Project

General

Profile

News

MicroTESK: MicroTESK 2.3.20 beta released

Added by Andrei Tatarnikov about 9 years ago

The new release contains the following changes:

  • Support for randomizing preparators (random selection between several preparator variants) was implemented (*)
  • Some issues in the branch generation engine were fixed

The MicroTESK distribution package can be downloaded from here: http://forge.ispras.ru/projects/microtesk/files

What's new (*):

Preparator Randomization

Now it is possible to specify several variants of a preparator, which can be selected at random when the preparator is instantiated. Syntax is the following:

preparator(:target => 'REG') {
  variant(:name => '1', :bias => 50) {
    comment 'Variant 1'
    lui target, value(16, 31)
    ori target, target, value(0, 15)
  }

  variant(:name => '2', :bias => 50) {
    comment 'Variant 2'
    ori target, zero, value(16, 31)
    sll target, target, 16
    addi target, target, value(0, 15)
  }
}

A variant description has two optional parameters: name and bias.

The name attribute allows specifying a preferred preparator variant when a preparator is explicitly inserted by using the prepare function. For example:

prepare t1, 0xdeadbeef, :variant => '1' # Preparator '1' is chosen.
prepare t2, 0xbaadf00d, :variant => '2' # Preparator '2' is chosen.

Search by variant name is performed after a suitable preparator is chosen according to value mask and argument values. So if a variant with the specified name is not defined for the chosen preparator, MicroTESK reports a warning and uses a randomly chosen variant of the given preparator.

The bias attribute is used set up a bias for the current variant. It works in a similar way as the dist construct which is for selecting random values. The atribute can be skipped if you want to define equally probable variants.

MicroTESK: MicroTESK 2.3.19 beta released

Added by Andrei Tatarnikov about 9 years ago

The new release contains the following changes:

  • Support for instruction permutations in test template blocks was implemented (*)

The MicroTESK distribution package can be downloaded from here: http://forge.ispras.ru/projects/microtesk/files

What's new (*):

Instruction Permutations

Test template blocks how have a new parameter permutator that specifies the permutation method to be applied to instruction sequences described by these blocks. Currently, two methods are supported:

  1. trivial - order of instruction is not changed
  2. random - instructions come in random order

For example, the following code produces a random permutation of the specified instructions:

atomic(:permutator => 'random') {
  Add reg(_), reg(_), reg(_)
  Sub reg(_), reg(_), reg(_)
  And reg(_), reg(_), reg(_)
  Or  reg(_), reg(_), reg(_)
  Nor reg(_), reg(_), reg(_)
  Xor reg(_), reg(_), reg(_)
}.run

MicroTESK: MicroTESK 2.3.15 beta released

Added by Andrei Tatarnikov over 9 years ago

The new release contains the following changes:

  • Support for explicit invocation of preparators in test templates (*)
  • Support for MMU-related constraints in test situations (*)
  • Some bug fixes and general improvements

The MicroTESK distribution package can be downloaded from here: http://forge.ispras.ru/projects/microtesk/files

What's new (*):

Explicit Preparators Invocation

A preparator can be explicitly inserted into code using the prepare method. Supported use cases:

1. Preparators can be invoked explicitly anywhere in a test template to assign constant or random values to registers. For example:

prepare t0, 0xDEADBEEF
prepare t1, rand(1, 63)

2. Preparators can be invoked explicitly inside of preparator and buffer_preparator blocks to initialize registers with values provided when the block is instantiated. For example:

buffer_preparator(:target => 'L1') {
  prepare t0, address
  lw t1, 0, t0
}

Constraint for Memory-Related Situations

Now test situations for MMU can be specified using constraints. There are two kinds of constraints: event-based (hit, mist and event with distribution) and equality-based (for a single value, array, range and distribution). Syntax is the following:

simple_dist = dist(range(:value => 0,      :bias => 50),
                   range(:value => 1..3,   :bias => 25),
                   range(:value => [5, 7], :bias => 25))

lw s0, 0, t0 do situation('address',
    :path => constraints(eq('L1.V', 0), hit('L1'))) end

lw s0, 0, t0 do situation('address',
    :path => constraints(eq('L1.V', [0, 2]), hit('L1'))) end

lw s1, 0, t1 do situation('address',
    :path => constraints(eq('L1.V', 0..3), miss('L1'))) end

lw s0, 0, t0 do situation('address',
    :path => constraints(eq('L1.V', simple_dist), hit('L1'))) end

lw s1, 0, t1 do situation('address',
    :path => constraints(eq('L1.V', 0), event('L1', :hit => 50, :miss => 50))) end
(191-200/365)

Also available in: Atom