llvm-project/llvm/lib/CodeGen/PseudoProbeInserter.cpp

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

147 lines
5.6 KiB
C++
Raw Normal View History

[CSSPGO] Pseudo probes for function calls. An indirect call site needs to be probed for its potential call targets. With CSSPGO a direct call also needs a probe so that a calling context can be represented by a stack of callsite probes. Unlike pseudo probes for basic blocks that are in form of standalone intrinsic call instructions, pseudo probes for callsites have to be attached to the call instruction, thus a separate instruction would not work. One possible way of attaching a probe to a call instruction is to use a special metadata that carries information about the probe. The special metadata will have to make its way through the optimization pipeline down to object emission. This requires additional efforts to maintain the metadata in various places. Given that the `!dbg` metadata is a first-class metadata and has all essential support in place , leveraging the `!dbg` metadata as a channel to encode pseudo probe information is probably the easiest solution. With the requirement of not inflating `!dbg` metadata that is allocated for almost every instruction, we found that the 32-bit DWARF discriminator field which mainly serves AutoFDO can be reused for pseudo probes. DWARF discriminators distinguish identical source locations between instructions and with pseudo probes such support is not required. In this change we are using the discriminator field to encode the ID and type of a callsite probe and the encoded value will be unpacked and consumed right before object emission. When a callsite is inlined, the callsite discriminator field will go with the inlined instructions. The `!dbg` metadata of an inlined instruction is in form of a scope stack. The top of the stack is the instruction's original `!dbg` metadata and the bottom of the stack is for the original callsite of the top-level inliner. Except for the top of the stack, all other elements of the stack actually refer to the nested inlined callsites whose discriminator field (which actually represents a calliste probe) can be used together to represent the inline context of an inlined PseudoProbeInst or CallInst. To avoid collision with the baseline AutoFDO in various places that handles dwarf discriminators where a check against the `-pseudo-probe-for-profiling` switch is not available, a special encoding scheme is used to tell apart a pseudo probe discriminator from a regular discriminator. For the regular discriminator, if all lowest 3 bits are non-zero, it means the discriminator is basically empty and all higher 29 bits can be reversed for pseudo probe use. Callsite pseudo probes are inserted in `SampleProfileProbePass` and a target-independent MIR pass `PseudoProbeInserter` is added to unpack the probe ID/type from `!dbg`. Note that with this work the switch -debug-info-for-profiling will not work with -pseudo-probe-for-profiling anymore. They cannot be used at the same time. Reviewed By: wmi Differential Revision: https://reviews.llvm.org/D91756
2020-12-02 13:44:06 +08:00
//===- PseudoProbeInserter.cpp - Insert annotation for callsite profiling -===//
//
// 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
//
//===----------------------------------------------------------------------===//
//
// This file implements PseudoProbeInserter pass, which inserts pseudo probe
// annotations for call instructions with a pseudo-probe-specific dwarf
// discriminator. such discriminator indicates that the call instruction comes
// with a pseudo probe, and the discriminator value holds information to
// identify the corresponding counter.
//===----------------------------------------------------------------------===//
#include "llvm/CodeGen/MachineBasicBlock.h"
#include "llvm/CodeGen/MachineFunctionPass.h"
#include "llvm/CodeGen/MachineInstr.h"
#include "llvm/CodeGen/TargetInstrInfo.h"
#include "llvm/IR/DebugInfoMetadata.h"
#include "llvm/IR/PseudoProbe.h"
#include "llvm/InitializePasses.h"
[CSSPGO] Introducing dangling pseudo probes. Dangling probes are the probes associated to an empty block. This usually happens when all real instructions are optimized away from the block. There is a problem with dangling probes during the offline counts processing. The way the sample profiler works is that samples collected on the first physical instruction following a probe will be counted towards the probe. This logically equals to treating the instruction next to a probe as if it is from the same block of the probe. In the dangling probe case, the real instruction following a dangling probe actually starts a new block, and samples collected on the new block may cause issues when counted towards the empty block. To mitigate this issue, we first try to move around a dangling probe inside its owning block. If there are still native instructions preceding the probe in the same block, we can then use them as a place holder to collect samples for the probe. A pass is added to walk each block backwards looking for probes not followed by any real instruction and moving them before the first real instruction. This is done right before the object emission. If we are unlucky to find such in-block preceding instructions for a probe, the solution we are taking is to tag such probe as dangling so that the samples reported for them will not be trusted by the compiler. We leave it up to the counts inference algorithm to get such probes a reasonable count. The number `UINT64_MAX` is used to mark sample count as collected for a dangling probe. Reviewed By: wmi Differential Revision: https://reviews.llvm.org/D95962
2021-02-25 16:43:17 +08:00
#include "llvm/MC/MCPseudoProbe.h"
[CSSPGO] Pseudo probes for function calls. An indirect call site needs to be probed for its potential call targets. With CSSPGO a direct call also needs a probe so that a calling context can be represented by a stack of callsite probes. Unlike pseudo probes for basic blocks that are in form of standalone intrinsic call instructions, pseudo probes for callsites have to be attached to the call instruction, thus a separate instruction would not work. One possible way of attaching a probe to a call instruction is to use a special metadata that carries information about the probe. The special metadata will have to make its way through the optimization pipeline down to object emission. This requires additional efforts to maintain the metadata in various places. Given that the `!dbg` metadata is a first-class metadata and has all essential support in place , leveraging the `!dbg` metadata as a channel to encode pseudo probe information is probably the easiest solution. With the requirement of not inflating `!dbg` metadata that is allocated for almost every instruction, we found that the 32-bit DWARF discriminator field which mainly serves AutoFDO can be reused for pseudo probes. DWARF discriminators distinguish identical source locations between instructions and with pseudo probes such support is not required. In this change we are using the discriminator field to encode the ID and type of a callsite probe and the encoded value will be unpacked and consumed right before object emission. When a callsite is inlined, the callsite discriminator field will go with the inlined instructions. The `!dbg` metadata of an inlined instruction is in form of a scope stack. The top of the stack is the instruction's original `!dbg` metadata and the bottom of the stack is for the original callsite of the top-level inliner. Except for the top of the stack, all other elements of the stack actually refer to the nested inlined callsites whose discriminator field (which actually represents a calliste probe) can be used together to represent the inline context of an inlined PseudoProbeInst or CallInst. To avoid collision with the baseline AutoFDO in various places that handles dwarf discriminators where a check against the `-pseudo-probe-for-profiling` switch is not available, a special encoding scheme is used to tell apart a pseudo probe discriminator from a regular discriminator. For the regular discriminator, if all lowest 3 bits are non-zero, it means the discriminator is basically empty and all higher 29 bits can be reversed for pseudo probe use. Callsite pseudo probes are inserted in `SampleProfileProbePass` and a target-independent MIR pass `PseudoProbeInserter` is added to unpack the probe ID/type from `!dbg`. Note that with this work the switch -debug-info-for-profiling will not work with -pseudo-probe-for-profiling anymore. They cannot be used at the same time. Reviewed By: wmi Differential Revision: https://reviews.llvm.org/D91756
2020-12-02 13:44:06 +08:00
#include "llvm/Target/TargetMachine.h"
#include <unordered_set>
[CSSPGO] Pseudo probes for function calls. An indirect call site needs to be probed for its potential call targets. With CSSPGO a direct call also needs a probe so that a calling context can be represented by a stack of callsite probes. Unlike pseudo probes for basic blocks that are in form of standalone intrinsic call instructions, pseudo probes for callsites have to be attached to the call instruction, thus a separate instruction would not work. One possible way of attaching a probe to a call instruction is to use a special metadata that carries information about the probe. The special metadata will have to make its way through the optimization pipeline down to object emission. This requires additional efforts to maintain the metadata in various places. Given that the `!dbg` metadata is a first-class metadata and has all essential support in place , leveraging the `!dbg` metadata as a channel to encode pseudo probe information is probably the easiest solution. With the requirement of not inflating `!dbg` metadata that is allocated for almost every instruction, we found that the 32-bit DWARF discriminator field which mainly serves AutoFDO can be reused for pseudo probes. DWARF discriminators distinguish identical source locations between instructions and with pseudo probes such support is not required. In this change we are using the discriminator field to encode the ID and type of a callsite probe and the encoded value will be unpacked and consumed right before object emission. When a callsite is inlined, the callsite discriminator field will go with the inlined instructions. The `!dbg` metadata of an inlined instruction is in form of a scope stack. The top of the stack is the instruction's original `!dbg` metadata and the bottom of the stack is for the original callsite of the top-level inliner. Except for the top of the stack, all other elements of the stack actually refer to the nested inlined callsites whose discriminator field (which actually represents a calliste probe) can be used together to represent the inline context of an inlined PseudoProbeInst or CallInst. To avoid collision with the baseline AutoFDO in various places that handles dwarf discriminators where a check against the `-pseudo-probe-for-profiling` switch is not available, a special encoding scheme is used to tell apart a pseudo probe discriminator from a regular discriminator. For the regular discriminator, if all lowest 3 bits are non-zero, it means the discriminator is basically empty and all higher 29 bits can be reversed for pseudo probe use. Callsite pseudo probes are inserted in `SampleProfileProbePass` and a target-independent MIR pass `PseudoProbeInserter` is added to unpack the probe ID/type from `!dbg`. Note that with this work the switch -debug-info-for-profiling will not work with -pseudo-probe-for-profiling anymore. They cannot be used at the same time. Reviewed By: wmi Differential Revision: https://reviews.llvm.org/D91756
2020-12-02 13:44:06 +08:00
#define DEBUG_TYPE "pseudo-probe-inserter"
using namespace llvm;
namespace {
class PseudoProbeInserter : public MachineFunctionPass {
public:
static char ID;
PseudoProbeInserter() : MachineFunctionPass(ID) {
initializePseudoProbeInserterPass(*PassRegistry::getPassRegistry());
}
StringRef getPassName() const override { return "Pseudo Probe Inserter"; }
void getAnalysisUsage(AnalysisUsage &AU) const override {
AU.setPreservesAll();
MachineFunctionPass::getAnalysisUsage(AU);
}
bool runOnMachineFunction(MachineFunction &MF) override {
const TargetInstrInfo *TII = MF.getSubtarget().getInstrInfo();
bool Changed = false;
for (MachineBasicBlock &MBB : MF) {
[CSSPGO] Introducing dangling pseudo probes. Dangling probes are the probes associated to an empty block. This usually happens when all real instructions are optimized away from the block. There is a problem with dangling probes during the offline counts processing. The way the sample profiler works is that samples collected on the first physical instruction following a probe will be counted towards the probe. This logically equals to treating the instruction next to a probe as if it is from the same block of the probe. In the dangling probe case, the real instruction following a dangling probe actually starts a new block, and samples collected on the new block may cause issues when counted towards the empty block. To mitigate this issue, we first try to move around a dangling probe inside its owning block. If there are still native instructions preceding the probe in the same block, we can then use them as a place holder to collect samples for the probe. A pass is added to walk each block backwards looking for probes not followed by any real instruction and moving them before the first real instruction. This is done right before the object emission. If we are unlucky to find such in-block preceding instructions for a probe, the solution we are taking is to tag such probe as dangling so that the samples reported for them will not be trusted by the compiler. We leave it up to the counts inference algorithm to get such probes a reasonable count. The number `UINT64_MAX` is used to mark sample count as collected for a dangling probe. Reviewed By: wmi Differential Revision: https://reviews.llvm.org/D95962
2021-02-25 16:43:17 +08:00
MachineInstr *FirstInstr = nullptr;
[CSSPGO] Pseudo probes for function calls. An indirect call site needs to be probed for its potential call targets. With CSSPGO a direct call also needs a probe so that a calling context can be represented by a stack of callsite probes. Unlike pseudo probes for basic blocks that are in form of standalone intrinsic call instructions, pseudo probes for callsites have to be attached to the call instruction, thus a separate instruction would not work. One possible way of attaching a probe to a call instruction is to use a special metadata that carries information about the probe. The special metadata will have to make its way through the optimization pipeline down to object emission. This requires additional efforts to maintain the metadata in various places. Given that the `!dbg` metadata is a first-class metadata and has all essential support in place , leveraging the `!dbg` metadata as a channel to encode pseudo probe information is probably the easiest solution. With the requirement of not inflating `!dbg` metadata that is allocated for almost every instruction, we found that the 32-bit DWARF discriminator field which mainly serves AutoFDO can be reused for pseudo probes. DWARF discriminators distinguish identical source locations between instructions and with pseudo probes such support is not required. In this change we are using the discriminator field to encode the ID and type of a callsite probe and the encoded value will be unpacked and consumed right before object emission. When a callsite is inlined, the callsite discriminator field will go with the inlined instructions. The `!dbg` metadata of an inlined instruction is in form of a scope stack. The top of the stack is the instruction's original `!dbg` metadata and the bottom of the stack is for the original callsite of the top-level inliner. Except for the top of the stack, all other elements of the stack actually refer to the nested inlined callsites whose discriminator field (which actually represents a calliste probe) can be used together to represent the inline context of an inlined PseudoProbeInst or CallInst. To avoid collision with the baseline AutoFDO in various places that handles dwarf discriminators where a check against the `-pseudo-probe-for-profiling` switch is not available, a special encoding scheme is used to tell apart a pseudo probe discriminator from a regular discriminator. For the regular discriminator, if all lowest 3 bits are non-zero, it means the discriminator is basically empty and all higher 29 bits can be reversed for pseudo probe use. Callsite pseudo probes are inserted in `SampleProfileProbePass` and a target-independent MIR pass `PseudoProbeInserter` is added to unpack the probe ID/type from `!dbg`. Note that with this work the switch -debug-info-for-profiling will not work with -pseudo-probe-for-profiling anymore. They cannot be used at the same time. Reviewed By: wmi Differential Revision: https://reviews.llvm.org/D91756
2020-12-02 13:44:06 +08:00
for (MachineInstr &MI : MBB) {
[CSSPGO] Introducing dangling pseudo probes. Dangling probes are the probes associated to an empty block. This usually happens when all real instructions are optimized away from the block. There is a problem with dangling probes during the offline counts processing. The way the sample profiler works is that samples collected on the first physical instruction following a probe will be counted towards the probe. This logically equals to treating the instruction next to a probe as if it is from the same block of the probe. In the dangling probe case, the real instruction following a dangling probe actually starts a new block, and samples collected on the new block may cause issues when counted towards the empty block. To mitigate this issue, we first try to move around a dangling probe inside its owning block. If there are still native instructions preceding the probe in the same block, we can then use them as a place holder to collect samples for the probe. A pass is added to walk each block backwards looking for probes not followed by any real instruction and moving them before the first real instruction. This is done right before the object emission. If we are unlucky to find such in-block preceding instructions for a probe, the solution we are taking is to tag such probe as dangling so that the samples reported for them will not be trusted by the compiler. We leave it up to the counts inference algorithm to get such probes a reasonable count. The number `UINT64_MAX` is used to mark sample count as collected for a dangling probe. Reviewed By: wmi Differential Revision: https://reviews.llvm.org/D95962
2021-02-25 16:43:17 +08:00
if (!MI.isPseudo())
FirstInstr = &MI;
[CSSPGO] Pseudo probes for function calls. An indirect call site needs to be probed for its potential call targets. With CSSPGO a direct call also needs a probe so that a calling context can be represented by a stack of callsite probes. Unlike pseudo probes for basic blocks that are in form of standalone intrinsic call instructions, pseudo probes for callsites have to be attached to the call instruction, thus a separate instruction would not work. One possible way of attaching a probe to a call instruction is to use a special metadata that carries information about the probe. The special metadata will have to make its way through the optimization pipeline down to object emission. This requires additional efforts to maintain the metadata in various places. Given that the `!dbg` metadata is a first-class metadata and has all essential support in place , leveraging the `!dbg` metadata as a channel to encode pseudo probe information is probably the easiest solution. With the requirement of not inflating `!dbg` metadata that is allocated for almost every instruction, we found that the 32-bit DWARF discriminator field which mainly serves AutoFDO can be reused for pseudo probes. DWARF discriminators distinguish identical source locations between instructions and with pseudo probes such support is not required. In this change we are using the discriminator field to encode the ID and type of a callsite probe and the encoded value will be unpacked and consumed right before object emission. When a callsite is inlined, the callsite discriminator field will go with the inlined instructions. The `!dbg` metadata of an inlined instruction is in form of a scope stack. The top of the stack is the instruction's original `!dbg` metadata and the bottom of the stack is for the original callsite of the top-level inliner. Except for the top of the stack, all other elements of the stack actually refer to the nested inlined callsites whose discriminator field (which actually represents a calliste probe) can be used together to represent the inline context of an inlined PseudoProbeInst or CallInst. To avoid collision with the baseline AutoFDO in various places that handles dwarf discriminators where a check against the `-pseudo-probe-for-profiling` switch is not available, a special encoding scheme is used to tell apart a pseudo probe discriminator from a regular discriminator. For the regular discriminator, if all lowest 3 bits are non-zero, it means the discriminator is basically empty and all higher 29 bits can be reversed for pseudo probe use. Callsite pseudo probes are inserted in `SampleProfileProbePass` and a target-independent MIR pass `PseudoProbeInserter` is added to unpack the probe ID/type from `!dbg`. Note that with this work the switch -debug-info-for-profiling will not work with -pseudo-probe-for-profiling anymore. They cannot be used at the same time. Reviewed By: wmi Differential Revision: https://reviews.llvm.org/D91756
2020-12-02 13:44:06 +08:00
if (MI.isCall()) {
if (DILocation *DL = MI.getDebugLoc()) {
auto Value = DL->getDiscriminator();
if (DILocation::isPseudoProbeDiscriminator(Value)) {
BuildMI(MBB, MI, DL, TII->get(TargetOpcode::PSEUDO_PROBE))
.addImm(getFuncGUID(MF.getFunction().getParent(), DL))
.addImm(
PseudoProbeDwarfDiscriminator::extractProbeIndex(Value))
.addImm(
PseudoProbeDwarfDiscriminator::extractProbeType(Value))
.addImm(PseudoProbeDwarfDiscriminator::extractProbeAttributes(
Value));
Changed = true;
}
}
}
}
[CSSPGO] Introducing dangling pseudo probes. Dangling probes are the probes associated to an empty block. This usually happens when all real instructions are optimized away from the block. There is a problem with dangling probes during the offline counts processing. The way the sample profiler works is that samples collected on the first physical instruction following a probe will be counted towards the probe. This logically equals to treating the instruction next to a probe as if it is from the same block of the probe. In the dangling probe case, the real instruction following a dangling probe actually starts a new block, and samples collected on the new block may cause issues when counted towards the empty block. To mitigate this issue, we first try to move around a dangling probe inside its owning block. If there are still native instructions preceding the probe in the same block, we can then use them as a place holder to collect samples for the probe. A pass is added to walk each block backwards looking for probes not followed by any real instruction and moving them before the first real instruction. This is done right before the object emission. If we are unlucky to find such in-block preceding instructions for a probe, the solution we are taking is to tag such probe as dangling so that the samples reported for them will not be trusted by the compiler. We leave it up to the counts inference algorithm to get such probes a reasonable count. The number `UINT64_MAX` is used to mark sample count as collected for a dangling probe. Reviewed By: wmi Differential Revision: https://reviews.llvm.org/D95962
2021-02-25 16:43:17 +08:00
// Walk the block backwards, move PSEUDO_PROBE before the first real
// instruction to fix out-of-order probes. There is a problem with probes
// as the terminator of the block. During the offline counts processing,
// the samples collected on the first physical instruction following a
// probe will be counted towards the probe. This logically equals to
// treating the instruction next to a probe as if it is from the same
// block of the probe. This is accurate most of the time unless the
// instruction can be reached from multiple flows, which means it actually
// starts a new block. Samples collected on such probes may cause
// imprecision with the counts inference algorithm. Fortunately, if
// there are still other native instructions preceding the probe we can
// use them as a place holder to collect samples for the probe.
if (FirstInstr) {
auto MII = MBB.rbegin();
while (MII != MBB.rend()) {
// Skip all pseudo probes followed by a real instruction since they
// are not dangling.
if (!MII->isPseudo())
break;
auto Cur = MII++;
if (Cur->getOpcode() != TargetOpcode::PSEUDO_PROBE)
continue;
// Move the dangling probe before FirstInstr.
auto *ProbeInstr = &*Cur;
MBB.remove(ProbeInstr);
MBB.insert(FirstInstr, ProbeInstr);
Changed = true;
}
} else {
// Probes not surrounded by any real instructions in the same block are
// called dangling probes. Since there's no good way to pick up a sample
// collection point for dangling probes at compile time, they are being
// removed so that the profile correlation tool will not report any
[CSSPGO] Introducing dangling pseudo probes. Dangling probes are the probes associated to an empty block. This usually happens when all real instructions are optimized away from the block. There is a problem with dangling probes during the offline counts processing. The way the sample profiler works is that samples collected on the first physical instruction following a probe will be counted towards the probe. This logically equals to treating the instruction next to a probe as if it is from the same block of the probe. In the dangling probe case, the real instruction following a dangling probe actually starts a new block, and samples collected on the new block may cause issues when counted towards the empty block. To mitigate this issue, we first try to move around a dangling probe inside its owning block. If there are still native instructions preceding the probe in the same block, we can then use them as a place holder to collect samples for the probe. A pass is added to walk each block backwards looking for probes not followed by any real instruction and moving them before the first real instruction. This is done right before the object emission. If we are unlucky to find such in-block preceding instructions for a probe, the solution we are taking is to tag such probe as dangling so that the samples reported for them will not be trusted by the compiler. We leave it up to the counts inference algorithm to get such probes a reasonable count. The number `UINT64_MAX` is used to mark sample count as collected for a dangling probe. Reviewed By: wmi Differential Revision: https://reviews.llvm.org/D95962
2021-02-25 16:43:17 +08:00
// samples collected for them and it's up to the counts inference tool
// to get them a reasonable count.
SmallVector<MachineInstr *, 4> ToBeRemoved;
[CSSPGO] Introducing dangling pseudo probes. Dangling probes are the probes associated to an empty block. This usually happens when all real instructions are optimized away from the block. There is a problem with dangling probes during the offline counts processing. The way the sample profiler works is that samples collected on the first physical instruction following a probe will be counted towards the probe. This logically equals to treating the instruction next to a probe as if it is from the same block of the probe. In the dangling probe case, the real instruction following a dangling probe actually starts a new block, and samples collected on the new block may cause issues when counted towards the empty block. To mitigate this issue, we first try to move around a dangling probe inside its owning block. If there are still native instructions preceding the probe in the same block, we can then use them as a place holder to collect samples for the probe. A pass is added to walk each block backwards looking for probes not followed by any real instruction and moving them before the first real instruction. This is done right before the object emission. If we are unlucky to find such in-block preceding instructions for a probe, the solution we are taking is to tag such probe as dangling so that the samples reported for them will not be trusted by the compiler. We leave it up to the counts inference algorithm to get such probes a reasonable count. The number `UINT64_MAX` is used to mark sample count as collected for a dangling probe. Reviewed By: wmi Differential Revision: https://reviews.llvm.org/D95962
2021-02-25 16:43:17 +08:00
for (MachineInstr &MI : MBB) {
if (MI.isPseudoProbe())
ToBeRemoved.push_back(&MI);
[CSSPGO] Introducing dangling pseudo probes. Dangling probes are the probes associated to an empty block. This usually happens when all real instructions are optimized away from the block. There is a problem with dangling probes during the offline counts processing. The way the sample profiler works is that samples collected on the first physical instruction following a probe will be counted towards the probe. This logically equals to treating the instruction next to a probe as if it is from the same block of the probe. In the dangling probe case, the real instruction following a dangling probe actually starts a new block, and samples collected on the new block may cause issues when counted towards the empty block. To mitigate this issue, we first try to move around a dangling probe inside its owning block. If there are still native instructions preceding the probe in the same block, we can then use them as a place holder to collect samples for the probe. A pass is added to walk each block backwards looking for probes not followed by any real instruction and moving them before the first real instruction. This is done right before the object emission. If we are unlucky to find such in-block preceding instructions for a probe, the solution we are taking is to tag such probe as dangling so that the samples reported for them will not be trusted by the compiler. We leave it up to the counts inference algorithm to get such probes a reasonable count. The number `UINT64_MAX` is used to mark sample count as collected for a dangling probe. Reviewed By: wmi Differential Revision: https://reviews.llvm.org/D95962
2021-02-25 16:43:17 +08:00
}
for (auto *MI : ToBeRemoved)
MI->eraseFromParent();
Changed |= !ToBeRemoved.empty();
}
}
[CSSPGO] Pseudo probes for function calls. An indirect call site needs to be probed for its potential call targets. With CSSPGO a direct call also needs a probe so that a calling context can be represented by a stack of callsite probes. Unlike pseudo probes for basic blocks that are in form of standalone intrinsic call instructions, pseudo probes for callsites have to be attached to the call instruction, thus a separate instruction would not work. One possible way of attaching a probe to a call instruction is to use a special metadata that carries information about the probe. The special metadata will have to make its way through the optimization pipeline down to object emission. This requires additional efforts to maintain the metadata in various places. Given that the `!dbg` metadata is a first-class metadata and has all essential support in place , leveraging the `!dbg` metadata as a channel to encode pseudo probe information is probably the easiest solution. With the requirement of not inflating `!dbg` metadata that is allocated for almost every instruction, we found that the 32-bit DWARF discriminator field which mainly serves AutoFDO can be reused for pseudo probes. DWARF discriminators distinguish identical source locations between instructions and with pseudo probes such support is not required. In this change we are using the discriminator field to encode the ID and type of a callsite probe and the encoded value will be unpacked and consumed right before object emission. When a callsite is inlined, the callsite discriminator field will go with the inlined instructions. The `!dbg` metadata of an inlined instruction is in form of a scope stack. The top of the stack is the instruction's original `!dbg` metadata and the bottom of the stack is for the original callsite of the top-level inliner. Except for the top of the stack, all other elements of the stack actually refer to the nested inlined callsites whose discriminator field (which actually represents a calliste probe) can be used together to represent the inline context of an inlined PseudoProbeInst or CallInst. To avoid collision with the baseline AutoFDO in various places that handles dwarf discriminators where a check against the `-pseudo-probe-for-profiling` switch is not available, a special encoding scheme is used to tell apart a pseudo probe discriminator from a regular discriminator. For the regular discriminator, if all lowest 3 bits are non-zero, it means the discriminator is basically empty and all higher 29 bits can be reversed for pseudo probe use. Callsite pseudo probes are inserted in `SampleProfileProbePass` and a target-independent MIR pass `PseudoProbeInserter` is added to unpack the probe ID/type from `!dbg`. Note that with this work the switch -debug-info-for-profiling will not work with -pseudo-probe-for-profiling anymore. They cannot be used at the same time. Reviewed By: wmi Differential Revision: https://reviews.llvm.org/D91756
2020-12-02 13:44:06 +08:00
return Changed;
}
private:
uint64_t getFuncGUID(Module *M, DILocation *DL) {
auto *SP = DL->getScope()->getSubprogram();
auto Name = SP->getLinkageName();
if (Name.empty())
Name = SP->getName();
return Function::getGUID(Name);
}
};
} // namespace
char PseudoProbeInserter::ID = 0;
INITIALIZE_PASS_BEGIN(PseudoProbeInserter, DEBUG_TYPE,
"Insert pseudo probe annotations for value profiling",
false, false)
INITIALIZE_PASS_DEPENDENCY(TargetPassConfig)
INITIALIZE_PASS_END(PseudoProbeInserter, DEBUG_TYPE,
"Insert pseudo probe annotations for value profiling",
false, false)
FunctionPass *llvm::createPseudoProbeInserter() {
return new PseudoProbeInserter();
}