Immediate load-use pairs¶

Generated workload catalogue · Execution contract

Purpose¶

Measure the dependency from a loaded value to its immediately following arithmetic consumer.

Setup¶

Property	Value
Workload ID	`load-use`
Selection	`--workload load-use`
Generator	`buildRV32IBenchmark() in RV32IBenchmarks.cpp`
End condition	EBREAK after the loop and result store

n seeded 32-bit words at 0x1000; checksum x10 and independent counter x7 start at zero.

Use the shared address map and parameter bounds. N includes any training iterations; n means size.

Implementation¶

For every word, issue LW into x5, add x5 to x10 immediately, then increment x7. Repeat the same unrolled sequence N times.

for (auto v : input) { sum += v; ++independent_counter; }

This is a behavioral summary of the generated assembly, not the host program being timed.

Expected result¶

Result = Nsum(input) modulo 2^32; x7=Nn. Input is unchanged. The same result is required from load-scheduled.

The runner checks the generator's expected result in addition to retirement-by-retirement oracle validation. Output-array assertions, where described, check more than a checksum.

Measurement and running¶

Compare directly with load-scheduled. The instruction multiset and data are identical; only the position of an independent instruction changes.

build/rv32i_experiment --core advanced --workload load-use \
  --iterations 2 --size 16 --seed 1  \
  --fidelity behavioral --period 4000 \
  --allocation-limit-mib 512 --max-cycles 4096

The explicit period is for a bounded functional check, not a hardware-speed result. For comparisons, use the serial campaign driver, identical case inputs, and a certified per-target clock.

Source and tests¶

src/analysis/RV32IBenchmarks.cpp: emitted instructions and independent expected results.
tests/cpp/RV32IBenchmarkTest.cpp: generator results, parameter bounds, and edge cases.
tests/checks/RV32IBenchmarkCLITest.py and tests/checks/BenchmarkCampaignCLITest.py: shared execution/campaign checks.