Floating Point Writing Fixes

specification/under_development/profiles/Adaptive_Profile.md

@@ -7,7 +7,7 @@
 well as contributors to the [targeting
 stage](../Compilation_And_Targeting.md#targeting) of the QIR compiler.
 
 The Adaptive Profile specifies supersets of the [Base
 Profile](./Base_Profile.md) that enable control flow based on mid-circuit
 measurements and classical computations while quantum resources remain coherent.
 A backend can support this profile by supporting a minimum set of features
@@ -20,7 +20,7 @@
 1. It can execute a sequence of quantum instructions that transform the quantum
    state.
 2. A backend must support applying a measurement operation at any point in the
    execution of the program. Qubits not undergoing the measurement should not
    have their state affected.
 3. A backend must be able to apply quantum instructions conditionally on a
    measurement outcome. Specifically, forward branching using the LLVM branch
@@ -84,16 +84,16 @@
 [`irreversible` attribute](./Base_Profile.md#quantum-instruction-set) that
 populates a value of type `%Result`. The available measurement functions are
 defined by the executing backend as part of the QIS. Unlike the Base Profile,
 the Adaptive Profile relieves a restriction that qubits can only be measured at
 the end of the program and that no quantum operations can be performed after or
 conditionally on a measurement outcome.
 
 Within the Adaptive Profile, there are no restrictions on when these
 measurements are performed during program execution. Correspondingly, it must be
 possible to measure individual qubits, or subsets of qubits depending on the
 supported QIS, without impacting the state of the non-measured qubits.
 Furthermore, it must be possible to use the measured qubit(s) afterwards and
 apply additional quantum instructions to the same qubit(s).
 
 ### Bullet 3: Forward Branching
 
@@ -115,9 +115,9 @@
 control flow graph and reject any program with a cycle as invalid unless support
 for backward branching is enabled.
 
 While the Adaptive Profile requires that it must be possible to use a qubit
 after it was measured, the availability of a `reset` instruction depends on the
 QIS supported by the backend. If a single-qubit measurement is supported as part
 of the QIS, it is always possible to reset that qubit, if needed for qubit reuse
 later in the program, using forward branching as illustrated for example by the
 following IR snippet:
@@ -660,6 +660,7 @@
 | `fsub`           | Subtracts two floating-point values. |                             |
 | `fmul`           | Multiplies two floating-point values.          |                             |
 | `fdiv`           | Divides two floating-point values. | Division by zero leads to undefined behavior, no support for `NaN`. |
+| `fcmp`           | Compares two floating-point | Comparision options are `olt`, `ole`, `ogt`, `oge`, `oeq`, `one`, `ord`, `ult`, `ule`, `ugt`, `uge`, `ueq`, `une`, `uno`, `false`, `true`. |
 | `fpext .. to`           | Casts a value of floating-point type to a larger floating-point type. | May be used at any point in the program if classical computations on both the input and the output type are supported. May only be used as part of a call to an output recording function if computations on the output type are not supported. |
 | `fptrunc .. to`  | Casts a value of floating-point type to a smaller floating-point type.         | May be used at any point in the program if classical computations on both the input and the output type are supported. May only be used as part of a call to an output recording function if computations on the output type are not supported. |
 
@@ -699,7 +700,7 @@
 
 | Function                            | Signature       | Description     |
 | :---------------------------------- | :-------------- | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| __quantum__rt__float_record_output | `void(double,i8*)` | Records a floating-point value in the generated output. The second parameter defines the string label for the value. Depending on the output schema, the label is included in the output or omitted. |
+| __quantum__rt__double_record_output | `void(double,i8*)` | Records a floating-point value in the generated output. The second parameter defines the string label for the value. Depending on the output schema, the label is included in the output or omitted. |
 
 ### Output Recording
 
@@ -813,8 +814,8 @@
   reference to another metadata node that contains a tuple of string literals
   of the supported/used floating-point precision(s). For example,
   `!0 = !{i32 5, !"float_computations", !1}` and `!1 = !{!"float", !"double"}`.
-  The precision must be one of the LLVM IEEE-754 floating-point types
-  ("half" (16-bit), "float" (32-bit), "double" (64-bit), "fp128" (128-bit)),
+  The precision must be one of the LLVM IEEE-754 floating-point types up to 64-bits
+  (`half` (16-bit), `float` (32-bit), `double` (64-bit)),
   and classical computations on floating point numbers of all listed
   precisions must be supported by the executing backend. An empty value
   indicates that no floating-point computations are supported/used.