xref: /openbmc/qemu/target/mips/tcg/mxu_translate.c (revision ffe98631)
1 /*
2  *  Ingenic XBurst Media eXtension Unit (MXU) translation routines.
3  *
4  *  Copyright (c) 2004-2005 Jocelyn Mayer
5  *  Copyright (c) 2006 Marius Groeger (FPU operations)
6  *  Copyright (c) 2006 Thiemo Seufer (MIPS32R2 support)
7  *  Copyright (c) 2009 CodeSourcery (MIPS16 and microMIPS support)
8  *  Copyright (c) 2012 Jia Liu & Dongxue Zhang (MIPS ASE DSP support)
9  *
10  * SPDX-License-Identifier: LGPL-2.1-or-later
11  *
12  * Datasheet:
13  *
14  *   "XBurst® Instruction Set Architecture MIPS eXtension/enhanced Unit
15  *   Programming Manual", Ingenic Semiconductor Co, Ltd., revision June 2, 2017
16  */
17 
18 #include "qemu/osdep.h"
19 #include "tcg/tcg-op.h"
20 #include "exec/helper-gen.h"
21 #include "translate.h"
22 
23 /*
24  *
25  *       AN OVERVIEW OF MXU EXTENSION INSTRUCTION SET
26  *       ============================================
27  *
28  *
29  * MXU (full name: MIPS eXtension/enhanced Unit) is a SIMD extension of MIPS32
30  * instructions set. It is designed to fit the needs of signal, graphical and
31  * video processing applications. MXU instruction set is used in Xburst family
32  * of microprocessors by Ingenic.
33  *
34  * MXU unit contains 17 registers called X0-X16. X0 is always zero, and X16 is
35  * the control register.
36  *
37  *
38  *     The notation used in MXU assembler mnemonics
39  *     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
40  *
41  *  Register operands:
42  *
43  *   XRa, XRb, XRc, XRd - MXU registers
44  *   Rb, Rc, Rd, Rs, Rt - general purpose MIPS registers
45  *
46  *  Non-register operands:
47  *
48  *   aptn1 - 1-bit accumulate add/subtract pattern
49  *   aptn2 - 2-bit accumulate add/subtract pattern
50  *   eptn2 - 2-bit execute add/subtract pattern
51  *   optn2 - 2-bit operand pattern
52  *   optn3 - 3-bit operand pattern
53  *   sft4  - 4-bit shift amount
54  *   strd2 - 2-bit stride amount
55  *
56  *  Prefixes:
57  *
58  *   Level of parallelism:                Operand size:
59  *    S - single operation at a time       32 - word
60  *    D - two operations in parallel       16 - half word
61  *    Q - four operations in parallel       8 - byte
62  *
63  *  Operations:
64  *
65  *   ADD   - Add or subtract
66  *   ADDC  - Add with carry-in
67  *   ACC   - Accumulate
68  *   ASUM  - Sum together then accumulate (add or subtract)
69  *   ASUMC - Sum together then accumulate (add or subtract) with carry-in
70  *   AVG   - Average between 2 operands
71  *   ABD   - Absolute difference
72  *   ALN   - Align data
73  *   AND   - Logical bitwise 'and' operation
74  *   CPS   - Copy sign
75  *   EXTR  - Extract bits
76  *   I2M   - Move from GPR register to MXU register
77  *   LDD   - Load data from memory to XRF
78  *   LDI   - Load data from memory to XRF (and increase the address base)
79  *   LUI   - Load unsigned immediate
80  *   MUL   - Multiply
81  *   MULU  - Unsigned multiply
82  *   MADD  - 64-bit operand add 32x32 product
83  *   MSUB  - 64-bit operand subtract 32x32 product
84  *   MAC   - Multiply and accumulate (add or subtract)
85  *   MAD   - Multiply and add or subtract
86  *   MAX   - Maximum between 2 operands
87  *   MIN   - Minimum between 2 operands
88  *   M2I   - Move from MXU register to GPR register
89  *   MOVZ  - Move if zero
90  *   MOVN  - Move if non-zero
91  *   NOR   - Logical bitwise 'nor' operation
92  *   OR    - Logical bitwise 'or' operation
93  *   STD   - Store data from XRF to memory
94  *   SDI   - Store data from XRF to memory (and increase the address base)
95  *   SLT   - Set of less than comparison
96  *   SAD   - Sum of absolute differences
97  *   SLL   - Logical shift left
98  *   SLR   - Logical shift right
99  *   SAR   - Arithmetic shift right
100  *   SAT   - Saturation
101  *   SFL   - Shuffle
102  *   SCOP  - Calculate x’s scope (-1, means x<0; 0, means x==0; 1, means x>0)
103  *   XOR   - Logical bitwise 'exclusive or' operation
104  *
105  *  Suffixes:
106  *
107  *   E - Expand results
108  *   F - Fixed point multiplication
109  *   L - Low part result
110  *   R - Doing rounding
111  *   V - Variable instead of immediate
112  *   W - Combine above L and V
113  *
114  *
115  *     The list of MXU instructions grouped by functionality
116  *     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
117  *
118  * Load/Store instructions           Multiplication instructions
119  * -----------------------           ---------------------------
120  *
121  *  S32LDD XRa, Rb, s12               S32MADD XRa, XRd, Rs, Rt
122  *  S32STD XRa, Rb, s12               S32MADDU XRa, XRd, Rs, Rt
123  *  S32LDDV XRa, Rb, rc, strd2        S32MSUB XRa, XRd, Rs, Rt
124  *  S32STDV XRa, Rb, rc, strd2        S32MSUBU XRa, XRd, Rs, Rt
125  *  S32LDI XRa, Rb, s12               S32MUL XRa, XRd, Rs, Rt
126  *  S32SDI XRa, Rb, s12               S32MULU XRa, XRd, Rs, Rt
127  *  S32LDIV XRa, Rb, rc, strd2        D16MUL XRa, XRb, XRc, XRd, optn2
128  *  S32SDIV XRa, Rb, rc, strd2        D16MULE XRa, XRb, XRc, optn2
129  *  S32LDDR XRa, Rb, s12              D16MULF XRa, XRb, XRc, optn2
130  *  S32STDR XRa, Rb, s12              D16MAC XRa, XRb, XRc, XRd, aptn2, optn2
131  *  S32LDDVR XRa, Rb, rc, strd2       D16MACE XRa, XRb, XRc, XRd, aptn2, optn2
132  *  S32STDVR XRa, Rb, rc, strd2       D16MACF XRa, XRb, XRc, XRd, aptn2, optn2
133  *  S32LDIR XRa, Rb, s12              D16MADL XRa, XRb, XRc, XRd, aptn2, optn2
134  *  S32SDIR XRa, Rb, s12              S16MAD XRa, XRb, XRc, XRd, aptn1, optn2
135  *  S32LDIVR XRa, Rb, rc, strd2       Q8MUL XRa, XRb, XRc, XRd
136  *  S32SDIVR XRa, Rb, rc, strd2       Q8MULSU XRa, XRb, XRc, XRd
137  *  S16LDD XRa, Rb, s10, eptn2        Q8MAC XRa, XRb, XRc, XRd, aptn2
138  *  S16STD XRa, Rb, s10, eptn2        Q8MACSU XRa, XRb, XRc, XRd, aptn2
139  *  S16LDI XRa, Rb, s10, eptn2        Q8MADL XRa, XRb, XRc, XRd, aptn2
140  *  S16SDI XRa, Rb, s10, eptn2
141  *  S8LDD XRa, Rb, s8, eptn3
142  *  S8STD XRa, Rb, s8, eptn3         Addition and subtraction instructions
143  *  S8LDI XRa, Rb, s8, eptn3         -------------------------------------
144  *  S8SDI XRa, Rb, s8, eptn3
145  *  LXW Rd, Rs, Rt, strd2             D32ADD XRa, XRb, XRc, XRd, eptn2
146  *  LXH Rd, Rs, Rt, strd2             D32ADDC XRa, XRb, XRc, XRd
147  *  LXHU Rd, Rs, Rt, strd2            D32ACC XRa, XRb, XRc, XRd, eptn2
148  *  LXB Rd, Rs, Rt, strd2             D32ACCM XRa, XRb, XRc, XRd, eptn2
149  *  LXBU Rd, Rs, Rt, strd2            D32ASUM XRa, XRb, XRc, XRd, eptn2
150  *                                    S32CPS XRa, XRb, XRc
151  *                                    Q16ADD XRa, XRb, XRc, XRd, eptn2, optn2
152  * Comparison instructions            Q16ACC XRa, XRb, XRc, XRd, eptn2
153  * -----------------------            Q16ACCM XRa, XRb, XRc, XRd, eptn2
154  *                                    D16ASUM XRa, XRb, XRc, XRd, eptn2
155  *  S32MAX XRa, XRb, XRc              D16CPS XRa, XRb,
156  *  S32MIN XRa, XRb, XRc              D16AVG XRa, XRb, XRc
157  *  S32SLT XRa, XRb, XRc              D16AVGR XRa, XRb, XRc
158  *  S32MOVZ XRa, XRb, XRc             Q8ADD XRa, XRb, XRc, eptn2
159  *  S32MOVN XRa, XRb, XRc             Q8ADDE XRa, XRb, XRc, XRd, eptn2
160  *  D16MAX XRa, XRb, XRc              Q8ACCE XRa, XRb, XRc, XRd, eptn2
161  *  D16MIN XRa, XRb, XRc              Q8ABD XRa, XRb, XRc
162  *  D16SLT XRa, XRb, XRc              Q8SAD XRa, XRb, XRc, XRd
163  *  D16MOVZ XRa, XRb, XRc             Q8AVG XRa, XRb, XRc
164  *  D16MOVN XRa, XRb, XRc             Q8AVGR XRa, XRb, XRc
165  *  Q8MAX XRa, XRb, XRc               D8SUM XRa, XRb, XRc, XRd
166  *  Q8MIN XRa, XRb, XRc               D8SUMC XRa, XRb, XRc, XRd
167  *  Q8SLT XRa, XRb, XRc
168  *  Q8SLTU XRa, XRb, XRc
169  *  Q8MOVZ XRa, XRb, XRc             Shift instructions
170  *  Q8MOVN XRa, XRb, XRc             ------------------
171  *
172  *                                    D32SLL XRa, XRb, XRc, XRd, sft4
173  * Bitwise instructions               D32SLR XRa, XRb, XRc, XRd, sft4
174  * --------------------               D32SAR XRa, XRb, XRc, XRd, sft4
175  *                                    D32SARL XRa, XRb, XRc, sft4
176  *  S32NOR XRa, XRb, XRc              D32SLLV XRa, XRb, Rb
177  *  S32AND XRa, XRb, XRc              D32SLRV XRa, XRb, Rb
178  *  S32XOR XRa, XRb, XRc              D32SARV XRa, XRb, Rb
179  *  S32OR XRa, XRb, XRc               D32SARW XRa, XRb, XRc, Rb
180  *                                    Q16SLL XRa, XRb, XRc, XRd, sft4
181  *                                    Q16SLR XRa, XRb, XRc, XRd, sft4
182  * Miscellaneous instructions         Q16SAR XRa, XRb, XRc, XRd, sft4
183  * -------------------------          Q16SLLV XRa, XRb, Rb
184  *                                    Q16SLRV XRa, XRb, Rb
185  *  S32SFL XRa, XRb, XRc, XRd, optn2  Q16SARV XRa, XRb, Rb
186  *  S32ALN XRa, XRb, XRc, Rb
187  *  S32ALNI XRa, XRb, XRc, s3
188  *  S32LUI XRa, s8, optn3            Move instructions
189  *  S32EXTR XRa, XRb, Rb, bits5      -----------------
190  *  S32EXTRV XRa, XRb, Rs, Rt
191  *  Q16SCOP XRa, XRb, XRc, XRd        S32M2I XRa, Rb
192  *  Q16SAT XRa, XRb, XRc              S32I2M XRa, Rb
193  *
194  *
195  *     The opcode organization of MXU instructions
196  *     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
197  *
198  * The bits 31..26 of all MXU instructions are equal to 0x1C (also referred
199  * as opcode SPECIAL2 in the base MIPS ISA). The organization and meaning of
200  * other bits up to the instruction level is as follows:
201  *
202  *              bits
203  *             05..00
204  *
205  *          ┌─ 000000 ─ OPC_MXU_S32MADD
206  *          ├─ 000001 ─ OPC_MXU_S32MADDU
207  *          ├─ 000010 ─ <not assigned>   (non-MXU OPC_MUL)
208  *          │
209  *          │                               20..18
210  *          ├─ 000011 ─ OPC_MXU__POOL00 ─┬─ 000 ─ OPC_MXU_S32MAX
211  *          │                            ├─ 001 ─ OPC_MXU_S32MIN
212  *          │                            ├─ 010 ─ OPC_MXU_D16MAX
213  *          │                            ├─ 011 ─ OPC_MXU_D16MIN
214  *          │                            ├─ 100 ─ OPC_MXU_Q8MAX
215  *          │                            ├─ 101 ─ OPC_MXU_Q8MIN
216  *          │                            ├─ 110 ─ OPC_MXU_Q8SLT
217  *          │                            └─ 111 ─ OPC_MXU_Q8SLTU
218  *          ├─ 000100 ─ OPC_MXU_S32MSUB
219  *          ├─ 000101 ─ OPC_MXU_S32MSUBU    20..18
220  *          ├─ 000110 ─ OPC_MXU__POOL01 ─┬─ 000 ─ OPC_MXU_S32SLT
221  *          │                            ├─ 001 ─ OPC_MXU_D16SLT
222  *          │                            ├─ 010 ─ OPC_MXU_D16AVG
223  *          │                            ├─ 011 ─ OPC_MXU_D16AVGR
224  *          │                            ├─ 100 ─ OPC_MXU_Q8AVG
225  *          │                            ├─ 101 ─ OPC_MXU_Q8AVGR
226  *          │                            └─ 111 ─ OPC_MXU_Q8ADD
227  *          │
228  *          │                               20..18
229  *          ├─ 000111 ─ OPC_MXU__POOL02 ─┬─ 000 ─ OPC_MXU_S32CPS
230  *          │                            ├─ 010 ─ OPC_MXU_D16CPS
231  *          │                            ├─ 100 ─ OPC_MXU_Q8ABD
232  *          │                            └─ 110 ─ OPC_MXU_Q16SAT
233  *          ├─ 001000 ─ OPC_MXU_D16MUL
234  *          │                               25..24
235  *          ├─ 001001 ─ OPC_MXU__POOL03 ─┬─ 00 ─ OPC_MXU_D16MULF
236  *          │                            └─ 01 ─ OPC_MXU_D16MULE
237  *          ├─ 001010 ─ OPC_MXU_D16MAC
238  *          ├─ 001011 ─ OPC_MXU_D16MACF
239  *          ├─ 001100 ─ OPC_MXU_D16MADL
240  *          ├─ 001101 ─ OPC_MXU_S16MAD
241  *          ├─ 001110 ─ OPC_MXU_Q16ADD
242  *          ├─ 001111 ─ OPC_MXU_D16MACE     23
243  *          │                            ┌─ 0 ─ OPC_MXU_S32LDD
244  *          ├─ 010000 ─ OPC_MXU__POOL04 ─┴─ 1 ─ OPC_MXU_S32LDDR
245  *          │
246  *          │                               23
247  *          ├─ 010001 ─ OPC_MXU__POOL05 ─┬─ 0 ─ OPC_MXU_S32STD
248  *          │                            └─ 1 ─ OPC_MXU_S32STDR
249  *          │
250  *          │                               13..10
251  *          ├─ 010010 ─ OPC_MXU__POOL06 ─┬─ 0000 ─ OPC_MXU_S32LDDV
252  *          │                            └─ 0001 ─ OPC_MXU_S32LDDVR
253  *          │
254  *          │                               13..10
255  *          ├─ 010011 ─ OPC_MXU__POOL07 ─┬─ 0000 ─ OPC_MXU_S32STDV
256  *          │                            └─ 0001 ─ OPC_MXU_S32STDVR
257  *          │
258  *          │                               23
259  *          ├─ 010100 ─ OPC_MXU__POOL08 ─┬─ 0 ─ OPC_MXU_S32LDI
260  *          │                            └─ 1 ─ OPC_MXU_S32LDIR
261  *          │
262  *          │                               23
263  *          ├─ 010101 ─ OPC_MXU__POOL09 ─┬─ 0 ─ OPC_MXU_S32SDI
264  *          │                            └─ 1 ─ OPC_MXU_S32SDIR
265  *          │
266  *          │                               13..10
267  *          ├─ 010110 ─ OPC_MXU__POOL10 ─┬─ 0000 ─ OPC_MXU_S32LDIV
268  *          │                            └─ 0001 ─ OPC_MXU_S32LDIVR
269  *          │
270  *          │                               13..10
271  *          ├─ 010111 ─ OPC_MXU__POOL11 ─┬─ 0000 ─ OPC_MXU_S32SDIV
272  *          │                            └─ 0001 ─ OPC_MXU_S32SDIVR
273  *          ├─ 011000 ─ OPC_MXU_D32ADD
274  *          │                               23..22
275  *   MXU    ├─ 011001 ─ OPC_MXU__POOL12 ─┬─ 00 ─ OPC_MXU_D32ACC
276  * opcodes ─┤                            ├─ 01 ─ OPC_MXU_D32ACCM
277  *          │                            └─ 10 ─ OPC_MXU_D32ASUM
278  *          ├─ 011010 ─ <not assigned>
279  *          │                               23..22
280  *          ├─ 011011 ─ OPC_MXU__POOL13 ─┬─ 00 ─ OPC_MXU_Q16ACC
281  *          │                            ├─ 01 ─ OPC_MXU_Q16ACCM
282  *          │                            └─ 10 ─ OPC_MXU_Q16ASUM
283  *          │
284  *          │                               23..22
285  *          ├─ 011100 ─ OPC_MXU__POOL14 ─┬─ 00 ─ OPC_MXU_Q8ADDE
286  *          │                            ├─ 01 ─ OPC_MXU_D8SUM
287  *          ├─ 011101 ─ OPC_MXU_Q8ACCE   └─ 10 ─ OPC_MXU_D8SUMC
288  *          ├─ 011110 ─ <not assigned>
289  *          ├─ 011111 ─ <not assigned>
290  *          ├─ 100000 ─ <not assigned>   (overlaps with CLZ)
291  *          ├─ 100001 ─ <not assigned>   (overlaps with CLO)
292  *          ├─ 100010 ─ OPC_MXU_S8LDD
293  *          ├─ 100011 ─ OPC_MXU_S8STD       15..14
294  *          ├─ 100100 ─ OPC_MXU_S8LDI    ┌─ 00 ─ OPC_MXU_S32MUL
295  *          ├─ 100101 ─ OPC_MXU_S8SDI    ├─ 00 ─ OPC_MXU_S32MULU
296  *          │                            ├─ 00 ─ OPC_MXU_S32EXTR
297  *          ├─ 100110 ─ OPC_MXU__POOL15 ─┴─ 00 ─ OPC_MXU_S32EXTRV
298  *          │
299  *          │                               20..18
300  *          ├─ 100111 ─ OPC_MXU__POOL16 ─┬─ 000 ─ OPC_MXU_D32SARW
301  *          │                            ├─ 001 ─ OPC_MXU_S32ALN
302  *          │                            ├─ 010 ─ OPC_MXU_S32ALNI
303  *          │                            ├─ 011 ─ OPC_MXU_S32LUI
304  *          │                            ├─ 100 ─ OPC_MXU_S32NOR
305  *          │                            ├─ 101 ─ OPC_MXU_S32AND
306  *          │                            ├─ 110 ─ OPC_MXU_S32OR
307  *          │                            └─ 111 ─ OPC_MXU_S32XOR
308  *          │
309  *          │                               7..5
310  *          ├─ 101000 ─ OPC_MXU__POOL17 ─┬─ 000 ─ OPC_MXU_LXB
311  *          │                            ├─ 001 ─ OPC_MXU_LXH
312  *          ├─ 101001 ─ <not assigned>   ├─ 011 ─ OPC_MXU_LXW
313  *          ├─ 101010 ─ OPC_MXU_S16LDD   ├─ 100 ─ OPC_MXU_LXBU
314  *          ├─ 101011 ─ OPC_MXU_S16STD   └─ 101 ─ OPC_MXU_LXHU
315  *          ├─ 101100 ─ OPC_MXU_S16LDI
316  *          ├─ 101101 ─ OPC_MXU_S16SDI
317  *          ├─ 101110 ─ OPC_MXU_S32M2I
318  *          ├─ 101111 ─ OPC_MXU_S32I2M
319  *          ├─ 110000 ─ OPC_MXU_D32SLL
320  *          ├─ 110001 ─ OPC_MXU_D32SLR      20..18
321  *          ├─ 110010 ─ OPC_MXU_D32SARL  ┌─ 000 ─ OPC_MXU_D32SLLV
322  *          ├─ 110011 ─ OPC_MXU_D32SAR   ├─ 001 ─ OPC_MXU_D32SLRV
323  *          ├─ 110100 ─ OPC_MXU_Q16SLL   ├─ 010 ─ OPC_MXU_D32SARV
324  *          ├─ 110101 ─ OPC_MXU_Q16SLR   ├─ 011 ─ OPC_MXU_Q16SLLV
325  *          │                            ├─ 100 ─ OPC_MXU_Q16SLRV
326  *          ├─ 110110 ─ OPC_MXU__POOL18 ─┴─ 101 ─ OPC_MXU_Q16SARV
327  *          │
328  *          ├─ 110111 ─ OPC_MXU_Q16SAR
329  *          │                               23..22
330  *          ├─ 111000 ─ OPC_MXU__POOL19 ─┬─ 00 ─ OPC_MXU_Q8MUL
331  *          │                            └─ 01 ─ OPC_MXU_Q8MULSU
332  *          │
333  *          │                               20..18
334  *          ├─ 111001 ─ OPC_MXU__POOL20 ─┬─ 000 ─ OPC_MXU_Q8MOVZ
335  *          │                            ├─ 001 ─ OPC_MXU_Q8MOVN
336  *          │                            ├─ 010 ─ OPC_MXU_D16MOVZ
337  *          │                            ├─ 011 ─ OPC_MXU_D16MOVN
338  *          │                            ├─ 100 ─ OPC_MXU_S32MOVZ
339  *          │                            └─ 101 ─ OPC_MXU_S32MOVN
340  *          │
341  *          │                               23..22
342  *          ├─ 111010 ─ OPC_MXU__POOL21 ─┬─ 00 ─ OPC_MXU_Q8MAC
343  *          │                            └─ 10 ─ OPC_MXU_Q8MACSU
344  *          ├─ 111011 ─ OPC_MXU_Q16SCOP
345  *          ├─ 111100 ─ OPC_MXU_Q8MADL
346  *          ├─ 111101 ─ OPC_MXU_S32SFL
347  *          ├─ 111110 ─ OPC_MXU_Q8SAD
348  *          └─ 111111 ─ <not assigned>   (overlaps with SDBBP)
349  *
350  *
351  * Compiled after:
352  *
353  *   "XBurst® Instruction Set Architecture MIPS eXtension/enhanced Unit
354  *   Programming Manual", Ingenic Semiconductor Co, Ltd., revision June 2, 2017
355  */
356 
357 enum {
358     OPC_MXU__POOL00  = 0x03,
359     OPC_MXU_D16MUL   = 0x08,
360     OPC_MXU_D16MAC   = 0x0A,
361     OPC_MXU__POOL04  = 0x10,
362     OPC_MXU_S8LDD    = 0x22,
363     OPC_MXU__POOL16  = 0x27,
364     OPC_MXU_S32M2I   = 0x2E,
365     OPC_MXU_S32I2M   = 0x2F,
366     OPC_MXU__POOL19  = 0x38,
367 };
368 
369 
370 /*
371  * MXU pool 00
372  */
373 enum {
374     OPC_MXU_S32MAX   = 0x00,
375     OPC_MXU_S32MIN   = 0x01,
376     OPC_MXU_D16MAX   = 0x02,
377     OPC_MXU_D16MIN   = 0x03,
378     OPC_MXU_Q8MAX    = 0x04,
379     OPC_MXU_Q8MIN    = 0x05,
380 };
381 
382 /*
383  * MXU pool 04
384  */
385 enum {
386     OPC_MXU_S32LDD   = 0x00,
387     OPC_MXU_S32LDDR  = 0x01,
388 };
389 
390 /*
391  * MXU pool 16
392  */
393 enum {
394     OPC_MXU_S32ALNI  = 0x02,
395     OPC_MXU_S32NOR   = 0x04,
396     OPC_MXU_S32AND   = 0x05,
397     OPC_MXU_S32OR    = 0x06,
398     OPC_MXU_S32XOR   = 0x07,
399 };
400 
401 /*
402  * MXU pool 19
403  */
404 enum {
405     OPC_MXU_Q8MUL    = 0x00,
406     OPC_MXU_Q8MULSU  = 0x01,
407 };
408 
409 /* MXU accumulate add/subtract 1-bit pattern 'aptn1' */
410 #define MXU_APTN1_A    0
411 #define MXU_APTN1_S    1
412 
413 /* MXU accumulate add/subtract 2-bit pattern 'aptn2' */
414 #define MXU_APTN2_AA    0
415 #define MXU_APTN2_AS    1
416 #define MXU_APTN2_SA    2
417 #define MXU_APTN2_SS    3
418 
419 /* MXU execute add/subtract 2-bit pattern 'eptn2' */
420 #define MXU_EPTN2_AA    0
421 #define MXU_EPTN2_AS    1
422 #define MXU_EPTN2_SA    2
423 #define MXU_EPTN2_SS    3
424 
425 /* MXU operand getting pattern 'optn2' */
426 #define MXU_OPTN2_PTN0  0
427 #define MXU_OPTN2_PTN1  1
428 #define MXU_OPTN2_PTN2  2
429 #define MXU_OPTN2_PTN3  3
430 /* alternative naming scheme for 'optn2' */
431 #define MXU_OPTN2_WW    0
432 #define MXU_OPTN2_LW    1
433 #define MXU_OPTN2_HW    2
434 #define MXU_OPTN2_XW    3
435 
436 /* MXU operand getting pattern 'optn3' */
437 #define MXU_OPTN3_PTN0  0
438 #define MXU_OPTN3_PTN1  1
439 #define MXU_OPTN3_PTN2  2
440 #define MXU_OPTN3_PTN3  3
441 #define MXU_OPTN3_PTN4  4
442 #define MXU_OPTN3_PTN5  5
443 #define MXU_OPTN3_PTN6  6
444 #define MXU_OPTN3_PTN7  7
445 
446 /* MXU registers */
447 static TCGv mxu_gpr[NUMBER_OF_MXU_REGISTERS - 1];
448 static TCGv mxu_CR;
449 
450 static const char mxuregnames[][4] = {
451     "XR1",  "XR2",  "XR3",  "XR4",  "XR5",  "XR6",  "XR7",  "XR8",
452     "XR9",  "XR10", "XR11", "XR12", "XR13", "XR14", "XR15", "XCR",
453 };
454 
455 void mxu_translate_init(void)
456 {
457     for (unsigned i = 0; i < NUMBER_OF_MXU_REGISTERS - 1; i++) {
458         mxu_gpr[i] = tcg_global_mem_new(cpu_env,
459                                         offsetof(CPUMIPSState, active_tc.mxu_gpr[i]),
460                                         mxuregnames[i]);
461     }
462 
463     mxu_CR = tcg_global_mem_new(cpu_env,
464                                 offsetof(CPUMIPSState, active_tc.mxu_cr),
465                                 mxuregnames[NUMBER_OF_MXU_REGISTERS - 1]);
466 }
467 
468 /* MXU General purpose registers moves. */
469 static inline void gen_load_mxu_gpr(TCGv t, unsigned int reg)
470 {
471     if (reg == 0) {
472         tcg_gen_movi_tl(t, 0);
473     } else if (reg <= 15) {
474         tcg_gen_mov_tl(t, mxu_gpr[reg - 1]);
475     }
476 }
477 
478 static inline void gen_store_mxu_gpr(TCGv t, unsigned int reg)
479 {
480     if (reg > 0 && reg <= 15) {
481         tcg_gen_mov_tl(mxu_gpr[reg - 1], t);
482     }
483 }
484 
485 /* MXU control register moves. */
486 static inline void gen_load_mxu_cr(TCGv t)
487 {
488     tcg_gen_mov_tl(t, mxu_CR);
489 }
490 
491 static inline void gen_store_mxu_cr(TCGv t)
492 {
493     /* TODO: Add handling of RW rules for MXU_CR. */
494     tcg_gen_mov_tl(mxu_CR, t);
495 }
496 
497 /*
498  * S32I2M XRa, rb - Register move from GRF to XRF
499  */
500 static void gen_mxu_s32i2m(DisasContext *ctx)
501 {
502     TCGv t0;
503     uint32_t XRa, Rb;
504 
505     t0 = tcg_temp_new();
506 
507     XRa = extract32(ctx->opcode, 6, 5);
508     Rb = extract32(ctx->opcode, 16, 5);
509 
510     gen_load_gpr(t0, Rb);
511     if (XRa <= 15) {
512         gen_store_mxu_gpr(t0, XRa);
513     } else if (XRa == 16) {
514         gen_store_mxu_cr(t0);
515     }
516 
517     tcg_temp_free(t0);
518 }
519 
520 /*
521  * S32M2I XRa, rb - Register move from XRF to GRF
522  */
523 static void gen_mxu_s32m2i(DisasContext *ctx)
524 {
525     TCGv t0;
526     uint32_t XRa, Rb;
527 
528     t0 = tcg_temp_new();
529 
530     XRa = extract32(ctx->opcode, 6, 5);
531     Rb = extract32(ctx->opcode, 16, 5);
532 
533     if (XRa <= 15) {
534         gen_load_mxu_gpr(t0, XRa);
535     } else if (XRa == 16) {
536         gen_load_mxu_cr(t0);
537     }
538 
539     gen_store_gpr(t0, Rb);
540 
541     tcg_temp_free(t0);
542 }
543 
544 /*
545  * S8LDD XRa, Rb, s8, optn3 - Load a byte from memory to XRF
546  */
547 static void gen_mxu_s8ldd(DisasContext *ctx)
548 {
549     TCGv t0, t1;
550     uint32_t XRa, Rb, s8, optn3;
551 
552     t0 = tcg_temp_new();
553     t1 = tcg_temp_new();
554 
555     XRa = extract32(ctx->opcode, 6, 4);
556     s8 = extract32(ctx->opcode, 10, 8);
557     optn3 = extract32(ctx->opcode, 18, 3);
558     Rb = extract32(ctx->opcode, 21, 5);
559 
560     gen_load_gpr(t0, Rb);
561     tcg_gen_addi_tl(t0, t0, (int8_t)s8);
562 
563     switch (optn3) {
564     /* XRa[7:0] = tmp8 */
565     case MXU_OPTN3_PTN0:
566         tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_UB);
567         gen_load_mxu_gpr(t0, XRa);
568         tcg_gen_deposit_tl(t0, t0, t1, 0, 8);
569         break;
570     /* XRa[15:8] = tmp8 */
571     case MXU_OPTN3_PTN1:
572         tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_UB);
573         gen_load_mxu_gpr(t0, XRa);
574         tcg_gen_deposit_tl(t0, t0, t1, 8, 8);
575         break;
576     /* XRa[23:16] = tmp8 */
577     case MXU_OPTN3_PTN2:
578         tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_UB);
579         gen_load_mxu_gpr(t0, XRa);
580         tcg_gen_deposit_tl(t0, t0, t1, 16, 8);
581         break;
582     /* XRa[31:24] = tmp8 */
583     case MXU_OPTN3_PTN3:
584         tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_UB);
585         gen_load_mxu_gpr(t0, XRa);
586         tcg_gen_deposit_tl(t0, t0, t1, 24, 8);
587         break;
588     /* XRa = {8'b0, tmp8, 8'b0, tmp8} */
589     case MXU_OPTN3_PTN4:
590         tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_UB);
591         tcg_gen_deposit_tl(t0, t1, t1, 16, 16);
592         break;
593     /* XRa = {tmp8, 8'b0, tmp8, 8'b0} */
594     case MXU_OPTN3_PTN5:
595         tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_UB);
596         tcg_gen_shli_tl(t1, t1, 8);
597         tcg_gen_deposit_tl(t0, t1, t1, 16, 16);
598         break;
599     /* XRa = {{8{sign of tmp8}}, tmp8, {8{sign of tmp8}}, tmp8} */
600     case MXU_OPTN3_PTN6:
601         tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_SB);
602         tcg_gen_mov_tl(t0, t1);
603         tcg_gen_andi_tl(t0, t0, 0xFF00FFFF);
604         tcg_gen_shli_tl(t1, t1, 16);
605         tcg_gen_or_tl(t0, t0, t1);
606         break;
607     /* XRa = {tmp8, tmp8, tmp8, tmp8} */
608     case MXU_OPTN3_PTN7:
609         tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_UB);
610         tcg_gen_deposit_tl(t1, t1, t1, 8, 8);
611         tcg_gen_deposit_tl(t0, t1, t1, 16, 16);
612         break;
613     }
614 
615     gen_store_mxu_gpr(t0, XRa);
616 
617     tcg_temp_free(t0);
618     tcg_temp_free(t1);
619 }
620 
621 /*
622  * D16MUL XRa, XRb, XRc, XRd, optn2 - Signed 16 bit pattern multiplication
623  */
624 static void gen_mxu_d16mul(DisasContext *ctx)
625 {
626     TCGv t0, t1, t2, t3;
627     uint32_t XRa, XRb, XRc, XRd, optn2;
628 
629     t0 = tcg_temp_new();
630     t1 = tcg_temp_new();
631     t2 = tcg_temp_new();
632     t3 = tcg_temp_new();
633 
634     XRa = extract32(ctx->opcode, 6, 4);
635     XRb = extract32(ctx->opcode, 10, 4);
636     XRc = extract32(ctx->opcode, 14, 4);
637     XRd = extract32(ctx->opcode, 18, 4);
638     optn2 = extract32(ctx->opcode, 22, 2);
639 
640     gen_load_mxu_gpr(t1, XRb);
641     tcg_gen_sextract_tl(t0, t1, 0, 16);
642     tcg_gen_sextract_tl(t1, t1, 16, 16);
643     gen_load_mxu_gpr(t3, XRc);
644     tcg_gen_sextract_tl(t2, t3, 0, 16);
645     tcg_gen_sextract_tl(t3, t3, 16, 16);
646 
647     switch (optn2) {
648     case MXU_OPTN2_WW: /* XRB.H*XRC.H == lop, XRB.L*XRC.L == rop */
649         tcg_gen_mul_tl(t3, t1, t3);
650         tcg_gen_mul_tl(t2, t0, t2);
651         break;
652     case MXU_OPTN2_LW: /* XRB.L*XRC.H == lop, XRB.L*XRC.L == rop */
653         tcg_gen_mul_tl(t3, t0, t3);
654         tcg_gen_mul_tl(t2, t0, t2);
655         break;
656     case MXU_OPTN2_HW: /* XRB.H*XRC.H == lop, XRB.H*XRC.L == rop */
657         tcg_gen_mul_tl(t3, t1, t3);
658         tcg_gen_mul_tl(t2, t1, t2);
659         break;
660     case MXU_OPTN2_XW: /* XRB.L*XRC.H == lop, XRB.H*XRC.L == rop */
661         tcg_gen_mul_tl(t3, t0, t3);
662         tcg_gen_mul_tl(t2, t1, t2);
663         break;
664     }
665     gen_store_mxu_gpr(t3, XRa);
666     gen_store_mxu_gpr(t2, XRd);
667 
668     tcg_temp_free(t0);
669     tcg_temp_free(t1);
670     tcg_temp_free(t2);
671     tcg_temp_free(t3);
672 }
673 
674 /*
675  * D16MAC XRa, XRb, XRc, XRd, aptn2, optn2 - Signed 16 bit pattern multiply
676  *                                           and accumulate
677  */
678 static void gen_mxu_d16mac(DisasContext *ctx)
679 {
680     TCGv t0, t1, t2, t3;
681     uint32_t XRa, XRb, XRc, XRd, optn2, aptn2;
682 
683     t0 = tcg_temp_new();
684     t1 = tcg_temp_new();
685     t2 = tcg_temp_new();
686     t3 = tcg_temp_new();
687 
688     XRa = extract32(ctx->opcode, 6, 4);
689     XRb = extract32(ctx->opcode, 10, 4);
690     XRc = extract32(ctx->opcode, 14, 4);
691     XRd = extract32(ctx->opcode, 18, 4);
692     optn2 = extract32(ctx->opcode, 22, 2);
693     aptn2 = extract32(ctx->opcode, 24, 2);
694 
695     gen_load_mxu_gpr(t1, XRb);
696     tcg_gen_sextract_tl(t0, t1, 0, 16);
697     tcg_gen_sextract_tl(t1, t1, 16, 16);
698 
699     gen_load_mxu_gpr(t3, XRc);
700     tcg_gen_sextract_tl(t2, t3, 0, 16);
701     tcg_gen_sextract_tl(t3, t3, 16, 16);
702 
703     switch (optn2) {
704     case MXU_OPTN2_WW: /* XRB.H*XRC.H == lop, XRB.L*XRC.L == rop */
705         tcg_gen_mul_tl(t3, t1, t3);
706         tcg_gen_mul_tl(t2, t0, t2);
707         break;
708     case MXU_OPTN2_LW: /* XRB.L*XRC.H == lop, XRB.L*XRC.L == rop */
709         tcg_gen_mul_tl(t3, t0, t3);
710         tcg_gen_mul_tl(t2, t0, t2);
711         break;
712     case MXU_OPTN2_HW: /* XRB.H*XRC.H == lop, XRB.H*XRC.L == rop */
713         tcg_gen_mul_tl(t3, t1, t3);
714         tcg_gen_mul_tl(t2, t1, t2);
715         break;
716     case MXU_OPTN2_XW: /* XRB.L*XRC.H == lop, XRB.H*XRC.L == rop */
717         tcg_gen_mul_tl(t3, t0, t3);
718         tcg_gen_mul_tl(t2, t1, t2);
719         break;
720     }
721     gen_load_mxu_gpr(t0, XRa);
722     gen_load_mxu_gpr(t1, XRd);
723 
724     switch (aptn2) {
725     case MXU_APTN2_AA:
726         tcg_gen_add_tl(t3, t0, t3);
727         tcg_gen_add_tl(t2, t1, t2);
728         break;
729     case MXU_APTN2_AS:
730         tcg_gen_add_tl(t3, t0, t3);
731         tcg_gen_sub_tl(t2, t1, t2);
732         break;
733     case MXU_APTN2_SA:
734         tcg_gen_sub_tl(t3, t0, t3);
735         tcg_gen_add_tl(t2, t1, t2);
736         break;
737     case MXU_APTN2_SS:
738         tcg_gen_sub_tl(t3, t0, t3);
739         tcg_gen_sub_tl(t2, t1, t2);
740         break;
741     }
742     gen_store_mxu_gpr(t3, XRa);
743     gen_store_mxu_gpr(t2, XRd);
744 
745     tcg_temp_free(t0);
746     tcg_temp_free(t1);
747     tcg_temp_free(t2);
748     tcg_temp_free(t3);
749 }
750 
751 /*
752  * Q8MUL   XRa, XRb, XRc, XRd - Parallel unsigned 8 bit pattern multiply
753  * Q8MULSU XRa, XRb, XRc, XRd - Parallel signed 8 bit pattern multiply
754  */
755 static void gen_mxu_q8mul_q8mulsu(DisasContext *ctx)
756 {
757     TCGv t0, t1, t2, t3, t4, t5, t6, t7;
758     uint32_t XRa, XRb, XRc, XRd, sel;
759 
760     t0 = tcg_temp_new();
761     t1 = tcg_temp_new();
762     t2 = tcg_temp_new();
763     t3 = tcg_temp_new();
764     t4 = tcg_temp_new();
765     t5 = tcg_temp_new();
766     t6 = tcg_temp_new();
767     t7 = tcg_temp_new();
768 
769     XRa = extract32(ctx->opcode, 6, 4);
770     XRb = extract32(ctx->opcode, 10, 4);
771     XRc = extract32(ctx->opcode, 14, 4);
772     XRd = extract32(ctx->opcode, 18, 4);
773     sel = extract32(ctx->opcode, 22, 2);
774 
775     gen_load_mxu_gpr(t3, XRb);
776     gen_load_mxu_gpr(t7, XRc);
777 
778     if (sel == 0x2) {
779         /* Q8MULSU */
780         tcg_gen_ext8s_tl(t0, t3);
781         tcg_gen_shri_tl(t3, t3, 8);
782         tcg_gen_ext8s_tl(t1, t3);
783         tcg_gen_shri_tl(t3, t3, 8);
784         tcg_gen_ext8s_tl(t2, t3);
785         tcg_gen_shri_tl(t3, t3, 8);
786         tcg_gen_ext8s_tl(t3, t3);
787     } else {
788         /* Q8MUL */
789         tcg_gen_ext8u_tl(t0, t3);
790         tcg_gen_shri_tl(t3, t3, 8);
791         tcg_gen_ext8u_tl(t1, t3);
792         tcg_gen_shri_tl(t3, t3, 8);
793         tcg_gen_ext8u_tl(t2, t3);
794         tcg_gen_shri_tl(t3, t3, 8);
795         tcg_gen_ext8u_tl(t3, t3);
796     }
797 
798     tcg_gen_ext8u_tl(t4, t7);
799     tcg_gen_shri_tl(t7, t7, 8);
800     tcg_gen_ext8u_tl(t5, t7);
801     tcg_gen_shri_tl(t7, t7, 8);
802     tcg_gen_ext8u_tl(t6, t7);
803     tcg_gen_shri_tl(t7, t7, 8);
804     tcg_gen_ext8u_tl(t7, t7);
805 
806     tcg_gen_mul_tl(t0, t0, t4);
807     tcg_gen_mul_tl(t1, t1, t5);
808     tcg_gen_mul_tl(t2, t2, t6);
809     tcg_gen_mul_tl(t3, t3, t7);
810 
811     tcg_gen_andi_tl(t0, t0, 0xFFFF);
812     tcg_gen_andi_tl(t1, t1, 0xFFFF);
813     tcg_gen_andi_tl(t2, t2, 0xFFFF);
814     tcg_gen_andi_tl(t3, t3, 0xFFFF);
815 
816     tcg_gen_shli_tl(t1, t1, 16);
817     tcg_gen_shli_tl(t3, t3, 16);
818 
819     tcg_gen_or_tl(t0, t0, t1);
820     tcg_gen_or_tl(t1, t2, t3);
821 
822     gen_store_mxu_gpr(t0, XRd);
823     gen_store_mxu_gpr(t1, XRa);
824 
825     tcg_temp_free(t0);
826     tcg_temp_free(t1);
827     tcg_temp_free(t2);
828     tcg_temp_free(t3);
829     tcg_temp_free(t4);
830     tcg_temp_free(t5);
831     tcg_temp_free(t6);
832     tcg_temp_free(t7);
833 }
834 
835 /*
836  * S32LDD  XRa, Rb, S12 - Load a word from memory to XRF
837  * S32LDDR XRa, Rb, S12 - Load a word from memory to XRF, reversed byte seq.
838  */
839 static void gen_mxu_s32ldd_s32lddr(DisasContext *ctx)
840 {
841     TCGv t0, t1;
842     uint32_t XRa, Rb, s12, sel;
843 
844     t0 = tcg_temp_new();
845     t1 = tcg_temp_new();
846 
847     XRa = extract32(ctx->opcode, 6, 4);
848     s12 = extract32(ctx->opcode, 10, 10);
849     sel = extract32(ctx->opcode, 20, 1);
850     Rb = extract32(ctx->opcode, 21, 5);
851 
852     gen_load_gpr(t0, Rb);
853 
854     tcg_gen_movi_tl(t1, s12);
855     tcg_gen_shli_tl(t1, t1, 2);
856     if (s12 & 0x200) {
857         tcg_gen_ori_tl(t1, t1, 0xFFFFF000);
858     }
859     tcg_gen_add_tl(t1, t0, t1);
860     tcg_gen_qemu_ld_tl(t1, t1, ctx->mem_idx, MO_TESL ^ (sel * MO_BSWAP));
861 
862     gen_store_mxu_gpr(t1, XRa);
863 
864     tcg_temp_free(t0);
865     tcg_temp_free(t1);
866 }
867 
868 
869 /*
870  *                 MXU instruction category: logic
871  *                 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
872  *
873  *               S32NOR    S32AND    S32OR    S32XOR
874  */
875 
876 /*
877  *  S32NOR XRa, XRb, XRc
878  *    Update XRa with the result of logical bitwise 'nor' operation
879  *    applied to the content of XRb and XRc.
880  */
881 static void gen_mxu_S32NOR(DisasContext *ctx)
882 {
883     uint32_t pad, XRc, XRb, XRa;
884 
885     pad = extract32(ctx->opcode, 21, 5);
886     XRc = extract32(ctx->opcode, 14, 4);
887     XRb = extract32(ctx->opcode, 10, 4);
888     XRa = extract32(ctx->opcode,  6, 4);
889 
890     if (unlikely(pad != 0)) {
891         /* opcode padding incorrect -> do nothing */
892     } else if (unlikely(XRa == 0)) {
893         /* destination is zero register -> do nothing */
894     } else if (unlikely((XRb == 0) && (XRc == 0))) {
895         /* both operands zero registers -> just set destination to all 1s */
896         tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0xFFFFFFFF);
897     } else if (unlikely(XRb == 0)) {
898         /* XRb zero register -> just set destination to the negation of XRc */
899         tcg_gen_not_i32(mxu_gpr[XRa - 1], mxu_gpr[XRc - 1]);
900     } else if (unlikely(XRc == 0)) {
901         /* XRa zero register -> just set destination to the negation of XRb */
902         tcg_gen_not_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
903     } else if (unlikely(XRb == XRc)) {
904         /* both operands same -> just set destination to the negation of XRb */
905         tcg_gen_not_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
906     } else {
907         /* the most general case */
908         tcg_gen_nor_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1], mxu_gpr[XRc - 1]);
909     }
910 }
911 
912 /*
913  *  S32AND XRa, XRb, XRc
914  *    Update XRa with the result of logical bitwise 'and' operation
915  *    applied to the content of XRb and XRc.
916  */
917 static void gen_mxu_S32AND(DisasContext *ctx)
918 {
919     uint32_t pad, XRc, XRb, XRa;
920 
921     pad = extract32(ctx->opcode, 21, 5);
922     XRc = extract32(ctx->opcode, 14, 4);
923     XRb = extract32(ctx->opcode, 10, 4);
924     XRa = extract32(ctx->opcode,  6, 4);
925 
926     if (unlikely(pad != 0)) {
927         /* opcode padding incorrect -> do nothing */
928     } else if (unlikely(XRa == 0)) {
929         /* destination is zero register -> do nothing */
930     } else if (unlikely((XRb == 0) || (XRc == 0))) {
931         /* one of operands zero register -> just set destination to all 0s */
932         tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
933     } else if (unlikely(XRb == XRc)) {
934         /* both operands same -> just set destination to one of them */
935         tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
936     } else {
937         /* the most general case */
938         tcg_gen_and_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1], mxu_gpr[XRc - 1]);
939     }
940 }
941 
942 /*
943  *  S32OR XRa, XRb, XRc
944  *    Update XRa with the result of logical bitwise 'or' operation
945  *    applied to the content of XRb and XRc.
946  */
947 static void gen_mxu_S32OR(DisasContext *ctx)
948 {
949     uint32_t pad, XRc, XRb, XRa;
950 
951     pad = extract32(ctx->opcode, 21, 5);
952     XRc = extract32(ctx->opcode, 14, 4);
953     XRb = extract32(ctx->opcode, 10, 4);
954     XRa = extract32(ctx->opcode,  6, 4);
955 
956     if (unlikely(pad != 0)) {
957         /* opcode padding incorrect -> do nothing */
958     } else if (unlikely(XRa == 0)) {
959         /* destination is zero register -> do nothing */
960     } else if (unlikely((XRb == 0) && (XRc == 0))) {
961         /* both operands zero registers -> just set destination to all 0s */
962         tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
963     } else if (unlikely(XRb == 0)) {
964         /* XRb zero register -> just set destination to the content of XRc */
965         tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRc - 1]);
966     } else if (unlikely(XRc == 0)) {
967         /* XRc zero register -> just set destination to the content of XRb */
968         tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
969     } else if (unlikely(XRb == XRc)) {
970         /* both operands same -> just set destination to one of them */
971         tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
972     } else {
973         /* the most general case */
974         tcg_gen_or_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1], mxu_gpr[XRc - 1]);
975     }
976 }
977 
978 /*
979  *  S32XOR XRa, XRb, XRc
980  *    Update XRa with the result of logical bitwise 'xor' operation
981  *    applied to the content of XRb and XRc.
982  */
983 static void gen_mxu_S32XOR(DisasContext *ctx)
984 {
985     uint32_t pad, XRc, XRb, XRa;
986 
987     pad = extract32(ctx->opcode, 21, 5);
988     XRc = extract32(ctx->opcode, 14, 4);
989     XRb = extract32(ctx->opcode, 10, 4);
990     XRa = extract32(ctx->opcode,  6, 4);
991 
992     if (unlikely(pad != 0)) {
993         /* opcode padding incorrect -> do nothing */
994     } else if (unlikely(XRa == 0)) {
995         /* destination is zero register -> do nothing */
996     } else if (unlikely((XRb == 0) && (XRc == 0))) {
997         /* both operands zero registers -> just set destination to all 0s */
998         tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
999     } else if (unlikely(XRb == 0)) {
1000         /* XRb zero register -> just set destination to the content of XRc */
1001         tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRc - 1]);
1002     } else if (unlikely(XRc == 0)) {
1003         /* XRc zero register -> just set destination to the content of XRb */
1004         tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
1005     } else if (unlikely(XRb == XRc)) {
1006         /* both operands same -> just set destination to all 0s */
1007         tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
1008     } else {
1009         /* the most general case */
1010         tcg_gen_xor_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1], mxu_gpr[XRc - 1]);
1011     }
1012 }
1013 
1014 
1015 /*
1016  *                   MXU instruction category max/min
1017  *                   ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1018  *
1019  *                     S32MAX     D16MAX     Q8MAX
1020  *                     S32MIN     D16MIN     Q8MIN
1021  */
1022 
1023 /*
1024  *  S32MAX XRa, XRb, XRc
1025  *    Update XRa with the maximum of signed 32-bit integers contained
1026  *    in XRb and XRc.
1027  *
1028  *  S32MIN XRa, XRb, XRc
1029  *    Update XRa with the minimum of signed 32-bit integers contained
1030  *    in XRb and XRc.
1031  */
1032 static void gen_mxu_S32MAX_S32MIN(DisasContext *ctx)
1033 {
1034     uint32_t pad, opc, XRc, XRb, XRa;
1035 
1036     pad = extract32(ctx->opcode, 21, 5);
1037     opc = extract32(ctx->opcode, 18, 3);
1038     XRc = extract32(ctx->opcode, 14, 4);
1039     XRb = extract32(ctx->opcode, 10, 4);
1040     XRa = extract32(ctx->opcode,  6, 4);
1041 
1042     if (unlikely(pad != 0)) {
1043         /* opcode padding incorrect -> do nothing */
1044     } else if (unlikely(XRa == 0)) {
1045         /* destination is zero register -> do nothing */
1046     } else if (unlikely((XRb == 0) && (XRc == 0))) {
1047         /* both operands zero registers -> just set destination to zero */
1048         tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
1049     } else if (unlikely((XRb == 0) || (XRc == 0))) {
1050         /* exactly one operand is zero register - find which one is not...*/
1051         uint32_t XRx = XRb ? XRb : XRc;
1052         /* ...and do max/min operation with one operand 0 */
1053         if (opc == OPC_MXU_S32MAX) {
1054             tcg_gen_smax_i32(mxu_gpr[XRa - 1], mxu_gpr[XRx - 1], 0);
1055         } else {
1056             tcg_gen_smin_i32(mxu_gpr[XRa - 1], mxu_gpr[XRx - 1], 0);
1057         }
1058     } else if (unlikely(XRb == XRc)) {
1059         /* both operands same -> just set destination to one of them */
1060         tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
1061     } else {
1062         /* the most general case */
1063         if (opc == OPC_MXU_S32MAX) {
1064             tcg_gen_smax_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1],
1065                                                mxu_gpr[XRc - 1]);
1066         } else {
1067             tcg_gen_smin_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1],
1068                                                mxu_gpr[XRc - 1]);
1069         }
1070     }
1071 }
1072 
1073 /*
1074  *  D16MAX
1075  *    Update XRa with the 16-bit-wise maximums of signed integers
1076  *    contained in XRb and XRc.
1077  *
1078  *  D16MIN
1079  *    Update XRa with the 16-bit-wise minimums of signed integers
1080  *    contained in XRb and XRc.
1081  */
1082 static void gen_mxu_D16MAX_D16MIN(DisasContext *ctx)
1083 {
1084     uint32_t pad, opc, XRc, XRb, XRa;
1085 
1086     pad = extract32(ctx->opcode, 21, 5);
1087     opc = extract32(ctx->opcode, 18, 3);
1088     XRc = extract32(ctx->opcode, 14, 4);
1089     XRb = extract32(ctx->opcode, 10, 4);
1090     XRa = extract32(ctx->opcode,  6, 4);
1091 
1092     if (unlikely(pad != 0)) {
1093         /* opcode padding incorrect -> do nothing */
1094     } else if (unlikely(XRa == 0)) {
1095         /* destination is zero register -> do nothing */
1096     } else if (unlikely((XRb == 0) && (XRc == 0))) {
1097         /* both operands zero registers -> just set destination to zero */
1098         tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
1099     } else if (unlikely((XRb == 0) || (XRc == 0))) {
1100         /* exactly one operand is zero register - find which one is not...*/
1101         uint32_t XRx = XRb ? XRb : XRc;
1102         /* ...and do half-word-wise max/min with one operand 0 */
1103         TCGv_i32 t0 = tcg_temp_new();
1104         TCGv_i32 t1 = tcg_const_i32(0);
1105 
1106         /* the left half-word first */
1107         tcg_gen_andi_i32(t0, mxu_gpr[XRx - 1], 0xFFFF0000);
1108         if (opc == OPC_MXU_D16MAX) {
1109             tcg_gen_smax_i32(mxu_gpr[XRa - 1], t0, t1);
1110         } else {
1111             tcg_gen_smin_i32(mxu_gpr[XRa - 1], t0, t1);
1112         }
1113 
1114         /* the right half-word */
1115         tcg_gen_andi_i32(t0, mxu_gpr[XRx - 1], 0x0000FFFF);
1116         /* move half-words to the leftmost position */
1117         tcg_gen_shli_i32(t0, t0, 16);
1118         /* t0 will be max/min of t0 and t1 */
1119         if (opc == OPC_MXU_D16MAX) {
1120             tcg_gen_smax_i32(t0, t0, t1);
1121         } else {
1122             tcg_gen_smin_i32(t0, t0, t1);
1123         }
1124         /* return resulting half-words to its original position */
1125         tcg_gen_shri_i32(t0, t0, 16);
1126         /* finally update the destination */
1127         tcg_gen_or_i32(mxu_gpr[XRa - 1], mxu_gpr[XRa - 1], t0);
1128 
1129         tcg_temp_free(t1);
1130         tcg_temp_free(t0);
1131     } else if (unlikely(XRb == XRc)) {
1132         /* both operands same -> just set destination to one of them */
1133         tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
1134     } else {
1135         /* the most general case */
1136         TCGv_i32 t0 = tcg_temp_new();
1137         TCGv_i32 t1 = tcg_temp_new();
1138 
1139         /* the left half-word first */
1140         tcg_gen_andi_i32(t0, mxu_gpr[XRb - 1], 0xFFFF0000);
1141         tcg_gen_andi_i32(t1, mxu_gpr[XRc - 1], 0xFFFF0000);
1142         if (opc == OPC_MXU_D16MAX) {
1143             tcg_gen_smax_i32(mxu_gpr[XRa - 1], t0, t1);
1144         } else {
1145             tcg_gen_smin_i32(mxu_gpr[XRa - 1], t0, t1);
1146         }
1147 
1148         /* the right half-word */
1149         tcg_gen_andi_i32(t0, mxu_gpr[XRb - 1], 0x0000FFFF);
1150         tcg_gen_andi_i32(t1, mxu_gpr[XRc - 1], 0x0000FFFF);
1151         /* move half-words to the leftmost position */
1152         tcg_gen_shli_i32(t0, t0, 16);
1153         tcg_gen_shli_i32(t1, t1, 16);
1154         /* t0 will be max/min of t0 and t1 */
1155         if (opc == OPC_MXU_D16MAX) {
1156             tcg_gen_smax_i32(t0, t0, t1);
1157         } else {
1158             tcg_gen_smin_i32(t0, t0, t1);
1159         }
1160         /* return resulting half-words to its original position */
1161         tcg_gen_shri_i32(t0, t0, 16);
1162         /* finally update the destination */
1163         tcg_gen_or_i32(mxu_gpr[XRa - 1], mxu_gpr[XRa - 1], t0);
1164 
1165         tcg_temp_free(t1);
1166         tcg_temp_free(t0);
1167     }
1168 }
1169 
1170 /*
1171  *  Q8MAX
1172  *    Update XRa with the 8-bit-wise maximums of signed integers
1173  *    contained in XRb and XRc.
1174  *
1175  *  Q8MIN
1176  *    Update XRa with the 8-bit-wise minimums of signed integers
1177  *    contained in XRb and XRc.
1178  */
1179 static void gen_mxu_Q8MAX_Q8MIN(DisasContext *ctx)
1180 {
1181     uint32_t pad, opc, XRc, XRb, XRa;
1182 
1183     pad = extract32(ctx->opcode, 21, 5);
1184     opc = extract32(ctx->opcode, 18, 3);
1185     XRc = extract32(ctx->opcode, 14, 4);
1186     XRb = extract32(ctx->opcode, 10, 4);
1187     XRa = extract32(ctx->opcode,  6, 4);
1188 
1189     if (unlikely(pad != 0)) {
1190         /* opcode padding incorrect -> do nothing */
1191     } else if (unlikely(XRa == 0)) {
1192         /* destination is zero register -> do nothing */
1193     } else if (unlikely((XRb == 0) && (XRc == 0))) {
1194         /* both operands zero registers -> just set destination to zero */
1195         tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
1196     } else if (unlikely((XRb == 0) || (XRc == 0))) {
1197         /* exactly one operand is zero register - make it be the first...*/
1198         uint32_t XRx = XRb ? XRb : XRc;
1199         /* ...and do byte-wise max/min with one operand 0 */
1200         TCGv_i32 t0 = tcg_temp_new();
1201         TCGv_i32 t1 = tcg_const_i32(0);
1202         int32_t i;
1203 
1204         /* the leftmost byte (byte 3) first */
1205         tcg_gen_andi_i32(t0, mxu_gpr[XRx - 1], 0xFF000000);
1206         if (opc == OPC_MXU_Q8MAX) {
1207             tcg_gen_smax_i32(mxu_gpr[XRa - 1], t0, t1);
1208         } else {
1209             tcg_gen_smin_i32(mxu_gpr[XRa - 1], t0, t1);
1210         }
1211 
1212         /* bytes 2, 1, 0 */
1213         for (i = 2; i >= 0; i--) {
1214             /* extract the byte */
1215             tcg_gen_andi_i32(t0, mxu_gpr[XRx - 1], 0xFF << (8 * i));
1216             /* move the byte to the leftmost position */
1217             tcg_gen_shli_i32(t0, t0, 8 * (3 - i));
1218             /* t0 will be max/min of t0 and t1 */
1219             if (opc == OPC_MXU_Q8MAX) {
1220                 tcg_gen_smax_i32(t0, t0, t1);
1221             } else {
1222                 tcg_gen_smin_i32(t0, t0, t1);
1223             }
1224             /* return resulting byte to its original position */
1225             tcg_gen_shri_i32(t0, t0, 8 * (3 - i));
1226             /* finally update the destination */
1227             tcg_gen_or_i32(mxu_gpr[XRa - 1], mxu_gpr[XRa - 1], t0);
1228         }
1229 
1230         tcg_temp_free(t1);
1231         tcg_temp_free(t0);
1232     } else if (unlikely(XRb == XRc)) {
1233         /* both operands same -> just set destination to one of them */
1234         tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
1235     } else {
1236         /* the most general case */
1237         TCGv_i32 t0 = tcg_temp_new();
1238         TCGv_i32 t1 = tcg_temp_new();
1239         int32_t i;
1240 
1241         /* the leftmost bytes (bytes 3) first */
1242         tcg_gen_andi_i32(t0, mxu_gpr[XRb - 1], 0xFF000000);
1243         tcg_gen_andi_i32(t1, mxu_gpr[XRc - 1], 0xFF000000);
1244         if (opc == OPC_MXU_Q8MAX) {
1245             tcg_gen_smax_i32(mxu_gpr[XRa - 1], t0, t1);
1246         } else {
1247             tcg_gen_smin_i32(mxu_gpr[XRa - 1], t0, t1);
1248         }
1249 
1250         /* bytes 2, 1, 0 */
1251         for (i = 2; i >= 0; i--) {
1252             /* extract corresponding bytes */
1253             tcg_gen_andi_i32(t0, mxu_gpr[XRb - 1], 0xFF << (8 * i));
1254             tcg_gen_andi_i32(t1, mxu_gpr[XRc - 1], 0xFF << (8 * i));
1255             /* move the bytes to the leftmost position */
1256             tcg_gen_shli_i32(t0, t0, 8 * (3 - i));
1257             tcg_gen_shli_i32(t1, t1, 8 * (3 - i));
1258             /* t0 will be max/min of t0 and t1 */
1259             if (opc == OPC_MXU_Q8MAX) {
1260                 tcg_gen_smax_i32(t0, t0, t1);
1261             } else {
1262                 tcg_gen_smin_i32(t0, t0, t1);
1263             }
1264             /* return resulting byte to its original position */
1265             tcg_gen_shri_i32(t0, t0, 8 * (3 - i));
1266             /* finally update the destination */
1267             tcg_gen_or_i32(mxu_gpr[XRa - 1], mxu_gpr[XRa - 1], t0);
1268         }
1269 
1270         tcg_temp_free(t1);
1271         tcg_temp_free(t0);
1272     }
1273 }
1274 
1275 
1276 /*
1277  *                 MXU instruction category: align
1278  *                 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1279  *
1280  *                       S32ALN     S32ALNI
1281  */
1282 
1283 /*
1284  *  S32ALNI XRc, XRb, XRa, optn3
1285  *    Arrange bytes from XRb and XRc according to one of five sets of
1286  *    rules determined by optn3, and place the result in XRa.
1287  */
1288 static void gen_mxu_S32ALNI(DisasContext *ctx)
1289 {
1290     uint32_t optn3, pad, XRc, XRb, XRa;
1291 
1292     optn3 = extract32(ctx->opcode,  23, 3);
1293     pad   = extract32(ctx->opcode,  21, 2);
1294     XRc   = extract32(ctx->opcode, 14, 4);
1295     XRb   = extract32(ctx->opcode, 10, 4);
1296     XRa   = extract32(ctx->opcode,  6, 4);
1297 
1298     if (unlikely(pad != 0)) {
1299         /* opcode padding incorrect -> do nothing */
1300     } else if (unlikely(XRa == 0)) {
1301         /* destination is zero register -> do nothing */
1302     } else if (unlikely((XRb == 0) && (XRc == 0))) {
1303         /* both operands zero registers -> just set destination to all 0s */
1304         tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
1305     } else if (unlikely(XRb == 0)) {
1306         /* XRb zero register -> just appropriatelly shift XRc into XRa */
1307         switch (optn3) {
1308         case MXU_OPTN3_PTN0:
1309             tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
1310             break;
1311         case MXU_OPTN3_PTN1:
1312         case MXU_OPTN3_PTN2:
1313         case MXU_OPTN3_PTN3:
1314             tcg_gen_shri_i32(mxu_gpr[XRa - 1], mxu_gpr[XRc - 1],
1315                              8 * (4 - optn3));
1316             break;
1317         case MXU_OPTN3_PTN4:
1318             tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRc - 1]);
1319             break;
1320         }
1321     } else if (unlikely(XRc == 0)) {
1322         /* XRc zero register -> just appropriatelly shift XRb into XRa */
1323         switch (optn3) {
1324         case MXU_OPTN3_PTN0:
1325             tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
1326             break;
1327         case MXU_OPTN3_PTN1:
1328         case MXU_OPTN3_PTN2:
1329         case MXU_OPTN3_PTN3:
1330             tcg_gen_shri_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1], 8 * optn3);
1331             break;
1332         case MXU_OPTN3_PTN4:
1333             tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
1334             break;
1335         }
1336     } else if (unlikely(XRb == XRc)) {
1337         /* both operands same -> just rotation or moving from any of them */
1338         switch (optn3) {
1339         case MXU_OPTN3_PTN0:
1340         case MXU_OPTN3_PTN4:
1341             tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
1342             break;
1343         case MXU_OPTN3_PTN1:
1344         case MXU_OPTN3_PTN2:
1345         case MXU_OPTN3_PTN3:
1346             tcg_gen_rotli_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1], 8 * optn3);
1347             break;
1348         }
1349     } else {
1350         /* the most general case */
1351         switch (optn3) {
1352         case MXU_OPTN3_PTN0:
1353             {
1354                 /*                                         */
1355                 /*         XRb                XRc          */
1356                 /*  +---------------+                      */
1357                 /*  | A   B   C   D |    E   F   G   H     */
1358                 /*  +-------+-------+                      */
1359                 /*          |                              */
1360                 /*         XRa                             */
1361                 /*                                         */
1362 
1363                 tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
1364             }
1365             break;
1366         case MXU_OPTN3_PTN1:
1367             {
1368                 /*                                         */
1369                 /*         XRb                 XRc         */
1370                 /*      +-------------------+              */
1371                 /*    A | B   C   D       E | F   G   H    */
1372                 /*      +---------+---------+              */
1373                 /*                |                        */
1374                 /*               XRa                       */
1375                 /*                                         */
1376 
1377                 TCGv_i32 t0 = tcg_temp_new();
1378                 TCGv_i32 t1 = tcg_temp_new();
1379 
1380                 tcg_gen_andi_i32(t0, mxu_gpr[XRb - 1], 0x00FFFFFF);
1381                 tcg_gen_shli_i32(t0, t0, 8);
1382 
1383                 tcg_gen_andi_i32(t1, mxu_gpr[XRc - 1], 0xFF000000);
1384                 tcg_gen_shri_i32(t1, t1, 24);
1385 
1386                 tcg_gen_or_i32(mxu_gpr[XRa - 1], t0, t1);
1387 
1388                 tcg_temp_free(t1);
1389                 tcg_temp_free(t0);
1390             }
1391             break;
1392         case MXU_OPTN3_PTN2:
1393             {
1394                 /*                                         */
1395                 /*         XRb                 XRc         */
1396                 /*          +-------------------+          */
1397                 /*    A   B | C   D       E   F | G   H    */
1398                 /*          +---------+---------+          */
1399                 /*                    |                    */
1400                 /*                   XRa                   */
1401                 /*                                         */
1402 
1403                 TCGv_i32 t0 = tcg_temp_new();
1404                 TCGv_i32 t1 = tcg_temp_new();
1405 
1406                 tcg_gen_andi_i32(t0, mxu_gpr[XRb - 1], 0x0000FFFF);
1407                 tcg_gen_shli_i32(t0, t0, 16);
1408 
1409                 tcg_gen_andi_i32(t1, mxu_gpr[XRc - 1], 0xFFFF0000);
1410                 tcg_gen_shri_i32(t1, t1, 16);
1411 
1412                 tcg_gen_or_i32(mxu_gpr[XRa - 1], t0, t1);
1413 
1414                 tcg_temp_free(t1);
1415                 tcg_temp_free(t0);
1416             }
1417             break;
1418         case MXU_OPTN3_PTN3:
1419             {
1420                 /*                                         */
1421                 /*         XRb                 XRc         */
1422                 /*              +-------------------+      */
1423                 /*    A   B   C | D       E   F   G | H    */
1424                 /*              +---------+---------+      */
1425                 /*                        |                */
1426                 /*                       XRa               */
1427                 /*                                         */
1428 
1429                 TCGv_i32 t0 = tcg_temp_new();
1430                 TCGv_i32 t1 = tcg_temp_new();
1431 
1432                 tcg_gen_andi_i32(t0, mxu_gpr[XRb - 1], 0x000000FF);
1433                 tcg_gen_shli_i32(t0, t0, 24);
1434 
1435                 tcg_gen_andi_i32(t1, mxu_gpr[XRc - 1], 0xFFFFFF00);
1436                 tcg_gen_shri_i32(t1, t1, 8);
1437 
1438                 tcg_gen_or_i32(mxu_gpr[XRa - 1], t0, t1);
1439 
1440                 tcg_temp_free(t1);
1441                 tcg_temp_free(t0);
1442             }
1443             break;
1444         case MXU_OPTN3_PTN4:
1445             {
1446                 /*                                         */
1447                 /*         XRb                 XRc         */
1448                 /*                     +---------------+   */
1449                 /*    A   B   C   D    | E   F   G   H |   */
1450                 /*                     +-------+-------+   */
1451                 /*                             |           */
1452                 /*                            XRa          */
1453                 /*                                         */
1454 
1455                 tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRc - 1]);
1456             }
1457             break;
1458         }
1459     }
1460 }
1461 
1462 
1463 /*
1464  * Decoding engine for MXU
1465  * =======================
1466  */
1467 
1468 static void decode_opc_mxu__pool00(DisasContext *ctx)
1469 {
1470     uint32_t opcode = extract32(ctx->opcode, 18, 3);
1471 
1472     switch (opcode) {
1473     case OPC_MXU_S32MAX:
1474     case OPC_MXU_S32MIN:
1475         gen_mxu_S32MAX_S32MIN(ctx);
1476         break;
1477     case OPC_MXU_D16MAX:
1478     case OPC_MXU_D16MIN:
1479         gen_mxu_D16MAX_D16MIN(ctx);
1480         break;
1481     case OPC_MXU_Q8MAX:
1482     case OPC_MXU_Q8MIN:
1483         gen_mxu_Q8MAX_Q8MIN(ctx);
1484         break;
1485     default:
1486         MIPS_INVAL("decode_opc_mxu");
1487         gen_reserved_instruction(ctx);
1488         break;
1489     }
1490 }
1491 
1492 static void decode_opc_mxu__pool04(DisasContext *ctx)
1493 {
1494     uint32_t opcode = extract32(ctx->opcode, 20, 1);
1495 
1496     switch (opcode) {
1497     case OPC_MXU_S32LDD:
1498     case OPC_MXU_S32LDDR:
1499         gen_mxu_s32ldd_s32lddr(ctx);
1500         break;
1501     default:
1502         MIPS_INVAL("decode_opc_mxu");
1503         gen_reserved_instruction(ctx);
1504         break;
1505     }
1506 }
1507 
1508 static void decode_opc_mxu__pool16(DisasContext *ctx)
1509 {
1510     uint32_t opcode = extract32(ctx->opcode, 18, 3);
1511 
1512     switch (opcode) {
1513     case OPC_MXU_S32ALNI:
1514         gen_mxu_S32ALNI(ctx);
1515         break;
1516     case OPC_MXU_S32NOR:
1517         gen_mxu_S32NOR(ctx);
1518         break;
1519     case OPC_MXU_S32AND:
1520         gen_mxu_S32AND(ctx);
1521         break;
1522     case OPC_MXU_S32OR:
1523         gen_mxu_S32OR(ctx);
1524         break;
1525     case OPC_MXU_S32XOR:
1526         gen_mxu_S32XOR(ctx);
1527         break;
1528     default:
1529         MIPS_INVAL("decode_opc_mxu");
1530         gen_reserved_instruction(ctx);
1531         break;
1532     }
1533 }
1534 
1535 static void decode_opc_mxu__pool19(DisasContext *ctx)
1536 {
1537     uint32_t opcode = extract32(ctx->opcode, 22, 2);
1538 
1539     switch (opcode) {
1540     case OPC_MXU_Q8MUL:
1541     case OPC_MXU_Q8MULSU:
1542         gen_mxu_q8mul_q8mulsu(ctx);
1543         break;
1544     default:
1545         MIPS_INVAL("decode_opc_mxu");
1546         gen_reserved_instruction(ctx);
1547         break;
1548     }
1549 }
1550 
1551 bool decode_ase_mxu(DisasContext *ctx, uint32_t insn)
1552 {
1553     uint32_t opcode = extract32(insn, 0, 6);
1554 
1555     if (opcode == OPC_MXU_S32M2I) {
1556         gen_mxu_s32m2i(ctx);
1557         return true;
1558     }
1559 
1560     if (opcode == OPC_MXU_S32I2M) {
1561         gen_mxu_s32i2m(ctx);
1562         return true;
1563     }
1564 
1565     {
1566         TCGv t_mxu_cr = tcg_temp_new();
1567         TCGLabel *l_exit = gen_new_label();
1568 
1569         gen_load_mxu_cr(t_mxu_cr);
1570         tcg_gen_andi_tl(t_mxu_cr, t_mxu_cr, MXU_CR_MXU_EN);
1571         tcg_gen_brcondi_tl(TCG_COND_NE, t_mxu_cr, MXU_CR_MXU_EN, l_exit);
1572 
1573         switch (opcode) {
1574         case OPC_MXU__POOL00:
1575             decode_opc_mxu__pool00(ctx);
1576             break;
1577         case OPC_MXU_D16MUL:
1578             gen_mxu_d16mul(ctx);
1579             break;
1580         case OPC_MXU_D16MAC:
1581             gen_mxu_d16mac(ctx);
1582             break;
1583         case OPC_MXU__POOL04:
1584             decode_opc_mxu__pool04(ctx);
1585             break;
1586         case OPC_MXU_S8LDD:
1587             gen_mxu_s8ldd(ctx);
1588             break;
1589         case OPC_MXU__POOL16:
1590             decode_opc_mxu__pool16(ctx);
1591             break;
1592         case OPC_MXU__POOL19:
1593             decode_opc_mxu__pool19(ctx);
1594             break;
1595         default:
1596             MIPS_INVAL("decode_opc_mxu");
1597             gen_reserved_instruction(ctx);
1598         }
1599 
1600         gen_set_label(l_exit);
1601         tcg_temp_free(t_mxu_cr);
1602     }
1603 
1604     return true;
1605 }
1606