From patchwork Wed Jan 17 16:14:24 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Richard Henderson X-Patchwork-Id: 124878 Delivered-To: patch@linaro.org Received: by 10.46.62.1 with SMTP id l1csp111670lja; Wed, 17 Jan 2018 08:54:15 -0800 (PST) X-Google-Smtp-Source: ACJfBouAt5vAYZZl9TdY6AEpYY8zixSAkG3CL645PpE1NWOxfrkM+dUAEpe6+qQOXx7cRF3FeLI4 X-Received: by 10.129.39.143 with SMTP id n137mr3271058ywn.53.1516208055207; Wed, 17 Jan 2018 08:54:15 -0800 (PST) ARC-Seal: i=1; a=rsa-sha256; t=1516208055; cv=none; d=google.com; s=arc-20160816; b=HtIdOlHsxuJxD6hKII/oCBp6QXqzBNEYJzoUpwbq4ESKSpmlz0qYFUhqZ/Yg0Ip7k9 lM10ScsUKc/g+mdq6sX82evofJteZhuXkyL8Mzrrs6Vp45Rgx7FL8JKUy99UcUx2CAfL bQqQ2jH39H87ucxH2ZcRGn/Yr0ynDNR3t2NX88j4Aozsk9zTbcWrFDspQL2405d0hpYA KqKYwEEeRj9IgEZRgV69oLZyjgDoAKOouTe8MeBpSslXEpxjIvpgO6uU9WmK66xVdvVl NrvkG+UlPm0Xd1u9SSqXJP80YPVFMfRgAvQyBDz+C9R5077puRrAd71TgNyLpIe80CtH aJ7A== 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:arc-authentication-results; bh=rOzLRr3Yx3U0sYXr+DTEVEL100T7Q4RXSZ7eTcxN/pI=; b=GIkGtkQHXTFu3qxhiLHPVtsaYaInQx5gp8Ba2lTJkrZUuzQSnFsxcKWlXJ1CP4MdZN frwB/Bu7C9pIUPo5rrXyxWd3WjLP7RLWJWPHql/v8Cgaw3NTQNC/Wswl1rLCHD74JYpg 5YVTjiH7J4zdgAN8UZ3lQJzIRYSe9lE67RDfbYxAm1KJq2sYyHL4CT976OqOLUsq+V3a dX0RTtjENvbx9/KViw6fYNtVMWoaVjtORoQesZyoJLwNVXxpDbYvPxb7vuSc2NPXv+JP spHmavzzbOGvS2R8bWASWkgWORwfKv/hyk2jcSX2HmEmGQg1FXAWI2OHL4jNZDFtA+R6 wlFA== ARC-Authentication-Results: i=1; mx.google.com; dkim=fail header.i=@linaro.org header.s=google header.b=Njkw8qsF; 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 c10si369602ybi.280.2018.01.17.08.54.14 for (version=TLS1 cipher=AES128-SHA bits=128/128); Wed, 17 Jan 2018 08:54:15 -0800 (PST) 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=Njkw8qsF; 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]:46873 helo=lists.gnu.org) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1ebqyc-00081U-EQ for patch@linaro.org; Wed, 17 Jan 2018 11:54:14 -0500 Received: from eggs.gnu.org ([2001:4830:134:3::10]:39062) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1ebqN0-0006nf-QK for qemu-devel@nongnu.org; Wed, 17 Jan 2018 11:15:26 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1ebqMx-0003kO-Ov for qemu-devel@nongnu.org; Wed, 17 Jan 2018 11:15:22 -0500 Received: from mail-pl0-x243.google.com ([2607:f8b0:400e:c01::243]:38306) by eggs.gnu.org with esmtps (TLS1.0:RSA_AES_128_CBC_SHA1:16) (Exim 4.71) (envelope-from ) id 1ebqMx-0003jn-F5 for qemu-devel@nongnu.org; Wed, 17 Jan 2018 11:15:19 -0500 Received: by mail-pl0-x243.google.com with SMTP id 13so9225673plb.5 for ; Wed, 17 Jan 2018 08:15:19 -0800 (PST) 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=rOzLRr3Yx3U0sYXr+DTEVEL100T7Q4RXSZ7eTcxN/pI=; b=Njkw8qsF1KLsEi8bglUEytoF8PrQqmlH4NQW9mlVqp+znErOIagwUJfHYEkJsSeUSH b29NQ6khVqngoJPvYpFtRvToUkn6ScTRrryg9mS3QQ13s8xg5aZbQdFBCq4ovplO3KWe dZO5sbVRFYkue+MWwyPqUSE32AZ2DEpzaHemw= 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=rOzLRr3Yx3U0sYXr+DTEVEL100T7Q4RXSZ7eTcxN/pI=; b=uTsXwYHgJatzGHctIQa4eaS4rC9b2CqQz+FpSR9Sqmi9bwtvhJCBOJMbcDqfn1AKZh 3D85sa5imOS11bq0+5EEtQ52nq9/Nl+JzjgoVwTfoOKsoiX7a5hJ/TcJhpBdZ4VQMSH7 x71ioqeeA5rBU/lLYMwRTsege7NprEwejIEDE8lXN5VlENVE8ym8Kfe3jtGIinFVjtFi DVTfLX6+Ocbp9wanVw7YJwhkwT1gkLXHqBbYotilh2qD4utTbjMrDQQnHZUwaEBDVl3j cCFJRzZNB02LKBGr+NK1KRLKHm+a1YahWxW+KAPF20gLMNzCycfjJecji468M+Ibej37 Cyww== X-Gm-Message-State: AKGB3mIxOFpC6b4VigKQ/Pmn5SZBa0KACb7BoWzbeVxStWXEvK0GEl9F e3czRpCLCzwxuDiQShq6dSvrHacRRlg= X-Received: by 10.84.244.72 with SMTP id e8mr37320542plt.420.1516205717901; Wed, 17 Jan 2018 08:15:17 -0800 (PST) Received: from cloudburst.twiddle.net.com (24-181-135-57.dhcp.knwc.wa.charter.com. [24.181.135.57]) by smtp.gmail.com with ESMTPSA id l10sm1099311pff.64.2018.01.17.08.15.13 (version=TLS1_2 cipher=ECDHE-RSA-CHACHA20-POLY1305 bits=256/256); Wed, 17 Jan 2018 08:15:17 -0800 (PST) From: Richard Henderson To: qemu-devel@nongnu.org Date: Wed, 17 Jan 2018 08:14:24 -0800 Message-Id: <20180117161435.28981-10-richard.henderson@linaro.org> X-Mailer: git-send-email 2.14.3 In-Reply-To: <20180117161435.28981-1-richard.henderson@linaro.org> References: <20180117161435.28981-1-richard.henderson@linaro.org> X-detected-operating-system: by eggs.gnu.org: Genre and OS details not recognized. X-Received-From: 2607:f8b0:400e:c01::243 Subject: [Qemu-devel] [PATCH v10.5 09/20] tcg: Add generic vector helpers with a scalar operand 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" Use dup to convert a non-constant scalar to a third vector. Add addition, multiplication, and logical operations with an immediate. Add addition and subtraction with a non-constant scalar. Allow for the front-end to build operations in which the scalar operand comes first. Signed-off-by: Richard Henderson --- accel/tcg/tcg-runtime.h | 19 +++ tcg/tcg-op-gvec.h | 50 +++++- accel/tcg/tcg-runtime-gvec.c | 180 ++++++++++++++++++++ tcg/tcg-op-gvec.c | 393 ++++++++++++++++++++++++++++++++++++++++++- 4 files changed, 640 insertions(+), 2 deletions(-) -- 2.14.3 diff --git a/accel/tcg/tcg-runtime.h b/accel/tcg/tcg-runtime.h index f224a975e8..d7e4583e7b 100644 --- a/accel/tcg/tcg-runtime.h +++ b/accel/tcg/tcg-runtime.h @@ -147,16 +147,31 @@ DEF_HELPER_FLAGS_4(gvec_add16, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_add32, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_add64, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_adds8, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(gvec_adds16, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(gvec_adds32, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(gvec_adds64, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) + DEF_HELPER_FLAGS_4(gvec_sub8, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_sub16, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_sub32, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_sub64, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_subs8, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(gvec_subs16, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(gvec_subs32, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(gvec_subs64, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) + DEF_HELPER_FLAGS_4(gvec_mul8, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_mul16, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_mul32, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_mul64, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_muls8, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(gvec_muls16, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(gvec_muls32, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(gvec_muls64, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) + DEF_HELPER_FLAGS_4(gvec_ssadd8, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_ssadd16, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_ssadd32, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) @@ -189,6 +204,10 @@ DEF_HELPER_FLAGS_4(gvec_xor, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_andc, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(gvec_orc, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(gvec_andi, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(gvec_xori, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) +DEF_HELPER_FLAGS_4(gvec_ori, TCG_CALL_NO_RWG, void, ptr, ptr, i64, i32) + DEF_HELPER_FLAGS_3(gvec_shl8i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) DEF_HELPER_FLAGS_3(gvec_shl16i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) DEF_HELPER_FLAGS_3(gvec_shl32i, TCG_CALL_NO_RWG, void, ptr, ptr, i32) diff --git a/tcg/tcg-op-gvec.h b/tcg/tcg-op-gvec.h index d0d43dcc62..f4ade86983 100644 --- a/tcg/tcg-op-gvec.h +++ b/tcg/tcg-op-gvec.h @@ -35,6 +35,12 @@ void tcg_gen_gvec_2_ool(uint32_t dofs, uint32_t aofs, uint32_t oprsz, uint32_t maxsz, int32_t data, gen_helper_gvec_2 *fn); +/* Similarly, passing an extra data value. */ +typedef void gen_helper_gvec_2i(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv_i32); +void tcg_gen_gvec_2i_ool(uint32_t dofs, uint32_t aofs, TCGv_i64 c, + uint32_t oprsz, uint32_t maxsz, int32_t data, + gen_helper_gvec_2i *fn); + /* Similarly, passing an extra pointer (e.g. env or float_status). */ typedef void gen_helper_gvec_2_ptr(TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_i32); void tcg_gen_gvec_2_ptr(uint32_t dofs, uint32_t aofs, @@ -102,8 +108,10 @@ typedef struct { void (*fni4)(TCGv_i32, TCGv_i32, int32_t); /* Expand inline with a host vector type. */ void (*fniv)(unsigned, TCGv_vec, TCGv_vec, int64_t); - /* Expand out-of-line helper w/descriptor. */ + /* Expand out-of-line helper w/descriptor, data in descriptor. */ gen_helper_gvec_2 *fno; + /* Expand out-of-line helper w/descriptor, data as argument. */ + gen_helper_gvec_2i *fnoi; /* The opcode, if any, to which this corresponds. */ TCGOpcode opc; /* The vector element size, if applicable. */ @@ -114,6 +122,27 @@ typedef struct { bool load_dest; } GVecGen2i; +typedef struct { + /* Expand inline as a 64-bit or 32-bit integer. + Only one of these will be non-NULL. */ + void (*fni8)(TCGv_i64, TCGv_i64, TCGv_i64); + void (*fni4)(TCGv_i32, TCGv_i32, TCGv_i32); + /* Expand inline with a host vector type. */ + void (*fniv)(unsigned, TCGv_vec, TCGv_vec, TCGv_vec); + /* Expand out-of-line helper w/descriptor. */ + gen_helper_gvec_2i *fno; + /* The opcode, if any, to which this corresponds. */ + TCGOpcode opc; + /* The data argument to the out-of-line helper. */ + uint32_t data; + /* The vector element size, if applicable. */ + uint8_t vece; + /* Prefer i64 to v64. */ + bool prefer_i64; + /* Load scalar as 1st source operand. */ + bool scalar_first; +} GVecGen2s; + typedef struct { /* Expand inline as a 64-bit or 32-bit integer. Only one of these will be non-NULL. */ @@ -158,6 +187,8 @@ void tcg_gen_gvec_2(uint32_t dofs, uint32_t aofs, uint32_t oprsz, uint32_t maxsz, const GVecGen2 *); void tcg_gen_gvec_2i(uint32_t dofs, uint32_t aofs, uint32_t oprsz, uint32_t maxsz, int64_t c, const GVecGen2i *); +void tcg_gen_gvec_2s(uint32_t dofs, uint32_t aofs, uint32_t oprsz, + uint32_t maxsz, TCGv_i64 c, const GVecGen2s *); void tcg_gen_gvec_3(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, uint32_t maxsz, const GVecGen3 *); void tcg_gen_gvec_4(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t cofs, @@ -179,6 +210,16 @@ void tcg_gen_gvec_sub(unsigned vece, uint32_t dofs, uint32_t aofs, void tcg_gen_gvec_mul(unsigned vece, uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, uint32_t maxsz); +void tcg_gen_gvec_addi(unsigned vece, uint32_t dofs, uint32_t aofs, + int64_t c, uint32_t oprsz, uint32_t maxsz); +void tcg_gen_gvec_muli(unsigned vece, uint32_t dofs, uint32_t aofs, + int64_t c, uint32_t oprsz, uint32_t maxsz); + +void tcg_gen_gvec_adds(unsigned vece, uint32_t dofs, uint32_t aofs, + TCGv_i64 c, uint32_t oprsz, uint32_t maxsz); +void tcg_gen_gvec_subs(unsigned vece, uint32_t dofs, uint32_t aofs, + TCGv_i64 c, uint32_t oprsz, uint32_t maxsz); + /* Saturated arithmetic. */ void tcg_gen_gvec_ssadd(unsigned vece, uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, uint32_t maxsz); @@ -200,6 +241,13 @@ void tcg_gen_gvec_andc(unsigned vece, uint32_t dofs, uint32_t aofs, void tcg_gen_gvec_orc(unsigned vece, uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, uint32_t maxsz); +void tcg_gen_gvec_andi(unsigned vece, uint32_t dofs, uint32_t aofs, + int64_t c, uint32_t oprsz, uint32_t maxsz); +void tcg_gen_gvec_xori(unsigned vece, uint32_t dofs, uint32_t aofs, + int64_t c, uint32_t oprsz, uint32_t maxsz); +void tcg_gen_gvec_ori(unsigned vece, uint32_t dofs, uint32_t aofs, + int64_t c, uint32_t oprsz, uint32_t maxsz); + void tcg_gen_gvec_dup_mem(unsigned vece, uint32_t dofs, uint32_t aofs, uint32_t s, uint32_t m); void tcg_gen_gvec_dup_i32(unsigned vece, uint32_t dofs, uint32_t s, diff --git a/accel/tcg/tcg-runtime-gvec.c b/accel/tcg/tcg-runtime-gvec.c index 70a23224c8..fdc9df0d66 100644 --- a/accel/tcg/tcg-runtime-gvec.c +++ b/accel/tcg/tcg-runtime-gvec.c @@ -122,6 +122,54 @@ void HELPER(gvec_add64)(void *d, void *a, void *b, uint32_t desc) clear_high(d, oprsz, desc); } +void HELPER(gvec_adds8)(void *d, void *a, uint64_t b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + vec8 vecb = (vec8)DUP16(b); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec8)) { + *(vec8 *)(d + i) = *(vec8 *)(a + i) + vecb; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_adds16)(void *d, void *a, uint64_t b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + vec16 vecb = (vec16)DUP8(b); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec16)) { + *(vec16 *)(d + i) = *(vec16 *)(a + i) + vecb; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_adds32)(void *d, void *a, uint64_t b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + vec32 vecb = (vec32)DUP4(b); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec32)) { + *(vec32 *)(d + i) = *(vec32 *)(a + i) + vecb; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_adds64)(void *d, void *a, uint64_t b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + vec64 vecb = (vec64)DUP2(b); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec64)) { + *(vec64 *)(d + i) = *(vec64 *)(a + i) + vecb; + } + clear_high(d, oprsz, desc); +} + void HELPER(gvec_sub8)(void *d, void *a, void *b, uint32_t desc) { intptr_t oprsz = simd_oprsz(desc); @@ -166,6 +214,54 @@ void HELPER(gvec_sub64)(void *d, void *a, void *b, uint32_t desc) clear_high(d, oprsz, desc); } +void HELPER(gvec_subs8)(void *d, void *a, uint64_t b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + vec8 vecb = (vec8)DUP16(b); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec8)) { + *(vec8 *)(d + i) = *(vec8 *)(a + i) - vecb; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_subs16)(void *d, void *a, uint64_t b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + vec16 vecb = (vec16)DUP8(b); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec16)) { + *(vec16 *)(d + i) = *(vec16 *)(a + i) - vecb; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_subs32)(void *d, void *a, uint64_t b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + vec32 vecb = (vec32)DUP4(b); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec32)) { + *(vec32 *)(d + i) = *(vec32 *)(a + i) - vecb; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_subs64)(void *d, void *a, uint64_t b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + vec64 vecb = (vec64)DUP2(b); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec64)) { + *(vec64 *)(d + i) = *(vec64 *)(a + i) - vecb; + } + clear_high(d, oprsz, desc); +} + void HELPER(gvec_mul8)(void *d, void *a, void *b, uint32_t desc) { intptr_t oprsz = simd_oprsz(desc); @@ -210,6 +306,54 @@ void HELPER(gvec_mul64)(void *d, void *a, void *b, uint32_t desc) clear_high(d, oprsz, desc); } +void HELPER(gvec_muls8)(void *d, void *a, uint64_t b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + vec8 vecb = (vec8)DUP16(b); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec8)) { + *(vec8 *)(d + i) = *(vec8 *)(a + i) * vecb; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_muls16)(void *d, void *a, uint64_t b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + vec16 vecb = (vec16)DUP8(b); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec16)) { + *(vec16 *)(d + i) = *(vec16 *)(a + i) * vecb; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_muls32)(void *d, void *a, uint64_t b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + vec32 vecb = (vec32)DUP4(b); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec32)) { + *(vec32 *)(d + i) = *(vec32 *)(a + i) * vecb; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_muls64)(void *d, void *a, uint64_t b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + vec64 vecb = (vec64)DUP2(b); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec64)) { + *(vec64 *)(d + i) = *(vec64 *)(a + i) * vecb; + } + clear_high(d, oprsz, desc); +} + void HELPER(gvec_neg8)(void *d, void *a, uint32_t desc) { intptr_t oprsz = simd_oprsz(desc); @@ -368,6 +512,42 @@ void HELPER(gvec_orc)(void *d, void *a, void *b, uint32_t desc) clear_high(d, oprsz, desc); } +void HELPER(gvec_andi)(void *d, void *a, uint64_t b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + vec64 vecb = (vec64)DUP2(b); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec64)) { + *(vec64 *)(d + i) = *(vec64 *)(a + i) & vecb; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_xori)(void *d, void *a, uint64_t b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + vec64 vecb = (vec64)DUP2(b); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec64)) { + *(vec64 *)(d + i) = *(vec64 *)(a + i) ^ vecb; + } + clear_high(d, oprsz, desc); +} + +void HELPER(gvec_ori)(void *d, void *a, uint64_t b, uint32_t desc) +{ + intptr_t oprsz = simd_oprsz(desc); + vec64 vecb = (vec64)DUP2(b); + intptr_t i; + + for (i = 0; i < oprsz; i += sizeof(vec64)) { + *(vec64 *)(d + i) = *(vec64 *)(a + i) | vecb; + } + clear_high(d, oprsz, desc); +} + void HELPER(gvec_shl8i)(void *d, void *a, uint32_t desc) { intptr_t oprsz = simd_oprsz(desc); diff --git a/tcg/tcg-op-gvec.c b/tcg/tcg-op-gvec.c index 368ab2f1de..3eedeaf15f 100644 --- a/tcg/tcg-op-gvec.c +++ b/tcg/tcg-op-gvec.c @@ -106,6 +106,28 @@ void tcg_gen_gvec_2_ool(uint32_t dofs, uint32_t aofs, tcg_temp_free_i32(desc); } +/* Generate a call to a gvec-style helper with two vector operands + and one scalar operand. */ +void tcg_gen_gvec_2i_ool(uint32_t dofs, uint32_t aofs, TCGv_i64 c, + uint32_t oprsz, uint32_t maxsz, int32_t data, + gen_helper_gvec_2i *fn) +{ + TCGv_ptr a0, a1; + TCGv_i32 desc = tcg_const_i32(simd_desc(oprsz, maxsz, data)); + + a0 = tcg_temp_new_ptr(); + a1 = tcg_temp_new_ptr(); + + tcg_gen_addi_ptr(a0, cpu_env, dofs); + tcg_gen_addi_ptr(a1, cpu_env, aofs); + + fn(a0, a1, c, desc); + + tcg_temp_free_ptr(a0); + tcg_temp_free_ptr(a1); + tcg_temp_free_i32(desc); +} + /* Generate a call to a gvec-style helper with three vector operands. */ void tcg_gen_gvec_3_ool(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, uint32_t maxsz, int32_t data, @@ -554,6 +576,27 @@ static void expand_2i_i32(uint32_t dofs, uint32_t aofs, uint32_t oprsz, tcg_temp_free_i32(t1); } +static void expand_2s_i32(uint32_t dofs, uint32_t aofs, uint32_t oprsz, + TCGv_i32 c, bool scalar_first, + void (*fni)(TCGv_i32, TCGv_i32, TCGv_i32)) +{ + TCGv_i32 t0 = tcg_temp_new_i32(); + TCGv_i32 t1 = tcg_temp_new_i32(); + uint32_t i; + + for (i = 0; i < oprsz; i += 4) { + tcg_gen_ld_i32(t0, cpu_env, aofs + i); + if (scalar_first) { + fni(t1, c, t0); + } else { + fni(t1, t0, c); + } + tcg_gen_st_i32(t1, cpu_env, dofs + i); + } + tcg_temp_free_i32(t0); + tcg_temp_free_i32(t1); +} + /* Expand OPSZ bytes worth of three-operand operations using i32 elements. */ static void expand_3_i32(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, bool load_dest, @@ -637,6 +680,27 @@ static void expand_2i_i64(uint32_t dofs, uint32_t aofs, uint32_t oprsz, tcg_temp_free_i64(t1); } +static void expand_2s_i64(uint32_t dofs, uint32_t aofs, uint32_t oprsz, + TCGv_i64 c, bool scalar_first, + void (*fni)(TCGv_i64, TCGv_i64, TCGv_i64)) +{ + TCGv_i64 t0 = tcg_temp_new_i64(); + TCGv_i64 t1 = tcg_temp_new_i64(); + uint32_t i; + + for (i = 0; i < oprsz; i += 8) { + tcg_gen_ld_i64(t0, cpu_env, aofs + i); + if (scalar_first) { + fni(t1, c, t0); + } else { + fni(t1, t0, c); + } + tcg_gen_st_i64(t1, cpu_env, dofs + i); + } + tcg_temp_free_i64(t0); + tcg_temp_free_i64(t1); +} + /* Expand OPSZ bytes worth of three-operand operations using i64 elements. */ static void expand_3_i64(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, bool load_dest, @@ -724,6 +788,28 @@ static void expand_2i_vec(unsigned vece, uint32_t dofs, uint32_t aofs, tcg_temp_free_vec(t1); } +static void expand_2s_vec(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t oprsz, uint32_t tysz, TCGType type, + TCGv_vec c, bool scalar_first, + void (*fni)(unsigned, TCGv_vec, TCGv_vec, TCGv_vec)) +{ + TCGv_vec t0 = tcg_temp_new_vec(type); + TCGv_vec t1 = tcg_temp_new_vec(type); + uint32_t i; + + for (i = 0; i < oprsz; i += tysz) { + tcg_gen_ld_vec(t0, cpu_env, aofs + i); + if (scalar_first) { + fni(vece, t1, c, t0); + } else { + fni(vece, t1, t0, c); + } + tcg_gen_st_vec(t1, cpu_env, dofs + i); + } + tcg_temp_free_vec(t0); + tcg_temp_free_vec(t1); +} + /* Expand OPSZ bytes worth of three-operand operations using host vectors. */ static void expand_3_vec(unsigned vece, uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, @@ -823,6 +909,7 @@ void tcg_gen_gvec_2(uint32_t dofs, uint32_t aofs, } } +/* Expand a vector operation with two vectors and an immediate. */ void tcg_gen_gvec_2i(uint32_t dofs, uint32_t aofs, uint32_t oprsz, uint32_t maxsz, int64_t c, const GVecGen2i *g) { @@ -859,7 +946,93 @@ void tcg_gen_gvec_2i(uint32_t dofs, uint32_t aofs, uint32_t oprsz, } else if (g->fni4 && check_size_impl(oprsz, 4)) { expand_2i_i32(dofs, aofs, oprsz, c, g->load_dest, g->fni4); } else { - tcg_gen_gvec_2_ool(dofs, aofs, oprsz, maxsz, c, g->fno); + if (g->fno) { + tcg_gen_gvec_2_ool(dofs, aofs, oprsz, maxsz, c, g->fno); + } else { + TCGv_i64 tcg_c = tcg_const_i64(c); + tcg_gen_gvec_2i_ool(dofs, aofs, tcg_c, oprsz, maxsz, c, g->fnoi); + tcg_temp_free_i64(tcg_c); + } + return; + } + + if (oprsz < maxsz) { + expand_clr(dofs + oprsz, maxsz - oprsz); + } +} + +/* Expand a vector operation with two vectors and a scalar. */ +void tcg_gen_gvec_2s(uint32_t dofs, uint32_t aofs, uint32_t oprsz, + uint32_t maxsz, TCGv_i64 c, const GVecGen2s *g) +{ + TCGType type; + + check_size_align(oprsz, maxsz, dofs | aofs); + check_overlap_2(dofs, aofs, maxsz); + + type = 0; + if (g->fniv) { + if (TCG_TARGET_HAS_v256 && check_size_impl(oprsz, 32)) { + type = TCG_TYPE_V256; + } else if (TCG_TARGET_HAS_v128 && check_size_impl(oprsz, 16)) { + type = TCG_TYPE_V128; + } else if (TCG_TARGET_HAS_v64 && !g->prefer_i64 + && check_size_impl(oprsz, 8)) { + type = TCG_TYPE_V64; + } + } + if (type != 0) { + TCGv_vec t_vec = tcg_temp_new_vec(type); + uint32_t done; + + tcg_gen_dup_i64_vec(g->vece, t_vec, c); + + /* Recall that ARM SVE allows vector sizes that are not a power of 2. + Expand with successively smaller host vector sizes. The intent is + that e.g. oprsz == 80 would be expanded with 2x32 + 1x16. */ + switch (type) { + case TCG_TYPE_V256: + done = QEMU_ALIGN_DOWN(oprsz, 32); + expand_2s_vec(g->vece, dofs, aofs, done, 32, TCG_TYPE_V256, + t_vec, g->scalar_first, g->fniv); + dofs += done; + aofs += done; + oprsz -= done; + maxsz -= done; + if (oprsz == 0) { + break; + } + /* fallthru */ + + case TCG_TYPE_V128: + expand_2s_vec(g->vece, dofs, aofs, oprsz, 16, TCG_TYPE_V128, + t_vec, g->scalar_first, g->fniv); + break; + + case TCG_TYPE_V64: + expand_2s_vec(g->vece, dofs, aofs, oprsz, 8, TCG_TYPE_V64, + t_vec, g->scalar_first, g->fniv); + break; + + default: + g_assert_not_reached(); + } + tcg_temp_free_vec(t_vec); + } else if (g->fni8 && check_size_impl(oprsz, 8)) { + TCGv_i64 t64 = tcg_temp_new_i64(); + + gen_dup_i64(g->vece, t64, c); + expand_2s_i64(dofs, aofs, oprsz, t64, g->scalar_first, g->fni8); + tcg_temp_free_i64(t64); + } else if (g->fni4 && check_size_impl(oprsz, 4)) { + TCGv_i32 t32 = tcg_temp_new_i32(); + + tcg_gen_extrl_i64_i32(t32, c); + gen_dup_i32(g->vece, t32, t32); + expand_2s_i32(dofs, aofs, oprsz, t32, g->scalar_first, g->fni4); + tcg_temp_free_i32(t32); + } else { + tcg_gen_gvec_2i_ool(dofs, aofs, c, oprsz, maxsz, 0, g->fno); return; } @@ -1183,6 +1356,121 @@ void tcg_gen_gvec_add(unsigned vece, uint32_t dofs, uint32_t aofs, tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g[vece]); } +static void tcg_gen_vec_addi8_i64(TCGv_i64 d, TCGv_i64 a, int64_t b) +{ + TCGv_i64 t = tcg_const_i64((b & 0xff) * (-1ull / 0xff)); + tcg_gen_vec_add8_i64(d, a, t); + tcg_temp_free_i64(t); +} + +static void tcg_gen_vec_addi16_i64(TCGv_i64 d, TCGv_i64 a, int64_t b) +{ + TCGv_i64 t = tcg_const_i64((b & 0xffff) * (-1ull / 0xffff)); + tcg_gen_vec_add16_i64(d, a, t); + tcg_temp_free_i64(t); +} + +static void tcg_gen_addi_vec(unsigned vece, TCGv_vec d, TCGv_vec a, int64_t b) +{ + TCGv_vec t = tcg_temp_new_vec_matching(d); + tcg_gen_dupi_vec(vece, t, b); + tcg_gen_add_vec(vece, d, a, t); + tcg_temp_free_vec(t); +} + +void tcg_gen_gvec_addi(unsigned vece, uint32_t dofs, uint32_t aofs, + int64_t c, uint32_t oprsz, uint32_t maxsz) +{ + static const GVecGen2i g[4] = { + { .fni8 = tcg_gen_vec_addi8_i64, + .fniv = tcg_gen_addi_vec, + .fnoi = gen_helper_gvec_adds8, + .opc = INDEX_op_add_vec, + .vece = MO_8 }, + { .fni8 = tcg_gen_vec_addi16_i64, + .fniv = tcg_gen_addi_vec, + .fnoi = gen_helper_gvec_adds16, + .opc = INDEX_op_add_vec, + .vece = MO_16 }, + { .fni4 = tcg_gen_addi_i32, + .fniv = tcg_gen_addi_vec, + .fnoi = gen_helper_gvec_adds32, + .opc = INDEX_op_add_vec, + .vece = MO_32 }, + { .fni8 = tcg_gen_addi_i64, + .fniv = tcg_gen_addi_vec, + .fnoi = gen_helper_gvec_adds64, + .opc = INDEX_op_add_vec, + .prefer_i64 = TCG_TARGET_REG_BITS == 64, + .vece = MO_64 }, + }; + + tcg_debug_assert(vece <= MO_64); + tcg_gen_gvec_2i(dofs, aofs, oprsz, maxsz, c, &g[vece]); +} + +void tcg_gen_gvec_adds(unsigned vece, uint32_t dofs, uint32_t aofs, + TCGv_i64 c, uint32_t oprsz, uint32_t maxsz) +{ + static const GVecGen2s g[4] = { + { .fni8 = tcg_gen_vec_add8_i64, + .fniv = tcg_gen_add_vec, + .fno = gen_helper_gvec_adds8, + .opc = INDEX_op_add_vec, + .vece = MO_8 }, + { .fni8 = tcg_gen_vec_add16_i64, + .fniv = tcg_gen_add_vec, + .fno = gen_helper_gvec_adds16, + .opc = INDEX_op_add_vec, + .vece = MO_16 }, + { .fni4 = tcg_gen_add_i32, + .fniv = tcg_gen_add_vec, + .fno = gen_helper_gvec_adds32, + .opc = INDEX_op_add_vec, + .vece = MO_32 }, + { .fni8 = tcg_gen_add_i64, + .fniv = tcg_gen_add_vec, + .fno = gen_helper_gvec_adds64, + .opc = INDEX_op_add_vec, + .prefer_i64 = TCG_TARGET_REG_BITS == 64, + .vece = MO_64 }, + }; + + tcg_debug_assert(vece <= MO_64); + tcg_gen_gvec_2s(dofs, aofs, oprsz, maxsz, c, &g[vece]); +} + +void tcg_gen_gvec_subs(unsigned vece, uint32_t dofs, uint32_t aofs, + TCGv_i64 c, uint32_t oprsz, uint32_t maxsz) +{ + static const GVecGen2s g[4] = { + { .fni8 = tcg_gen_vec_sub8_i64, + .fniv = tcg_gen_sub_vec, + .fno = gen_helper_gvec_subs8, + .opc = INDEX_op_sub_vec, + .vece = MO_8 }, + { .fni8 = tcg_gen_vec_sub16_i64, + .fniv = tcg_gen_sub_vec, + .fno = gen_helper_gvec_subs16, + .opc = INDEX_op_sub_vec, + .vece = MO_16 }, + { .fni4 = tcg_gen_sub_i32, + .fniv = tcg_gen_sub_vec, + .fno = gen_helper_gvec_subs32, + .opc = INDEX_op_sub_vec, + .vece = MO_32 }, + { .fni8 = tcg_gen_sub_i64, + .fniv = tcg_gen_sub_vec, + .fno = gen_helper_gvec_subs64, + .opc = INDEX_op_sub_vec, + .prefer_i64 = TCG_TARGET_REG_BITS == 64, + .vece = MO_64 }, + }; + + tcg_debug_assert(vece <= MO_64); + tcg_gen_gvec_2s(dofs, aofs, oprsz, maxsz, c, &g[vece]); +} + /* Perform a vector subtraction using normal subtraction and a mask. Compare gen_addv_mask above. */ static void gen_subv_mask(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b, TCGv_i64 m) @@ -1291,6 +1579,43 @@ void tcg_gen_gvec_mul(unsigned vece, uint32_t dofs, uint32_t aofs, tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g[vece]); } +static void tcg_gen_muli_vec(unsigned vece, TCGv_vec d, TCGv_vec a, int64_t b) +{ + TCGv_vec t = tcg_temp_new_vec_matching(d); + tcg_gen_dupi_vec(vece, t, b); + tcg_gen_mul_vec(vece, d, a, t); + tcg_temp_free_vec(t); +} + +void tcg_gen_gvec_muli(unsigned vece, uint32_t dofs, uint32_t aofs, + int64_t c, uint32_t oprsz, uint32_t maxsz) +{ + static const GVecGen2i g[4] = { + { .fniv = tcg_gen_muli_vec, + .fnoi = gen_helper_gvec_muls8, + .opc = INDEX_op_mul_vec, + .vece = MO_8 }, + { .fniv = tcg_gen_muli_vec, + .fnoi = gen_helper_gvec_muls16, + .opc = INDEX_op_mul_vec, + .vece = MO_16 }, + { .fni4 = tcg_gen_muli_i32, + .fniv = tcg_gen_muli_vec, + .fnoi = gen_helper_gvec_muls32, + .opc = INDEX_op_mul_vec, + .vece = MO_32 }, + { .fni8 = tcg_gen_muli_i64, + .fniv = tcg_gen_muli_vec, + .fnoi = gen_helper_gvec_muls64, + .opc = INDEX_op_mul_vec, + .prefer_i64 = TCG_TARGET_REG_BITS == 64, + .vece = MO_64 }, + }; + + tcg_debug_assert(vece <= MO_64); + tcg_gen_gvec_2i(dofs, aofs, oprsz, maxsz, c, &g[vece]); +} + void tcg_gen_gvec_ssadd(unsigned vece, uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t oprsz, uint32_t maxsz) { @@ -1523,6 +1848,72 @@ void tcg_gen_gvec_orc(unsigned vece, uint32_t dofs, uint32_t aofs, tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g); } +static void tcg_gen_andi_vec(unsigned vece, TCGv_vec d, TCGv_vec a, int64_t b) +{ + TCGv_vec t = tcg_temp_new_vec_matching(d); + tcg_gen_dupi_vec(vece, t, b); + tcg_gen_and_vec(vece, d, a, t); + tcg_temp_free_vec(t); +} + +void tcg_gen_gvec_andi(unsigned vece, uint32_t dofs, uint32_t aofs, + int64_t c, uint32_t oprsz, uint32_t maxsz) +{ + static const GVecGen2i g = { + .fni8 = tcg_gen_andi_i64, + .fniv = tcg_gen_andi_vec, + .fnoi = gen_helper_gvec_andi, + .opc = INDEX_op_and_vec, + .prefer_i64 = TCG_TARGET_REG_BITS == 64, + .vece = MO_64 + }; + tcg_gen_gvec_2i(dofs, aofs, oprsz, maxsz, c, &g); +} + +static void tcg_gen_xori_vec(unsigned vece, TCGv_vec d, TCGv_vec a, int64_t b) +{ + TCGv_vec t = tcg_temp_new_vec_matching(d); + tcg_gen_dupi_vec(vece, t, b); + tcg_gen_xor_vec(vece, d, a, t); + tcg_temp_free_vec(t); +} + +void tcg_gen_gvec_xori(unsigned vece, uint32_t dofs, uint32_t aofs, + int64_t c, uint32_t oprsz, uint32_t maxsz) +{ + static const GVecGen2i g = { + .fni8 = tcg_gen_xori_i64, + .fniv = tcg_gen_xori_vec, + .fnoi = gen_helper_gvec_xori, + .opc = INDEX_op_xor_vec, + .prefer_i64 = TCG_TARGET_REG_BITS == 64, + .vece = MO_64 + }; + tcg_gen_gvec_2i(dofs, aofs, oprsz, maxsz, c, &g); +} + +static void tcg_gen_ori_vec(unsigned vece, TCGv_vec d, TCGv_vec a, int64_t b) +{ + TCGv_vec t = tcg_temp_new_vec_matching(d); + tcg_gen_dupi_vec(vece, t, b); + tcg_gen_or_vec(vece, d, a, t); + tcg_temp_free_vec(t); +} + +void tcg_gen_gvec_ori(unsigned vece, uint32_t dofs, uint32_t aofs, + int64_t c, uint32_t oprsz, uint32_t maxsz) +{ + static const GVecGen2i g = { + .fni8 = tcg_gen_ori_i64, + .fniv = tcg_gen_ori_vec, + .fnoi = gen_helper_gvec_ori, + .opc = INDEX_op_or_vec, + .prefer_i64 = TCG_TARGET_REG_BITS == 64, + .vece = MO_64 + }; + tcg_gen_gvec_2i(dofs, aofs, oprsz, maxsz, c, &g); +} + void tcg_gen_vec_shl8i_i64(TCGv_i64 d, TCGv_i64 a, int64_t c) { uint64_t mask = ((0xff << c) & 0xff) * (-1ull / 0xff);