Right now, our codegen pattern for is
i32.const <R_WASM_MEMORY_ADDR_SLEB>
i32.load
i32.load
call_indirect
We should utilize the WASM_MEMORY_ADD_REL_LEB reloc type to embed the reloc as a load offset, to reduce the instruction footprint for calls, like so:
i32.load offset=<R_WASM_MEMORY_ADDR_LEB>
i32.load
call_indirect
Right now, our codegen pattern for is
We should utilize the
WASM_MEMORY_ADD_REL_LEBreloc type to embed the reloc as a load offset, to reduce the instruction footprint for calls, like so: