[Bug target/112936] New: LoongArch: Wrong instruction costs

[xry111 at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112936

            Bug ID: 112936
           Summary: LoongArch: Wrong instruction costs
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: xry111 at gcc dot gnu.org
  Target Milestone: ---

Separated from PR112935 comment 3.

The current loongarch_rtx_cost_data constructor (for default cost model, used
by LA464 and LA664 for now) seems completely wrong.  Per my micro-benchmark the
values in reality should be:

loongarch_rtx_cost_data::loongarch_rtx_cost_data ()
  : fp_add (COSTS_N_INSNS (5)),
    fp_mult_sf (COSTS_N_INSNS (5)),
    fp_mult_df (COSTS_N_INSNS (5)),
    fp_div_sf (COSTS_N_INSNS (8)),
    fp_div_df (COSTS_N_INSNS (8)),
    int_mult_si (COSTS_N_INSNS (4)),
    int_mult_di (COSTS_N_INSNS (4)),
    int_div_si (COSTS_N_INSNS (5)),
    int_div_di (COSTS_N_INSNS (5)),
    branch_cost (6),
    memory_latency (4) {}