Moving features into experimental

[troelsfr]

All experimental and non-experimental features are listed in the help section:

% ./qir/qat/Apps/qat
Usage: /Users/tfr/Documents/Projects/qat/Debug/qir/qat/Apps/qat [options] filename

Component configuration - Used to disable or enable components

--disable-validation-configuration                Disables Validation configuration. Default: false
--disable-llvm-optimisation                       Disables LLVM optimisations. Default: false
--disable-transformation-rules                    Disables Pass configuration. Default: false
--disable-grouping                                Disables Grouping quantum instructions. Default: false

Base configuration - Configuration of the quantum adoption tool to execute a specific behaviour.

--load                                            EXPERIMENTAL. Load component. Default:
--apply                                           Applies a profile to transform the IR in correspondence with the profile. Default: false
--validate                                        Executes the validation procedure. Default: false
--profile                                         Sets the profile. Default: generic
--S                                               Emits LLVM IR to the standard output. Default: false
--O0                                              Optimisation level 0. Default: false
--O1                                              Optimisation level 1. Default: false
--O2                                              Optimisation level 2. Default: false
--O3                                              Optimisation level 3. Default: false
--verify-module                                   Verifies the module after transformation. Default: false
--experimental                                    Enables experimental features. Default: false
--dump-config                                     Prints the configuration to the standard output. Default: false

Transformation rules - Rules used to transform instruction sequences in the QIR.

--[no-]optimise-result-one                        EXPERIMENTAL. Maps branching based on quantum measurements compared to one to base profile type measurement. Default: true
--[no-]optimise-result-zero                       EXPERIMENTAL. Maps branching based on quantum measurements compared to zero to base profile type measurement. Default: true
--[no-]use-static-qubit-array-allocation          Maps allocation of qubit arrays to static array allocation. Default: true
--[no-]use-static-qubit-allocation                Maps qubit allocation to static allocation. Default: true
--[no-]use-static-result-allocation               Maps result allocation to static allocation. Default: true
--[no-]disable-reference-counting                 Disables reference counting by instruction removal. Default: true
--[no-]disable-alias-counting                     Disables alias counting by instruction removal. Default: true
--[no-]disable-string-support                     Disables string support by instruction removal. Default: true

Validation configuration -

--[no-]allow-internal-calls                       Whether or not internal calls are allowed. Default: true
--save-validation-report                          Saves the validation report to specified filename. Default:

LLVM optimisations - Enables specific LLVM optimisations before and after transformation.

--unroll-loops                                    Aggressively unroll loops. Default: false
--[no-]allow-partial                              Enables or disables partial unrolling. When disabled only full unrolling is allowed. Default: true
--[no-]allow-peeling                              Enables or disables loop peeling. Default: true
--[no-]allow-runtime                              Enables or disables unrolling of loops with runtime trip count. Default: true
--[no-]allow-upper-bound                          Enables or disables the use of trip count upper bound in loop unrolling. Default: true
--[no-]allow-profile-based-peeling                Enables or disables loop peeling basing on profile. Default: true
--full-unroll-count                               Sets the max full unroll count. Default: 1024
--unroll-opt-level                                Sets the optimisation level for loop unrolling. Default: 3
--only-when-forced                                If true, only loops that explicitly request unrolling via metadata are considered. Default: false
--forget-scev                                     If true, forget all loops when unrolling. If false, forget top-most loop of the currently processed loops. Default: false
--always-inline                                   Aggressively inline function calls. Default: false
--inlining-parameter                              Number of code lines acceptable when inlining. Default: 2147483647
--[no-]eliminate-constants                        Uses LLVM pass to eliminate constants. Default: true
--[no-]eliminate-dead-code                        Uses LLVMs aggressive dead code elimination. Default: true
--[no-]eliminate-memory                           Maps memory into registers where feasible. Default: true
--use-llvm-opt-pipeline                           EXPERIMENTAL. Disables the the default pipeline. Default: false
--opt-pipeline-config                             EXPERIMENTAL. LLVM passes pipeline to use upon applying this component. Default:

Pass configuration - Configuration of the pass and its corresponding optimisations.

--[no-]delete-dead-code                           EXPERIMENTAL. Deleted dead code. Default: true
--[no-]clone-functions                            EXPERIMENTAL. Clone functions to ensure correct qubit allocation. Default: true
--[no-]transform-execution-path-only              EXPERIMENTAL. Transform execution paths only. Default: true
--max-recursion                                   EXPERIMENTAL. Defines the maximum recursion when unrolling the execution path Default: 512
--assume-no-except                                EXPERIMENTAL. Assumes that no exception will occur during runtime. Default: false
--reuse-qubits                                    EXPERIMENTAL. Use to define whether or not to reuse qubits. Default: false
--reuse-results                                   EXPERIMENTAL. Use to define whether or not to reuse results. Default: false
--entry-point-attr                                Specifies the attribute indicating the entry point. Default: InteropFriendly
--[no-]annotate-qubit-use                         Annotate the number of qubits used Default: true
--[no-]annotate-result-use                        Annotate the number of results used Default: true

Grouping quantum instructions -

--[no-]separate-circuits                          EXPERIMENTAL. Whether or not to separate quantum and classical circuits Default: true

[bettinaheim]

@troelsfr Could you please ensure that all APIs (and also doc comments, but I don't mind if these get gradually updated) follow American standard spelling? While that is not a reflection of who the tool is for, overall it is nicer and less surprising for users to have some consistency across all tools/packages that we maintain.

[troelsfr]

Sure. I will pick this up on Monday.

On Sat, 26 Feb 2022, 04:11 Bettina Heim, @.***> wrote:

@troelsfr https://github.com/troelsfr Could you please ensure that all APIs (and also doc comments, but I don't mind if these get gradually updated) follow American standard spelling? While that is not a reflection of who the tool is for, overall it is nicer and less surprising for users to have some consistency across all out tools.

— Reply to this email directly, view it on GitHub https://github.com/qir-alliance/qat/pull/58#issuecomment-1051527011, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAEMHDXQGH3D6WUWXIBZZBDU5BAGPANCNFSM5PD27RJQ . You are receiving this because you were mentioned.Message ID: @.***>

[troelsfr]

Code changes look good, but one thing that took me a while to piece together is that experimental features effectively have two default settings: one default when experimental is not provided and one when it is. The help description of these parameters is a little confusing in that regard, because it describes some parameters as default true and it makes it seem as if some experimental features are enabled by default for normal operation. Perhaps some additional text on the --experimental parameter itself to clarify that any features that are marked EXPERIMENTAL are disabled if the top level --experimental param is not provided?

Yes, this is a good suggestion. I will make this change today.