forked from OSchip/llvm-project
a1532ed275
Change `CountersPtr` in `__profd_` to a label difference, which is a link-time
constant. On ELF, when linking a shared object, this requires that `__profc_` is
either private or linkonce/linkonce_odr hidden. On COFF, we need D104564 so that
`.quad a-b` (64-bit label difference) can lower to a 32-bit PC-relative relocation.
```
# ELF: R_X86_64_PC64 (PC-relative)
.quad .L__profc_foo-.L__profd_foo
# Mach-O: a pair of 8-byte X86_64_RELOC_UNSIGNED and X86_64_RELOC_SUBTRACTOR
.quad l___profc_foo-l___profd_foo
# COFF: we actually use IMAGE_REL_AMD64_REL32/IMAGE_REL_ARM64_REL32 so
# the high 32-bit value is zero even if .L__profc_foo < .L__profd_foo
# As compensation, we truncate CountersDelta in the header so that
# __llvm_profile_merge_from_buffer and llvm-profdata reader keep working.
.quad .L__profc_foo-.L__profd_foo
```
(Note: link.exe sorts `.lprfc` before `.lprfd` even if the object writer
has `.lprfd` before `.lprfc`, so we cannot work around by reordering
`.lprfc` and `.lprfd`.)
With this change, a stage 2 (`-DLLVM_TARGETS_TO_BUILD=X86 -DLLVM_BUILD_INSTRUMENTED=IR`)
`ld -pie` linked clang is 1.74% smaller due to fewer R_X86_64_RELATIVE relocations.
```
% readelf -r pie | awk '$3~/R.*/{s[$3]++} END {for (k in s) print k, s[k]}'
R_X86_64_JUMP_SLO 331
R_X86_64_TPOFF64 2
R_X86_64_RELATIVE 476059 # was: 607712
R_X86_64_64 2616
R_X86_64_GLOB_DAT 31
```
The absolute function address (used by llvm-profdata to collect indirect call
targets) can be converted to relative as well, but is not done in this patch.
Differential Revision: https://reviews.llvm.org/D104556
|
||
|---|---|---|
| .. | ||
| Inputs | ||
| X86 | ||
| PR28219.ll | ||
| PR41279.ll | ||
| PR41279_2.ll | ||
| bfi_verification.ll | ||
| branch1.ll | ||
| branch2.ll | ||
| callbr.ll | ||
| chr.ll | ||
| comdat_internal.ll | ||
| comdat_rename.ll | ||
| consecutive-zeros.ll | ||
| counter_promo.ll | ||
| counter_promo_exit_catchswitch.ll | ||
| counter_promo_exit_merge.ll | ||
| counter_promo_mexits.ll | ||
| counter_promo_nest-inseltpoison.ll | ||
| counter_promo_nest.ll | ||
| criticaledge.ll | ||
| cspgo_profile_summary.ll | ||
| diag_FE_profile.ll | ||
| diag_mismatch.ll | ||
| diag_no_funcprofdata.ll | ||
| diag_no_profile.ll | ||
| diag_no_value_sites.ll | ||
| do-not-instrument.ll | ||
| fix_bfi.ll | ||
| fix_entry_count.ll | ||
| func_entry.ll | ||
| hash_mismatch_metadata.ll | ||
| icp_covariant_call_return.ll | ||
| icp_covariant_invoke_return.ll | ||
| icp_invoke.ll | ||
| icp_invoke_nouse.ll | ||
| icp_mismatch_msg.ll | ||
| icp_sample.ll | ||
| icp_vararg.ll | ||
| icp_vararg_sret.ll | ||
| indirect_call_annotation.ll | ||
| indirect_call_profile.ll | ||
| indirect_call_profile_funclet.ll | ||
| indirect_call_promotion.ll | ||
| indirect_call_promotion_byval.ll | ||
| indirect_call_promotion_musttail.ll | ||
| indirect_call_promotion_unique.ll | ||
| indirect_call_promotion_vla.ll | ||
| indirectbr.ll | ||
| infinite_loop.ll | ||
| infinite_loop_gen.ll | ||
| instr_entry_bb.ll | ||
| irreducible.ll | ||
| landingpad.ll | ||
| large_count_remarks.ll | ||
| loop1.ll | ||
| loop2.ll | ||
| memcpy.ll | ||
| memop_clone.ll | ||
| memop_hash.ll | ||
| memop_profile_funclet.ll | ||
| memop_size_annotation.ll | ||
| memop_size_from_strlen.ll | ||
| memop_size_opt.ll | ||
| memop_size_opt_skip_ranges_promote_three.ll | ||
| memop_size_opt_zero.ll | ||
| multiple_hash_profile.ll | ||
| noprofile.ll | ||
| noreturncall.ll | ||
| not_promote_ret_exit.ll | ||
| preinline.ll | ||
| remap.ll | ||
| select1.ll | ||
| select2.ll | ||
| select_hash_conflict.ll | ||
| single_bb.ll | ||
| split-indirectbr-critical-edges.ll | ||
| statics_counter_naming.ll | ||
| suppl-profile.ll | ||
| switch.ll | ||
| thinlto_cspgo_gen.ll | ||
| thinlto_cspgo_use.ll | ||
| thinlto_indirect_call_promotion.ll | ||
| thinlto_samplepgo_icp.ll | ||
| thinlto_samplepgo_icp2.ll | ||
| thinlto_samplepgo_icp3.ll | ||
| thinlto_samplepgo_icp_droppeddead.ll | ||
| unreachable_bb.ll | ||