On 1/10/26 00:26, Richard Henderson wrote:
On 1/10/26 04:23, Paolo Bonzini wrote:
Il ven 9 gen 2026, 17:51 Philippe Mathieu-Daudé <[email protected]
<mailto:[email protected]>> ha scritto:
Introduce load/store helpers which take a MemOp argument.
Inspired-by: Paolo Bonzini <[email protected]
<mailto:[email protected]>>
That's a new one. :)
I think they should be inline and so should be address_space_{ld,st}
m_internal (maybe even always_inline). The amount of optimization
enabled by constant MemOp is huge.
When are we going to have constant MemOp?
Every time address_space_{ld,st}m_internal is called, it's done with
constant MemOp. For example:
uint16_t ADDRESS_SPACE_LD(uw)(ARG1_DECL, hwaddr addr,
MemTxAttrs attrs, MemTxResult *result)
{
return ADDRESS_SPACE_LD_INTERNAL(m)(ARG1, MO_ENDIAN | MO_16,
addr, attrs, result);
}
Paolo
