Generating Test Cases with WizToolKit

The following test-case generation scripts are provided

Matrix Multiplication

The matrix_prod.py script generates an IR relation demonstrating a matrix multiplication in the IR. The circuit inputs three matrixes, M, N, and C, of sizes a*b, b*c, and a*c. It computes C' := M x N, and then checks that C' == C. M and C are read from the public input stream, while N is read from the private input stream.

The script accepts the following parameters

python3 src/main/python/matrix_prod.py <ir> <a> <b> <c> <p> <relation> <public_inputs> <private_inputs>

  • <ir> A keyword indicating how the relation will be generated.

    • flat_tb A flat matrix-multiply function is generated (no helper functions), using the textbook matrix-multiply algorithm

    • mem_flat_tb Same as flat_tb, however memory management gates are used more aggressively.

    • flat_pt A flat matrix-multiply function is generated (no helper functions), using a derivation of the textbook algorithm which transposes each column in N one at a time. (pt is short for Partial Transpose).

    • mem_flat_pt Same as flat_pt, however memory management gates are used more aggressively.

    • dotprod_tb A matrix-multiply function is generated to take advantage of a dot-product helper function. The textbook algorithm is used.

    • mem_dotprod_tb Same as dotprod_tb, however memory management gates are used more aggressively.

    • dotprod_pt A matrix-multiply function is generated to take advantage of a dot-product helper function, and a partial-transpose helper function.

    • mem_dotprod_pt Same as dotprod_pt, however memory management gates are used more aggressively.

  • <a>, <b>, and <c> matrix size parameters.

  • <p> the field/prime.

  • <relation> the filename of the relation

  • <public_inputs> the filename of the public inputs

  • <private_inputs> the filename of the private inputs

Multiplexer 1-type

The multiplexer 1-type script generates a function implementing an arithmetic multiplexer either in the IR or with a plugin.

The script accepts the following parameters

python3 src/main/python/multiplexer_2type.py <ir> <F> <G> <cases> <items> <output>

  • <ir> A keyword indicating how the relation will be generated

    • no_plugins indicates the multiplexer should be generated without the use of any plugins.

    • mux_v0 use the plugin.

  • <F> the prime used in the multiplexer test relation

  • <cases> the number of cases

  • <item> the number of items in each case

  • <output> the base name of all output files

Multiplexer 2-type

The multiplexer 2-type script generates a function implementing an arithmetic multiplexer. The function has an input wire in field F as the case select wire. It also hase one input range for every case in the multiplexer using field G. The outputs are provided in field G.

The script accepts the following parameters

python3 src/main/python/multiplexer_2type.py <ir> <F> <G> <cases> <items> <output>

  • <ir> A keyword indicating how the relation will be generated

    • no_plugins indicates the multiplexer should be generated without the use of any plugins.

    • The current mux_v0 plugin does not support multiple types.

  • <F> and <G> the primes/fields used for the case selector and item lists respectively.

  • <cases> the number of cases

  • <item> the number of items in each case

  • <output> the base name of all output files

Memchk

The memchk script generates a relation using the ram_arith_v0 and iter_v0 plugins. It emulates the classic "memchk" program writing values to each location in memory and asserting that the same value is read. Then it reads a sequence of random reads and writes from the witness, and asserts the correctness of those too.

The script accepts the following parameters

python3 src/main/python/memchk.py <size> <prime> <output>

  • <size> the size of the RAM buffer to test

  • <prime> the prime used for the index/element type

  • <output> the base name for all output files