From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-sender-0.a4lg.com (mail-sender-0.a4lg.com [IPv6:2401:2500:203:30b:4000:6bfe:4757:0]) by sourceware.org (Postfix) with ESMTPS id 07538382D39F for ; Tue, 15 Nov 2022 04:58:59 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.1 sourceware.org 07538382D39F Authentication-Results: sourceware.org; dmarc=pass (p=none dis=none) header.from=irq.a4lg.com Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=irq.a4lg.com Received: from [127.0.0.1] (localhost [127.0.0.1]) by mail-sender-0.a4lg.com (Postfix) with ESMTPSA id 5A277300089; Tue, 15 Nov 2022 04:58:57 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=irq.a4lg.com; s=2017s01; t=1668488337; bh=rH8mxvEGOFghryGPjYWDBEhxXyhVT3aW5oEj3Y+hF9s=; h=From:To:Cc:Subject:Date:Message-Id:In-Reply-To:References: Mime-Version:Content-Transfer-Encoding; b=qda1ADuBuC8y06n9wxkrpla99o2OHeCGVFz7G/BT7TZrfddfkJ6yd3AXIcP7Jmzd4 gYRyWqR7XgXmeDfgBOV3lPSnYcdn8mAdB/pGSMpuEuPcr57xP82xmifP+NesdW3P9N 7ViGtNJVnlCZH616kDCehnPLmDEIkIYvwZJj71ms= From: Tsukasa OI To: Tsukasa OI , Nelson Chu , Kito Cheng , Palmer Dabbelt Cc: binutils@sourceware.org Subject: [PATCH 08/11] RISC-V: Split match/print steps on disassembler Date: Tue, 15 Nov 2022 04:52:51 +0000 Message-Id: <1352fb8c63539727204df94651f371ed09bbce4c.1668487922.git.research_trasio@irq.a4lg.com> In-Reply-To: References: Mime-Version: 1.0 Content-Transfer-Encoding: 8bit X-Spam-Status: No, score=-12.1 required=5.0 tests=BAYES_00,DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,GIT_PATCH_0,SPF_HELO_NONE,SPF_PASS,TXREP autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on server2.sourceware.org List-Id: For further optimization and more disassembler features, we may need to change the core RISC-V instruction matching. For this purpose, it is inconvenient to have "match" and "print" steps in the same loop. This commit rewrites riscv_disassemble_insn function so that we store matched_op for matching RISC-V opcode and then print it (if not NULL). Although it looks a bit inefficient, it also lowers the indent of opcode matching loop to clarify the opcode matching changes on the next optimization commit. Unfortunately, this commit alone will impose some performance penalty (<5% on most cases but sometimes about 15% worse) but it can be easily paid back by other optimizations. opcodes/ChangeLog: * riscv-dis.c (riscv_disassemble_insn): Split instruction handling to two separate steps - opcode matching and printing. --- opcodes/riscv-dis.c | 151 +++++++++++++++++++++++--------------------- 1 file changed, 79 insertions(+), 72 deletions(-) diff --git a/opcodes/riscv-dis.c b/opcodes/riscv-dis.c index d9c16dad615..316c6c97607 100644 --- a/opcodes/riscv-dis.c +++ b/opcodes/riscv-dis.c @@ -646,7 +646,7 @@ print_insn_args (const char *oparg, insn_t l, bfd_vma pc, disassemble_info *info static int riscv_disassemble_insn (bfd_vma memaddr, insn_t word, disassemble_info *info) { - const struct riscv_opcode *op; + const struct riscv_opcode *op, *matched_op; static bool init = false; static const struct riscv_opcode *riscv_hash[OP_MASK_OP + 1]; struct riscv_private_data *pd; @@ -702,85 +702,92 @@ riscv_disassemble_insn (bfd_vma memaddr, insn_t word, disassemble_info *info) info->target = 0; info->target2 = 0; + matched_op = NULL; op = riscv_hash[OP_HASH_IDX (word)]; - if (op != NULL) + + /* If XLEN is not known, get its value from the ELF class. */ + if (info->mach == bfd_mach_riscv64) + xlen = 64; + else if (info->mach == bfd_mach_riscv32) + xlen = 32; + else if (info->section != NULL) { - /* If XLEN is not known, get its value from the ELF class. */ - if (info->mach == bfd_mach_riscv64) - xlen = 64; - else if (info->mach == bfd_mach_riscv32) - xlen = 32; - else if (info->section != NULL) - { - Elf_Internal_Ehdr *ehdr = elf_elfheader (info->section->owner); - xlen = ehdr->e_ident[EI_CLASS] == ELFCLASS64 ? 64 : 32; - } + Elf_Internal_Ehdr *ehdr = elf_elfheader (info->section->owner); + xlen = ehdr->e_ident[EI_CLASS] == ELFCLASS64 ? 64 : 32; + } - /* If arch has the Zfinx extension, replace FPR with GPR. */ - if (riscv_subset_supports (&riscv_rps_dis, "zfinx")) - riscv_fpr_names = riscv_gpr_names; - else - riscv_fpr_names = riscv_gpr_names == riscv_gpr_names_abi ? - riscv_fpr_names_abi : riscv_fpr_names_numeric; + /* If arch has the Zfinx extension, replace FPR with GPR. */ + if (riscv_subset_supports (&riscv_rps_dis, "zfinx")) + riscv_fpr_names = riscv_gpr_names; + else + riscv_fpr_names = riscv_gpr_names == riscv_gpr_names_abi + ? riscv_fpr_names_abi + : riscv_fpr_names_numeric; - for (; op->name; op++) - { - /* Does the opcode match? */ - if (! (op->match_func) (op, word)) - continue; - /* Is this a pseudo-instruction and may we print it as such? */ - if (no_aliases && (op->pinfo & INSN_ALIAS)) - continue; - /* Is this instruction restricted to a certain value of XLEN? */ - if ((op->xlen_requirement != 0) && (op->xlen_requirement != xlen)) - continue; - /* Is this instruction supported by the current architecture? */ - if (!riscv_multi_subset_supports (&riscv_rps_dis, op->insn_class)) - continue; - - /* It's a match. */ - (*info->fprintf_styled_func) (info->stream, dis_style_mnemonic, - "%s", op->name); - print_insn_args (op->args, word, memaddr, info); - - /* Try to disassemble multi-instruction addressing sequences. */ - if (pd->to_print_addr) - { - info->target = pd->print_addr; - (*info->fprintf_styled_func) - (info->stream, dis_style_comment_start, " # "); - (*info->print_address_func) (info->target, info); - pd->to_print_addr = false; - } + for (; op && op->name; op++) + { + /* Does the opcode match? */ + if (!(op->match_func) (op, word)) + continue; + /* Is this a pseudo-instruction and may we print it as such? */ + if (no_aliases && (op->pinfo & INSN_ALIAS)) + continue; + /* Is this instruction restricted to a certain value of XLEN? */ + if ((op->xlen_requirement != 0) && (op->xlen_requirement != xlen)) + continue; + /* Is this instruction supported by the current architecture? */ + if (!riscv_multi_subset_supports (&riscv_rps_dis, op->insn_class)) + continue; - /* Finish filling out insn_info fields. */ - switch (op->pinfo & INSN_TYPE) - { - case INSN_BRANCH: - info->insn_type = dis_branch; - break; - case INSN_CONDBRANCH: - info->insn_type = dis_condbranch; - break; - case INSN_JSR: - info->insn_type = dis_jsr; - break; - case INSN_DREF: - info->insn_type = dis_dref; - break; - default: - break; - } + matched_op = op; + break; + } - if (op->pinfo & INSN_DATA_SIZE) - { - int size = ((op->pinfo & INSN_DATA_SIZE) - >> INSN_DATA_SIZE_SHIFT); - info->data_size = 1 << (size - 1); - } + if (matched_op != NULL) + { + /* There is a match. */ + op = matched_op; + + (*info->fprintf_styled_func) (info->stream, dis_style_mnemonic, + "%s", op->name); + print_insn_args (op->args, word, memaddr, info); - return insnlen; + /* Try to disassemble multi-instruction addressing sequences. */ + if (pd->to_print_addr) + { + info->target = pd->print_addr; + (*info->fprintf_styled_func) (info->stream, dis_style_comment_start, + " # "); + (*info->print_address_func) (info->target, info); + pd->to_print_addr = false; } + + /* Finish filling out insn_info fields. */ + switch (op->pinfo & INSN_TYPE) + { + case INSN_BRANCH: + info->insn_type = dis_branch; + break; + case INSN_CONDBRANCH: + info->insn_type = dis_condbranch; + break; + case INSN_JSR: + info->insn_type = dis_jsr; + break; + case INSN_DREF: + info->insn_type = dis_dref; + break; + default: + break; + } + + if (op->pinfo & INSN_DATA_SIZE) + { + int size = ((op->pinfo & INSN_DATA_SIZE) >> INSN_DATA_SIZE_SHIFT); + info->data_size = 1 << (size - 1); + } + + return insnlen; } /* We did not find a match, so just print the instruction bits. */ -- 2.37.2