[Bug target/87832] New: AMD pipeline models are very costly size-wise

[amonakov at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87832

            Bug ID: 87832
           Summary: AMD pipeline models are very costly size-wise
           Product: gcc
           Version: 9.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: amonakov at gcc dot gnu.org
  Target Milestone: ---
            Target: x86_64-*-*, i?86-*-*

Looking at i386 insn-automata.o, out of its 2.2M rodata size almost all is due
to very large tables for AMD CPU models. Note how znver additions are more than
half of overall size.

What is causing that and can it be improved?

2176 core2_core_transitions
2496 slm_base
2527 bdver3_load_min_issue_delay
2746 glm_base
3892 bdver1_fp_base
4261 insn_latency(rtx_insn*, rtx_insn*)
4444 bdver1_ieu_min_issue_delay
4492 geode_base
4608 bdver3_ieu_transitions
6402 bdver1_load_transitions
7862 athlon_fp_check
7862 athlon_fp_transitions
9433 internal_min_issue_delay(int, DFA_chip*)
10108 bdver3_load_transitions
10360 print_reservation(_IO_FILE*, rtx_insn*)::reservation_names
10498 geode_check
10498 geode_transitions
12575 athlon_fp_min_issue_delay
12599 internal_state_transition(int, DFA_chip*)
12742 btver2_fp_check
12742 btver2_fp_transitions
13896 slm_transitions
13896 slm_check
17776 bdver1_ieu_transitions
20068 bdver1_fp_check
20068 bdver1_fp_transitions
26208 slm_min_issue_delay
27244 bdver1_fp_min_issue_delay
28518 glm_transitions
28518 glm_check
33690 geode_min_issue_delay
46980 bdver3_fp_min_issue_delay
49428 glm_min_issue_delay
53730 btver2_fp_min_issue_delay
68160 znver1_ieu_min_issue_delay
93960 bdver3_fp_transitions
136320 znver1_ieu_transitions
428108 znver1_fp_min_issue_delay
856216 znver1_fp_transitions