From patchwork Tue Apr 25 07:02:19 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: gaosong X-Patchwork-Id: 1773238 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@legolas.ozlabs.org Authentication-Results: legolas.ozlabs.org; spf=pass (sender SPF authorized) smtp.mailfrom=nongnu.org (client-ip=209.51.188.17; helo=lists.gnu.org; envelope-from=qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org; receiver=) Received: from lists.gnu.org (lists.gnu.org [209.51.188.17]) (using TLSv1.2 with cipher ECDHE-ECDSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by legolas.ozlabs.org (Postfix) with ESMTPS id 4Q5Ckm0l6qz23tF for ; Tue, 25 Apr 2023 17:08:00 +1000 (AEST) Received: from localhost ([::1] helo=lists1p.gnu.org) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1prChw-0001NL-Ht; Tue, 25 Apr 2023 03:03:24 -0400 Received: from eggs.gnu.org ([2001:470:142:3::10]) by lists.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1prCht-0001Ls-FK for qemu-devel@nongnu.org; Tue, 25 Apr 2023 03:03:21 -0400 Received: from mail.loongson.cn ([114.242.206.163] helo=loongson.cn) by eggs.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1prChp-00084U-Po for qemu-devel@nongnu.org; Tue, 25 Apr 2023 03:03:21 -0400 Received: from loongson.cn (unknown [10.2.5.185]) by gateway (Coremail) with SMTP id _____8CxE_Aoe0dkg18AAA--.677S3; Tue, 25 Apr 2023 15:03:04 +0800 (CST) Received: from localhost.localdomain (unknown [10.2.5.185]) by localhost.localdomain (Coremail) with SMTP id AQAAf8Cxeb0Ye0dk3Eo6AA--.4591S17; Tue, 25 Apr 2023 15:03:03 +0800 (CST) From: Song Gao To: qemu-devel@nongnu.org Cc: richard.henderson@linaro.org, gaosong@loongson.cn Subject: [RFC PATCH v4 15/44] target/loongarch: Implement vmadd/vmsub/vmaddw{ev/od} Date: Tue, 25 Apr 2023 15:02:19 +0800 Message-Id: <20230425070248.2550028-16-gaosong@loongson.cn> X-Mailer: git-send-email 2.31.1 In-Reply-To: <20230425070248.2550028-1-gaosong@loongson.cn> References: <20230425070248.2550028-1-gaosong@loongson.cn> MIME-Version: 1.0 X-CM-TRANSID: AQAAf8Cxeb0Ye0dk3Eo6AA--.4591S17 X-CM-SenderInfo: 5jdr20tqj6z05rqj20fqof0/ X-Coremail-Antispam: 1Uk129KBjvAXoWfCrykWrWxKr4Dtw48Cry5CFg_yoW5GFW8uo W3Gw4UArWxKr13Cr1UW39rXFWxtrW0va4DJ34jvrsFqFyrAFy7tas8K3ZxJaykArsI9F9x JFy7AF45J3Z0qrWDn29KB7ZKAUJUUUU8529EdanIXcx71UUUUU7KY7ZEXasCq-sGcSsGvf J3Ic02F40EFcxC0VAKzVAqx4xG6I80ebIjqfuFe4nvWSU5nxnvy29KBjDU0xBIdaVrnRJU UUqm1xkIjI8I6I8E6xAIw20EY4v20xvaj40_Wr0E3s1l8cAvFVAK0II2c7xJM28CjxkF64 kEwVA0rcxSw2x7M28EF7xvwVC0I7IYx2IY67AKxVW7JVWDJwA2z4x0Y4vE2Ix0cI8IcVCY 1x0267AKxVWxJVW8Jr1l84ACjcxK6I8E87Iv67AKxVW8Jr0_Cr1UM28EF7xvwVC2z280aV CY1x0267AKxVW8Jr0_Cr1UM2AIxVAIcxkEcVAq07x20xvEncxIr21l57IF6xkI12xvs2x2 6I8E6xACxx1l5I8CrVACY4xI64kE6c02F40Ex7xfMcIj6x8ErcxFaVAv8VWrMcvjeVCFs4 IE7xkEbVWUJVW8JwACjcxG0xvY0x0EwIxGrwCF04k20xvY0x0EwIxGrwCF04k20xvE74AG Y7Cv6cx26rWl4I8I3I0E4IkC6x0Yz7v_Jr0_Gr1lx2IqxVAqx4xG67AKxVWUJVWUGwC20s 026x8GjcxK67AKxVWUGVWUWwC2zVAF1VAY17CE14v26r1Y6r17MIIYrxkI7VAKI48JMIIF 0xvE2Ix0cI8IcVAFwI0_Xr0_Ar1lIxAIcVC0I7IYx2IY6xkF7I0E14v26r4j6F4UMIIF0x vE42xK8VAvwI8IcIk0rVWUJVWUCwCI42IY6I8E87Iv67AKxVW8JVWxJwCI42IY6I8E87Iv 6xkF7I0E14v26r4j6r4UJbIYCTnIWIevJa73UjIFyTuYvj4RC_MaUUUUU Received-SPF: pass client-ip=114.242.206.163; envelope-from=gaosong@loongson.cn; helo=loongson.cn X-Spam_score_int: -18 X-Spam_score: -1.9 X-Spam_bar: - X-Spam_report: (-1.9 / 5.0 requ) BAYES_00=-1.9, SPF_HELO_PASS=-0.001, SPF_PASS=-0.001, T_SCC_BODY_TEXT_LINE=-0.01 autolearn=ham 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: , Errors-To: qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org Sender: qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org This patch includes: - VMADD.{B/H/W/D}; - VMSUB.{B/H/W/D}; - VMADDW{EV/OD}.{H.B/W.H/D.W/Q.D}[U]; - VMADDW{EV/OD}.{H.BU.B/W.HU.H/D.WU.W/Q.DU.D}. Reviewed-by: Richard Henderson Signed-off-by: Song Gao --- target/loongarch/disas.c | 34 ++ target/loongarch/helper.h | 30 + target/loongarch/insn_trans/trans_lsx.c.inc | 612 ++++++++++++++++++++ target/loongarch/insns.decode | 34 ++ target/loongarch/lsx_helper.c | 107 ++++ 5 files changed, 817 insertions(+) diff --git a/target/loongarch/disas.c b/target/loongarch/disas.c index 48e6ef5309..980e6e6375 100644 --- a/target/loongarch/disas.c +++ b/target/loongarch/disas.c @@ -1010,3 +1010,37 @@ INSN_LSX(vmulwod_h_bu_b, vvv) INSN_LSX(vmulwod_w_hu_h, vvv) INSN_LSX(vmulwod_d_wu_w, vvv) INSN_LSX(vmulwod_q_du_d, vvv) + +INSN_LSX(vmadd_b, vvv) +INSN_LSX(vmadd_h, vvv) +INSN_LSX(vmadd_w, vvv) +INSN_LSX(vmadd_d, vvv) +INSN_LSX(vmsub_b, vvv) +INSN_LSX(vmsub_h, vvv) +INSN_LSX(vmsub_w, vvv) +INSN_LSX(vmsub_d, vvv) + +INSN_LSX(vmaddwev_h_b, vvv) +INSN_LSX(vmaddwev_w_h, vvv) +INSN_LSX(vmaddwev_d_w, vvv) +INSN_LSX(vmaddwev_q_d, vvv) +INSN_LSX(vmaddwod_h_b, vvv) +INSN_LSX(vmaddwod_w_h, vvv) +INSN_LSX(vmaddwod_d_w, vvv) +INSN_LSX(vmaddwod_q_d, vvv) +INSN_LSX(vmaddwev_h_bu, vvv) +INSN_LSX(vmaddwev_w_hu, vvv) +INSN_LSX(vmaddwev_d_wu, vvv) +INSN_LSX(vmaddwev_q_du, vvv) +INSN_LSX(vmaddwod_h_bu, vvv) +INSN_LSX(vmaddwod_w_hu, vvv) +INSN_LSX(vmaddwod_d_wu, vvv) +INSN_LSX(vmaddwod_q_du, vvv) +INSN_LSX(vmaddwev_h_bu_b, vvv) +INSN_LSX(vmaddwev_w_hu_h, vvv) +INSN_LSX(vmaddwev_d_wu_w, vvv) +INSN_LSX(vmaddwev_q_du_d, vvv) +INSN_LSX(vmaddwod_h_bu_b, vvv) +INSN_LSX(vmaddwod_w_hu_h, vvv) +INSN_LSX(vmaddwod_d_wu_w, vvv) +INSN_LSX(vmaddwod_q_du_d, vvv) diff --git a/target/loongarch/helper.h b/target/loongarch/helper.h index 830d8cbe62..8cf9620702 100644 --- a/target/loongarch/helper.h +++ b/target/loongarch/helper.h @@ -273,3 +273,33 @@ DEF_HELPER_FLAGS_4(vmulwev_d_wu_w, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(vmulwod_h_bu_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(vmulwod_w_hu_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) DEF_HELPER_FLAGS_4(vmulwod_d_wu_w, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(vmadd_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmadd_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmadd_w, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmadd_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmsub_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmsub_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmsub_w, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmsub_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(vmaddwev_h_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmaddwev_w_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmaddwev_d_w, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmaddwod_h_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmaddwod_w_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmaddwod_d_w, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(vmaddwev_h_bu, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmaddwev_w_hu, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmaddwev_d_wu, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmaddwod_h_bu, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmaddwod_w_hu, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmaddwod_d_wu, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) + +DEF_HELPER_FLAGS_4(vmaddwev_h_bu_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmaddwev_w_hu_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmaddwev_d_wu_w, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmaddwod_h_bu_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmaddwod_w_hu_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_4(vmaddwod_d_wu_w, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32) diff --git a/target/loongarch/insn_trans/trans_lsx.c.inc b/target/loongarch/insn_trans/trans_lsx.c.inc index 9a36f942e2..400c3a0339 100644 --- a/target/loongarch/insn_trans/trans_lsx.c.inc +++ b/target/loongarch/insn_trans/trans_lsx.c.inc @@ -2064,3 +2064,615 @@ static void do_vmulwod_u_s(unsigned vece, uint32_t vd_ofs, uint32_t vj_ofs, TRANS(vmulwod_h_bu_b, gvec_vvv, MO_8, do_vmulwod_u_s) TRANS(vmulwod_w_hu_h, gvec_vvv, MO_16, do_vmulwod_u_s) TRANS(vmulwod_d_wu_w, gvec_vvv, MO_32, do_vmulwod_u_s) + +static void gen_vmadd(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b) +{ + TCGv_vec t1; + + t1 = tcg_temp_new_vec_matching(t); + tcg_gen_mul_vec(vece, t1, a, b); + tcg_gen_add_vec(vece, t, t, t1); +} + +static void gen_vmadd_w(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b) +{ + TCGv_i32 t1; + + t1 = tcg_temp_new_i32(); + tcg_gen_mul_i32(t1, a, b); + tcg_gen_add_i32(t, t, t1); +} + +static void gen_vmadd_d(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b) +{ + TCGv_i64 t1; + + t1 = tcg_temp_new_i64(); + tcg_gen_mul_i64(t1, a, b); + tcg_gen_add_i64(t, t, t1); +} + +static void do_vmadd(unsigned vece, uint32_t vd_ofs, uint32_t vj_ofs, + uint32_t vk_ofs, uint32_t oprsz, uint32_t maxsz) +{ + static const TCGOpcode vecop_list[] = { + INDEX_op_mul_vec, INDEX_op_add_vec, 0 + }; + static const GVecGen3 op[4] = { + { + .fniv = gen_vmadd, + .fno = gen_helper_vmadd_b, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_8 + }, + { + .fniv = gen_vmadd, + .fno = gen_helper_vmadd_h, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_16 + }, + { + .fni4 = gen_vmadd_w, + .fniv = gen_vmadd, + .fno = gen_helper_vmadd_w, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_32 + }, + { + .fni8 = gen_vmadd_d, + .fniv = gen_vmadd, + .fno = gen_helper_vmadd_d, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_64 + }, + }; + + tcg_gen_gvec_3(vd_ofs, vj_ofs, vk_ofs, oprsz, maxsz, &op[vece]); +} + +TRANS(vmadd_b, gvec_vvv, MO_8, do_vmadd) +TRANS(vmadd_h, gvec_vvv, MO_16, do_vmadd) +TRANS(vmadd_w, gvec_vvv, MO_32, do_vmadd) +TRANS(vmadd_d, gvec_vvv, MO_64, do_vmadd) + +static void gen_vmsub(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b) +{ + TCGv_vec t1; + + t1 = tcg_temp_new_vec_matching(t); + tcg_gen_mul_vec(vece, t1, a, b); + tcg_gen_sub_vec(vece, t, t, t1); +} + +static void gen_vmsub_w(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b) +{ + TCGv_i32 t1; + + t1 = tcg_temp_new_i32(); + tcg_gen_mul_i32(t1, a, b); + tcg_gen_sub_i32(t, t, t1); +} + +static void gen_vmsub_d(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b) +{ + TCGv_i64 t1; + + t1 = tcg_temp_new_i64(); + tcg_gen_mul_i64(t1, a, b); + tcg_gen_sub_i64(t, t, t1); +} + +static void do_vmsub(unsigned vece, uint32_t vd_ofs, uint32_t vj_ofs, + uint32_t vk_ofs, uint32_t oprsz, uint32_t maxsz) +{ + static const TCGOpcode vecop_list[] = { + INDEX_op_mul_vec, INDEX_op_sub_vec, 0 + }; + static const GVecGen3 op[4] = { + { + .fniv = gen_vmsub, + .fno = gen_helper_vmsub_b, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_8 + }, + { + .fniv = gen_vmsub, + .fno = gen_helper_vmsub_h, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_16 + }, + { + .fni4 = gen_vmsub_w, + .fniv = gen_vmsub, + .fno = gen_helper_vmsub_w, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_32 + }, + { + .fni8 = gen_vmsub_d, + .fniv = gen_vmsub, + .fno = gen_helper_vmsub_d, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_64 + }, + }; + + tcg_gen_gvec_3(vd_ofs, vj_ofs, vk_ofs, oprsz, maxsz, &op[vece]); +} + +TRANS(vmsub_b, gvec_vvv, MO_8, do_vmsub) +TRANS(vmsub_h, gvec_vvv, MO_16, do_vmsub) +TRANS(vmsub_w, gvec_vvv, MO_32, do_vmsub) +TRANS(vmsub_d, gvec_vvv, MO_64, do_vmsub) + +static void gen_vmaddwev_s(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b) +{ + TCGv_vec t1, t2, t3; + int halfbits = 4 << vece; + + t1 = tcg_temp_new_vec_matching(a); + t2 = tcg_temp_new_vec_matching(b); + t3 = tcg_temp_new_vec_matching(t); + tcg_gen_shli_vec(vece, t1, a, halfbits); + tcg_gen_sari_vec(vece, t1, t1, halfbits); + tcg_gen_shli_vec(vece, t2, b, halfbits); + tcg_gen_sari_vec(vece, t2, t2, halfbits); + tcg_gen_mul_vec(vece, t3, t1, t2); + tcg_gen_add_vec(vece, t, t, t3); +} + +static void gen_vmaddwev_w_h(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b) +{ + TCGv_i32 t1; + + t1 = tcg_temp_new_i32(); + gen_vmulwev_w_h(t1, a, b); + tcg_gen_add_i32(t, t, t1); +} + +static void gen_vmaddwev_d_w(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b) +{ + TCGv_i64 t1; + + t1 = tcg_temp_new_i64(); + gen_vmulwev_d_w(t1, a, b); + tcg_gen_add_i64(t, t, t1); +} + +static void do_vmaddwev_s(unsigned vece, uint32_t vd_ofs, uint32_t vj_ofs, + uint32_t vk_ofs, uint32_t oprsz, uint32_t maxsz) +{ + static const TCGOpcode vecop_list[] = { + INDEX_op_shli_vec, INDEX_op_sari_vec, + INDEX_op_mul_vec, INDEX_op_add_vec, 0 + }; + static const GVecGen3 op[3] = { + { + .fniv = gen_vmaddwev_s, + .fno = gen_helper_vmaddwev_h_b, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_16 + }, + { + .fni4 = gen_vmaddwev_w_h, + .fniv = gen_vmaddwev_s, + .fno = gen_helper_vmaddwev_w_h, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_32 + }, + { + .fni8 = gen_vmaddwev_d_w, + .fniv = gen_vmaddwev_s, + .fno = gen_helper_vmaddwev_d_w, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_64 + }, + }; + + tcg_gen_gvec_3(vd_ofs, vj_ofs, vk_ofs, oprsz, maxsz, &op[vece]); +} + +TRANS(vmaddwev_h_b, gvec_vvv, MO_8, do_vmaddwev_s) +TRANS(vmaddwev_w_h, gvec_vvv, MO_16, do_vmaddwev_s) +TRANS(vmaddwev_d_w, gvec_vvv, MO_32, do_vmaddwev_s) + +#define VMADD_Q(NAME, FN, idx1, idx2) \ +static bool trans_## NAME (DisasContext *ctx, arg_vvv *a) \ +{ \ + TCGv_i64 rh, rl, arg1, arg2, th, tl; \ + \ + rh = tcg_temp_new_i64(); \ + rl = tcg_temp_new_i64(); \ + arg1 = tcg_temp_new_i64(); \ + arg2 = tcg_temp_new_i64(); \ + th = tcg_temp_new_i64(); \ + tl = tcg_temp_new_i64(); \ + \ + get_vreg64(arg1, a->vj, idx1); \ + get_vreg64(arg2, a->vk, idx2); \ + get_vreg64(rh, a->vd, 1); \ + get_vreg64(rl, a->vd, 0); \ + \ + tcg_gen_## FN ##_i64(tl, th, arg1, arg2); \ + tcg_gen_add2_i64(rl, rh, rl, rh, tl, th); \ + \ + set_vreg64(rh, a->vd, 1); \ + set_vreg64(rl, a->vd, 0); \ + \ + return true; \ +} + +VMADD_Q(vmaddwev_q_d, muls2, 0, 0) +VMADD_Q(vmaddwod_q_d, muls2, 1, 1) +VMADD_Q(vmaddwev_q_du, mulu2, 0, 0) +VMADD_Q(vmaddwod_q_du, mulu2, 1, 1) +VMADD_Q(vmaddwev_q_du_d, mulus2, 0, 0) +VMADD_Q(vmaddwod_q_du_d, mulus2, 1, 1) + +static void gen_vmaddwod_s(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b) +{ + TCGv_vec t1, t2, t3; + int halfbits = 4 << vece; + + t1 = tcg_temp_new_vec_matching(a); + t2 = tcg_temp_new_vec_matching(b); + t3 = tcg_temp_new_vec_matching(t); + tcg_gen_sari_vec(vece, t1, a, halfbits); + tcg_gen_sari_vec(vece, t2, b, halfbits); + tcg_gen_mul_vec(vece, t3, t1, t2); + tcg_gen_add_vec(vece, t, t, t3); +} + +static void gen_vmaddwod_w_h(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b) +{ + TCGv_i32 t1; + + t1 = tcg_temp_new_i32(); + gen_vmulwod_w_h(t1, a, b); + tcg_gen_add_i32(t, t, t1); +} + +static void gen_vmaddwod_d_w(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b) +{ + TCGv_i64 t1; + + t1 = tcg_temp_new_i64(); + gen_vmulwod_d_w(t1, a, b); + tcg_gen_add_i64(t, t, t1); +} + +static void do_vmaddwod_s(unsigned vece, uint32_t vd_ofs, uint32_t vj_ofs, + uint32_t vk_ofs, uint32_t oprsz, uint32_t maxsz) +{ + static const TCGOpcode vecop_list[] = { + INDEX_op_sari_vec, INDEX_op_mul_vec, INDEX_op_add_vec, 0 + }; + static const GVecGen3 op[3] = { + { + .fniv = gen_vmaddwod_s, + .fno = gen_helper_vmaddwod_h_b, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_16 + }, + { + .fni4 = gen_vmaddwod_w_h, + .fniv = gen_vmaddwod_s, + .fno = gen_helper_vmaddwod_w_h, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_32 + }, + { + .fni8 = gen_vmaddwod_d_w, + .fniv = gen_vmaddwod_s, + .fno = gen_helper_vmaddwod_d_w, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_64 + }, + }; + + tcg_gen_gvec_3(vd_ofs, vj_ofs, vk_ofs, oprsz, maxsz, &op[vece]); +} + +TRANS(vmaddwod_h_b, gvec_vvv, MO_8, do_vmaddwod_s) +TRANS(vmaddwod_w_h, gvec_vvv, MO_16, do_vmaddwod_s) +TRANS(vmaddwod_d_w, gvec_vvv, MO_32, do_vmaddwod_s) + +static void gen_vmaddwev_u(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b) +{ + TCGv_vec t1, t2, mask; + + t1 = tcg_temp_new_vec_matching(t); + t2 = tcg_temp_new_vec_matching(b); + mask = tcg_constant_vec_matching(t, vece, MAKE_64BIT_MASK(0, 4 << vece)); + tcg_gen_and_vec(vece, t1, a, mask); + tcg_gen_and_vec(vece, t2, b, mask); + tcg_gen_mul_vec(vece, t1, t1, t2); + tcg_gen_add_vec(vece, t, t, t1); +} + +static void gen_vmaddwev_w_hu(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b) +{ + TCGv_i32 t1; + + t1 = tcg_temp_new_i32(); + gen_vmulwev_w_hu(t1, a, b); + tcg_gen_add_i32(t, t, t1); +} + +static void gen_vmaddwev_d_wu(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b) +{ + TCGv_i64 t1; + + t1 = tcg_temp_new_i64(); + gen_vmulwev_d_wu(t1, a, b); + tcg_gen_add_i64(t, t, t1); +} + +static void do_vmaddwev_u(unsigned vece, uint32_t vd_ofs, uint32_t vj_ofs, + uint32_t vk_ofs, uint32_t oprsz, uint32_t maxsz) +{ + static const TCGOpcode vecop_list[] = { + INDEX_op_mul_vec, INDEX_op_add_vec, 0 + }; + static const GVecGen3 op[3] = { + { + .fniv = gen_vmaddwev_u, + .fno = gen_helper_vmaddwev_h_bu, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_16 + }, + { + .fni4 = gen_vmaddwev_w_hu, + .fniv = gen_vmaddwev_u, + .fno = gen_helper_vmaddwev_w_hu, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_32 + }, + { + .fni8 = gen_vmaddwev_d_wu, + .fniv = gen_vmaddwev_u, + .fno = gen_helper_vmaddwev_d_wu, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_64 + }, + }; + + tcg_gen_gvec_3(vd_ofs, vj_ofs, vk_ofs, oprsz, maxsz, &op[vece]); +} + +TRANS(vmaddwev_h_bu, gvec_vvv, MO_8, do_vmaddwev_u) +TRANS(vmaddwev_w_hu, gvec_vvv, MO_16, do_vmaddwev_u) +TRANS(vmaddwev_d_wu, gvec_vvv, MO_32, do_vmaddwev_u) + +static void gen_vmaddwod_u(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b) +{ + TCGv_vec t1, t2, t3; + int halfbits = 4 << vece; + + t1 = tcg_temp_new_vec_matching(a); + t2 = tcg_temp_new_vec_matching(b); + t3 = tcg_temp_new_vec_matching(t); + tcg_gen_shri_vec(vece, t1, a, halfbits); + tcg_gen_shri_vec(vece, t2, b, halfbits); + tcg_gen_mul_vec(vece, t3, t1, t2); + tcg_gen_add_vec(vece, t, t, t3); +} + +static void gen_vmaddwod_w_hu(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b) +{ + TCGv_i32 t1; + + t1 = tcg_temp_new_i32(); + gen_vmulwod_w_hu(t1, a, b); + tcg_gen_add_i32(t, t, t1); +} + +static void gen_vmaddwod_d_wu(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b) +{ + TCGv_i64 t1; + + t1 = tcg_temp_new_i64(); + gen_vmulwod_d_wu(t1, a, b); + tcg_gen_add_i64(t, t, t1); +} + +static void do_vmaddwod_u(unsigned vece, uint32_t vd_ofs, uint32_t vj_ofs, + uint32_t vk_ofs, uint32_t oprsz, uint32_t maxsz) +{ + static const TCGOpcode vecop_list[] = { + INDEX_op_shri_vec, INDEX_op_mul_vec, INDEX_op_add_vec, 0 + }; + static const GVecGen3 op[3] = { + { + .fniv = gen_vmaddwod_u, + .fno = gen_helper_vmaddwod_h_bu, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_16 + }, + { + .fni4 = gen_vmaddwod_w_hu, + .fniv = gen_vmaddwod_u, + .fno = gen_helper_vmaddwod_w_hu, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_32 + }, + { + .fni8 = gen_vmaddwod_d_wu, + .fniv = gen_vmaddwod_u, + .fno = gen_helper_vmaddwod_d_wu, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_64 + }, + }; + + tcg_gen_gvec_3(vd_ofs, vj_ofs, vk_ofs, oprsz, maxsz, &op[vece]); +} + +TRANS(vmaddwod_h_bu, gvec_vvv, MO_8, do_vmaddwod_u) +TRANS(vmaddwod_w_hu, gvec_vvv, MO_16, do_vmaddwod_u) +TRANS(vmaddwod_d_wu, gvec_vvv, MO_32, do_vmaddwod_u) + +static void gen_vmaddwev_u_s(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b) +{ + TCGv_vec t1, t2, mask; + int halfbits = 4 << vece; + + t1 = tcg_temp_new_vec_matching(a); + t2 = tcg_temp_new_vec_matching(b); + mask = tcg_constant_vec_matching(t, vece, MAKE_64BIT_MASK(0, 4 << vece)); + tcg_gen_and_vec(vece, t1, a, mask); + tcg_gen_shli_vec(vece, t2, b, halfbits); + tcg_gen_sari_vec(vece, t2, t2, halfbits); + tcg_gen_mul_vec(vece, t1, t1, t2); + tcg_gen_add_vec(vece, t, t, t1); +} + +static void gen_vmaddwev_w_hu_h(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b) +{ + TCGv_i32 t1; + + t1 = tcg_temp_new_i32(); + gen_vmulwev_w_hu_h(t1, a, b); + tcg_gen_add_i32(t, t, t1); +} + +static void gen_vmaddwev_d_wu_w(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b) +{ + TCGv_i64 t1; + + t1 = tcg_temp_new_i64(); + gen_vmulwev_d_wu_w(t1, a, b); + tcg_gen_add_i64(t, t, t1); +} + +static void do_vmaddwev_u_s(unsigned vece, uint32_t vd_ofs, uint32_t vj_ofs, + uint32_t vk_ofs, uint32_t oprsz, uint32_t maxsz) +{ + static const TCGOpcode vecop_list[] = { + INDEX_op_shli_vec, INDEX_op_sari_vec, + INDEX_op_mul_vec, INDEX_op_add_vec, 0 + }; + static const GVecGen3 op[3] = { + { + .fniv = gen_vmaddwev_u_s, + .fno = gen_helper_vmaddwev_h_bu_b, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_16 + }, + { + .fni4 = gen_vmaddwev_w_hu_h, + .fniv = gen_vmaddwev_u_s, + .fno = gen_helper_vmaddwev_w_hu_h, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_32 + }, + { + .fni8 = gen_vmaddwev_d_wu_w, + .fniv = gen_vmaddwev_u_s, + .fno = gen_helper_vmaddwev_d_wu_w, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_64 + }, + }; + + tcg_gen_gvec_3(vd_ofs, vj_ofs, vk_ofs, oprsz, maxsz, &op[vece]); +} + +TRANS(vmaddwev_h_bu_b, gvec_vvv, MO_8, do_vmaddwev_u_s) +TRANS(vmaddwev_w_hu_h, gvec_vvv, MO_16, do_vmaddwev_u_s) +TRANS(vmaddwev_d_wu_w, gvec_vvv, MO_32, do_vmaddwev_u_s) + +static void gen_vmaddwod_u_s(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b) +{ + TCGv_vec t1, t2, t3; + int halfbits = 4 << vece; + + t1 = tcg_temp_new_vec_matching(a); + t2 = tcg_temp_new_vec_matching(b); + t3 = tcg_temp_new_vec_matching(t); + tcg_gen_shri_vec(vece, t1, a, halfbits); + tcg_gen_sari_vec(vece, t2, b, halfbits); + tcg_gen_mul_vec(vece, t3, t1, t2); + tcg_gen_add_vec(vece, t, t, t3); +} + +static void gen_vmaddwod_w_hu_h(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b) +{ + TCGv_i32 t1; + + t1 = tcg_temp_new_i32(); + gen_vmulwod_w_hu_h(t1, a, b); + tcg_gen_add_i32(t, t, t1); +} + +static void gen_vmaddwod_d_wu_w(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b) +{ + TCGv_i64 t1; + + t1 = tcg_temp_new_i64(); + gen_vmulwod_d_wu_w(t1, a, b); + tcg_gen_add_i64(t, t, t1); +} + +static void do_vmaddwod_u_s(unsigned vece, uint32_t vd_ofs, uint32_t vj_ofs, + uint32_t vk_ofs, uint32_t oprsz, uint32_t maxsz) +{ + static const TCGOpcode vecop_list[] = { + INDEX_op_shri_vec, INDEX_op_sari_vec, + INDEX_op_mul_vec, INDEX_op_add_vec, 0 + }; + static const GVecGen3 op[3] = { + { + .fniv = gen_vmaddwod_u_s, + .fno = gen_helper_vmaddwod_h_bu_b, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_16 + }, + { + .fni4 = gen_vmaddwod_w_hu_h, + .fniv = gen_vmaddwod_u_s, + .fno = gen_helper_vmaddwod_w_hu_h, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_32 + }, + { + .fni8 = gen_vmaddwod_d_wu_w, + .fniv = gen_vmaddwod_u_s, + .fno = gen_helper_vmaddwod_d_wu_w, + .load_dest = true, + .opt_opc = vecop_list, + .vece = MO_64 + }, + }; + + tcg_gen_gvec_3(vd_ofs, vj_ofs, vk_ofs, oprsz, maxsz, &op[vece]); +} + +TRANS(vmaddwod_h_bu_b, gvec_vvv, MO_8, do_vmaddwod_u_s) +TRANS(vmaddwod_w_hu_h, gvec_vvv, MO_16, do_vmaddwod_u_s) +TRANS(vmaddwod_d_wu_w, gvec_vvv, MO_32, do_vmaddwod_u_s) diff --git a/target/loongarch/insns.decode b/target/loongarch/insns.decode index 64e8042c9c..df23d4ee1e 100644 --- a/target/loongarch/insns.decode +++ b/target/loongarch/insns.decode @@ -706,3 +706,37 @@ vmulwod_h_bu_b 0111 00001010 00100 ..... ..... ..... @vvv vmulwod_w_hu_h 0111 00001010 00101 ..... ..... ..... @vvv vmulwod_d_wu_w 0111 00001010 00110 ..... ..... ..... @vvv vmulwod_q_du_d 0111 00001010 00111 ..... ..... ..... @vvv + +vmadd_b 0111 00001010 10000 ..... ..... ..... @vvv +vmadd_h 0111 00001010 10001 ..... ..... ..... @vvv +vmadd_w 0111 00001010 10010 ..... ..... ..... @vvv +vmadd_d 0111 00001010 10011 ..... ..... ..... @vvv +vmsub_b 0111 00001010 10100 ..... ..... ..... @vvv +vmsub_h 0111 00001010 10101 ..... ..... ..... @vvv +vmsub_w 0111 00001010 10110 ..... ..... ..... @vvv +vmsub_d 0111 00001010 10111 ..... ..... ..... @vvv + +vmaddwev_h_b 0111 00001010 11000 ..... ..... ..... @vvv +vmaddwev_w_h 0111 00001010 11001 ..... ..... ..... @vvv +vmaddwev_d_w 0111 00001010 11010 ..... ..... ..... @vvv +vmaddwev_q_d 0111 00001010 11011 ..... ..... ..... @vvv +vmaddwod_h_b 0111 00001010 11100 ..... ..... ..... @vvv +vmaddwod_w_h 0111 00001010 11101 ..... ..... ..... @vvv +vmaddwod_d_w 0111 00001010 11110 ..... ..... ..... @vvv +vmaddwod_q_d 0111 00001010 11111 ..... ..... ..... @vvv +vmaddwev_h_bu 0111 00001011 01000 ..... ..... ..... @vvv +vmaddwev_w_hu 0111 00001011 01001 ..... ..... ..... @vvv +vmaddwev_d_wu 0111 00001011 01010 ..... ..... ..... @vvv +vmaddwev_q_du 0111 00001011 01011 ..... ..... ..... @vvv +vmaddwod_h_bu 0111 00001011 01100 ..... ..... ..... @vvv +vmaddwod_w_hu 0111 00001011 01101 ..... ..... ..... @vvv +vmaddwod_d_wu 0111 00001011 01110 ..... ..... ..... @vvv +vmaddwod_q_du 0111 00001011 01111 ..... ..... ..... @vvv +vmaddwev_h_bu_b 0111 00001011 11000 ..... ..... ..... @vvv +vmaddwev_w_hu_h 0111 00001011 11001 ..... ..... ..... @vvv +vmaddwev_d_wu_w 0111 00001011 11010 ..... ..... ..... @vvv +vmaddwev_q_du_d 0111 00001011 11011 ..... ..... ..... @vvv +vmaddwod_h_bu_b 0111 00001011 11100 ..... ..... ..... @vvv +vmaddwod_w_hu_h 0111 00001011 11101 ..... ..... ..... @vvv +vmaddwod_d_wu_w 0111 00001011 11110 ..... ..... ..... @vvv +vmaddwod_q_du_d 0111 00001011 11111 ..... ..... ..... @vvv diff --git a/target/loongarch/lsx_helper.c b/target/loongarch/lsx_helper.c index 3b565277fd..8f82091302 100644 --- a/target/loongarch/lsx_helper.c +++ b/target/loongarch/lsx_helper.c @@ -446,3 +446,110 @@ DO_EVEN_U_S(vmulwev_d_wu_w, 64, D, UD, W, UW, DO_MUL) DO_ODD_U_S(vmulwod_h_bu_b, 16, H, UH, B, UB, DO_MUL) DO_ODD_U_S(vmulwod_w_hu_h, 32, W, UW, H, UH, DO_MUL) DO_ODD_U_S(vmulwod_d_wu_w, 64, D, UD, W, UW, DO_MUL) + +#define DO_MADD(a, b, c) (a + b * c) +#define DO_MSUB(a, b, c) (a - b * c) + +#define VMADDSUB(NAME, BIT, E, DO_OP) \ +void HELPER(NAME)(void *vd, void *vj, void *vk, uint32_t v) \ +{ \ + int i; \ + VReg *Vd = (VReg *)vd; \ + VReg *Vj = (VReg *)vj; \ + VReg *Vk = (VReg *)vk; \ + for (i = 0; i < LSX_LEN/BIT; i++) { \ + Vd->E(i) = DO_OP(Vd->E(i), Vj->E(i) ,Vk->E(i)); \ + } \ +} + +VMADDSUB(vmadd_b, 8, B, DO_MADD) +VMADDSUB(vmadd_h, 16, H, DO_MADD) +VMADDSUB(vmadd_w, 32, W, DO_MADD) +VMADDSUB(vmadd_d, 64, D, DO_MADD) +VMADDSUB(vmsub_b, 8, B, DO_MSUB) +VMADDSUB(vmsub_h, 16, H, DO_MSUB) +VMADDSUB(vmsub_w, 32, W, DO_MSUB) +VMADDSUB(vmsub_d, 64, D, DO_MSUB) + +#define VMADDWEV(NAME, BIT, E1, E2, DO_OP) \ +void HELPER(NAME)(void *vd, void *vj, void *vk, uint32_t v) \ +{ \ + int i; \ + VReg *Vd = (VReg *)vd; \ + VReg *Vj = (VReg *)vj; \ + VReg *Vk = (VReg *)vk; \ + typedef __typeof(Vd->E1(0)) TD; \ + \ + for (i = 0; i < LSX_LEN/BIT; i++) { \ + Vd->E1(i) += DO_OP((TD)Vj->E2(2 * i), (TD)Vk->E2(2 * i)); \ + } \ +} + +VMADDWEV(vmaddwev_h_b, 16, H, B, DO_MUL) +VMADDWEV(vmaddwev_w_h, 32, W, H, DO_MUL) +VMADDWEV(vmaddwev_d_w, 64, D, W, DO_MUL) +VMADDWEV(vmaddwev_h_bu, 16, UH, UB, DO_MUL) +VMADDWEV(vmaddwev_w_hu, 32, UW, UH, DO_MUL) +VMADDWEV(vmaddwev_d_wu, 64, UD, UW, DO_MUL) + +#define VMADDWOD(NAME, BIT, E1, E2, DO_OP) \ +void HELPER(NAME)(void *vd, void *vj, void *vk, uint32_t v) \ +{ \ + int i; \ + VReg *Vd = (VReg *)vd; \ + VReg *Vj = (VReg *)vj; \ + VReg *Vk = (VReg *)vk; \ + typedef __typeof(Vd->E1(0)) TD; \ + \ + for (i = 0; i < LSX_LEN/BIT; i++) { \ + Vd->E1(i) += DO_OP((TD)Vj->E2(2 * i + 1), \ + (TD)Vk->E2(2 * i + 1)); \ + } \ +} + +VMADDWOD(vmaddwod_h_b, 16, H, B, DO_MUL) +VMADDWOD(vmaddwod_w_h, 32, W, H, DO_MUL) +VMADDWOD(vmaddwod_d_w, 64, D, W, DO_MUL) +VMADDWOD(vmaddwod_h_bu, 16, UH, UB, DO_MUL) +VMADDWOD(vmaddwod_w_hu, 32, UW, UH, DO_MUL) +VMADDWOD(vmaddwod_d_wu, 64, UD, UW, DO_MUL) + +#define VMADDWEV_U_S(NAME, BIT, ES1, EU1, ES2, EU2, DO_OP) \ +void HELPER(NAME)(void *vd, void *vj, void *vk, uint32_t v) \ +{ \ + int i; \ + VReg *Vd = (VReg *)vd; \ + VReg *Vj = (VReg *)vj; \ + VReg *Vk = (VReg *)vk; \ + typedef __typeof(Vd->ES1(0)) TS1; \ + typedef __typeof(Vd->EU1(0)) TU1; \ + \ + for (i = 0; i < LSX_LEN/BIT; i++) { \ + Vd->ES1(i) += DO_OP((TU1)Vj->EU2(2 * i), \ + (TS1)Vk->ES2(2 * i)); \ + } \ +} + +VMADDWEV_U_S(vmaddwev_h_bu_b, 16, H, UH, B, UB, DO_MUL) +VMADDWEV_U_S(vmaddwev_w_hu_h, 32, W, UW, H, UH, DO_MUL) +VMADDWEV_U_S(vmaddwev_d_wu_w, 64, D, UD, W, UW, DO_MUL) + +#define VMADDWOD_U_S(NAME, BIT, ES1, EU1, ES2, EU2, DO_OP) \ +void HELPER(NAME)(void *vd, void *vj, void *vk, uint32_t v) \ +{ \ + int i; \ + VReg *Vd = (VReg *)vd; \ + VReg *Vj = (VReg *)vj; \ + VReg *Vk = (VReg *)vk; \ + typedef __typeof(Vd->ES1(0)) TS1; \ + typedef __typeof(Vd->EU1(0)) TU1; \ + \ + for (i = 0; i < LSX_LEN/BIT; i++) { \ + Vd->ES1(i) += DO_OP((TU1)Vj->EU2(2 * i + 1), \ + (TS1)Vk->ES2(2 * i + 1)); \ + } \ +} + +VMADDWOD_U_S(vmaddwod_h_bu_b, 16, H, UH, B, UB, DO_MUL) +VMADDWOD_U_S(vmaddwod_w_hu_h, 32, W, UW, H, UH, DO_MUL) +VMADDWOD_U_S(vmaddwod_d_wu_w, 64, D, UD, W, UW, DO_MUL)