openpower/sv/int_fp_mv/cvt_fp_to_int_overview.mdwn

   1 ## Floating-point to Integer Conversion Overview
   2
   3 <div id="fpr-to-gpr-conversion-mode"></div>
   4
   5 IEEE 754 doesn't specify what results are obtained when converting a NaN
   6 or out-of-range floating-point value to integer, so different programming
   7 languages and ISAs have made different choices.  The different conversion
   8 modes supported by the `cffpr` instruction are as follows:
   9
  10 * P-Type:
  11     This type of conversion is that used by most other PowerISA instructions,
  12     as well as commonly used floating-point to integer conversions on x86.
  13
  14 * S-Type:
  15     This type of conversion is used for WebAssembly's
  16     [`trunc_sat_u`](https://webassembly.github.io/spec/core/exec/numerics.html#op-trunc-sat-u)
  17     and
  18     [trunc_sat_s](https://webassembly.github.io/spec/core/exec/numerics.html#op-trunc-sat-s)
  19     instructions, as well as several notable programming languages:
  20
  21     * Java's conversion from
  22         [`float`/`double` to `long`/`int`](https://docs.oracle.com/javase/specs/jls/se16/html/jls-5.html#jls-5.1.3)
  23     * Rust's [`as` operator](https://doc.rust-lang.org/reference/expressions/operator-expr.html#semantics)
  24     * LLVM's [`llvm.fptosi.sat`](https://llvm.org/docs/LangRef.html#llvm-fptosi-sat-intrinsic) and
  25         [`llvm.fptoui.sat`](https://llvm.org/docs/LangRef.html#llvm-fptoui-sat-intrinsic) intrinsics
  26     * SPIR-V's OpenCL dialect's
  27         [`OpConvertFToU`](https://www.khronos.org/registry/spir-v/specs/unified1/SPIRV.html#OpConvertFToU) and
  28         [`OpConvertFToS`](https://www.khronos.org/registry/spir-v/specs/unified1/SPIRV.html#OpConvertFToS)
  29         instructions when decorated with
  30         [the `SaturatedConversion` decorator](https://www.khronos.org/registry/spir-v/specs/unified1/SPIRV.html#_a_id_decoration_a_decoration).
  31
  32 * E-Type:
  33     This type of conversion is used for
  34     [ECMAScript's `ToInt32` abstract operation](https://262.ecma-international.org/14.0/#sec-toint32).
  35
  36     This type of conversion is also implemented in ARMv8.3A as the `FJCVTZS`
  37     instruction.
  38     <https://developer.arm.com/documentation/dui0801/g/hko1477562192868>
  39
  40 ### Floating-point to Integer Conversion Semantics Summary
  41
  42 Let `rounded` be the result of `bfp_ROUND_TO_INTEGER(rmode, input)`.
  43 Let `w` be the number of bits in the result's type.
  44 The result of Floating-point to Integer conversion is as follows:
  45
  46 ```
  47 +--------+------------+-----------------------------------------------------------------------+
  48 | Kind   | Result's   | Category of rounded                                                   |
  49 |        | Signedness +-----------+-----------+-----------+-----------+-----------+-----------+
  50 |        |            | NaN       | +Infinity | -Infinity | > Maximum | < Minimum | Otherwise |
  51 |        |            |           |           |           | Possible  | Possible  |           |
  52 |        |            |           |           |           | Result    | Result    |           |
  53 +--------+------------+-----------+-----------+-----------+-----------+-----------+-----------+
  54 | P-Type | Unsigned   | 0         | 2^w - 1   | 0         | 2^w - 1   | 0         | rounded   |
  55 |        +------------+-----------+-----------+-----------+-----------+-----------+-----------+
  56 |        | Signed     | -2^(w-1)  | 2^(w-1)-1 | -2^(w-1)  | 2^(w-1)-1 | -2^(w-1)  | rounded   |
  57 +--------+------------+-----------+-----------+-----------+-----------+-----------+-----------+
  58 | S-Type | Unsigned   | 0         | 2^w - 1   | 0         | 2^w - 1   | 0         | rounded   |
  59 |        +------------+-----------+-----------+-----------+-----------+-----------+-----------+
  60 |        | Signed     | 0         | 2^(w-1)-1 | -2^(w-1)  | 2^(w-1)-1 | -2^(w-1)  | rounded   |
  61 +--------+------------+-----------+-----------+-----------+-----------+-----------+-----------+
  62 | E-Type | Either     | 0                                 | rounded & (2^w - 1)               |
  63 +--------+------------+-----------------------------------+-----------------------------------+
  64 ```