From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from [140.186.70.92] (port=42305 helo=eggs.gnu.org)
	by lists.gnu.org with esmtp (Exim 4.43) id 1Pi9zc-0005ta-Cj
	for qemu-devel@nongnu.org; Wed, 26 Jan 2011 13:21:22 -0500
Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71)
	(envelope-from <rth@twiddle.net>) id 1Pi9zY-0007Qs-7C
	for qemu-devel@nongnu.org; Wed, 26 Jan 2011 13:21:20 -0500
Received: from b.mail.sonic.net ([64.142.19.5]:45940)
	by eggs.gnu.org with esmtp (Exim 4.71)
	(envelope-from <rth@twiddle.net>) id 1Pi9zX-0007QT-W2
	for qemu-devel@nongnu.org; Wed, 26 Jan 2011 13:21:16 -0500
Message-ID: <4D40660F.9070004@twiddle.net>
Date: Wed, 26 Jan 2011 10:21:03 -0800
From: Richard Henderson <rth@twiddle.net>
MIME-Version: 1.0
Subject: Re: [Qemu-devel] [PATCH 5/7] tcg-i386: Implement deposit operation.
References: <1294716228-9299-1-git-send-email-rth@twiddle.net>
	<1294716228-9299-6-git-send-email-rth@twiddle.net>
	<20110125122749.GA19736@edde.se.axis.com>
	<4D3EF6C1.3080502@twiddle.net> <20110125164816.GA23569@laped.lan>
	<4D3F4993.4010109@twiddle.net> <20110126085338.GA26088@laped.lan>
	<CB925427-781D-4117-804A-B705B322397E@suse.de>
	<4D4042B4.4020805@twiddle.net> <4D404533.3060708@suse.de>
	<4D404E66.4020109@twiddle.net> <4D4050E3.9010207@suse.de>
In-Reply-To: <4D4050E3.9010207@suse.de>
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
List-Id: qemu-devel.nongnu.org
List-Unsubscribe: <http://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=unsubscribe>
List-Archive: <http://lists.nongnu.org/archive/html/qemu-devel>
List-Post: <mailto:qemu-devel@nongnu.org>
List-Help: <mailto:qemu-devel-request@nongnu.org?subject=help>
List-Subscribe: <http://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=subscribe>
To: Alexander Graf <agraf@suse.de>
Cc: "Edgar E.Iglesias" <edgar.iglesias@gmail.com>, qemu-devel@nongnu.org, aurelien@aurel32.net

On 01/26/2011 08:50 AM, Alexander Graf wrote:
> Oh, you mean basically to have the following:
> 
> TCGv_i32 regs32[16];
> TCGv_i64 regs[16];
> 
> Then declare both as globals with offset and just switch between the
> access type using a disas struct variable. Once the TB ends, I'd
> obviously have to sync back to 64 bit again.

Maybe?  I don't think that regs32 can overlap with regs in the real
cpu structure.  Otherwise you have the same sort of sync problems as
was originally rejected.

I had been picturing regs32 as a member of DisasContext, and it would
hold tcg_temp_new_i32 variables as-needed.  Something like

#if TCG_TARGET_REG_BITS == 64

static void writeback_reg32(DisasContext *dc, int r)
{
    // ??? This macro should really exist to match TCGV_UNUSED_I32 etc.
    if (!TCGV_IS_UNUSED_I32 (dc->regs32[r])) {
        tcg_gen_deposit_tl(cpu_regs[r], cpu_regs[r],
                           MAKE_TCGV_I64 (GET_TCGV_I32 (dc->regs32[r])),
                           0, 32);
    }
}

static void writeback_all_reg32(DisasContext *dc)
{
    int i;
    for (i = 0; i < 16; ++i) {
        flush_reg32(dc, i);
    }
}

static void flush_all_reg32(DisasContext *dc)
{
    int i;
    for (i = 0; i < 16; ++i) {
        if (!TCGV_IS_UNUSED_I32 (dc->regs32[r])) {
            tcg_temp_free_i32 (dc->regs32[r]);
            TCGV_UNUSED_I32 (dc->regs32[r]);
        }
    }
}

static TCGv_i32 get_reg32(DisasContext *dc, int r)
{
    if (TCGV_IS_UNUSED_I32 (dc->regs32[r])) {
        dc->regs32[r] = tcg_temp_new_i32();
        tcg_gen_trunc_i64_i32(dc->regs32[r], cpu_regs[r]);
    }
    return dc->regs32[r];
}

static TCGv_i64 get_reg64(DisasContext *dc, int r)
{
    writeback_reg32(dc, r);
    return cpu_regs[r];
}

#elif TCG_TARGET_REG_BITS == 32

static void writeback_reg32(DisasContext *dc, int r) { }
static void writeback_all_reg32(DisasContext *dc) { }
static void flush_all_reg32(DisasContext *dc) { }

static TCGv_i32 get_reg32(DisasContext *dc, int r)
{
    return TCGV_LOW(cpu_regs[r]);
}

static TCGv_i64 get_reg64(DisasContext *dc, int r)
{
    return cpu_regs[r];
}

#endif


r~