ztrans_proposal.mdwn

   1 # Ztrans - transcendental operations
   2
   3 See:
   4
   5 * <http://bugs.libre-riscv.org/show_bug.cgi?id=127>
   6 * <https://www.khronos.org/registry/spir-v/specs/unified1/OpenCL.ExtendedInstructionSet.100.html>
   7
   8 Extension subsets:
   9
  10 * **Ztrans**: standard transcendentals (best suited to 3D)
  11 * **ZtransExt**: extra functions (useful, not generally needed for 3D)
  12 * Ztrigpi: trig. *-pi sinpi cospi tanpi
  13 * Ztrignpi: trig non-*pi sin cos tan
  14 * Zarctrigpi: arc-trig. *pi atan2pi asinpi acospi
  15 * Zarctrignpi: arc-trig. non-*pi
  16 * **ZtransAdv**: much more complex to implement in hardware
  17
  18 [[!toc levels=2]]
  19
  20 # List of 2-arg opcodes
  21
  22 [[!table  data="""
  23 opcode    | Description           | pseudo-code                | Extension |
  24 FATAN2    | atan2 arc tangent     | rd = atan2(rs2, rs1)       | Ztrans    |
  25 FATAN2PI  | atan arc tangent / pi | rd = atan2(rs2, rs1) / pi  | ZtransExt |
  26 FPOW      | x power of y          | rd = pow(rs1, rs2)         | ZtransAdv |
  27 FROOT     | x power 1/y           | rd = pow(rs1, 1/rs2)       | ZtransAdv |
  28 """]]
  29
  30 # List of 1-arg opcodes
  31
  32 [[!table  data="""
  33 opcode   | Description              | pseudo-code             | Extension |
  34 FCBRT    | Cube Root                | rd = pow(rs1, 3)        | Ztrans    |
  35 FEXP2    | power-of-2               | rd = pow(2, rs1)        | Ztrans    |
  36 FLOG2    | log2                     | rd = log2(rs1)          | Ztrans    |
  37 FEXPM1   | exponent minus 1         | rd = pow(e, rs1) - 1.0  | Ztrans    |
  38 FLOG1P   | log plus 1               | rd = log(e, 1 + rs1)    | Ztrans    |
  39 FEXP     | exponent                 | rd = pow(e, rs1)        | ZtransExt |
  40 FLOG     | natural log (base e)     | rd = log(e, rs1)        | ZtransExt |
  41 FEXP10   | power-of-10              | rd = pow(10, rs1)       | ZtransExt |
  42 FLOG10   | log base 10              | rd = log10(rs1)         | ZtransExt |
  43 FSIN     | sin (radians)            |                         | Ztrignpi    |
  44 FCOS     | cos (radians)            |                         | Ztrignpi    |
  45 FTAN     | tan (radians)            |                         | Ztrignpi    |
  46 FASIN    | arcsin (radians)         | rd = asin(rs1)          | Zarctrignpi   |
  47 FACOS    | arccos (radians)         | rd = acos(rs1)          | Zarctrignpi    |
  48 FSINPI   | sin times pi             | rd = sin(pi * rs1)      | Ztrigpi |
  49 FCOSPI   | cos times pi             | rd = cos(pi * rs1)      | Ztrigpi |
  50 FSINH    | hyperbolic sin (radians) |                         | ZtransExt |
  51 FCOSH    | hyperbolic cos (radians) |                         | ZtransExt |
  52 FTANH    | hyperbolic tan (radians) |                         | ZtransExt |
  53 FASINH   | inverse hyperbolic sin   |                         | ZtransExt |
  54 FACOSH   | inverse hyperbolic cos   |                         | ZtransExt |
  55 FATANH   | inverse hyperbolic tan   |                         | ZtransExt |
  56 """]]
  57
  58 # Pseudo-code ops and macro-ops
  59
  60 * FRCP rd, rs1 - pseudo-code alias for rd = 1.0 / rs1
  61 * FATAN - pseudo-code alias for rd = atan2(rs1, 1.0) - FATAN2
  62 * FATANPI - pseudo alias for rd = atan2pi(rs1, 1.0) - FATAN2PI
  63 * FSINCOS - fused macro-op between FSIN and FCOS (issued in that order).
  64 * FSINCOSPI - fused macro-op between FSINPI and FCOSPI (issued in that order).
  65
  66 FATANPI example pseudo-code:
  67
  68     lui t0, 0x3F800 // upper bits of f32 1.0
  69     fmv.x.s ft0, t0
  70     fatan2pi.s rd, rs1, ft0
  71