From mboxrd@z Thu Jan 1 00:00:00 1970 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=-18.7 required=3.0 tests=BAYES_00,DKIM_SIGNED, DKIM_VALID,DKIM_VALID_AU,HEADER_FROM_DIFFERENT_DOMAINS,INCLUDES_CR_TRAILER, INCLUDES_PATCH,MAILING_LIST_MULTI,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 1E161C433ED for ; Fri, 16 Apr 2021 21:33:27 +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 9A0B0610A6 for ; Fri, 16 Apr 2021 21:33:26 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 9A0B0610A6 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]:49050 helo=lists1p.gnu.org) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1lXW5d-00010U-IY for qemu-devel@archiver.kernel.org; Fri, 16 Apr 2021 17:33:25 -0400 Received: from eggs.gnu.org ([2001:470:142:3::10]:44620) by lists.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1lXVd9-0005f1-MR for qemu-devel@nongnu.org; Fri, 16 Apr 2021 17:03:59 -0400 Received: from mail-pf1-x432.google.com ([2607:f8b0:4864:20::432]:34415) by eggs.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_128_GCM_SHA256:128) (Exim 4.90_1) (envelope-from ) id 1lXVd5-0001cj-Pw for qemu-devel@nongnu.org; Fri, 16 Apr 2021 17:03:59 -0400 Received: by mail-pf1-x432.google.com with SMTP id 10so10257834pfl.1 for ; Fri, 16 Apr 2021 14:03:55 -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=SGOOa0zT0FE1S29kkqgncHYFmVqbSwpz80PVFtDGE1o=; b=CDZPdi8BdgOl/Hzv8n2SJ0mxWsZgKydcpMIFMV6Uc6dm+blpdfBJtzNdBdzhLf+QPQ mEniCCsIHezCKXBCvhOuZPXDVmU5uaQkB8dO95q6Pn+vVKugKngx8set92/tkZgAyWdb k73dH1BqFzjVEkeosk52TohEnDIdQvDcQ9p90i7c/VGOxFKXZ/e+hvxhfNmCLersUYUE ObVKqokTgWUIZzraYck5M2KQ0mEFJfnEDLU5E3uqZeSofXbQSLRnpSjppPn0rEsxZWht mqRRFbMKkLNZjlMQTEv0/m9RrWIY8MH5JA0Aady/MabARXklzWfiw+Jc9XjuqwUMXgad xvpA== 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=SGOOa0zT0FE1S29kkqgncHYFmVqbSwpz80PVFtDGE1o=; b=rf1N8mLFO8UpSb+gpb5dp3YdULLNy5az65VxGxHiVXL000bPsRnhphuwcoqYoqfY8v SzjBn7HknlmeXRiOmP/TUFzED4T5pOrlJAKMJkML4ZWo2p4ROS/Dsmt36//KFey0Yx05 2Pf23P+8jYS+zv8lPuD58WRZJoKFBhOlaGYFy7Gjr42Cty1Sh9fTbP2d1H9ZsZmdmTyh Wj4qRWN+wUOUNYaTZaTU/jdGzOWZc+zTwsJwmc2Q7aJxDo7Wp/P5K+hG0yvp25/Je+E7 ZAkvP6dwzbqR3xwY6pIxukm1caGbPmUHJ+uI6QCJdNrMZO72hbwDd0fOqcNl+6b9EXQn lTJA== X-Gm-Message-State: AOAM532QmUP+j7tU7ATRk6hh2FwInWCVTh7ZkORWtT2mxtqdi7DR91vb 8E0Jr4/aFuMWKqi1tOwflpDEKEcdY+CZLA== X-Google-Smtp-Source: ABdhPJyDVpcsWqLqvJKfmjkQwuqjl3696bkaDkVmzYgU6RrVr2/5Y6+uc2wPI2VeP5o6vO0KTbKfBw== X-Received: by 2002:aa7:84c7:0:b029:259:d7ed:a8dd with SMTP id x7-20020aa784c70000b0290259d7eda8ddmr5030310pfn.20.1618607034253; Fri, 16 Apr 2021 14:03:54 -0700 (PDT) Received: from localhost.localdomain ([71.212.131.83]) by smtp.gmail.com with ESMTPSA id m9sm5766734pgt.65.2021.04.16.14.03.53 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Fri, 16 Apr 2021 14:03:53 -0700 (PDT) From: Richard Henderson To: qemu-devel@nongnu.org Subject: [PATCH v5 42/81] target/arm: Implement SVE2 HISTCNT, HISTSEG Date: Fri, 16 Apr 2021 14:02:01 -0700 Message-Id: <20210416210240.1591291-43-richard.henderson@linaro.org> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20210416210240.1591291-1-richard.henderson@linaro.org> References: <20210416210240.1591291-1-richard.henderson@linaro.org> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Received-SPF: pass client-ip=2607:f8b0:4864:20::432; envelope-from=richard.henderson@linaro.org; helo=mail-pf1-x432.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, 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=unavailable 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: qemu-arm@nongnu.org, Stephen Long Errors-To: qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org Sender: "Qemu-devel" From: Stephen Long Signed-off-by: Stephen Long Message-Id: <20200416173109.8856-1-steplong@quicinc.com> Signed-off-by: Richard Henderson --- v2: Fix overlap between output and input vectors. v4: Fix histseg counting (zhiwei). --- target/arm/helper-sve.h | 7 ++ target/arm/sve.decode | 6 ++ target/arm/sve_helper.c | 131 +++++++++++++++++++++++++++++++++++++ target/arm/translate-sve.c | 19 ++++++ 4 files changed, 163 insertions(+) diff --git a/target/arm/helper-sve.h b/target/arm/helper-sve.h index 98e6b57e38..507a2fea8e 100644 --- a/target/arm/helper-sve.h +++ b/target/arm/helper-sve.h @@ -2551,6 +2551,13 @@ DEF_HELPER_FLAGS_5(sve2_nmatch_ppzz_b, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_5(sve2_nmatch_ppzz_h, TCG_CALL_NO_RWG, i32, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve2_histcnt_s, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(sve2_histcnt_d, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(sve2_histseg, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_6(sve2_faddp_zpzz_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_6(sve2_faddp_zpzz_s, TCG_CALL_NO_RWG, diff --git a/target/arm/sve.decode b/target/arm/sve.decode index 388bf92acf..8f501a083c 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -146,6 +146,7 @@ &rprrr_esz rn=%reg_movprfx @rdn_pg_rm_ra ........ esz:2 . ra:5 ... pg:3 rm:5 rd:5 \ &rprrr_esz rn=%reg_movprfx +@rd_pg_rn_rm ........ esz:2 . rm:5 ... pg:3 rn:5 rd:5 &rprr_esz # One register operand, with governing predicate, vector element size @rd_pg_rn ........ esz:2 ... ... ... pg:3 rn:5 rd:5 &rpr_esz @@ -1336,6 +1337,11 @@ RSUBHNT 01000101 .. 1 ..... 011 111 ..... ..... @rd_rn_rm MATCH 01000101 .. 1 ..... 100 ... ..... 0 .... @pd_pg_rn_rm NMATCH 01000101 .. 1 ..... 100 ... ..... 1 .... @pd_pg_rn_rm +### SVE2 Histogram Computation + +HISTCNT 01000101 .. 1 ..... 110 ... ..... ..... @rd_pg_rn_rm +HISTSEG 01000101 .. 1 ..... 101 000 ..... ..... @rd_rn_rm + ## SVE2 floating-point pairwise operations FADDP 01100100 .. 010 00 0 100 ... ..... ..... @rdn_pg_rm diff --git a/target/arm/sve_helper.c b/target/arm/sve_helper.c index 12acc4fb0b..8d002fdb65 100644 --- a/target/arm/sve_helper.c +++ b/target/arm/sve_helper.c @@ -7062,3 +7062,134 @@ DO_PPZZ_MATCH(sve2_nmatch_ppzz_b, MO_8, true) DO_PPZZ_MATCH(sve2_nmatch_ppzz_h, MO_16, true) #undef DO_PPZZ_MATCH + +void HELPER(sve2_histcnt_s)(void *vd, void *vn, void *vm, void *vg, + uint32_t desc) +{ + ARMVectorReg scratch; + intptr_t i, j; + intptr_t opr_sz = simd_oprsz(desc); + uint32_t *d = vd, *n = vn, *m = vm; + uint8_t *pg = vg; + + if (d == n) { + n = memcpy(&scratch, n, opr_sz); + if (d == m) { + m = n; + } + } else if (d == m) { + m = memcpy(&scratch, m, opr_sz); + } + + for (i = 0; i < opr_sz; i += 4) { + uint64_t count = 0; + uint8_t pred; + + pred = pg[H1(i >> 3)] >> (i & 7); + if (pred & 1) { + uint32_t nn = n[H4(i >> 2)]; + + for (j = 0; j <= i; j += 4) { + pred = pg[H1(j >> 3)] >> (j & 7); + if ((pred & 1) && nn == m[H4(j >> 2)]) { + ++count; + } + } + } + d[H4(i >> 2)] = count; + } +} + +void HELPER(sve2_histcnt_d)(void *vd, void *vn, void *vm, void *vg, + uint32_t desc) +{ + ARMVectorReg scratch; + intptr_t i, j; + intptr_t opr_sz = simd_oprsz(desc); + uint64_t *d = vd, *n = vn, *m = vm; + uint8_t *pg = vg; + + if (d == n) { + n = memcpy(&scratch, n, opr_sz); + if (d == m) { + m = n; + } + } else if (d == m) { + m = memcpy(&scratch, m, opr_sz); + } + + for (i = 0; i < opr_sz / 8; ++i) { + uint64_t count = 0; + if (pg[H1(i)] & 1) { + uint64_t nn = n[i]; + for (j = 0; j <= i; ++j) { + if ((pg[H1(j)] & 1) && nn == m[j]) { + ++count; + } + } + } + d[i] = count; + } +} + +/* + * Returns the number of bytes in m0 and m1 that match n. + * Unlike do_match2 we don't just need true/false, we need an exact count. + * This requires two extra logical operations. + */ +static inline uint64_t do_histseg_cnt(uint8_t n, uint64_t m0, uint64_t m1) +{ + const uint64_t mask = dup_const(MO_8, 0x7f); + uint64_t cmp0, cmp1; + + cmp1 = dup_const(MO_8, n); + cmp0 = cmp1 ^ m0; + cmp1 = cmp1 ^ m1; + + /* + * 1: clear msb of each byte to avoid carry to next byte (& mask) + * 2: carry in to msb if byte != 0 (+ mask) + * 3: set msb if cmp has msb set (| cmp) + * 4: set ~msb to ignore them (| mask) + * We now have 0xff for byte != 0 or 0x7f for byte == 0. + * 5: invert, resulting in 0x80 if and only if byte == 0. + */ + cmp0 = ~(((cmp0 & mask) + mask) | cmp0 | mask); + cmp1 = ~(((cmp1 & mask) + mask) | cmp1 | mask); + + /* + * Combine the two compares in a way that the bits do + * not overlap, and so preserves the count of set bits. + * If the host has an efficient instruction for ctpop, + * then ctpop(x) + ctpop(y) has the same number of + * operations as ctpop(x | (y >> 1)). If the host does + * not have an efficient ctpop, then we only want to + * use it once. + */ + return ctpop64(cmp0 | (cmp1 >> 1)); +} + +void HELPER(sve2_histseg)(void *vd, void *vn, void *vm, uint32_t desc) +{ + intptr_t i, j; + intptr_t opr_sz = simd_oprsz(desc); + + for (i = 0; i < opr_sz; i += 16) { + uint64_t n0 = *(uint64_t *)(vn + i); + uint64_t m0 = *(uint64_t *)(vm + i); + uint64_t n1 = *(uint64_t *)(vn + i + 8); + uint64_t m1 = *(uint64_t *)(vm + i + 8); + uint64_t out0 = 0; + uint64_t out1 = 0; + + for (j = 0; j < 64; j += 8) { + uint64_t cnt0 = do_histseg_cnt(n0 >> j, m0, m1); + uint64_t cnt1 = do_histseg_cnt(n1 >> j, m0, m1); + out0 |= cnt0 << j; + out1 |= cnt1 << j; + } + + *(uint64_t *)(vd + i) = out0; + *(uint64_t *)(vd + i + 8) = out1; + } +} diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index 49d7a45a50..0cbca8d715 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -7504,6 +7504,25 @@ static bool trans_##NAME(DisasContext *s, arg_rprr_esz *a) \ DO_SVE2_PPZZ_MATCH(MATCH, match) DO_SVE2_PPZZ_MATCH(NMATCH, nmatch) +static bool trans_HISTCNT(DisasContext *s, arg_rprr_esz *a) +{ + static gen_helper_gvec_4 * const fns[2] = { + gen_helper_sve2_histcnt_s, gen_helper_sve2_histcnt_d + }; + if (a->esz < 2) { + return false; + } + return do_sve2_zpzz_ool(s, a, fns[a->esz - 2]); +} + +static bool trans_HISTSEG(DisasContext *s, arg_rrr_esz *a) +{ + if (a->esz != 0) { + return false; + } + return do_sve2_zzz_ool(s, a, gen_helper_sve2_histseg); +} + static bool do_sve2_zpzz_fp(DisasContext *s, arg_rprr_esz *a, gen_helper_gvec_4_ptr *fn) { -- 2.25.1