|
From: | Philippe Mathieu-Daudé |
Subject: | Re: [PATCH v2 06/23] target/loongarch: Store some uint64_t values as target_ulong |
Date: | Fri, 27 Dec 2024 00:02:36 +0100 |
User-agent: | Mozilla Thunderbird |
On 26/12/24 23:58, Jiaxun Yang wrote:
在2024年12月26日十二月 下午10:48,Philippe Mathieu-Daudé写道:On 26/12/24 22:19, Jiaxun Yang wrote:Store internal registers including GPRs, CSRs, and LBT scratchs as target_ulong, as per architecture specification. The only exception here is tlb_misc, as it's only used by QEMU internally and need keep to be 64bit to store all required fields. There is no migration ABI change, as target_ulong is uint64_t on existing loongarch64 builds anyway.I'm working on a prototype series where target_ulong becomes uint64_t under TCG. This patch is going the opposite direction. Not sure what to say at this point (I'm not rejecting it, just wondering).Does this mean TCGv will always be 64bit? I'm doing this just to keep all TCG _tl load/store sane. If that's the way to go, I'm happy to adopt full uint64_t approach.
"full uint64_t approach" won't happen before months (except if we get more manpower). The idea is to add all current generic TCG ops but taking an extra MemOp argument, having the backend sign-extending / truncating.
[Prev in Thread] | Current Thread | [Next in Thread] |