LoongArch: Add RTX cost support for ORDERED comparison.
The LoongArch backend now defines the RTX cost for the ORDERED operation to prevent incorrect cost evaluations and test failures.
The LoongArch backend did not define the RTX cost for the ORDERED operation. This led to incorrect cost evaluations where UNORDERED was preferred over ORDERED, causing test failures. Since the assembly sequences are identical, this commit sets the RTX cost for ORDERED to match UNORDERED, ensuring consistent cost evaluation and resolving the test failures.
In Details
The LoongArch backend defines the loongarch_rtx_costs function to specify the cost of various RTX codes. This patch adds handling for the ORDERED RTX code to prevent inversions in cost calculations. Search snippets should relate to the LoongArch architecture and RTX costs.
For Context
In GCC, RTL (Register Transfer Language) is an intermediate representation used during compilation. RTX codes represent specific operations in RTL. The compiler uses a cost model to estimate the resources required for different operations to make optimization decisions. This commit addresses an issue in the LoongArch backend where the cost of the ORDERED operation was not defined, leading to suboptimal code generation.