xref: /qemu/target/mips/tcg/mxu_translate.c (revision 83ecdb18)
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 
518 /*
519  * S32M2I XRa, rb - Register move from XRF to GRF
520  */
521 static void gen_mxu_s32m2i(DisasContext *ctx)
522 {
523     TCGv t0;
524     uint32_t XRa, Rb;
525 
526     t0 = tcg_temp_new();
527 
528     XRa = extract32(ctx->opcode, 6, 5);
529     Rb = extract32(ctx->opcode, 16, 5);
530 
531     if (XRa <= 15) {
532         gen_load_mxu_gpr(t0, XRa);
533     } else if (XRa == 16) {
534         gen_load_mxu_cr(t0);
535     }
536 
537     gen_store_gpr(t0, Rb);
538 }
539 
540 /*
541  * S8LDD XRa, Rb, s8, optn3 - Load a byte from memory to XRF
542  */
543 static void gen_mxu_s8ldd(DisasContext *ctx)
544 {
545     TCGv t0, t1;
546     uint32_t XRa, Rb, s8, optn3;
547 
548     t0 = tcg_temp_new();
549     t1 = tcg_temp_new();
550 
551     XRa = extract32(ctx->opcode, 6, 4);
552     s8 = extract32(ctx->opcode, 10, 8);
553     optn3 = extract32(ctx->opcode, 18, 3);
554     Rb = extract32(ctx->opcode, 21, 5);
555 
556     gen_load_gpr(t0, Rb);
557     tcg_gen_addi_tl(t0, t0, (int8_t)s8);
558 
559     switch (optn3) {
560     /* XRa[7:0] = tmp8 */
561     case MXU_OPTN3_PTN0:
562         tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_UB);
563         gen_load_mxu_gpr(t0, XRa);
564         tcg_gen_deposit_tl(t0, t0, t1, 0, 8);
565         break;
566     /* XRa[15:8] = tmp8 */
567     case MXU_OPTN3_PTN1:
568         tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_UB);
569         gen_load_mxu_gpr(t0, XRa);
570         tcg_gen_deposit_tl(t0, t0, t1, 8, 8);
571         break;
572     /* XRa[23:16] = tmp8 */
573     case MXU_OPTN3_PTN2:
574         tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_UB);
575         gen_load_mxu_gpr(t0, XRa);
576         tcg_gen_deposit_tl(t0, t0, t1, 16, 8);
577         break;
578     /* XRa[31:24] = tmp8 */
579     case MXU_OPTN3_PTN3:
580         tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_UB);
581         gen_load_mxu_gpr(t0, XRa);
582         tcg_gen_deposit_tl(t0, t0, t1, 24, 8);
583         break;
584     /* XRa = {8'b0, tmp8, 8'b0, tmp8} */
585     case MXU_OPTN3_PTN4:
586         tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_UB);
587         tcg_gen_deposit_tl(t0, t1, t1, 16, 16);
588         break;
589     /* XRa = {tmp8, 8'b0, tmp8, 8'b0} */
590     case MXU_OPTN3_PTN5:
591         tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_UB);
592         tcg_gen_shli_tl(t1, t1, 8);
593         tcg_gen_deposit_tl(t0, t1, t1, 16, 16);
594         break;
595     /* XRa = {{8{sign of tmp8}}, tmp8, {8{sign of tmp8}}, tmp8} */
596     case MXU_OPTN3_PTN6:
597         tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_SB);
598         tcg_gen_mov_tl(t0, t1);
599         tcg_gen_andi_tl(t0, t0, 0xFF00FFFF);
600         tcg_gen_shli_tl(t1, t1, 16);
601         tcg_gen_or_tl(t0, t0, t1);
602         break;
603     /* XRa = {tmp8, tmp8, tmp8, tmp8} */
604     case MXU_OPTN3_PTN7:
605         tcg_gen_qemu_ld_tl(t1, t0, ctx->mem_idx, MO_UB);
606         tcg_gen_deposit_tl(t1, t1, t1, 8, 8);
607         tcg_gen_deposit_tl(t0, t1, t1, 16, 16);
608         break;
609     }
610 
611     gen_store_mxu_gpr(t0, XRa);
612 }
613 
614 /*
615  * D16MUL XRa, XRb, XRc, XRd, optn2 - Signed 16 bit pattern multiplication
616  */
617 static void gen_mxu_d16mul(DisasContext *ctx)
618 {
619     TCGv t0, t1, t2, t3;
620     uint32_t XRa, XRb, XRc, XRd, optn2;
621 
622     t0 = tcg_temp_new();
623     t1 = tcg_temp_new();
624     t2 = tcg_temp_new();
625     t3 = tcg_temp_new();
626 
627     XRa = extract32(ctx->opcode, 6, 4);
628     XRb = extract32(ctx->opcode, 10, 4);
629     XRc = extract32(ctx->opcode, 14, 4);
630     XRd = extract32(ctx->opcode, 18, 4);
631     optn2 = extract32(ctx->opcode, 22, 2);
632 
633     gen_load_mxu_gpr(t1, XRb);
634     tcg_gen_sextract_tl(t0, t1, 0, 16);
635     tcg_gen_sextract_tl(t1, t1, 16, 16);
636     gen_load_mxu_gpr(t3, XRc);
637     tcg_gen_sextract_tl(t2, t3, 0, 16);
638     tcg_gen_sextract_tl(t3, t3, 16, 16);
639 
640     switch (optn2) {
641     case MXU_OPTN2_WW: /* XRB.H*XRC.H == lop, XRB.L*XRC.L == rop */
642         tcg_gen_mul_tl(t3, t1, t3);
643         tcg_gen_mul_tl(t2, t0, t2);
644         break;
645     case MXU_OPTN2_LW: /* XRB.L*XRC.H == lop, XRB.L*XRC.L == rop */
646         tcg_gen_mul_tl(t3, t0, t3);
647         tcg_gen_mul_tl(t2, t0, t2);
648         break;
649     case MXU_OPTN2_HW: /* XRB.H*XRC.H == lop, XRB.H*XRC.L == rop */
650         tcg_gen_mul_tl(t3, t1, t3);
651         tcg_gen_mul_tl(t2, t1, t2);
652         break;
653     case MXU_OPTN2_XW: /* XRB.L*XRC.H == lop, XRB.H*XRC.L == rop */
654         tcg_gen_mul_tl(t3, t0, t3);
655         tcg_gen_mul_tl(t2, t1, t2);
656         break;
657     }
658     gen_store_mxu_gpr(t3, XRa);
659     gen_store_mxu_gpr(t2, XRd);
660 }
661 
662 /*
663  * D16MAC XRa, XRb, XRc, XRd, aptn2, optn2 - Signed 16 bit pattern multiply
664  *                                           and accumulate
665  */
666 static void gen_mxu_d16mac(DisasContext *ctx)
667 {
668     TCGv t0, t1, t2, t3;
669     uint32_t XRa, XRb, XRc, XRd, optn2, aptn2;
670 
671     t0 = tcg_temp_new();
672     t1 = tcg_temp_new();
673     t2 = tcg_temp_new();
674     t3 = tcg_temp_new();
675 
676     XRa = extract32(ctx->opcode, 6, 4);
677     XRb = extract32(ctx->opcode, 10, 4);
678     XRc = extract32(ctx->opcode, 14, 4);
679     XRd = extract32(ctx->opcode, 18, 4);
680     optn2 = extract32(ctx->opcode, 22, 2);
681     aptn2 = extract32(ctx->opcode, 24, 2);
682 
683     gen_load_mxu_gpr(t1, XRb);
684     tcg_gen_sextract_tl(t0, t1, 0, 16);
685     tcg_gen_sextract_tl(t1, t1, 16, 16);
686 
687     gen_load_mxu_gpr(t3, XRc);
688     tcg_gen_sextract_tl(t2, t3, 0, 16);
689     tcg_gen_sextract_tl(t3, t3, 16, 16);
690 
691     switch (optn2) {
692     case MXU_OPTN2_WW: /* XRB.H*XRC.H == lop, XRB.L*XRC.L == rop */
693         tcg_gen_mul_tl(t3, t1, t3);
694         tcg_gen_mul_tl(t2, t0, t2);
695         break;
696     case MXU_OPTN2_LW: /* XRB.L*XRC.H == lop, XRB.L*XRC.L == rop */
697         tcg_gen_mul_tl(t3, t0, t3);
698         tcg_gen_mul_tl(t2, t0, t2);
699         break;
700     case MXU_OPTN2_HW: /* XRB.H*XRC.H == lop, XRB.H*XRC.L == rop */
701         tcg_gen_mul_tl(t3, t1, t3);
702         tcg_gen_mul_tl(t2, t1, t2);
703         break;
704     case MXU_OPTN2_XW: /* XRB.L*XRC.H == lop, XRB.H*XRC.L == rop */
705         tcg_gen_mul_tl(t3, t0, t3);
706         tcg_gen_mul_tl(t2, t1, t2);
707         break;
708     }
709     gen_load_mxu_gpr(t0, XRa);
710     gen_load_mxu_gpr(t1, XRd);
711 
712     switch (aptn2) {
713     case MXU_APTN2_AA:
714         tcg_gen_add_tl(t3, t0, t3);
715         tcg_gen_add_tl(t2, t1, t2);
716         break;
717     case MXU_APTN2_AS:
718         tcg_gen_add_tl(t3, t0, t3);
719         tcg_gen_sub_tl(t2, t1, t2);
720         break;
721     case MXU_APTN2_SA:
722         tcg_gen_sub_tl(t3, t0, t3);
723         tcg_gen_add_tl(t2, t1, t2);
724         break;
725     case MXU_APTN2_SS:
726         tcg_gen_sub_tl(t3, t0, t3);
727         tcg_gen_sub_tl(t2, t1, t2);
728         break;
729     }
730     gen_store_mxu_gpr(t3, XRa);
731     gen_store_mxu_gpr(t2, XRd);
732 }
733 
734 /*
735  * Q8MUL   XRa, XRb, XRc, XRd - Parallel unsigned 8 bit pattern multiply
736  * Q8MULSU XRa, XRb, XRc, XRd - Parallel signed 8 bit pattern multiply
737  */
738 static void gen_mxu_q8mul_q8mulsu(DisasContext *ctx)
739 {
740     TCGv t0, t1, t2, t3, t4, t5, t6, t7;
741     uint32_t XRa, XRb, XRc, XRd, sel;
742 
743     t0 = tcg_temp_new();
744     t1 = tcg_temp_new();
745     t2 = tcg_temp_new();
746     t3 = tcg_temp_new();
747     t4 = tcg_temp_new();
748     t5 = tcg_temp_new();
749     t6 = tcg_temp_new();
750     t7 = tcg_temp_new();
751 
752     XRa = extract32(ctx->opcode, 6, 4);
753     XRb = extract32(ctx->opcode, 10, 4);
754     XRc = extract32(ctx->opcode, 14, 4);
755     XRd = extract32(ctx->opcode, 18, 4);
756     sel = extract32(ctx->opcode, 22, 2);
757 
758     gen_load_mxu_gpr(t3, XRb);
759     gen_load_mxu_gpr(t7, XRc);
760 
761     if (sel == 0x2) {
762         /* Q8MULSU */
763         tcg_gen_ext8s_tl(t0, t3);
764         tcg_gen_shri_tl(t3, t3, 8);
765         tcg_gen_ext8s_tl(t1, t3);
766         tcg_gen_shri_tl(t3, t3, 8);
767         tcg_gen_ext8s_tl(t2, t3);
768         tcg_gen_shri_tl(t3, t3, 8);
769         tcg_gen_ext8s_tl(t3, t3);
770     } else {
771         /* Q8MUL */
772         tcg_gen_ext8u_tl(t0, t3);
773         tcg_gen_shri_tl(t3, t3, 8);
774         tcg_gen_ext8u_tl(t1, t3);
775         tcg_gen_shri_tl(t3, t3, 8);
776         tcg_gen_ext8u_tl(t2, t3);
777         tcg_gen_shri_tl(t3, t3, 8);
778         tcg_gen_ext8u_tl(t3, t3);
779     }
780 
781     tcg_gen_ext8u_tl(t4, t7);
782     tcg_gen_shri_tl(t7, t7, 8);
783     tcg_gen_ext8u_tl(t5, t7);
784     tcg_gen_shri_tl(t7, t7, 8);
785     tcg_gen_ext8u_tl(t6, t7);
786     tcg_gen_shri_tl(t7, t7, 8);
787     tcg_gen_ext8u_tl(t7, t7);
788 
789     tcg_gen_mul_tl(t0, t0, t4);
790     tcg_gen_mul_tl(t1, t1, t5);
791     tcg_gen_mul_tl(t2, t2, t6);
792     tcg_gen_mul_tl(t3, t3, t7);
793 
794     tcg_gen_andi_tl(t0, t0, 0xFFFF);
795     tcg_gen_andi_tl(t1, t1, 0xFFFF);
796     tcg_gen_andi_tl(t2, t2, 0xFFFF);
797     tcg_gen_andi_tl(t3, t3, 0xFFFF);
798 
799     tcg_gen_shli_tl(t1, t1, 16);
800     tcg_gen_shli_tl(t3, t3, 16);
801 
802     tcg_gen_or_tl(t0, t0, t1);
803     tcg_gen_or_tl(t1, t2, t3);
804 
805     gen_store_mxu_gpr(t0, XRd);
806     gen_store_mxu_gpr(t1, XRa);
807 }
808 
809 /*
810  * S32LDD  XRa, Rb, S12 - Load a word from memory to XRF
811  * S32LDDR XRa, Rb, S12 - Load a word from memory to XRF, reversed byte seq.
812  */
813 static void gen_mxu_s32ldd_s32lddr(DisasContext *ctx)
814 {
815     TCGv t0, t1;
816     uint32_t XRa, Rb, s12, sel;
817 
818     t0 = tcg_temp_new();
819     t1 = tcg_temp_new();
820 
821     XRa = extract32(ctx->opcode, 6, 4);
822     s12 = extract32(ctx->opcode, 10, 10);
823     sel = extract32(ctx->opcode, 20, 1);
824     Rb = extract32(ctx->opcode, 21, 5);
825 
826     gen_load_gpr(t0, Rb);
827 
828     tcg_gen_movi_tl(t1, s12);
829     tcg_gen_shli_tl(t1, t1, 2);
830     if (s12 & 0x200) {
831         tcg_gen_ori_tl(t1, t1, 0xFFFFF000);
832     }
833     tcg_gen_add_tl(t1, t0, t1);
834     tcg_gen_qemu_ld_tl(t1, t1, ctx->mem_idx, MO_TESL ^ (sel * MO_BSWAP));
835 
836     gen_store_mxu_gpr(t1, XRa);
837 }
838 
839 
840 /*
841  *                 MXU instruction category: logic
842  *                 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
843  *
844  *               S32NOR    S32AND    S32OR    S32XOR
845  */
846 
847 /*
848  *  S32NOR XRa, XRb, XRc
849  *    Update XRa with the result of logical bitwise 'nor' operation
850  *    applied to the content of XRb and XRc.
851  */
852 static void gen_mxu_S32NOR(DisasContext *ctx)
853 {
854     uint32_t pad, XRc, XRb, XRa;
855 
856     pad = extract32(ctx->opcode, 21, 5);
857     XRc = extract32(ctx->opcode, 14, 4);
858     XRb = extract32(ctx->opcode, 10, 4);
859     XRa = extract32(ctx->opcode,  6, 4);
860 
861     if (unlikely(pad != 0)) {
862         /* opcode padding incorrect -> do nothing */
863     } else if (unlikely(XRa == 0)) {
864         /* destination is zero register -> do nothing */
865     } else if (unlikely((XRb == 0) && (XRc == 0))) {
866         /* both operands zero registers -> just set destination to all 1s */
867         tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0xFFFFFFFF);
868     } else if (unlikely(XRb == 0)) {
869         /* XRb zero register -> just set destination to the negation of XRc */
870         tcg_gen_not_i32(mxu_gpr[XRa - 1], mxu_gpr[XRc - 1]);
871     } else if (unlikely(XRc == 0)) {
872         /* XRa zero register -> just set destination to the negation of XRb */
873         tcg_gen_not_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
874     } else if (unlikely(XRb == XRc)) {
875         /* both operands same -> just set destination to the negation of XRb */
876         tcg_gen_not_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
877     } else {
878         /* the most general case */
879         tcg_gen_nor_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1], mxu_gpr[XRc - 1]);
880     }
881 }
882 
883 /*
884  *  S32AND XRa, XRb, XRc
885  *    Update XRa with the result of logical bitwise 'and' operation
886  *    applied to the content of XRb and XRc.
887  */
888 static void gen_mxu_S32AND(DisasContext *ctx)
889 {
890     uint32_t pad, XRc, XRb, XRa;
891 
892     pad = extract32(ctx->opcode, 21, 5);
893     XRc = extract32(ctx->opcode, 14, 4);
894     XRb = extract32(ctx->opcode, 10, 4);
895     XRa = extract32(ctx->opcode,  6, 4);
896 
897     if (unlikely(pad != 0)) {
898         /* opcode padding incorrect -> do nothing */
899     } else if (unlikely(XRa == 0)) {
900         /* destination is zero register -> do nothing */
901     } else if (unlikely((XRb == 0) || (XRc == 0))) {
902         /* one of operands zero register -> just set destination to all 0s */
903         tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
904     } else if (unlikely(XRb == XRc)) {
905         /* both operands same -> just set destination to one of them */
906         tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
907     } else {
908         /* the most general case */
909         tcg_gen_and_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1], mxu_gpr[XRc - 1]);
910     }
911 }
912 
913 /*
914  *  S32OR XRa, XRb, XRc
915  *    Update XRa with the result of logical bitwise 'or' operation
916  *    applied to the content of XRb and XRc.
917  */
918 static void gen_mxu_S32OR(DisasContext *ctx)
919 {
920     uint32_t pad, XRc, XRb, XRa;
921 
922     pad = extract32(ctx->opcode, 21, 5);
923     XRc = extract32(ctx->opcode, 14, 4);
924     XRb = extract32(ctx->opcode, 10, 4);
925     XRa = extract32(ctx->opcode,  6, 4);
926 
927     if (unlikely(pad != 0)) {
928         /* opcode padding incorrect -> do nothing */
929     } else if (unlikely(XRa == 0)) {
930         /* destination is zero register -> do nothing */
931     } else if (unlikely((XRb == 0) && (XRc == 0))) {
932         /* both operands zero registers -> just set destination to all 0s */
933         tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
934     } else if (unlikely(XRb == 0)) {
935         /* XRb zero register -> just set destination to the content of XRc */
936         tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRc - 1]);
937     } else if (unlikely(XRc == 0)) {
938         /* XRc zero register -> just set destination to the content of XRb */
939         tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
940     } else if (unlikely(XRb == XRc)) {
941         /* both operands same -> just set destination to one of them */
942         tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
943     } else {
944         /* the most general case */
945         tcg_gen_or_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1], mxu_gpr[XRc - 1]);
946     }
947 }
948 
949 /*
950  *  S32XOR XRa, XRb, XRc
951  *    Update XRa with the result of logical bitwise 'xor' operation
952  *    applied to the content of XRb and XRc.
953  */
954 static void gen_mxu_S32XOR(DisasContext *ctx)
955 {
956     uint32_t pad, XRc, XRb, XRa;
957 
958     pad = extract32(ctx->opcode, 21, 5);
959     XRc = extract32(ctx->opcode, 14, 4);
960     XRb = extract32(ctx->opcode, 10, 4);
961     XRa = extract32(ctx->opcode,  6, 4);
962 
963     if (unlikely(pad != 0)) {
964         /* opcode padding incorrect -> do nothing */
965     } else if (unlikely(XRa == 0)) {
966         /* destination is zero register -> do nothing */
967     } else if (unlikely((XRb == 0) && (XRc == 0))) {
968         /* both operands zero registers -> just set destination to all 0s */
969         tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
970     } else if (unlikely(XRb == 0)) {
971         /* XRb zero register -> just set destination to the content of XRc */
972         tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRc - 1]);
973     } else if (unlikely(XRc == 0)) {
974         /* XRc zero register -> just set destination to the content of XRb */
975         tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
976     } else if (unlikely(XRb == XRc)) {
977         /* both operands same -> just set destination to all 0s */
978         tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
979     } else {
980         /* the most general case */
981         tcg_gen_xor_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1], mxu_gpr[XRc - 1]);
982     }
983 }
984 
985 
986 /*
987  *                   MXU instruction category max/min
988  *                   ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
989  *
990  *                     S32MAX     D16MAX     Q8MAX
991  *                     S32MIN     D16MIN     Q8MIN
992  */
993 
994 /*
995  *  S32MAX XRa, XRb, XRc
996  *    Update XRa with the maximum of signed 32-bit integers contained
997  *    in XRb and XRc.
998  *
999  *  S32MIN XRa, XRb, XRc
1000  *    Update XRa with the minimum of signed 32-bit integers contained
1001  *    in XRb and XRc.
1002  */
1003 static void gen_mxu_S32MAX_S32MIN(DisasContext *ctx)
1004 {
1005     uint32_t pad, opc, XRc, XRb, XRa;
1006 
1007     pad = extract32(ctx->opcode, 21, 5);
1008     opc = extract32(ctx->opcode, 18, 3);
1009     XRc = extract32(ctx->opcode, 14, 4);
1010     XRb = extract32(ctx->opcode, 10, 4);
1011     XRa = extract32(ctx->opcode,  6, 4);
1012 
1013     if (unlikely(pad != 0)) {
1014         /* opcode padding incorrect -> do nothing */
1015     } else if (unlikely(XRa == 0)) {
1016         /* destination is zero register -> do nothing */
1017     } else if (unlikely((XRb == 0) && (XRc == 0))) {
1018         /* both operands zero registers -> just set destination to zero */
1019         tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
1020     } else if (unlikely((XRb == 0) || (XRc == 0))) {
1021         /* exactly one operand is zero register - find which one is not...*/
1022         uint32_t XRx = XRb ? XRb : XRc;
1023         /* ...and do max/min operation with one operand 0 */
1024         if (opc == OPC_MXU_S32MAX) {
1025             tcg_gen_smax_i32(mxu_gpr[XRa - 1], mxu_gpr[XRx - 1], 0);
1026         } else {
1027             tcg_gen_smin_i32(mxu_gpr[XRa - 1], mxu_gpr[XRx - 1], 0);
1028         }
1029     } else if (unlikely(XRb == XRc)) {
1030         /* both operands same -> just set destination to one of them */
1031         tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
1032     } else {
1033         /* the most general case */
1034         if (opc == OPC_MXU_S32MAX) {
1035             tcg_gen_smax_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1],
1036                                                mxu_gpr[XRc - 1]);
1037         } else {
1038             tcg_gen_smin_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1],
1039                                                mxu_gpr[XRc - 1]);
1040         }
1041     }
1042 }
1043 
1044 /*
1045  *  D16MAX
1046  *    Update XRa with the 16-bit-wise maximums of signed integers
1047  *    contained in XRb and XRc.
1048  *
1049  *  D16MIN
1050  *    Update XRa with the 16-bit-wise minimums of signed integers
1051  *    contained in XRb and XRc.
1052  */
1053 static void gen_mxu_D16MAX_D16MIN(DisasContext *ctx)
1054 {
1055     uint32_t pad, opc, XRc, XRb, XRa;
1056 
1057     pad = extract32(ctx->opcode, 21, 5);
1058     opc = extract32(ctx->opcode, 18, 3);
1059     XRc = extract32(ctx->opcode, 14, 4);
1060     XRb = extract32(ctx->opcode, 10, 4);
1061     XRa = extract32(ctx->opcode,  6, 4);
1062 
1063     if (unlikely(pad != 0)) {
1064         /* opcode padding incorrect -> do nothing */
1065     } else if (unlikely(XRa == 0)) {
1066         /* destination is zero register -> do nothing */
1067     } else if (unlikely((XRb == 0) && (XRc == 0))) {
1068         /* both operands zero registers -> just set destination to zero */
1069         tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
1070     } else if (unlikely((XRb == 0) || (XRc == 0))) {
1071         /* exactly one operand is zero register - find which one is not...*/
1072         uint32_t XRx = XRb ? XRb : XRc;
1073         /* ...and do half-word-wise max/min with one operand 0 */
1074         TCGv_i32 t0 = tcg_temp_new();
1075         TCGv_i32 t1 = tcg_constant_i32(0);
1076 
1077         /* the left half-word first */
1078         tcg_gen_andi_i32(t0, mxu_gpr[XRx - 1], 0xFFFF0000);
1079         if (opc == OPC_MXU_D16MAX) {
1080             tcg_gen_smax_i32(mxu_gpr[XRa - 1], t0, t1);
1081         } else {
1082             tcg_gen_smin_i32(mxu_gpr[XRa - 1], t0, t1);
1083         }
1084 
1085         /* the right half-word */
1086         tcg_gen_andi_i32(t0, mxu_gpr[XRx - 1], 0x0000FFFF);
1087         /* move half-words to the leftmost position */
1088         tcg_gen_shli_i32(t0, t0, 16);
1089         /* t0 will be max/min of t0 and t1 */
1090         if (opc == OPC_MXU_D16MAX) {
1091             tcg_gen_smax_i32(t0, t0, t1);
1092         } else {
1093             tcg_gen_smin_i32(t0, t0, t1);
1094         }
1095         /* return resulting half-words to its original position */
1096         tcg_gen_shri_i32(t0, t0, 16);
1097         /* finally update the destination */
1098         tcg_gen_or_i32(mxu_gpr[XRa - 1], mxu_gpr[XRa - 1], t0);
1099     } else if (unlikely(XRb == XRc)) {
1100         /* both operands same -> just set destination to one of them */
1101         tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
1102     } else {
1103         /* the most general case */
1104         TCGv_i32 t0 = tcg_temp_new();
1105         TCGv_i32 t1 = tcg_temp_new();
1106 
1107         /* the left half-word first */
1108         tcg_gen_andi_i32(t0, mxu_gpr[XRb - 1], 0xFFFF0000);
1109         tcg_gen_andi_i32(t1, mxu_gpr[XRc - 1], 0xFFFF0000);
1110         if (opc == OPC_MXU_D16MAX) {
1111             tcg_gen_smax_i32(mxu_gpr[XRa - 1], t0, t1);
1112         } else {
1113             tcg_gen_smin_i32(mxu_gpr[XRa - 1], t0, t1);
1114         }
1115 
1116         /* the right half-word */
1117         tcg_gen_andi_i32(t0, mxu_gpr[XRb - 1], 0x0000FFFF);
1118         tcg_gen_andi_i32(t1, mxu_gpr[XRc - 1], 0x0000FFFF);
1119         /* move half-words to the leftmost position */
1120         tcg_gen_shli_i32(t0, t0, 16);
1121         tcg_gen_shli_i32(t1, t1, 16);
1122         /* t0 will be max/min of t0 and t1 */
1123         if (opc == OPC_MXU_D16MAX) {
1124             tcg_gen_smax_i32(t0, t0, t1);
1125         } else {
1126             tcg_gen_smin_i32(t0, t0, t1);
1127         }
1128         /* return resulting half-words to its original position */
1129         tcg_gen_shri_i32(t0, t0, 16);
1130         /* finally update the destination */
1131         tcg_gen_or_i32(mxu_gpr[XRa - 1], mxu_gpr[XRa - 1], t0);
1132     }
1133 }
1134 
1135 /*
1136  *  Q8MAX
1137  *    Update XRa with the 8-bit-wise maximums of signed integers
1138  *    contained in XRb and XRc.
1139  *
1140  *  Q8MIN
1141  *    Update XRa with the 8-bit-wise minimums of signed integers
1142  *    contained in XRb and XRc.
1143  */
1144 static void gen_mxu_Q8MAX_Q8MIN(DisasContext *ctx)
1145 {
1146     uint32_t pad, opc, XRc, XRb, XRa;
1147 
1148     pad = extract32(ctx->opcode, 21, 5);
1149     opc = extract32(ctx->opcode, 18, 3);
1150     XRc = extract32(ctx->opcode, 14, 4);
1151     XRb = extract32(ctx->opcode, 10, 4);
1152     XRa = extract32(ctx->opcode,  6, 4);
1153 
1154     if (unlikely(pad != 0)) {
1155         /* opcode padding incorrect -> do nothing */
1156     } else if (unlikely(XRa == 0)) {
1157         /* destination is zero register -> do nothing */
1158     } else if (unlikely((XRb == 0) && (XRc == 0))) {
1159         /* both operands zero registers -> just set destination to zero */
1160         tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
1161     } else if (unlikely((XRb == 0) || (XRc == 0))) {
1162         /* exactly one operand is zero register - make it be the first...*/
1163         uint32_t XRx = XRb ? XRb : XRc;
1164         /* ...and do byte-wise max/min with one operand 0 */
1165         TCGv_i32 t0 = tcg_temp_new();
1166         TCGv_i32 t1 = tcg_constant_i32(0);
1167         int32_t i;
1168 
1169         /* the leftmost byte (byte 3) first */
1170         tcg_gen_andi_i32(t0, mxu_gpr[XRx - 1], 0xFF000000);
1171         if (opc == OPC_MXU_Q8MAX) {
1172             tcg_gen_smax_i32(mxu_gpr[XRa - 1], t0, t1);
1173         } else {
1174             tcg_gen_smin_i32(mxu_gpr[XRa - 1], t0, t1);
1175         }
1176 
1177         /* bytes 2, 1, 0 */
1178         for (i = 2; i >= 0; i--) {
1179             /* extract the byte */
1180             tcg_gen_andi_i32(t0, mxu_gpr[XRx - 1], 0xFF << (8 * i));
1181             /* move the byte to the leftmost position */
1182             tcg_gen_shli_i32(t0, t0, 8 * (3 - i));
1183             /* t0 will be max/min of t0 and t1 */
1184             if (opc == OPC_MXU_Q8MAX) {
1185                 tcg_gen_smax_i32(t0, t0, t1);
1186             } else {
1187                 tcg_gen_smin_i32(t0, t0, t1);
1188             }
1189             /* return resulting byte to its original position */
1190             tcg_gen_shri_i32(t0, t0, 8 * (3 - i));
1191             /* finally update the destination */
1192             tcg_gen_or_i32(mxu_gpr[XRa - 1], mxu_gpr[XRa - 1], t0);
1193         }
1194     } else if (unlikely(XRb == XRc)) {
1195         /* both operands same -> just set destination to one of them */
1196         tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
1197     } else {
1198         /* the most general case */
1199         TCGv_i32 t0 = tcg_temp_new();
1200         TCGv_i32 t1 = tcg_temp_new();
1201         int32_t i;
1202 
1203         /* the leftmost bytes (bytes 3) first */
1204         tcg_gen_andi_i32(t0, mxu_gpr[XRb - 1], 0xFF000000);
1205         tcg_gen_andi_i32(t1, mxu_gpr[XRc - 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 corresponding bytes */
1215             tcg_gen_andi_i32(t0, mxu_gpr[XRb - 1], 0xFF << (8 * i));
1216             tcg_gen_andi_i32(t1, mxu_gpr[XRc - 1], 0xFF << (8 * i));
1217             /* move the bytes to the leftmost position */
1218             tcg_gen_shli_i32(t0, t0, 8 * (3 - i));
1219             tcg_gen_shli_i32(t1, t1, 8 * (3 - i));
1220             /* t0 will be max/min of t0 and t1 */
1221             if (opc == OPC_MXU_Q8MAX) {
1222                 tcg_gen_smax_i32(t0, t0, t1);
1223             } else {
1224                 tcg_gen_smin_i32(t0, t0, t1);
1225             }
1226             /* return resulting byte to its original position */
1227             tcg_gen_shri_i32(t0, t0, 8 * (3 - i));
1228             /* finally update the destination */
1229             tcg_gen_or_i32(mxu_gpr[XRa - 1], mxu_gpr[XRa - 1], t0);
1230         }
1231     }
1232 }
1233 
1234 
1235 /*
1236  *                 MXU instruction category: align
1237  *                 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1238  *
1239  *                       S32ALN     S32ALNI
1240  */
1241 
1242 /*
1243  *  S32ALNI XRc, XRb, XRa, optn3
1244  *    Arrange bytes from XRb and XRc according to one of five sets of
1245  *    rules determined by optn3, and place the result in XRa.
1246  */
1247 static void gen_mxu_S32ALNI(DisasContext *ctx)
1248 {
1249     uint32_t optn3, pad, XRc, XRb, XRa;
1250 
1251     optn3 = extract32(ctx->opcode,  23, 3);
1252     pad   = extract32(ctx->opcode,  21, 2);
1253     XRc   = extract32(ctx->opcode, 14, 4);
1254     XRb   = extract32(ctx->opcode, 10, 4);
1255     XRa   = extract32(ctx->opcode,  6, 4);
1256 
1257     if (unlikely(pad != 0)) {
1258         /* opcode padding incorrect -> do nothing */
1259     } else if (unlikely(XRa == 0)) {
1260         /* destination is zero register -> do nothing */
1261     } else if (unlikely((XRb == 0) && (XRc == 0))) {
1262         /* both operands zero registers -> just set destination to all 0s */
1263         tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
1264     } else if (unlikely(XRb == 0)) {
1265         /* XRb zero register -> just appropriatelly shift XRc into XRa */
1266         switch (optn3) {
1267         case MXU_OPTN3_PTN0:
1268             tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
1269             break;
1270         case MXU_OPTN3_PTN1:
1271         case MXU_OPTN3_PTN2:
1272         case MXU_OPTN3_PTN3:
1273             tcg_gen_shri_i32(mxu_gpr[XRa - 1], mxu_gpr[XRc - 1],
1274                              8 * (4 - optn3));
1275             break;
1276         case MXU_OPTN3_PTN4:
1277             tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRc - 1]);
1278             break;
1279         }
1280     } else if (unlikely(XRc == 0)) {
1281         /* XRc zero register -> just appropriatelly shift XRb into XRa */
1282         switch (optn3) {
1283         case MXU_OPTN3_PTN0:
1284             tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
1285             break;
1286         case MXU_OPTN3_PTN1:
1287         case MXU_OPTN3_PTN2:
1288         case MXU_OPTN3_PTN3:
1289             tcg_gen_shri_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1], 8 * optn3);
1290             break;
1291         case MXU_OPTN3_PTN4:
1292             tcg_gen_movi_i32(mxu_gpr[XRa - 1], 0);
1293             break;
1294         }
1295     } else if (unlikely(XRb == XRc)) {
1296         /* both operands same -> just rotation or moving from any of them */
1297         switch (optn3) {
1298         case MXU_OPTN3_PTN0:
1299         case MXU_OPTN3_PTN4:
1300             tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
1301             break;
1302         case MXU_OPTN3_PTN1:
1303         case MXU_OPTN3_PTN2:
1304         case MXU_OPTN3_PTN3:
1305             tcg_gen_rotli_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1], 8 * optn3);
1306             break;
1307         }
1308     } else {
1309         /* the most general case */
1310         switch (optn3) {
1311         case MXU_OPTN3_PTN0:
1312             {
1313                 /*                                         */
1314                 /*         XRb                XRc          */
1315                 /*  +---------------+                      */
1316                 /*  | A   B   C   D |    E   F   G   H     */
1317                 /*  +-------+-------+                      */
1318                 /*          |                              */
1319                 /*         XRa                             */
1320                 /*                                         */
1321 
1322                 tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRb - 1]);
1323             }
1324             break;
1325         case MXU_OPTN3_PTN1:
1326             {
1327                 /*                                         */
1328                 /*         XRb                 XRc         */
1329                 /*      +-------------------+              */
1330                 /*    A | B   C   D       E | F   G   H    */
1331                 /*      +---------+---------+              */
1332                 /*                |                        */
1333                 /*               XRa                       */
1334                 /*                                         */
1335 
1336                 TCGv_i32 t0 = tcg_temp_new();
1337                 TCGv_i32 t1 = tcg_temp_new();
1338 
1339                 tcg_gen_andi_i32(t0, mxu_gpr[XRb - 1], 0x00FFFFFF);
1340                 tcg_gen_shli_i32(t0, t0, 8);
1341 
1342                 tcg_gen_andi_i32(t1, mxu_gpr[XRc - 1], 0xFF000000);
1343                 tcg_gen_shri_i32(t1, t1, 24);
1344 
1345                 tcg_gen_or_i32(mxu_gpr[XRa - 1], t0, t1);
1346             }
1347             break;
1348         case MXU_OPTN3_PTN2:
1349             {
1350                 /*                                         */
1351                 /*         XRb                 XRc         */
1352                 /*          +-------------------+          */
1353                 /*    A   B | C   D       E   F | G   H    */
1354                 /*          +---------+---------+          */
1355                 /*                    |                    */
1356                 /*                   XRa                   */
1357                 /*                                         */
1358 
1359                 TCGv_i32 t0 = tcg_temp_new();
1360                 TCGv_i32 t1 = tcg_temp_new();
1361 
1362                 tcg_gen_andi_i32(t0, mxu_gpr[XRb - 1], 0x0000FFFF);
1363                 tcg_gen_shli_i32(t0, t0, 16);
1364 
1365                 tcg_gen_andi_i32(t1, mxu_gpr[XRc - 1], 0xFFFF0000);
1366                 tcg_gen_shri_i32(t1, t1, 16);
1367 
1368                 tcg_gen_or_i32(mxu_gpr[XRa - 1], t0, t1);
1369             }
1370             break;
1371         case MXU_OPTN3_PTN3:
1372             {
1373                 /*                                         */
1374                 /*         XRb                 XRc         */
1375                 /*              +-------------------+      */
1376                 /*    A   B   C | D       E   F   G | H    */
1377                 /*              +---------+---------+      */
1378                 /*                        |                */
1379                 /*                       XRa               */
1380                 /*                                         */
1381 
1382                 TCGv_i32 t0 = tcg_temp_new();
1383                 TCGv_i32 t1 = tcg_temp_new();
1384 
1385                 tcg_gen_andi_i32(t0, mxu_gpr[XRb - 1], 0x000000FF);
1386                 tcg_gen_shli_i32(t0, t0, 24);
1387 
1388                 tcg_gen_andi_i32(t1, mxu_gpr[XRc - 1], 0xFFFFFF00);
1389                 tcg_gen_shri_i32(t1, t1, 8);
1390 
1391                 tcg_gen_or_i32(mxu_gpr[XRa - 1], t0, t1);
1392             }
1393             break;
1394         case MXU_OPTN3_PTN4:
1395             {
1396                 /*                                         */
1397                 /*         XRb                 XRc         */
1398                 /*                     +---------------+   */
1399                 /*    A   B   C   D    | E   F   G   H |   */
1400                 /*                     +-------+-------+   */
1401                 /*                             |           */
1402                 /*                            XRa          */
1403                 /*                                         */
1404 
1405                 tcg_gen_mov_i32(mxu_gpr[XRa - 1], mxu_gpr[XRc - 1]);
1406             }
1407             break;
1408         }
1409     }
1410 }
1411 
1412 
1413 /*
1414  * Decoding engine for MXU
1415  * =======================
1416  */
1417 
1418 static void decode_opc_mxu__pool00(DisasContext *ctx)
1419 {
1420     uint32_t opcode = extract32(ctx->opcode, 18, 3);
1421 
1422     switch (opcode) {
1423     case OPC_MXU_S32MAX:
1424     case OPC_MXU_S32MIN:
1425         gen_mxu_S32MAX_S32MIN(ctx);
1426         break;
1427     case OPC_MXU_D16MAX:
1428     case OPC_MXU_D16MIN:
1429         gen_mxu_D16MAX_D16MIN(ctx);
1430         break;
1431     case OPC_MXU_Q8MAX:
1432     case OPC_MXU_Q8MIN:
1433         gen_mxu_Q8MAX_Q8MIN(ctx);
1434         break;
1435     default:
1436         MIPS_INVAL("decode_opc_mxu");
1437         gen_reserved_instruction(ctx);
1438         break;
1439     }
1440 }
1441 
1442 static void decode_opc_mxu__pool04(DisasContext *ctx)
1443 {
1444     uint32_t opcode = extract32(ctx->opcode, 20, 1);
1445 
1446     switch (opcode) {
1447     case OPC_MXU_S32LDD:
1448     case OPC_MXU_S32LDDR:
1449         gen_mxu_s32ldd_s32lddr(ctx);
1450         break;
1451     default:
1452         MIPS_INVAL("decode_opc_mxu");
1453         gen_reserved_instruction(ctx);
1454         break;
1455     }
1456 }
1457 
1458 static void decode_opc_mxu__pool16(DisasContext *ctx)
1459 {
1460     uint32_t opcode = extract32(ctx->opcode, 18, 3);
1461 
1462     switch (opcode) {
1463     case OPC_MXU_S32ALNI:
1464         gen_mxu_S32ALNI(ctx);
1465         break;
1466     case OPC_MXU_S32NOR:
1467         gen_mxu_S32NOR(ctx);
1468         break;
1469     case OPC_MXU_S32AND:
1470         gen_mxu_S32AND(ctx);
1471         break;
1472     case OPC_MXU_S32OR:
1473         gen_mxu_S32OR(ctx);
1474         break;
1475     case OPC_MXU_S32XOR:
1476         gen_mxu_S32XOR(ctx);
1477         break;
1478     default:
1479         MIPS_INVAL("decode_opc_mxu");
1480         gen_reserved_instruction(ctx);
1481         break;
1482     }
1483 }
1484 
1485 static void decode_opc_mxu__pool19(DisasContext *ctx)
1486 {
1487     uint32_t opcode = extract32(ctx->opcode, 22, 2);
1488 
1489     switch (opcode) {
1490     case OPC_MXU_Q8MUL:
1491     case OPC_MXU_Q8MULSU:
1492         gen_mxu_q8mul_q8mulsu(ctx);
1493         break;
1494     default:
1495         MIPS_INVAL("decode_opc_mxu");
1496         gen_reserved_instruction(ctx);
1497         break;
1498     }
1499 }
1500 
1501 bool decode_ase_mxu(DisasContext *ctx, uint32_t insn)
1502 {
1503     uint32_t opcode = extract32(insn, 0, 6);
1504 
1505     if (opcode == OPC_MXU_S32M2I) {
1506         gen_mxu_s32m2i(ctx);
1507         return true;
1508     }
1509 
1510     if (opcode == OPC_MXU_S32I2M) {
1511         gen_mxu_s32i2m(ctx);
1512         return true;
1513     }
1514 
1515     {
1516         TCGv t_mxu_cr = tcg_temp_new();
1517         TCGLabel *l_exit = gen_new_label();
1518 
1519         gen_load_mxu_cr(t_mxu_cr);
1520         tcg_gen_andi_tl(t_mxu_cr, t_mxu_cr, MXU_CR_MXU_EN);
1521         tcg_gen_brcondi_tl(TCG_COND_NE, t_mxu_cr, MXU_CR_MXU_EN, l_exit);
1522 
1523         switch (opcode) {
1524         case OPC_MXU__POOL00:
1525             decode_opc_mxu__pool00(ctx);
1526             break;
1527         case OPC_MXU_D16MUL:
1528             gen_mxu_d16mul(ctx);
1529             break;
1530         case OPC_MXU_D16MAC:
1531             gen_mxu_d16mac(ctx);
1532             break;
1533         case OPC_MXU__POOL04:
1534             decode_opc_mxu__pool04(ctx);
1535             break;
1536         case OPC_MXU_S8LDD:
1537             gen_mxu_s8ldd(ctx);
1538             break;
1539         case OPC_MXU__POOL16:
1540             decode_opc_mxu__pool16(ctx);
1541             break;
1542         case OPC_MXU__POOL19:
1543             decode_opc_mxu__pool19(ctx);
1544             break;
1545         default:
1546             MIPS_INVAL("decode_opc_mxu");
1547             gen_reserved_instruction(ctx);
1548         }
1549 
1550         gen_set_label(l_exit);
1551     }
1552 
1553     return true;
1554 }
1555