From patchwork Sat Oct 17 02:28:58 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Richard Henderson X-Patchwork-Id: 271093 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-10.0 required=3.0 tests=BAYES_00,DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, HEADER_FROM_DIFFERENT_DOMAINS, INCLUDES_PATCH, MAILING_LIST_MULTI,SIGNED_OFF_BY,SPF_HELO_NONE,SPF_PASS, UNWANTED_LANGUAGE_BODY,URIBL_BLOCKED,USER_AGENT_GIT autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id AAC5FC43457 for ; Sat, 17 Oct 2020 02:30:18 +0000 (UTC) Received: from lists.gnu.org (lists.gnu.org [209.51.188.17]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPS id 20B72206CB for ; Sat, 17 Oct 2020 02:30:18 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (2048-bit key) header.d=linaro.org header.i=@linaro.org header.b="UtUIn0NT" DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 20B72206CB Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=linaro.org Authentication-Results: mail.kernel.org; spf=pass smtp.mailfrom=qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org Received: from localhost ([::1]:34160 helo=lists1p.gnu.org) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1kTbz7-0008SG-1r for qemu-devel@archiver.kernel.org; Fri, 16 Oct 2020 22:30:17 -0400 Received: from eggs.gnu.org ([2001:470:142:3::10]:57154) by lists.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1kTby1-0006n9-UH for qemu-devel@nongnu.org; Fri, 16 Oct 2020 22:29:09 -0400 Received: from mail-pl1-x642.google.com ([2607:f8b0:4864:20::642]:33753) by eggs.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_128_GCM_SHA256:128) (Exim 4.90_1) (envelope-from ) id 1kTbxz-0007uT-Iy for qemu-devel@nongnu.org; Fri, 16 Oct 2020 22:29:09 -0400 Received: by mail-pl1-x642.google.com with SMTP id b19so2231693pld.0 for ; Fri, 16 Oct 2020 19:29:07 -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 :mime-version:content-transfer-encoding; bh=QE7oXE3Y7Ps/NfN2zozWUpmNMqzopUkfLZJ74iw9yws=; b=UtUIn0NTTxPc2tzksP3Jp78BJniKv/Jpe6SyiCVcfjMflRrdSBDyX9bDGf81K2d3t+ Qmte3tPghan8zcGiPQG5pQu1kIhcZbaSKc3D/1dt0REr1w2+4PyOmTr76rTzUIQfi9Lk 7lSYW/tm8rrNHZuQh+6RxcjRFxID6im07pAPqtJClKmiRbhsAP/bR8ytJqOFhl0M4t6T tyGzyNpQLNPTwGPtAfpfX7CYD37rUIaCc2WJW/3nubV+I0zZXWHWXXSf2J7kGiAoGmwq c3lu8qLZEcawUZ1gEf3JjUdRGNRpvqnfmTO3XixTjHirrtKgEQW6wevoYy3TQB6rpx9+ nITA== 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:mime-version:content-transfer-encoding; bh=QE7oXE3Y7Ps/NfN2zozWUpmNMqzopUkfLZJ74iw9yws=; b=h8tMpb4d90nUt8Fro3rxfdZnyUsba1EEio2N3GtdXmxxRUjX3UOpJImSPDpYU4yNrM w1y9Y8Vg+ZuE7qht2nV3qGlLRhQZfhJjemFFPU9Hor7PUBPiOt0weFpAQ3swHEWpqtHR 119Wx1Q2hE3K23JNIq5+UyjoZMH0R1Hp24XBzEelRJgYxXpfjfWfQGovPa5NDNFoSOUA WLSiW8ZMgUyYRMT7J2ARXWZpgGQHrkr/DQj2ne/ZqfY9JFdulmVOx2gbEePEDUK8sC7h 3DNJMmyxQrCbUaxHBcPIw8RHV8Qt1WMLisxS4i3kAxsd1BdG054sjzvftYXCsnFtS6Z2 DneQ== X-Gm-Message-State: AOAM533Yti7/sWI5jyTj3ZGzTO1D8lYvNdacO4toFSR8oT3cazsO9BpP PKc5vQkvixuOW96wF6VZqZ6Ao9jWznBQmA== X-Google-Smtp-Source: ABdhPJwmPomjDVnlN5ABHsmMW23i551+5md52bXYXfDdkilhLwMy0/wO/OCp1unWoouaaSSPZjERFQ== X-Received: by 2002:a17:90a:7d09:: with SMTP id g9mr6693280pjl.63.1602901745160; Fri, 16 Oct 2020 19:29:05 -0700 (PDT) Received: from localhost.localdomain ([71.212.141.89]) by smtp.gmail.com with ESMTPSA id k9sm4150818pgt.72.2020.10.16.19.29.04 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Fri, 16 Oct 2020 19:29:04 -0700 (PDT) From: Richard Henderson To: qemu-devel@nongnu.org Subject: [PATCH 1/4] target/s390x: Improve cc computation for ADD LOGICAL Date: Fri, 16 Oct 2020 19:28:58 -0700 Message-Id: <20201017022901.78425-2-richard.henderson@linaro.org> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20201017022901.78425-1-richard.henderson@linaro.org> References: <20201017022901.78425-1-richard.henderson@linaro.org> MIME-Version: 1.0 Received-SPF: pass client-ip=2607:f8b0:4864:20::642; envelope-from=richard.henderson@linaro.org; helo=mail-pl1-x642.google.com X-detected-operating-system: by eggs.gnu.org: No matching host in p0f cache. That's all we know. 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, RCVD_IN_DNSWL_NONE=-0.0001, SPF_HELO_NONE=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: 2.1.23 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: david@redhat.com Errors-To: qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org Sender: "Qemu-devel" The resulting cc is only dependent on the result and the carry-out. So save those things rather than the inputs. Carry-out for 64-bit inputs is had via tcg_gen_add2_i64 directly into cc_src. Carry-out for 32-bit inputs is had via extraction from a normal 64-bit add (with zero-extended inputs). Signed-off-by: Richard Henderson Reviewed-by: David Hildenbrand --- target/s390x/internal.h | 4 +- target/s390x/cc_helper.c | 25 ++++----- target/s390x/helper.c | 3 +- target/s390x/translate.c | 103 ++++++++++++++++++++++++------------- target/s390x/insn-data.def | 36 ++++++------- 5 files changed, 97 insertions(+), 74 deletions(-) diff --git a/target/s390x/internal.h b/target/s390x/internal.h index 64602660ae..55c5442102 100644 --- a/target/s390x/internal.h +++ b/target/s390x/internal.h @@ -160,6 +160,8 @@ enum cc_op { CC_OP_STATIC, /* CC value is env->cc_op */ CC_OP_NZ, /* env->cc_dst != 0 */ + CC_OP_ADDU, /* dst != 0, src = carry out (0,1) */ + CC_OP_LTGT_32, /* signed less/greater than (32bit) */ CC_OP_LTGT_64, /* signed less/greater than (64bit) */ CC_OP_LTUGTU_32, /* unsigned less/greater than (32bit) */ @@ -168,7 +170,6 @@ enum cc_op { CC_OP_LTGT0_64, /* signed less/greater than 0 (64bit) */ CC_OP_ADD_64, /* overflow on add (64bit) */ - CC_OP_ADDU_64, /* overflow on unsigned add (64bit) */ CC_OP_ADDC_64, /* overflow on unsigned add-carry (64bit) */ CC_OP_SUB_64, /* overflow on subtraction (64bit) */ CC_OP_SUBU_64, /* overflow on unsigned subtraction (64bit) */ @@ -178,7 +179,6 @@ enum cc_op { CC_OP_MULS_64, /* overflow on signed multiply (64bit) */ CC_OP_ADD_32, /* overflow on add (32bit) */ - CC_OP_ADDU_32, /* overflow on unsigned add (32bit) */ CC_OP_ADDC_32, /* overflow on unsigned add-carry (32bit) */ CC_OP_SUB_32, /* overflow on subtraction (32bit) */ CC_OP_SUBU_32, /* overflow on unsigned subtraction (32bit) */ diff --git a/target/s390x/cc_helper.c b/target/s390x/cc_helper.c index 5432aeeed4..59da4d1cc2 100644 --- a/target/s390x/cc_helper.c +++ b/target/s390x/cc_helper.c @@ -123,6 +123,12 @@ static uint32_t cc_calc_nz(uint64_t dst) return !!dst; } +static uint32_t cc_calc_addu(uint64_t carry_out, uint64_t result) +{ + g_assert(carry_out <= 1); + return (result != 0) + 2 * carry_out; +} + static uint32_t cc_calc_add_64(int64_t a1, int64_t a2, int64_t ar) { if ((a1 > 0 && a2 > 0 && ar < 0) || (a1 < 0 && a2 < 0 && ar > 0)) { @@ -138,11 +144,6 @@ static uint32_t cc_calc_add_64(int64_t a1, int64_t a2, int64_t ar) } } -static uint32_t cc_calc_addu_64(uint64_t a1, uint64_t a2, uint64_t ar) -{ - return (ar != 0) + 2 * (ar < a1); -} - static uint32_t cc_calc_addc_64(uint64_t a1, uint64_t a2, uint64_t ar) { /* Recover a2 + carry_in. */ @@ -239,11 +240,6 @@ static uint32_t cc_calc_add_32(int32_t a1, int32_t a2, int32_t ar) } } -static uint32_t cc_calc_addu_32(uint32_t a1, uint32_t a2, uint32_t ar) -{ - return (ar != 0) + 2 * (ar < a1); -} - static uint32_t cc_calc_addc_32(uint32_t a1, uint32_t a2, uint32_t ar) { /* Recover a2 + carry_in. */ @@ -483,12 +479,12 @@ static uint32_t do_calc_cc(CPUS390XState *env, uint32_t cc_op, case CC_OP_NZ: r = cc_calc_nz(dst); break; + case CC_OP_ADDU: + r = cc_calc_addu(src, dst); + break; case CC_OP_ADD_64: r = cc_calc_add_64(src, dst, vr); break; - case CC_OP_ADDU_64: - r = cc_calc_addu_64(src, dst, vr); - break; case CC_OP_ADDC_64: r = cc_calc_addc_64(src, dst, vr); break; @@ -517,9 +513,6 @@ static uint32_t do_calc_cc(CPUS390XState *env, uint32_t cc_op, case CC_OP_ADD_32: r = cc_calc_add_32(src, dst, vr); break; - case CC_OP_ADDU_32: - r = cc_calc_addu_32(src, dst, vr); - break; case CC_OP_ADDC_32: r = cc_calc_addc_32(src, dst, vr); break; diff --git a/target/s390x/helper.c b/target/s390x/helper.c index b877690845..db87a62a57 100644 --- a/target/s390x/helper.c +++ b/target/s390x/helper.c @@ -395,6 +395,7 @@ const char *cc_name(enum cc_op cc_op) [CC_OP_DYNAMIC] = "CC_OP_DYNAMIC", [CC_OP_STATIC] = "CC_OP_STATIC", [CC_OP_NZ] = "CC_OP_NZ", + [CC_OP_ADDU] = "CC_OP_ADDU", [CC_OP_LTGT_32] = "CC_OP_LTGT_32", [CC_OP_LTGT_64] = "CC_OP_LTGT_64", [CC_OP_LTUGTU_32] = "CC_OP_LTUGTU_32", @@ -402,7 +403,6 @@ const char *cc_name(enum cc_op cc_op) [CC_OP_LTGT0_32] = "CC_OP_LTGT0_32", [CC_OP_LTGT0_64] = "CC_OP_LTGT0_64", [CC_OP_ADD_64] = "CC_OP_ADD_64", - [CC_OP_ADDU_64] = "CC_OP_ADDU_64", [CC_OP_ADDC_64] = "CC_OP_ADDC_64", [CC_OP_SUB_64] = "CC_OP_SUB_64", [CC_OP_SUBU_64] = "CC_OP_SUBU_64", @@ -410,7 +410,6 @@ const char *cc_name(enum cc_op cc_op) [CC_OP_ABS_64] = "CC_OP_ABS_64", [CC_OP_NABS_64] = "CC_OP_NABS_64", [CC_OP_ADD_32] = "CC_OP_ADD_32", - [CC_OP_ADDU_32] = "CC_OP_ADDU_32", [CC_OP_ADDC_32] = "CC_OP_ADDC_32", [CC_OP_SUB_32] = "CC_OP_SUB_32", [CC_OP_SUBU_32] = "CC_OP_SUBU_32", diff --git a/target/s390x/translate.c b/target/s390x/translate.c index ac10f42f10..9bf4c14f66 100644 --- a/target/s390x/translate.c +++ b/target/s390x/translate.c @@ -600,13 +600,11 @@ static void gen_op_calc_cc(DisasContext *s) dummy = tcg_const_i64(0); /* FALLTHRU */ case CC_OP_ADD_64: - case CC_OP_ADDU_64: case CC_OP_ADDC_64: case CC_OP_SUB_64: case CC_OP_SUBU_64: case CC_OP_SUBB_64: case CC_OP_ADD_32: - case CC_OP_ADDU_32: case CC_OP_ADDC_32: case CC_OP_SUB_32: case CC_OP_SUBU_32: @@ -650,6 +648,7 @@ static void gen_op_calc_cc(DisasContext *s) /* 1 argument */ gen_helper_calc_cc(cc_op, cpu_env, local_cc_op, dummy, cc_dst, dummy); break; + case CC_OP_ADDU: case CC_OP_ICM: case CC_OP_LTGT_32: case CC_OP_LTGT_64: @@ -666,13 +665,11 @@ static void gen_op_calc_cc(DisasContext *s) gen_helper_calc_cc(cc_op, cpu_env, local_cc_op, cc_src, cc_dst, dummy); break; case CC_OP_ADD_64: - case CC_OP_ADDU_64: case CC_OP_ADDC_64: case CC_OP_SUB_64: case CC_OP_SUBU_64: case CC_OP_SUBB_64: case CC_OP_ADD_32: - case CC_OP_ADDU_32: case CC_OP_ADDC_32: case CC_OP_SUB_32: case CC_OP_SUBU_32: @@ -849,20 +846,19 @@ static void disas_jcc(DisasContext *s, DisasCompare *c, uint32_t mask) account_inline_branch(s, old_cc_op); break; - case CC_OP_ADDU_32: - case CC_OP_ADDU_64: + case CC_OP_ADDU: switch (mask) { - case 8 | 2: /* vr == 0 */ + case 8 | 2: /* result == 0 */ cond = TCG_COND_EQ; break; - case 4 | 1: /* vr != 0 */ + case 4 | 1: /* result != 0 */ cond = TCG_COND_NE; break; - case 8 | 4: /* no carry -> vr >= src */ - cond = TCG_COND_GEU; + case 8 | 4: /* no carry */ + cond = TCG_COND_EQ; break; - case 2 | 1: /* carry -> vr < src */ - cond = TCG_COND_LTU; + case 2 | 1: /* carry */ + cond = TCG_COND_NE; break; default: goto do_dynamic; @@ -950,26 +946,21 @@ static void disas_jcc(DisasContext *s, DisasCompare *c, uint32_t mask) tcg_gen_and_i64(c->u.s64.a, cc_src, cc_dst); break; - case CC_OP_ADDU_32: - c->is_64 = false; - c->u.s32.a = tcg_temp_new_i32(); - c->u.s32.b = tcg_temp_new_i32(); - tcg_gen_extrl_i64_i32(c->u.s32.a, cc_vr); - if (cond == TCG_COND_EQ || cond == TCG_COND_NE) { - tcg_gen_movi_i32(c->u.s32.b, 0); - } else { - tcg_gen_extrl_i64_i32(c->u.s32.b, cc_src); - } - break; - - case CC_OP_ADDU_64: - c->u.s64.a = cc_vr; + case CC_OP_ADDU: + c->is_64 = true; + c->u.s64.b = tcg_const_i64(0); c->g1 = true; - if (cond == TCG_COND_EQ || cond == TCG_COND_NE) { - c->u.s64.b = tcg_const_i64(0); - } else { - c->u.s64.b = cc_src; - c->g2 = true; + switch (mask) { + case 8 | 2: + case 4 | 1: /* result */ + c->u.s64.a = cc_dst; + break; + case 8 | 4: + case 2 | 1: /* carry */ + c->u.s64.a = cc_src; + break; + default: + g_assert_not_reached(); } break; @@ -1444,6 +1435,13 @@ static DisasJumpType op_add(DisasContext *s, DisasOps *o) return DISAS_NEXT; } +static DisasJumpType op_addu64(DisasContext *s, DisasOps *o) +{ + tcg_gen_movi_i64(cc_src, 0); + tcg_gen_add2_i64(o->out, cc_src, o->in1, cc_src, o->in2, cc_src); + return DISAS_NEXT; +} + static DisasJumpType op_addc(DisasContext *s, DisasOps *o) { DisasCompare cmp; @@ -1473,9 +1471,10 @@ static DisasJumpType op_addc(DisasContext *s, DisasOps *o) static DisasJumpType op_asi(DisasContext *s, DisasOps *o) { - o->in1 = tcg_temp_new_i64(); + bool non_atomic = !s390_has_feat(S390_FEAT_STFLE_45); - if (!s390_has_feat(S390_FEAT_STFLE_45)) { + o->in1 = tcg_temp_new_i64(); + if (non_atomic) { tcg_gen_qemu_ld_tl(o->in1, o->addr1, get_mem_index(s), s->insn->data); } else { /* Perform the atomic addition in memory. */ @@ -1486,7 +1485,30 @@ static DisasJumpType op_asi(DisasContext *s, DisasOps *o) /* Recompute also for atomic case: needed for setting CC. */ tcg_gen_add_i64(o->out, o->in1, o->in2); - if (!s390_has_feat(S390_FEAT_STFLE_45)) { + if (non_atomic) { + tcg_gen_qemu_st_tl(o->out, o->addr1, get_mem_index(s), s->insn->data); + } + return DISAS_NEXT; +} + +static DisasJumpType op_asiu64(DisasContext *s, DisasOps *o) +{ + bool non_atomic = !s390_has_feat(S390_FEAT_STFLE_45); + + o->in1 = tcg_temp_new_i64(); + if (non_atomic) { + tcg_gen_qemu_ld_tl(o->in1, o->addr1, get_mem_index(s), s->insn->data); + } else { + /* Perform the atomic addition in memory. */ + tcg_gen_atomic_fetch_add_i64(o->in1, o->addr1, o->in2, get_mem_index(s), + s->insn->data); + } + + /* Recompute also for atomic case: needed for setting CC. */ + tcg_gen_movi_i64(cc_src, 0); + tcg_gen_add2_i64(o->out, cc_src, o->in1, cc_src, o->in2, cc_src); + + if (non_atomic) { tcg_gen_qemu_st_tl(o->out, o->addr1, get_mem_index(s), s->insn->data); } return DISAS_NEXT; @@ -5184,12 +5206,14 @@ static void cout_adds64(DisasContext *s, DisasOps *o) static void cout_addu32(DisasContext *s, DisasOps *o) { - gen_op_update3_cc_i64(s, CC_OP_ADDU_32, o->in1, o->in2, o->out); + tcg_gen_shri_i64(cc_src, o->out, 32); + tcg_gen_ext32u_i64(cc_dst, o->out); + gen_op_update2_cc_i64(s, CC_OP_ADDU, cc_src, cc_dst); } static void cout_addu64(DisasContext *s, DisasOps *o) { - gen_op_update3_cc_i64(s, CC_OP_ADDU_64, o->in1, o->in2, o->out); + gen_op_update2_cc_i64(s, CC_OP_ADDU, cc_src, o->out); } static void cout_addc32(DisasContext *s, DisasOps *o) @@ -5636,6 +5660,13 @@ static void in1_r2_sr32(DisasContext *s, DisasOps *o) } #define SPEC_in1_r2_sr32 0 +static void in1_r2_32u(DisasContext *s, DisasOps *o) +{ + o->in1 = tcg_temp_new_i64(); + tcg_gen_ext32u_i64(o->in1, regs[get_field(s, r2)]); +} +#define SPEC_in1_r2_32u 0 + static void in1_r3(DisasContext *s, DisasOps *o) { o->in1 = load_reg(get_field(s, r3)); diff --git a/target/s390x/insn-data.def b/target/s390x/insn-data.def index d3bcdfd67b..b9ca9aeff5 100644 --- a/target/s390x/insn-data.def +++ b/target/s390x/insn-data.def @@ -58,29 +58,29 @@ C(0xa70b, AGHI, RI_a, Z, r1, i2, r1, 0, add, adds64) /* ADD LOGICAL */ - C(0x1e00, ALR, RR_a, Z, r1, r2, new, r1_32, add, addu32) - C(0xb9fa, ALRK, RRF_a, DO, r2, r3, new, r1_32, add, addu32) - C(0x5e00, AL, RX_a, Z, r1, m2_32u, new, r1_32, add, addu32) - C(0xe35e, ALY, RXY_a, LD, r1, m2_32u, new, r1_32, add, addu32) - C(0xb90a, ALGR, RRE, Z, r1, r2, r1, 0, add, addu64) - C(0xb91a, ALGFR, RRE, Z, r1, r2_32u, r1, 0, add, addu64) - C(0xb9ea, ALGRK, RRF_a, DO, r2, r3, r1, 0, add, addu64) - C(0xe30a, ALG, RXY_a, Z, r1, m2_64, r1, 0, add, addu64) - C(0xe31a, ALGF, RXY_a, Z, r1, m2_32u, r1, 0, add, addu64) + C(0x1e00, ALR, RR_a, Z, r1_32u, r2_32u, new, r1_32, add, addu32) + C(0xb9fa, ALRK, RRF_a, DO, r2_32u, r3_32u, new, r1_32, add, addu32) + C(0x5e00, AL, RX_a, Z, r1_32u, m2_32u, new, r1_32, add, addu32) + C(0xe35e, ALY, RXY_a, LD, r1_32u, m2_32u, new, r1_32, add, addu32) + C(0xb90a, ALGR, RRE, Z, r1, r2, r1, 0, addu64, addu64) + C(0xb91a, ALGFR, RRE, Z, r1, r2_32u, r1, 0, addu64, addu64) + C(0xb9ea, ALGRK, RRF_a, DO, r2, r3, r1, 0, addu64, addu64) + C(0xe30a, ALG, RXY_a, Z, r1, m2_64, r1, 0, addu64, addu64) + C(0xe31a, ALGF, RXY_a, Z, r1, m2_32u, r1, 0, addu64, addu64) /* ADD LOGICAL HIGH */ C(0xb9ca, ALHHHR, RRF_a, HW, r2_sr32, r3_sr32, new, r1_32h, add, addu32) - C(0xb9da, ALHHLR, RRF_a, HW, r2_sr32, r3, new, r1_32h, add, addu32) + C(0xb9da, ALHHLR, RRF_a, HW, r2_sr32, r3_32u, new, r1_32h, add, addu32) /* ADD LOGICAL IMMEDIATE */ - C(0xc20b, ALFI, RIL_a, EI, r1, i2_32u, new, r1_32, add, addu32) - C(0xc20a, ALGFI, RIL_a, EI, r1, i2_32u, r1, 0, add, addu64) + C(0xc20b, ALFI, RIL_a, EI, r1_32u, i2_32u, new, r1_32, add, addu32) + C(0xc20a, ALGFI, RIL_a, EI, r1, i2_32u, r1, 0, addu64, addu64) /* ADD LOGICAL WITH SIGNED IMMEDIATE */ - D(0xeb6e, ALSI, SIY, GIE, la1, i2, new, 0, asi, addu32, MO_TEUL) - C(0xecda, ALHSIK, RIE_d, DO, r3, i2, new, r1_32, add, addu32) - D(0xeb7e, ALGSI, SIY, GIE, la1, i2, new, 0, asi, addu64, MO_TEQ) - C(0xecdb, ALGHSIK, RIE_d, DO, r3, i2, r1, 0, add, addu64) + D(0xeb6e, ALSI, SIY, GIE, la1, i2_32u, new, 0, asi, addu32, MO_TEUL) + C(0xecda, ALHSIK, RIE_d, DO, r3_32u, i2_32u, new, r1_32, add, addu32) + C(0xeb7e, ALGSI, SIY, GIE, la1, i2, r1, 0, asiu64, addu64) + C(0xecdb, ALGHSIK, RIE_d, DO, r3, i2, r1, 0, addu64, addu64) /* ADD LOGICAL WITH SIGNED IMMEDIATE HIGH */ - C(0xcc0a, ALSIH, RIL_a, HW, r1_sr32, i2, new, r1_32h, add, addu32) - C(0xcc0b, ALSIHN, RIL_a, HW, r1_sr32, i2, new, r1_32h, add, 0) + C(0xcc0a, ALSIH, RIL_a, HW, r1_sr32, i2_32u, new, r1_32h, add, addu32) + C(0xcc0b, ALSIHN, RIL_a, HW, r1_sr32, i2_32u, new, r1_32h, add, 0) /* ADD LOGICAL WITH CARRY */ C(0xb998, ALCR, RRE, Z, r1, r2, new, r1_32, addc, addc32) C(0xb988, ALCGR, RRE, Z, r1, r2, r1, 0, addc, addc64) From patchwork Sat Oct 17 02:28:59 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Richard Henderson X-Patchwork-Id: 302697 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-12.8 required=3.0 tests=BAYES_00,DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, HEADER_FROM_DIFFERENT_DOMAINS, INCLUDES_PATCH, MAILING_LIST_MULTI, SIGNED_OFF_BY, SPF_HELO_NONE, SPF_PASS, URIBL_BLOCKED, USER_AGENT_GIT autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 78D22C433DF for ; Sat, 17 Oct 2020 02:30:16 +0000 (UTC) Received: from lists.gnu.org (lists.gnu.org [209.51.188.17]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPS id EBE4E206CB for ; Sat, 17 Oct 2020 02:30:15 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (2048-bit key) header.d=linaro.org header.i=@linaro.org header.b="SomVDVXe" DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org EBE4E206CB Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=linaro.org Authentication-Results: mail.kernel.org; spf=pass smtp.mailfrom=qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org Received: from localhost ([::1]:34018 helo=lists1p.gnu.org) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1kTbz4-0008PD-W4 for qemu-devel@archiver.kernel.org; Fri, 16 Oct 2020 22:30:15 -0400 Received: from eggs.gnu.org ([2001:470:142:3::10]:57156) by lists.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1kTby2-0006nm-Ip for qemu-devel@nongnu.org; Fri, 16 Oct 2020 22:29:10 -0400 Received: from mail-pl1-x644.google.com ([2607:f8b0:4864:20::644]:38269) by eggs.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_128_GCM_SHA256:128) (Exim 4.90_1) (envelope-from ) id 1kTby0-0007uW-0l for qemu-devel@nongnu.org; Fri, 16 Oct 2020 22:29:10 -0400 Received: by mail-pl1-x644.google.com with SMTP id p11so2215363pld.5 for ; Fri, 16 Oct 2020 19:29:07 -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 :mime-version:content-transfer-encoding; bh=iu8QEUTMk7QdqWKe4PiXTbPYVEKOLM/zzjRIHVs9ZJg=; b=SomVDVXet+nW4ioe2qqDObI6t8qkqzLY7WwwYwvr9Gt6ggEvqw/Q+DpopdmvMlq16a Ii0Cvy8gNmFOCFdMiIdTuKjeEXAsi+VBJTV/MK+uNpDWuHuRPip8sbvZQsd9dL6W2pfz o45Uk6Z2qZUYaVNpUSG5t3REmWqYg/k5+I0LawEXLuYA276oFS4Ohuke0G0iUe6OeNKD Us7JcMUmWXHOL/OMas4/OWP8UPGwQj6t0n16UgJqp0SoXb2wzAD8eFX5zKILGOirLZsI XXl03LBrHK5+hSZB0bJVC8bnPN3oOYq0MP7WjcoM6Zo1Ar5cxxI2PF6Fzbx/A5LfwBbc sbEg== 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:mime-version:content-transfer-encoding; bh=iu8QEUTMk7QdqWKe4PiXTbPYVEKOLM/zzjRIHVs9ZJg=; b=JXFPhmwG0JZvEAFHEMS5D4zTuo5VVmoz2qbsJI+e8Lb+CC81+ffR/leyZdSv9rhZ/B aHeJVYlua8HvxDyNv0eVh8ogNQPAT/+jTuHpRSlidbFfdvgkOnXvxIJ/kz1ImUqUdN8n vbOMVbxvfKg/etxtHt4hX+DfT4b/yKvUyT/hMQyXXx+FHs20KbrVNM3Ed/mingq7UpFp RMbHuE07ELlWLjW/RcVtGmjIbw8SVNVHpPMNl54OKrS5ftP11/gbGQshmOGhYP8yiN1u z4Z6YrJqmP6f0baA/fp/L/qTIbfciYP5Hwk99ali/MJaBPprbwPtLKfynHZG79Be5jRv AxpQ== X-Gm-Message-State: AOAM530QF2nw4poqEqAhv/2xrG0frYY1dhj9hmtVPnHce4S9wguWd8GI 8NJO4lW7EZBoFGJr3370dFHSNqMMQFCPtA== X-Google-Smtp-Source: ABdhPJyAhVO0ffYRtr+9uksCNWk/ZLZXTxC5L6SmcECb5e6unwKU4JEzyNxKuJ9AoJohsLmdk6iKRQ== X-Received: by 2002:a17:90a:e453:: with SMTP id jp19mr7214136pjb.34.1602901746203; Fri, 16 Oct 2020 19:29:06 -0700 (PDT) Received: from localhost.localdomain ([71.212.141.89]) by smtp.gmail.com with ESMTPSA id k9sm4150818pgt.72.2020.10.16.19.29.05 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Fri, 16 Oct 2020 19:29:05 -0700 (PDT) From: Richard Henderson To: qemu-devel@nongnu.org Subject: [PATCH 2/4] target/s390x: Improve ADD LOGICAL WITH CARRY Date: Fri, 16 Oct 2020 19:28:59 -0700 Message-Id: <20201017022901.78425-3-richard.henderson@linaro.org> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20201017022901.78425-1-richard.henderson@linaro.org> References: <20201017022901.78425-1-richard.henderson@linaro.org> MIME-Version: 1.0 Received-SPF: pass client-ip=2607:f8b0:4864:20::644; envelope-from=richard.henderson@linaro.org; helo=mail-pl1-x644.google.com X-detected-operating-system: by eggs.gnu.org: No matching host in p0f cache. That's all we know. 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, RCVD_IN_DNSWL_NONE=-0.0001, SPF_HELO_NONE=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: 2.1.23 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: david@redhat.com Errors-To: qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org Sender: "Qemu-devel" Now that ADD LOGICAL outputs carry, we can use that as input directly. It also means we can re-use CC_OP_ZC and produce an output carry directly from ADD LOGICAL WITH CARRY. Signed-off-by: Richard Henderson Reviewed-by: David Hildenbrand --- target/s390x/internal.h | 2 -- target/s390x/cc_helper.c | 26 --------------- target/s390x/helper.c | 2 -- target/s390x/translate.c | 66 ++++++++++++++++++-------------------- target/s390x/insn-data.def | 8 ++--- 5 files changed, 35 insertions(+), 69 deletions(-) diff --git a/target/s390x/internal.h b/target/s390x/internal.h index 55c5442102..f5f3ae063e 100644 --- a/target/s390x/internal.h +++ b/target/s390x/internal.h @@ -170,7 +170,6 @@ enum cc_op { CC_OP_LTGT0_64, /* signed less/greater than 0 (64bit) */ CC_OP_ADD_64, /* overflow on add (64bit) */ - CC_OP_ADDC_64, /* overflow on unsigned add-carry (64bit) */ CC_OP_SUB_64, /* overflow on subtraction (64bit) */ CC_OP_SUBU_64, /* overflow on unsigned subtraction (64bit) */ CC_OP_SUBB_64, /* overflow on unsigned sub-borrow (64bit) */ @@ -179,7 +178,6 @@ enum cc_op { CC_OP_MULS_64, /* overflow on signed multiply (64bit) */ CC_OP_ADD_32, /* overflow on add (32bit) */ - CC_OP_ADDC_32, /* overflow on unsigned add-carry (32bit) */ CC_OP_SUB_32, /* overflow on subtraction (32bit) */ CC_OP_SUBU_32, /* overflow on unsigned subtraction (32bit) */ CC_OP_SUBB_32, /* overflow on unsigned sub-borrow (32bit) */ diff --git a/target/s390x/cc_helper.c b/target/s390x/cc_helper.c index 59da4d1cc2..cd2c5c4b39 100644 --- a/target/s390x/cc_helper.c +++ b/target/s390x/cc_helper.c @@ -144,16 +144,6 @@ static uint32_t cc_calc_add_64(int64_t a1, int64_t a2, int64_t ar) } } -static uint32_t cc_calc_addc_64(uint64_t a1, uint64_t a2, uint64_t ar) -{ - /* Recover a2 + carry_in. */ - uint64_t a2c = ar - a1; - /* Check for a2+carry_in overflow, then a1+a2c overflow. */ - int carry_out = (a2c < a2) || (ar < a1); - - return (ar != 0) + 2 * carry_out; -} - static uint32_t cc_calc_sub_64(int64_t a1, int64_t a2, int64_t ar) { if ((a1 > 0 && a2 < 0 && ar < 0) || (a1 < 0 && a2 > 0 && ar > 0)) { @@ -240,16 +230,6 @@ static uint32_t cc_calc_add_32(int32_t a1, int32_t a2, int32_t ar) } } -static uint32_t cc_calc_addc_32(uint32_t a1, uint32_t a2, uint32_t ar) -{ - /* Recover a2 + carry_in. */ - uint32_t a2c = ar - a1; - /* Check for a2+carry_in overflow, then a1+a2c overflow. */ - int carry_out = (a2c < a2) || (ar < a1); - - return (ar != 0) + 2 * carry_out; -} - static uint32_t cc_calc_sub_32(int32_t a1, int32_t a2, int32_t ar) { if ((a1 > 0 && a2 < 0 && ar < 0) || (a1 < 0 && a2 > 0 && ar > 0)) { @@ -485,9 +465,6 @@ static uint32_t do_calc_cc(CPUS390XState *env, uint32_t cc_op, case CC_OP_ADD_64: r = cc_calc_add_64(src, dst, vr); break; - case CC_OP_ADDC_64: - r = cc_calc_addc_64(src, dst, vr); - break; case CC_OP_SUB_64: r = cc_calc_sub_64(src, dst, vr); break; @@ -513,9 +490,6 @@ static uint32_t do_calc_cc(CPUS390XState *env, uint32_t cc_op, case CC_OP_ADD_32: r = cc_calc_add_32(src, dst, vr); break; - case CC_OP_ADDC_32: - r = cc_calc_addc_32(src, dst, vr); - break; case CC_OP_SUB_32: r = cc_calc_sub_32(src, dst, vr); break; diff --git a/target/s390x/helper.c b/target/s390x/helper.c index db87a62a57..4f4561bc64 100644 --- a/target/s390x/helper.c +++ b/target/s390x/helper.c @@ -403,14 +403,12 @@ const char *cc_name(enum cc_op cc_op) [CC_OP_LTGT0_32] = "CC_OP_LTGT0_32", [CC_OP_LTGT0_64] = "CC_OP_LTGT0_64", [CC_OP_ADD_64] = "CC_OP_ADD_64", - [CC_OP_ADDC_64] = "CC_OP_ADDC_64", [CC_OP_SUB_64] = "CC_OP_SUB_64", [CC_OP_SUBU_64] = "CC_OP_SUBU_64", [CC_OP_SUBB_64] = "CC_OP_SUBB_64", [CC_OP_ABS_64] = "CC_OP_ABS_64", [CC_OP_NABS_64] = "CC_OP_NABS_64", [CC_OP_ADD_32] = "CC_OP_ADD_32", - [CC_OP_ADDC_32] = "CC_OP_ADDC_32", [CC_OP_SUB_32] = "CC_OP_SUB_32", [CC_OP_SUBU_32] = "CC_OP_SUBU_32", [CC_OP_SUBB_32] = "CC_OP_SUBB_32", diff --git a/target/s390x/translate.c b/target/s390x/translate.c index 9bf4c14f66..570b3c88c8 100644 --- a/target/s390x/translate.c +++ b/target/s390x/translate.c @@ -600,12 +600,10 @@ static void gen_op_calc_cc(DisasContext *s) dummy = tcg_const_i64(0); /* FALLTHRU */ case CC_OP_ADD_64: - case CC_OP_ADDC_64: case CC_OP_SUB_64: case CC_OP_SUBU_64: case CC_OP_SUBB_64: case CC_OP_ADD_32: - case CC_OP_ADDC_32: case CC_OP_SUB_32: case CC_OP_SUBU_32: case CC_OP_SUBB_32: @@ -665,12 +663,10 @@ static void gen_op_calc_cc(DisasContext *s) gen_helper_calc_cc(cc_op, cpu_env, local_cc_op, cc_src, cc_dst, dummy); break; case CC_OP_ADD_64: - case CC_OP_ADDC_64: case CC_OP_SUB_64: case CC_OP_SUBU_64: case CC_OP_SUBB_64: case CC_OP_ADD_32: - case CC_OP_ADDC_32: case CC_OP_SUB_32: case CC_OP_SUBU_32: case CC_OP_SUBB_32: @@ -1442,30 +1438,40 @@ static DisasJumpType op_addu64(DisasContext *s, DisasOps *o) return DISAS_NEXT; } -static DisasJumpType op_addc(DisasContext *s, DisasOps *o) +/* Compute carry into cc_src. */ +static void compute_carry(DisasContext *s) { - DisasCompare cmp; - TCGv_i64 carry; - - tcg_gen_add_i64(o->out, o->in1, o->in2); - - /* The carry flag is the msb of CC, therefore the branch mask that would - create that comparison is 3. Feeding the generated comparison to - setcond produces the carry flag that we desire. */ - disas_jcc(s, &cmp, 3); - carry = tcg_temp_new_i64(); - if (cmp.is_64) { - tcg_gen_setcond_i64(cmp.cond, carry, cmp.u.s64.a, cmp.u.s64.b); - } else { - TCGv_i32 t = tcg_temp_new_i32(); - tcg_gen_setcond_i32(cmp.cond, t, cmp.u.s32.a, cmp.u.s32.b); - tcg_gen_extu_i32_i64(carry, t); - tcg_temp_free_i32(t); + switch (s->cc_op) { + case CC_OP_ADDU: + break; + default: + gen_op_calc_cc(s); + /* fall through */ + case CC_OP_STATIC: + /* The carry flag is the msb of CC; compute into cc_src. */ + tcg_gen_extu_i32_i64(cc_src, cc_op); + tcg_gen_shri_i64(cc_src, cc_src, 1); + break; } - free_compare(&cmp); +} + +static DisasJumpType op_addc32(DisasContext *s, DisasOps *o) +{ + compute_carry(s); + tcg_gen_add_i64(o->out, o->in1, o->in2); + tcg_gen_add_i64(o->out, o->out, cc_src); + return DISAS_NEXT; +} + +static DisasJumpType op_addc64(DisasContext *s, DisasOps *o) +{ + compute_carry(s); + + TCGv_i64 zero = tcg_const_i64(0); + tcg_gen_add2_i64(o->out, cc_src, o->in1, zero, cc_src, zero); + tcg_gen_add2_i64(o->out, cc_src, o->out, cc_src, o->in2, zero); + tcg_temp_free_i64(zero); - tcg_gen_add_i64(o->out, o->out, carry); - tcg_temp_free_i64(carry); return DISAS_NEXT; } @@ -5216,16 +5222,6 @@ static void cout_addu64(DisasContext *s, DisasOps *o) gen_op_update2_cc_i64(s, CC_OP_ADDU, cc_src, o->out); } -static void cout_addc32(DisasContext *s, DisasOps *o) -{ - gen_op_update3_cc_i64(s, CC_OP_ADDC_32, o->in1, o->in2, o->out); -} - -static void cout_addc64(DisasContext *s, DisasOps *o) -{ - gen_op_update3_cc_i64(s, CC_OP_ADDC_64, o->in1, o->in2, o->out); -} - static void cout_cmps32(DisasContext *s, DisasOps *o) { gen_op_update2_cc_i64(s, CC_OP_LTGT_32, o->in1, o->in2); diff --git a/target/s390x/insn-data.def b/target/s390x/insn-data.def index b9ca9aeff5..d9e65a0380 100644 --- a/target/s390x/insn-data.def +++ b/target/s390x/insn-data.def @@ -82,10 +82,10 @@ C(0xcc0a, ALSIH, RIL_a, HW, r1_sr32, i2_32u, new, r1_32h, add, addu32) C(0xcc0b, ALSIHN, RIL_a, HW, r1_sr32, i2_32u, new, r1_32h, add, 0) /* ADD LOGICAL WITH CARRY */ - C(0xb998, ALCR, RRE, Z, r1, r2, new, r1_32, addc, addc32) - C(0xb988, ALCGR, RRE, Z, r1, r2, r1, 0, addc, addc64) - C(0xe398, ALC, RXY_a, Z, r1, m2_32u, new, r1_32, addc, addc32) - C(0xe388, ALCG, RXY_a, Z, r1, m2_64, r1, 0, addc, addc64) + C(0xb998, ALCR, RRE, Z, r1_32u, r2_32u, new, r1_32, addc32, addu32) + C(0xb988, ALCGR, RRE, Z, r1, r2, r1, 0, addc64, addu64) + C(0xe398, ALC, RXY_a, Z, r1_32u, m2_32u, new, r1_32, addc32, addu32) + C(0xe388, ALCG, RXY_a, Z, r1, m2_64, r1, 0, addc64, addu64) /* AND */ C(0x1400, NR, RR_a, Z, r1, r2, new, r1_32, and, nz32) From patchwork Sat Oct 17 02:29:00 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Richard Henderson X-Patchwork-Id: 271092 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-10.0 required=3.0 tests=BAYES_00,DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, HEADER_FROM_DIFFERENT_DOMAINS, INCLUDES_PATCH, MAILING_LIST_MULTI,SIGNED_OFF_BY,SPF_HELO_NONE,SPF_PASS, UNWANTED_LANGUAGE_BODY,URIBL_BLOCKED,USER_AGENT_GIT autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 1F138C433DF for ; Sat, 17 Oct 2020 02:33:03 +0000 (UTC) Received: from lists.gnu.org (lists.gnu.org [209.51.188.17]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPS id 99A732073A for ; Sat, 17 Oct 2020 02:33:02 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (2048-bit key) header.d=linaro.org header.i=@linaro.org header.b="J1SPbKLn" DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 99A732073A Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=linaro.org Authentication-Results: mail.kernel.org; spf=pass smtp.mailfrom=qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org Received: from localhost ([::1]:41026 helo=lists1p.gnu.org) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1kTc1l-000354-PO for qemu-devel@archiver.kernel.org; Fri, 16 Oct 2020 22:33:01 -0400 Received: from eggs.gnu.org ([2001:470:142:3::10]:57178) by lists.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1kTby3-0006pS-UX for qemu-devel@nongnu.org; Fri, 16 Oct 2020 22:29:11 -0400 Received: from mail-pl1-x635.google.com ([2607:f8b0:4864:20::635]:43212) by eggs.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_128_GCM_SHA256:128) (Exim 4.90_1) (envelope-from ) id 1kTby1-0007uf-Js for qemu-devel@nongnu.org; Fri, 16 Oct 2020 22:29:11 -0400 Received: by mail-pl1-x635.google.com with SMTP id o9so2206493plx.10 for ; Fri, 16 Oct 2020 19:29:09 -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 :mime-version:content-transfer-encoding; bh=HwBhEtwUMo2n1jlgRpkoHJK7IbT+pEwOIJ1XBIp7rtA=; b=J1SPbKLnf+qNxRU7TXHF/uasSM8DAgKYYEJTvfwPKeAq6KxBgAvPfralqEHJZ1iWyn dExM10CUqxvQJqTxLAM6CL/Hw1L2BKFWlL8zya7cDXgQkYh3GKzpCoIa6SVG/oQlP1R4 62cz2NI+qoutubNuBCHcLivyUgePP/kOnjIQKF5jtKHlMsfb1P2NouVl+u6C9Pt/x69L oDOpsE3ahLTqjsUZ1uDWgsWqmgNVBDwhNImyU30A+l44QFCl9ebql1DB5C19/ZUVPEC6 V27zVs+it7CsyluuAcIPrnjG3QLfDxFTVxgwDlTY/gwZPhChT/DCnBYSBG5U66JfMomg +kgg== 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:mime-version:content-transfer-encoding; bh=HwBhEtwUMo2n1jlgRpkoHJK7IbT+pEwOIJ1XBIp7rtA=; b=f5N0GeHpvVdSj2oAyLPn+O2Y36mbjfX3Kgkmzt2NZIRGXT6vuNF2HMWnEiemtGraf7 RRmb9kccDWNavD4g4eCmAeWkhULOUMsFMpRnmxXJH9LbEnpfh+qManWGpuaj6yOt6RjH RyWqBK7hssv/y//NOjyPeYaFFa+ZzKwvgWTqS8dzW2g8jthJubQMlUVsIczr6FFI7Q6s ERXglPWHBEuydGtCLRAWd5ct59yC/DdPj3nUehU2qHztGNv9GEFmgNETicW2As8inrtE oC4qH9FUaDO+gRkBl5yFfcWoluRYlnccB5yKNQGl3i1Nmu+/ylohEMxQeCE84O2zb7F5 tJCQ== X-Gm-Message-State: AOAM531S/HBcdaUv26hdi7cTXhqe+UmIORTElqjvgeDreEu9GRgLvxG3 nPDdax3Oo/JxGDeAw95NjyxjqiKOduNXBw== X-Google-Smtp-Source: ABdhPJyQjCAnT0jzu6OBdCN/eO7bGGr/uXfaDSzstybxyntz0KFH5TRvzMK3GXaUfDCjWsLA9F7rPQ== X-Received: by 2002:a17:90a:a394:: with SMTP id x20mr6906694pjp.213.1602901747477; Fri, 16 Oct 2020 19:29:07 -0700 (PDT) Received: from localhost.localdomain ([71.212.141.89]) by smtp.gmail.com with ESMTPSA id k9sm4150818pgt.72.2020.10.16.19.29.06 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Fri, 16 Oct 2020 19:29:06 -0700 (PDT) From: Richard Henderson To: qemu-devel@nongnu.org Subject: [PATCH 3/4] target/s390x: Improve cc computation for SUBTRACT LOGICAL Date: Fri, 16 Oct 2020 19:29:00 -0700 Message-Id: <20201017022901.78425-4-richard.henderson@linaro.org> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20201017022901.78425-1-richard.henderson@linaro.org> References: <20201017022901.78425-1-richard.henderson@linaro.org> MIME-Version: 1.0 Received-SPF: pass client-ip=2607:f8b0:4864:20::635; envelope-from=richard.henderson@linaro.org; helo=mail-pl1-x635.google.com X-detected-operating-system: by eggs.gnu.org: No matching host in p0f cache. That's all we know. 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, RCVD_IN_DNSWL_NONE=-0.0001, SPF_HELO_NONE=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: 2.1.23 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: david@redhat.com Errors-To: qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org Sender: "Qemu-devel" The resulting cc is only dependent on the result and the borrow-out. So save those things rather than the inputs. Borrow-out for 64-bit inputs is had via tcg_gen_sub2_i64 directly into cc_src. Borrow-out for 32-bit inputs is had via extraction from a normal 64-bit sub (with zero-extended inputs). Signed-off-by: Richard Henderson Reviewed-by: David Hildenbrand --- target/s390x/internal.h | 3 +-- target/s390x/cc_helper.c | 40 ++++++--------------------- target/s390x/helper.c | 3 +-- target/s390x/translate.c | 55 +++++++++++++++----------------------- target/s390x/insn-data.def | 24 ++++++++--------- 5 files changed, 43 insertions(+), 82 deletions(-) diff --git a/target/s390x/internal.h b/target/s390x/internal.h index f5f3ae063e..4077047494 100644 --- a/target/s390x/internal.h +++ b/target/s390x/internal.h @@ -161,6 +161,7 @@ enum cc_op { CC_OP_NZ, /* env->cc_dst != 0 */ CC_OP_ADDU, /* dst != 0, src = carry out (0,1) */ + CC_OP_SUBU, /* dst != 0, src = borrow out (0,-1) */ CC_OP_LTGT_32, /* signed less/greater than (32bit) */ CC_OP_LTGT_64, /* signed less/greater than (64bit) */ @@ -171,7 +172,6 @@ enum cc_op { CC_OP_ADD_64, /* overflow on add (64bit) */ CC_OP_SUB_64, /* overflow on subtraction (64bit) */ - CC_OP_SUBU_64, /* overflow on unsigned subtraction (64bit) */ CC_OP_SUBB_64, /* overflow on unsigned sub-borrow (64bit) */ CC_OP_ABS_64, /* sign eval on abs (64bit) */ CC_OP_NABS_64, /* sign eval on nabs (64bit) */ @@ -179,7 +179,6 @@ enum cc_op { CC_OP_ADD_32, /* overflow on add (32bit) */ CC_OP_SUB_32, /* overflow on subtraction (32bit) */ - CC_OP_SUBU_32, /* overflow on unsigned subtraction (32bit) */ CC_OP_SUBB_32, /* overflow on unsigned sub-borrow (32bit) */ CC_OP_ABS_32, /* sign eval on abs (64bit) */ CC_OP_NABS_32, /* sign eval on nabs (64bit) */ diff --git a/target/s390x/cc_helper.c b/target/s390x/cc_helper.c index cd2c5c4b39..c7728d1225 100644 --- a/target/s390x/cc_helper.c +++ b/target/s390x/cc_helper.c @@ -129,6 +129,11 @@ static uint32_t cc_calc_addu(uint64_t carry_out, uint64_t result) return (result != 0) + 2 * carry_out; } +static uint32_t cc_calc_subu(uint64_t borrow_out, uint64_t result) +{ + return cc_calc_addu(borrow_out + 1, result); +} + static uint32_t cc_calc_add_64(int64_t a1, int64_t a2, int64_t ar) { if ((a1 > 0 && a2 > 0 && ar < 0) || (a1 < 0 && a2 < 0 && ar > 0)) { @@ -159,19 +164,6 @@ static uint32_t cc_calc_sub_64(int64_t a1, int64_t a2, int64_t ar) } } -static uint32_t cc_calc_subu_64(uint64_t a1, uint64_t a2, uint64_t ar) -{ - if (ar == 0) { - return 2; - } else { - if (a2 > a1) { - return 1; - } else { - return 3; - } - } -} - static uint32_t cc_calc_subb_64(uint64_t a1, uint64_t a2, uint64_t ar) { int borrow_out; @@ -245,19 +237,6 @@ static uint32_t cc_calc_sub_32(int32_t a1, int32_t a2, int32_t ar) } } -static uint32_t cc_calc_subu_32(uint32_t a1, uint32_t a2, uint32_t ar) -{ - if (ar == 0) { - return 2; - } else { - if (a2 > a1) { - return 1; - } else { - return 3; - } - } -} - static uint32_t cc_calc_subb_32(uint32_t a1, uint32_t a2, uint32_t ar) { int borrow_out; @@ -462,15 +441,15 @@ static uint32_t do_calc_cc(CPUS390XState *env, uint32_t cc_op, case CC_OP_ADDU: r = cc_calc_addu(src, dst); break; + case CC_OP_SUBU: + r = cc_calc_subu(src, dst); + break; case CC_OP_ADD_64: r = cc_calc_add_64(src, dst, vr); break; case CC_OP_SUB_64: r = cc_calc_sub_64(src, dst, vr); break; - case CC_OP_SUBU_64: - r = cc_calc_subu_64(src, dst, vr); - break; case CC_OP_SUBB_64: r = cc_calc_subb_64(src, dst, vr); break; @@ -493,9 +472,6 @@ static uint32_t do_calc_cc(CPUS390XState *env, uint32_t cc_op, case CC_OP_SUB_32: r = cc_calc_sub_32(src, dst, vr); break; - case CC_OP_SUBU_32: - r = cc_calc_subu_32(src, dst, vr); - break; case CC_OP_SUBB_32: r = cc_calc_subb_32(src, dst, vr); break; diff --git a/target/s390x/helper.c b/target/s390x/helper.c index 4f4561bc64..fa3aa500e5 100644 --- a/target/s390x/helper.c +++ b/target/s390x/helper.c @@ -396,6 +396,7 @@ const char *cc_name(enum cc_op cc_op) [CC_OP_STATIC] = "CC_OP_STATIC", [CC_OP_NZ] = "CC_OP_NZ", [CC_OP_ADDU] = "CC_OP_ADDU", + [CC_OP_SUBU] = "CC_OP_SUBU", [CC_OP_LTGT_32] = "CC_OP_LTGT_32", [CC_OP_LTGT_64] = "CC_OP_LTGT_64", [CC_OP_LTUGTU_32] = "CC_OP_LTUGTU_32", @@ -404,13 +405,11 @@ const char *cc_name(enum cc_op cc_op) [CC_OP_LTGT0_64] = "CC_OP_LTGT0_64", [CC_OP_ADD_64] = "CC_OP_ADD_64", [CC_OP_SUB_64] = "CC_OP_SUB_64", - [CC_OP_SUBU_64] = "CC_OP_SUBU_64", [CC_OP_SUBB_64] = "CC_OP_SUBB_64", [CC_OP_ABS_64] = "CC_OP_ABS_64", [CC_OP_NABS_64] = "CC_OP_NABS_64", [CC_OP_ADD_32] = "CC_OP_ADD_32", [CC_OP_SUB_32] = "CC_OP_SUB_32", - [CC_OP_SUBU_32] = "CC_OP_SUBU_32", [CC_OP_SUBB_32] = "CC_OP_SUBB_32", [CC_OP_ABS_32] = "CC_OP_ABS_32", [CC_OP_NABS_32] = "CC_OP_NABS_32", diff --git a/target/s390x/translate.c b/target/s390x/translate.c index 570b3c88c8..48494a86cc 100644 --- a/target/s390x/translate.c +++ b/target/s390x/translate.c @@ -601,11 +601,9 @@ static void gen_op_calc_cc(DisasContext *s) /* FALLTHRU */ case CC_OP_ADD_64: case CC_OP_SUB_64: - case CC_OP_SUBU_64: case CC_OP_SUBB_64: case CC_OP_ADD_32: case CC_OP_SUB_32: - case CC_OP_SUBU_32: case CC_OP_SUBB_32: local_cc_op = tcg_const_i32(s->cc_op); break; @@ -656,6 +654,7 @@ static void gen_op_calc_cc(DisasContext *s) case CC_OP_TM_64: case CC_OP_SLA_32: case CC_OP_SLA_64: + case CC_OP_SUBU: case CC_OP_NZ_F128: case CC_OP_VC: case CC_OP_MULS_64: @@ -664,11 +663,9 @@ static void gen_op_calc_cc(DisasContext *s) break; case CC_OP_ADD_64: case CC_OP_SUB_64: - case CC_OP_SUBU_64: case CC_OP_SUBB_64: case CC_OP_ADD_32: case CC_OP_SUB_32: - case CC_OP_SUBU_32: case CC_OP_SUBB_32: /* 3 arguments */ gen_helper_calc_cc(cc_op, cpu_env, local_cc_op, cc_src, cc_dst, cc_vr); @@ -843,6 +840,7 @@ static void disas_jcc(DisasContext *s, DisasCompare *c, uint32_t mask) break; case CC_OP_ADDU: + case CC_OP_SUBU: switch (mask) { case 8 | 2: /* result == 0 */ cond = TCG_COND_EQ; @@ -850,33 +848,11 @@ static void disas_jcc(DisasContext *s, DisasCompare *c, uint32_t mask) case 4 | 1: /* result != 0 */ cond = TCG_COND_NE; break; - case 8 | 4: /* no carry */ - cond = TCG_COND_EQ; + case 8 | 4: /* !carry (borrow) */ + cond = old_cc_op == CC_OP_ADDU ? TCG_COND_EQ : TCG_COND_NE; break; - case 2 | 1: /* carry */ - cond = TCG_COND_NE; - break; - default: - goto do_dynamic; - } - account_inline_branch(s, old_cc_op); - break; - - case CC_OP_SUBU_32: - case CC_OP_SUBU_64: - /* Note that CC=0 is impossible; treat it as dont-care. */ - switch (mask & 7) { - case 2: /* zero -> op1 == op2 */ - cond = TCG_COND_EQ; - break; - case 4 | 1: /* !zero -> op1 != op2 */ - cond = TCG_COND_NE; - break; - case 4: /* borrow (!carry) -> op1 < op2 */ - cond = TCG_COND_LTU; - break; - case 2 | 1: /* !borrow (carry) -> op1 >= op2 */ - cond = TCG_COND_GEU; + case 2 | 1: /* carry (!borrow) */ + cond = old_cc_op == CC_OP_ADDU ? TCG_COND_NE : TCG_COND_EQ; break; default: goto do_dynamic; @@ -911,7 +887,6 @@ static void disas_jcc(DisasContext *s, DisasCompare *c, uint32_t mask) break; case CC_OP_LTGT_32: case CC_OP_LTUGTU_32: - case CC_OP_SUBU_32: c->is_64 = false; c->u.s32.a = tcg_temp_new_i32(); tcg_gen_extrl_i64_i32(c->u.s32.a, cc_src); @@ -928,7 +903,6 @@ static void disas_jcc(DisasContext *s, DisasCompare *c, uint32_t mask) break; case CC_OP_LTGT_64: case CC_OP_LTUGTU_64: - case CC_OP_SUBU_64: c->u.s64.a = cc_src; c->u.s64.b = cc_dst; c->g1 = c->g2 = true; @@ -943,6 +917,7 @@ static void disas_jcc(DisasContext *s, DisasCompare *c, uint32_t mask) break; case CC_OP_ADDU: + case CC_OP_SUBU: c->is_64 = true; c->u.s64.b = tcg_const_i64(0); c->g1 = true; @@ -1444,6 +1419,9 @@ static void compute_carry(DisasContext *s) switch (s->cc_op) { case CC_OP_ADDU: break; + case CC_OP_SUBU: + tcg_gen_addi_i64(cc_src, cc_src, 1); + break; default: gen_op_calc_cc(s); /* fall through */ @@ -4759,6 +4737,13 @@ static DisasJumpType op_sub(DisasContext *s, DisasOps *o) return DISAS_NEXT; } +static DisasJumpType op_subu64(DisasContext *s, DisasOps *o) +{ + tcg_gen_movi_i64(cc_src, 0); + tcg_gen_sub2_i64(o->out, cc_src, o->in1, cc_src, o->in2, cc_src); + return DISAS_NEXT; +} + static DisasJumpType op_subb(DisasContext *s, DisasOps *o) { DisasCompare cmp; @@ -5310,12 +5295,14 @@ static void cout_subs64(DisasContext *s, DisasOps *o) static void cout_subu32(DisasContext *s, DisasOps *o) { - gen_op_update3_cc_i64(s, CC_OP_SUBU_32, o->in1, o->in2, o->out); + tcg_gen_sari_i64(cc_src, o->out, 32); + tcg_gen_ext32u_i64(cc_dst, o->out); + gen_op_update2_cc_i64(s, CC_OP_SUBU, cc_src, cc_dst); } static void cout_subu64(DisasContext *s, DisasOps *o) { - gen_op_update3_cc_i64(s, CC_OP_SUBU_64, o->in1, o->in2, o->out); + gen_op_update2_cc_i64(s, CC_OP_SUBU, cc_src, o->out); } static void cout_subb32(DisasContext *s, DisasOps *o) diff --git a/target/s390x/insn-data.def b/target/s390x/insn-data.def index d9e65a0380..65ee998484 100644 --- a/target/s390x/insn-data.def +++ b/target/s390x/insn-data.def @@ -900,21 +900,21 @@ C(0xb9c9, SHHHR, RRF_a, HW, r2_sr32, r3_sr32, new, r1_32h, sub, subs32) C(0xb9d9, SHHLR, RRF_a, HW, r2_sr32, r3, new, r1_32h, sub, subs32) /* SUBTRACT LOGICAL */ - C(0x1f00, SLR, RR_a, Z, r1, r2, new, r1_32, sub, subu32) - C(0xb9fb, SLRK, RRF_a, DO, r2, r3, new, r1_32, sub, subu32) - C(0x5f00, SL, RX_a, Z, r1, m2_32u, new, r1_32, sub, subu32) - C(0xe35f, SLY, RXY_a, LD, r1, m2_32u, new, r1_32, sub, subu32) - C(0xb90b, SLGR, RRE, Z, r1, r2, r1, 0, sub, subu64) - C(0xb91b, SLGFR, RRE, Z, r1, r2_32u, r1, 0, sub, subu64) - C(0xb9eb, SLGRK, RRF_a, DO, r2, r3, r1, 0, sub, subu64) - C(0xe30b, SLG, RXY_a, Z, r1, m2_64, r1, 0, sub, subu64) - C(0xe31b, SLGF, RXY_a, Z, r1, m2_32u, r1, 0, sub, subu64) + C(0x1f00, SLR, RR_a, Z, r1_32u, r2_32u, new, r1_32, sub, subu32) + C(0xb9fb, SLRK, RRF_a, DO, r2_32u, r3_32u, new, r1_32, sub, subu32) + C(0x5f00, SL, RX_a, Z, r1_32u, m2_32u, new, r1_32, sub, subu32) + C(0xe35f, SLY, RXY_a, LD, r1_32u, m2_32u, new, r1_32, sub, subu32) + C(0xb90b, SLGR, RRE, Z, r1, r2, r1, 0, subu64, subu64) + C(0xb91b, SLGFR, RRE, Z, r1, r2_32u, r1, 0, subu64, subu64) + C(0xb9eb, SLGRK, RRF_a, DO, r2, r3, r1, 0, subu64, subu64) + C(0xe30b, SLG, RXY_a, Z, r1, m2_64, r1, 0, subu64, subu64) + C(0xe31b, SLGF, RXY_a, Z, r1, m2_32u, r1, 0, subu64, subu64) /* SUBTRACT LOCICAL HIGH */ C(0xb9cb, SLHHHR, RRF_a, HW, r2_sr32, r3_sr32, new, r1_32h, sub, subu32) - C(0xb9db, SLHHLR, RRF_a, HW, r2_sr32, r3, new, r1_32h, sub, subu32) + C(0xb9db, SLHHLR, RRF_a, HW, r2_sr32, r3_32u, new, r1_32h, sub, subu32) /* SUBTRACT LOGICAL IMMEDIATE */ - C(0xc205, SLFI, RIL_a, EI, r1, i2_32u, new, r1_32, sub, subu32) - C(0xc204, SLGFI, RIL_a, EI, r1, i2_32u, r1, 0, sub, subu64) + C(0xc205, SLFI, RIL_a, EI, r1_32u, i2_32u, new, r1_32, sub, subu32) + C(0xc204, SLGFI, RIL_a, EI, r1, i2_32u, r1, 0, subu64, subu64) /* SUBTRACT LOGICAL WITH BORROW */ C(0xb999, SLBR, RRE, Z, r1, r2, new, r1_32, subb, subb32) C(0xb989, SLBGR, RRE, Z, r1, r2, r1, 0, subb, subb64) From patchwork Sat Oct 17 02:29:01 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Richard Henderson X-Patchwork-Id: 302696 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-12.8 required=3.0 tests=BAYES_00,DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, HEADER_FROM_DIFFERENT_DOMAINS, INCLUDES_PATCH, MAILING_LIST_MULTI, SIGNED_OFF_BY, SPF_HELO_NONE, SPF_PASS, URIBL_BLOCKED, USER_AGENT_GIT autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 99E9FC433E7 for ; Sat, 17 Oct 2020 02:31:41 +0000 (UTC) Received: from lists.gnu.org (lists.gnu.org [209.51.188.17]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPS id 3BF10206CB for ; Sat, 17 Oct 2020 02:31:41 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (2048-bit key) header.d=linaro.org header.i=@linaro.org header.b="x6j7qZ9h" DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 3BF10206CB Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=linaro.org Authentication-Results: mail.kernel.org; spf=pass smtp.mailfrom=qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org Received: from localhost ([::1]:37178 helo=lists1p.gnu.org) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1kTc0S-0001VW-Db for qemu-devel@archiver.kernel.org; Fri, 16 Oct 2020 22:31:40 -0400 Received: from eggs.gnu.org ([2001:470:142:3::10]:57180) by lists.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1kTby4-0006qw-JZ for qemu-devel@nongnu.org; Fri, 16 Oct 2020 22:29:12 -0400 Received: from mail-pf1-x444.google.com ([2607:f8b0:4864:20::444]:34330) by eggs.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_128_GCM_SHA256:128) (Exim 4.90_1) (envelope-from ) id 1kTby2-0007ul-Gi for qemu-devel@nongnu.org; Fri, 16 Oct 2020 22:29:12 -0400 Received: by mail-pf1-x444.google.com with SMTP id e10so2554585pfj.1 for ; Fri, 16 Oct 2020 19:29:10 -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 :mime-version:content-transfer-encoding; bh=7uoZ/ov1iV24cdkp01wD4WHng15fvh7IXfD1X0fJdkY=; b=x6j7qZ9hhQ4HHVOpGPITlHBd1H2NSiGpicBLsYsK+mboCZWGOyEjLtbQ+WB1DLfNv5 zJEqNAtbccdvXwiTZa7ny3ACiDThTDKBBMpGANvQCFHnY5BDKuaA7lwsVyhzwYDXpWkh ySURDCdanWrknY92gUHEazBu9Jf/mQw5vefh8l/ySXZQsdvZWPCClR1JqIHg6n2q0lkW 4yPcXUSydjn1grUBBePbExGpS+h2ZfqEQoPJJarf8Hq6pBh8z+UTGqLdEdwxn24UIT2r CPH+YpvdTuWFxo1QSX7QGIAIfTGO3lzGnQi31eG+97Jua16Kxw5TCC/5pB6mdG6MLcxi SanA== 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:mime-version:content-transfer-encoding; bh=7uoZ/ov1iV24cdkp01wD4WHng15fvh7IXfD1X0fJdkY=; b=TvjQ7W9FxGunnKFGDnr+faJB6EjhQs0vELNg3mqsxt4FoPJiRFauPNdOzN1tPJPpoU zpBIKedQ5RcwTcFkdzxskyIV9JMmuoQRDiUAHGxQEpYdOTgGzZ8VBUDP0E9cifd6eJRq hsfcpF8eDoaimLdylEkILpXPYlPECdFh7k3DBTA8p4Hexr6Lz3KW9d7b1uQKurX6+CFd CJ9iVGrxK+8f3z99SNrI/kMdNCLcMVOJ7jUaxDvB4+wu2732+efw9e0tBAn23jFgvWcT lMN4ffHfA8Pn/2cKwPysu3GDzxzXHlYDgmiM05vKC3XmUNQu+Z/6XkXxBpS8Tbuz3Ged T3sw== X-Gm-Message-State: AOAM533fc6As4PNsesCF5Mp4fC+8UxaecqFAmOZds58Sqi+k0FKLqyu0 Rq4hYxU9mHov7hVC5irJXCoiLnT0nHqLmg== X-Google-Smtp-Source: ABdhPJx88MvT3vORUWrCvkUg9HNhkbyKLcjljYitxa3HnRXq15ePDNFRxvHuhbQ4L3eXjBqJSNMfgA== X-Received: by 2002:a63:f343:: with SMTP id t3mr5496864pgj.86.1602901748527; Fri, 16 Oct 2020 19:29:08 -0700 (PDT) Received: from localhost.localdomain ([71.212.141.89]) by smtp.gmail.com with ESMTPSA id k9sm4150818pgt.72.2020.10.16.19.29.07 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Fri, 16 Oct 2020 19:29:07 -0700 (PDT) From: Richard Henderson To: qemu-devel@nongnu.org Subject: [PATCH 4/4] target/s390x: Improve SUB LOGICAL WITH BORROW Date: Fri, 16 Oct 2020 19:29:01 -0700 Message-Id: <20201017022901.78425-5-richard.henderson@linaro.org> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20201017022901.78425-1-richard.henderson@linaro.org> References: <20201017022901.78425-1-richard.henderson@linaro.org> MIME-Version: 1.0 Received-SPF: pass client-ip=2607:f8b0:4864:20::444; envelope-from=richard.henderson@linaro.org; helo=mail-pf1-x444.google.com X-detected-operating-system: by eggs.gnu.org: No matching host in p0f cache. That's all we know. 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, RCVD_IN_DNSWL_NONE=-0.0001, SPF_HELO_NONE=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: 2.1.23 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: david@redhat.com Errors-To: qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org Sender: "Qemu-devel" Now that SUB LOGICAL outputs carry, we can use that as input directly. It also means we can re-use CC_OP_ZC and produce an output carry directly from SUB LOGICAL WITH BORROW. Signed-off-by: Richard Henderson --- target/s390x/internal.h | 2 -- target/s390x/cc_helper.c | 32 ----------------- target/s390x/helper.c | 2 -- target/s390x/translate.c | 74 ++++++++++++++++++++------------------ target/s390x/insn-data.def | 8 ++--- 5 files changed, 44 insertions(+), 74 deletions(-) diff --git a/target/s390x/internal.h b/target/s390x/internal.h index 4077047494..11515bb617 100644 --- a/target/s390x/internal.h +++ b/target/s390x/internal.h @@ -172,14 +172,12 @@ enum cc_op { CC_OP_ADD_64, /* overflow on add (64bit) */ CC_OP_SUB_64, /* overflow on subtraction (64bit) */ - CC_OP_SUBB_64, /* overflow on unsigned sub-borrow (64bit) */ CC_OP_ABS_64, /* sign eval on abs (64bit) */ CC_OP_NABS_64, /* sign eval on nabs (64bit) */ CC_OP_MULS_64, /* overflow on signed multiply (64bit) */ CC_OP_ADD_32, /* overflow on add (32bit) */ CC_OP_SUB_32, /* overflow on subtraction (32bit) */ - CC_OP_SUBB_32, /* overflow on unsigned sub-borrow (32bit) */ CC_OP_ABS_32, /* sign eval on abs (64bit) */ CC_OP_NABS_32, /* sign eval on nabs (64bit) */ CC_OP_MULS_32, /* overflow on signed multiply (32bit) */ diff --git a/target/s390x/cc_helper.c b/target/s390x/cc_helper.c index c7728d1225..e7039d0d18 100644 --- a/target/s390x/cc_helper.c +++ b/target/s390x/cc_helper.c @@ -164,19 +164,6 @@ static uint32_t cc_calc_sub_64(int64_t a1, int64_t a2, int64_t ar) } } -static uint32_t cc_calc_subb_64(uint64_t a1, uint64_t a2, uint64_t ar) -{ - int borrow_out; - - if (ar != a1 - a2) { /* difference means borrow-in */ - borrow_out = (a2 >= a1); - } else { - borrow_out = (a2 > a1); - } - - return (ar != 0) + 2 * !borrow_out; -} - static uint32_t cc_calc_abs_64(int64_t dst) { if ((uint64_t)dst == 0x8000000000000000ULL) { @@ -237,19 +224,6 @@ static uint32_t cc_calc_sub_32(int32_t a1, int32_t a2, int32_t ar) } } -static uint32_t cc_calc_subb_32(uint32_t a1, uint32_t a2, uint32_t ar) -{ - int borrow_out; - - if (ar != a1 - a2) { /* difference means borrow-in */ - borrow_out = (a2 >= a1); - } else { - borrow_out = (a2 > a1); - } - - return (ar != 0) + 2 * !borrow_out; -} - static uint32_t cc_calc_abs_32(int32_t dst) { if ((uint32_t)dst == 0x80000000UL) { @@ -450,9 +424,6 @@ static uint32_t do_calc_cc(CPUS390XState *env, uint32_t cc_op, case CC_OP_SUB_64: r = cc_calc_sub_64(src, dst, vr); break; - case CC_OP_SUBB_64: - r = cc_calc_subb_64(src, dst, vr); - break; case CC_OP_ABS_64: r = cc_calc_abs_64(dst); break; @@ -472,9 +443,6 @@ static uint32_t do_calc_cc(CPUS390XState *env, uint32_t cc_op, case CC_OP_SUB_32: r = cc_calc_sub_32(src, dst, vr); break; - case CC_OP_SUBB_32: - r = cc_calc_subb_32(src, dst, vr); - break; case CC_OP_ABS_32: r = cc_calc_abs_32(dst); break; diff --git a/target/s390x/helper.c b/target/s390x/helper.c index fa3aa500e5..7678994feb 100644 --- a/target/s390x/helper.c +++ b/target/s390x/helper.c @@ -405,12 +405,10 @@ const char *cc_name(enum cc_op cc_op) [CC_OP_LTGT0_64] = "CC_OP_LTGT0_64", [CC_OP_ADD_64] = "CC_OP_ADD_64", [CC_OP_SUB_64] = "CC_OP_SUB_64", - [CC_OP_SUBB_64] = "CC_OP_SUBB_64", [CC_OP_ABS_64] = "CC_OP_ABS_64", [CC_OP_NABS_64] = "CC_OP_NABS_64", [CC_OP_ADD_32] = "CC_OP_ADD_32", [CC_OP_SUB_32] = "CC_OP_SUB_32", - [CC_OP_SUBB_32] = "CC_OP_SUBB_32", [CC_OP_ABS_32] = "CC_OP_ABS_32", [CC_OP_NABS_32] = "CC_OP_NABS_32", [CC_OP_COMP_32] = "CC_OP_COMP_32", diff --git a/target/s390x/translate.c b/target/s390x/translate.c index 48494a86cc..0d8235a5fb 100644 --- a/target/s390x/translate.c +++ b/target/s390x/translate.c @@ -601,10 +601,8 @@ static void gen_op_calc_cc(DisasContext *s) /* FALLTHRU */ case CC_OP_ADD_64: case CC_OP_SUB_64: - case CC_OP_SUBB_64: case CC_OP_ADD_32: case CC_OP_SUB_32: - case CC_OP_SUBB_32: local_cc_op = tcg_const_i32(s->cc_op); break; case CC_OP_CONST0: @@ -663,10 +661,8 @@ static void gen_op_calc_cc(DisasContext *s) break; case CC_OP_ADD_64: case CC_OP_SUB_64: - case CC_OP_SUBB_64: case CC_OP_ADD_32: case CC_OP_SUB_32: - case CC_OP_SUBB_32: /* 3 arguments */ gen_helper_calc_cc(cc_op, cpu_env, local_cc_op, cc_src, cc_dst, cc_vr); break; @@ -4744,29 +4740,49 @@ static DisasJumpType op_subu64(DisasContext *s, DisasOps *o) return DISAS_NEXT; } -static DisasJumpType op_subb(DisasContext *s, DisasOps *o) +/* Compute borrow (0, -1) into cc_src. */ +static void compute_borrow(DisasContext *s) { - DisasCompare cmp; - TCGv_i64 borrow; - - tcg_gen_sub_i64(o->out, o->in1, o->in2); - - /* The !borrow flag is the msb of CC. Since we want the inverse of - that, we ask for a comparison of CC=0 | CC=1 -> mask of 8 | 4. */ - disas_jcc(s, &cmp, 8 | 4); - borrow = tcg_temp_new_i64(); - if (cmp.is_64) { - tcg_gen_setcond_i64(cmp.cond, borrow, cmp.u.s64.a, cmp.u.s64.b); - } else { - TCGv_i32 t = tcg_temp_new_i32(); - tcg_gen_setcond_i32(cmp.cond, t, cmp.u.s32.a, cmp.u.s32.b); - tcg_gen_extu_i32_i64(borrow, t); - tcg_temp_free_i32(t); + switch (s->cc_op) { + case CC_OP_SUBU: + break; + default: + gen_op_calc_cc(s); + /* fall through */ + case CC_OP_STATIC: + /* The carry flag is the msb of CC; compute into cc_src. */ + tcg_gen_extu_i32_i64(cc_src, cc_op); + tcg_gen_shri_i64(cc_src, cc_src, 1); + /* fall through */ + case CC_OP_ADDU: + tcg_gen_subi_i64(cc_src, cc_src, 1); + break; } - free_compare(&cmp); +} + +static DisasJumpType op_subb32(DisasContext *s, DisasOps *o) +{ + compute_borrow(s); + + /* Borrow is {0, -1}, so add to subtract. */ + tcg_gen_add_i64(o->out, o->in1, cc_src); + tcg_gen_sub_i64(o->out, o->out, o->in2); + return DISAS_NEXT; +} + +static DisasJumpType op_subb64(DisasContext *s, DisasOps *o) +{ + compute_borrow(s); + + /* + * Borrow is {0, -1}, so add to subtract; replicate the + * borrow input to produce 128-bit -1 for the addition. + */ + TCGv_i64 zero = tcg_const_i64(0); + tcg_gen_add2_i64(o->out, cc_src, o->in1, zero, cc_src, cc_src); + tcg_gen_sub2_i64(o->out, cc_src, o->out, cc_src, o->in2, zero); + tcg_temp_free_i64(zero); - tcg_gen_sub_i64(o->out, o->out, borrow); - tcg_temp_free_i64(borrow); return DISAS_NEXT; } @@ -5305,16 +5321,6 @@ static void cout_subu64(DisasContext *s, DisasOps *o) gen_op_update2_cc_i64(s, CC_OP_SUBU, cc_src, o->out); } -static void cout_subb32(DisasContext *s, DisasOps *o) -{ - gen_op_update3_cc_i64(s, CC_OP_SUBB_32, o->in1, o->in2, o->out); -} - -static void cout_subb64(DisasContext *s, DisasOps *o) -{ - gen_op_update3_cc_i64(s, CC_OP_SUBB_64, o->in1, o->in2, o->out); -} - static void cout_tm32(DisasContext *s, DisasOps *o) { gen_op_update2_cc_i64(s, CC_OP_TM_32, o->in1, o->in2); diff --git a/target/s390x/insn-data.def b/target/s390x/insn-data.def index 65ee998484..d91051843c 100644 --- a/target/s390x/insn-data.def +++ b/target/s390x/insn-data.def @@ -916,10 +916,10 @@ C(0xc205, SLFI, RIL_a, EI, r1_32u, i2_32u, new, r1_32, sub, subu32) C(0xc204, SLGFI, RIL_a, EI, r1, i2_32u, r1, 0, subu64, subu64) /* SUBTRACT LOGICAL WITH BORROW */ - C(0xb999, SLBR, RRE, Z, r1, r2, new, r1_32, subb, subb32) - C(0xb989, SLBGR, RRE, Z, r1, r2, r1, 0, subb, subb64) - C(0xe399, SLB, RXY_a, Z, r1, m2_32u, new, r1_32, subb, subb32) - C(0xe389, SLBG, RXY_a, Z, r1, m2_64, r1, 0, subb, subb64) + C(0xb999, SLBR, RRE, Z, r1_32u, r2_32u, new, r1_32, subb32, subu32) + C(0xb989, SLBGR, RRE, Z, r1, r2, r1, 0, subb64, subu64) + C(0xe399, SLB, RXY_a, Z, r1_32u, m2_32u, new, r1_32, subb32, subu32) + C(0xe389, SLBG, RXY_a, Z, r1, m2_64, r1, 0, subb64, subu64) /* SUPERVISOR CALL */ C(0x0a00, SVC, I, Z, 0, 0, 0, 0, svc, 0)