1 SimpleV Prefix (SVprefix) Proposal v0.3
2 =======================================
4 * Copyright (c) Jacob Lifshay, 2019
5 * Copyright (c) Luke Kenneth Casson Leighton, 2019
7 This proposal is designed to be able to operate without SVorig, but not to
8 require the absence of SVorig See Specification_.
10 .. _Specification: http://libre-riscv.org/simple_v_extension/specification/
17 Conventions used in this document:
19 * Bits are numbered starting from 0 at the LSB, so bit 3 is 1 in the integer 8.
20 * Bit ranges are inclusive on both ends, so 5:3 means bits 5, 4, and 3.
21 * Operations work on variable-length vectors of sub-vectors up to *VL* in length,
23 has a length *svlen*, and *svlen* elements of type *etype*.
24 * The actual total number of elements is therefore *svlen* times *VL*.
25 * When the vectors are stored
26 in registers, all elements are packed so that there is no padding in-between
27 elements of the same vector.
28 * The register file itself is thus best viewed as a byte-level
29 SRAM that is typecast to an array of *etype*s
30 * The number of bytes in a sub-vector, *svsz*, is the
31 product of *svlen* and the element size in bytes.
36 The following partial / full implementation options are possible:
38 * SVPrefix augments the main Specification_
39 * SVPregix operates independently, without the main spec VL (and MVL) CSRs (in any priv level)
40 * SVPrefix operates independently, without the main spec SUBVL CSRs (in any priv level)
41 * SVPrefix operates independently, with no support for VL (or MVL) overrides in the 64 bit instruction format either (VLtyp=0 as the only legal permitted value)
42 * SVPrefix operates independently, with no support for svlen overrides in either the 48 or 64 bit instruction format either (svlen=0 as the only legal permitted value).
44 All permutations of the above options are permitted, and the UNIX platform must raise illegal instruction exceptions on implementations that do not support each option. For example, an implementation that has no support for VLtyp that sees a nonzero VLtyp must raise an illegal instruction exception.
46 Note that SVPrefix (VLtyp and svlen) and the main spec share (modify) the STATE CSR. P48 and P64 opcodes must **NOT** set VLtyp or svlen inside loops that also use VL or SUBVL. Doing so will result in undefined behaviour, as STATE will be affected by doing so.
48 However, using VLtyp or svlen in standalone operations, or pushing (and restoring) the contents of the STATE CSR to the stack, or just storing its contents in a temporary register whilst executing a sequence of P48 or P64 opcodes, is perfectly fine.
50 If the main Specification_ CSRs are to be supported, the STATE, VL, MVL and SUBVL CSRs all operate according to the main specification. Under the options above, hypothetically an implementor could choose not to support setting of VL, MVL or SUBVL (only allowing them to be set to a value of 1). Under such circumstances, where *neither* VL/MVL *nor* SUBVL are supported, STATE would then not be required either.
52 If however support for SUBVL is to be provided, storing of the sub-vector offsets and SUBVL itself (and context switching of the same) in the STATE CSRs are mandatory.
54 Likewise if support for VL is to be provided, storing of VL, MVL and the dest and src offsets (and context switching of the same) in the STATE CSRs are mandatory.
57 Half-Precision Floating Point (FP16)
58 ====================================
60 If the F extension is supported, SVprefix adds support for FP16 in the
61 base FP instructions by using 10 (H) in the floating-point format field *fmt*
62 and using 001 (H) in the floating-point load/store *width* field.
64 Compressed Instructions
65 =======================
66 This proposal does not include any prefixed RVC instructions, instead, it will
67 include 32-bit instructions that are compressed forms of SVprefix 48-bit
68 instructions, in the same manner that RVC instructions are compressed forms of
69 RVI instructions. The compressed instructions will be defined later by
70 considering which 48-bit instructions are the most common.
72 48-bit Prefixed Instructions
73 ============================
74 All 48-bit prefixed instructions contain a 32-bit "base" instruction as the
75 last 4 bytes. Since all 32-bit instructions have bits 1:0 set to 11, those bits
76 are reused for additional encoding space in the 48-bit instructions.
78 64-bit Prefixed Instructions
79 ============================
81 The 48 bit format is further extended with the full 128-bit range on all source
82 and destination registers, and the option to set both VL and MVL is provided.
84 48-bit Instruction Encodings
85 ============================
87 In the following table, *Reserved* entries must be zero. RV32 equivalent encodings
88 included for side-by-side comparison (and listed below, separately).
92 +---------------+--------+------------+------------+-----+------------+-------------+------+------------+--------+
93 | Encoding | 17 | 16 | 15 | 14 | 13 | 12 | 11:7 | 6 | 5:0 |
94 +---------------+--------+------------+------------+-----+------------+-------------+------+------------+--------+
95 | P48-LD-type | rd[5] | rs1[5] | vitp7[6] | vd | vs1 | vitp7[5:0] | *Reserved* | 011111 |
96 +---------------+--------+------------+------------+-----+------------+-------------+------+------------+--------+
97 | P48-ST-type |vitp7[6]| rs1[5] | rs2[5] | vs2 | vs1 | vitp7[5:0] | *Reserved* | 011111 |
98 +---------------+--------+------------+------------+-----+------------+-------------+------+------------+--------+
99 | P48-R-type | rd[5] | rs1[5] | rs2[5] | vs2 | vs1 | vitp6 | *Reserved* | 011111 |
100 +---------------+--------+------------+------------+-----+------------+--------------------+------------+--------+
101 | P48-I-type | rd[5] | rs1[5] | vitp7[6] | vd | vs1 | vitp7[5:0] | *Reserved* | 011111 |
102 +---------------+--------+------------+------------+-----+------------+--------------------+------------+--------+
103 | P48-U-type | rd[5] | *Reserved* | *Reserved* | vd | *Reserved* | vitp6 | *Reserved* | 011111 |
104 +---------------+--------+------------+------------+-----+------------+-------------+------+------------+--------+
105 | P48-FR-type | rd[5] | rs1[5] | rs2[5] | vs2 | vs1 | *Reserved* | vtp5 | *Reserved* | 011111 |
106 +---------------+--------+------------+------------+-----+------------+-------------+------+------------+--------+
107 | P48-FI-type | rd[5] | rs1[5] | vitp7[6] | vd | vs1 | vitp7[5:0] | *Reserved* | 011111 |
108 +---------------+--------+------------+------------+-----+------------+-------------+------+------------+--------+
109 | P48-FR4-type | rd[5] | rs1[5] | rs2[5] | vs2 | rs3[5] | vs3 [#fr4]_ | vtp5 | *Reserved* | 011111 |
110 +---------------+--------+------------+------------+-----+------------+-------------+------+------------+--------+
112 .. [#fr4] Only vs2 and vs3 are included in the P48-FR4-type encoding because
113 there is not enough space for vs1 as well, and because it is more
114 useful to have a scalar argument for each of the multiplication and
115 addition portions of fmadd than to have two scalars on the
116 multiplication portion.
118 Table showing correspondance between P48-*-type and RV32-*-type. These are
119 bits 47:18 (RV32 shifted up by 16 bits):
121 +---------------+---------------+
123 +---------------+---------------+
124 | RV32 Encoding | 31:2 |
125 +---------------+---------------+
126 | P48-LD-type | RV32-I-type |
127 +---------------+---------------+
128 | P48-ST-type | RV32-S-Type |
129 +---------------+---------------+
130 | P48-R-type | RV32-R-Type |
131 +---------------+---------------+
132 | P48-I-type | RV32-I-Type |
133 +---------------+---------------+
134 | P48-U-type | RV32-U-Type |
135 +---------------+---------------+
136 | P48-FR-type | RV32-FR-Type |
137 +---------------+---------------+
138 | P48-FI-type | RV32-I-Type |
139 +---------------+---------------+
140 | P48-FR4-type | RV32-FR-type |
141 +---------------+---------------+
143 Table showing Standard RV32 encodings:
145 +---------------+-------------+-------+----------+----------+--------+----------+--------+--------+------------+
146 | Encoding | 31:27 | 26:25 | 24:20 | 19:15 | 14:12 | 11:7 | 6:2 | 1 | 0 |
147 +---------------+-------------+-------+----------+----------+--------+----------+--------+--------+------------+
148 | RV32-R-type + funct7 + rs2[4:0] + rs1[4:0] + funct3 | rd[4:0] + opcode + 1 + 1 |
149 +---------------+-------------+-------+----------+----------+--------+----------+--------+--------+------------+
150 | RV32-S-type + imm[11:5] + rs2[4:0] + rs1[4:0] + funct3 | imm[4:0] + opcode + 1 + 1 |
151 +---------------+-------------+-------+----------+----------+--------+----------+--------+--------+------------+
152 | RV32-I-type + imm[11:0] + rs1[4:0] + funct3 | rd[4:0] + opcode + 1 + 1 |
153 +---------------+-------------+-------+----------+----------+--------+----------+--------+--------+------------+
154 | RV32-U-type + imm[31:12] | rd[4:0] + opcode + 1 + 1 |
155 +---------------+-------------+-------+----------+----------+--------+----------+--------+--------+------------+
156 | RV32-FR4-type + rs3[4:0] + fmt + rs2[4:0] + rs1[4:0] + funct3 | rd[4:0] + opcode + 1 + 1 |
157 +---------------+-------------+-------+----------+----------+--------+----------+--------+--------+------------+
158 | RV32-FR-type + funct5 + fmt + rs2[4:0] + rs1[4:0] + rm | rd[4:0] + opcode + 1 + 1 |
159 +---------------+-------------+-------+----------+----------+--------+----------+--------+--------+------------+
161 64-bit Instruction Encodings
162 ============================
164 Where in the 48 bit format the prefix is "0b0011111" in bits 0 to 6, this is
165 now set to "0b0111111".
167 +---------------+---------------+--------------+-----------+
168 | 63:48 | 47:18 | 17:7 | 6:0 |
169 +---------------+---------------+--------------+-----------+
170 | 64 bit prefix | RV32[31:3] | P48[17:7] | 0b0111111 |
171 +---------------+---------------+--------------+-----------+
173 * The 64 bit prefix format is below
174 * Bits 18 to 47 contain bits 3 to 31 of a standard RV32 format
175 * Bits 7 to 17 contain bits 7 through 17 of the P48 format
176 * Bits 0 to 6 contain the standard RV 64-bit prefix 0b0111111
178 64 bit prefix format:
180 +--------------+-------+--------+--------+--------+--------+
181 | Encoding | 63 | 62 | 61 | 60 | 59:48 |
182 +--------------+-------+--------+--------+--------+--------+
183 | P64-LD-type | rd[6] | rs1[6] | | | VLtyp |
184 +--------------+-------+--------+--------+--------+--------+
185 | P64-ST-type | | rs1[6] | rs2[6] | | VLtyp |
186 +--------------+-------+--------+--------+--------+--------+
187 | P64-R-type | rd[6] | rs1[6] | rs2[6] | | VLtyp |
188 +--------------+-------+--------+--------+--------+--------+
189 | P64-I-type | rd[6] | rs1[6] | | | VLtyp |
190 +--------------+-------+--------+--------+--------+--------+
191 | P64-U-type | rd[6] | | | | VLtyp |
192 +--------------+-------+--------+--------+--------+--------+
193 | P64-FR-type | | rs1[6] | rs2[6] | | VLtyp |
194 +--------------+-------+--------+--------+--------+--------+
195 | P64-FI-type | rd[6] | rs1[6] | rs2[6] | | VLtyp |
196 +--------------+-------+--------+--------+--------+--------+
197 | P64-FR4-type | rd[6] | rs1[6] | rs2[6] | rs3[6] | VLtyp |
198 +--------------+-------+--------+--------+--------+--------+
200 The extra bit for src and dest registers provides the full range of
201 up to 128 registers, when combined with the extra bit from the 48 bit
202 prefix as well. VLtyp encodes how (whether) to set VL and MAXVL.
207 NOTE: VL and MVL below are modified (potentially damaging) and so is the STATE CSR. It is the responsibility of the programmer to ensure that modifications to STATE do not compromise loops or VLIW Group opetations, by saving and restoring the STATE CSR (if needed).
209 +-----------+-------------+--------------+----------+----------------------+
210 | VLtyp[11] | VLtyp[10:6] | VLtyp[5:1] | VLtyp[0] | comment |
211 +-----------+-------------+--------------+----------+----------------------+
212 | 0 | 000000 | 00000 | 0 | no change to VL/MVL |
213 +-----------+-------------+--------------+----------+----------------------+
214 | 0 | VLdest | VLEN | vlt | VL imm/reg mode (vlt)|
215 +-----------+-------------+--------------+----------+----------------------+
216 | 1 | VLdest | MVL+VL-immed | 0 | MVL+VL immed mode |
217 +-----------+-------------+--------------+----------+----------------------+
218 | 1 | VLdest | MVL-immed | 1 | MVL immed mode |
219 +-----------+-------------+--------------+----------+----------------------+
221 Note: when VLtyp is all zeros, neither VL nor MVL are changed.
223 Just as in the VLIW format, when bit 11 of VLtyp is zero:
225 * if vlt is zero, bits 1 to 5 specify the VLEN as a 5 bit immediate
226 (offset by 1: 0b00000 represents VL=1, 0b00001 represents VL=2 etc.)
227 * if vlt is 1, bits 1 to 5 specify the scalar (RV standard) register
228 from which VL is set. x0 is not permitted
229 * VL goes into the scalar register VLdest (if VLdest is not x0)
231 When bit 11 of VLtype is 1:
233 * if VLtyp[0] is zero, both MAXVL and VL are set to (imm+1). The same
234 value goes into the scalar register VLdest (if VLdest is not x0)
235 * if VLtyp[0] is 1, MAXVL is set to (imm+1).
236 VL will be truncated to within the new range (if VL was greater
237 than the new MAXVL). The new VL goes into the scalar register VLdest
238 (if VLdest is not x0).
240 This gives the option to set up VL in a "loop mode" (VLtype[11]=0) or
241 in a "one-off" mode (VLtype[11]=1) which sets both MVL and VL to the
242 same immediate value. This may be most useful for one-off Vectorised
243 operations such as LOAD-MULTI / STORE-MULTI, for saving and restoration
244 of large batches of registers in context-switches or function calls.
246 Note that VLtyp's VL and MVL are the same as the main Specification_ VL or MVL, and that loops will also alter srcoffs and destoffs. It is the programmer's responsibility to ensure that STATE is not compromised (e.g saved to a temp reg or to the stack).
248 Furthermore, the execution order and exception handling must be exactly the same as in the main spec.
250 vs#/vd Fields' Encoding
251 =======================
253 +--------+----------+----------------------------------------------------------+
254 | vs#/vd | Mnemonic | Meaning |
255 +========+==========+==========================================================+
256 | 0 | S | the rs#/rd field specifies a scalar (single sub-vector); |
257 | | | the rs#/rd field is zero-extended to get the actual |
258 | | | 7-bit register number |
259 +--------+----------+----------------------------------------------------------+
260 | 1 | V | the rs#/rd field specifies a vector; the rs#/rd field is |
261 | | | decoded using the `Vector Register Number Encoding`_ to |
262 | | | get the actual 7-bit register number |
263 +--------+----------+----------------------------------------------------------+
265 If a vs#/vd field is not present, it is as if it was present with a value that
266 is the bitwise-or of all present vs#/vd fields.
268 * scalar register numbers do NOT increment when allocated in the
269 hardware for-loop. the same scalar register number is handed
272 * vector register numbers *DO* increase when allocated in the
273 hardware for-loop. sequentially-increasing register data
274 is handed to sequential ALUs.
276 Vector Register Number Encoding
277 ===============================
279 For the 48 bit format, when vs#/vd is 1, the actual 7-bit register number is derived from the
280 corresponding 6-bit rs#/rd field:
282 +---------------------------------+
283 | Actual 7-bit register number |
284 +===========+=============+=======+
285 | Bit 6 | Bits 5:1 | Bit 0 |
286 +-----------+-------------+-------+
287 | rs#/rd[0] | rs#/rd[5:1] | 0 |
288 +-----------+-------------+-------+
290 For the 64 bit format, the 7 bit register is constructed from the 7 bit fields: bits 0 to 4 from the 32 bit RV Standard format, bit 5 from the 48 bit prefix and bit 6 from the 64 bit prefix. Thus in the 64 bit format the full range of up to 128 registers is directly available. This for both when either scalar or vector mode is set.
292 Load/Store Kind (lsk) Field Encoding
293 ====================================
295 +--------+-----+--------------------------------------------------------------------------------+
296 | vd/vs2 | vs1 | Meaning |
297 +========+=====+================================================================================+
298 | 0 | 0 | srcbase is scalar, LD/ST is pure scalar. |
299 +--------+-----+--------------------------------------------------------------------------------+
300 | 1 | 0 | srcbase is scalar, LD/ST is unit strided |
301 +--------+-----+--------------------------------------------------------------------------------+
302 | 0 | 1 | srcbase is a vector (gather/scatter aka array of srcbases). VSPLAT and VSELECT |
303 +--------+-----+--------------------------------------------------------------------------------+
304 | 1 | 1 | srcbase is a vector, LD/ST is a full vector LD/ST. |
305 +--------+-----+--------------------------------------------------------------------------------+
309 * A register strided LD/ST would require *5* registers. srcbase, vd/vs2, predicate 1, predicate 2 and the stride register.
310 * Complex strides may all be done with a general purpose vector of srcbases.
311 * Twin predication may be used even when vd/vs1 is a scalar, to give VSPLAT and VSELECT, because the hardware loop ends on the first occurrence of a 1 in the predicate when a predicate is applied to a scalar.
312 * Full vectorised gather/scatter is enabled when both registers are marked as vectorised, however unlike e.g Intel AVX512, twin predication can be applied.
314 Open question: RVV overloads the width field of LOAD-FP/STORE-FP using the bit 2 to indicate additional interpretation of the 11 bit immediate. Should this be considered?
317 Sub-Vector Length (svlen) Field Encoding
318 ========================================
320 NOTE: svlen is the same as the main spec SUBVL, and modifies the STATE CSR. The same caveats apply to svlen as do to SUBVL.
322 Bitwidth, from VL's perspective, is a multiple of the elwidth times svlen. So within each loop of VL there are svlen sub-elements of elwidth in size, just like in a SIMD architecture. When svlen is set to 0b00 (indicating svlen=1) no such SIMD-like behaviour exists and the subvectoring is disabled.
324 Predicate bits do not apply to the individual sub-vector elements, they apply to the entire subvector group. This saves instructions on setup of the predicate.
326 +----------------+-------+
327 | svlen Encoding | Value |
328 +================+=======+
330 +----------------+-------+
332 +----------------+-------+
334 +----------------+-------+
336 +----------------+-------+
338 In independent standalone implementations that do not implement the main specification, the value of SUBVL in the above table (svtyp=0b00) is set to 1, such that svlen is also 1.
340 Behaviour of operations that set svlen are identical to those of the main spec. See section on VLtyp, above.
342 Predication (pred) Field Encoding
343 =================================
345 +------+------------+--------------------+----------------------------------------+
346 | pred | Mnemonic | Predicate Register | Meaning |
347 +======+============+====================+========================================+
348 | 000 | *None* | *None* | The instruction is unpredicated |
349 +------+------------+--------------------+----------------------------------------+
350 | 001 | *Reserved* | *Reserved* | |
351 +------+------------+--------------------+----------------------------------------+
352 | 010 | !x9 | x9 (s1) | execute vector op[0..i] on x9[i] == 0 |
353 +------+------------+ +----------------------------------------+
354 | 011 | x9 | | execute vector op[0..i] on x9[i] == 1 |
355 +------+------------+--------------------+----------------------------------------+
356 | 100 | !x10 | x10 (a0) | execute vector op[0..i] on x10[i] == 0 |
357 +------+------------+ +----------------------------------------+
358 | 101 | x10 | | execute vector op[0..i] on x10[i] == 1 |
359 +------+------------+--------------------+----------------------------------------+
360 | 110 | !x11 | x11 (a1) | execute vector op[0..i] on x11[i] == 0 |
361 +------+------------+ +----------------------------------------+
362 | 111 | x11 | | execute vector op[0..i] on x11[i] == 1 |
363 +------+------------+--------------------+----------------------------------------+
365 Twin-predication (tpred) Field Encoding
366 =======================================
368 +-------+------------+--------------------+----------------------------------------------+
369 | tpred | Mnemonic | Predicate Register | Meaning |
370 +=======+============+====================+==============================================+
371 | 000 | *None* | *None* | The instruction is unpredicated |
372 +-------+------------+--------------------+----------------------------------------------+
373 | 001 | x9,off | src=x9, dest=none | src[0..i] uses x9[i], dest unpredicated |
374 +-------+------------+ +----------------------------------------------+
375 | 010 | off,x10 | src=none, dest=x10 | dest[0..i] uses x10[i], src unpredicated |
376 +-------+------------+ +----------------------------------------------+
377 | 011 | x9,10 | src=x9, dest=x10 | src[0..i] uses x9[i], dest[0..i] uses x10[i] |
378 +-------+------------+--------------------+----------------------------------------------+
379 | 100 | *None* | *RESERVED* | Instruction is unpredicated (TBD) |
380 +-------+------------+--------------------+----------------------------------------------+
381 | 101 | !x9,off | src=!x9, dest=none | |
382 +-------+------------+ +----------------------------------------------+
383 | 110 | off,!x10 | src=none, dest=!x10| |
384 +-------+------------+ +----------------------------------------------+
385 | 111 | !x9,!x10 | src=!x9, dest=!x10 | |
386 +-------+------------+--------------------+----------------------------------------------+
388 Integer Element Type (itype) Field Encoding
389 ===========================================
391 +------------+-------+--------------+--------------+-----------------+-------------------+
392 | Signedness | itype | Element Type | Mnemonic in | Mnemonic in FP | Meaning (INT may |
393 | [#sgn_def]_| | | Integer | Instructions | be un/signed, FP |
394 | [#sgn_def]_| | | Instructions | (such as fmv.x) | just re-sized |
395 +============+=======+==============+==============+=================+===================+
396 | Unsigned | 01 | u8 | BU | BU | Unsigned 8-bit |
397 | +-------+--------------+--------------+-----------------+-------------------+
398 | | 10 | u16 | HU | HU | Unsigned 16-bit |
399 | +-------+--------------+--------------+-----------------+-------------------+
400 | | 11 | u32 | WU | WU | Unsigned 32-bit |
401 | +-------+--------------+--------------+-----------------+-------------------+
402 | | 00 | uXLEN | WU/DU/QU | WU/LU/TU | Unsigned XLEN-bit |
403 +------------+-------+--------------+--------------+-----------------+-------------------+
404 | Signed | 01 | i8 | BS | BS | Signed 8-bit |
405 | +-------+--------------+--------------+-----------------+-------------------+
406 | | 10 | i16 | HS | HS | Signed 16-bit |
407 | +-------+--------------+--------------+-----------------+-------------------+
408 | | 11 | i32 | W | W | Signed 32-bit |
409 | +-------+--------------+--------------+-----------------+-------------------+
410 | | 00 | iXLEN | W/D/Q | W/L/T | Signed XLEN-bit |
411 +------------+-------+--------------+--------------+-----------------+-------------------+
413 .. [#sgn_def] Signedness is defined in `Signedness Decision Procedure`_
415 Note: vector mode is effectively a type-cast of the register file
416 as if it was a sequential array being typecast to typedef itype[]
417 (c syntax). The starting point of the "typecast" is the vector
420 Example: if itype=0b10 (u16), and rd is set to "vector", and
421 VL is set to 4, the 64-bit register at rd is subdivided into
422 *FOUR* 16-bit destination elements. It is *NOT* four
423 separate 64-bit destination registers (rd+0, rd+1, rd+2, rd+3)
424 that are sign-extended from the source width size out to 64-bit,
425 because that is itype=0b00 (uXLEN).
427 Note also: changing elwidth creates packed elements that, depending on VL, may create vectors that do not fit perfectly onto XLEN sized registry file bit-boundaries. This does NOT result in the destruction of the MSBs of the last register written to at the end of a VL loop. More details on how to handle this are described in the main Specification_.
429 Signedness Decision Procedure
430 =============================
432 1. If the opcode field is either OP or OP-IMM, then
433 1. Signedness is Unsigned.
434 2. If the opcode field is either OP-32 or OP-IMM-32, then
435 1. Signedness is Signed.
436 3. If Signedness is encoded in a field of the base instruction, [#sign_enc]_ then
437 1. Signedness uses the encoded value.
439 1. Signedness is Unsigned.
441 .. [#sign_enc] Like in fcvt.d.l[u], but unlike in fmv.x.w, since there is no
444 Vector Type and Predication 5-bit (vtp5) Field Encoding
445 =========================================================
447 In the following table, X denotes a wildcard that is 0 or 1 and can be a
448 different value for every occurrence.
450 +-------+-----------+-----------+
451 | vtp5 | pred | svlen |
452 +=======+===========+===========+
453 | 1XXXX | vtp5[4:2] | vtp5[1:0] |
458 +-------+-----------+-----------+
459 | 001XX | *Reserved* |
460 +-------+-----------------------+
462 Vector Integer Type and Predication 6-bit (vitp6) Field Encoding
463 =================================================================
465 In the following table, X denotes a wildcard that is 0 or 1 and can be a
466 different value for every occurrence.
468 +--------+------------+---------+------------+------------+
469 | vitp6 | itype | pred[2] | pred[0:1] | svlen |
470 +========+============+=========+============+============+
471 | XX1XXX | vitp6[5:4] | 0 | vitp6[3:2] | vitp6[1:0] |
474 +--------+------------+---------+------------+------------+
475 | XX01XX | *Reserved* |
476 +--------+------------------------------------------------+
478 vitp7 field: only tpred
480 +---------+------------+----------+-------------+------------+
481 | vitp7 | itype | tpred[2] | tpred[0:1] | svlen |
482 +=========+============+==========+=============+============+
483 | XXXXXXX | vitp7[5:4] | vitp7[6] | vitp7[3:2] | vitp7[1:0] |
484 +---------+------------+----------+-------------+------------+
486 48-bit Instruction Encoding Decision Procedure
487 ==============================================
489 In the following decision procedure, *Reserved* means that there is not yet a
490 defined 48-bit instruction encoding for the base instruction.
492 1. If the base instruction is a load instruction, then
493 a. If the base instruction is an I-type instruction, then
494 1. The encoding is P48-LD-type.
496 1. The encoding is *Reserved*.
497 2. If the base instruction is a store instruction, then
498 a. If the base instruction is an S-type instruction, then
499 1. The encoding is P48-ST-type.
501 1. The encoding is *Reserved*.
502 3. If the base instruction is a SYSTEM instruction, then
503 a. The encoding is *Reserved*.
504 4. If the base instruction is an integer instruction, then
505 a. If the base instruction is an R-type instruction, then
506 1. The encoding is P48-R-type.
507 b. If the base instruction is an I-type instruction, then
508 1. The encoding is P48-I-type.
509 c. If the base instruction is an S-type instruction, then
510 1. The encoding is *Reserved*.
511 d. If the base instruction is an B-type instruction, then
512 1. The encoding is *Reserved*.
513 e. If the base instruction is an U-type instruction, then
514 1. The encoding is P48-U-type.
515 f. If the base instruction is an J-type instruction, then
516 1. The encoding is *Reserved*.
518 1. The encoding is *Reserved*.
519 5. If the base instruction is a floating-point instruction, then
520 a. If the base instruction is an R-type instruction, then
521 1. The encoding is P48-FR-type.
522 b. If the base instruction is an I-type instruction, then
523 1. The encoding is P48-FI-type.
524 c. If the base instruction is an S-type instruction, then
525 1. The encoding is *Reserved*.
526 d. If the base instruction is an B-type instruction, then
527 1. The encoding is *Reserved*.
528 e. If the base instruction is an U-type instruction, then
529 1. The encoding is *Reserved*.
530 f. If the base instruction is an J-type instruction, then
531 1. The encoding is *Reserved*.
532 g. If the base instruction is an R4-type instruction, then
533 1. The encoding is P48-FR4-type.
535 1. The encoding is *Reserved*.
537 a. The encoding is *Reserved*.
542 CSRs are the same as in the main Specification_, if associated functionality is implemented. They have the exact same meaning as in the main specification.
549 Associated SET and GET on the CSRs is exactly as in the main spec as well (including CSRRWI and CSRRW differences).
551 Note that if all of VL/MVL, SUBVL, VLtyp and svlen are all chosen by an implementor not to be implemented, the STATE CSR is not required.
553 However if partial functionality is implemented, the unimplemented bits in STATE must be zero, and, in the UNIX Platform, an illegal exception **MUST** be raised if unsupported bits are written to.
555 Additional Instructions
556 =======================
558 Add instructions to convert between integer types.
560 Add instructions to `swizzle`_ elements in sub-vectors. Note that the sub-vector
561 lengths of the source and destination won't necessarily match.
563 .. _swizzle: https://www.khronos.org/opengl/wiki/Data_Type_(GLSL)#Swizzling
565 Add instructions to transpose (2-4)x(2-4) element matrices.
567 Add instructions to insert or extract a sub-vector from a vector, with the index
568 allowed to be both immediate and from a register (*immediate can be covered
569 by twin-predication, register might be, by virtue of predicates being registers*)
571 Add a register gather instruction (aka MV.X: regfile[rd] = regfile[regfile[rs1]])
575 Confirmation needed as to whether subvector extraction can be covered by twin predication (it probably can, it is one of the many purposes it is for).
579 What is SUBVL and how does it work
583 SVorig goes to a lot of effort to make VL 1<= MAXVL and MAXVL 1..64 where both CSRs may be stored internally in only 6 bits.
585 Thus, CSRRWI can reach 1..32 for VL and MAXVL.
587 In addition, setting a hardware loop to zero turning instructions into NOPs, um, just branch over them, to start the first loop at the end, on the test for loop variable being zero, a la c "while do" instead of "do while".
589 Or, does it not matter that VL only goes up to 31 on a CSRRWI, and that it only goes to a max of 63 rather than 64?
593 Should these questions be moved to Discussion subpage
597 Is MV.X good enough a substitute for swizzle?
601 Is vectorised srcbase ok as a gather scatter and ok substitute for register stride? 5 dependency registers (reg stride being the 5th) is quite scary
605 Why are integer conversion instructions needed, when the main SV spec covers them by allowing elwidth to be set on both src and dest regs?
609 Why are the SETVL rules so complex? What is the reason, how are loops carried out?
613 With SUBVL (sub vector len) being both a CSR and also part of the 48/64 bit opcode, how does that work?
617 What are the interaction rules when a 48/64 prefix opcode has a rd/rs that already has a Vector Context for either predication or a register?
619 It would perhaps make sense (and for svlen as well) to make 48/64 isolated and unaffected by VLIW context, with the exception of VL/MVL.
621 MVL and VL should be modifiable by 64 bit prefix as they are global in nature.
623 Possible solution, svlen and VLtyp allowed to share STATE CSR however programmer becomes responsible for push and pop of state during use of a sequence of P48 and P64 ops.