From patchwork Thu Oct 11 20:51:50 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Richard Henderson X-Patchwork-Id: 148684 Delivered-To: patch@linaro.org Received: by 2002:a2e:8595:0:0:0:0:0 with SMTP id b21-v6csp2604812lji; Thu, 11 Oct 2018 14:04:07 -0700 (PDT) X-Google-Smtp-Source: ACcGV638+NnTKKZi++DSqDlLXYA2O6SK/cuu7IBn6ROo6+1fLYrPly1xFI20dYp9OpOgn58huyWj X-Received: by 2002:ae9:ec05:: with SMTP id h5-v6mr3205113qkg.32.1539291846934; Thu, 11 Oct 2018 14:04:06 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1539291846; cv=none; d=google.com; s=arc-20160816; b=ewM2MR9W4MZD6fl9FiXff4YVNgkQb2TV/xYss5Ugo7tlVjDVhTH8aChO10U2LxWAeQ ZL4J4sgyrtIg82gj0zlroQCEfVoCHnA2iUnDS556SWTeSfJvfI8cax3lpp/VU8+7tq6W Ry9HhbWlu5ws7g2LSRapNSVbBNnzloZxEdCowI7PAbGQV0Fzqxz//KKGZ18FY5wx+hna 6hsMaaqfkS0S5w5YEQ4eOTQ1h4EKIzaVPmgdOc00XrYJklc6gGw7j5jmNFvNV7LoYg44 05RwG71TC7J4IbKUSrW/BX51Yx+l58CwqC9fOD/ISnEcH+P74oEdLlPJG4BDp/AK8Nh0 cPmg== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:cc:list-subscribe:list-help:list-post:list-archive :list-unsubscribe:list-id:precedence:subject:references:in-reply-to :message-id:date:to:from:dkim-signature; bh=Smv7/EDiE77NiuFaVj/NIlvKHR40VLp0vXf5RsdOZ4o=; b=jNt2SutE6GST9nbFISUhzuqL3rx4MI/FZU+oEpAQgww49iz7w8AcC/SHQ71TcEYe3u 5OOSJFSbeZxQp87g13jnrSs+zzT433C17cyWhYy7iVH6eOaSxlekA6KIJa6fvsdBl1AI s8C986QQoQHzIQsOkUbzFPujAfqhVfry4hzgxDfl2EfOJO6wdKQeq6ZMy6nDxPtz2/c+ jVaJMIY1Bdzh+HM6/D+hrDYGznsBdGO95Y0JpSStXEFTKijNBtGHzFDCQL1HH5GHqVF9 nxPr2OxfYGErwSwD8Z6BalGMAsdpOxCPyjF2TONSDkJo+mqY69frxMNPzBIWcfizbWOT itVA== ARC-Authentication-Results: i=1; mx.google.com; dkim=fail header.i=@linaro.org header.s=google header.b=ArsT5d70; spf=pass (google.com: domain of qemu-devel-bounces+patch=linaro.org@nongnu.org designates 2001:4830:134:3::11 as permitted sender) smtp.mailfrom="qemu-devel-bounces+patch=linaro.org@nongnu.org"; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=linaro.org Return-Path: Received: from lists.gnu.org (lists.gnu.org. [2001:4830:134:3::11]) by mx.google.com with ESMTPS id v12-v6si9763452qkf.160.2018.10.11.14.04.06 for (version=TLS1 cipher=AES128-SHA bits=128/128); Thu, 11 Oct 2018 14:04:06 -0700 (PDT) Received-SPF: pass (google.com: domain of qemu-devel-bounces+patch=linaro.org@nongnu.org designates 2001:4830:134:3::11 as permitted sender) client-ip=2001:4830:134:3::11; Authentication-Results: mx.google.com; dkim=fail header.i=@linaro.org header.s=google header.b=ArsT5d70; spf=pass (google.com: domain of qemu-devel-bounces+patch=linaro.org@nongnu.org designates 2001:4830:134:3::11 as permitted sender) smtp.mailfrom="qemu-devel-bounces+patch=linaro.org@nongnu.org"; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=linaro.org Received: from localhost ([::1]:37129 helo=lists.gnu.org) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1gAi7q-0007CC-AS for patch@linaro.org; Thu, 11 Oct 2018 17:04:06 -0400 Received: from eggs.gnu.org ([2001:4830:134:3::10]:45197) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1gAhwq-0006sC-2O for qemu-devel@nongnu.org; Thu, 11 Oct 2018 16:52:48 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1gAhwm-0005zz-72 for qemu-devel@nongnu.org; Thu, 11 Oct 2018 16:52:43 -0400 Received: from mail-pf1-x443.google.com ([2607:f8b0:4864:20::443]:40800) by eggs.gnu.org with esmtps (TLS1.0:RSA_AES_128_CBC_SHA1:16) (Exim 4.71) (envelope-from ) id 1gAhwl-0003SW-Mi for qemu-devel@nongnu.org; Thu, 11 Oct 2018 16:52:39 -0400 Received: by mail-pf1-x443.google.com with SMTP id s5-v6so5005658pfj.7 for ; Thu, 11 Oct 2018 13:52:15 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=linaro.org; s=google; h=from:to:cc:subject:date:message-id:in-reply-to:references; bh=Smv7/EDiE77NiuFaVj/NIlvKHR40VLp0vXf5RsdOZ4o=; b=ArsT5d70H3y9YxQJvWJMKD5lyxLZBdthWvekNUo3yoUPzC7GnsoIjhE6Gz+w4gXVIh 5QAOtiTHlz49Yzpm4PiUCYGVyFJ67NWOpMtxcITgX8hUZ+g+rSBss64JGm6KyeS3VRs+ wjWd3rKX23RkzDzQAjjQYjspqK+Hnycate/Y8= X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references; bh=Smv7/EDiE77NiuFaVj/NIlvKHR40VLp0vXf5RsdOZ4o=; b=snMNOb4F9Q9AhojIPCUzaQiQSQaRlYMvP77xPHT9m1rZqzJwghkGsWrXh7pghY87gK +SDcynnawSBeILnK6nDfVzv6ZGZfDztfpFuzJJJz8XDImtJ+ov/dXExGa++QnV0GpfL3 atcboTqI646Wjb+FC+KICfxrwO1AnZWXXYl4U/aGXgAVfFB8ujeZvQy9Ksq/jhimyObD 4iQW4of7xGg/g+vQnRTRz43zEpPN8m4MkE2X0MZVqnfqo4Wbv0Pf3WbWbvYs4e2lnsUV xe4evY36xgn0Anfn0VN9JrV5MTdW40rOmRYegHf+/vya5bvZDuPTcvK95qbmsVK1nAoB VwbQ== X-Gm-Message-State: ABuFfoh7SOSSSz0DwmBEkpDW+qlMGXYuq5RuMjqYDhMfQ+9VEJNfHiHu 6uKW8RJInbDSgZZ2hI7qeElWOmg2VmE= X-Received: by 2002:aa7:86cc:: with SMTP id h12-v6mr3054589pfo.58.1539291134647; Thu, 11 Oct 2018 13:52:14 -0700 (PDT) Received: from cloudburst.twiddle.net (97-113-8-179.tukw.qwest.net. [97.113.8.179]) by smtp.gmail.com with ESMTPSA id h87-v6sm34707866pfj.78.2018.10.11.13.52.13 (version=TLS1_2 cipher=ECDHE-RSA-CHACHA20-POLY1305 bits=256/256); Thu, 11 Oct 2018 13:52:13 -0700 (PDT) From: Richard Henderson To: qemu-devel@nongnu.org Date: Thu, 11 Oct 2018 13:51:50 -0700 Message-Id: <20181011205206.3552-5-richard.henderson@linaro.org> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20181011205206.3552-1-richard.henderson@linaro.org> References: <20181011205206.3552-1-richard.henderson@linaro.org> X-detected-operating-system: by eggs.gnu.org: Genre and OS details not recognized. X-Received-From: 2607:f8b0:4864:20::443 Subject: [Qemu-devel] [PATCH 04/20] target/arm: Promote consecutive memory ops for aa64 X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.21 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: peter.maydell@linaro.org Errors-To: qemu-devel-bounces+patch=linaro.org@nongnu.org Sender: "Qemu-devel" For a sequence of loads or stores from a single register, little-endian operations can be promoted to an 8-byte op. This can reduce the number of operations by a factor of 8. Signed-off-by: Richard Henderson --- target/arm/translate-a64.c | 66 +++++++++++++++++++++++--------------- 1 file changed, 40 insertions(+), 26 deletions(-) -- 2.17.1 diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index fff99ca303..2f4041462e 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -1200,25 +1200,23 @@ static void write_vec_element_i32(DisasContext *s, TCGv_i32 tcg_src, /* Store from vector register to memory */ static void do_vec_st(DisasContext *s, int srcidx, int element, - TCGv_i64 tcg_addr, int size) + TCGv_i64 tcg_addr, int size, TCGMemOp endian) { - TCGMemOp memop = s->be_data + size; TCGv_i64 tcg_tmp = tcg_temp_new_i64(); read_vec_element(s, tcg_tmp, srcidx, element, size); - tcg_gen_qemu_st_i64(tcg_tmp, tcg_addr, get_mem_index(s), memop); + tcg_gen_qemu_st_i64(tcg_tmp, tcg_addr, get_mem_index(s), endian | size); tcg_temp_free_i64(tcg_tmp); } /* Load from memory to vector register */ static void do_vec_ld(DisasContext *s, int destidx, int element, - TCGv_i64 tcg_addr, int size) + TCGv_i64 tcg_addr, int size, TCGMemOp endian) { - TCGMemOp memop = s->be_data + size; TCGv_i64 tcg_tmp = tcg_temp_new_i64(); - tcg_gen_qemu_ld_i64(tcg_tmp, tcg_addr, get_mem_index(s), memop); + tcg_gen_qemu_ld_i64(tcg_tmp, tcg_addr, get_mem_index(s), endian | size); write_vec_element(s, tcg_tmp, destidx, element, size); tcg_temp_free_i64(tcg_tmp); @@ -3013,9 +3011,10 @@ static void disas_ldst_multiple_struct(DisasContext *s, uint32_t insn) bool is_postidx = extract32(insn, 23, 1); bool is_q = extract32(insn, 30, 1); TCGv_i64 tcg_addr, tcg_rn, tcg_ebytes; + TCGMemOp endian = s->be_data; - int ebytes = 1 << size; - int elements = (is_q ? 128 : 64) / (8 << size); + int ebytes; /* bytes per element */ + int elements; /* elements per vector */ int rpt; /* num iterations */ int selem; /* structure elements */ int r; @@ -3074,6 +3073,20 @@ static void disas_ldst_multiple_struct(DisasContext *s, uint32_t insn) gen_check_sp_alignment(s); } + /* For our purposes, bytes are always little-endian. */ + if (size == 0) { + endian = MO_LE; + } + + /* Consecutive little-endian elements from a single register + * can be promoted to a larger little-endian operation. + */ + if (selem == 1 && endian == MO_LE) { + size = 3; + } + ebytes = 1 << size; + elements = (is_q ? 16 : 8) / ebytes; + tcg_rn = cpu_reg_sp(s, rn); tcg_addr = tcg_temp_new_i64(); tcg_gen_mov_i64(tcg_addr, tcg_rn); @@ -3082,32 +3095,33 @@ static void disas_ldst_multiple_struct(DisasContext *s, uint32_t insn) for (r = 0; r < rpt; r++) { int e; for (e = 0; e < elements; e++) { - int tt = (rt + r) % 32; int xs; for (xs = 0; xs < selem; xs++) { + int tt = (rt + r + xs) % 32; if (is_store) { - do_vec_st(s, tt, e, tcg_addr, size); + do_vec_st(s, tt, e, tcg_addr, size, endian); } else { - do_vec_ld(s, tt, e, tcg_addr, size); - - /* For non-quad operations, setting a slice of the low - * 64 bits of the register clears the high 64 bits (in - * the ARM ARM pseudocode this is implicit in the fact - * that 'rval' is a 64 bit wide variable). - * For quad operations, we might still need to zero the - * high bits of SVE. We optimize by noticing that we only - * need to do this the first time we touch a register. - */ - if (e == 0 && (r == 0 || xs == selem - 1)) { - clear_vec_high(s, is_q, tt); - } + do_vec_ld(s, tt, e, tcg_addr, size, endian); } tcg_gen_add_i64(tcg_addr, tcg_addr, tcg_ebytes); - tt = (tt + 1) % 32; } } } + if (!is_store) { + /* For non-quad operations, setting a slice of the low + * 64 bits of the register clears the high 64 bits (in + * the ARM ARM pseudocode this is implicit in the fact + * that 'rval' is a 64 bit wide variable). + * For quad operations, we might still need to zero the + * high bits of SVE. + */ + for (r = 0; r < rpt * selem; r++) { + int tt = (rt + r) % 32; + clear_vec_high(s, is_q, tt); + } + } + if (is_postidx) { int rm = extract32(insn, 16, 5); if (rm == 31) { @@ -3228,9 +3242,9 @@ static void disas_ldst_single_struct(DisasContext *s, uint32_t insn) } else { /* Load/store one element per register */ if (is_load) { - do_vec_ld(s, rt, index, tcg_addr, scale); + do_vec_ld(s, rt, index, tcg_addr, scale, s->be_data); } else { - do_vec_st(s, rt, index, tcg_addr, scale); + do_vec_st(s, rt, index, tcg_addr, scale, s->be_data); } } tcg_gen_add_i64(tcg_addr, tcg_addr, tcg_ebytes);