1 # Rewrite of SVP64 for OpenPower ISA v3.1
5 The plan is to create an encoding for SVP64, then to create an encoding for
6 SVP48, then to reorganize them both to improve field overlap, reducing the
7 amount of decoder hardware necessary.
9 All bit numbers are in MSB0 form (the bits are numbered from 0 at the MSB and
10 counting up as you move to the LSB end). All bit ranges are inclusive (so
11 `4:6` means bits 4, 5, and 6).
13 64-bit instructions are split into two 32-bit words, the prefix and the suffix. The prefix always comes before the suffix in PC order.
15 ## Definition of Reserved in this spec.
17 For the new fields added in SVP64, instructions that have any of their fields set to a reserved value must cause an illegal instruction trap, to allow emulation of future instruction sets.
19 This is unlike OpenPower ISA v3.1, which doesn't require a CPU to trap.
21 ## Remapped Encoding (`RM[0:23]`)
23 To allow relatively easy remapping of which portions of the Prefix Opcode Map
24 are used for SVP64 without needing to rewrite a large portion of the SVP64
25 spec, a mapping is defined from the OpenPower v3.1 prefix bits to a new 24-bit
26 Remapped Encoding denoted `RM[0]` at the MSB to `RM[23]` at the LSB.
28 The mapping from the OpenPower v3.1 prefix bits to the Remapped Encoding is
29 defined in the Prefix Fields section.
31 ## Remapped Encoding Fields
33 Shows all fields in the Remapped Encoding `RM[0:23]` for all instruction variants. There are two categories: Single and Twin Predication. Due to space considerations further subdivision of Single Predication is based on whether the number of src operands is 2 or 3.
35 ### Single Predication (N(src) > 1)
38 | Field Name | Field bits | Description |
39 |------------|------------|------------------------------------------------|
40 | MASK_KIND | `0` | Execution Mask Kind |
41 | MASK | `1:3` | Execution Mask |
42 | ELWIDTH | `4:5` | Element Width |
43 | SUBVL | `6:7` | Sub-vector length |
44 | EXTRA | `8:16` | Extra fields qualifying registers |
45 | MODE | `19:23` | see [[discussion]] |
47 Extra2: applies to 4-operand instructions (fmadd)
49 | Field Name | Field bits | Description |
50 |--------------|---------|--------------------------------------------------|
51 | Rdest_EXTRA2 | `8:9` | extra bits for Rdest (R\*_EXTRA2 Encoding) |
52 | Rsrc1_EXTRA2 | `10:11` | extra bits for Rsrc1 (R\*_EXTRA2 Encoding) |
53 | Rsrc2_EXTRA2 | `12:13` | extra bits for Rsrc2 (R\*_EXTRA2 Encoding) |
54 | Rsrc3_EXTRA2 | `14:15` | extra bits for Rsrc3 (R\*_EXTRA2 Encoding|
55 | reserved | `16` | reserved |
57 Extra3: applies to 3-operand instructions (src1 src2 dest)
60 | Field Name | Field bits | Description |
61 |--------------|---------|--------------------------------------------------|
62 | Rdest_EXTRA3 | `8:10` | extra bits for Rdest (Uses R\*_EXTRA3 Encoding) |
63 | Rsrc1_EXTRA3 | `11:13` | extra bits for Rsrc1 (Uses R\*_EXTRA3 Encoding) |
64 | Rsrc2_EXTRA3 | `14:16` | extra bits for Rsrc3 (Uses R\*_EXTRA3 Encoding) |
66 ### Twin Predication (src=1, dest=1)
68 | Remapped Encoding Field Name | Field bits | Description |
69 |------------------------------|------------|---------------------------------------------------------------------------|
70 | MASK_KIND | `0` | Execution Mask Kind |
71 | MASK | `1:3` | Execution Mask |
72 | ELWIDTH | `4:5` | Element Width |
73 | SUBVL | `6:7` | Sub-vector length |
74 | Rdest_EXTRA3 | `8:10` | extra bits for Rdest (Uses R\*_EXTRA Encoding) |
75 | Rsrc1_EXTRA3 | `11:13` | extra bits for Rsrc1 (Uses R\*_EXTRA Encoding) |
76 | MASK_SRC | `14:16` | Execution Mask for Source (only on instructions with twin-predication) |
77 | ELWIDTH_SRC | `17:18` | Element Width for Source (only on instructions with twin-predication) |
78 | MODE | `19:23` | see [[discussion]] |
80 note in [[discussion]]: TODO, evaluate if 2nd SUBVL should be added. conclusion: no. 2nd SUBVL makes no sense except for mv, and that is covered by [[mv.vec]]
82 ## R\*_EXTRA2 and R\*_EXTRA3 Encoding
84 (**TODO: 2-bit version of the table, just like in the original SVPrefix. This is important, to save bits on 4-operand instructions such as fmadd**)
86 In the following table, `<N>` denotes the value of the corresponding register field in the SVP64 suffix word.
88 (**Jacob: these tables are not in the slightest bit understandable due to the use of register names that are impossible to interpret clearly**)
92 | R\*_EXTRA3 | Vector/Scalar<br/>Mode | CR Register | Int/FP<br/>Register |
93 |-----------|------------------------|---------------|---------------------|
94 | 000 | Scalar | `SVCR<N>_000` | `SV[F]R<N>_00` |
95 | 001 | Scalar | `SVCR<N>_010` | `SV[F]R<N>_01` |
96 | 010 | Scalar | `SVCR<N>_100` | `SV[F]R<N>_10` |
97 | 011 | Scalar | `SVCR<N>_110` | `SV[F]R<N>_11` |
98 | 100 | Vector | `SVCR<N>_000` | `SV[F]R<N>_00` |
99 | 101 | Vector | `SVCR<N>_010` | `SV[F]R<N>_01` |
100 | 110 | Vector | `SVCR<N>_100` | `SV[F]R<N>_10` |
101 | 111 | Vector | `SVCR<N>_110` | `SV[F]R<N>_11` |
105 (**TODO, i simply cannot interpret the names, they have absolutely zero meaning to me so i have no idea how to fill in the table. this is a bad sign, indicative that the names have to go, to be replaced by something xlear snd obvious**)
107 | R\*_EXTRA2 | Vector/Scalar<br/>Mode | CR Register | Int/FP<br/>Register |
108 |-----------|------------------------|---------------|---------------------|
109 | 00 | Scalar | `SVCR<N>_000` | `SV[F]R<N>_00` |
110 | 01 | Scalar | `SVCR<N>_100` | `SV[F]R<N>_10` |
111 | 10 | Vector | `SVCR<N>_000` | `SV[F]R<N>_00` |
112 | 11 | Vector | `SVCR<N>_100` | `SV[F]R<N>_10` |
114 | R\*_EXTRA2 | Mode | CR Register | Int/FP<br/>Register |
115 |-----------|-------|---------------|---------------------|
116 | 00 | Scalar | `` | `RA` |
117 | 01 | Scalar | `` | `RA || 0b00` |
118 | 10 | Vector | `` | `RA || 0b00` |
119 | 11 | Vector | `` | `RA || 0b10` |
123 | Instruction Kind | ELWIDTH Value | Mnemonic | Description |
124 |------------------|---------------|---------------------------|-------------------------------------------------------------------------------------|
125 | Integer | 00 | `ELWIDTH=b` | Byte: 8-bit integer |
126 | Integer | 01 | `ELWIDTH=h` | Halfword: 16-bit integer |
127 | Integer | 10 | `ELWIDTH=w` | Word: 32-bit integer |
128 | Integer | 11 | `ELWIDTH=d` | Doubleword: 64-bit integer |
129 | FP | 00 | `ELWIDTH=bf16` (Reserved) | Reserved for [`bf16`](https://en.wikipedia.org/wiki/Bfloat16_floating-point_format) |
130 | FP | 01 | `ELWIDTH=f16` | 16-bit IEEE 754 Half floating-point |
131 | FP | 10 | `ELWIDTH=f32` | 32-bit IEEE 754 Single floating-point |
132 | FP | 11 | `ELWIDTH=f64` | 64-bit IEEE 754 Double floating-point |
136 | SUBVL Value | Mnemonic | Description |
137 |-------------|---------------------|------------------------|
138 | 00 | `SUBVL=4` | Sub-vector length of 4 |
139 | 01 | `SUBVL=1` (default) | Sub-vector length of 1 |
140 | 10 | `SUBVL=2` | Sub-vector length of 2 |
141 | 11 | `SUBVL=3` | Sub-vector length of 3 |
143 ## MASK/MASK_SRC & MASK_KIND Encoding
145 One bit (`MASKMODE`) indicates the mode: CR or Int predication. The two types may not be mixed.
147 | MASK_KIND Value | Description |
148 |-----------------|------------------------------------------------------|
149 | 0 | MASK/MASK_SRC are encoded using Integer Predication |
150 | 1 | MASK/MASK_SRC are encoded using CR-based Predication |
152 Integer Twin predication has a second set if 3 bits that uses the same encoding thus allowing either the same register (r3 or r10) to be used for both src and dest, or different regs (one for src, one for dest).
154 Likewise CR based twin predication has a second set of 3 bits, allowing a different test to be applied.
156 ### Integer Predication (MASK_KIND=0)
158 When the predicate mode bit is zero the 3 bits are interpreted as below.
159 Twin predication has an identical 3 bit field similarly encoded.
161 | MASK/MASK_SRC<br/>Value | Mnemonic | Description |
162 |-------------------------|----------|--------------------------------------------------------|
163 | 000 | ALWAYS | Operation is not masked (mask set to all 1s) |
164 | 001 | 1 << R3 | Element `i` is enabled if `i == R3` |
165 | 010 | R3 | Element `i` is enabled if `R3 & (1 << i)` is non-zero |
166 | 011 | ~R3 | Element `i` is enabled if `R3 & (1 << i)` is zero |
167 | 100 | R10 | Element `i` is enabled if `R10 & (1 << i)` is non-zero |
168 | 101 | ~R10 | Element `i` is enabled if `R10 & (1 << i)` is zero |
169 | 110 | R30 | Element `i` is enabled if `R30 & (1 << i)` is non-zero |
170 | 111 | ~R30 | Element `i` is enabled if `R30 & (1 << i)` is zero |
172 ### CR-based Predication (MASK_KIND=1)
174 When the predicate mode bit is one the 3 bits are interpreted as below. Twin predication has an identical 3 bit field similarly encoded
176 | MASK/MASK_SRC<br/>Value | Mnemonic | Description |
177 |-------------------------|----------|-------------------------------------------------|
178 | 000 | lt | Element `i` is enabled if `CR[6+i].LT` is set |
179 | 001 | nl/ge | Element `i` is enabled if `CR[6+i].LT` is clear |
180 | 010 | gt | Element `i` is enabled if `CR[6+i].GT` is set |
181 | 011 | ng/le | Element `i` is enabled if `CR[6+i].GT` is clear |
182 | 100 | eq | Element `i` is enabled if `CR[6+i].EQ` is set |
183 | 101 | ne | Element `i` is enabled if `CR[6+i].EQ` is clear |
184 | 110 | so/un | Element `i` is enabled if `CR[6+i].FU` is set |
185 | 111 | ns/nu | Element `i` is enabled if `CR[6+i].FU` is clear |
187 CR based predication. TODO: select alternate CR for twin predication? see [[discussion]] Overlap of the two CR based predicates must be taken into account, so the starting point for one of them must be suitably high, or accept that for twin predication VL must not exceed the range where overlap will occur, *or* that they use the same starting point but select different *bits* of the same CRs
190 ## Prefix Opcode Map (64-bit instruction encoding) (prefix bits 6:11)
192 (shows both PowerISA v3.1 instructions as well as new SVP instructions; empty spaces are yet-to-be-allocated Illegal Instructions)
194 | bits 6:11 | ---000 | ---001 | ---010 | ---011 | ---100 | ---101 | ---110 | ---111 |
195 |-----------|----------|------------|----------|----------|----------|----------|----------|----------|
196 | 000--- | 8LS-form | 8LS-form | 8LS-form | 8LS-form | 8LS-form | 8LS-form | 8LS-form | 8LS-form |
197 | 001--- | | | | | | | | |
198 | 010--- | 8RR-form | | | | SVP64 | SVP64 | SVP64 | SVP64 |
199 | 011--- | | | | | SVP64 | SVP64 | SVP64 | SVP64 |
200 | 100--- | MLS-form | MLS-form | MLS-form | MLS-form | MLS-form | MLS-form | MLS-form | MLS-form |
201 | 101--- | | | | | | | | |
202 | 110--- | MRR-form | | | | SVP64 | SVP64 | SVP64 | SVP64 |
203 | 111--- | | MMIRR-form | | | SVP64 | SVP64 | SVP64 | SVP64 |
207 | Prefix Field Name | Field bits | Constant Value | Description |
208 |---------------------|------------|----------------|--------------------------------------------|
209 | PO (Primary Opcode) | `0:5` | `1` | Indicates this is a 64-bit instruction |
210 | `RM[0]` | `6` | | Bit 0 of the Remapped Encoding |
211 | SVP64_7 | `7` | `1` | Indicates this is a SVP64 instruction |
212 | `RM[1]` | `8` | | Bit 1 of the Remapped Encoding |
213 | SVP64_9 | `9` | `1` | Indicates this is a SVP64 instruction |
214 | `RM[2:23]` | `10:31` | | Bits 2 through 23 of the Remapped Encoding |
218 This is a novel concept that allows predication to be applied to a single source and a single dest register. The following types of traditional Vector operations may be encoded with it, *without requiring explicit opcodes to do so*
220 * VSPLAT (a single scalar distributed across a vector)
221 * VEXTRACT (like LLVM IR [`extractelement`](https://releases.llvm.org/11.0.0/docs/LangRef.html#extractelement-instruction))
222 * VINSERT (like LLVM IR [`insertelement`](https://releases.llvm.org/11.0.0/docs/LangRef.html#insertelement-instruction))
223 * VCOMPRESS (like LLVM IR [`llvm.masked.compressstore.*`](https://releases.llvm.org/11.0.0/docs/LangRef.html#llvm-masked-compressstore-intrinsics))
224 * VEXPAND (like LLVM IR [`llvm.masked.expandload.*`](https://releases.llvm.org/11.0.0/docs/LangRef.html#llvm-masked-expandload-intrinsics))
226 Those patterns (and more) may be applied to:
228 * mv (the usual way that V\* operations are created)
229 * exts\* sign-extension
230 * rwlinm and other RS-RA shift operations
231 * LD and ST (treating AGEN as one source)
232 * FP fclass, fsgn, fneg, fabs, fcvt, frecip, fsqrt etc.
233 * Condition Register ops mfcr, mtcr and other similar
235 This is a huge list that creates extremely powerful combinations, particularly given that one of the predicate options is `(1<<r3)`
237 Additional unusual capabilities of Twin Predication include a back-to-back version of VCOMPRESS-VEXPAND which is effectively the ability to do an ordered multiple VINSERT.
241 There are two different encodings: single-predication (typically arithmetic operations i.e. with more than one source register) and twin-predication (one source, one destination). They require different encodings
245 SV Registers are numbered using the notation `SV[F|C]R<N>_<M>` where `<N>` is a decimal integer and `<M>` is a binary integer. Two integers are used to enable future register expansions to add more registers by appending more LSB bits to `<M>`.
247 For all `SV[F|C]R<N>_<M>` registers, the N is the
248 upper bits in decimal and the M is the lower bits in binary, so `SVR5_01` is
249 SV integer register `(5 << 2) + 0b01`, `SVCR6_011` is SV condition register
250 `(6 << 3) + 0b011`, and `SVFR20_10` is SV floating-point register
255 a vectorized 32-bit add:
257 add SVR3_01, SVR6_10, SVR10_00, elwidth=w, subvl=1, mask=lt
261 const size_t start_cr = (6 << 3) + 0b000; // starting at SVCR6_000
262 // pretend for the moment that type-punning actually works in C/C++
263 uint32_t *rt = (uint32_t *)®s[(3 << 2) + 0b01]; // SVR3_01
264 uint32_t *ra = (uint32_t *)®s[(6 << 2) + 0b10]; // SVR6_10
265 uint32_t *rb = (uint32_t *)®s[(10 << 2) + 0b00]; // SVR10_00
266 for(size_t i = 0; i < VL; i++) {
267 if(CRs[(start_cr + i) % 64].lt) {
268 rt[i] = ra[i] + rb[i];
275 add r20, r25, r30, elwidth=64, subvl=1
277 where `r20`, `r25`, and `r30` are standard OpenPower register names.
278 Those names correspond to `SVR20_00`, `SVR25_00`, and `SVR30_00`.
282 const size_t STD_TO_SV_SHIFT = 2; // gets bigger as reg files expand to 256, 512, ... registers
284 VL = 7; // setvli (omitting maxvl here)
286 for(size_t i = 0; i < VL; i++) {
287 regs[(20 << STD_TO_SV_SHIFT) + i] = regs[(25 << STD_TO_SV_SHIFT) + i]
288 + regs[(30 << STD_TO_SV_SHIFT) + i];
291 Standard PowerISA Integer registers are aliased to some of the SV integer registers:
293 (**Jacob these names are impossible to interpret due to them not being sequential numbering and there being no compact algorithm given that shows how they're created. the original SVPrefix was dead easy to understand**)
295 | Integer<br/>Register | SV Integer<br/>Register | Integer<br/>Register | SV Integer<br/>Register | Integer<br/>Register | SV Integer<br/>Register | Integer<br/>Register | SV Integer<br/>Register |
296 |----------------------|-------------------------|----------------------|-------------------------|----------------------|-------------------------|----------------------|-------------------------|
297 | R0 | SVR0_00 | R8 | SVR8_00 | R16 | SVR16_00 | R24 | SVR24_00 |
298 | | SVR0_01 | | SVR8_01 | | SVR16_01 | | SVR24_01 |
299 | | SVR0_10 | | SVR8_10 | | SVR16_10 | | SVR24_10 |
300 | | SVR0_11 | | SVR8_11 | | SVR16_11 | | SVR24_11 |
301 | R1 | SVR1_00 | R9 | SVR9_00 | R17 | SVR17_00 | R25 | SVR25_00 |
302 | | SVR1_01 | | SVR9_01 | | SVR17_01 | | SVR25_01 |
303 | | SVR1_10 | | SVR9_10 | | SVR17_10 | | SVR25_10 |
304 | | SVR1_11 | | SVR9_11 | | SVR17_11 | | SVR25_11 |
305 | R2 | SVR2_00 | R10 | SVR10_00 | R18 | SVR18_00 | R26 | SVR26_00 |
306 | | SVR2_01 | | SVR10_01 | | SVR18_01 | | SVR26_01 |
307 | | SVR2_10 | | SVR10_10 | | SVR18_10 | | SVR26_10 |
308 | | SVR2_11 | | SVR10_11 | | SVR18_11 | | SVR26_11 |
309 | R3 | SVR3_00 | R11 | SVR11_00 | R19 | SVR19_00 | R27 | SVR27_00 |
310 | | SVR3_01 | | SVR11_01 | | SVR19_01 | | SVR27_01 |
311 | | SVR3_10 | | SVR11_10 | | SVR19_10 | | SVR27_10 |
312 | | SVR3_11 | | SVR11_11 | | SVR19_11 | | SVR27_11 |
313 | R4 | SVR4_00 | R12 | SVR12_00 | R20 | SVR20_00 | R28 | SVR28_00 |
314 | | SVR4_01 | | SVR12_01 | | SVR20_01 | | SVR28_01 |
315 | | SVR4_10 | | SVR12_10 | | SVR20_10 | | SVR28_10 |
316 | | SVR4_11 | | SVR12_11 | | SVR20_11 | | SVR28_11 |
317 | R5 | SVR5_00 | R13 | SVR13_00 | R21 | SVR21_00 | R29 | SVR29_00 |
318 | | SVR5_01 | | SVR13_01 | | SVR21_01 | | SVR29_01 |
319 | | SVR5_10 | | SVR13_10 | | SVR21_10 | | SVR29_10 |
320 | | SVR5_11 | | SVR13_11 | | SVR21_11 | | SVR29_11 |
321 | R6 | SVR6_00 | R14 | SVR14_00 | R22 | SVR22_00 | R30 | SVR30_00 |
322 | | SVR6_01 | | SVR14_01 | | SVR22_01 | | SVR30_01 |
323 | | SVR6_10 | | SVR14_10 | | SVR22_10 | | SVR30_10 |
324 | | SVR6_11 | | SVR14_11 | | SVR22_11 | | SVR30_11 |
325 | R7 | SVR7_00 | R15 | SVR15_00 | R23 | SVR23_00 | R31 | SVR31_00 |
326 | | SVR7_01 | | SVR15_01 | | SVR23_01 | | SVR31_01 |
327 | | SVR7_10 | | SVR15_10 | | SVR23_10 | | SVR31_10 |
328 | | SVR7_11 | | SVR15_11 | | SVR23_11 | | SVR31_11 |
330 ## Floating-Point Registers
332 Standard PowerISA floating-point and VSX registers are aliased to some of the SV floating-point registers:
334 (**Jacob these names are impossible to interpret due to them not being sequential numbering and there being no compact algorithm given that shows how they're created. the original SVPrefix was dead easy to understand**)
336 | FP<br/>Register | VSX Register | SV FP<br/>Register | FP<br/>Register | VSX Register | SV FP<br/>Register |
337 |-----------------|-----------------------|--------------------|-----------------|-----------------------|--------------------|
338 | FPR\[0\] | VSR\[0\]\.dword\[0\] | SVFR0\_00 | FPR\[16\] | VSR\[16\]\.dword\[0\] | SVFR16\_00 |
339 | | VSR\[0\]\.dword\[1\] | SVFR0\_01 | | VSR\[16\]\.dword\[1\] | SVFR16\_01 |
340 | | VSR\[32\]\.dword\[0\] | SVFR0\_10 | | VSR\[48\]\.dword\[0\] | SVFR16\_10 |
341 | | VSR\[32\]\.dword\[1\] | SVFR0\_11 | | VSR\[48\]\.dword\[1\] | SVFR16\_11 |
342 | FPR\[1\] | VSR\[1\]\.dword\[0\] | SVFR1\_00 | FPR\[17\] | VSR\[17\]\.dword\[0\] | SVFR17\_00 |
343 | | VSR\[1\]\.dword\[1\] | SVFR1\_01 | | VSR\[17\]\.dword\[1\] | SVFR17\_01 |
344 | | VSR\[33\]\.dword\[0\] | SVFR1\_10 | | VSR\[49\]\.dword\[0\] | SVFR17\_10 |
345 | | VSR\[33\]\.dword\[1\] | SVFR1\_11 | | VSR\[49\]\.dword\[1\] | SVFR17\_11 |
346 | FPR\[2\] | VSR\[2\]\.dword\[0\] | SVFR2\_00 | FPR\[18\] | VSR\[18\]\.dword\[0\] | SVFR18\_00 |
347 | | VSR\[2\]\.dword\[1\] | SVFR2\_01 | | VSR\[18\]\.dword\[1\] | SVFR18\_01 |
348 | | VSR\[34\]\.dword\[0\] | SVFR2\_10 | | VSR\[50\]\.dword\[0\] | SVFR18\_10 |
349 | | VSR\[34\]\.dword\[1\] | SVFR2\_11 | | VSR\[50\]\.dword\[1\] | SVFR18\_11 |
350 | FPR\[3\] | VSR\[3\]\.dword\[0\] | SVFR3\_00 | FPR\[19\] | VSR\[19\]\.dword\[0\] | SVFR19\_00 |
351 | | VSR\[3\]\.dword\[1\] | SVFR3\_01 | | VSR\[19\]\.dword\[1\] | SVFR19\_01 |
352 | | VSR\[35\]\.dword\[0\] | SVFR3\_10 | | VSR\[51\]\.dword\[0\] | SVFR19\_10 |
353 | | VSR\[35\]\.dword\[1\] | SVFR3\_11 | | VSR\[51\]\.dword\[1\] | SVFR19\_11 |
354 | FPR\[4\] | VSR\[4\]\.dword\[0\] | SVFR4\_00 | FPR\[20\] | VSR\[20\]\.dword\[0\] | SVFR20\_00 |
355 | | VSR\[4\]\.dword\[1\] | SVFR4\_01 | | VSR\[20\]\.dword\[1\] | SVFR20\_01 |
356 | | VSR\[36\]\.dword\[0\] | SVFR4\_10 | | VSR\[52\]\.dword\[0\] | SVFR20\_10 |
357 | | VSR\[36\]\.dword\[1\] | SVFR4\_11 | | VSR\[52\]\.dword\[1\] | SVFR20\_11 |
358 | FPR\[5\] | VSR\[5\]\.dword\[0\] | SVFR5\_00 | FPR\[21\] | VSR\[21\]\.dword\[0\] | SVFR21\_00 |
359 | | VSR\[5\]\.dword\[1\] | SVFR5\_01 | | VSR\[21\]\.dword\[1\] | SVFR21\_01 |
360 | | VSR\[37\]\.dword\[0\] | SVFR5\_10 | | VSR\[53\]\.dword\[0\] | SVFR21\_10 |
361 | | VSR\[37\]\.dword\[1\] | SVFR5\_11 | | VSR\[53\]\.dword\[1\] | SVFR21\_11 |
362 | FPR\[6\] | VSR\[6\]\.dword\[0\] | SVFR6\_00 | FPR\[22\] | VSR\[22\]\.dword\[0\] | SVFR22\_00 |
363 | | VSR\[6\]\.dword\[1\] | SVFR6\_01 | | VSR\[22\]\.dword\[1\] | SVFR22\_01 |
364 | | VSR\[38\]\.dword\[0\] | SVFR6\_10 | | VSR\[54\]\.dword\[0\] | SVFR22\_10 |
365 | | VSR\[38\]\.dword\[1\] | SVFR6\_11 | | VSR\[54\]\.dword\[1\] | SVFR22\_11 |
366 | FPR\[7\] | VSR\[7\]\.dword\[0\] | SVFR7\_00 | FPR\[23\] | VSR\[23\]\.dword\[0\] | SVFR23\_00 |
367 | | VSR\[7\]\.dword\[1\] | SVFR7\_01 | | VSR\[23\]\.dword\[1\] | SVFR23\_01 |
368 | | VSR\[39\]\.dword\[0\] | SVFR7\_10 | | VSR\[55\]\.dword\[0\] | SVFR23\_10 |
369 | | VSR\[39\]\.dword\[1\] | SVFR7\_11 | | VSR\[55\]\.dword\[1\] | SVFR23\_11 |
370 | FPR\[8\] | VSR\[8\]\.dword\[0\] | SVFR8\_00 | FPR\[24\] | VSR\[24\]\.dword\[0\] | SVFR24\_00 |
371 | | VSR\[8\]\.dword\[1\] | SVFR8\_01 | | VSR\[24\]\.dword\[1\] | SVFR24\_01 |
372 | | VSR\[40\]\.dword\[0\] | SVFR8\_10 | | VSR\[56\]\.dword\[0\] | SVFR24\_10 |
373 | | VSR\[40\]\.dword\[1\] | SVFR8\_11 | | VSR\[56\]\.dword\[1\] | SVFR24\_11 |
374 | FPR\[9\] | VSR\[9\]\.dword\[0\] | SVFR9\_00 | FPR\[25\] | VSR\[25\]\.dword\[0\] | SVFR25\_00 |
375 | | VSR\[9\]\.dword\[1\] | SVFR9\_01 | | VSR\[25\]\.dword\[1\] | SVFR25\_01 |
376 | | VSR\[41\]\.dword\[0\] | SVFR9\_10 | | VSR\[57\]\.dword\[0\] | SVFR25\_10 |
377 | | VSR\[41\]\.dword\[1\] | SVFR9\_11 | | VSR\[57\]\.dword\[1\] | SVFR25\_11 |
378 | FPR\[10\] | VSR\[10\]\.dword\[0\] | SVFR10\_00 | FPR\[26\] | VSR\[26\]\.dword\[0\] | SVFR26\_00 |
379 | | VSR\[10\]\.dword\[1\] | SVFR10\_01 | | VSR\[26\]\.dword\[1\] | SVFR26\_01 |
380 | | VSR\[42\]\.dword\[0\] | SVFR10\_10 | | VSR\[58\]\.dword\[0\] | SVFR26\_10 |
381 | | VSR\[42\]\.dword\[1\] | SVFR10\_11 | | VSR\[58\]\.dword\[1\] | SVFR26\_11 |
382 | FPR\[11\] | VSR\[11\]\.dword\[0\] | SVFR11\_00 | FPR\[27\] | VSR\[27\]\.dword\[0\] | SVFR27\_00 |
383 | | VSR\[11\]\.dword\[1\] | SVFR11\_01 | | VSR\[27\]\.dword\[1\] | SVFR27\_01 |
384 | | VSR\[43\]\.dword\[0\] | SVFR11\_10 | | VSR\[59\]\.dword\[0\] | SVFR27\_10 |
385 | | VSR\[43\]\.dword\[1\] | SVFR11\_11 | | VSR\[59\]\.dword\[1\] | SVFR27\_11 |
386 | FPR\[12\] | VSR\[12\]\.dword\[0\] | SVFR12\_00 | FPR\[28\] | VSR\[28\]\.dword\[0\] | SVFR28\_00 |
387 | | VSR\[12\]\.dword\[1\] | SVFR12\_01 | | VSR\[28\]\.dword\[1\] | SVFR28\_01 |
388 | | VSR\[44\]\.dword\[0\] | SVFR12\_10 | | VSR\[60\]\.dword\[0\] | SVFR28\_10 |
389 | | VSR\[44\]\.dword\[1\] | SVFR12\_11 | | VSR\[60\]\.dword\[1\] | SVFR28\_11 |
390 | FPR\[13\] | VSR\[13\]\.dword\[0\] | SVFR13\_00 | FPR\[29\] | VSR\[29\]\.dword\[0\] | SVFR29\_00 |
391 | | VSR\[13\]\.dword\[1\] | SVFR13\_01 | | VSR\[29\]\.dword\[1\] | SVFR29\_01 |
392 | | VSR\[45\]\.dword\[0\] | SVFR13\_10 | | VSR\[61\]\.dword\[0\] | SVFR29\_10 |
393 | | VSR\[45\]\.dword\[1\] | SVFR13\_11 | | VSR\[61\]\.dword\[1\] | SVFR29\_11 |
394 | FPR\[14\] | VSR\[14\]\.dword\[0\] | SVFR14\_00 | FPR\[30\] | VSR\[30\]\.dword\[0\] | SVFR30\_00 |
395 | | VSR\[14\]\.dword\[1\] | SVFR14\_01 | | VSR\[30\]\.dword\[1\] | SVFR30\_01 |
396 | | VSR\[46\]\.dword\[0\] | SVFR14\_10 | | VSR\[62\]\.dword\[0\] | SVFR30\_10 |
397 | | VSR\[46\]\.dword\[1\] | SVFR14\_11 | | VSR\[62\]\.dword\[1\] | SVFR30\_11 |
398 | FPR\[15\] | VSR\[15\]\.dword\[0\] | SVFR15\_00 | FPR\[31\] | VSR\[31\]\.dword\[0\] | SVFR31\_00 |
399 | | VSR\[15\]\.dword\[1\] | SVFR15\_01 | | VSR\[31\]\.dword\[1\] | SVFR31\_01 |
400 | | VSR\[47\]\.dword\[0\] | SVFR15\_10 | | VSR\[63\]\.dword\[0\] | SVFR31\_10 |
401 | | VSR\[47\]\.dword\[1\] | SVFR15\_11 | | VSR\[63\]\.dword\[1\] | SVFR31\_11 |
405 ## CR fields as inputs/outputs of vector operations
407 When vectorized, the CR inputs/outputs are read/written to 4-bit CR fields
408 starting from SVCR6_000 and incrementing from there. If SVCR7_111 is reached, the next CR
409 field used wraps around to SVCR0_000, then incrementing from there.
410 (see [[discussion]]. some alternative schemes are described there)
412 SVCR6_000 was chosen to balance avoiding needing to save CR2-CR4 (which are
413 callee-saved) just to use SV vectors with VL <= 61 as well as having the first
414 vector CR field readily accessible to standard CR instructions and branches.
415 Additionally, SVCR6_000 is used as the implicit result of a OpenPower ISA v3.1
416 standard vector (SIMD) instruction with Rc=1.
418 ## Table of CR fields
420 CR[i] is the notation used by the OpenPower spec to refer to CR field #i,
421 so FP instructions with Rc=1 write to CR[1] aka SVCR1_000.
423 There are 3 new SPRs for holding CRs: CR_EXT1, CR_EXT2, and CR_EXT3.
425 The 64 SV CRs are arranged similarly to the way the 128 integer registers are arranged:
427 (**Jacob these names are impossible to interpret due to them not being sequential numbering and there being no compact algorithm given that shows how they're created. the original SVPrefix was dead easy to understand**)
429 | CR<br/>Register | SPR<br/>Field | SV CR<br/>Register | CR<br/>Register | SPR<br/>Field | SV CR<br/>Register |
430 |-----------------|----------------|--------------------|-----------------|----------------|--------------------|
431 | CR[0] | CR[32:35] | SVCR0_000 | CR[4] | CR[48:51] | SVCR4_000 |
432 | | CR_EXT1[32:35] | SVCR0_001 | | CR_EXT1[48:51] | SVCR4_001 |
433 | | CR_EXT2[32:35] | SVCR0_010 | | CR_EXT2[48:51] | SVCR4_010 |
434 | | CR_EXT3[32:35] | SVCR0_011 | | CR_EXT3[48:51] | SVCR4_011 |
435 | *CR[-8]* | CR[0:3] | SVCR0_100 | *CR[-4]* | CR[16:19] | SVCR4_100 |
436 | | CR_EXT1[0:3] | SVCR0_101 | | CR_EXT1[16:19] | SVCR4_101 |
437 | | CR_EXT2[0:3] | SVCR0_110 | | CR_EXT2[16:19] | SVCR4_110 |
438 | | CR_EXT3[0:3] | SVCR0_111 | | CR_EXT3[16:19] | SVCR4_111 |
439 | CR[1] | CR[36:39] | SVCR1_000 | CR[5] | CR[52:55] | SVCR5_000 |
440 | | CR_EXT1[36:39] | SVCR1_001 | | CR_EXT1[52:55] | SVCR5_001 |
441 | | CR_EXT2[36:39] | SVCR1_010 | | CR_EXT2[52:55] | SVCR5_010 |
442 | | CR_EXT3[36:39] | SVCR1_011 | | CR_EXT3[52:55] | SVCR5_011 |
443 | *CR[-7]* | CR[4:7] | SVCR1_100 | *CR[-3]* | CR[20:23] | SVCR5_100 |
444 | | CR_EXT1[4:7] | SVCR1_101 | | CR_EXT1[20:23] | SVCR5_101 |
445 | | CR_EXT2[4:7] | SVCR1_110 | | CR_EXT2[20:23] | SVCR5_110 |
446 | | CR_EXT3[4:7] | SVCR1_111 | | CR_EXT3[20:23] | SVCR5_111 |
447 | CR[2] | CR[40:43] | SVCR2_000 | CR[6] | CR[56:59] | SVCR6_000 |
448 | | CR_EXT1[40:43] | SVCR2_001 | | CR_EXT1[56:59] | SVCR6_001 |
449 | | CR_EXT2[40:43] | SVCR2_010 | | CR_EXT2[56:59] | SVCR6_010 |
450 | | CR_EXT3[40:43] | SVCR2_011 | | CR_EXT3[56:59] | SVCR6_011 |
451 | *CR[-6]* | CR[8:11] | SVCR2_100 | *CR[-2]* | CR[24:27] | SVCR6_100 |
452 | | CR_EXT1[8:11] | SVCR2_101 | | CR_EXT1[24:27] | SVCR6_101 |
453 | | CR_EXT2[8:11] | SVCR2_110 | | CR_EXT2[24:27] | SVCR6_110 |
454 | | CR_EXT3[8:11] | SVCR2_111 | | CR_EXT3[24:27] | SVCR6_111 |
455 | CR[3] | CR[44:47] | SVCR3_000 | CR[7] | CR[60:63] | SVCR7_000 |
456 | | CR_EXT1[44:47] | SVCR3_001 | | CR_EXT1[60:63] | SVCR7_001 |
457 | | CR_EXT2[44:47] | SVCR3_010 | | CR_EXT2[60:63] | SVCR7_010 |
458 | | CR_EXT3[44:47] | SVCR3_011 | | CR_EXT3[60:63] | SVCR7_011 |
459 | *CR[-5]* | CR[12:15] | SVCR3_100 | *CR[-1]* | CR[28:31] | SVCR7_100 |
460 | | CR_EXT1[12:15] | SVCR3_101 | | CR_EXT1[28:31] | SVCR7_101 |
461 | | CR_EXT2[12:15] | SVCR3_110 | | CR_EXT2[28:31] | SVCR7_110 |
462 | | CR_EXT3[12:15] | SVCR3_111 | | CR_EXT3[28:31] | SVCR7_111 |
464 Note: CR[-8] through CR[-1] are not part of OpenPower v3.1, they are the MSB half of the 64-bit CR SPR.
468 Instructions are broken down by Register Profiles as listed in the following auto-generated page:
469 [[opcode_regs_deduped]]. "Non-SV" indicates that the operations with this Register Profile cannot be Vectorised (mtspr, bc, dcbz, twi)
493 Remapped Encoding Fields:
495 | `0` | `1:3` | `4:5` | `6:7` | `8:10` | `11:13` | `14:16` | `17:23` |
496 |-----------|-------|---------|-------|-------------|-------------|-------------|---------|
497 | MASK_KIND | MASK | ELWIDTH | SUBVL | Rdest_EXTRA | Rsrc1_EXTRA | Rsrc2_EXTRA | TBD |
501 Remapped Encoding Fields:
503 | `0` | `1:3` | `4:5` | `6:7` | `8:10` | `11:13` | `14:16` | `17:18` | `19:20` | `21:23` |
504 |-----------|-------|---------|-------|-------------|-------------|----------|-------------|-----------|---------|
505 | MASK_KIND | MASK | ELWIDTH | SUBVL | Rdest_EXTRA | Rsrc1_EXTRA | MASK_SRC | ELWIDTH_SRC | SUBVL_SRC | TBD |
509 Remapped Encoding Fields:
511 | `0` | `1:3` | `4:5` | `6:7` | `8:10` | `11:13` | `14:16` | `17:18` | `19:20` | `21:23` |
512 |-----------|-------|---------|-------|-------------|-------------|----------|-------------|-----------|---------|
513 | MASK_KIND | MASK | ELWIDTH | SUBVL | Rdest_EXTRA | Rsrc1_EXTRA | MASK_SRC | ELWIDTH_SRC | SUBVL_SRC | TBD |
517 Remapped Encoding Fields:
519 | `0` | `1:3` | `4:5` | `6:7` | `8:10` | `11:13` | `14:16` | `17:18` | `19:20` | `21:23` |
520 |-----------|-------|---------|-------|-------------|-------------|----------|-------------|-----------|---------|
521 | MASK_KIND | MASK | ELWIDTH | SUBVL | Rdest_EXTRA | Rsrc1_EXTRA | MASK_SRC | ELWIDTH_SRC | SUBVL_SRC | TBD |
525 Remapped Encoding Fields:
527 | `0` | `1:3` | `4:5` | `6:7` | `8:10` | `11:13` | `14:16` | `17:18` | `19:20` | `21:23` |
528 |-----------|-------|---------|-------|-------------|-------------|----------|-------------|-----------|---------|
529 | MASK_KIND | MASK | ELWIDTH | SUBVL | Rdest_EXTRA | Rsrc1_EXTRA | MASK_SRC | ELWIDTH_SRC | SUBVL_SRC | TBD |
533 Remapped Encoding Fields:
535 | `0` | `1:3` | `4:5` | `6:7` | `8:10` | `11:13` | `14:16` | `17:18` | `19:20` | `21:23` |
536 |-----------|-------|---------|-------|-------------|-------------|----------|-------------|-----------|---------|
537 | MASK_KIND | MASK | ELWIDTH | SUBVL | Rdest_EXTRA | Rsrc1_EXTRA | MASK_SRC | ELWIDTH_SRC | SUBVL_SRC | TBD |
541 Remapped Encoding Fields:
543 | `0` | `1:3` | `4:5` | `6:7` | `8:10` | `11:13` | `14:16` | `17:18` | `19:20` | `21:23` |
544 |-----------|-------|---------|-------|-------------|-------------|----------|-------------|-----------|---------|
545 | MASK_KIND | MASK | ELWIDTH | SUBVL | Rdest_EXTRA | Rsrc1_EXTRA | MASK_SRC | ELWIDTH_SRC | SUBVL_SRC | TBD |
549 Remapped Encoding Fields:
551 | `0` | `1:3` | `4:5` | `6:7` | `8:10` | `11:13` | `14:16` | `17:18` | `19:20` | `21:23` |
552 |-----------|-------|---------|-------|-------------|-------------|----------|-------------|-----------|---------|
553 | MASK_KIND | MASK | ELWIDTH | SUBVL | Rdest_EXTRA | Rsrc1_EXTRA | MASK_SRC | ELWIDTH_SRC | SUBVL_SRC | TBD |
557 Remapped Encoding Fields:
559 | `0` | `1:3` | `4:5` | `6:7` | `8:10` | `11:13` | `14:16` | `17:23` |
560 |-----------|-------|---------|-------|-------------|-------------|-------------|---------|
561 | MASK_KIND | MASK | ELWIDTH | SUBVL | Rdest_EXTRA | Rsrc1_EXTRA | Rsrc2_EXTRA | TBD |
565 Remapped Encoding Fields:
567 | `0` | `1:3` | `4:5` | `6:7` | `8:10` | `11:13` | `14:16` | `17:23` |
568 |-----------|-------|---------|-------|-------------|-------------|-------------|---------|
569 | MASK_KIND | MASK | ELWIDTH | SUBVL | Rdest_EXTRA | Rsrc1_EXTRA | Rsrc2_EXTRA | TBD |
573 Remapped Encoding Fields:
575 | `0` | `1:3` | `4:5` | `6:7` | `8:10` | `11:13` | `14:16` | `17:23` |
576 |-----------|-------|---------|-------|-------------|-------------|-------------|---------|
577 | MASK_KIND | MASK | ELWIDTH | SUBVL | Rdest_EXTRA | Rsrc1_EXTRA | Rsrc2_EXTRA | TBD |
581 Remapped Encoding Fields:
583 | `0` | `1:3` | `4:5` | `6:7` | `8:10` | `11:13` | `14:16` | `17:23` |
584 |-----------|-------|---------|-------|-------------|-------------|-------------|---------|
585 | MASK_KIND | MASK | ELWIDTH | SUBVL | Rdest_EXTRA | Rsrc1_EXTRA | Rsrc2_EXTRA | TBD |
587 <!-- comment needed to stop ikiwiki markdown from mis-parsing table -->
589 ## 2R-1W-CRo (rl(w|d)imi)
591 Remapped Encoding Fields:
593 | `0` | `1:3` | `4:5` | `6:7` | `8:10` | `11:13` | `14:23` |
594 |-----------|-------|---------|-------|-------------|-------------|---------|
595 | MASK_KIND | MASK | ELWIDTH | SUBVL | Rdest_EXTRA | Rsrc1_EXTRA | TBD |
601 Remapped Encoding Fields:
603 | `0` | `1:3` | `4:5` | `6:7` | `8:10` | `11:13` | `14:16` | `17:23` |
604 |-----------|-------|---------|-------|-------------|-------------|-------------|---------|
605 | MASK_KIND | MASK | ELWIDTH | SUBVL | Rdest_EXTRA | Rsrc1_EXTRA | Rsrc2_EXTRA | TBD |
609 Remapped Encoding Fields:
611 | `0` | `1:3` | `4:5` | `6:7` | `8:10` | `11:13` | `14:16` | `17:19` | `20:23` |
612 |-----------|-------|---------|-------|-------------|-------------|-------------|-------------|----------|
613 | MASK_KIND | MASK | ELWIDTH | SUBVL | Rdest_EXTRA | Rsrc1_EXTRA | Rsrc2_EXTRA | Rsrc3_EXTRA | Reserved |