From: Richard Henderson <richard.henderson@linaro.org>
To: Song Gao <gaosong@loongson.cn>, qemu-devel@nongnu.org
Cc: peter.maydell@linaro.org, thuth@redhat.com, chenhuacai@gmail.com,
philmd@redhat.com, yangxiaojuan@loongson.cn, laurent@vivier.eu,
maobibo@loongson.cn, alistair.francis@wdc.com,
pbonzini@redhat.com, alex.bennee@linaro.org
Subject: Re: [PATCH v2 03/22] target/loongarch: Add core definition
Date: Thu, 22 Jul 2021 12:43:45 -1000 [thread overview]
Message-ID: <ea7be587-1bc6-f770-1c9f-4f57e1c3c7eb@linaro.org> (raw)
In-Reply-To: <1626861198-6133-4-git-send-email-gaosong@loongson.cn>
On 7/20/21 11:52 PM, Song Gao wrote:
> This patch add target state header, target definitions
> and initialization routines.
>
> Signed-off-by: Song Gao <gaosong@loongson.cn>
> ---
> target/loongarch/cpu-param.h | 21 ++++
> target/loongarch/cpu-qom.h | 40 ++++++
> target/loongarch/cpu.c | 293 +++++++++++++++++++++++++++++++++++++++++++
> target/loongarch/cpu.h | 265 ++++++++++++++++++++++++++++++++++++++
> 4 files changed, 619 insertions(+)
> create mode 100644 target/loongarch/cpu-param.h
> create mode 100644 target/loongarch/cpu-qom.h
> create mode 100644 target/loongarch/cpu.c
> create mode 100644 target/loongarch/cpu.h
>
> diff --git a/target/loongarch/cpu-param.h b/target/loongarch/cpu-param.h
> new file mode 100644
> index 0000000..582ee29
> --- /dev/null
> +++ b/target/loongarch/cpu-param.h
> @@ -0,0 +1,21 @@
> +/*
> + * LoongArch cpu parameters for qemu.
> + *
> + * Copyright (c) 2021 Loongson Technology Corporation Limited
> + *
> + * SPDX-License-Identifier: LGPL-2.1+
> + */
> +
> +#ifndef LOONGARCH_CPU_PARAM_H
> +#define LOONGARCH_CPU_PARAM_H 1
> +
> +#ifdef TARGET_LOONGARCH64
> +#define TARGET_LONG_BITS 64
Why the ifdef for TARGET_LOONGARCH64?
Nothing will compile without that set.
> +#ifdef CONFIG_TCG
> +static void loongarch_cpu_synchronize_from_tb(CPUState *cs,
> + const TranslationBlock *tb)
> +{
> + LoongArchCPU *cpu = LOONGARCH_CPU(cs);
> + CPULoongArchState *env = &cpu->env;
> +
> + env->active_tc.PC = tb->pc;
> + env->hflags &= ~LOONGARCH_HFLAG_BMASK;
> + env->hflags |= tb->flags & LOONGARCH_HFLAG_BMASK;
> +}
Loongarch has no branch delay slots, so you should not have replicated the mips branch
delay slot handling. There should be no BMASK at all.
> +#ifdef CONFIG_TCG
> +#include "hw/core/tcg-cpu-ops.h"
> +
> +static struct TCGCPUOps loongarch_tcg_ops = {
> + .initialize = loongarch_tcg_init,
> + .synchronize_from_tb = loongarch_cpu_synchronize_from_tb,
> +};
> +#endif /* CONFIG_TCG */
May I presume that Loongarch has virtualization hardware, and will eventually support KVM?
If not, there is no need for CONFIG_TCG anywhere.
> +#define TCG_GUEST_DEFAULT_MO (0)
> +#define UNASSIGNED_CPU_ID 0xFFFFFFFF
> +
> +typedef union fpr_t fpr_t;
> +union fpr_t {
> + float64 fd; /* ieee double precision */
> + float32 fs[2];/* ieee single precision */
> + uint64_t d; /* binary double fixed-point */
> + uint32_t w[2]; /* binary single fixed-point */
> +};
For what it's worth, we already have a CPU_DoubleU type that could be used. But frankly,
float64 *is* uint64_t, so there's very little use in putting them together into a union.
It would seem that you don't even use fs and w for more than fpu_dump_state, and you're
even doing it wrong there.
> +typedef struct CPULoongArchFPUContext CPULoongArchFPUContext;
> +struct CPULoongArchFPUContext {
> + /* Floating point registers */
> + fpr_t fpr[32];
> + float_status fp_status;
> +
> + bool cf[8];
> + /*
> + * fcsr0
> + * 31:29 |28:24 |23:21 |20:16 |15:10 |9:8 |7 |6 |5 |4:0
> + * Cause Flags RM DAE TM Enables
> + */
> + uint32_t fcsr0;
> + uint32_t fcsr0_mask;
> + uint32_t vcsr16;
> +
> +#define FCSR0_M1 0xdf /* FCSR1 mask, DAE, TM and Enables */
> +#define FCSR0_M2 0x1f1f0000 /* FCSR2 mask, Cause and Flags */
> +#define FCSR0_M3 0x300 /* FCSR3 mask, Round Mode */
> +#define FCSR0_RM 8 /* Round Mode bit num on fcsr0 */
> +#define GET_FP_CAUSE(reg) (((reg) >> 24) & 0x1f)
> +#define GET_FP_ENABLE(reg) (((reg) >> 0) & 0x1f)
> +#define GET_FP_FLAGS(reg) (((reg) >> 16) & 0x1f)
> +#define SET_FP_CAUSE(reg, v) do { (reg) = ((reg) & ~(0x1f << 24)) | \
> + ((v & 0x1f) << 24); \
> + } while (0)
> +#define SET_FP_ENABLE(reg, v) do { (reg) = ((reg) & ~(0x1f << 0)) | \
> + ((v & 0x1f) << 0); \
> + } while (0)
> +#define SET_FP_FLAGS(reg, v) do { (reg) = ((reg) & ~(0x1f << 16)) | \
> + ((v & 0x1f) << 16); \
> + } while (0)
> +#define UPDATE_FP_FLAGS(reg, v) do { (reg) |= ((v & 0x1f) << 16); } while (0)
> +#define FP_INEXACT 1
> +#define FP_UNDERFLOW 2
> +#define FP_OVERFLOW 4
> +#define FP_DIV0 8
> +#define FP_INVALID 16
> +};
> +
> +#define TARGET_INSN_START_EXTRA_WORDS 2
> +#define LOONGARCH_FPU_MAX 1
> +#define N_IRQS 14
> +
> +enum loongarch_feature {
> + LA_FEATURE_3A5000,
> +};
> +
> +typedef struct TCState TCState;
> +struct TCState {
> + target_ulong gpr[32];
> + target_ulong PC;
> +};
> +
> +typedef struct CPULoongArchState CPULoongArchState;
> +struct CPULoongArchState {
> + TCState active_tc;
> + CPULoongArchFPUContext active_fpu;
Please don't replicate the mips foolishness with active_tc and active_fpu. There is no
inactive_fpu with which to contrast this. Just include these fields directly into the
main CPULoongArchState structure.
> +
> + uint32_t current_tc;
> + uint64_t scr[4];
> + uint32_t current_fpu;
> +
> + /* LoongArch CSR register */
> + CPU_LOONGARCH_CSR
> + target_ulong lladdr; /* LL virtual address compared against SC */
> + target_ulong llval;
> +
> + CPULoongArchFPUContext fpus[LOONGARCH_FPU_MAX];
More copying from MIPS? What is this for?
> +
> + /* QEMU */
> + int error_code;
> + uint32_t hflags; /* CPU State */
> +#define TLB_NOMATCH 0x1
> +#define INST_INAVAIL 0x2 /* Invalid instruction word for BadInstr */
> + /* TMASK defines different execution modes */
> +#define LOONGARCH_HFLAG_TMASK 0x1F5807FF
> +#define LOONGARCH_HFLAG_KU 0x00003 /* kernel/supervisor/user mode mask */
> +#define LOONGARCH_HFLAG_UM 0x00003 /* user mode flag */
> +#define LOONGARCH_HFLAG_KM 0x00000 /* kernel mode flag */
> +#define LOONGARCH_HFLAG_64 0x00008 /* 64-bit instructions enabled */
Is there a 32-bit mode for LoongArch? I don't see this big in CRMD. This big overlaps
the "Direct address translation mode enable bit". Which does sound like it should be
present in tb->flags,
> +#define LOONGARCH_HFLAG_FPU 0x00020 /* FPU enabled */
> +#define LOONGARCH_HFLAG_F64 0x00040 /* 64-bit FPU enabled */
I don't see that there is a mode-switch for a 32-bit fpu either.
> +#define LOONGARCH_HFLAG_BMASK 0x3800
> +#define LOONGARCH_HFLAG_B 0x00800 /* Unconditional branch */
> +#define LOONGARCH_HFLAG_BC 0x01000 /* Conditional branch */
> +#define LOONGARCH_HFLAG_BR 0x02000 /* branch to register (can't link TB) */
None of the BMASK stuff applies to LoongArch.
> +#define LOONGARCH_HFLAG_FRE 0x2000000 /* FRE enabled */
> +#define LOONGARCH_HFLAG_ELPA 0x4000000
> + target_ulong btarget; /* Jump / branch target */
> + target_ulong bcond; /* Branch condition (if needed) */
Nor this.
> +static inline LoongArchCPU *loongarch_env_get_cpu(CPULoongArchState *env)
> +{
> + return container_of(env, LoongArchCPU, env);
> +}
> +
> +#define ENV_GET_CPU(e) CPU(loongarch_env_get_cpu(e))
You have copied this from a very old version of qemu. These were replaced by generic
functions in include/exec/cpu-all.h.
> +void loongarch_tcg_init(void);
> +
> +void loongarch_cpu_dump_state(CPUState *cpu, FILE *f, int flags);
> +
> +void QEMU_NORETURN do_raise_exception_err(CPULoongArchState *env,
> + uint32_t exception,
> + int error_code,
> + uintptr_t pc);
> +
> +static inline void QEMU_NORETURN do_raise_exception(CPULoongArchState *env,
> + uint32_t exception,
> + uintptr_t pc)
> +{
> + do_raise_exception_err(env, exception, 0, pc);
> +}
> +
> +static inline void compute_hflags(CPULoongArchState *env)
> +{
> + env->hflags &= ~(LOONGARCH_HFLAG_64 | LOONGARCH_HFLAG_FPU |
> + LOONGARCH_HFLAG_KU | LOONGARCH_HFLAG_ELPA);
> +
> + env->hflags |= (env->CSR_CRMD & CSR_CRMD_PLV);
> + env->hflags |= LOONGARCH_HFLAG_64;
> +
> + if (env->CSR_EUEN & CSR_EUEN_FPEN) {
> + env->hflags |= LOONGARCH_HFLAG_FPU;
> + }
> +}
> +
> +const char *loongarch_exception_name(int32_t exception);
These should not be declared in cpu.h.
r~
next prev parent reply other threads:[~2021-07-22 22:44 UTC|newest]
Thread overview: 76+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-07-21 9:52 [PATCH v2 00/22] Add LoongArch linux-user emulation support Song Gao
2021-07-21 9:52 ` [PATCH v2 01/22] target/loongarch: Add README Song Gao
2021-07-21 9:52 ` [PATCH v2 02/22] target/loongarch: Add CSR registers definition Song Gao
2021-07-21 9:52 ` [PATCH v2 03/22] target/loongarch: Add core definition Song Gao
2021-07-22 22:43 ` Richard Henderson [this message]
2021-07-26 8:47 ` Song Gao
2021-07-26 15:32 ` Richard Henderson
2021-07-21 9:53 ` [PATCH v2 04/22] target/loongarch: Add interrupt handling support Song Gao
2021-07-22 22:47 ` Richard Henderson
2021-07-26 9:23 ` Song Gao
2021-07-21 9:53 ` [PATCH v2 05/22] target/loongarch: Add memory management support Song Gao
2021-07-22 22:48 ` Richard Henderson
2021-07-26 9:25 ` Song Gao
2021-07-21 9:53 ` [PATCH v2 06/22] target/loongarch: Add main translation routines Song Gao
2021-07-22 23:50 ` Richard Henderson
2021-07-26 9:39 ` Song Gao
2021-07-26 15:35 ` Richard Henderson
2021-07-21 9:53 ` [PATCH v2 07/22] target/loongarch: Add fixed point arithmetic instruction translation Song Gao
2021-07-21 17:38 ` Philippe Mathieu-Daudé
2021-07-21 17:49 ` Philippe Mathieu-Daudé
2021-07-22 7:41 ` Song Gao
2021-07-23 0:46 ` Richard Henderson
2021-07-26 11:56 ` Song Gao
2021-07-26 15:53 ` Richard Henderson
2021-07-27 1:51 ` Song Gao
2021-07-21 9:53 ` [PATCH v2 08/22] target/loongarch: Add fixed point shift " Song Gao
2021-07-23 0:51 ` Richard Henderson
2021-07-26 11:57 ` Song Gao
2021-07-21 9:53 ` [PATCH v2 09/22] target/loongarch: Add fixed point bit " Song Gao
2021-07-21 17:46 ` Philippe Mathieu-Daudé
2021-07-22 8:17 ` Song Gao
2021-07-23 1:29 ` Richard Henderson
2021-07-26 12:22 ` Song Gao
2021-07-26 16:39 ` Richard Henderson
2021-07-21 9:53 ` [PATCH v2 10/22] target/loongarch: Add fixed point load/store " Song Gao
2021-07-23 1:45 ` Richard Henderson
2021-07-26 12:25 ` Song Gao
2021-07-21 9:53 ` [PATCH v2 11/22] target/loongarch: Add fixed point atomic " Song Gao
2021-07-23 1:49 ` Richard Henderson
2021-07-26 12:25 ` Song Gao
2021-07-21 9:53 ` [PATCH v2 12/22] target/loongarch: Add fixed point extra " Song Gao
2021-07-23 5:12 ` Richard Henderson
2021-07-26 12:57 ` Song Gao
2021-07-26 16:42 ` Richard Henderson
2021-07-27 1:46 ` Song Gao
2021-08-04 7:40 ` Song Gao
2021-08-04 7:51 ` Song Gao
2021-07-21 9:53 ` [PATCH v2 13/22] target/loongarch: Add floating point arithmetic " Song Gao
2021-07-23 5:44 ` Richard Henderson
2021-07-27 7:17 ` Song Gao
2021-07-27 16:12 ` Richard Henderson
2021-07-28 1:18 ` Song Gao
2021-07-21 9:53 ` [PATCH v2 14/22] target/loongarch: Add floating point comparison " Song Gao
2021-07-23 6:11 ` Richard Henderson
2021-07-27 7:56 ` Song Gao
2021-07-21 9:53 ` [PATCH v2 15/22] target/loongarch: Add floating point conversion " Song Gao
2021-07-23 6:16 ` Richard Henderson
2021-07-27 7:57 ` Song Gao
2021-07-21 9:53 ` [PATCH v2 16/22] target/loongarch: Add floating point move " Song Gao
2021-07-23 6:29 ` Richard Henderson
2021-07-27 8:06 ` Song Gao
2021-08-12 9:20 ` Song Gao
2021-08-12 19:31 ` Richard Henderson
2021-07-21 9:53 ` [PATCH v2 17/22] target/loongarch: Add floating point load/store " Song Gao
2021-07-23 6:34 ` Richard Henderson
2021-07-27 8:07 ` Song Gao
2021-07-21 9:53 ` [PATCH v2 18/22] target/loongarch: Add branch " Song Gao
2021-07-23 6:38 ` Richard Henderson
2021-07-27 8:07 ` Song Gao
2021-07-21 9:53 ` [PATCH v2 19/22] target/loongarch: Add disassembler Song Gao
2021-07-23 6:40 ` Richard Henderson
2021-08-12 10:33 ` Philippe Mathieu-Daudé
2021-07-21 9:53 ` [PATCH v2 20/22] LoongArch Linux User Emulation Song Gao
2021-07-21 9:53 ` [PATCH v2 21/22] configs: Add loongarch linux-user config Song Gao
2021-07-23 6:43 ` Richard Henderson
2021-07-21 9:53 ` [PATCH v2 22/22] target/loongarch: Add target build suport Song Gao
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=ea7be587-1bc6-f770-1c9f-4f57e1c3c7eb@linaro.org \
--to=richard.henderson@linaro.org \
--cc=alex.bennee@linaro.org \
--cc=alistair.francis@wdc.com \
--cc=chenhuacai@gmail.com \
--cc=gaosong@loongson.cn \
--cc=laurent@vivier.eu \
--cc=maobibo@loongson.cn \
--cc=pbonzini@redhat.com \
--cc=peter.maydell@linaro.org \
--cc=philmd@redhat.com \
--cc=qemu-devel@nongnu.org \
--cc=thuth@redhat.com \
--cc=yangxiaojuan@loongson.cn \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).