On 9/30/24 00:34, Philippe Mathieu-Daudé wrote:
Introduce the ld/st_endian_phys() API, which takes an extra
boolean argument to dispatch to ld/st_{be,le}_phys() methods.
Signed-off-by: Philippe Mathieu-Daudé <phi...@linaro.org>
---
TODO: Update docstring regexp
---
include/exec/memory_ldst_phys.h.inc | 66 +++++++++++++++++++++++++++++
1 file changed, 66 insertions(+)
diff --git a/include/exec/memory_ldst_phys.h.inc
b/include/exec/memory_ldst_phys.h.inc
index ecd678610d..8ea162b40d 100644
--- a/include/exec/memory_ldst_phys.h.inc
+++ b/include/exec/memory_ldst_phys.h.inc
@@ -74,6 +74,16 @@ static inline uint16_t glue(lduw_be_phys, SUFFIX)(ARG1_DECL,
hwaddr addr)
MEMTXATTRS_UNSPECIFIED, NULL);
}
+static inline uint16_t glue(lduw_endian_phys, SUFFIX)(bool big_endian,
+ ARG1_DECL, hwaddr addr)
+{
+ return big_endian
+ ? glue(address_space_lduw_le, SUFFIX)(ARG1, addr,
+ MEMTXATTRS_UNSPECIFIED, NULL)
+ : glue(address_space_lduw_be, SUFFIX)(ARG1, addr,
+ MEMTXATTRS_UNSPECIFIED, NULL);
+}
Endian swap aside, I think you should expose this at the address_space_* level first,
where the internals already have
return glue(address_space_lduw_internal, SUFFIX)(ARG1, addr, attrs, result,
DEVICE_LITTLE_ENDIAN);
then you can pass big_endian directly (perhaps frobbed into DEVICE_* space).
That leaves one unconditional function call instead of two conditional calls.
r~
