zBhiFUb5+3mCSiCDa/3WUh7mnaEsqUonbAxMVt59Q+VJ4Amp MI=; To: qemu-devel@nongnu.org Cc: ale@rev.ng, brian.cain@oss.qualcomm.com, pierrick.bouvier@oss.qualcomm.com, philmd@mailo.com, Anton Johansson Subject: [PATCH v2 21/50] helper-to-tcg: Add guest vector layout description Date: Thu, 30 Jul 2026 05:09:55 +0200 Message-ID: <20260730031025.12926-22-anjo@rev.ng> In-Reply-To: <20260730031025.12926-1-anjo@rev.ng> References: <20260730031025.12926-1-anjo@rev.ng> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Received-SPF: pass client-ip=94.130.142.21; envelope-from=anjo@rev.ng; helo=rev.ng X-Spam_score_int: -20 X-Spam_score: -2.1 X-Spam_bar: -- X-Spam_report: (-2.1 / 5.0 requ) BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, SPF_HELO_PASS=-0.001, SPF_PASS=-0.001 autolearn=ham autolearn_force=no X-Spam_action: no action X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: qemu development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Reply-to: Anton Johansson From: Anton Johansson via qemu development Errors-To: qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org Sender: qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org Vectors are represented as arrays of host-endian blocks at least 64 bits in size, the size of which is configurable. Lane 0 in each block may be placed either at the most- or least-significant bytes. This covers a wide range of QEMU targets' vector representations. Vector layout information is used later when mapping constant expressions such as constant vectors, and when emitting helper functions for performing size-changing operations. Signed-off-by: Anton Johansson --- .../helper-to-tcg/include/VectorLayout.hpp | 116 ++++++++++++++++++ 1 file changed, 116 insertions(+) create mode 100644 subprojects/helper-to-tcg/include/VectorLayout.hpp diff --git a/subprojects/helper-to-tcg/include/VectorLayout.hpp b/subprojects/helper-to-tcg/include/VectorLayout.hpp new file mode 100644 index 0000000000..e85f5029d5 --- /dev/null +++ b/subprojects/helper-to-tcg/include/VectorLayout.hpp @@ -0,0 +1,116 @@ +// +// Copyright(c) 2026 rev.ng Labs Srl. All Rights Reserved. +// +// This program is free software; you can redistribute it and/or modify +// it under the terms of the GNU General Public License as published by +// the Free Software Foundation; either version 2 of the License, or +// (at your option) any later version. +// +// This program is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +// GNU General Public License for more details. +// +// You should have received a copy of the GNU General Public License +// along with this program; if not, see . +// + +#pragma once + +#include +#include +#include + +// Vector Layout +// +// Vector layouts among QEMU guests is quite diverse, this header aims to be +// simple but still capture a large subset of targets. +// +// Guest vectors are assumed to be divided into host-endian blocks of a +// configurable size (>= 64-bit and power-of-two), with block 0 always being at +// the lowest address. Blocks in turn are divided into 64-bit columns which are +// used when interfacing with QEMUs `gvec` API. +// +// Below is a 32-byte <32 x i8> guest vector consisting of 16-byte blocks, +// 8-byte columns, and 1-byte lanes for each element in the vector. Elements +// count from 1-32 in hex to indicate their position in memory. Note this +// example is for a little-endian host with `LeastSignificant` lane preference +// meaning lane 0 is at the least significant bytes of each block. +// +// Guest vector <32 x i8> {1, 2, 3, ..., 32} +// +---------------------------------+---------------------------------+ +// | BLOCK 0 | BLOCK 1 | +// +----------------+----------------+----------------+----------------+ +// | COLUMN 0 | COLUMN 1 | COLUMN 2 | COLUMN 3 | +// |0102030405060708|090A0B0C0D0E0F10|1112131415161718|191A1B1C1D1E1F20| +// +// LOW ADDRESS HIGH ADDRESS +// +// On a big-endian host, the vector would instead be expressed in memory as +// +// Guest vector <32 x i8> {1, 2, 3, ..., 32} +// +---------------------------------+---------------------------------+ +// | BLOCK 0 | BLOCK 1 | +// +----------------+----------------+----------------+----------------+ +// | COLUMN 1 | COLUMN 0 | COLUMN 3 | COLUMN 2 | +// |100F0E0D0C0B0A09|0807060504030201|201F1E1D1C1B1A19|1817161514131211| +// +// LOW ADDRESS HIGH ADDRESS +// +// with the column order having shifted to make sure the host-endian blocks +// retain the same value. Blocks are also allowed to be equal or greater in +// size that the guest vector, at which point the entire memory view of the +// vector would reverse when going from a little to big-endian host. +// +// The lane preference may also be changed to `MostSignificant` which for the +// little-endian example places lane 0 and the highest address of block 0. +// Note, that `MostSignificant` on little-endian and `LeastSignificant` on +// big-endian hosts are equivalent in their memory representation, but will +// differ in how they are emitted in C: +// +// // MostSignificant, little endian +// uint64_t vec[] = {0x90A0B0C0D0E0F10, 0x102030405060708, ...}; +// +// // LeastSignificant, big endian +// uint64_t vec[] = {0x100F0E0D0C0B0A09, 0x807060504030201, ...}; +// + +enum LanePreference { + LeastSignificant, + MostSignificant, +}; + +struct VectorLayout { + bool HostBigEndian; + LanePreference Lane0; + size_t BlockBytes; + + inline size_t index(bool Reverse, size_t Count, size_t I, size_t J) const { + if (Reverse) { + return Count * I + (J ^ (Count - 1)); + } else { + return Count * I + J; + } + } + + // Return the index into a linear array of lane L in column C + inline size_t indexLane(size_t LanesPerColumn, size_t C, size_t L) const { + const bool Reverse = (Lane0 == MostSignificant); + return index(Reverse, LanesPerColumn, C, L); + } + + // Return the index into a linear array of + inline size_t indexBlock(size_t BlocksPerVec, size_t I) const { + const bool Reverse = (HostBigEndian ^ (Lane0 == MostSignificant)); + return index(Reverse, BlocksPerVec, 0, I); + } + + //inline uint64_t column(llvm::ArrayRef Lanes) { + // uint64_t Column = 0; + // const size_t LanesPerColumn = 64 / ElementSize; + // for (size_t L = 0; L < LanesPerColumn; ++L) { + // const size_t Index = VL.indexLane(LanesPerColumn, C, L); + // Column |= Ints[Index] << (ElementSize * L); + // } + //} +}; -- 2.52.0