decoupled direct memory accessomutlu/pub/decoupled... · decoupled direct memory access cpu access....
TRANSCRIPT
![Page 1: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/1.jpg)
IsolatingCPUandIOTrafficbyLeveragingaDual-Data-PortDRAM
DonghyukLeeLavanya Subramanian,Rachata Ausavarungnirun,
Jongmoo Choi,Onur Mutlu
DecoupledDirectMemoryAccess
![Page 2: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/2.jpg)
2
processor
LogicalSystemOrganization
mainmemory
IOdevices
CPUaccess
IOaccess
MainmemoryconnectsprocessorandIOdevicesasanintermediatelayer
![Page 3: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/3.jpg)
3
processor
PhysicalSystemImplementation
mainmemory
IOdevices
CPUaccess
IOaccess
IOaccess
HighPinCostinProcessor
HighContentioninMemoryChannel
![Page 4: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/4.jpg)
4
processor
OurApproach
mainmemory
IOdevices
CPUaccess
EnablingIOchannel,decoupled & isolated fromCPUchannel
IOaccess
IOaccess
![Page 5: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/5.jpg)
5
ExecutiveSummary• Problem
– CPUandIOaccessescontendforthesharedmemorychannel
• OurApproach:DecoupledDirectMemoryAccess(DDMA)– DesignnewDRAMarchitecturewithtwoindependentdataports
àDual-Data-PortDRAM– ConnectoneporttoCPUandtheotherporttoIOdevices
àDecoupleCPUandIOaccesses
• Application– Communicationbetweencomputeunits(e.g.,CPU–GPU)– In-memorycommunication(e.g.,bulkin-memorycopy/init.)– Memory-storagecommunication(e.g.,pagefault,IOprefetch)
• Result– Significantperformanceimprovement(20%in2ch.&2ranksystem)– CPUpincountreduction(4.5%)
![Page 6: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/6.jpg)
6
Outline1.Problem
3.Dual-Data-PortDRAM
5.Evaluation
4.ApplicationsforDDMA
2.OurApproach
1.Problem
![Page 7: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/7.jpg)
7
mainmemory
CPU
DMA
graphics
network
storage
USB
IOinterfacememorycontroller
MemoryChannelContentionDRAMChip
ProcessorChip
Problem1:MemoryChannelContention
DMAIOinterface
![Page 8: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/8.jpg)
8
0%
20%
40%
60%
80%
100%TimeSpentonCPU-GPUCommunication
Benchmarks
33.5%onaverage
Fractio
nofExecutio
nTime
AlargefractionoftheexecutiontimeisspentonIOaccesses
Problem1:MemoryChannelContention
![Page 9: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/9.jpg)
9
IntegratingIOinterfaceontheprocessorchipleadstohighareacost
ProcessorPinCount(w/opowerpins)
power memory(2ch)
IOinterface(10.6%)
IOinterface(28.4%)
others
memory(2ch)
(w/powerpins)ProcessorPinCount
959pinsintotal 359pinsintotal
Problem2:HighCostforIOInterfaces
![Page 10: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/10.jpg)
10
SharedMemoryChannel
• MemorychannelcontentionforIOaccessandCPUaccess
• HighareacostforintegratingIOinterfacesonprocessorchip
![Page 11: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/11.jpg)
11
Outline1.Problem
3.Dual-Data-PortDRAM
5.Evaluation
4.ApplicationsforDDMA
2.OurApproach
![Page 12: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/12.jpg)
12
OurApproach
CPU
DMA
graphics
network
storage
USB
DRAMChip
mainmemory
?
DMACTRL.
DMAcontrol
ProcessorChip
controlchannel
Dual-Data-PortDRAM
Port1
Port2
memorycontroller IOinterface
DMAChip DMAIOinterface
![Page 13: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/13.jpg)
13
OurApproach
?
CPU
graphics
network
storage
USB
DRAMChip
DMACTRL.
DMAcontrol
ProcessorChip
controlchannel
Dual-Data-PortDRAM
Port1
Port2
memorycontroller
DMAChip DMAIOinterface
IOACCESS
DecoupledDirectMemoryAccess
CPUACCESS
![Page 14: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/14.jpg)
14
Outline1.Problem
3.Dual-Data-PortDRAM
5.Evaluation
4.ApplicationsforDDMA
2.OurApproach
![Page 15: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/15.jpg)
15
peripherallogic
bank
Background:DRAMOperation
mem
orychannel
datachannel controlchannel
control
port
dataport
control
port
dataport
bank
activateread
bankbankREADY
DRAMperipherallogic:i)controlsbanks,andii)transfersdataovermemorychannel
memorycontrolleratCPU
![Page 16: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/16.jpg)
16
bank
Problem:SingleDataPort
periphery
Requestsareservedseriallyduetosingledataport
datachannel controlchannel
control
port
dataport
read
control
port
dataport
bankREADY
bankREADY
dataport
read
ManyBanks
SingleDataPort
memorycontrolleratCPU
![Page 17: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/17.jpg)
17
Problem:SingleDataPort
RD
DATA
RD
DATA
ControlPort
DataPort
time
RD
DATA
RDControlPort
DataPort1
time
DATADataPort2
WhataboutaDRAMwithtwodataports?
![Page 18: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/18.jpg)
18
bank
periphery
twicethebandwidth&independentdataportswithlowoverhead
datachannel controlchanneldataport1
bank
bank
control
port
toPort1(upper)
toPort2(lower)
bankdatabus
portse
lectsignal
dataport2
datachannel
mux
mux
OverheadArea:1.6%↑Pins:20↑
Dual-Data-PortDRAM
![Page 19: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/19.jpg)
19
DDP-DRAMMemorySystem
bank
periphery
CPUchannel controlchannelwithportselectdata
port1
bank
bank
control
port
dataport2
IOchannel
mux
mux
DDMAIOinterface
memorycontrolleratCPU
![Page 20: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/20.jpg)
20
ThreeDataTransferModes
• CPUAccess:AccessthroughCPUchannel– DRAMread/writewithCPUportselection
• IOAccess:AccessthroughIOchannel– DRAMread/writewithIOportselection
• PortBypass:Directtransferbetweenchannels– DRAMaccesswithportbypassselection
![Page 21: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/21.jpg)
21
1.CPUAccessMode
bank
periphery
CPUchannel
bank
control
port
dataport2
IOchannel
DDMAIOinterface
controlchannelwithportselect
mux
mux
dataport
bankREADY
memorycontrolleratCPU
read
control
port
CPUchanneldataport1
controlchannelwithCPUchannel
![Page 22: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/22.jpg)
22
2.IOAccessMode
bank
periphery
CPUchannel
bank
control
portIOchannel
DDMAIOinterface
controlchannelwithportselect
mux
mux
dataport1
controlchannelwith IOchannel
memorycontrolleratCPU
IOchannel
dataportdataport2
bankREADY
read
control
port
![Page 23: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/23.jpg)
23
3.PortBypassMode
bank
periphery
CPUchannel
bank
control
portIOchannel
controlchannelwithportselect
mux
mux
controlchannelwith portbypass
IOchannel
bank
dataport
dataport
dataport2
dataport1
CPUchannel
DDMAIOinterface
memorycontrolleratCPU
![Page 24: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/24.jpg)
24
Outline1.Problem
3.Dual-Data-PortDRAM
5.Evaluation
4.ApplicationsforDDMA
2.OurApproach
![Page 25: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/25.jpg)
25
ThreeApplicationsforDDMA
• Communicationb/wComputeUnits– CPU-GPUcommunication
• In-MemoryCommunicationandInitialization– Bulkpagecopy/initialization
• Communicationb/wMemoryandStorage– Servingpagefault/fileread&write
![Page 26: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/26.jpg)
26
ctrl.channel
DDMActrl.
read
with
IOse
l.
CPU→
GPU
1.ComputeUnit↔ComputeUnitCPU
DDMActrl.
memorycontroller
DDP-DRAM
DDMAIOinterface
GPU
DDMActrl.
memorycontroller
DDP-DRAM
DDMAIOinterface
ctrl.channel
DDMActrl.
destination
DDMAIOinterface
source Ack.destination
DDMAIOinterface
write
with
IOse
l.
TransferdatathroughDDMAwithoutinterferingw/CPU/GPUmemoryaccesses
CPU
memorycontroller
GPU
memorycontroller
![Page 27: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/27.jpg)
27
ctrl.chan.
read
with
IOse
l.write
with
IOse
l.
2.In-MemoryCommunication
DDMActrl.
CPU
DDMActrl.
memorycontroller
DDP-DRAM
DDMAIOinterface
sourcedestination
TransferdatainDRAMthroughDDAMwithoutinterferingwithCPUmemoryaccesses
CPU
memorycontroller
![Page 28: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/28.jpg)
28
DDMActrl.
Acc.Storage
Ack.
3.Memory↔Storage
ctrl.chan.
write
with
IOse
l.
CPU
DDMActrl.
memorycontroller
DDP-DRAM
DDMAIOinterface StorageStorage(source)
destination
DDMAIOinterface
TransferdatafromstoragethroughDDMAwithoutinterferingwithCPUmemoryaccesses
destination
CPU
memorycontroller
![Page 29: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/29.jpg)
29
Outline1.Problem
3.Dual-Data-PortDRAM
5.Evaluation
4.ApplicationsforDDMA
2.OurApproach
![Page 30: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/30.jpg)
30
EvaluationMethods• System
– Processor:4– 16cores– LLC:16-wayassociative,512KBprivatecache-slice/core– Memory:1– 4ranksand1– 4channels
• Workloads– Memoryintensive:SPECCPU2006,TPC,stream(31benchmarks)
– CPU-GPUcommunicationintensive:polybench (8benchmarks)
– In-memorycommunicationintensive:apache,bootup,compiler,filecopy,mysql,fork,shell,memcached (8intotal)
![Page 31: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/31.jpg)
31
0%
5%
10%
15%
20%
25%
4-Core 8-Core 16-Core0%
5%
10%
15%
20%
25%
4-Core 8-Core 16-Core
Perfo
rmanceIm
provem
ent
Perfo
rmanceIm
provem
ent
CPU-GPUComm.-Intensive In-MemoryComm.-Intensive
More performance improvementathighercorecountHighperformanceimprovement
Performance(2Channel,2Rank)
![Page 32: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/32.jpg)
32
PerformanceonVariousSystems
0%5%10%15%20%25%30%35%40%
1rank 2rank 4rank0%5%10%15%20%25%30%35%40%
1ch 2ch 4ch
ChannelCount RankCount
Perfo
rmanceIm
provem
ent
Perfo
rmanceIm
provem
ent
Performanceincreaseswithrankcount
![Page 33: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/33.jpg)
33
0
200
400
600
800
1000
1200
1ch 1chDDMA
2ch0%20%40%60%80%100%120%140%160%180%
1ch 1chDDMA
2ch
Perfo
rmance
ProcessorP
inCou
nt
DDMAachieveshigherperformanceatlowerprocessorpincount
959 915
1103
DDMAvs.DoublingChannel
![Page 34: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/34.jpg)
34
Conclusion• Problem
– CPUandIOaccessescontendforthesharedmemorychannel
• OurApproach:DecoupledDirectMemoryAccess(DDMA)– DesignnewDRAMarchitecturewithtwoindependentdataports
àDual-Data-PortDRAM– ConnectoneporttoCPUandtheotherporttoIOdevices
àDecoupleCPUandIOaccesses
• Application– Communicationbetweencomputeunits(e.g.,CPU–GPU)– In-memorycommunication(e.g.,bulkin-memorycopy/init.)– Memory-storagecommunication(e.g.,pagefault,IOprefetch)
• Result– Significantperformanceimprovement(20%in2ch.&2ranksystem)– CPUpincountreduction(4.5%)
![Page 35: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/35.jpg)
IsolatingCPUandIOTrafficbyLeveragingaDual-Data-PortDRAM
DonghyukLeeLavanya Subramanian,Rachata Ausavarungnirun,
Jongmoo Choi,Onur Mutlu
DecoupledDirectMemoryAccess
![Page 36: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/36.jpg)
36
SystemOverhead
DDMAreducesmoreexpensiveon-chiparea,whileincreasinglessexpensiveoff-chiparea
processor
DRAM
IOdevices
ConventionalSystem
processor
DDP-DRAM
IOdevicesDDMA-IO
ProposedSystem
LowC
ost
High
![Page 37: Decoupled Direct Memory Accessomutlu/pub/decoupled... · Decoupled Direct Memory Access CPU ACCESS. 14 Outline 1. Problem 3. Dual-Data-Port DRAM 5. Evaluation 4. Applications for](https://reader034.vdocuments.site/reader034/viewer/2022051808/6009570eed8f311c707280bf/html5/thumbnails/37.jpg)
37
0%10%20%30%40%50%60%70%80%90%100%
1Channel 2Channel 2Channel 1Rank 2Rank 4Rank
ChannelUtilizationAnalysis
SimultaneousChannelUtilizationàPerformanceImprovement
CPU-GPUCommunication-Intensive
ChannelU
tiliza
tion
CPU IO
CPU IO
CPU IO
CPU IO
CPU IO
CPU IO
0%10%20%30%40%50%60%70%80%90%100%
1Channel 2Channel 2Channel 1Rank 2Rank 4Rank
0%10%20%30%40%50%60%70%80%90%100%
1Channel 2Channel 2Channel 1Rank 2Rank 4Rank
0%10%20%30%40%50%60%70%80%90%100%
1Channel 2Channel 2Channel 1Rank 2Rank 4Rank
BothChannelsBusy SingleChannelBusy
4