From patchwork Fri Feb 28 16:43:47 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Taylor Simpson X-Patchwork-Id: 11413001 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id E98BF13A4 for ; Fri, 28 Feb 2020 17:18:24 +0000 (UTC) Received: from lists.gnu.org (lists.gnu.org [209.51.188.17]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPS id B0D78246A3 for ; Fri, 28 Feb 2020 17:18:24 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=fail reason="signature verification failed" (1024-bit key) header.d=quicinc.com header.i=@quicinc.com header.b="Onr30Cir" DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org B0D78246A3 Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=quicinc.com Authentication-Results: mail.kernel.org; spf=pass smtp.mailfrom=qemu-devel-bounces+patchwork-qemu-devel=patchwork.kernel.org@nongnu.org Received: from localhost ([::1]:51140 helo=lists1p.gnu.org) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1j7jHL-0006Hn-Rr for patchwork-qemu-devel@patchwork.kernel.org; Fri, 28 Feb 2020 12:18:23 -0500 Received: from eggs.gnu.org ([2001:470:142:3::10]:58626) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1j7imS-0005sI-Hh for qemu-devel@nongnu.org; Fri, 28 Feb 2020 11:46:31 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1j7imP-0007Sr-0n for qemu-devel@nongnu.org; Fri, 28 Feb 2020 11:46:28 -0500 Received: from alexa-out-sd-01.qualcomm.com ([199.106.114.38]:13235) by eggs.gnu.org with esmtps (TLS1.0:RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1j7imL-0005U6-5f for qemu-devel@nongnu.org; Fri, 28 Feb 2020 11:46:22 -0500 DKIM-Signature: v=1; a=rsa-sha256; c=simple/simple; d=quicinc.com; i=@quicinc.com; q=dns/txt; s=qcdkim; t=1582908381; x=1614444381; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=/+qcbcQjmqrzLTqTjCZoiRd0csQBkuFop74v+70w62M=; b=Onr30Cir3uHSewAfbVEomShPty80U23fWSv/evhqYmhwXOKVEYftOkU4 0Bejc4h+wUIeAy8t14eYdqDGXCGbST0XiIiPCmqdQusBf+i38hoZcQSq5 kvNsM+Im+mbhEjC5dbVbUo0L6yA0LEQQQtajoZp4axUvPlg/Ay3Ka030k 4=; Received: from unknown (HELO ironmsg-SD-alpha.qualcomm.com) ([10.53.140.30]) by alexa-out-sd-01.qualcomm.com with ESMTP; 28 Feb 2020 08:44:32 -0800 Received: from vu-tsimpson-aus.qualcomm.com (HELO vu-tsimpson1-aus.qualcomm.com) ([10.222.150.1]) by ironmsg-SD-alpha.qualcomm.com with ESMTP; 28 Feb 2020 08:44:32 -0800 Received: by vu-tsimpson1-aus.qualcomm.com (Postfix, from userid 47164) id 3821E1191; Fri, 28 Feb 2020 10:44:32 -0600 (CST) From: Taylor Simpson To: qemu-devel@nongnu.org Subject: [RFC PATCH v2 51/67] Hexagon translation Date: Fri, 28 Feb 2020 10:43:47 -0600 Message-Id: <1582908244-304-52-git-send-email-tsimpson@quicinc.com> X-Mailer: git-send-email 2.7.4 In-Reply-To: <1582908244-304-1-git-send-email-tsimpson@quicinc.com> References: <1582908244-304-1-git-send-email-tsimpson@quicinc.com> MIME-Version: 1.0 X-detected-operating-system: by eggs.gnu.org: FreeBSD 9.x [fuzzy] X-Received-From: 199.106.114.38 X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.23 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: riku.voipio@iki.fi, richard.henderson@linaro.org, laurent@vivier.eu, Taylor Simpson , philmd@redhat.com, aleksandar.m.mail@gmail.com Errors-To: qemu-devel-bounces+patchwork-qemu-devel=patchwork.kernel.org@nongnu.org Sender: "Qemu-devel" Read the instruction memory Create a packet data structure Generate TCG code for the start of the packet Invoke the generate function for each instruction Generate TCG code for the end of the packet Signed-off-by: Taylor Simpson --- target/hexagon/translate.h | 82 +++++ target/hexagon/translate.c | 728 +++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 810 insertions(+) create mode 100644 target/hexagon/translate.h create mode 100644 target/hexagon/translate.c diff --git a/target/hexagon/translate.h b/target/hexagon/translate.h new file mode 100644 index 0000000..65b721e --- /dev/null +++ b/target/hexagon/translate.h @@ -0,0 +1,82 @@ +/* + * Copyright(c) 2019-2020 Qualcomm Innovation Center, Inc. All Rights Reserved. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see . + */ + +#ifndef HEXAGON_TRANSLATE_H +#define HEXAGON_TRANSLATE_H + +#include "cpu.h" +#include "exec/translator.h" +#include "exec/helper-proto.h" +#include "exec/helper-gen.h" +#include "internal.h" + +typedef struct DisasContext { + DisasContextBase base; + uint32_t mem_idx; + int ctx_reg_log[REG_WRITES_MAX]; + int ctx_reg_log_idx; + int ctx_preg_log[PRED_WRITES_MAX]; + int ctx_preg_log_idx; + uint8_t ctx_store_width[STORES_MAX]; +} DisasContext; + +static inline void ctx_log_reg_write(DisasContext *ctx, int rnum) +{ +#if HEX_DEBUG + int i; + for (i = 0; i < ctx->ctx_reg_log_idx; i++) { + if (ctx->ctx_reg_log[i] == rnum) { + HEX_DEBUG_LOG("WARNING: Multiple writes to r%d\n", rnum); + } + } +#endif + ctx->ctx_reg_log[ctx->ctx_reg_log_idx] = rnum; + ctx->ctx_reg_log_idx++; +} + +static inline void ctx_log_pred_write(DisasContext *ctx, int pnum) +{ + ctx->ctx_preg_log[ctx->ctx_preg_log_idx] = pnum; + ctx->ctx_preg_log_idx++; +} + +extern TCGv hex_gpr[TOTAL_PER_THREAD_REGS]; +extern TCGv hex_pred[NUM_PREGS]; +extern TCGv hex_next_PC; +extern TCGv hex_this_PC; +extern TCGv hex_slot_cancelled; +extern TCGv hex_branch_taken; +extern TCGv hex_new_value[TOTAL_PER_THREAD_REGS]; +extern TCGv hex_reg_written[TOTAL_PER_THREAD_REGS]; +extern TCGv hex_new_pred_value[NUM_PREGS]; +extern TCGv hex_pred_written; +extern TCGv hex_store_addr[STORES_MAX]; +extern TCGv hex_store_width[STORES_MAX]; +extern TCGv hex_store_val32[STORES_MAX]; +extern TCGv_i64 hex_store_val64[STORES_MAX]; +extern TCGv hex_dczero_addr; +extern TCGv llsc_addr; +extern TCGv llsc_val; +extern TCGv_i64 llsc_val_i64; +extern TCGv hex_is_gather_store_insn; +extern TCGv hex_gather_issued; + +void hexagon_translate_init(void); +extern void gen_exception(int excp); +extern void gen_exception_debug(void); + +#endif diff --git a/target/hexagon/translate.c b/target/hexagon/translate.c new file mode 100644 index 0000000..57ab294 --- /dev/null +++ b/target/hexagon/translate.c @@ -0,0 +1,728 @@ +/* + * Copyright(c) 2019-2020 Qualcomm Innovation Center, Inc. All Rights Reserved. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see . + */ + +#define QEMU_GENERATE +#include "qemu/osdep.h" +#include "cpu.h" +#include "tcg/tcg-op.h" +#include "exec/cpu_ldst.h" +#include "exec/log.h" +#include "internal.h" +#include "attribs.h" +#include "insn.h" +#include "decode.h" +#include "translate.h" +#include "printinsn.h" +#include "macros.h" + +TCGv hex_gpr[TOTAL_PER_THREAD_REGS]; +TCGv hex_pred[NUM_PREGS]; +TCGv hex_next_PC; +TCGv hex_this_PC; +TCGv hex_slot_cancelled; +TCGv hex_branch_taken; +TCGv hex_new_value[TOTAL_PER_THREAD_REGS]; +#if HEX_DEBUG +TCGv hex_reg_written[TOTAL_PER_THREAD_REGS]; +#endif +TCGv hex_new_pred_value[NUM_PREGS]; +TCGv hex_pred_written; +TCGv hex_store_addr[STORES_MAX]; +TCGv hex_store_width[STORES_MAX]; +TCGv hex_store_val32[STORES_MAX]; +TCGv_i64 hex_store_val64[STORES_MAX]; +TCGv hex_dczero_addr; +TCGv llsc_addr; +TCGv llsc_val; +TCGv_i64 llsc_val_i64; +TCGv hex_is_gather_store_insn; +TCGv hex_gather_issued; + +static const char * const hexagon_prednames[] = { + "p0", "p1", "p2", "p3" +}; + +void gen_exception(int excp) +{ + TCGv_i32 helper_tmp = tcg_const_i32(excp); + gen_helper_raise_exception(cpu_env, helper_tmp); + tcg_temp_free_i32(helper_tmp); +} + +void gen_exception_debug(void) +{ + gen_exception(EXCP_DEBUG); +} + +#if HEX_DEBUG +#define PACKET_BUFFER_LEN 1028 +static void print_pkt(packet_t *pkt) +{ + char buf[PACKET_BUFFER_LEN]; + snprint_a_pkt(buf, PACKET_BUFFER_LEN, pkt); + HEX_DEBUG_LOG("%s", buf); +} +#define HEX_DEBUG_PRINT_PKT(pkt) print_pkt(pkt) +#else +#define HEX_DEBUG_PRINT_PKT(pkt) /* nothing */ +#endif + +static int read_packet_words(CPUHexagonState *env, DisasContext *ctx, + uint32_t words[]) +{ + bool found_end = false; + int max_words; + int nwords; + int i; + + /* Make sure we don't cross a page boundary */ + max_words = -(ctx->base.pc_next | TARGET_PAGE_MASK) / sizeof(uint32_t); + if (max_words < PACKET_WORDS_MAX) { + /* Might cross a page boundary */ + if (ctx->base.num_insns == 1) { + /* OK if it's the first packet in the TB */ + max_words = PACKET_WORDS_MAX; + } + } else { + max_words = PACKET_WORDS_MAX; + } + + memset(words, 0, PACKET_WORDS_MAX * sizeof(uint32_t)); + for (nwords = 0; !found_end && nwords < max_words; nwords++) { + words[nwords] = cpu_ldl_code(env, + ctx->base.pc_next + nwords * sizeof(uint32_t)); + found_end = is_packet_end(words[nwords]); + } + if (!found_end) { + if (nwords == PACKET_WORDS_MAX) { + /* Read too many words without finding the end */ + gen_exception(HEX_EXCP_INVALID_PACKET); + ctx->base.is_jmp = DISAS_NORETURN; + return 0; + } + /* Crosses page boundary - defer to next TB */ + ctx->base.is_jmp = DISAS_TOO_MANY; + return 0; + } + + HEX_DEBUG_LOG("decode_packet: pc = 0x%x\n", ctx->base.pc_next); + HEX_DEBUG_LOG(" words = { "); + for (i = 0; i < nwords; i++) { + HEX_DEBUG_LOG("0x%x, ", words[i]); + } + HEX_DEBUG_LOG("}\n"); + + return nwords; +} + +static void gen_start_packet(DisasContext *ctx, packet_t *pkt) +{ + target_ulong next_PC = ctx->base.pc_next + pkt->encod_pkt_size_in_bytes; + int i; + + /* Clear out the disassembly context */ + ctx->ctx_reg_log_idx = 0; + ctx->ctx_preg_log_idx = 0; + for (i = 0; i < STORES_MAX; i++) { + ctx->ctx_store_width[i] = 0; + } + +#if HEX_DEBUG + /* Handy place to set a breakpoint before the packet executes */ + gen_helper_debug_start_packet(cpu_env); + tcg_gen_movi_tl(hex_this_PC, ctx->base.pc_next); +#endif + + /* Initialize the runtime state for packet semantics */ + tcg_gen_movi_tl(hex_gpr[HEX_REG_PC], ctx->base.pc_next); + tcg_gen_movi_tl(hex_slot_cancelled, 0); + if (pkt->pkt_has_cof) { + tcg_gen_movi_tl(hex_branch_taken, 0); + tcg_gen_movi_tl(hex_next_PC, next_PC); + } + tcg_gen_movi_tl(hex_pred_written, 0); +} + +static int is_gather_store_insn(insn_t *insn) +{ + int check = GET_ATTRIB(insn->opcode, A_CVI_NEW); + check &= (insn->new_value_producer_slot == 1); + return check; +} + +/* + * The LOG_*_WRITE macros mark most of the writes in a packet + * However, there are some implicit writes marked as attributes + * of the applicable instructions. + */ +static void mark_implicit_reg_write(DisasContext *ctx, insn_t *insn, + int attrib, int rnum) +{ + if (GET_ATTRIB(insn->opcode, attrib)) { + ctx->ctx_reg_log[ctx->ctx_reg_log_idx] = rnum; + ctx->ctx_reg_log_idx++; + } +} + +static void mark_implicit_pred_write(DisasContext *ctx, insn_t *insn, + int attrib, int pnum) +{ + if (GET_ATTRIB(insn->opcode, attrib)) { + ctx->ctx_preg_log[ctx->ctx_preg_log_idx] = pnum; + ctx->ctx_preg_log_idx++; + } +} + +static void mark_implicit_writes(DisasContext *ctx, insn_t *insn) +{ + mark_implicit_reg_write(ctx, insn, A_IMPLICIT_WRITES_LR, HEX_REG_LR); + mark_implicit_reg_write(ctx, insn, A_IMPLICIT_WRITES_LC0, HEX_REG_LC0); + mark_implicit_reg_write(ctx, insn, A_IMPLICIT_WRITES_SA0, HEX_REG_SA0); + mark_implicit_reg_write(ctx, insn, A_IMPLICIT_WRITES_LC1, HEX_REG_LC1); + mark_implicit_reg_write(ctx, insn, A_IMPLICIT_WRITES_SA1, HEX_REG_SA1); + + mark_implicit_pred_write(ctx, insn, A_IMPLICIT_WRITES_P0, 0); + mark_implicit_pred_write(ctx, insn, A_IMPLICIT_WRITES_P1, 1); + mark_implicit_pred_write(ctx, insn, A_IMPLICIT_WRITES_P2, 2); + mark_implicit_pred_write(ctx, insn, A_IMPLICIT_WRITES_P3, 3); +} + +static void gen_insn(CPUHexagonState *env, DisasContext *ctx, insn_t *insn) +{ + if (insn->generate) { + bool is_gather_store = is_gather_store_insn(insn); + if (is_gather_store) { + tcg_gen_movi_tl(hex_is_gather_store_insn, 1); + } + insn->generate(env, ctx, insn); + mark_implicit_writes(ctx, insn); + if (is_gather_store) { + tcg_gen_movi_tl(hex_is_gather_store_insn, 0); + } + } else { + gen_exception(HEX_EXCP_INVALID_OPCODE); + ctx->base.is_jmp = DISAS_NORETURN; + } +} + +/* + * Helpers for generating the packet commit + */ +static void gen_reg_writes(DisasContext *ctx) +{ + int i; + + for (i = 0; i < ctx->ctx_reg_log_idx; i++) { + int reg_num = ctx->ctx_reg_log[i]; + + tcg_gen_mov_tl(hex_gpr[reg_num], hex_new_value[reg_num]); + } +} + +static void gen_pred_writes(DisasContext *ctx, packet_t *pkt) +{ + /* Early exit if the log is empty */ + if (!ctx->ctx_preg_log_idx) { + return; + } + + TCGv zero = tcg_const_tl(0); + TCGv control_reg = tcg_temp_new(); + TCGv pval = tcg_temp_new(); + int i; + + /* + * Only endloop instructions will conditionally + * write a predicate. If there are no endloop + * instructions, we can use the non-conditional + * write of the predicates. + */ + if (pkt->pkt_has_endloop) { + TCGv pred_written = tcg_temp_new(); + for (i = 0; i < ctx->ctx_preg_log_idx; i++) { + int pred_num = ctx->ctx_preg_log[i]; + + tcg_gen_andi_tl(pred_written, hex_pred_written, 1 << pred_num); + tcg_gen_movcond_tl(TCG_COND_NE, hex_pred[pred_num], + pred_written, zero, + hex_new_pred_value[pred_num], + hex_pred[pred_num]); + } + tcg_temp_free(pred_written); + } else { + for (i = 0; i < ctx->ctx_preg_log_idx; i++) { + int pred_num = ctx->ctx_preg_log[i]; + tcg_gen_mov_tl(hex_pred[pred_num], hex_new_pred_value[pred_num]); +#if HEX_DEBUG + /* Do this so HELPER(debug_commit_end) will know */ + tcg_gen_ori_tl(hex_pred_written, hex_pred_written, 1 << pred_num); +#endif + } + } + + tcg_temp_free(zero); + tcg_temp_free(control_reg); + tcg_temp_free(pval); +} + +#if HEX_DEBUG +static inline void gen_check_store_width(DisasContext *ctx, int slot_num) +{ + TCGv slot = tcg_const_tl(slot_num); + TCGv check = tcg_const_tl(ctx->ctx_store_width[slot_num]); + gen_helper_debug_check_store_width(cpu_env, slot, check); + tcg_temp_free(slot); + tcg_temp_free(check); +} +#define HEX_DEBUG_GEN_CHECK_STORE_WIDTH(ctx, slot_num) \ + gen_check_store_width(ctx, slot_num) +#else +#define HEX_DEBUG_GEN_CHECK_STORE_WIDTH(ctx, slot_num) /* nothing */ +#endif + +static void process_store(DisasContext *ctx, int slot_num) +{ + TCGv cancelled = tcg_temp_local_new(); + TCGLabel *label_end = gen_new_label(); + + /* Don't do anything if the slot was cancelled */ + gen_slot_cancelled_check(cancelled, slot_num); + tcg_gen_brcondi_tl(TCG_COND_NE, cancelled, 0, label_end); + { + int ctx_width = ctx->ctx_store_width[slot_num]; + TCGv address = tcg_temp_local_new(); + tcg_gen_mov_tl(address, hex_store_addr[slot_num]); + + /* + * If we know the width from the DisasContext, we can + * generate much cleaner code. + * Unfortunately, not all instructions execute the fSTORE + * macro during code generation. Anything that uses the + * generic helper will have this problem. Instructions + * that use fWRAP to generate proper TCG code will be OK. + */ + if (ctx_width == 1) { + HEX_DEBUG_GEN_CHECK_STORE_WIDTH(ctx, slot_num); + TCGv value = tcg_temp_new(); + tcg_gen_mov_tl(value, hex_store_val32[slot_num]); + tcg_gen_qemu_st8(value, address, ctx->mem_idx); + tcg_temp_free(value); + } else if (ctx_width == 2) { + HEX_DEBUG_GEN_CHECK_STORE_WIDTH(ctx, slot_num); + TCGv value = tcg_temp_new(); + tcg_gen_mov_tl(value, hex_store_val32[slot_num]); + tcg_gen_qemu_st16(value, address, ctx->mem_idx); + tcg_temp_free(value); + } else if (ctx_width == 4) { + HEX_DEBUG_GEN_CHECK_STORE_WIDTH(ctx, slot_num); + TCGv value = tcg_temp_new(); + tcg_gen_mov_tl(value, hex_store_val32[slot_num]); + tcg_gen_qemu_st32(value, address, ctx->mem_idx); + tcg_temp_free(value); + } else if (ctx_width == 8) { + HEX_DEBUG_GEN_CHECK_STORE_WIDTH(ctx, slot_num); + TCGv_i64 value = tcg_temp_new_i64(); + tcg_gen_mov_i64(value, hex_store_val64[slot_num]); + tcg_gen_qemu_st64(value, address, ctx->mem_idx); + tcg_temp_free_i64(value); + } else { + /* + * If we get to here, we don't know the width at + * TCG generation time, we'll generate branching + * based on the width at runtime. + */ + TCGLabel *label_w2 = gen_new_label(); + TCGLabel *label_w4 = gen_new_label(); + TCGLabel *label_w8 = gen_new_label(); + TCGv width = tcg_temp_local_new(); + + tcg_gen_mov_tl(width, hex_store_width[slot_num]); + tcg_gen_brcondi_tl(TCG_COND_NE, width, 1, label_w2); + { + /* Width is 1 byte */ + TCGv value = tcg_temp_new(); + tcg_gen_mov_tl(value, hex_store_val32[slot_num]); + tcg_gen_qemu_st8(value, address, ctx->mem_idx); + tcg_gen_br(label_end); + tcg_temp_free(value); + } + gen_set_label(label_w2); + tcg_gen_brcondi_tl(TCG_COND_NE, width, 2, label_w4); + { + /* Width is 2 bytes */ + TCGv value = tcg_temp_new(); + tcg_gen_mov_tl(value, hex_store_val32[slot_num]); + tcg_gen_qemu_st16(value, address, ctx->mem_idx); + tcg_gen_br(label_end); + tcg_temp_free(value); + } + gen_set_label(label_w4); + tcg_gen_brcondi_tl(TCG_COND_NE, width, 4, label_w8); + { + /* Width is 4 bytes */ + TCGv value = tcg_temp_new(); + tcg_gen_mov_tl(value, hex_store_val32[slot_num]); + tcg_gen_qemu_st32(value, address, ctx->mem_idx); + tcg_gen_br(label_end); + tcg_temp_free(value); + } + gen_set_label(label_w8); + { + /* Width is 8 bytes */ + TCGv_i64 value = tcg_temp_new_i64(); + tcg_gen_mov_i64(value, hex_store_val64[slot_num]); + tcg_gen_qemu_st64(value, address, ctx->mem_idx); + tcg_gen_br(label_end); + tcg_temp_free_i64(value); + } + + tcg_temp_free(width); + } + tcg_temp_free(address); + } + gen_set_label(label_end); + + tcg_temp_free(cancelled); +} + +static void process_store_log(DisasContext *ctx, packet_t *pkt) +{ + /* + * When a packet has two stores, the hardware processes + * slot 1 and then slot 2. This will be important when + * the memory accesses overlap. + */ + if (pkt->pkt_has_store_s1 && !pkt->pkt_has_dczeroa) { + process_store(ctx, 1); + } + if (pkt->pkt_has_store_s0 && !pkt->pkt_has_dczeroa) { + process_store(ctx, 0); + } +} + +/* Zero out a 32-bit cache line */ +static void process_dczeroa(DisasContext *ctx, packet_t *pkt) +{ + if (pkt->pkt_has_dczeroa) { + /* Store 32 bytes of zero starting at (addr & ~0x1f) */ + TCGv addr = tcg_temp_new(); + TCGv_i64 zero = tcg_const_i64(0); + + tcg_gen_andi_tl(addr, hex_dczero_addr, ~0x1f); + tcg_gen_qemu_st64(zero, addr, ctx->mem_idx); + tcg_gen_addi_tl(addr, addr, 8); + tcg_gen_qemu_st64(zero, addr, ctx->mem_idx); + tcg_gen_addi_tl(addr, addr, 8); + tcg_gen_qemu_st64(zero, addr, ctx->mem_idx); + tcg_gen_addi_tl(addr, addr, 8); + tcg_gen_qemu_st64(zero, addr, ctx->mem_idx); + + tcg_temp_free(addr); + tcg_temp_free_i64(zero); + } +} + +static bool process_change_of_flow(DisasContext *ctx, packet_t *pkt) +{ + if (pkt->pkt_has_cof) { + tcg_gen_mov_tl(hex_gpr[HEX_REG_PC], hex_next_PC); + return true; + } + return false; +} + +static void gen_exec_counters(packet_t *pkt) +{ + int num_insns = pkt->num_insns; + int num_real_insns = 0; + int i; + + for (i = 0; i < num_insns; i++) { + if (!pkt->insn[i].is_endloop && + !pkt->insn[i].part1 && + !GET_ATTRIB(pkt->insn[i].opcode, A_IT_NOP)) { + num_real_insns++; + } + } + + tcg_gen_addi_tl(hex_gpr[HEX_REG_QEMU_PKT_CNT], + hex_gpr[HEX_REG_QEMU_PKT_CNT], 1); + if (num_real_insns) { + tcg_gen_addi_tl(hex_gpr[HEX_REG_QEMU_INSN_CNT], + hex_gpr[HEX_REG_QEMU_INSN_CNT], num_real_insns); + } +} + +static void gen_commit_packet(DisasContext *ctx, packet_t *pkt) +{ + bool end_tb = false; + + gen_reg_writes(ctx); + gen_pred_writes(ctx, pkt); + process_store_log(ctx, pkt); + process_dczeroa(ctx, pkt); + end_tb |= process_change_of_flow(ctx, pkt); + gen_exec_counters(pkt); +#if HEX_DEBUG + { + TCGv has_st0 = + tcg_const_tl(pkt->pkt_has_store_s0 && !pkt->pkt_has_dczeroa); + TCGv has_st1 = + tcg_const_tl(pkt->pkt_has_store_s1 && !pkt->pkt_has_dczeroa); + + /* Handy place to set a breakpoint at the end of execution */ + gen_helper_debug_commit_end(cpu_env, has_st0, has_st1); + + tcg_temp_free(has_st0); + tcg_temp_free(has_st1); + } +#endif + + if (end_tb) { + tcg_gen_exit_tb(NULL, 0); + ctx->base.is_jmp = DISAS_NORETURN; + } +} + +static void decode_and_translate_packet(CPUHexagonState *env, DisasContext *ctx) +{ + uint32_t words[PACKET_WORDS_MAX]; + int nwords; + packet_t pkt; + int i; + + nwords = read_packet_words(env, ctx, words); + if (!nwords) { + return; + } + + if (decode_this(nwords, words, &pkt)) { + HEX_DEBUG_PRINT_PKT(&pkt); + gen_start_packet(ctx, &pkt); + for (i = 0; i < pkt.num_insns; i++) { + gen_insn(env, ctx, &pkt.insn[i]); + } + gen_commit_packet(ctx, &pkt); + ctx->base.pc_next += pkt.encod_pkt_size_in_bytes; + } else { + gen_exception(HEX_EXCP_INVALID_PACKET); + ctx->base.is_jmp = DISAS_NORETURN; + } +} + +static void hexagon_tr_init_disas_context(DisasContextBase *dcbase, + CPUState *cs) +{ + DisasContext *ctx = container_of(dcbase, DisasContext, base); + + ctx->mem_idx = MMU_USER_IDX; +} + +static void hexagon_tr_tb_start(DisasContextBase *db, CPUState *cpu) +{ +} + +static void hexagon_tr_insn_start(DisasContextBase *dcbase, CPUState *cpu) +{ + DisasContext *ctx = container_of(dcbase, DisasContext, base); + + tcg_gen_insn_start(ctx->base.pc_next); +} + +static bool hexagon_tr_breakpoint_check(DisasContextBase *dcbase, CPUState *cpu, + const CPUBreakpoint *bp) +{ + DisasContext *ctx = container_of(dcbase, DisasContext, base); + + tcg_gen_movi_tl(hex_gpr[HEX_REG_PC], ctx->base.pc_next); + ctx->base.is_jmp = DISAS_NORETURN; + gen_exception_debug(); + /* + * The address covered by the breakpoint must be included in + * [tb->pc, tb->pc + tb->size) in order to for it to be + * properly cleared -- thus we increment the PC here so that + * the logic setting tb->size below does the right thing. + */ + ctx->base.pc_next += 4; + return true; +} + + +static void hexagon_tr_translate_packet(DisasContextBase *dcbase, CPUState *cpu) +{ + DisasContext *ctx = container_of(dcbase, DisasContext, base); + CPUHexagonState *env = cpu->env_ptr; + + decode_and_translate_packet(env, ctx); + + if (ctx->base.is_jmp == DISAS_NEXT) { + target_ulong page_start; + + page_start = ctx->base.pc_first & TARGET_PAGE_MASK; + if (ctx->base.pc_next - page_start >= TARGET_PAGE_SIZE) { + ctx->base.is_jmp = DISAS_TOO_MANY; + } + + /* + * The CPU log is used to compare against LLDB single stepping, + * so end the TLB after every packet. + */ + if (qemu_loglevel_mask(CPU_LOG_TB_CPU)) { + ctx->base.is_jmp = DISAS_TOO_MANY; + } +#if HEX_DEBUG + /* When debugging, only put one packet per TB */ + ctx->base.is_jmp = DISAS_TOO_MANY; +#endif + } +} + +static void hexagon_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu) +{ + DisasContext *ctx = container_of(dcbase, DisasContext, base); + + switch (ctx->base.is_jmp) { + case DISAS_TOO_MANY: + tcg_gen_movi_tl(hex_gpr[HEX_REG_PC], ctx->base.pc_next); + if (ctx->base.singlestep_enabled) { + gen_exception_debug(); + } else { + tcg_gen_exit_tb(NULL, 0); + } + break; + case DISAS_NORETURN: + break; + default: + g_assert_not_reached(); + } +} + +static void hexagon_tr_disas_log(const DisasContextBase *dcbase, CPUState *cpu) +{ + qemu_log("IN: %s\n", lookup_symbol(dcbase->pc_first)); + log_target_disas(cpu, dcbase->pc_first, dcbase->tb->size); +} + + +static const TranslatorOps hexagon_tr_ops = { + .init_disas_context = hexagon_tr_init_disas_context, + .tb_start = hexagon_tr_tb_start, + .insn_start = hexagon_tr_insn_start, + .breakpoint_check = hexagon_tr_breakpoint_check, + .translate_insn = hexagon_tr_translate_packet, + .tb_stop = hexagon_tr_tb_stop, + .disas_log = hexagon_tr_disas_log, +}; + +void gen_intermediate_code(CPUState *cs, TranslationBlock *tb, int max_insns) +{ + DisasContext ctx; + + translator_loop(&hexagon_tr_ops, &ctx.base, cs, tb, max_insns); +} + +#define NAME_LEN 64 +static char new_value_names[TOTAL_PER_THREAD_REGS][NAME_LEN]; +#if HEX_DEBUG +static char reg_written_names[TOTAL_PER_THREAD_REGS][NAME_LEN]; +#endif +static char new_pred_value_names[NUM_PREGS][NAME_LEN]; +static char store_addr_names[STORES_MAX][NAME_LEN]; +static char store_width_names[STORES_MAX][NAME_LEN]; +static char store_val32_names[STORES_MAX][NAME_LEN]; +static char store_val64_names[STORES_MAX][NAME_LEN]; + +void hexagon_translate_init(void) +{ + int i; + + opcode_init(); + + for (i = 0; i < TOTAL_PER_THREAD_REGS; i++) { + hex_gpr[i] = tcg_global_mem_new(cpu_env, + offsetof(CPUHexagonState, gpr[i]), + hexagon_regnames[i]); + + sprintf(new_value_names[i], "new_%s", hexagon_regnames[i]); + hex_new_value[i] = tcg_global_mem_new(cpu_env, + offsetof(CPUHexagonState, new_value[i]), + new_value_names[i]); + +#if HEX_DEBUG + sprintf(reg_written_names[i], "reg_written_%s", hexagon_regnames[i]); + hex_reg_written[i] = tcg_global_mem_new(cpu_env, + offsetof(CPUHexagonState, reg_written[i]), + reg_written_names[i]); +#endif + } + for (i = 0; i < NUM_PREGS; i++) { + hex_pred[i] = tcg_global_mem_new(cpu_env, + offsetof(CPUHexagonState, pred[i]), + hexagon_prednames[i]); + + sprintf(new_pred_value_names[i], "new_pred_%s", hexagon_prednames[i]); + hex_new_pred_value[i] = tcg_global_mem_new(cpu_env, + offsetof(CPUHexagonState, new_pred_value[i]), + new_pred_value_names[i]); + } + hex_pred_written = tcg_global_mem_new(cpu_env, + offsetof(CPUHexagonState, pred_written), "pred_written"); + hex_next_PC = tcg_global_mem_new(cpu_env, + offsetof(CPUHexagonState, next_PC), "next_PC"); + hex_this_PC = tcg_global_mem_new(cpu_env, + offsetof(CPUHexagonState, this_PC), "this_PC"); + hex_slot_cancelled = tcg_global_mem_new(cpu_env, + offsetof(CPUHexagonState, slot_cancelled), "slot_cancelled"); + hex_branch_taken = tcg_global_mem_new(cpu_env, + offsetof(CPUHexagonState, branch_taken), "branch_taken"); + hex_dczero_addr = tcg_global_mem_new(cpu_env, + offsetof(CPUHexagonState, dczero_addr), "dczero_addr"); + llsc_addr = tcg_global_mem_new(cpu_env, + offsetof(CPUHexagonState, llsc_addr), "llsc_addr"); + llsc_val = tcg_global_mem_new(cpu_env, + offsetof(CPUHexagonState, llsc_val), "llsc_val"); + llsc_val_i64 = tcg_global_mem_new_i64(cpu_env, + offsetof(CPUHexagonState, llsc_val_i64), "llsc_val_i64"); + hex_is_gather_store_insn = tcg_global_mem_new(cpu_env, + offsetof(CPUHexagonState, is_gather_store_insn), + "is_gather_store_insn"); + hex_gather_issued = tcg_global_mem_new(cpu_env, + offsetof(CPUHexagonState, gather_issued), "gather_issued"); + for (i = 0; i < STORES_MAX; i++) { + sprintf(store_addr_names[i], "store_addr_%d", i); + hex_store_addr[i] = tcg_global_mem_new(cpu_env, + offsetof(CPUHexagonState, mem_log_stores[i].va), + store_addr_names[i]); + + sprintf(store_width_names[i], "store_width_%d", i); + hex_store_width[i] = tcg_global_mem_new(cpu_env, + offsetof(CPUHexagonState, mem_log_stores[i].width), + store_width_names[i]); + + sprintf(store_val32_names[i], "store_val32_%d", i); + hex_store_val32[i] = tcg_global_mem_new(cpu_env, + offsetof(CPUHexagonState, mem_log_stores[i].data32), + store_val32_names[i]); + + sprintf(store_val64_names[i], "store_val64_%d", i); + hex_store_val64[i] = tcg_global_mem_new_i64(cpu_env, + offsetof(CPUHexagonState, mem_log_stores[i].data64), + store_val64_names[i]); + } + + init_genptr(); +}