From nobody Tue Feb 10 13:36:55 2026 Delivered-To: importer@patchew.org Authentication-Results: mx.zohomail.com; spf=pass (zohomail.com: domain of gnu.org designates 209.51.188.17 as permitted sender) smtp.mailfrom=qemu-devel-bounces+importer=patchew.org@nongnu.org Return-Path: Received: from lists.gnu.org (lists.gnu.org [209.51.188.17]) by mx.zohomail.com with SMTPS id 164450157798298.3056874953428; Thu, 10 Feb 2022 05:59:37 -0800 (PST) Received: from localhost ([::1]:55198 helo=lists1p.gnu.org) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1nI9yz-00086f-Mt for importer@patchew.org; Thu, 10 Feb 2022 08:59:37 -0500 Received: from eggs.gnu.org ([209.51.188.92]:33800) by lists.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1nI8ow-0004mr-6x; Thu, 10 Feb 2022 07:45:17 -0500 Received: from [187.72.171.209] (port=6986 helo=outlook.eldorado.org.br) by eggs.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1nI8or-0000nd-22; Thu, 10 Feb 2022 07:45:09 -0500 Received: from p9ibm ([10.10.71.235]) by outlook.eldorado.org.br over TLS secured channel with Microsoft SMTPSVC(8.5.9600.16384); Thu, 10 Feb 2022 09:35:31 -0300 Received: from eldorado.org.br (unknown [10.10.70.45]) by p9ibm (Postfix) with ESMTP id 278E1800172; Thu, 10 Feb 2022 09:35:31 -0300 (-03) From: matheus.ferst@eldorado.org.br To: qemu-devel@nongnu.org, qemu-ppc@nongnu.org Subject: [PATCH v3 24/37] tcg/tcg-op-gvec.c: Introduce tcg_gen_gvec_4i Date: Thu, 10 Feb 2022 09:34:34 -0300 Message-Id: <20220210123447.3933301-25-matheus.ferst@eldorado.org.br> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20220210123447.3933301-1-matheus.ferst@eldorado.org.br> References: <20220210123447.3933301-1-matheus.ferst@eldorado.org.br> MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable X-OriginalArrivalTime: 10 Feb 2022 12:35:31.0519 (UTC) FILETIME=[B1331CF0:01D81E7A] X-Host-Lookup-Failed: Reverse DNS lookup failed for 187.72.171.209 (failed) Received-SPF: pass (zohomail.com: domain of gnu.org designates 209.51.188.17 as permitted sender) client-ip=209.51.188.17; envelope-from=qemu-devel-bounces+importer=patchew.org@nongnu.org; helo=lists.gnu.org; Received-SPF: pass client-ip=187.72.171.209; envelope-from=matheus.ferst@eldorado.org.br; helo=outlook.eldorado.org.br X-Spam_score_int: 0 X-Spam_score: -0.1 X-Spam_bar: / X-Spam_report: (-0.1 / 5.0 requ) BAYES_00=-1.9, OBFU_UNSUB_UL=1, PDS_HP_HELO_NORDNS=0.001, RDNS_NONE=0.793, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, T_SCC_BODY_TEXT_LINE=-0.01 autolearn=no autolearn_force=no X-Spam_action: no action X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: danielhb413@gmail.com, richard.henderson@linaro.org, groug@kaod.org, clg@kaod.org, Matheus Ferst , david@gibson.dropbear.id.au Errors-To: qemu-devel-bounces+importer=patchew.org@nongnu.org Sender: "Qemu-devel" X-ZM-MESSAGEID: 1644501608775100001 Content-Type: text/plain; charset="utf-8" From: Matheus Ferst Following the implementation of tcg_gen_gvec_3i, add a four-vector and immediate operand expansion method. Signed-off-by: Matheus Ferst --- include/tcg/tcg-op-gvec.h | 22 ++++++ tcg/tcg-op-gvec.c | 146 ++++++++++++++++++++++++++++++++++++++ 2 files changed, 168 insertions(+) diff --git a/include/tcg/tcg-op-gvec.h b/include/tcg/tcg-op-gvec.h index da55fed870..28cafbcc5c 100644 --- a/include/tcg/tcg-op-gvec.h +++ b/include/tcg/tcg-op-gvec.h @@ -218,6 +218,25 @@ typedef struct { bool write_aofs; } GVecGen4; =20 +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, TCGv_i64, int64_t); + void (*fni4)(TCGv_i32, TCGv_i32, TCGv_i32, TCGv_i32, int32_t); + /* Expand inline with a host vector type. */ + void (*fniv)(unsigned, TCGv_vec, TCGv_vec, TCGv_vec, TCGv_vec, int64_t= ); + /* Expand out-of-line helper w/descriptor, data in descriptor. */ + gen_helper_gvec_4 *fno; + /* The optional opcodes, if any, utilized by .fniv. */ + const TCGOpcode *opt_opc; + /* The vector element size, if applicable. */ + uint8_t vece; + /* Prefer i64 to v64. */ + bool prefer_i64; +} GVecGen4i; + 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, @@ -231,6 +250,9 @@ void tcg_gen_gvec_3i(uint32_t dofs, uint32_t aofs, uint= 32_t bofs, const GVecGen3i *); void tcg_gen_gvec_4(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t = cofs, uint32_t oprsz, uint32_t maxsz, const GVecGen4 *); +void tcg_gen_gvec_4i(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t= cofs, + uint32_t oprsz, uint32_t maxsz, int64_t c, + const GVecGen4i *); =20 /* Expand a specific vector operation. */ =20 diff --git a/tcg/tcg-op-gvec.c b/tcg/tcg-op-gvec.c index ffe55e908f..079a761b04 100644 --- a/tcg/tcg-op-gvec.c +++ b/tcg/tcg-op-gvec.c @@ -836,6 +836,30 @@ static void expand_4_i32(uint32_t dofs, uint32_t aofs,= uint32_t bofs, tcg_temp_free_i32(t0); } =20 +static void expand_4i_i32(uint32_t dofs, uint32_t aofs, uint32_t bofs, + uint32_t cofs, uint32_t oprsz, int32_t c, + void (*fni)(TCGv_i32, TCGv_i32, TCGv_i32, TCGv_i= 32, + int32_t)) +{ + TCGv_i32 t0 =3D tcg_temp_new_i32(); + TCGv_i32 t1 =3D tcg_temp_new_i32(); + TCGv_i32 t2 =3D tcg_temp_new_i32(); + TCGv_i32 t3 =3D tcg_temp_new_i32(); + uint32_t i; + + for (i =3D 0; i < oprsz; i +=3D 4) { + tcg_gen_ld_i32(t1, cpu_env, aofs + i); + tcg_gen_ld_i32(t2, cpu_env, bofs + i); + tcg_gen_ld_i32(t3, cpu_env, cofs + i); + fni(t0, t1, t2, t3, c); + tcg_gen_st_i32(t0, cpu_env, dofs + i); + } + tcg_temp_free_i32(t3); + tcg_temp_free_i32(t2); + tcg_temp_free_i32(t1); + tcg_temp_free_i32(t0); +} + /* Expand OPSZ bytes worth of two-operand operations using i64 elements. = */ static void expand_2_i64(uint32_t dofs, uint32_t aofs, uint32_t oprsz, bool load_dest, void (*fni)(TCGv_i64, TCGv_i64)) @@ -971,6 +995,30 @@ static void expand_4_i64(uint32_t dofs, uint32_t aofs,= uint32_t bofs, tcg_temp_free_i64(t0); } =20 +static void expand_4i_i64(uint32_t dofs, uint32_t aofs, uint32_t bofs, + uint32_t cofs, uint32_t oprsz, int64_t c, + void (*fni)(TCGv_i64, TCGv_i64, TCGv_i64, TCGv_i= 64, + int64_t)) +{ + TCGv_i64 t0 =3D tcg_temp_new_i64(); + TCGv_i64 t1 =3D tcg_temp_new_i64(); + TCGv_i64 t2 =3D tcg_temp_new_i64(); + TCGv_i64 t3 =3D tcg_temp_new_i64(); + uint32_t i; + + for (i =3D 0; i < oprsz; i +=3D 8) { + tcg_gen_ld_i64(t1, cpu_env, aofs + i); + tcg_gen_ld_i64(t2, cpu_env, bofs + i); + tcg_gen_ld_i64(t3, cpu_env, cofs + i); + fni(t0, t1, t2, t3, c); + tcg_gen_st_i64(t0, cpu_env, dofs + i); + } + tcg_temp_free_i64(t3); + tcg_temp_free_i64(t2); + tcg_temp_free_i64(t1); + tcg_temp_free_i64(t0); +} + /* Expand OPSZ bytes worth of two-operand operations using host vectors. = */ static void expand_2_vec(unsigned vece, uint32_t dofs, uint32_t aofs, uint32_t oprsz, uint32_t tysz, TCGType type, @@ -1121,6 +1169,35 @@ static void expand_4_vec(unsigned vece, uint32_t dof= s, uint32_t aofs, tcg_temp_free_vec(t0); } =20 +/* + * Expand OPSZ bytes worth of four-vector operands and an immediate operand + * using host vectors. + */ +static void expand_4i_vec(unsigned vece, uint32_t dofs, uint32_t aofs, + uint32_t bofs, uint32_t cofs, uint32_t oprsz, + uint32_t tysz, TCGType type, int64_t c, + void (*fni)(unsigned, TCGv_vec, TCGv_vec, + TCGv_vec, TCGv_vec, int64_t)) +{ + TCGv_vec t0 =3D tcg_temp_new_vec(type); + TCGv_vec t1 =3D tcg_temp_new_vec(type); + TCGv_vec t2 =3D tcg_temp_new_vec(type); + TCGv_vec t3 =3D tcg_temp_new_vec(type); + uint32_t i; + + for (i =3D 0; i < oprsz; i +=3D tysz) { + tcg_gen_ld_vec(t1, cpu_env, aofs + i); + tcg_gen_ld_vec(t2, cpu_env, bofs + i); + tcg_gen_ld_vec(t3, cpu_env, cofs + i); + fni(vece, t0, t1, t2, t3, c); + tcg_gen_st_vec(t0, cpu_env, dofs + i); + } + tcg_temp_free_vec(t3); + tcg_temp_free_vec(t2); + tcg_temp_free_vec(t1); + tcg_temp_free_vec(t0); +} + /* Expand a vector two-operand operation. */ void tcg_gen_gvec_2(uint32_t dofs, uint32_t aofs, uint32_t oprsz, uint32_t maxsz, const GVecGen2 *g) @@ -1533,6 +1610,75 @@ void tcg_gen_gvec_4(uint32_t dofs, uint32_t aofs, ui= nt32_t bofs, uint32_t cofs, } } =20 +/* Expand a vector four-operand operation. */ +void tcg_gen_gvec_4i(uint32_t dofs, uint32_t aofs, uint32_t bofs, uint32_t= cofs, + uint32_t oprsz, uint32_t maxsz, int64_t c, + const GVecGen4i *g) +{ + const TCGOpcode *this_list =3D g->opt_opc ? : vecop_list_empty; + const TCGOpcode *hold_list =3D tcg_swap_vecop_list(this_list); + TCGType type; + uint32_t some; + + check_size_align(oprsz, maxsz, dofs | aofs | bofs | cofs); + check_overlap_4(dofs, aofs, bofs, cofs, maxsz); + + type =3D 0; + if (g->fniv) { + type =3D choose_vector_type(g->opt_opc, g->vece, oprsz, g->prefer_= i64); + } + switch (type) { + case TCG_TYPE_V256: + /* + * Recall that ARM SVE allows vector sizes that are not a + * power of 2, but always a multiple of 16. The intent is + * that e.g. size =3D=3D 80 would be expanded with 2x32 + 1x16. + */ + some =3D QEMU_ALIGN_DOWN(oprsz, 32); + expand_4i_vec(g->vece, dofs, aofs, bofs, cofs, some, + 32, TCG_TYPE_V256, c, g->fniv); + if (some =3D=3D oprsz) { + break; + } + dofs +=3D some; + aofs +=3D some; + bofs +=3D some; + cofs +=3D some; + oprsz -=3D some; + maxsz -=3D some; + /* fallthru */ + case TCG_TYPE_V128: + expand_4i_vec(g->vece, dofs, aofs, bofs, cofs, oprsz, + 16, TCG_TYPE_V128, c, g->fniv); + break; + case TCG_TYPE_V64: + expand_4i_vec(g->vece, dofs, aofs, bofs, cofs, oprsz, + 8, TCG_TYPE_V64, c, g->fniv); + break; + + case 0: + if (g->fni8 && check_size_impl(oprsz, 8)) { + expand_4i_i64(dofs, aofs, bofs, cofs, oprsz, c, g->fni8); + } else if (g->fni4 && check_size_impl(oprsz, 4)) { + expand_4i_i32(dofs, aofs, bofs, cofs, oprsz, c, g->fni4); + } else { + assert(g->fno !=3D NULL); + tcg_gen_gvec_4_ool(dofs, aofs, bofs, cofs, + oprsz, maxsz, c, g->fno); + oprsz =3D maxsz; + } + break; + + default: + g_assert_not_reached(); + } + tcg_swap_vecop_list(hold_list); + + if (oprsz < maxsz) { + expand_clr(dofs + oprsz, maxsz - oprsz); + } +} + /* * Expand specific vector operations. */ --=20 2.31.1