From patchwork Wed Jun 27 04:33:26 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Patchwork-Submitter: Richard Henderson X-Patchwork-Id: 935300 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@bilbo.ozlabs.org Authentication-Results: ozlabs.org; spf=pass (mailfrom) smtp.mailfrom=nongnu.org (client-ip=2001:4830:134:3::11; helo=lists.gnu.org; envelope-from=qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org; receiver=) Authentication-Results: ozlabs.org; dmarc=fail (p=none dis=none) header.from=linaro.org Authentication-Results: ozlabs.org; dkim=fail reason="signature verification failed" (1024-bit key; unprotected) header.d=linaro.org header.i=@linaro.org header.b="KDe8YeJP"; dkim-atps=neutral Received: from lists.gnu.org (lists.gnu.org [IPv6:2001:4830:134:3::11]) (using TLSv1 with cipher AES256-SHA (256/256 bits)) (No client certificate requested) by ozlabs.org (Postfix) with ESMTPS id 41FrNW3VYVz9s0w for ; Wed, 27 Jun 2018 15:01:31 +1000 (AEST) Received: from localhost ([::1]:56666 helo=lists.gnu.org) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1fY2a9-0001gh-6X for incoming@patchwork.ozlabs.org; Wed, 27 Jun 2018 01:01:29 -0400 Received: from eggs.gnu.org ([2001:4830:134:3::10]:32952) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1fY29s-0004j2-RC for qemu-devel@nongnu.org; Wed, 27 Jun 2018 00:34:22 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1fY29r-00019q-4L for qemu-devel@nongnu.org; Wed, 27 Jun 2018 00:34:20 -0400 Received: from mail-pg0-x22f.google.com ([2607:f8b0:400e:c05::22f]:40253) by eggs.gnu.org with esmtps (TLS1.0:RSA_AES_128_CBC_SHA1:16) (Exim 4.71) (envelope-from ) id 1fY29q-00018l-Rm for qemu-devel@nongnu.org; Wed, 27 Jun 2018 00:34:19 -0400 Received: by mail-pg0-x22f.google.com with SMTP id w8-v6so363067pgp.7 for ; Tue, 26 Jun 2018 21:34:18 -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=k58LLRl/ubomF3QyWjTfEzFxj0hKy1XlanWjizd0aaM=; b=KDe8YeJPzp4V6bDofkISoXbfUvSgNaoDKI3JVQPI5vaszV7gbR6dhCKfl/CFHJOCNh kdFKhYHOAEBDLJO5AcRZJH1ekwI+EEkZD9ShzFJ89rTGktF80rXk7oMBVuWZSq4gMpoj yVHSGbENKGxkH1il+nxLeoPvhbKgGzt+rbGsM= 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=k58LLRl/ubomF3QyWjTfEzFxj0hKy1XlanWjizd0aaM=; b=jfr4H5sUnRpqiNaLII1AV+xxwYn14BHJaJAHqhr3LzeRREz32i+IS1yVkHAauAaBkW t4xsyRTgPO3YT6KmgXSAsSNwMRpj64C6k+Q0XSRND2YYFO1N4O2cQhx/tDAEuSU+T10+ y21dWsX8GQUyfshkonP+yC+8jsGVKw+NWT2mhWm0XoZ62MHGA08QB0SBIGdWjpCEim0T ZvQXQj0ywE1GyTVS6UaA7HxCv/PDghfpJ0z0bOVgy+9KqHmMfKMz6+eFXrDC4Uxqq5XC M4F2WYmzKJCHeDnm/NjBxuKXzE5EzSe1SQfg3sskE8cke73kau5NS2pHcbENEFxJNHHz bWQA== X-Gm-Message-State: APt69E2aKMq9ETiPm2Q3mgU4Hsf5m+fcf4UYFzxDd1TnUZueiDG4Bxr5 JWrsl8YJCHVJ96ZVqA9bz2/sN3Urv5w= X-Google-Smtp-Source: AAOMgpcOBPv/gVRUW4/7qCoaAkea3qc7YlhjUSJ6e8CdrRhzn0jGvGFHxBMKE8AX+OzBOgbETBszOA== X-Received: by 2002:a62:b02:: with SMTP id t2-v6mr4306504pfi.36.1530074057444; Tue, 26 Jun 2018 21:34:17 -0700 (PDT) Received: from cloudburst.twiddle.net (97-126-112-211.tukw.qwest.net. [97.126.112.211]) by smtp.gmail.com with ESMTPSA id p20-v6sm4577638pff.90.2018.06.26.21.34.16 (version=TLS1_2 cipher=ECDHE-RSA-CHACHA20-POLY1305 bits=256/256); Tue, 26 Jun 2018 21:34:16 -0700 (PDT) From: Richard Henderson To: qemu-devel@nongnu.org Date: Tue, 26 Jun 2018 21:33:26 -0700 Message-Id: <20180627043328.11531-34-richard.henderson@linaro.org> X-Mailer: git-send-email 2.17.1 In-Reply-To: <20180627043328.11531-1-richard.henderson@linaro.org> References: <20180627043328.11531-1-richard.henderson@linaro.org> X-detected-operating-system: by eggs.gnu.org: Genre and OS details not recognized. X-Received-From: 2607:f8b0:400e:c05::22f Subject: [Qemu-devel] [PATCH v6 33/35] target/arm: Implement SVE dot product (indexed) 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, qemu-arm@nongnu.org Errors-To: qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org Sender: "Qemu-devel" Signed-off-by: Richard Henderson Reviewed-by: Peter Maydell Reviewed-by: Alex Bennée --- v6: Rearrange the loops. The compiler does well with this form and hopefully they are also easier to read. --- target/arm/helper.h | 5 ++ target/arm/translate-sve.c | 18 ++++++ target/arm/vec_helper.c | 124 +++++++++++++++++++++++++++++++++++++ target/arm/sve.decode | 8 ++- 4 files changed, 154 insertions(+), 1 deletion(-) diff --git a/target/arm/helper.h b/target/arm/helper.h index e23ce7ff19..59e8c3bd1b 100644 --- a/target/arm/helper.h +++ b/target/arm/helper.h @@ -588,6 +588,11 @@ DEF_HELPER_FLAGS_4(gvec_udot_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_sdot_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_udot_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_sdot_idx_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_udot_idx_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_sdot_idx_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_udot_idx_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_5(gvec_fcaddh, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_5(gvec_fcadds, TCG_CALL_NO_RWG, diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index 8a2bd1f8c5..3cff71cae8 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -3440,6 +3440,24 @@ static bool trans_DOT_zzz(DisasContext *s, arg_DOT_zzz *a, uint32_t insn) return true; } +static bool trans_DOT_zzx(DisasContext *s, arg_DOT_zzx *a, uint32_t insn) +{ + static gen_helper_gvec_3 * const fns[2][2] = { + { gen_helper_gvec_sdot_idx_b, gen_helper_gvec_sdot_idx_h }, + { gen_helper_gvec_udot_idx_b, gen_helper_gvec_udot_idx_h } + }; + + if (sve_access_check(s)) { + unsigned vsz = vec_full_reg_size(s); + tcg_gen_gvec_3_ool(vec_full_reg_offset(s, a->rd), + vec_full_reg_offset(s, a->rn), + vec_full_reg_offset(s, a->rm), + vsz, vsz, a->index, fns[a->u][a->sz]); + } + return true; +} + + /* *** SVE Floating Point Multiply-Add Indexed Group */ diff --git a/target/arm/vec_helper.c b/target/arm/vec_helper.c index c16a30c3b5..37f338732e 100644 --- a/target/arm/vec_helper.c +++ b/target/arm/vec_helper.c @@ -261,6 +261,130 @@ void HELPER(gvec_udot_h)(void *vd, void *vn, void *vm, uint32_t desc) clear_tail(d, opr_sz, simd_maxsz(desc)); } +void HELPER(gvec_sdot_idx_b)(void *vd, void *vn, void *vm, uint32_t desc) +{ + intptr_t i, segend, opr_sz = simd_oprsz(desc), opr_sz_4 = opr_sz / 4; + intptr_t index = simd_data(desc); + uint32_t *d = vd; + int8_t *n = vn; + int8_t *m_indexed = (int8_t *)vm + index * 4; + + /* Notice the special case of opr_sz == 8, from aa64/aa32 advsimd. + * Otherwise opr_sz is a multiple of 16. + */ + segend = MIN(4, opr_sz_4); + i = 0; + do { + int8_t m0 = m_indexed[i * 4 + 0]; + int8_t m1 = m_indexed[i * 4 + 1]; + int8_t m2 = m_indexed[i * 4 + 2]; + int8_t m3 = m_indexed[i * 4 + 3]; + + do { + d[i] += n[i * 4 + 0] * m0 + + n[i * 4 + 1] * m1 + + n[i * 4 + 2] * m2 + + n[i * 4 + 3] * m3; + } while (++i < segend); + segend = i + 4; + } while (i < opr_sz_4); + + clear_tail(d, opr_sz, simd_maxsz(desc)); +} + +void HELPER(gvec_udot_idx_b)(void *vd, void *vn, void *vm, uint32_t desc) +{ + intptr_t i, segend, opr_sz = simd_oprsz(desc), opr_sz_4 = opr_sz / 4; + intptr_t index = simd_data(desc); + uint32_t *d = vd; + uint8_t *n = vn; + uint8_t *m_indexed = (uint8_t *)vm + index * 4; + + /* Notice the special case of opr_sz == 8, from aa64/aa32 advsimd. + * Otherwise opr_sz is a multiple of 16. + */ + segend = MIN(4, opr_sz_4); + i = 0; + do { + uint8_t m0 = m_indexed[i * 4 + 0]; + uint8_t m1 = m_indexed[i * 4 + 1]; + uint8_t m2 = m_indexed[i * 4 + 2]; + uint8_t m3 = m_indexed[i * 4 + 3]; + + do { + d[i] += n[i * 4 + 0] * m0 + + n[i * 4 + 1] * m1 + + n[i * 4 + 2] * m2 + + n[i * 4 + 3] * m3; + } while (++i < segend); + segend = i + 4; + } while (i < opr_sz_4); + + clear_tail(d, opr_sz, simd_maxsz(desc)); +} + +void HELPER(gvec_sdot_idx_h)(void *vd, void *vn, void *vm, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc), opr_sz_8 = opr_sz / 8; + intptr_t index = simd_data(desc); + uint64_t *d = vd; + int16_t *n = vn; + int16_t *m_indexed = (int16_t *)vm + index * 4; + + /* This is supported by SVE only, so opr_sz is always a multiple of 16. + * Process the entire segment all at once, writing back the results + * only after we've consumed all of the inputs. + */ + for (i = 0; i < opr_sz_8 ; i += 2) { + uint64_t d0, d1; + + d0 = n[i * 4 + 0] * (int64_t)m_indexed[i * 4 + 0]; + d0 += n[i * 4 + 1] * (int64_t)m_indexed[i * 4 + 1]; + d0 += n[i * 4 + 2] * (int64_t)m_indexed[i * 4 + 2]; + d0 += n[i * 4 + 3] * (int64_t)m_indexed[i * 4 + 3]; + d1 = n[i * 4 + 4] * (int64_t)m_indexed[i * 4 + 0]; + d1 += n[i * 4 + 5] * (int64_t)m_indexed[i * 4 + 1]; + d1 += n[i * 4 + 6] * (int64_t)m_indexed[i * 4 + 2]; + d1 += n[i * 4 + 7] * (int64_t)m_indexed[i * 4 + 3]; + + d[i + 0] += d0; + d[i + 1] += d1; + } + + clear_tail(d, opr_sz, simd_maxsz(desc)); +} + +void HELPER(gvec_udot_idx_h)(void *vd, void *vn, void *vm, uint32_t desc) +{ + intptr_t i, opr_sz = simd_oprsz(desc), opr_sz_8 = opr_sz / 8; + intptr_t index = simd_data(desc); + uint64_t *d = vd; + uint16_t *n = vn; + uint16_t *m_indexed = (uint16_t *)vm + index * 4; + + /* This is supported by SVE only, so opr_sz is always a multiple of 16. + * Process the entire segment all at once, writing back the results + * only after we've consumed all of the inputs. + */ + for (i = 0; i < opr_sz_8 ; i += 2) { + uint64_t d0, d1; + + d0 = n[i * 4 + 0] * (uint64_t)m_indexed[i * 4 + 0]; + d0 += n[i * 4 + 1] * (uint64_t)m_indexed[i * 4 + 1]; + d0 += n[i * 4 + 2] * (uint64_t)m_indexed[i * 4 + 2]; + d0 += n[i * 4 + 3] * (uint64_t)m_indexed[i * 4 + 3]; + d1 = n[i * 4 + 4] * (uint64_t)m_indexed[i * 4 + 0]; + d1 += n[i * 4 + 5] * (uint64_t)m_indexed[i * 4 + 1]; + d1 += n[i * 4 + 6] * (uint64_t)m_indexed[i * 4 + 2]; + d1 += n[i * 4 + 7] * (uint64_t)m_indexed[i * 4 + 3]; + + d[i + 0] += d0; + d[i + 1] += d1; + } + + clear_tail(d, opr_sz, simd_maxsz(desc)); +} + void HELPER(gvec_fcaddh)(void *vd, void *vn, void *vm, void *vfpst, uint32_t desc) { diff --git a/target/arm/sve.decode b/target/arm/sve.decode index 35415bfb6c..e10b689454 100644 --- a/target/arm/sve.decode +++ b/target/arm/sve.decode @@ -726,7 +726,13 @@ UMIN_zzi 00100101 .. 101 011 110 ........ ..... @rdn_i8u MUL_zzi 00100101 .. 110 000 110 ........ ..... @rdn_i8s # SVE integer dot product (unpredicated) -DOT_zzz 01000100 1 sz:1 0 rm:5 00000 u:1 rn:5 rd:5 +DOT_zzz 01000100 1 sz:1 0 rm:5 00000 u:1 rn:5 rd:5 ra=%reg_movprfx + +# SVE integer dot product (indexed) +DOT_zzx 01000100 101 index:2 rm:3 00000 u:1 rn:5 rd:5 \ + sz=0 ra=%reg_movprfx +DOT_zzx 01000100 111 index:1 rm:4 00000 u:1 rn:5 rd:5 \ + sz=1 ra=%reg_movprfx # SVE floating-point complex add (predicated) FCADD 01100100 esz:2 00000 rot:1 100 pg:3 rm:5 rd:5 \