[CSSPGO] LTO option for pseudo probe

[hoy]

Summary

Adding a lld option to support emitting pseudo probe metadata in LTO mode.

[hoy]

Rebasing.

[MaskRay]

lld/ELF/Config.h

178

Omit initializer

lld/test/ELF/lto/pseudo-probe-lto.ll

2

This is not satisfied.

; REQUIRES: x86

22

; CHECK:      .section
; CHECK-NEXT:   .quad [[#GUID]]

33

can some metadata be simplified?

[hoy]

lld/test/ELF/lto/pseudo-probe-lto.ll

33

Yes. I'll reduce unnecessary debug data.

[hoy]

Addressing MaskRay's feedbacks.

[wmi]

lld/test/ELF/lto/pseudo-probe-lto.ll

4

Add the check for the alias option plugin-opt=pseudo-probe-for-profiling?

[hoy]

lld/test/ELF/lto/pseudo-probe-lto.ll

4

Sounds good.

[hoy]

Addressing Wei's feedback.

[wenlei]

lld/ELF/Options.td

553

nit: Encode -> Emit? the clang driver help messages says "Emit pseudo probes for sample profiler", would be good change to "profiling" there too.

[hoy]

lld/ELF/Options.td

553

Good catch, changed Encode and profiling.

[hoy]

Addressing Wenlei's feedback.

[wenlei]

lgtm with minor comment.

lld/ELF/Options.td

553

I actually meant to suggest both lld and clang driver use "Emit pseudo probes for sample profiling", the switch name uses profiling.

[hoy]

Changing help message for the switch.

[wmi]

LGTM.

[MaskRay]

lld/ELF/Options.td

553

Drop period

628

- -> --

[hoy]

Addressing MaskRay's feedbacks.

[MaskRay]

Thanks!

lld/ELF/Config.h

struct Configuration {
   bool gnuUnique;
   bool hasDynSymTab;
   bool ignoreDataAddressEquality;
   bool ignoreFunctionAddressEquality;
   bool ltoCSProfileGenerate;
   bool ltoDebugPassManager;
   bool ltoEmitAsm;
   bool ltoNewPassManager;
+  bool ltoPseudoProbeForProfiling = false;
   bool ltoUniqueBasicBlockSectionNames;
   bool ltoWholeProgramVisibility;
   bool mergeArmExidx;
   bool mipsN32Abi = false;
   bool mmapOutputFile;
   bool nmagic;
   bool noDynamicLinker = false;
   bool noinhibitExec;

lld/ELF/Driver.cpp

config->ltoNewPassManager =
                    LLVM_ENABLE_NEW_PASS_MANAGER);
   config->ltoNewPmPasses = args.getLastArgValue(OPT_lto_newpm_passes);
   config->ltoWholeProgramVisibility =
       args.hasFlag(OPT_lto_whole_program_visibility,
                    OPT_no_lto_whole_program_visibility, false);
   config->ltoo = args::getInteger(args, OPT_lto_O, 2);
   config->ltoObjPath = args.getLastArgValue(OPT_lto_obj_path_eq);
   config->ltoPartitions = args::getInteger(args, OPT_lto_partitions, 1);
+  config->ltoPseudoProbeForProfiling =
+      args.hasArg(OPT_lto_pseudo_probe_for_profiling);
   config->ltoSampleProfile = args.getLastArgValue(OPT_lto_sample_profile);
   config->ltoBasicBlockSections =
       args.getLastArgValue(OPT_lto_basic_block_sections);
   config->ltoUniqueBasicBlockSectionNames =
       args.hasFlag(OPT_lto_unique_basic_block_section_names,
                    OPT_no_lto_unique_basic_block_section_names, false);
   config->mapFile = args.getLastArgValue(OPT_Map);
   config->mipsGotSize = args::getInteger(args, OPT_mips_got_size, 0xfff0);

lld/ELF/LTO.cpp

if (config->ltoBasicBlockSections == "all") {
               MBOrErr.getError().message());
       } else {
         c.Options.BBSectionsFuncListBuf = std::move(*MBOrErr);
       }
       c.Options.BBSections = BasicBlockSection::List;
     }
   }
 
+  c.Options.PseudoProbeForProfiling = config->ltoPseudoProbeForProfiling;
   c.Options.UniqueBasicBlockSectionNames =
       config->ltoUniqueBasicBlockSectionNames;
 
   if (auto relocModel = getRelocModelFromCMModel())
     c.RelocModel = *relocModel;
   else if (config->relocatable)
     c.RelocModel = None;
   else if (config->isPic)

lld/ELF/Options.td

 def lto_cs_profile_file: JJ<"lto-cs-profile-file=">,
   HelpText<"Context sensitive profile file path">;
 def lto_obj_path_eq: JJ<"lto-obj-path=">;
 def lto_sample_profile: JJ<"lto-sample-profile=">,
   HelpText<"Sample profile file path">;
 defm lto_whole_program_visibility: BB<"lto-whole-program-visibility",
   "Asserts that the LTO link has whole program visibility",
   "Asserts that the LTO link does not have whole program visibility">;
+def lto_pseudo_probe_for_profiling: F<"lto-pseudo-probe-for-profiling">,
+  HelpText<"Encode pseudo probes for sample profiling.">;
 def disable_verify: F<"disable-verify">;
 defm mllvm: Eq<"mllvm", "Additional arguments to forward to LLVM's option processing">;
 def opt_remarks_filename: Separate<["--"], "opt-remarks-filename">,
   HelpText<"YAML output file for optimization remarks">;
 defm opt_remarks_hotness_threshold: EEq<"opt-remarks-hotness-threshold",
   "Minimum profile count required for an optimization remark to be output."
   " Use 'auto' to apply the threshold from profile summary.">,
   MetaVarName<"<value>">;
def: J<"plugin-opt=opt-remarks-format=">,
   Alias<opt_remarks_format>,
   HelpText<"Alias for --opt-remarks-format">;
 def: F<"plugin-opt=opt-remarks-with-hotness">,
   Alias<opt_remarks_with_hotness>,
   HelpText<"Alias for --opt-remarks-with_hotness">;
 def: J<"plugin-opt=opt-remarks-hotness-threshold=">,
   Alias<opt_remarks_hotness_threshold>,
   HelpText<"Alias for --opt-remarks-hotness-threshold">;
+def: J<"plugin-opt=pseudo-probe-for-profiling">,
+  Alias<lto_pseudo_probe_for_profiling>, HelpText<"Alias for -lto-pseudo-probe-for-profiling">;
 def: J<"plugin-opt=sample-profile=">,
   Alias<lto_sample_profile>, HelpText<"Alias for --lto-sample-profile">;
 def: F<"plugin-opt=save-temps">, Alias<save_temps>, HelpText<"Alias for --save-temps">;
 def: F<"plugin-opt=thinlto-emit-imports-files">,
   Alias<thinlto_emit_imports_files>,
   HelpText<"Alias for --thinlto-emit-imports-files">;
 def: F<"plugin-opt=thinlto-index-only">,
   Alias<thinlto_index_only>,

lld/test/ELF/lto/pseudo-probe-lto.ll

+; REQUIRES: x86_64-linux
+; RUN: opt < %s -passes=pseudo-probe -function-sections -o %t.o
+; RUN: ld.lld %t.o -shared --lto-pseudo-probe-for-profiling --lto-emit-asm -o - | FileCheck %s
+
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-scei-ps4"
+
+@g = dso_local global i32 3, align 4
+
+define void @foo(void (i32)* %f) !dbg !4 {
+entry:
+; CHECK: .pseudoprobe [[#GUID:]] 1 0 0
+; CHECK: .pseudoprobe [[#GUID]] 2 1 0
+  call void %f(i32 1), !dbg !13
+  %0 = load i32, i32* @g, align 4
+  %inc = add nsw i32 %0, 1
+  store i32 %inc, i32* @g, align 4
+  ret void
+}
+
+; CHECK: .section .pseudo_probe_desc,"G",@progbits,.pseudo_probe_desc_foo,comdat
+; CHECK-NEXT: .quad [[#GUID]]
+; CHECK-NEXT: .quad [[#HASH:]]
+; CHECK-NEXT: .byte  3
+; CHECK-NEXT: .ascii  "foo"
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!9, !10}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1)
+!1 = !DIFile(filename: "test.c", directory: "")
+!2 = !{}
+!4 = distinct !DISubprogram(name: "foo", scope: !1, file: !1, line: 2, type: !5, unit: !0, retainedNodes: !2)
+!5 = !DISubroutineType(types: !6)
+!6 = !{!7}
+!7 = !DIBasicType(name: "int", size: 32, align: 32, encoding: DW_ATE_signed)
+!9 = !{i32 2, !"Dwarf Version", i32 4}
+!10 = !{i32 2, !"Debug Info Version", i32 3}
+!13 = !DILocation(line: 2, column: 20, scope: !4)