-
Peter Zijlstra authored
Replace the instruction::list by allocating instructions in arrays of 256 entries and stringing them together by (amortized) find_insn(). This shrinks instruction by 16 bytes and brings it down to 128. struct instruction { - struct list_head list; /* 0 16 */ - struct hlist_node hash; /* 16 16 */ - struct list_head call_node; /* 32 16 */ - struct section * sec; /* 48 8 */ - long unsigned int offset; /* 56 8 */ - /* --- cacheline 1 boundary (64 bytes) --- */ - long unsigned int immediate; /* 64 8 */ - unsigned int len; /* 72 4 */ - u8 type; /* 76 1 */ - - /* Bitfield combined with previous fields */ + struct hlist_node hash; /* 0 16 */ + struct list_head call_node; /* 16 16 */ + struct section * sec; /* 32 8 */ + long unsigned int offset; /* 40 8 */ + long unsigned int immediate; /* 48 8 */ + u8 len; /* 56 1 */ + u8 prev_len; /* 57 1 */ + u8 type; /* 58 1 */ + s8 instr; /* 59 1 */ + u32 idx:8; /* 60: 0 4 */ + u32 dead_end:1; /* 60: 8 4 */ + u32 ignore:1; /* 60: 9 4 */ + u32 ignore_alts:1; /* 60:10 4 */ + u32 hint:1; /* 60:11 4 */ + u32 save:1; /* 60:12 4 */ + u32 restore:1; /* 60:13 4 */ + u32 retpoline_safe:1; /* 60:14 4 */ + u32 noendbr:1; /* 60:15 4 */ + u32 entry:1; /* 60:16 4 */ + u32 visited:4; /* 60:17 4 */ + u32 no_reloc:1; /* 60:21 4 */ - u16 dead_end:1; /* 76: 8 2 */ - u16 ignore:1; /* 76: 9 2 */ - u16 ignore_alts:1; /* 76:10 2 */ - u16 hint:1; /* 76:11 2 */ - u16 save:1; /* 76:12 2 */ - u16 restore:1; /* 76:13 2 */ - u16 retpoline_safe:1; /* 76:14 2 */ - u16 noendbr:1; /* 76:15 2 */ - u16 entry:1; /* 78: 0 2 */ - u16 visited:4; /* 78: 1 2 */ - u16 no_reloc:1; /* 78: 5 2 */ + /* XXX 10 bits hole, try to pack */ - /* XXX 2 bits hole, try to pack */ - /* Bitfield combined with next fields */ - - s8 instr; /* 79 1 */ - struct alt_group * alt_group; /* 80 8 */ - struct instruction * jump_dest; /* 88 8 */ - struct instruction * first_jump_src; /* 96 8 */ + /* --- cacheline 1 boundary (64 bytes) --- */ + struct alt_group * alt_group; /* 64 8 */ + struct instruction * jump_dest; /* 72 8 */ + struct instruction * first_jump_src; /* 80 8 */ union { - struct symbol * _call_dest; /* 104 8 */ - struct reloc * _jump_table; /* 104 8 */ - }; /* 104 8 */ - struct alternative * alts; /* 112 8 */ - struct symbol * sym; /* 120 8 */ - /* --- cacheline 2 boundary (128 bytes) --- */ - struct stack_op * stack_ops; /* 128 8 */ - struct cfi_state * cfi; /* 136 8 */ + struct symbol * _call_dest; /* 88 8 */ + struct reloc * _jump_table; /* 88 8 */ + }; /* 88 8 */ + struct alternative * alts; /* 96 8 */ + struct symbol * sym; /* 104 8 */ + struct stack_op * stack_ops; /* 112 8 */ + struct cfi_state * cfi; /* 120 8 */ - /* size: 144, cachelines: 3, members: 28 */ - /* sum members: 142 */ - /* sum bitfield members: 14 bits, bit holes: 1, sum bit holes: 2 bits */ - /* last cacheline: 16 bytes */ + /* size: 128, cachelines: 2, members: 29 */ + /* sum members: 124 */ + /* sum bitfield members: 22 bits, bit holes: 1, sum bit holes: 10 bits */ }; pre: 5:38.18 real, 213.25 user, 124.90 sys, 23449040 mem post: 5:03.34 real, 210.75 user, 88.80 sys, 20241232 mem Signed-off-by:Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by:
Ingo Molnar <mingo@kernel.org> Acked-by:
Josh Poimboeuf <jpoimboe@kernel.org> Tested-by: Nathan Chancellor <nathan@kernel.org> # build only Tested-by: Thomas Weißschuh <linux@weissschuh.net> # compile and run Link: https://lore.kernel.org/r/20230208172245.851307606@infradead.org
1c34496e
