llvm-project/llvm/tools/llvm-profgen/ProfiledBinary.h

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

393 lines
15 KiB
C
Raw Normal View History

//===-- ProfiledBinary.h - Binary decoder -----------------------*- C++ -*-===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
#ifndef LLVM_TOOLS_LLVM_PROFGEN_PROFILEDBINARY_H
#define LLVM_TOOLS_LLVM_PROFGEN_PROFILEDBINARY_H
[CSSPGO][llvm-profgen] Instruction symbolization This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change adds the support of instruction symbolization. Given the RVA on an instruction pointer, a full calling context can be printed side-by-side with the disassembly code. E.g. ``` Disassembly of section .text [0x0, 0x4a]: <funcA>: 0: mov eax, edi funcA:0 2: mov ecx, dword ptr [rip] funcLeaf:2 @ funcA:1 8: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 @ funcA:1 b: cmp ecx, 3 fib:2 @ funcLeaf:2 @ funcA:1 e: cmovl edx, ecx fib:2 @ funcLeaf:2 @ funcA:1 11: sub eax, edx funcLeaf:2 @ funcA:1 13: ret funcA:2 14: nop word ptr cs:[rax + rax] 1e: nop <funcLeaf>: 20: mov eax, edi funcLeaf:1 22: mov ecx, dword ptr [rip] funcLeaf:2 28: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 2b: cmp ecx, 3 fib:2 @ funcLeaf:2 2e: cmovl edx, ecx fib:2 @ funcLeaf:2 31: sub eax, edx funcLeaf:2 33: ret funcLeaf:3 34: nop word ptr cs:[rax + rax] 3e: nop <fib>: 40: lea eax, [rdi + 3] fib:2 43: cmp edi, 3 fib:2 46: cmovl eax, edi fib:2 49: ret fib:8 ``` Test Plan: ninja check-llvm Reviewed By: wenlei, wmi Differential Revision: https://reviews.llvm.org/D89715
2020-10-19 10:19:08 -07:00
#include "CallContext.h"
#include "llvm/ADT/Optional.h"
#include "llvm/ADT/StringRef.h"
[CSSPGO][llvm-profgen] Instruction symbolization This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change adds the support of instruction symbolization. Given the RVA on an instruction pointer, a full calling context can be printed side-by-side with the disassembly code. E.g. ``` Disassembly of section .text [0x0, 0x4a]: <funcA>: 0: mov eax, edi funcA:0 2: mov ecx, dword ptr [rip] funcLeaf:2 @ funcA:1 8: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 @ funcA:1 b: cmp ecx, 3 fib:2 @ funcLeaf:2 @ funcA:1 e: cmovl edx, ecx fib:2 @ funcLeaf:2 @ funcA:1 11: sub eax, edx funcLeaf:2 @ funcA:1 13: ret funcA:2 14: nop word ptr cs:[rax + rax] 1e: nop <funcLeaf>: 20: mov eax, edi funcLeaf:1 22: mov ecx, dword ptr [rip] funcLeaf:2 28: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 2b: cmp ecx, 3 fib:2 @ funcLeaf:2 2e: cmovl edx, ecx fib:2 @ funcLeaf:2 31: sub eax, edx funcLeaf:2 33: ret funcLeaf:3 34: nop word ptr cs:[rax + rax] 3e: nop <fib>: 40: lea eax, [rdi + 3] fib:2 43: cmp edi, 3 fib:2 46: cmovl eax, edi fib:2 49: ret fib:8 ``` Test Plan: ninja check-llvm Reviewed By: wenlei, wmi Differential Revision: https://reviews.llvm.org/D89715
2020-10-19 10:19:08 -07:00
#include "llvm/DebugInfo/Symbolize/Symbolize.h"
#include "llvm/MC/MCAsmInfo.h"
#include "llvm/MC/MCContext.h"
#include "llvm/MC/MCDisassembler/MCDisassembler.h"
#include "llvm/MC/MCInst.h"
#include "llvm/MC/MCInstPrinter.h"
#include "llvm/MC/MCInstrAnalysis.h"
#include "llvm/MC/MCInstrInfo.h"
#include "llvm/MC/MCObjectFileInfo.h"
#include "llvm/MC/MCPseudoProbe.h"
#include "llvm/MC/MCRegisterInfo.h"
#include "llvm/MC/MCSubtargetInfo.h"
#include "llvm/MC/MCTargetOptions.h"
#include "llvm/Object/ELFObjectFile.h"
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
#include "llvm/ProfileData/SampleProf.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Path.h"
#include "llvm/Transforms/IPO/SampleContextTracker.h"
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
#include <list>
#include <map>
#include <set>
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
#include <sstream>
#include <string>
#include <unordered_map>
#include <unordered_set>
#include <vector>
extern cl::opt<bool> EnableCSPreInliner;
extern cl::opt<bool> UseContextCostForPreInliner;
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
using namespace llvm;
using namespace sampleprof;
using namespace llvm::object;
namespace llvm {
namespace sampleprof {
[CSSPGO][llvm-profgen] Instruction symbolization This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change adds the support of instruction symbolization. Given the RVA on an instruction pointer, a full calling context can be printed side-by-side with the disassembly code. E.g. ``` Disassembly of section .text [0x0, 0x4a]: <funcA>: 0: mov eax, edi funcA:0 2: mov ecx, dword ptr [rip] funcLeaf:2 @ funcA:1 8: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 @ funcA:1 b: cmp ecx, 3 fib:2 @ funcLeaf:2 @ funcA:1 e: cmovl edx, ecx fib:2 @ funcLeaf:2 @ funcA:1 11: sub eax, edx funcLeaf:2 @ funcA:1 13: ret funcA:2 14: nop word ptr cs:[rax + rax] 1e: nop <funcLeaf>: 20: mov eax, edi funcLeaf:1 22: mov ecx, dword ptr [rip] funcLeaf:2 28: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 2b: cmp ecx, 3 fib:2 @ funcLeaf:2 2e: cmovl edx, ecx fib:2 @ funcLeaf:2 31: sub eax, edx funcLeaf:2 33: ret funcLeaf:3 34: nop word ptr cs:[rax + rax] 3e: nop <fib>: 40: lea eax, [rdi + 3] fib:2 43: cmp edi, 3 fib:2 46: cmovl eax, edi fib:2 49: ret fib:8 ``` Test Plan: ninja check-llvm Reviewed By: wenlei, wmi Differential Revision: https://reviews.llvm.org/D89715
2020-10-19 10:19:08 -07:00
class ProfiledBinary;
struct InstructionPointer {
[llvm-profgen] Support LBR only perf script This change aims at supporting LBR only sample perf script which is used for regular(Non-CS) profile generation. A LBR perf script includes a batch of LBR sample which starts with a frame pointer and a group of 32 LBR entries is followed. The FROM/TO LBR pair and the range between two consecutive entries (the former entry's TO and the latter entry's FROM) will be used to infer function profile info. An example of LBR perf script(created by `perf script -F ip,brstack -i perf.data`) ``` 40062f 0x40062f/0x4005b0/P/-/-/9 0x400645/0x4005ff/P/-/-/1 0x400637/0x400645/P/-/-/1 ... 4005d7 0x4005d7/0x4005e5/P/-/-/8 0x40062f/0x4005b0/P/-/-/6 0x400645/0x4005ff/P/-/-/1 ... ... ``` For implementation: - Extended a new child class `LBRPerfReader` for the sample parsing, reused all the functionalities in `extractLBRStack` except for an extension to parsing leading instruction pointer. - `HybridSample` is reused(just leave the call stack empty) and the parsed samples is still aggregated in `AggregatedSamples`. After that, range samples, branch sample, address samples are computed and recorded. - Reused `ContextSampleCounterMap` to store the raw profile, since it's no need to aggregation by context, here it just registered one sample counter with a fake context key. - Unified to use `show-raw-profile` instead of `show-unwinder-output` to dump the intermediate raw profile, see the comments of the format of the raw profile. For CS profile, it remains to output the unwinder output. Profile generation part will come soon. Differential Revision: https://reviews.llvm.org/D108153
2021-08-31 13:27:42 -07:00
const ProfiledBinary *Binary;
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
union {
// Offset of the executable segment of the binary.
uint64_t Offset = 0;
// Also used as address in unwinder
uint64_t Address;
};
[CSSPGO][llvm-profgen] Instruction symbolization This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change adds the support of instruction symbolization. Given the RVA on an instruction pointer, a full calling context can be printed side-by-side with the disassembly code. E.g. ``` Disassembly of section .text [0x0, 0x4a]: <funcA>: 0: mov eax, edi funcA:0 2: mov ecx, dword ptr [rip] funcLeaf:2 @ funcA:1 8: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 @ funcA:1 b: cmp ecx, 3 fib:2 @ funcLeaf:2 @ funcA:1 e: cmovl edx, ecx fib:2 @ funcLeaf:2 @ funcA:1 11: sub eax, edx funcLeaf:2 @ funcA:1 13: ret funcA:2 14: nop word ptr cs:[rax + rax] 1e: nop <funcLeaf>: 20: mov eax, edi funcLeaf:1 22: mov ecx, dword ptr [rip] funcLeaf:2 28: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 2b: cmp ecx, 3 fib:2 @ funcLeaf:2 2e: cmovl edx, ecx fib:2 @ funcLeaf:2 31: sub eax, edx funcLeaf:2 33: ret funcLeaf:3 34: nop word ptr cs:[rax + rax] 3e: nop <fib>: 40: lea eax, [rdi + 3] fib:2 43: cmp edi, 3 fib:2 46: cmovl eax, edi fib:2 49: ret fib:8 ``` Test Plan: ninja check-llvm Reviewed By: wenlei, wmi Differential Revision: https://reviews.llvm.org/D89715
2020-10-19 10:19:08 -07:00
// Index to the sorted code address array of the binary.
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
uint64_t Index = 0;
[llvm-profgen] Support LBR only perf script This change aims at supporting LBR only sample perf script which is used for regular(Non-CS) profile generation. A LBR perf script includes a batch of LBR sample which starts with a frame pointer and a group of 32 LBR entries is followed. The FROM/TO LBR pair and the range between two consecutive entries (the former entry's TO and the latter entry's FROM) will be used to infer function profile info. An example of LBR perf script(created by `perf script -F ip,brstack -i perf.data`) ``` 40062f 0x40062f/0x4005b0/P/-/-/9 0x400645/0x4005ff/P/-/-/1 0x400637/0x400645/P/-/-/1 ... 4005d7 0x4005d7/0x4005e5/P/-/-/8 0x40062f/0x4005b0/P/-/-/6 0x400645/0x4005ff/P/-/-/1 ... ... ``` For implementation: - Extended a new child class `LBRPerfReader` for the sample parsing, reused all the functionalities in `extractLBRStack` except for an extension to parsing leading instruction pointer. - `HybridSample` is reused(just leave the call stack empty) and the parsed samples is still aggregated in `AggregatedSamples`. After that, range samples, branch sample, address samples are computed and recorded. - Reused `ContextSampleCounterMap` to store the raw profile, since it's no need to aggregation by context, here it just registered one sample counter with a fake context key. - Unified to use `show-raw-profile` instead of `show-unwinder-output` to dump the intermediate raw profile, see the comments of the format of the raw profile. For CS profile, it remains to output the unwinder output. Profile generation part will come soon. Differential Revision: https://reviews.llvm.org/D108153
2021-08-31 13:27:42 -07:00
InstructionPointer(const ProfiledBinary *Binary, uint64_t Address,
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
bool RoundToNext = false);
void advance();
void backward();
void update(uint64_t Addr);
};
// PrologEpilog offset tracker, used to filter out broken stack samples
// Currently we use a heuristic size (two) to infer prolog and epilog
// based on the start address and return address. In the future,
// we will switch to Dwarf CFI based tracker
struct PrologEpilogTracker {
// A set of prolog and epilog offsets. Used by virtual unwinding.
std::unordered_set<uint64_t> PrologEpilogSet;
ProfiledBinary *Binary;
PrologEpilogTracker(ProfiledBinary *Bin) : Binary(Bin){};
// Take the two addresses from the start of function as prolog
void inferPrologOffsets(std::map<uint64_t, std::pair<std::string, uint64_t>>
&FuncStartOffsetMap) {
for (auto I : FuncStartOffsetMap) {
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
PrologEpilogSet.insert(I.first);
InstructionPointer IP(Binary, I.first);
IP.advance();
PrologEpilogSet.insert(IP.Offset);
}
}
[CSSPGO][llvm-profgen] Instruction symbolization This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change adds the support of instruction symbolization. Given the RVA on an instruction pointer, a full calling context can be printed side-by-side with the disassembly code. E.g. ``` Disassembly of section .text [0x0, 0x4a]: <funcA>: 0: mov eax, edi funcA:0 2: mov ecx, dword ptr [rip] funcLeaf:2 @ funcA:1 8: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 @ funcA:1 b: cmp ecx, 3 fib:2 @ funcLeaf:2 @ funcA:1 e: cmovl edx, ecx fib:2 @ funcLeaf:2 @ funcA:1 11: sub eax, edx funcLeaf:2 @ funcA:1 13: ret funcA:2 14: nop word ptr cs:[rax + rax] 1e: nop <funcLeaf>: 20: mov eax, edi funcLeaf:1 22: mov ecx, dword ptr [rip] funcLeaf:2 28: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 2b: cmp ecx, 3 fib:2 @ funcLeaf:2 2e: cmovl edx, ecx fib:2 @ funcLeaf:2 31: sub eax, edx funcLeaf:2 33: ret funcLeaf:3 34: nop word ptr cs:[rax + rax] 3e: nop <fib>: 40: lea eax, [rdi + 3] fib:2 43: cmp edi, 3 fib:2 46: cmovl eax, edi fib:2 49: ret fib:8 ``` Test Plan: ninja check-llvm Reviewed By: wenlei, wmi Differential Revision: https://reviews.llvm.org/D89715
2020-10-19 10:19:08 -07:00
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
// Take the last two addresses before the return address as epilog
void inferEpilogOffsets(std::unordered_set<uint64_t> &RetAddrs) {
for (auto Addr : RetAddrs) {
PrologEpilogSet.insert(Addr);
InstructionPointer IP(Binary, Addr);
IP.backward();
PrologEpilogSet.insert(IP.Offset);
}
[CSSPGO][llvm-profgen] Instruction symbolization This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change adds the support of instruction symbolization. Given the RVA on an instruction pointer, a full calling context can be printed side-by-side with the disassembly code. E.g. ``` Disassembly of section .text [0x0, 0x4a]: <funcA>: 0: mov eax, edi funcA:0 2: mov ecx, dword ptr [rip] funcLeaf:2 @ funcA:1 8: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 @ funcA:1 b: cmp ecx, 3 fib:2 @ funcLeaf:2 @ funcA:1 e: cmovl edx, ecx fib:2 @ funcLeaf:2 @ funcA:1 11: sub eax, edx funcLeaf:2 @ funcA:1 13: ret funcA:2 14: nop word ptr cs:[rax + rax] 1e: nop <funcLeaf>: 20: mov eax, edi funcLeaf:1 22: mov ecx, dword ptr [rip] funcLeaf:2 28: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 2b: cmp ecx, 3 fib:2 @ funcLeaf:2 2e: cmovl edx, ecx fib:2 @ funcLeaf:2 31: sub eax, edx funcLeaf:2 33: ret funcLeaf:3 34: nop word ptr cs:[rax + rax] 3e: nop <fib>: 40: lea eax, [rdi + 3] fib:2 43: cmp edi, 3 fib:2 46: cmovl eax, edi fib:2 49: ret fib:8 ``` Test Plan: ninja check-llvm Reviewed By: wenlei, wmi Differential Revision: https://reviews.llvm.org/D89715
2020-10-19 10:19:08 -07:00
}
};
// Track function byte size under different context (outlined version as well as
// various inlined versions). It also provides query support to get function
// size with the best matching context, which is used to help pre-inliner use
// accurate post-optimization size to make decisions.
// TODO: If an inlinee is completely optimized away, ideally we should have zero
// for its context size, currently we would misss such context since it doesn't
// have instructions. To fix this, we need to mark all inlinee with entry probe
// but without instructions as having zero size.
class BinarySizeContextTracker {
public:
// Add instruction with given size to a context
[CSSPGO] Split context string to deduplicate function name used in the context. Currently context strings contain a lot of duplicated function names and that significantly increase the profile size. This change split the context into a series of {name, offset, discriminator} tuples so function names used in the context can be replaced by the index into the name table and that significantly reduce the size consumed by context. A follow-up improvement made in the compiler and profiling tools is to avoid reconstructing full context strings which is time- and memory- consuming. Instead a context vector of `StringRef` is adopted to represent the full context in all scenarios. As a result, the previous prevalent profile map which was implemented as a `StringRef` is now engineered as an unordered map keyed by `SampleContext`. `SampleContext` is reshaped to using an `ArrayRef` to represent a full context for CS profile. For non-CS profile, it falls back to use `StringRef` to represent a contextless function name. Both the `ArrayRef` and `StringRef` objects are underpinned by real array and string objects that are stored in producer buffers. For compiler, they are maintained by the sample reader. For llvm-profgen, they are maintained in `ProfiledBinary` and `ProfileGenerator`. Full context strings can be generated only in those cases of debugging and printing. When it comes to profile format, nothing has changed to the text format, though internally CS context is implemented as a vector. Extbinary format is only changed for CS profile, with an additional `SecCSNameTable` section which stores all full contexts logically in the form of `vector<int>`, which each element as an offset points to `SecNameTable`. All occurrences of contexts elsewhere are redirected to using the offset of `SecCSNameTable`. Testing This is no-diff change in terms of code quality and profile content (for text profile). For our internal large service (aka ads), the profile generation is cut to half, with a 20x smaller string-based extbinary format generated. The compile time of ads is dropped by 25%. Differential Revision: https://reviews.llvm.org/D107299
2021-08-25 11:40:34 -07:00
void addInstructionForContext(const SampleContextFrameVector &Context,
uint32_t InstrSize);
// Get function size with a specific context. When there's no exact match
// for the given context, try to retrieve the size of that function from
// closest matching context.
uint32_t getFuncSizeForContext(const SampleContext &Context);
// For inlinees that are full optimized away, we can establish zero size using
// their remaining probes.
void trackInlineesOptimizedAway(MCPseudoProbeDecoder &ProbeDecoder);
void dump() { RootContext.dumpTree(); }
private:
using ProbeFrameStack = SmallVector<std::pair<StringRef, uint32_t>>;
void trackInlineesOptimizedAway(MCPseudoProbeDecoder &ProbeDecoder,
MCDecodedPseudoProbeInlineTree &ProbeNode,
ProbeFrameStack &Context);
// Root node for context trie tree, node that this is a reverse context trie
// with callee as parent and caller as child. This way we can traverse from
// root to find the best/longest matching context if an exact match does not
// exist. It gives us the best possible estimate for function's post-inline,
// post-optimization byte size.
ContextTrieNode RootContext;
};
using OffsetRange = std::pair<uint64_t, uint64_t>;
class ProfiledBinary {
// Absolute path of the binary.
std::string Path;
// The target triple.
Triple TheTriple;
// The runtime base address that the first executable segment is loaded at.
uint64_t BaseAddress = 0;
// The preferred load address of each executable segment.
std::vector<uint64_t> PreferredTextSegmentAddresses;
// The file offset of each executable segment.
std::vector<uint64_t> TextSegmentOffsets;
// Mutiple MC component info
std::unique_ptr<const MCRegisterInfo> MRI;
std::unique_ptr<const MCAsmInfo> AsmInfo;
std::unique_ptr<const MCSubtargetInfo> STI;
std::unique_ptr<const MCInstrInfo> MII;
std::unique_ptr<MCDisassembler> DisAsm;
std::unique_ptr<const MCInstrAnalysis> MIA;
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
std::unique_ptr<MCInstPrinter> IPrinter;
// A list of text sections sorted by start RVA and size. Used to check
// if a given RVA is a valid code address.
std::set<std::pair<uint64_t, uint64_t>> TextSections;
// An ordered map of mapping function's start offset to its name and
// end offset.
std::map<uint64_t, std::pair<std::string, uint64_t>> FuncStartOffsetMap;
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
// Offset to context location map. Used to expand the context.
[CSSPGO] Split context string to deduplicate function name used in the context. Currently context strings contain a lot of duplicated function names and that significantly increase the profile size. This change split the context into a series of {name, offset, discriminator} tuples so function names used in the context can be replaced by the index into the name table and that significantly reduce the size consumed by context. A follow-up improvement made in the compiler and profiling tools is to avoid reconstructing full context strings which is time- and memory- consuming. Instead a context vector of `StringRef` is adopted to represent the full context in all scenarios. As a result, the previous prevalent profile map which was implemented as a `StringRef` is now engineered as an unordered map keyed by `SampleContext`. `SampleContext` is reshaped to using an `ArrayRef` to represent a full context for CS profile. For non-CS profile, it falls back to use `StringRef` to represent a contextless function name. Both the `ArrayRef` and `StringRef` objects are underpinned by real array and string objects that are stored in producer buffers. For compiler, they are maintained by the sample reader. For llvm-profgen, they are maintained in `ProfiledBinary` and `ProfileGenerator`. Full context strings can be generated only in those cases of debugging and printing. When it comes to profile format, nothing has changed to the text format, though internally CS context is implemented as a vector. Extbinary format is only changed for CS profile, with an additional `SecCSNameTable` section which stores all full contexts logically in the form of `vector<int>`, which each element as an offset points to `SecNameTable`. All occurrences of contexts elsewhere are redirected to using the offset of `SecCSNameTable`. Testing This is no-diff change in terms of code quality and profile content (for text profile). For our internal large service (aka ads), the profile generation is cut to half, with a 20x smaller string-based extbinary format generated. The compile time of ads is dropped by 25%. Differential Revision: https://reviews.llvm.org/D107299
2021-08-25 11:40:34 -07:00
std::unordered_map<uint64_t, SampleContextFrameVector> Offset2LocStackMap;
// An array of offsets of all instructions sorted in increasing order. The
// sorting is needed to fast advance to the next forward/backward instruction.
std::vector<uint64_t> CodeAddrs;
// A set of call instruction offsets. Used by virtual unwinding.
std::unordered_set<uint64_t> CallAddrs;
// A set of return instruction offsets. Used by virtual unwinding.
std::unordered_set<uint64_t> RetAddrs;
// Estimate and track function prolog and epilog ranges.
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
PrologEpilogTracker ProEpilogTracker;
// Track function sizes under different context
BinarySizeContextTracker FuncSizeTracker;
[CSSPGO][llvm-profgen] Instruction symbolization This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change adds the support of instruction symbolization. Given the RVA on an instruction pointer, a full calling context can be printed side-by-side with the disassembly code. E.g. ``` Disassembly of section .text [0x0, 0x4a]: <funcA>: 0: mov eax, edi funcA:0 2: mov ecx, dword ptr [rip] funcLeaf:2 @ funcA:1 8: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 @ funcA:1 b: cmp ecx, 3 fib:2 @ funcLeaf:2 @ funcA:1 e: cmovl edx, ecx fib:2 @ funcLeaf:2 @ funcA:1 11: sub eax, edx funcLeaf:2 @ funcA:1 13: ret funcA:2 14: nop word ptr cs:[rax + rax] 1e: nop <funcLeaf>: 20: mov eax, edi funcLeaf:1 22: mov ecx, dword ptr [rip] funcLeaf:2 28: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 2b: cmp ecx, 3 fib:2 @ funcLeaf:2 2e: cmovl edx, ecx fib:2 @ funcLeaf:2 31: sub eax, edx funcLeaf:2 33: ret funcLeaf:3 34: nop word ptr cs:[rax + rax] 3e: nop <fib>: 40: lea eax, [rdi + 3] fib:2 43: cmp edi, 3 fib:2 46: cmovl eax, edi fib:2 49: ret fib:8 ``` Test Plan: ninja check-llvm Reviewed By: wenlei, wmi Differential Revision: https://reviews.llvm.org/D89715
2020-10-19 10:19:08 -07:00
// The symbolizer used to get inline context for an instruction.
std::unique_ptr<symbolize::LLVMSymbolizer> Symbolizer;
[CSSPGO][llvm-profgen] Pseudo probe decoding and disassembling This change implements pseudo probe decoding and disassembling for llvm-profgen/CSSPGO. Please see https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s and https://reviews.llvm.org/D89707 for more context about CSSPGO and llvm-profgen. **ELF section format** Please see the encoding patch(https://reviews.llvm.org/D91878) for more details of the format, just copy the example here: Two section(`.pseudo_probe_desc` and  `.pseudoprobe` ) is emitted in ELF to support pseudo probe. The format of `.pseudo_probe_desc` section looks like: ``` .section .pseudo_probe_desc,"",@progbits .quad 6309742469962978389 // Func GUID .quad 4294967295 // Func Hash .byte 9 // Length of func name .ascii "_Z5funcAi" // Func name .quad 7102633082150537521 .quad 138828622701 .byte 12 .ascii "_Z8funcLeafi" .quad 446061515086924981 .quad 4294967295 .byte 9 .ascii "_Z5funcBi" .quad -2016976694713209516 .quad 72617220756 .byte 7 .ascii "_Z3fibi" ``` For each `.pseudoprobe` section, the encoded binary data consists of a single function record corresponding to an outlined function (i.e, a function with a code entry in the `.text` section). A function record has the following format : ``` FUNCTION BODY (one for each outlined function present in the text section) GUID (uint64) GUID of the function NPROBES (ULEB128) Number of probes originating from this function. NUM_INLINED_FUNCTIONS (ULEB128) Number of callees inlined into this function, aka number of first-level inlinees PROBE RECORDS A list of NPROBES entries. Each entry contains: INDEX (ULEB128) TYPE (uint4) 0 - block probe, 1 - indirect call, 2 - direct call ATTRIBUTE (uint3) reserved ADDRESS_TYPE (uint1) 0 - code address, 1 - address delta CODE_ADDRESS (uint64 or ULEB128) code address or address delta, depending on ADDRESS_TYPE INLINED FUNCTION RECORDS A list of NUM_INLINED_FUNCTIONS entries describing each of the inlined callees. Each record contains: INLINE SITE GUID of the inlinee (uint64) ID of the callsite probe (ULEB128) FUNCTION BODY A FUNCTION BODY entry describing the inlined function. ``` **Disassembling** A switch `--show-pseudo-probe` is added to use along with `--show-disassembly` to print disassembly code with pseudo probe directives. For example: ``` 00000000002011a0 <foo2>: 2011a0: 50 push rax 2011a1: 85 ff test edi,edi [Probe]: FUNC: foo2 Index: 1 Type: Block 2011a3: 74 02 je 2011a7 <foo2+0x7> [Probe]: FUNC: foo2 Index: 3 Type: Block [Probe]: FUNC: foo2 Index: 4 Type: Block [Probe]: FUNC: foo Index: 1 Type: Block Inlined: @ foo2:6 2011a5: 58 pop rax 2011a6: c3 ret [Probe]: FUNC: foo2 Index: 2 Type: Block 2011a7: bf 01 00 00 00 mov edi,0x1 [Probe]: FUNC: foo2 Index: 5 Type: IndirectCall 2011ac: ff d6 call rsi [Probe]: FUNC: foo2 Index: 4 Type: Block 2011ae: 58 pop rax 2011af: c3 ret ``` **Implementation** - `PseudoProbeDecoder` is added in ProfiledBinary as an infra for the decoding. It decoded the two section and generate two map: `GUIDProbeFunctionMap` stores all the `PseudoProbeFunction` which is the abstraction of a general function. `AddressProbesMap` stores all the pseudo probe info indexed by its address. - All the inline info is encoded into binary as a trie(`PseudoProbeInlineTree`) and will be constructed from the decoding. Each pseudo probe can get its inline context(`getInlineContext`) by traversing its inline tree node backwards. Test Plan: ninja & ninja check-llvm Differential Revision: https://reviews.llvm.org/D92334
2020-11-23 20:33:23 -08:00
[CSSPGO] Split context string to deduplicate function name used in the context. Currently context strings contain a lot of duplicated function names and that significantly increase the profile size. This change split the context into a series of {name, offset, discriminator} tuples so function names used in the context can be replaced by the index into the name table and that significantly reduce the size consumed by context. A follow-up improvement made in the compiler and profiling tools is to avoid reconstructing full context strings which is time- and memory- consuming. Instead a context vector of `StringRef` is adopted to represent the full context in all scenarios. As a result, the previous prevalent profile map which was implemented as a `StringRef` is now engineered as an unordered map keyed by `SampleContext`. `SampleContext` is reshaped to using an `ArrayRef` to represent a full context for CS profile. For non-CS profile, it falls back to use `StringRef` to represent a contextless function name. Both the `ArrayRef` and `StringRef` objects are underpinned by real array and string objects that are stored in producer buffers. For compiler, they are maintained by the sample reader. For llvm-profgen, they are maintained in `ProfiledBinary` and `ProfileGenerator`. Full context strings can be generated only in those cases of debugging and printing. When it comes to profile format, nothing has changed to the text format, though internally CS context is implemented as a vector. Extbinary format is only changed for CS profile, with an additional `SecCSNameTable` section which stores all full contexts logically in the form of `vector<int>`, which each element as an offset points to `SecNameTable`. All occurrences of contexts elsewhere are redirected to using the offset of `SecCSNameTable`. Testing This is no-diff change in terms of code quality and profile content (for text profile). For our internal large service (aka ads), the profile generation is cut to half, with a 20x smaller string-based extbinary format generated. The compile time of ads is dropped by 25%. Differential Revision: https://reviews.llvm.org/D107299
2021-08-25 11:40:34 -07:00
// String table owning function name strings created from the symbolizer.
std::unordered_set<std::string> NameStrings;
// A collection of functions to print disassembly for.
StringSet<> DisassembleFunctionSet;
[CSSPGO][llvm-profgen] Pseudo probe decoding and disassembling This change implements pseudo probe decoding and disassembling for llvm-profgen/CSSPGO. Please see https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s and https://reviews.llvm.org/D89707 for more context about CSSPGO and llvm-profgen. **ELF section format** Please see the encoding patch(https://reviews.llvm.org/D91878) for more details of the format, just copy the example here: Two section(`.pseudo_probe_desc` and  `.pseudoprobe` ) is emitted in ELF to support pseudo probe. The format of `.pseudo_probe_desc` section looks like: ``` .section .pseudo_probe_desc,"",@progbits .quad 6309742469962978389 // Func GUID .quad 4294967295 // Func Hash .byte 9 // Length of func name .ascii "_Z5funcAi" // Func name .quad 7102633082150537521 .quad 138828622701 .byte 12 .ascii "_Z8funcLeafi" .quad 446061515086924981 .quad 4294967295 .byte 9 .ascii "_Z5funcBi" .quad -2016976694713209516 .quad 72617220756 .byte 7 .ascii "_Z3fibi" ``` For each `.pseudoprobe` section, the encoded binary data consists of a single function record corresponding to an outlined function (i.e, a function with a code entry in the `.text` section). A function record has the following format : ``` FUNCTION BODY (one for each outlined function present in the text section) GUID (uint64) GUID of the function NPROBES (ULEB128) Number of probes originating from this function. NUM_INLINED_FUNCTIONS (ULEB128) Number of callees inlined into this function, aka number of first-level inlinees PROBE RECORDS A list of NPROBES entries. Each entry contains: INDEX (ULEB128) TYPE (uint4) 0 - block probe, 1 - indirect call, 2 - direct call ATTRIBUTE (uint3) reserved ADDRESS_TYPE (uint1) 0 - code address, 1 - address delta CODE_ADDRESS (uint64 or ULEB128) code address or address delta, depending on ADDRESS_TYPE INLINED FUNCTION RECORDS A list of NUM_INLINED_FUNCTIONS entries describing each of the inlined callees. Each record contains: INLINE SITE GUID of the inlinee (uint64) ID of the callsite probe (ULEB128) FUNCTION BODY A FUNCTION BODY entry describing the inlined function. ``` **Disassembling** A switch `--show-pseudo-probe` is added to use along with `--show-disassembly` to print disassembly code with pseudo probe directives. For example: ``` 00000000002011a0 <foo2>: 2011a0: 50 push rax 2011a1: 85 ff test edi,edi [Probe]: FUNC: foo2 Index: 1 Type: Block 2011a3: 74 02 je 2011a7 <foo2+0x7> [Probe]: FUNC: foo2 Index: 3 Type: Block [Probe]: FUNC: foo2 Index: 4 Type: Block [Probe]: FUNC: foo Index: 1 Type: Block Inlined: @ foo2:6 2011a5: 58 pop rax 2011a6: c3 ret [Probe]: FUNC: foo2 Index: 2 Type: Block 2011a7: bf 01 00 00 00 mov edi,0x1 [Probe]: FUNC: foo2 Index: 5 Type: IndirectCall 2011ac: ff d6 call rsi [Probe]: FUNC: foo2 Index: 4 Type: Block 2011ae: 58 pop rax 2011af: c3 ret ``` **Implementation** - `PseudoProbeDecoder` is added in ProfiledBinary as an infra for the decoding. It decoded the two section and generate two map: `GUIDProbeFunctionMap` stores all the `PseudoProbeFunction` which is the abstraction of a general function. `AddressProbesMap` stores all the pseudo probe info indexed by its address. - All the inline info is encoded into binary as a trie(`PseudoProbeInlineTree`) and will be constructed from the decoding. Each pseudo probe can get its inline context(`getInlineContext`) by traversing its inline tree node backwards. Test Plan: ninja & ninja check-llvm Differential Revision: https://reviews.llvm.org/D92334
2020-11-23 20:33:23 -08:00
// Pseudo probe decoder
MCPseudoProbeDecoder ProbeDecoder;
[CSSPGO][llvm-profgen] Pseudo probe decoding and disassembling This change implements pseudo probe decoding and disassembling for llvm-profgen/CSSPGO. Please see https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s and https://reviews.llvm.org/D89707 for more context about CSSPGO and llvm-profgen. **ELF section format** Please see the encoding patch(https://reviews.llvm.org/D91878) for more details of the format, just copy the example here: Two section(`.pseudo_probe_desc` and  `.pseudoprobe` ) is emitted in ELF to support pseudo probe. The format of `.pseudo_probe_desc` section looks like: ``` .section .pseudo_probe_desc,"",@progbits .quad 6309742469962978389 // Func GUID .quad 4294967295 // Func Hash .byte 9 // Length of func name .ascii "_Z5funcAi" // Func name .quad 7102633082150537521 .quad 138828622701 .byte 12 .ascii "_Z8funcLeafi" .quad 446061515086924981 .quad 4294967295 .byte 9 .ascii "_Z5funcBi" .quad -2016976694713209516 .quad 72617220756 .byte 7 .ascii "_Z3fibi" ``` For each `.pseudoprobe` section, the encoded binary data consists of a single function record corresponding to an outlined function (i.e, a function with a code entry in the `.text` section). A function record has the following format : ``` FUNCTION BODY (one for each outlined function present in the text section) GUID (uint64) GUID of the function NPROBES (ULEB128) Number of probes originating from this function. NUM_INLINED_FUNCTIONS (ULEB128) Number of callees inlined into this function, aka number of first-level inlinees PROBE RECORDS A list of NPROBES entries. Each entry contains: INDEX (ULEB128) TYPE (uint4) 0 - block probe, 1 - indirect call, 2 - direct call ATTRIBUTE (uint3) reserved ADDRESS_TYPE (uint1) 0 - code address, 1 - address delta CODE_ADDRESS (uint64 or ULEB128) code address or address delta, depending on ADDRESS_TYPE INLINED FUNCTION RECORDS A list of NUM_INLINED_FUNCTIONS entries describing each of the inlined callees. Each record contains: INLINE SITE GUID of the inlinee (uint64) ID of the callsite probe (ULEB128) FUNCTION BODY A FUNCTION BODY entry describing the inlined function. ``` **Disassembling** A switch `--show-pseudo-probe` is added to use along with `--show-disassembly` to print disassembly code with pseudo probe directives. For example: ``` 00000000002011a0 <foo2>: 2011a0: 50 push rax 2011a1: 85 ff test edi,edi [Probe]: FUNC: foo2 Index: 1 Type: Block 2011a3: 74 02 je 2011a7 <foo2+0x7> [Probe]: FUNC: foo2 Index: 3 Type: Block [Probe]: FUNC: foo2 Index: 4 Type: Block [Probe]: FUNC: foo Index: 1 Type: Block Inlined: @ foo2:6 2011a5: 58 pop rax 2011a6: c3 ret [Probe]: FUNC: foo2 Index: 2 Type: Block 2011a7: bf 01 00 00 00 mov edi,0x1 [Probe]: FUNC: foo2 Index: 5 Type: IndirectCall 2011ac: ff d6 call rsi [Probe]: FUNC: foo2 Index: 4 Type: Block 2011ae: 58 pop rax 2011af: c3 ret ``` **Implementation** - `PseudoProbeDecoder` is added in ProfiledBinary as an infra for the decoding. It decoded the two section and generate two map: `GUIDProbeFunctionMap` stores all the `PseudoProbeFunction` which is the abstraction of a general function. `AddressProbesMap` stores all the pseudo probe info indexed by its address. - All the inline info is encoded into binary as a trie(`PseudoProbeInlineTree`) and will be constructed from the decoding. Each pseudo probe can get its inline context(`getInlineContext`) by traversing its inline tree node backwards. Test Plan: ninja & ninja check-llvm Differential Revision: https://reviews.llvm.org/D92334
2020-11-23 20:33:23 -08:00
bool UsePseudoProbes = false;
// Whether we need to symbolize all instructions to get function context size.
bool TrackFuncContextSize = false;
// Indicate if the base loading address is parsed from the mmap event or uses
// the preferred address
bool IsLoadedByMMap = false;
// Use to avoid redundant warning.
bool MissingMMapWarned = false;
void setPreferredTextSegmentAddresses(const ELFObjectFileBase *O);
template <class ELFT>
void setPreferredTextSegmentAddresses(const ELFFile<ELFT> &Obj, StringRef FileName);
[CSSPGO][llvm-profgen] Pseudo probe decoding and disassembling This change implements pseudo probe decoding and disassembling for llvm-profgen/CSSPGO. Please see https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s and https://reviews.llvm.org/D89707 for more context about CSSPGO and llvm-profgen. **ELF section format** Please see the encoding patch(https://reviews.llvm.org/D91878) for more details of the format, just copy the example here: Two section(`.pseudo_probe_desc` and  `.pseudoprobe` ) is emitted in ELF to support pseudo probe. The format of `.pseudo_probe_desc` section looks like: ``` .section .pseudo_probe_desc,"",@progbits .quad 6309742469962978389 // Func GUID .quad 4294967295 // Func Hash .byte 9 // Length of func name .ascii "_Z5funcAi" // Func name .quad 7102633082150537521 .quad 138828622701 .byte 12 .ascii "_Z8funcLeafi" .quad 446061515086924981 .quad 4294967295 .byte 9 .ascii "_Z5funcBi" .quad -2016976694713209516 .quad 72617220756 .byte 7 .ascii "_Z3fibi" ``` For each `.pseudoprobe` section, the encoded binary data consists of a single function record corresponding to an outlined function (i.e, a function with a code entry in the `.text` section). A function record has the following format : ``` FUNCTION BODY (one for each outlined function present in the text section) GUID (uint64) GUID of the function NPROBES (ULEB128) Number of probes originating from this function. NUM_INLINED_FUNCTIONS (ULEB128) Number of callees inlined into this function, aka number of first-level inlinees PROBE RECORDS A list of NPROBES entries. Each entry contains: INDEX (ULEB128) TYPE (uint4) 0 - block probe, 1 - indirect call, 2 - direct call ATTRIBUTE (uint3) reserved ADDRESS_TYPE (uint1) 0 - code address, 1 - address delta CODE_ADDRESS (uint64 or ULEB128) code address or address delta, depending on ADDRESS_TYPE INLINED FUNCTION RECORDS A list of NUM_INLINED_FUNCTIONS entries describing each of the inlined callees. Each record contains: INLINE SITE GUID of the inlinee (uint64) ID of the callsite probe (ULEB128) FUNCTION BODY A FUNCTION BODY entry describing the inlined function. ``` **Disassembling** A switch `--show-pseudo-probe` is added to use along with `--show-disassembly` to print disassembly code with pseudo probe directives. For example: ``` 00000000002011a0 <foo2>: 2011a0: 50 push rax 2011a1: 85 ff test edi,edi [Probe]: FUNC: foo2 Index: 1 Type: Block 2011a3: 74 02 je 2011a7 <foo2+0x7> [Probe]: FUNC: foo2 Index: 3 Type: Block [Probe]: FUNC: foo2 Index: 4 Type: Block [Probe]: FUNC: foo Index: 1 Type: Block Inlined: @ foo2:6 2011a5: 58 pop rax 2011a6: c3 ret [Probe]: FUNC: foo2 Index: 2 Type: Block 2011a7: bf 01 00 00 00 mov edi,0x1 [Probe]: FUNC: foo2 Index: 5 Type: IndirectCall 2011ac: ff d6 call rsi [Probe]: FUNC: foo2 Index: 4 Type: Block 2011ae: 58 pop rax 2011af: c3 ret ``` **Implementation** - `PseudoProbeDecoder` is added in ProfiledBinary as an infra for the decoding. It decoded the two section and generate two map: `GUIDProbeFunctionMap` stores all the `PseudoProbeFunction` which is the abstraction of a general function. `AddressProbesMap` stores all the pseudo probe info indexed by its address. - All the inline info is encoded into binary as a trie(`PseudoProbeInlineTree`) and will be constructed from the decoding. Each pseudo probe can get its inline context(`getInlineContext`) by traversing its inline tree node backwards. Test Plan: ninja & ninja check-llvm Differential Revision: https://reviews.llvm.org/D92334
2020-11-23 20:33:23 -08:00
void decodePseudoProbe(const ELFObjectFileBase *Obj);
// Set up disassembler and related components.
void setUpDisassembler(const ELFObjectFileBase *Obj);
[CSSPGO][llvm-profgen] Instruction symbolization This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change adds the support of instruction symbolization. Given the RVA on an instruction pointer, a full calling context can be printed side-by-side with the disassembly code. E.g. ``` Disassembly of section .text [0x0, 0x4a]: <funcA>: 0: mov eax, edi funcA:0 2: mov ecx, dword ptr [rip] funcLeaf:2 @ funcA:1 8: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 @ funcA:1 b: cmp ecx, 3 fib:2 @ funcLeaf:2 @ funcA:1 e: cmovl edx, ecx fib:2 @ funcLeaf:2 @ funcA:1 11: sub eax, edx funcLeaf:2 @ funcA:1 13: ret funcA:2 14: nop word ptr cs:[rax + rax] 1e: nop <funcLeaf>: 20: mov eax, edi funcLeaf:1 22: mov ecx, dword ptr [rip] funcLeaf:2 28: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 2b: cmp ecx, 3 fib:2 @ funcLeaf:2 2e: cmovl edx, ecx fib:2 @ funcLeaf:2 31: sub eax, edx funcLeaf:2 33: ret funcLeaf:3 34: nop word ptr cs:[rax + rax] 3e: nop <fib>: 40: lea eax, [rdi + 3] fib:2 43: cmp edi, 3 fib:2 46: cmovl eax, edi fib:2 49: ret fib:8 ``` Test Plan: ninja check-llvm Reviewed By: wenlei, wmi Differential Revision: https://reviews.llvm.org/D89715
2020-10-19 10:19:08 -07:00
void setupSymbolizer();
/// Dissassemble the text section and build various address maps.
void disassemble(const ELFObjectFileBase *O);
/// Helper function to dissassemble the symbol and extract info for unwinding
bool dissassembleSymbol(std::size_t SI, ArrayRef<uint8_t> Bytes,
SectionSymbolsTy &Symbols, const SectionRef &Section);
[CSSPGO][llvm-profgen] Instruction symbolization This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change adds the support of instruction symbolization. Given the RVA on an instruction pointer, a full calling context can be printed side-by-side with the disassembly code. E.g. ``` Disassembly of section .text [0x0, 0x4a]: <funcA>: 0: mov eax, edi funcA:0 2: mov ecx, dword ptr [rip] funcLeaf:2 @ funcA:1 8: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 @ funcA:1 b: cmp ecx, 3 fib:2 @ funcLeaf:2 @ funcA:1 e: cmovl edx, ecx fib:2 @ funcLeaf:2 @ funcA:1 11: sub eax, edx funcLeaf:2 @ funcA:1 13: ret funcA:2 14: nop word ptr cs:[rax + rax] 1e: nop <funcLeaf>: 20: mov eax, edi funcLeaf:1 22: mov ecx, dword ptr [rip] funcLeaf:2 28: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 2b: cmp ecx, 3 fib:2 @ funcLeaf:2 2e: cmovl edx, ecx fib:2 @ funcLeaf:2 31: sub eax, edx funcLeaf:2 33: ret funcLeaf:3 34: nop word ptr cs:[rax + rax] 3e: nop <fib>: 40: lea eax, [rdi + 3] fib:2 43: cmp edi, 3 fib:2 46: cmovl eax, edi fib:2 49: ret fib:8 ``` Test Plan: ninja check-llvm Reviewed By: wenlei, wmi Differential Revision: https://reviews.llvm.org/D89715
2020-10-19 10:19:08 -07:00
/// Symbolize a given instruction pointer and return a full call context.
[CSSPGO] Split context string to deduplicate function name used in the context. Currently context strings contain a lot of duplicated function names and that significantly increase the profile size. This change split the context into a series of {name, offset, discriminator} tuples so function names used in the context can be replaced by the index into the name table and that significantly reduce the size consumed by context. A follow-up improvement made in the compiler and profiling tools is to avoid reconstructing full context strings which is time- and memory- consuming. Instead a context vector of `StringRef` is adopted to represent the full context in all scenarios. As a result, the previous prevalent profile map which was implemented as a `StringRef` is now engineered as an unordered map keyed by `SampleContext`. `SampleContext` is reshaped to using an `ArrayRef` to represent a full context for CS profile. For non-CS profile, it falls back to use `StringRef` to represent a contextless function name. Both the `ArrayRef` and `StringRef` objects are underpinned by real array and string objects that are stored in producer buffers. For compiler, they are maintained by the sample reader. For llvm-profgen, they are maintained in `ProfiledBinary` and `ProfileGenerator`. Full context strings can be generated only in those cases of debugging and printing. When it comes to profile format, nothing has changed to the text format, though internally CS context is implemented as a vector. Extbinary format is only changed for CS profile, with an additional `SecCSNameTable` section which stores all full contexts logically in the form of `vector<int>`, which each element as an offset points to `SecNameTable`. All occurrences of contexts elsewhere are redirected to using the offset of `SecCSNameTable`. Testing This is no-diff change in terms of code quality and profile content (for text profile). For our internal large service (aka ads), the profile generation is cut to half, with a 20x smaller string-based extbinary format generated. The compile time of ads is dropped by 25%. Differential Revision: https://reviews.llvm.org/D107299
2021-08-25 11:40:34 -07:00
SampleContextFrameVector symbolize(const InstructionPointer &IP,
bool UseCanonicalFnName = false,
bool UseProbeDiscriminator = false);
/// Decode the interesting parts of the binary and build internal data
/// structures. On high level, the parts of interest are:
/// 1. Text sections, including the main code section and the PLT
/// entries that will be used to handle cross-module call transitions.
/// 2. The .debug_line section, used by Dwarf-based profile generation.
/// 3. Pseudo probe related sections, used by probe-based profile
/// generation.
void load();
public:
ProfiledBinary(const StringRef Path)
: Path(Path), ProEpilogTracker(this),
TrackFuncContextSize(EnableCSPreInliner &&
UseContextCostForPreInliner) {
[CSSPGO][llvm-profgen] Instruction symbolization This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change adds the support of instruction symbolization. Given the RVA on an instruction pointer, a full calling context can be printed side-by-side with the disassembly code. E.g. ``` Disassembly of section .text [0x0, 0x4a]: <funcA>: 0: mov eax, edi funcA:0 2: mov ecx, dword ptr [rip] funcLeaf:2 @ funcA:1 8: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 @ funcA:1 b: cmp ecx, 3 fib:2 @ funcLeaf:2 @ funcA:1 e: cmovl edx, ecx fib:2 @ funcLeaf:2 @ funcA:1 11: sub eax, edx funcLeaf:2 @ funcA:1 13: ret funcA:2 14: nop word ptr cs:[rax + rax] 1e: nop <funcLeaf>: 20: mov eax, edi funcLeaf:1 22: mov ecx, dword ptr [rip] funcLeaf:2 28: lea edx, [rcx + 3] fib:2 @ funcLeaf:2 2b: cmp ecx, 3 fib:2 @ funcLeaf:2 2e: cmovl edx, ecx fib:2 @ funcLeaf:2 31: sub eax, edx funcLeaf:2 33: ret funcLeaf:3 34: nop word ptr cs:[rax + rax] 3e: nop <fib>: 40: lea eax, [rdi + 3] fib:2 43: cmp edi, 3 fib:2 46: cmovl eax, edi fib:2 49: ret fib:8 ``` Test Plan: ninja check-llvm Reviewed By: wenlei, wmi Differential Revision: https://reviews.llvm.org/D89715
2020-10-19 10:19:08 -07:00
setupSymbolizer();
load();
}
uint64_t virtualAddrToOffset(uint64_t VirtualAddress) const {
return VirtualAddress - BaseAddress;
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
}
uint64_t offsetToVirtualAddr(uint64_t Offset) const {
return Offset + BaseAddress;
}
StringRef getPath() const { return Path; }
StringRef getName() const { return llvm::sys::path::filename(Path); }
uint64_t getBaseAddress() const { return BaseAddress; }
void setBaseAddress(uint64_t Address) { BaseAddress = Address; }
// Return the preferred load address for the first executable segment.
uint64_t getPreferredBaseAddress() const { return PreferredTextSegmentAddresses[0]; }
// Return the file offset for the first executable segment.
uint64_t getTextSegmentOffset() const { return TextSegmentOffsets[0]; }
const std::vector<uint64_t> &getPreferredTextSegmentAddresses() const {
return PreferredTextSegmentAddresses;
}
const std::vector<uint64_t> &getTextSegmentOffsets() const {
return TextSegmentOffsets;
}
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
bool addressIsCode(uint64_t Address) const {
uint64_t Offset = virtualAddrToOffset(Address);
return Offset2LocStackMap.find(Offset) != Offset2LocStackMap.end();
}
bool addressIsCall(uint64_t Address) const {
uint64_t Offset = virtualAddrToOffset(Address);
return CallAddrs.count(Offset);
}
bool addressIsReturn(uint64_t Address) const {
uint64_t Offset = virtualAddrToOffset(Address);
return RetAddrs.count(Offset);
}
bool addressInPrologEpilog(uint64_t Address) const {
uint64_t Offset = virtualAddrToOffset(Address);
return ProEpilogTracker.PrologEpilogSet.count(Offset);
}
uint64_t getAddressforIndex(uint64_t Index) const {
return offsetToVirtualAddr(CodeAddrs[Index]);
}
bool usePseudoProbes() const { return UsePseudoProbes; }
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
// Get the index in CodeAddrs for the address
// As we might get an address which is not the code
// here it would round to the next valid code address by
// using lower bound operation
uint32_t getIndexForAddr(uint64_t Address) const {
uint64_t Offset = virtualAddrToOffset(Address);
auto Low = llvm::lower_bound(CodeAddrs, Offset);
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
return Low - CodeAddrs.begin();
}
uint64_t getCallAddrFromFrameAddr(uint64_t FrameAddr) const {
auto I = getIndexForAddr(FrameAddr);
FrameAddr = I ? getAddressforIndex(I - 1) : 0;
if (FrameAddr && addressIsCall(FrameAddr))
return FrameAddr;
return 0;
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
}
StringRef getFuncFromStartOffset(uint64_t Offset) {
auto I = FuncStartOffsetMap.find(Offset);
if (I == FuncStartOffsetMap.end())
return StringRef();
return I->second.first;
}
OffsetRange findFuncOffsetRange(uint64_t Offset) {
auto I = FuncStartOffsetMap.upper_bound(Offset);
if (I == FuncStartOffsetMap.begin())
return {0, 0};
I--;
return {I->first, I->second.second};
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
}
uint32_t getFuncSizeForContext(SampleContext &Context) {
return FuncSizeTracker.getFuncSizeForContext(Context);
}
const SampleContextFrameVector &getFrameLocationStack(uint64_t Offset) const {
auto I = Offset2LocStackMap.find(Offset);
assert(I != Offset2LocStackMap.end() &&
"Can't find location for offset in the binary");
return I->second;
}
[CSSPGO] Split context string to deduplicate function name used in the context. Currently context strings contain a lot of duplicated function names and that significantly increase the profile size. This change split the context into a series of {name, offset, discriminator} tuples so function names used in the context can be replaced by the index into the name table and that significantly reduce the size consumed by context. A follow-up improvement made in the compiler and profiling tools is to avoid reconstructing full context strings which is time- and memory- consuming. Instead a context vector of `StringRef` is adopted to represent the full context in all scenarios. As a result, the previous prevalent profile map which was implemented as a `StringRef` is now engineered as an unordered map keyed by `SampleContext`. `SampleContext` is reshaped to using an `ArrayRef` to represent a full context for CS profile. For non-CS profile, it falls back to use `StringRef` to represent a contextless function name. Both the `ArrayRef` and `StringRef` objects are underpinned by real array and string objects that are stored in producer buffers. For compiler, they are maintained by the sample reader. For llvm-profgen, they are maintained in `ProfiledBinary` and `ProfileGenerator`. Full context strings can be generated only in those cases of debugging and printing. When it comes to profile format, nothing has changed to the text format, though internally CS context is implemented as a vector. Extbinary format is only changed for CS profile, with an additional `SecCSNameTable` section which stores all full contexts logically in the form of `vector<int>`, which each element as an offset points to `SecNameTable`. All occurrences of contexts elsewhere are redirected to using the offset of `SecCSNameTable`. Testing This is no-diff change in terms of code quality and profile content (for text profile). For our internal large service (aka ads), the profile generation is cut to half, with a 20x smaller string-based extbinary format generated. The compile time of ads is dropped by 25%. Differential Revision: https://reviews.llvm.org/D107299
2021-08-25 11:40:34 -07:00
Optional<SampleContextFrame> getInlineLeafFrameLoc(uint64_t Offset) {
const auto &Stack = getFrameLocationStack(Offset);
if (Stack.empty())
return {};
return Stack.back();
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
}
// Compare two addresses' inline context
bool inlineContextEqual(uint64_t Add1, uint64_t Add2) const;
[CSSPGO] Split context string to deduplicate function name used in the context. Currently context strings contain a lot of duplicated function names and that significantly increase the profile size. This change split the context into a series of {name, offset, discriminator} tuples so function names used in the context can be replaced by the index into the name table and that significantly reduce the size consumed by context. A follow-up improvement made in the compiler and profiling tools is to avoid reconstructing full context strings which is time- and memory- consuming. Instead a context vector of `StringRef` is adopted to represent the full context in all scenarios. As a result, the previous prevalent profile map which was implemented as a `StringRef` is now engineered as an unordered map keyed by `SampleContext`. `SampleContext` is reshaped to using an `ArrayRef` to represent a full context for CS profile. For non-CS profile, it falls back to use `StringRef` to represent a contextless function name. Both the `ArrayRef` and `StringRef` objects are underpinned by real array and string objects that are stored in producer buffers. For compiler, they are maintained by the sample reader. For llvm-profgen, they are maintained in `ProfiledBinary` and `ProfileGenerator`. Full context strings can be generated only in those cases of debugging and printing. When it comes to profile format, nothing has changed to the text format, though internally CS context is implemented as a vector. Extbinary format is only changed for CS profile, with an additional `SecCSNameTable` section which stores all full contexts logically in the form of `vector<int>`, which each element as an offset points to `SecNameTable`. All occurrences of contexts elsewhere are redirected to using the offset of `SecCSNameTable`. Testing This is no-diff change in terms of code quality and profile content (for text profile). For our internal large service (aka ads), the profile generation is cut to half, with a 20x smaller string-based extbinary format generated. The compile time of ads is dropped by 25%. Differential Revision: https://reviews.llvm.org/D107299
2021-08-25 11:40:34 -07:00
// Get the full context of the current stack with inline context filled in.
[CSSPGO][llvm-profgen] Context-sensitive profile data generation This stack of changes introduces `llvm-profgen` utility which generates a profile data file from given perf script data files for sample-based PGO. It’s part of(not only) the CSSPGO work. Specifically to support context-sensitive with/without pseudo probe profile, it implements a series of functionalities including perf trace parsing, instruction symbolization, LBR stack/call frame stack unwinding, pseudo probe decoding, etc. Also high throughput is achieved by multiple levels of sample aggregation and compatible format with one stop is generated at the end. Please refer to: https://groups.google.com/g/llvm-dev/c/1p1rdYbL93s for the CSSPGO RFC. This change supports context-sensitive profile data generation into llvm-profgen. With simultaneous sampling for LBR and call stack, we can identify leaf of LBR sample with calling context from stack sample . During the process of deriving fall through path from LBR entries, we unwind LBR by replaying all the calls and returns (including implicit calls/returns due to inlining) backwards on top of the sampled call stack. Then the state of call stack as we unwind through LBR always represents the calling context of current fall through path. we have two types of virtual unwinding 1) LBR unwinding and 2) linear range unwinding. Specifically, for each LBR entry which can be classified into call, return, regular branch, LBR unwinding will replay the operation by pushing, popping or switching leaf frame towards the call stack and since the initial call stack is most recently sampled, the replay should be in anti-execution order, i.e. for the regular case, pop the call stack when LBR is call, push frame on call stack when LBR is return. After each LBR processed, it also needs to align with the next LBR by going through instructions from previous LBR's target to current LBR's source, which we named linear unwinding. As instruction from linear range can come from different function by inlining, linear unwinding will do the range splitting and record counters through the range with same inline context. With each fall through path from LBR unwinding, we aggregate each sample into counters by the calling context and eventually generate full context sensitive profile (without relying on inlining) to driver compiler's PGO/FDO. A breakdown of noteworthy changes: - Added `HybridSample` class as the abstraction perf sample including LBR stack and call stack * Extended `PerfReader` to implement auto-detect whether input perf script output contains CS profile, then do the parsing. Multiple `HybridSample` are extracted * Speed up by aggregating `HybridSample` into `AggregatedSamples` * Added VirtualUnwinder that consumes aggregated `HybridSample` and implements unwinding of calls, returns, and linear path that contains implicit call/return from inlining. Ranges and branches counters are aggregated by the calling context.
 Here calling context is string type, each context is a pair of function name and callsite location info, the whole context is like `main:1 @ foo:2 @ bar`. * Added PorfileGenerater that accumulates counters by ranges unfolding or branch target mapping, then generates context-sensitive function profile including function body, inferring callee's head sample, callsite target samples, eventually records into ProfileMap.
 * Leveraged LLVM build-in(`SampleProfWriter`) writer to support different serialization format with no stop - `getCanonicalFnName` for callee name and name from ELF section - Added regression test for both unwinding and profile generation Test Plan: ninja & ninja check-llvm Reviewed By: hoy, wenlei, wmi Differential Revision: https://reviews.llvm.org/D89723
2020-10-19 12:55:59 -07:00
// It will search the disassembling info stored in Offset2LocStackMap. This is
// used as the key of function sample map
[CSSPGO] Split context string to deduplicate function name used in the context. Currently context strings contain a lot of duplicated function names and that significantly increase the profile size. This change split the context into a series of {name, offset, discriminator} tuples so function names used in the context can be replaced by the index into the name table and that significantly reduce the size consumed by context. A follow-up improvement made in the compiler and profiling tools is to avoid reconstructing full context strings which is time- and memory- consuming. Instead a context vector of `StringRef` is adopted to represent the full context in all scenarios. As a result, the previous prevalent profile map which was implemented as a `StringRef` is now engineered as an unordered map keyed by `SampleContext`. `SampleContext` is reshaped to using an `ArrayRef` to represent a full context for CS profile. For non-CS profile, it falls back to use `StringRef` to represent a contextless function name. Both the `ArrayRef` and `StringRef` objects are underpinned by real array and string objects that are stored in producer buffers. For compiler, they are maintained by the sample reader. For llvm-profgen, they are maintained in `ProfiledBinary` and `ProfileGenerator`. Full context strings can be generated only in those cases of debugging and printing. When it comes to profile format, nothing has changed to the text format, though internally CS context is implemented as a vector. Extbinary format is only changed for CS profile, with an additional `SecCSNameTable` section which stores all full contexts logically in the form of `vector<int>`, which each element as an offset points to `SecNameTable`. All occurrences of contexts elsewhere are redirected to using the offset of `SecCSNameTable`. Testing This is no-diff change in terms of code quality and profile content (for text profile). For our internal large service (aka ads), the profile generation is cut to half, with a 20x smaller string-based extbinary format generated. The compile time of ads is dropped by 25%. Differential Revision: https://reviews.llvm.org/D107299
2021-08-25 11:40:34 -07:00
SampleContextFrameVector
getExpandedContext(const SmallVectorImpl<uint64_t> &Stack,
bool &WasLeafInlined) const;
const MCDecodedPseudoProbe *getCallProbeForAddr(uint64_t Address) const {
return ProbeDecoder.getCallProbeForAddr(Address);
}
[CSSPGO] Split context string to deduplicate function name used in the context. Currently context strings contain a lot of duplicated function names and that significantly increase the profile size. This change split the context into a series of {name, offset, discriminator} tuples so function names used in the context can be replaced by the index into the name table and that significantly reduce the size consumed by context. A follow-up improvement made in the compiler and profiling tools is to avoid reconstructing full context strings which is time- and memory- consuming. Instead a context vector of `StringRef` is adopted to represent the full context in all scenarios. As a result, the previous prevalent profile map which was implemented as a `StringRef` is now engineered as an unordered map keyed by `SampleContext`. `SampleContext` is reshaped to using an `ArrayRef` to represent a full context for CS profile. For non-CS profile, it falls back to use `StringRef` to represent a contextless function name. Both the `ArrayRef` and `StringRef` objects are underpinned by real array and string objects that are stored in producer buffers. For compiler, they are maintained by the sample reader. For llvm-profgen, they are maintained in `ProfiledBinary` and `ProfileGenerator`. Full context strings can be generated only in those cases of debugging and printing. When it comes to profile format, nothing has changed to the text format, though internally CS context is implemented as a vector. Extbinary format is only changed for CS profile, with an additional `SecCSNameTable` section which stores all full contexts logically in the form of `vector<int>`, which each element as an offset points to `SecNameTable`. All occurrences of contexts elsewhere are redirected to using the offset of `SecCSNameTable`. Testing This is no-diff change in terms of code quality and profile content (for text profile). For our internal large service (aka ads), the profile generation is cut to half, with a 20x smaller string-based extbinary format generated. The compile time of ads is dropped by 25%. Differential Revision: https://reviews.llvm.org/D107299
2021-08-25 11:40:34 -07:00
void getInlineContextForProbe(const MCDecodedPseudoProbe *Probe,
SampleContextFrameVector &InlineContextStack,
bool IncludeLeaf = false) const {
SmallVector<MCPseduoProbeFrameLocation, 16> ProbeInlineContext;
ProbeDecoder.getInlineContextForProbe(Probe, ProbeInlineContext,
IncludeLeaf);
for (auto &Callsite : ProbeInlineContext) {
InlineContextStack.emplace_back(Callsite.first,
LineLocation(Callsite.second, 0));
}
}
2021-01-11 09:08:39 -08:00
const AddressProbesMap &getAddress2ProbesMap() const {
return ProbeDecoder.getAddress2ProbesMap();
}
const MCPseudoProbeFuncDesc *getFuncDescForGUID(uint64_t GUID) {
2021-01-11 09:08:39 -08:00
return ProbeDecoder.getFuncDescForGUID(GUID);
}
const MCPseudoProbeFuncDesc *
getInlinerDescForProbe(const MCDecodedPseudoProbe *Probe) {
2021-01-11 09:08:39 -08:00
return ProbeDecoder.getInlinerDescForProbe(Probe);
}
bool getIsLoadedByMMap() { return IsLoadedByMMap; }
void setIsLoadedByMMap(bool Value) { IsLoadedByMMap = Value; }
bool getMissingMMapWarned() { return MissingMMapWarned; }
void setMissingMMapWarned(bool Value) { MissingMMapWarned = Value; }
};
} // end namespace sampleprof
} // end namespace llvm
#endif