openpower/sv/tern_bin.mdwn

   1 ## GPR Ternary Logic Immediate
   2
   3 Add this section to Book I 3.3.13
   4
   5 TLI-form
   6
   7 | 0-5 | 6-10 | 11-15 | 16-20 | 21-28 | 29-30 | 31 | Form     |
   8 |-----|------|-------|-------|-------|-------|----|----------|
   9 | PO  | RT   |  RA   |  RB   | TLI   | XO    | Rc | TLI-Form |
  10
  11 * `ternlogi RT, RA, RB, TLI` (`Rc=0`)
  12 * `ternlogi. RT, RA, RB, TLI` (`Rc=1`)
  13
  14 Pseudocode:
  15
  16 ```
  17     result <- (~RT & ~RA & ~RB & TLI[0]*64) | # 64 copies of TLI[0]
  18               (~RT & ~RA &  RB & TLI[1]*64) | # ...
  19               (~RT &  RA & ~RB & TLI[2]*64) |
  20               (~RT &  RA &  RB & TLI[3]*64) |
  21               ( RT & ~RA & ~RB & TLI[4]*64) |
  22               ( RT & ~RA &  RB & TLI[5]*64) |
  23               ( RT &  RA & ~RB & TLI[6]*64) | # ...
  24               ( RT &  RA &  RB & TLI[7]*64)   # 64 copies of TLI[7]
  25     RT <- result
  26 ```
  27
  28 For each integer value i, 0 to 63, do the following.
  29
  30 ```
  31     Let j be the value of the concatenation of the
  32     contents of bit i of RT, bit i of RB, bit i of RT.
  33     The value of bit j of TLI is placed into bit i of RT.
  34
  35     See Table 145, "xxeval(A, B, C, TLI) Equivalent
  36     Functions," on page 968 for the equivalent function
  37     evaluated by this instruction for any given value of TLI.
  38 ```
  39
  40 *Programmer's Note: this is a Read-Modify-Write instruction on RT.
  41 A simple copy instruction may be used to achieve the effect of
  42 3-in 1-out. The copy instruction should come immediately before
  43 `ternlogi` so that hardware may optionally Macro-Op Fuse them*
  44
  45 *Programmer's note: This instruction is useful when combined with Matrix REMAP
  46 in "Inner Product" Mode, creating Warshall Transitive Closure that has many
  47 applications in Computer Science.*
  48
  49 Special registers altered:
  50
  51 ```
  52     CR0                    (if Rc=1)
  53 ```
  54
  55 ----------
  56
  57 \newpage{}
  58
  59 ## Condition Register Ternary Logic Immediate
  60
  61 Add this section to Book I 2.5.1
  62
  63 CRB-form
  64
  65 | 0.5|6.8 |9.10|11.13|14.15|16.18|19.25|26.30| 31| Form     |
  66 |----|----|----|-----|-----|-----|-----|-----|---|----------|
  67 | PO | BF | msk|BFA  | msk | BFB | TLI | XO  |TLI| CRB-Form |
  68
  69 * `crternlogi BF, BFA, BFB, BFC, TLI, msk`
  70
  71 Pseudocode:
  72
  73 ```
  74     a <- CR[4*BF+32:4*BF+35]
  75     b <- CR[4*BFA+32:4*BFA+35]
  76     c <- CR[4*BFB+32:4*BFB+35]
  77     ternary <- (~a & ~b & ~c & TLI[0]*4) | # 4 copies of TLI[0]
  78                (~a & ~b &  c & TLI[1]*4) | # 4 copies of TLI[1]
  79                (~a &  b & ~c & TLI[2]*4) | # ...
  80                (~a &  b &  c & TLI[3]*4) |
  81                ( a & ~b & ~c & TLI[4]*4) |
  82                ( a & ~b &  c & TLI[5]*4) |
  83                ( a &  b & ~c & TLI[6]*4) | # ...
  84                ( a &  b &  c & TLI[7]*4))  # 4 copies of TLI[7]
  85     do i = 0 to 3
  86         if msk[i] = 1 then
  87             CR[4*BF+32+i] <- ternary[i]
  88 ```
  89
  90 For each integer value i, 0 to 3, do the following.
  91
  92 ```
  93     Let j be the value of the concatenation of the
  94     contents of bit i of CR Field BF, bit i of CR Field BFA,
  95     bit i of CR Field BFB.
  96
  97     If bit i of msk is set to 1 then the value of bit j of TLI
  98     is placed into bit i of CR Field BF.
  99
 100     Otherwise, if bit i of msk is a zero then bit i of
 101     CR Field BF is unchanged.
 102
 103     See Table 145, "xxeval(A, B, C, TLI) Equivalent
 104     Functions," on page 968 for the equivalent function
 105     evaluated by this instruction for any given value of TLI.
 106 ```
 107
 108 If `msk` is zero an Illegal Instruction trap is raised.
 109
 110 *Programmer's Note: this instruction is a "masked" overwrite on CR Field
 111 BF. For each bit set in msk a Write is performed but for each bit clear
 112 in msk the corresponding bit of BF is preserved. Overall this makes
 113 crbinlog a conditionally Read-Modify-Write instruction on CR Field BF.
 114 A simple copy instruction may be used to achieve the effect of
 115 3-in 1-out. The copy instruction should come immediately before
 116 `crternlogi` so that hardware may optionally Macro-Op Fuse them*
 117
 118 Special registers altered:
 119
 120 ```
 121     CR field BF
 122 ```
 123
 124 ----------
 125
 126 \newpage{}
 127
 128 ## GPR Dynamic Binary Logic
 129
 130 Add this section to Book I 3.3.13
 131
 132 VA-form
 133
 134 | 0-5 | 6-10 | 11-15 | 16-20 | 21-25 | 26 | 27-31 | Form    |
 135 |-----|------|-------|-------|-------|----|-------|---------|
 136 | PO  |  RT  |  RA   |  RB   |  RC   | nh | XO    | VA-Form |
 137
 138 * `binlog RT, RA, RB, RC, nh`
 139
 140 Pseudocode:
 141
 142 ```
 143     if nh = 1 then lut <- (RC)[56:59]
 144     else           lut <- (RC)[60:63]
 145     result <- (~RA & ~RB & lut[0]*64) |
 146               (~RA &  RB & lut[1]*64) |
 147               ( RA & ~RB & lut[2]*64) |
 148               ( RA &  RB & lut[3]*64))
 149     RT <- result
 150 ```
 151
 152 For each integer value i, 0 to 63, do the following.
 153
 154 ```
 155     If nh contains a 0, let lut be the four LSBs of RC
 156     (bits 60 to 63).  Otherwise let lut be the next
 157     four LSBs of RC (bits 56 to 59).
 158
 159     Let j be the value of the concatenation of the
 160     contents of bit i of RT with bit i of RB.
 161
 162     The value of bit j of lut is placed into bit i of RT.
 163 ```
 164
 165 Special registers altered:
 166
 167 ```
 168     None
 169 ```
 170
 171 **Programmer's Note**:
 172
 173 Dynamic (non-immediate-based) Ternary Logic, suitable for FPGA-style LUT3
 174 dynamic lookups and for JIT runtime acceleration, may be emulated by
 175 appropriate combination of `binlog` and `ternlogi`, using the `nh`
 176 (next half) operand to select first and second nibble:
 177
 178 ```
 179     # compute r3 = ternlog(r4, r5, r6, table=r7)
 180     # compute the values for when r6[i] = 0:
 181     binlog r3, r4, r5, r7, 0  # takes look-up-table from LSB 4 bits
 182     # compute the values for when r6[i] = 1:
 183     binlog r4, r4, r5, r7, 1  # takes look-up-table from second-to-LSB 4 bits
 184     # mux the two results together: r3 = (r3 & ~r6) | (r4 & r6)
 185     ternlogi r3, r4, r6, 0b11011000
 186 ```
 187
 188 ----------
 189
 190 \newpage{}
 191
 192 ## Condition Register Field Dynamic Binary Logic
 193
 194 Add this section to Book I 2.5.1
 195
 196 CRB-form
 197
 198 | 0.5|6.8 |9.10|11.13|14.15|16.18|19.25|26.30| 31| Form     |
 199 |----|----|----|-----|-----|-----|-----|-----|---|----------|
 200 | PO | BF | msk|BFA  | msk | BFB | //  | XO  |// | CRB-Form |
 201
 202 * `crbinlog BF, BFA, BFB, msk`
 203
 204 Pseudocode:
 205
 206 ```
 207     a <- CR[4*BF+32:4*BFA+35]
 208     b <- CR[4*BFA+32:4*BFA+35]
 209     lut <- CR[4*BFB+32:4*BFB+35]
 210     binary <- (~a & ~b & lut[0]*4) |
 211               (~a &  b & lut[1]*4) |
 212               ( a & ~b & lut[2]*4) |
 213               ( a &  b & lut[3]*4))
 214     do i = 0 to 3
 215         if msk[i] = 1 then
 216             CR[4*BF+32+i] <- binary[i]
 217 ```
 218
 219 For each integer value i, 0 to 3, do the following.
 220
 221 ```
 222     Let j be the value of the concatenation of the
 223     contents of bit i of CR Field BF with bit i of CR Field BFA.
 224
 225     If bit i of msk is set to 1 then the value of bit j of
 226     CR Field BFB is placed into bit i of CR Field BF.
 227
 228     Otherwise, if bit i of msk is a zero then bit i of
 229     CR Field BF is unchanged.
 230 ```
 231
 232 If `msk` is zero an Illegal Instruction trap is raised.
 233
 234 Special registers altered:
 235
 236 ```
 237     CR field BF
 238 ```
 239
 240 *Programmer's Note: just as with binlut and ternlogi, a pair
 241  of crbinlog instructions followed by a merging crternlogi may
 242  be deployed to synthesise dynamic ternary (LUT3) CR Field
 243  manipulation*
 244
 245 *Programmer's Note: this instruction is a "masked" overwrite on CR
 246 Field BF. For each bit set in `msk` a Write is performed
 247 but for each bit clear in `msk` the corresponding bit of BF is
 248 preserved. Overall this makes `crbinlog` a conditionally
 249 Read-Modify-Write instruction on CR Field BF.
 250 A simple copy instruction may be used to achieve the effect of
 251 3-in 1-out. The copy instruction should come immediately before
 252 `crternlogi` so that hardware may optionally Macro-Op Fuse them*
 253
 254 [[!tag standards]]
 255
 256 ----------
 257
 258 \newpage{}
 259