[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] TCG native 32->64 concatenation
From: |
Paul Brook |
Subject: |
Re: [Qemu-devel] TCG native 32->64 concatenation |
Date: |
Sun, 7 Sep 2008 19:43:03 +0100 |
User-agent: |
KMail/1.9.9 |
On Sunday 07 September 2008, Blue Swirl wrote:
> On 9/7/08, Paul Brook <address@hidden> wrote:
> > The patch below adds a new concat_i32_i64 TCG op. This allows a pair of
> > 32-bit values to be efficiently combined to form a 64-bit value. I've
> > converted all the cases I could find to use this, and tested the arm
> > code on both 32 and 64-bit hosts.
>
> Sparc's helper_pack64 does the same thing, I used it to reduce
> register pressure on i386 for 64 bit stores.
They're slightly different because helper_pack64 takes target_ulong arguments.
Replacing helper_pack64 with concat_i32_i64 would need explicit truncation on
64-bit targets. The current TCG implementation will work if you use the wrong
register type, however I don't want targets relying on that.
Paul