xref: /llvm-project/bolt/test/X86/tail-duplication-cache.s (revision 11791ae7b0b05b8bd8d806331ff51da618912cf8)
15904836bSspupyrev# REQUIRES: system-linux
25904836bSspupyrev
35904836bSspupyrev# RUN: llvm-mc -filetype=obj -triple x86_64-unknown-unknown \
45904836bSspupyrev# RUN:   %s -o %t.o
55904836bSspupyrev# RUN: link_fdata %s %t.o %t.fdata
65904836bSspupyrev# RUN: link_fdata %s %t.o %t.fdata2 "FDATA2"
75904836bSspupyrev# RUN: %clang %cflags %t.o -o %t.exe -Wl,-q
8d648aa1bSMaksim Panchenko# RUN: llvm-bolt %t.exe --data %t.fdata --reorder-blocks=none \
9d648aa1bSMaksim Panchenko# RUN:    --print-finalized --tail-duplication=cache -o %t.out | FileCheck %s
10d648aa1bSMaksim Panchenko# RUN: llvm-bolt %t.exe --data %t.fdata2 --reorder-blocks=none \
11d648aa1bSMaksim Panchenko# RUN:    --print-finalized --tail-duplication=cache -o %t.out2 \
12d648aa1bSMaksim Panchenko# RUN:    | FileCheck --check-prefix="CHECK2" %s
135904836bSspupyrev
14*11791ae7SSayhaan Siddiqui## A test where the tail is duplicated to eliminate an unconditional jump
155904836bSspupyrev# FDATA: 1 main #.BB0_br# 1 main #.BB4# 0 100
165904836bSspupyrev# FDATA: 1 main #.BB0_br# 1 main #.BB1# 0 100
175904836bSspupyrev# FDATA: 1 main #.BB1_br# 1 main #.BB3# 0 50
185904836bSspupyrev# FDATA: 1 main #.BB1_br# 1 main #.BB2# 0 50
195904836bSspupyrev# FDATA: 1 main #.BB3_br# 1 main #.BB2# 0 50
205904836bSspupyrev# CHECK: BOLT-INFO: tail duplication modified 1 ({{.*}}%) functions; duplicated 1 blocks (13 bytes) responsible for 50 dynamic executions ({{.*}}% of all block executions)
215904836bSspupyrev# CHECK: BB Layout   : .LBB00, .Ltmp0, .Ltmp1, .Ltmp2, .Ltmp3, .Ltmp4, .Ltmp5, .Ltail-dup0, .Ltmp6
225904836bSspupyrev
23*11791ae7SSayhaan Siddiqui## A test where the tail is not duplicated due to the cache score
245904836bSspupyrev# FDATA2: 1 main #.BB0_br# 1 main #.BB4# 0 100
255904836bSspupyrev# FDATA2: 1 main #.BB0_br# 1 main #.BB1# 0 2
265904836bSspupyrev# FDATA2: 1 main #.BB1_br# 1 main #.BB3# 0 1
275904836bSspupyrev# FDATA2: 1 main #.BB1_br# 1 main #.BB2# 0 1
285904836bSspupyrev# FDATA2: 1 main #.BB3_br# 1 main #.BB2# 0 1
295904836bSspupyrev# CHECK2: BOLT-INFO: tail duplication modified 0 (0.00%) functions; duplicated 0 blocks (0 bytes) responsible for 0 dynamic executions (0.00% of all block executions)
305904836bSspupyrev# CHECK2: BB Layout   : .LBB00, .Ltmp0, .Ltmp1, .Ltmp2, .Ltmp3, .Ltmp4, .Ltmp5, .Ltmp6
315904836bSspupyrev
325904836bSspupyrev    .text
335904836bSspupyrev    .globl main
345904836bSspupyrev    .type main, %function
355904836bSspupyrev    .size main, .Lend-main
365904836bSspupyrevmain:
375904836bSspupyrev.BB0:
385904836bSspupyrev    xor %eax, %eax
395904836bSspupyrev  	cmpl	%eax, %ebx
405904836bSspupyrev.BB0_br:
415904836bSspupyrev    je      .BB4
425904836bSspupyrev.BB1:
435904836bSspupyrev    inc %rax
445904836bSspupyrev.BB1_br:
455904836bSspupyrev    je .BB3
465904836bSspupyrev.BB2:
475904836bSspupyrev    inc %rax
485904836bSspupyrev    inc %rax
495904836bSspupyrev    inc %rax
505904836bSspupyrev    inc %rax
515904836bSspupyrev    retq
525904836bSspupyrev.BB3:
535904836bSspupyrev    inc %rax
545904836bSspupyrev.BB3_br:
555904836bSspupyrev    jmp .BB2
565904836bSspupyrev.BB4:
575904836bSspupyrev    retq
585904836bSspupyrev# For relocations against .text
595904836bSspupyrev    call exit
605904836bSspupyrev.Lend:
61