Patch Detail
get:
Show a patch.
patch:
Update a patch.
put:
Update a patch.
GET /api/1.2/patches/2224536/?format=api
{ "id": 2224536, "url": "http://patchwork.ozlabs.org/api/1.2/patches/2224536/?format=api", "web_url": "http://patchwork.ozlabs.org/project/qemu-devel/patch/8ab885ce490cede31b8e9e7f6a67295c1aa353f7.1776437127.git.chao.liu.zevorn@gmail.com/", "project": { "id": 14, "url": "http://patchwork.ozlabs.org/api/1.2/projects/14/?format=api", "name": "QEMU Development", "link_name": "qemu-devel", "list_id": "qemu-devel.nongnu.org", "list_email": "qemu-devel@nongnu.org", "web_url": "", "scm_url": "", "webscm_url": "", "list_archive_url": "", "list_archive_url_format": "", "commit_url_format": "" }, "msgid": "<8ab885ce490cede31b8e9e7f6a67295c1aa353f7.1776437127.git.chao.liu.zevorn@gmail.com>", "list_archive_url": null, "date": "2026-04-17T14:49:31", "name": "[v10,1/2] target/riscv: Use tcg nodes for strided vector ld/st generation", "commit_ref": null, "pull_url": null, "state": "new", "archived": false, "hash": "ea645c1abf8e8d851cf3177a61da46060e3c9fca", "submitter": { "id": 92265, "url": "http://patchwork.ozlabs.org/api/1.2/people/92265/?format=api", "name": "Chao Liu", "email": "chao.liu.zevorn@gmail.com" }, "delegate": null, "mbox": "http://patchwork.ozlabs.org/project/qemu-devel/patch/8ab885ce490cede31b8e9e7f6a67295c1aa353f7.1776437127.git.chao.liu.zevorn@gmail.com/mbox/", "series": [ { "id": 500357, "url": "http://patchwork.ozlabs.org/api/1.2/series/500357/?format=api", "web_url": "http://patchwork.ozlabs.org/project/qemu-devel/list/?series=500357", "date": "2026-04-17T14:49:31", "name": "Generate strided vector loads/stores with tcg nodes", "version": 10, "mbox": "http://patchwork.ozlabs.org/series/500357/mbox/" } ], "comments": "http://patchwork.ozlabs.org/api/patches/2224536/comments/", "check": "pending", "checks": "http://patchwork.ozlabs.org/api/patches/2224536/checks/", "tags": {}, "related": [], "headers": { "Return-Path": "<qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org>", "X-Original-To": "incoming@patchwork.ozlabs.org", "Delivered-To": "patchwork-incoming@legolas.ozlabs.org", "Authentication-Results": [ "legolas.ozlabs.org;\n\tdkim=pass (2048-bit key;\n unprotected) header.d=gmail.com header.i=@gmail.com header.a=rsa-sha256\n header.s=20251104 header.b=RQ+vtk0S;\n\tdkim-atps=neutral", "legolas.ozlabs.org;\n spf=pass (sender SPF authorized) smtp.mailfrom=nongnu.org\n (client-ip=209.51.188.17; helo=lists1p.gnu.org;\n envelope-from=qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org;\n receiver=patchwork.ozlabs.org)" ], "Received": [ "from lists1p.gnu.org (lists1p.gnu.org [209.51.188.17])\n\t(using TLSv1.2 with cipher ECDHE-ECDSA-AES256-GCM-SHA384 (256/256 bits))\n\t(No client certificate requested)\n\tby legolas.ozlabs.org (Postfix) with ESMTPS id 4fxyVL0Z64z1yDF\n\tfor <incoming@patchwork.ozlabs.org>; Sat, 18 Apr 2026 00:50:34 +1000 (AEST)", "from localhost ([::1] helo=lists1p.gnu.org)\n\tby lists1p.gnu.org with esmtp (Exim 4.90_1)\n\t(envelope-from <qemu-devel-bounces@nongnu.org>)\n\tid 1wDkWB-000693-VG; Fri, 17 Apr 2026 10:50:04 -0400", "from eggs.gnu.org ([2001:470:142:3::10])\n by lists1p.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256)\n (Exim 4.90_1) (envelope-from <chao.liu.zevorn@gmail.com>)\n id 1wDkWA-00065d-0g\n for qemu-devel@nongnu.org; Fri, 17 Apr 2026 10:50:02 -0400", "from mail-vk1-xa41.google.com ([2607:f8b0:4864:20::a41])\n by eggs.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_128_GCM_SHA256:128)\n (Exim 4.90_1) (envelope-from <chao.liu.zevorn@gmail.com>)\n id 1wDkW7-0007K7-GC\n for qemu-devel@nongnu.org; Fri, 17 Apr 2026 10:50:01 -0400", "by mail-vk1-xa41.google.com with SMTP id\n 71dfb90a1353d-56faf1cfe04so75970e0c.3\n for <qemu-devel@nongnu.org>; Fri, 17 Apr 2026 07:49:59 -0700 (PDT)", "from ZEVORN-PC.bbrouter ([162.244.208.119])\n by smtp.gmail.com with ESMTPSA id\n 71dfb90a1353d-56fa92f21bbsm954365e0c.11.2026.04.17.07.49.52\n (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256);\n Fri, 17 Apr 2026 07:49:57 -0700 (PDT)" ], "DKIM-Signature": "v=1; a=rsa-sha256; c=relaxed/relaxed;\n d=gmail.com; s=20251104; t=1776437398; x=1777042198; darn=nongnu.org;\n h=content-transfer-encoding:mime-version:references:in-reply-to\n :message-id:date:subject:cc:to:from:from:to:cc:subject:date\n :message-id:reply-to;\n bh=Re5DyUwcF8JZxjBhQgkUafRPBAnFADGG1yW9O0XHGKI=;\n b=RQ+vtk0SSDnqSCsp/CN62o7VNiHuiD9I8dQRrADc2ML45llhM/MbDMVEz93c5tWrIC\n gKxCfMDpzcecSnhhWuzq/HWFf1YLkwDHojqbXPekI3FL6iLCmVXAaxeazx1MiDrj0tFJ\n SsCYaNATRJgzVuemGoY3PUMCeNpxeigsY+nm+9kUSJo9JU3ucgZi6Wln7t9wFYzrE2Um\n zYnB3G8nhuLJMLtUKfCP4uBh0wzHXYQXiQQhruMQUI/t8OCY2M+n0FSp5KRMfz2mQP58\n iaqmRbUzrTprd+yF05f8CdrSwTm9X+2WhJQTEUYtTVuNfd5ZkuOasTEV8kzZiTjSG4Q7\n JOqQ==", "X-Google-DKIM-Signature": "v=1; a=rsa-sha256; c=relaxed/relaxed;\n d=1e100.net; s=20251104; t=1776437398; x=1777042198;\n h=content-transfer-encoding:mime-version:references:in-reply-to\n :message-id:date:subject:cc:to:from:x-gm-gg:x-gm-message-state:from\n :to:cc:subject:date:message-id:reply-to;\n bh=Re5DyUwcF8JZxjBhQgkUafRPBAnFADGG1yW9O0XHGKI=;\n b=s+jnrfZ/21cPDLVpKBd4tTKtrVFwfnCQgl9yohEWcNbRM2WdnhSaAkC5gioUXGiiXv\n VCLkPmF1X6ZKPVIKAJ+6zL8XUtUCIq+/Lkx9DZM1WbdI2xdoAmpRKM8fS/56E685r54e\n a5i/jGnJoKkX/9Br9CvlyEDqdsaQiNVRjd7mZre/F+J9I03ZDowWvkndX9oUWYh8BJV/\n eJ1TM6LpFj/fkLSzxPyI3hw5jSdOtRTKoV71OXRuwTSQ9HZQFSyCHQw4fgRMRVgQDKxa\n N1NBe2DadUauxhU0W6bbZkJvn/KqT+OFogPNYDsJeFmBoJLA5OejEylP7KgQ+uiSvT72\n mILQ==", "X-Gm-Message-State": "AOJu0YyB1l4itAEqX5tSGkXlGyieZZkjPdXVoJG8E6/O3brQcPwH9YlC\n VS805BiipQzoLH5x5+nudUGGKgAD25+3r0imIL95SiOm564/rOFp4lbo", "X-Gm-Gg": "AeBDieuHJC8tJH3nGEOL6F7zaf7axJFgJa2x8Ds0JiE6jEH+BGIGdadYpMMvlBdet3H\n o0aH77Ymk0Sqn/Qa0QmyM8BY1Su9i8kXCw1avD5KgWlwaHUdqPMBp1O31H7Ct3hx2CAelTwVPvD\n s+WCwGmE9oNoxrfiEo/Uf+txnng+9KP2torIiRa56CzZfpprZzRqUMx+atJxjcHaBCSH0B2xv2a\n h1HJP1KXfcqrju0LYk5NkMBh3HyF5+Aecg22aPBks6Qbi5NZtXdVH/H03AuXRyYFCq0/NepjrnC\n uUkcvYlwIqVIeG9RBUpAiR+EQES8+v6yjXykbb1aIJW5WvphItSX18sQWy4bsWttYmZgx5gYG/Q\n P9yexIT7jmChubRxHEfV+/2N22IKIPvklcXuR1UcCdO9az2OcHYHCBoj1iT6rOQ0L/A8HcqqB20\n YhvUQg/NBrcFy/4BXTYcus1pLotOPeTcbjOIcq+N8i7dYh2c9gU4ejaQw8KNWZRwtFVyN0eRFhv\n g5u9rw=", "X-Received": "by 2002:a05:6122:3b0d:b0:56f:7300:37c9 with SMTP id\n 71dfb90a1353d-56fa5996dc2mr1325521e0c.10.1776437398232;\n Fri, 17 Apr 2026 07:49:58 -0700 (PDT)", "From": "Chao Liu <chao.liu.zevorn@gmail.com>", "To": "Palmer Dabbelt <palmer@dabbelt.com>,\n Alistair Francis <alistair.francis@wdc.com>,\n Weiwei Li <liwei1518@gmail.com>,\n Daniel Henrique Barboza <daniel.barboza@oss.qualcomm.com>,\n Liu Zhiwei <zhiwei_liu@linux.alibaba.com>,\n Chao Liu <chao.liu.zevorn@gmail.com>", "Cc": "qemu-devel@nongnu.org, qemu-riscv@nongnu.org,\n Paolo Savini <paolo.savini@embecosm.com>,\n Nicholas Piggin <npiggin@gmail.com>, Eric Biggers <ebiggers@kernel.org>,\n Daniel Henrique Barboza <dbarboza@ventanamicro.com>", "Subject": "[PATCH v10 1/2] target/riscv: Use tcg nodes for strided vector ld/st\n generation", "Date": "Fri, 17 Apr 2026 22:49:31 +0800", "Message-ID": "\n <8ab885ce490cede31b8e9e7f6a67295c1aa353f7.1776437127.git.chao.liu.zevorn@gmail.com>", "X-Mailer": "git-send-email 2.53.0", "In-Reply-To": "<cover.1776437127.git.chao.liu.zevorn@gmail.com>", "References": "<cover.1776437127.git.chao.liu.zevorn@gmail.com>", "MIME-Version": "1.0", "Content-Transfer-Encoding": "8bit", "Received-SPF": "pass client-ip=2607:f8b0:4864:20::a41;\n envelope-from=chao.liu.zevorn@gmail.com; helo=mail-vk1-xa41.google.com", "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,\n DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, FREEMAIL_FROM=0.001,\n RCVD_IN_DNSWL_NONE=-0.0001, SPF_HELO_NONE=0.001,\n SPF_PASS=-0.001 autolearn=unavailable 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 <qemu-devel.nongnu.org>", "List-Unsubscribe": "<https://lists.nongnu.org/mailman/options/qemu-devel>,\n <mailto:qemu-devel-request@nongnu.org?subject=unsubscribe>", "List-Archive": "<https://lists.nongnu.org/archive/html/qemu-devel>", "List-Post": "<mailto:qemu-devel@nongnu.org>", "List-Help": "<mailto:qemu-devel-request@nongnu.org?subject=help>", "List-Subscribe": "<https://lists.nongnu.org/mailman/listinfo/qemu-devel>,\n <mailto:qemu-devel-request@nongnu.org?subject=subscribe>", "Errors-To": "qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org", "Sender": "qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org" }, "content": "This commit improves the performance of QEMU when emulating strided vector\nloads and stores by substituting the call for the helper function with the\ngeneration of equivalent TCG operations.\n\nPS:\n\nAn implementation is permitted to cause an illegal instruction if vstart\nis not 0 and it is set to a value that can not be produced implicitly by\nthe implementation, but memory accesses will generally always need to\ndeal with page faults.\n\nSo, if a strided vector memory access instruction has non-zero vstart,\ncheck it through vlse/vsse helpers function.\n\nSigned-off-by: Paolo Savini <paolo.savini@embecosm.com>\nSigned-off-by: Chao Liu <chao.liu.zevorn@gmail.com>\nSigned-off-by: Nicholas Piggin <npiggin@gmail.com>\nTested-by: Eric Biggers <ebiggers@kernel.org>\nReviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>\n---\n target/riscv/insn_trans/trans_rvv.c.inc | 354 ++++++++++++++++++++++--\n 1 file changed, 337 insertions(+), 17 deletions(-)", "diff": "diff --git a/target/riscv/insn_trans/trans_rvv.c.inc b/target/riscv/insn_trans/trans_rvv.c.inc\nindex 5b72926b3c..4cf3643cbd 100644\n--- a/target/riscv/insn_trans/trans_rvv.c.inc\n+++ b/target/riscv/insn_trans/trans_rvv.c.inc\n@@ -16,6 +16,7 @@\n */\n #include \"tcg/tcg-op-gvec.h\"\n #include \"tcg/tcg-gvec-desc.h\"\n+#include \"tcg/tcg-temp-internal.h\"\n #include \"internals.h\"\n \n static inline bool is_overlapped(const int8_t astart, int8_t asize,\n@@ -872,15 +873,290 @@ static bool st_us_mask_check(DisasContext *s, arg_vsm_v *a, uint8_t eew)\n GEN_VEXT_TRANS(vlm_v, MO_8, vlm_v, ld_us_mask_op, ld_us_mask_check)\n GEN_VEXT_TRANS(vsm_v, MO_8, vsm_v, st_us_mask_op, st_us_mask_check)\n \n+/*\n+ * MAXSZ returns the maximum vector size can be operated in bytes,\n+ * which is used in GVEC IR when vl_eq_vlmax flag is set to true\n+ * to accelerate vector operation.\n+ */\n+static inline uint32_t MAXSZ(DisasContext *s)\n+{\n+ int max_sz = s->cfg_ptr->vlenb << 3;\n+ return max_sz >> (3 - s->lmul);\n+}\n+\n+static inline uint32_t get_log2(uint32_t a)\n+{\n+ assert(is_power_of_2(a));\n+ return ctz32(a);\n+}\n+\n+typedef void gen_tl_ldst(TCGv, TCGv_ptr, tcg_target_long);\n+\n+static void gen_ldst_vreg(DisasContext *s, TCGv_i64 dest_offs, TCGv_i64 addr,\n+ gen_tl_ldst *ld_fn, gen_tl_ldst *st_fn, bool is_load)\n+{\n+ MemOp atomicity = (s->sew == 0) ? MO_ATOM_NONE : MO_ATOM_IFALIGN_PAIR;\n+ TCGv_ptr dest_ptr = tcg_temp_new_ptr();\n+ TCGv_i64 vreg = tcg_temp_new_i64();\n+ tcg_gen_trunc_i64_ptr(dest_ptr, dest_offs);\n+\n+ if (is_load) {\n+ tcg_gen_qemu_ld_tl(vreg, addr, s->mem_idx, MO_LE | s->sew | atomicity);\n+ st_fn(vreg, dest_ptr, 0);\n+ } else {\n+ ld_fn(vreg, dest_ptr, 0);\n+ tcg_gen_qemu_st_tl(vreg, addr, s->mem_idx, MO_LE | s->sew | atomicity);\n+ }\n+ tcg_temp_free_ptr(dest_ptr);\n+ tcg_temp_free_i64(vreg);\n+}\n+\n+/*\n+ * Check whether the i bit of the mask is 0 or 1.\n+ *\n+ * static inline int vext_elem_mask(void *v0, int index)\n+ * {\n+ * int idx = index / 64;\n+ * int pos = index % 64;\n+ * return (((uint64_t *)v0)[idx] >> pos) & 1;\n+ * }\n+ *\n+ * And\n+ *\n+ * if (vext_elem_mask(v0, i) != 0) {\n+ * goto label;\n+ * }\n+ */\n+static void gen_check_vext_elem_mask(DisasContext *s, TCGLabel *label,\n+ TCGv_i64 mask_offs)\n+{\n+ TCGv_i64 temp = tcg_temp_new_i64();\n+ TCGv_ptr ptr = tcg_temp_new_ptr();\n+ TCGv_i64 elem = tcg_temp_new_i64();\n+\n+ tcg_gen_shri_tl(temp, mask_offs, 3);\n+ tcg_gen_trunc_i64_ptr(ptr, temp);\n+ tcg_gen_add_ptr(ptr, ptr, tcg_env);\n+\n+ tcg_gen_ld8u_i64(elem, ptr, 0);\n+ tcg_gen_andi_tl(temp, mask_offs, 7);\n+ tcg_gen_shr_tl(elem, elem, temp);\n+ tcg_gen_brcond_i64(TCG_COND_TSTNE, elem, tcg_constant_i64(1), label);\n+\n+ tcg_temp_free_i64(temp);\n+ tcg_temp_free_ptr(ptr);\n+ tcg_temp_free_i64(elem);\n+}\n+\n+static void gen_vext_set_elems_1s(TCGv dest, TCGv_i64 mask_offs, int sew,\n+ gen_tl_ldst *st_fn, bool is_load)\n+{\n+ if (is_load) {\n+ TCGv_ptr ptr = tcg_temp_new_ptr();\n+ tcg_gen_shli_tl(mask_offs, mask_offs, sew);\n+ tcg_gen_add_tl(mask_offs, mask_offs, dest);\n+ tcg_gen_trunc_i64_ptr(ptr, mask_offs);\n+ st_fn(tcg_constant_tl(-1), ptr, 0);\n+ tcg_temp_free_ptr(ptr);\n+ }\n+}\n+\n+/*\n+ * Simulate the strided load/store main loop:\n+ *\n+ * for (i = env->vstart; i < env->vl; env->vstart = ++i) {\n+ * k = 0;\n+ * while (k < nf) {\n+ * if (!vm && !vext_elem_mask(v0, i)) {\n+ * vext_set_elems_1s(vd, vma, (i + k * max_elems) * esz,\n+ * (i + k * max_elems + 1) * esz);\n+ * k++;\n+ * continue;\n+ * }\n+ * target_ulong addr = base + stride * i + (k << log2_esz);\n+ * ldst(env, adjust_addr(env, addr), i + k * max_elems, vd, ra);\n+ * k++;\n+ * }\n+ * }\n+ */\n+static void gen_ldst_stride_main_loop(DisasContext *s, TCGv dest, uint32_t rs1,\n+ uint32_t rs2, uint32_t vm, uint32_t nf,\n+ gen_tl_ldst *ld_fn, gen_tl_ldst *st_fn,\n+ bool is_load)\n+{\n+ TCGv_i64 addr = tcg_temp_new_i64();\n+ TCGv base = get_gpr(s, rs1, EXT_NONE);\n+ TCGv stride = get_gpr(s, rs2, EXT_NONE);\n+\n+ TCGv i = tcg_temp_new();\n+ TCGv i_esz = tcg_temp_new();\n+ TCGv k = tcg_temp_new();\n+ TCGv k_esz = tcg_temp_new();\n+ TCGv k_max = tcg_temp_new();\n+ TCGv_i64 mask_offs = tcg_temp_new_i64();\n+ TCGv_i64 dest_offs = tcg_temp_new_i64();\n+ TCGv_i64 stride_offs = tcg_temp_new_i64();\n+\n+ uint32_t max_elems = MAXSZ(s) >> s->sew;\n+\n+ TCGLabel *start = gen_new_label();\n+ TCGLabel *end = gen_new_label();\n+ TCGLabel *start_k = gen_new_label();\n+ TCGLabel *inc_k = gen_new_label();\n+ TCGLabel *end_k = gen_new_label();\n+\n+ /* Start of outer loop. */\n+ tcg_gen_mov_tl(i, cpu_vstart);\n+ gen_set_label(start);\n+ tcg_gen_brcond_tl(TCG_COND_GE, i, cpu_vl, end);\n+ tcg_gen_shli_tl(i_esz, i, s->sew);\n+\n+ /* Start of inner loop. */\n+ tcg_gen_movi_tl(k, 0);\n+ gen_set_label(start_k);\n+ tcg_gen_brcond_tl(TCG_COND_GE, k, tcg_constant_tl(nf), end_k);\n+\n+ /*\n+ * If we are in mask agnostic regime and the operation is not unmasked we\n+ * set the inactive elements to 1.\n+ */\n+ if (!vm && s->vma) {\n+ TCGLabel *active_element = gen_new_label();\n+ /* (i + k * max_elems) * esz */\n+ tcg_gen_shli_tl(mask_offs, k, get_log2(max_elems << s->sew));\n+ tcg_gen_add_tl(mask_offs, mask_offs, i_esz);\n+\n+ /*\n+ * Check whether the i bit of the mask is 0 or 1.\n+ * If it is 0, set masked-off elements;\n+ * otherwise, directly load/store the vector register.\n+ */\n+ gen_check_vext_elem_mask(s, active_element, mask_offs);\n+\n+ /*\n+ * Set masked-off elements in the destination vector register to 1s.\n+ * Store instructions simply skip this bit as memory ops access memory\n+ * only for active elements.\n+ */\n+ gen_vext_set_elems_1s(dest, mask_offs, s->sew, st_fn, is_load);\n+\n+ tcg_gen_br(inc_k);\n+ gen_set_label(active_element);\n+ }\n+\n+ /*\n+ * The element is active, calculate the address with stride:\n+ * target_ulong addr = base + stride * i + (k << log2_esz);\n+ */\n+ tcg_gen_mul_tl(stride_offs, stride, i);\n+ tcg_gen_shli_tl(k_esz, k, s->sew);\n+ tcg_gen_add_tl(stride_offs, stride_offs, k_esz);\n+ tcg_gen_add_tl(addr, base, stride_offs);\n+\n+ /* Calculate the offset in the dst/src vector register. */\n+ tcg_gen_shli_tl(k_max, k, get_log2(max_elems));\n+ tcg_gen_add_tl(dest_offs, i, k_max);\n+ tcg_gen_shli_tl(dest_offs, dest_offs, s->sew);\n+ tcg_gen_add_tl(dest_offs, dest_offs, dest);\n+\n+ /* Load/Store vector register. */\n+ gen_ldst_vreg(s, dest_offs, addr, ld_fn, st_fn, is_load);\n+\n+ /*\n+ * We don't execute the load/store above if the element was inactive.\n+ * We jump instead directly to incrementing k and continuing the loop.\n+ */\n+ if (!vm && s->vma) {\n+ gen_set_label(inc_k);\n+ }\n+ tcg_gen_addi_tl(k, k, 1);\n+ tcg_gen_br(start_k);\n+\n+ /* End of the inner loop. */\n+ gen_set_label(end_k);\n+\n+ tcg_gen_addi_tl(i, i, 1);\n+ tcg_gen_mov_tl(cpu_vstart, i);\n+ tcg_gen_br(start);\n+\n+ /* End of the outer loop. */\n+ gen_set_label(end);\n+\n+ return;\n+}\n+\n+/*\n+ * Set the tail bytes of the strided loads/stores to 1:\n+ *\n+ * for (k = 0; k < nf; ++k) {\n+ * cnt = (k * max_elems + vl) * esz;\n+ * tot = (k * max_elems + max_elems) * esz;\n+ * for (i = cnt; i < tot; i += esz) {\n+ * store_1s(-1, vd[vl+i]);\n+ * }\n+ * }\n+ */\n+static void gen_ldst_stride_tail_loop(DisasContext *s, TCGv dest, uint32_t nf,\n+ gen_tl_ldst *st_fn)\n+{\n+ TCGv i = tcg_temp_new();\n+ TCGv k = tcg_temp_new();\n+ TCGv tail_cnt = tcg_temp_new();\n+ TCGv tail_tot = tcg_temp_new();\n+ TCGv tail_addr = tcg_temp_new();\n+\n+ TCGLabel *start = gen_new_label();\n+ TCGLabel *end = gen_new_label();\n+ TCGLabel *start_i = gen_new_label();\n+ TCGLabel *end_i = gen_new_label();\n+\n+ uint32_t max_elems_b = MAXSZ(s);\n+ uint32_t esz = 1 << s->sew;\n+\n+ /* Start of the outer loop. */\n+ tcg_gen_movi_tl(k, 0);\n+ tcg_gen_shli_tl(tail_cnt, cpu_vl, s->sew);\n+ tcg_gen_movi_tl(tail_tot, max_elems_b);\n+ tcg_gen_add_tl(tail_addr, dest, tail_cnt);\n+ gen_set_label(start);\n+ tcg_gen_brcond_tl(TCG_COND_GE, k, tcg_constant_tl(nf), end);\n+\n+ /* Start of the inner loop. */\n+ tcg_gen_mov_tl(i, tail_cnt);\n+ gen_set_label(start_i);\n+ tcg_gen_brcond_tl(TCG_COND_GE, i, tail_tot, end_i);\n+\n+ /* store_1s(-1, vd[vl+i]); */\n+ st_fn(tcg_constant_tl(-1), (TCGv_ptr)tail_addr, 0);\n+ tcg_gen_addi_tl(tail_addr, tail_addr, esz);\n+ tcg_gen_addi_tl(i, i, esz);\n+ tcg_gen_br(start_i);\n+\n+ /* End of the inner loop. */\n+ gen_set_label(end_i);\n+\n+ /* Update the counts */\n+ tcg_gen_addi_tl(tail_cnt, tail_cnt, max_elems_b);\n+ tcg_gen_addi_tl(tail_tot, tail_cnt, max_elems_b);\n+ tcg_gen_addi_tl(k, k, 1);\n+ tcg_gen_br(start);\n+\n+ /* End of the outer loop. */\n+ gen_set_label(end);\n+\n+ return;\n+}\n+\n /*\n *** stride load and store\n */\n typedef void gen_helper_ldst_stride(TCGv_ptr, TCGv_ptr, TCGv,\n TCGv, TCGv_env, TCGv_i32);\n \n-static bool ldst_stride_trans(uint32_t vd, uint32_t rs1, uint32_t rs2,\n- uint32_t data, gen_helper_ldst_stride *fn,\n- DisasContext *s)\n+static\n+bool gen_call_helper_ldst_stride(uint32_t vd, uint32_t rs1, uint32_t rs2,\n+ uint32_t data, gen_helper_ldst_stride *fn,\n+ DisasContext *s)\n {\n TCGv_ptr dest, mask;\n TCGv base, stride;\n@@ -904,11 +1180,66 @@ static bool ldst_stride_trans(uint32_t vd, uint32_t rs1, uint32_t rs2,\n return true;\n }\n \n+static bool ldst_stride_trans(uint32_t vd, uint32_t rs1, uint32_t rs2,\n+ uint32_t data, gen_helper_ldst_stride *fn,\n+ DisasContext *s, bool is_load)\n+{\n+ if (!s->vstart_eq_zero) {\n+ /* vstart != 0 helper slowpath */\n+ return gen_call_helper_ldst_stride(vd, rs1, rs2, data, fn, s);\n+ }\n+\n+ TCGv dest = tcg_temp_new();\n+\n+ uint32_t nf = FIELD_EX32(data, VDATA, NF);\n+ uint32_t vm = FIELD_EX32(data, VDATA, VM);\n+\n+ /* Destination register and mask register */\n+ tcg_gen_addi_tl(dest, (TCGv)tcg_env, vreg_ofs(s, vd));\n+\n+ /*\n+ * Select the appropriate load/store to retrieve data from the vector\n+ * register given a specific sew.\n+ */\n+ static gen_tl_ldst * const ld_fns[4] = {\n+ tcg_gen_ld8u_tl, tcg_gen_ld16u_tl,\n+ tcg_gen_ld32u_tl, tcg_gen_ld_tl\n+ };\n+\n+ static gen_tl_ldst * const st_fns[4] = {\n+ tcg_gen_st8_tl, tcg_gen_st16_tl,\n+ tcg_gen_st32_tl, tcg_gen_st_tl\n+ };\n+\n+ gen_tl_ldst *ld_fn = ld_fns[s->sew];\n+ gen_tl_ldst *st_fn = st_fns[s->sew];\n+\n+ if (ld_fn == NULL || st_fn == NULL) {\n+ return false;\n+ }\n+\n+ mark_vs_dirty(s);\n+\n+ gen_ldst_stride_main_loop(s, dest, rs1, rs2, vm, nf, ld_fn, st_fn, is_load);\n+\n+ tcg_gen_movi_tl(cpu_vstart, 0);\n+\n+ /*\n+ * Set the tail bytes to 1 if tail agnostic:\n+ */\n+ if (s->vta != 0 && is_load) {\n+ gen_ldst_stride_tail_loop(s, dest, nf, st_fn);\n+ }\n+\n+ finalize_rvv_inst(s);\n+ return true;\n+}\n+\n static bool ld_stride_op(DisasContext *s, arg_rnfvm *a, uint8_t eew)\n {\n uint32_t data = 0;\n gen_helper_ldst_stride *fn;\n- static gen_helper_ldst_stride * const fns[4] = {\n+ static gen_helper_ldst_stride *const fns[4] = {\n gen_helper_vlse8_v, gen_helper_vlse16_v,\n gen_helper_vlse32_v, gen_helper_vlse64_v\n };\n@@ -924,7 +1255,7 @@ static bool ld_stride_op(DisasContext *s, arg_rnfvm *a, uint8_t eew)\n data = FIELD_DP32(data, VDATA, NF, a->nf);\n data = FIELD_DP32(data, VDATA, VTA, s->vta);\n data = FIELD_DP32(data, VDATA, VMA, s->vma);\n- return ldst_stride_trans(a->rd, a->rs1, a->rs2, data, fn, s);\n+ return ldst_stride_trans(a->rd, a->rs1, a->rs2, data, fn, s, true);\n }\n \n static bool ld_stride_check(DisasContext *s, arg_rnfvm* a, uint8_t eew)\n@@ -958,7 +1289,7 @@ static bool st_stride_op(DisasContext *s, arg_rnfvm *a, uint8_t eew)\n return false;\n }\n \n- return ldst_stride_trans(a->rd, a->rs1, a->rs2, data, fn, s);\n+ return ldst_stride_trans(a->rd, a->rs1, a->rs2, data, fn, s, false);\n }\n \n static bool st_stride_check(DisasContext *s, arg_rnfvm* a, uint8_t eew)\n@@ -1281,17 +1612,6 @@ GEN_LDST_WHOLE_TRANS(vs8r_v, int8_t, 8, false)\n *** Vector Integer Arithmetic Instructions\n */\n \n-/*\n- * MAXSZ returns the maximum vector size can be operated in bytes,\n- * which is used in GVEC IR when vl_eq_vlmax flag is set to true\n- * to accelerate vector operation.\n- */\n-static inline uint32_t MAXSZ(DisasContext *s)\n-{\n- int max_sz = s->cfg_ptr->vlenb * 8;\n- return max_sz >> (3 - s->lmul);\n-}\n-\n static bool opivv_check(DisasContext *s, arg_rmrr *a)\n {\n return require_rvv(s) &&\n", "prefixes": [ "v10", "1/2" ] }