[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] SH: Implement MOVCO.L and MOVLI.L
From: |
Vladimir Prus |
Subject: |
Re: [Qemu-devel] SH: Implement MOVCO.L and MOVLI.L |
Date: |
Thu, 12 Feb 2009 00:05:15 +0300 |
User-agent: |
KMail/1.9.10 |
On Saturday 07 February 2009 17:24:59 Aurelien Jarno wrote:
> On Thu, Dec 11, 2008 at 09:48:00PM +0300, Vladimir Prus wrote:
> > On Thursday 11 December 2008 21:36:13 Paul Brook wrote:
> > > On Thursday 11 December 2008, Vladimir Prus wrote:
> > > > This patch implements a couple more instructions
> > > > specific to SH4A.
> > >
> > > > + For now, do unconditional move. FIXME.
> > >
> > > I can't make sense of this FIXME.
> >
> > At it happens, it's stale. The entire line you quoted can be removed.
> > I attach the revised patch.
> >
> > - Volodya
> >
> >
>
> > commit 9ba7a6294a809dc11a0348ccca501513dd90ae6e
> > Author: Vladimir Prus <address@hidden>
> > Date: Sun Oct 12 19:36:22 2008 +0400
> >
> > SH: Implement MOVCO.L and MOVLI.L
> >
> > * target-sh4/cpu.h (struct CPUSH4State): New field ldst.
> > * target-sh4/helper.c (do_interrupt): Initialize ldst.
> > * target-sh4/translate.c (cpu_ldst): New.
> > (sh4_translate_init): Initialize cpu_ldst.
> > (_decode_opc): Support MOVCO.L and MOVLI.L.
> >
> > diff --git a/target-sh4/cpu.h b/target-sh4/cpu.h
> > index a0108b4..ae434d1 100644
> > --- a/target-sh4/cpu.h
> > +++ b/target-sh4/cpu.h
> > @@ -135,6 +135,8 @@ typedef struct CPUSH4State {
> > uint32_t prr; /* Processor Revision Register */
> > uint32_t cvr; /* Cache Version Register */
> >
> > + uint32_t ldst;
> > +
> > CPU_COMMON tlb_t utlb[UTLB_SIZE]; /* unified translation table */
> > tlb_t itlb[ITLB_SIZE]; /* instruction translation table */
> > void *intc_handle;
> > diff --git a/target-sh4/helper.c b/target-sh4/helper.c
> > index 882bc9c..61b668b 100644
> > --- a/target-sh4/helper.c
> > +++ b/target-sh4/helper.c
> > @@ -160,6 +160,7 @@ void do_interrupt(CPUState * env)
> > env->spc = env->pc;
> > env->sgr = env->gregs[15];
> > env->sr |= SR_BL | SR_MD | SR_RB;
> > + env->ldst = 0;
> >
> > if (env->flags & (DELAY_SLOT | DELAY_SLOT_CONDITIONAL)) {
> > /* Branch instruction should be executed again before delay slot.
> > */
> > diff --git a/target-sh4/translate.c b/target-sh4/translate.c
> > index d38e95f..e4a1b28 100644
> > --- a/target-sh4/translate.c
> > +++ b/target-sh4/translate.c
> > @@ -72,7 +72,7 @@ static TCGv_ptr cpu_env;
> > static TCGv cpu_gregs[24];
> > static TCGv cpu_pc, cpu_sr, cpu_ssr, cpu_spc, cpu_gbr;
> > static TCGv cpu_vbr, cpu_sgr, cpu_dbr, cpu_mach, cpu_macl;
> > -static TCGv cpu_pr, cpu_fpscr, cpu_fpul;
> > +static TCGv cpu_pr, cpu_fpscr, cpu_fpul, cpu_ldst;
> > static TCGv cpu_fregs[32];
> >
> > /* internal register indexes */
> > @@ -144,6 +144,8 @@ static void sh4_translate_init(void)
> > cpu_delayed_pc = tcg_global_mem_new_i32(TCG_AREG0,
> > offsetof(CPUState, delayed_pc),
> > "_delayed_pc_");
> > + cpu_ldst = tcg_global_mem_new_i32(TCG_AREG0,
> > + offsetof(CPUState, ldst), "_ldst_");
> >
> > for (i = 0; i < 32; i++)
> > cpu_fregs[i] = tcg_global_mem_new_i32(TCG_AREG0,
> > @@ -1543,6 +1545,36 @@ static void _decode_opc(DisasContext * ctx)
> > case 0x0029: /* movt Rn */
> > tcg_gen_andi_i32(REG(B11_8), cpu_sr, SR_T);
> > return;
> > + case 0x0073:
> > + /* MOVCO.L
> > + LDST -> T
> > + If (T == 1) R0 -> (Rn)
> > + 0 -> LDST
> > + */
> > + if (ctx->features & SH_FEATURE_SH4A) {
> > + int label = gen_new_label();
> > + gen_clr_t();
> > + tcg_gen_or_i32(cpu_sr, cpu_sr, cpu_ldst);
> > + tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_ldst, 0, label);
> > + tcg_gen_qemu_st32(REG(0), REG(B11_8), ctx->memidx);
> > + gen_set_label(label);
> > + tcg_gen_movi_i32(cpu_ldst, 0);
> > + return;
> > + } else
> > + break;
> > + case 0x0063:
> > + /* MOVLI.L @Rm,R0
> > + 1 -> LDST
> > + (Rm) -> R0
> > + When interrupt/exception
> > + occurred 0 -> LDST
> > + */
> > + if (ctx->features & SH_FEATURE_SH4A) {
> > + tcg_gen_movi_i32(cpu_ldst, 1);
> > + tcg_gen_qemu_ld32s(REG(0), REG(B11_8), ctx->memidx);
> > + return;
>
> This TCG code doesn't seems to match the comments. With the current
> code, if a load exception occurs, LDST end with a value of 1 instead of
> 0.
I might be missing something -- it was my intention that if any exception
or interrupt happens, do_interrupt will set ldst to zero. And if the exception
happens during load, there's nothing else that will set ldst back to 1. Did
I miss something?
- Volodya