Test vectors are located in the testVectors directory. Each test vector contains parameters and data arrays defined in the cuMAC API structures (aerial_sdk/cuMAC/src/api.h): cumacCellGrpUeStatus, cumacCellGrpPrms, and cumacSchdSol.
Parameter configurations can be specified the aerial_sdk/cuMAC/examples/parameters.h file.
Use the multiCellSchedulerUeSelection testbench (aerial_sdk/cuMAC/examples/multiCellSchedulerUeSelection) to create TVs:
DL TV:
./aerial_sdk/cuMAC/build/examples/multiCellSchedulerUeSelection/multiCellSchedulerUeSelection -t 1
UL TV:
./aerial_sdk/cuMAC/build/examples/multiCellSchedulerUeSelection/multiCellSchedulerUeSelection -d 0 -t 1
An H5 TV is created after the last simulated TTI. The assumption is that the simulation duration is long enough so that the scheduler algorithm’s performance converges.
Given the same input parameters of a single TTI, GPU and CPU implementations of the same scheduler algorithms should give the same output solution.
Two types of tests:
Per scheduler module tests: DL/UL UE selection, DL/UL PRG allocation, DL/UL layer selection, and DL/UL MCS selection
Complete DL/UL scheduler pipeline tests
TV loading-based single-TTI testbench (aerial_sdk/cuMAC/examples/tvLoadingTest).
After building cumac, use the following commend to check input arguments of the testbench: ./aerial_sdk/cuMAC/build/examples/tvLoadingTest/tvLoadingTest -h
Per scheduler module tests:
DL UE selection:
./aerial_sdk/cuMAC/build/examples/tvLoadingTest/tvLoadingTest -i [path to TV] -g 2 -d 1 -m 01000
DL PRG allocation:
./aerial_sdk/cuMAC/build/examples/tvLoadingTest/tvLoadingTest -i [path to TV] -g 2 -d 1 -m 00100
DL layer selection:
./aerial_sdk/cuMAC/build/examples/tvLoadingTest/tvLoadingTest -i [path to TV] -g 2 -d 1 -m 00010
DL MCS selection:
./aerial_sdk/cuMAC/build/examples/tvLoadingTest/tvLoadingTest -i [path to TV] -g 2 -d 1 -m 00001
UL scheduler modules can be tested by setting input argument:
-d 0
Complete DL/UL scheduler pipeline tests
DL/UL scheduler modules executed sequentially: UE selection > PRG allocation > layer selection > MCS selection
DL scheduler pipeline:
./aerial_sdk/cuMAC/build/examples/tvLoadingTest/tvLoadingTest -i [path to TV] -g 2 -d 1 -m 01111
UL scheduler pipeline:
./aerial_sdk/cuMAC/build/examples/tvLoadingTest/tvLoadingTest -i [path to TV] -g 2 -d 0 -m 01111
Passing criteria:
Solutions computed by CPU and GPU should match exactly: testbench returns 1 (PASS) or 0 (FAIL)
With the same initial sate, GPU and CPU implementations of the same scheduler algorithms should achieve similar performance curves when running for a period of time.
Complete DL/UL scheduler pipeline tests
Continuous-time testbench (
aerial_sdk/cuMAC/examples/multiCellSchedulerUeSelection)After building cumac, use the following commend to check input arguments of the testbench:
./aerial_sdk/cuMAC/build/examples/multiCellSchedulerUeSelection/multiCellSchedulerUeSelection -h
No need to use pre-generated H5 TVs. All parameters are computed using cuMAC internal simulator.
Simulator configuration can be specified using the
aerial_sdk/cuMAC/examples/parameters.hfile.DL/UL scheduler modules executed sequentially: UE selection > PRG allocation > layer selection > MCS selection
DL scheduler pipeline test:
./aerial_sdk/cuMAC/build/examples/multiCellSchedulerUeSelection/multiCellSchedulerUeSelection
UL scheduler pipeline test:
./aerial_sdk/cuMAC/build/examples/multiCellSchedulerUeSelection/multiCellSchedulerUeSelection -d 0
Passing criteria:
Performance curves achieved by GPU and CPU scheduler implementations should match: testbench returns 1 (PASS) or 0 (FAIL)
Two types of performance curves:
Sum throughput of all cells
CDF of per-UE throughput