openpower/sv/shape_table_format.mdwn

   1 Shape is 32-bits.  When SHAPE is set entirely to zeros, remapping is
   2 disabled: the register's elements are a linear (1D) vector.
   3
   4 | 31..30   | 29..24 | 23..21  | 20..18  | 17..12  | 11..6   | 5..0    |
   5 | -------- | ------ | ------- | ------- | ------- | ------- | ------- |
   6 | mode     | offset | invxyz  | permute | zdimsz  | ydimsz  | xdimsz  |
   7
   8 mode sets different behaviours (straight matrix multiply, FFT, DCT).
   9 matrix multiply sets mode=0b00
  10
  11 invxyz will invert the start index of each of x, y or z. If invxyz[0] is
  12 zero then x-dimensional counting begins from 0 and increments, otherwise
  13 it begins from xdimsz-1 and iterates down to zero. Likewise for y and z.
  14
  15 offset will have the effect equivalent to the sequential element loop
  16 to appear to run for offset (additional) iterations prior to actually
  17 generating output.  in pseudo-code the loop would be:
  18
  19     for index in offset to (offset+VL-1)
  20
  21 xdimsz, ydimsz and zdimsz are offset by 1, such that a value of 0 indicates
  22 that the array dimensionality for that dimension is 1. any dimension
  23 not intended to be used must have its value set to 0 (dimensionality
  24 of 1).  A value of xdimsz=2 would indicate that in the first dimension
  25 there are 3 elements in the array.  For example, to create a 2D array
  26 X,Y of dimensionality X=3 and Y=2, set xdimsz=2, ydimsz=1 and zdimsz=0
  27
  28 The format of the array is therefore as follows:
  29
  30     array[xdimsz+1][ydimsz+1][zdimsz+1]
  31
  32 However whilst illustrative of the dimensionality, that does not take the
  33 "permute" setting into account.  "permute" may be any one of six values
  34 (0-5, with values of 6 and 7 being reserved, and not legal).  The table
  35 below shows how the permutation dimensionality order works:
  36
  37 | permute | order | array format             |
  38 | ------- | ----- | ------------------------ |
  39 | 000     | 0,1,2 | (xdim+1)(ydim+1)(zdim+1) |
  40 | 001     | 0,2,1 | (xdim+1)(zdim+1)(ydim+1) |
  41 | 010     | 1,0,2 | (ydim+1)(xdim+1)(zdim+1) |
  42 | 011     | 1,2,0 | (ydim+1)(zdim+1)(xdim+1) |
  43 | 100     | 2,0,1 | (zdim+1)(xdim+1)(ydim+1) |
  44 | 101     | 2,1,0 | (zdim+1)(ydim+1)(xdim+1) |
  45
  46 In other words, the "permute" option changes the order in which
  47 nested for-loops over the array would be done.  See executable
  48 python reference code for further details.
  49