1/*
2 * translate/vmx-impl.c
3 *
4 * Altivec/VMX translation
5 */
6
7/***                      Altivec vector extension                         ***/
8/* Altivec registers moves */
9
10static inline TCGv_ptr gen_avr_ptr(int reg)
11{
12    TCGv_ptr r = tcg_temp_new_ptr();
13    tcg_gen_addi_ptr(r, cpu_env, avr_full_offset(reg));
14    return r;
15}
16
17#define GEN_VR_LDX(name, opc2, opc3)                                          \
18static void glue(gen_, name)(DisasContext *ctx)                               \
19{                                                                             \
20    TCGv EA;                                                                  \
21    TCGv_i64 avr;                                                             \
22    if (unlikely(!ctx->altivec_enabled)) {                                    \
23        gen_exception(ctx, POWERPC_EXCP_VPU);                                 \
24        return;                                                               \
25    }                                                                         \
26    gen_set_access_type(ctx, ACCESS_INT);                                     \
27    avr = tcg_temp_new_i64();                                                 \
28    EA = tcg_temp_new();                                                      \
29    gen_addr_reg_index(ctx, EA);                                              \
30    tcg_gen_andi_tl(EA, EA, ~0xf);                                            \
31    /*                                                                        \
32     * We only need to swap high and low halves. gen_qemu_ld64_i64            \
33     * does necessary 64-bit byteswap already.                                \
34     */                                                                       \
35    if (ctx->le_mode) {                                                       \
36        gen_qemu_ld64_i64(ctx, avr, EA);                                      \
37        set_avr64(rD(ctx->opcode), avr, false);                               \
38        tcg_gen_addi_tl(EA, EA, 8);                                           \
39        gen_qemu_ld64_i64(ctx, avr, EA);                                      \
40        set_avr64(rD(ctx->opcode), avr, true);                                \
41    } else {                                                                  \
42        gen_qemu_ld64_i64(ctx, avr, EA);                                      \
43        set_avr64(rD(ctx->opcode), avr, true);                                \
44        tcg_gen_addi_tl(EA, EA, 8);                                           \
45        gen_qemu_ld64_i64(ctx, avr, EA);                                      \
46        set_avr64(rD(ctx->opcode), avr, false);                               \
47    }                                                                         \
48}
49
50#define GEN_VR_STX(name, opc2, opc3)                                          \
51static void gen_st##name(DisasContext *ctx)                                   \
52{                                                                             \
53    TCGv EA;                                                                  \
54    TCGv_i64 avr;                                                             \
55    if (unlikely(!ctx->altivec_enabled)) {                                    \
56        gen_exception(ctx, POWERPC_EXCP_VPU);                                 \
57        return;                                                               \
58    }                                                                         \
59    gen_set_access_type(ctx, ACCESS_INT);                                     \
60    avr = tcg_temp_new_i64();                                                 \
61    EA = tcg_temp_new();                                                      \
62    gen_addr_reg_index(ctx, EA);                                              \
63    tcg_gen_andi_tl(EA, EA, ~0xf);                                            \
64    /*                                                                        \
65     * We only need to swap high and low halves. gen_qemu_st64_i64            \
66     * does necessary 64-bit byteswap already.                                \
67     */                                                                       \
68    if (ctx->le_mode) {                                                       \
69        get_avr64(avr, rD(ctx->opcode), false);                               \
70        gen_qemu_st64_i64(ctx, avr, EA);                                      \
71        tcg_gen_addi_tl(EA, EA, 8);                                           \
72        get_avr64(avr, rD(ctx->opcode), true);                                \
73        gen_qemu_st64_i64(ctx, avr, EA);                                      \
74    } else {                                                                  \
75        get_avr64(avr, rD(ctx->opcode), true);                                \
76        gen_qemu_st64_i64(ctx, avr, EA);                                      \
77        tcg_gen_addi_tl(EA, EA, 8);                                           \
78        get_avr64(avr, rD(ctx->opcode), false);                               \
79        gen_qemu_st64_i64(ctx, avr, EA);                                      \
80    }                                                                         \
81}
82
83#define GEN_VR_LVE(name, opc2, opc3, size)                              \
84static void gen_lve##name(DisasContext *ctx)                            \
85    {                                                                   \
86        TCGv EA;                                                        \
87        TCGv_ptr rs;                                                    \
88        if (unlikely(!ctx->altivec_enabled)) {                          \
89            gen_exception(ctx, POWERPC_EXCP_VPU);                       \
90            return;                                                     \
91        }                                                               \
92        gen_set_access_type(ctx, ACCESS_INT);                           \
93        EA = tcg_temp_new();                                            \
94        gen_addr_reg_index(ctx, EA);                                    \
95        if (size > 1) {                                                 \
96            tcg_gen_andi_tl(EA, EA, ~(size - 1));                       \
97        }                                                               \
98        rs = gen_avr_ptr(rS(ctx->opcode));                              \
99        gen_helper_lve##name(cpu_env, rs, EA);                          \
100    }
101
102#define GEN_VR_STVE(name, opc2, opc3, size)                             \
103static void gen_stve##name(DisasContext *ctx)                           \
104    {                                                                   \
105        TCGv EA;                                                        \
106        TCGv_ptr rs;                                                    \
107        if (unlikely(!ctx->altivec_enabled)) {                          \
108            gen_exception(ctx, POWERPC_EXCP_VPU);                       \
109            return;                                                     \
110        }                                                               \
111        gen_set_access_type(ctx, ACCESS_INT);                           \
112        EA = tcg_temp_new();                                            \
113        gen_addr_reg_index(ctx, EA);                                    \
114        if (size > 1) {                                                 \
115            tcg_gen_andi_tl(EA, EA, ~(size - 1));                       \
116        }                                                               \
117        rs = gen_avr_ptr(rS(ctx->opcode));                              \
118        gen_helper_stve##name(cpu_env, rs, EA);                         \
119    }
120
121GEN_VR_LDX(lvx, 0x07, 0x03);
122/* As we don't emulate the cache, lvxl is strictly equivalent to lvx */
123GEN_VR_LDX(lvxl, 0x07, 0x0B);
124
125GEN_VR_LVE(bx, 0x07, 0x00, 1);
126GEN_VR_LVE(hx, 0x07, 0x01, 2);
127GEN_VR_LVE(wx, 0x07, 0x02, 4);
128
129GEN_VR_STX(svx, 0x07, 0x07);
130/* As we don't emulate the cache, stvxl is strictly equivalent to stvx */
131GEN_VR_STX(svxl, 0x07, 0x0F);
132
133GEN_VR_STVE(bx, 0x07, 0x04, 1);
134GEN_VR_STVE(hx, 0x07, 0x05, 2);
135GEN_VR_STVE(wx, 0x07, 0x06, 4);
136
137static void gen_mfvscr(DisasContext *ctx)
138{
139    TCGv_i32 t;
140    TCGv_i64 avr;
141    if (unlikely(!ctx->altivec_enabled)) {
142        gen_exception(ctx, POWERPC_EXCP_VPU);
143        return;
144    }
145    avr = tcg_temp_new_i64();
146    tcg_gen_movi_i64(avr, 0);
147    set_avr64(rD(ctx->opcode), avr, true);
148    t = tcg_temp_new_i32();
149    gen_helper_mfvscr(t, cpu_env);
150    tcg_gen_extu_i32_i64(avr, t);
151    set_avr64(rD(ctx->opcode), avr, false);
152}
153
154static void gen_mtvscr(DisasContext *ctx)
155{
156    TCGv_i32 val;
157    int bofs;
158
159    if (unlikely(!ctx->altivec_enabled)) {
160        gen_exception(ctx, POWERPC_EXCP_VPU);
161        return;
162    }
163
164    val = tcg_temp_new_i32();
165    bofs = avr_full_offset(rB(ctx->opcode));
166#if HOST_BIG_ENDIAN
167    bofs += 3 * 4;
168#endif
169
170    tcg_gen_ld_i32(val, cpu_env, bofs);
171    gen_helper_mtvscr(cpu_env, val);
172}
173
174static void gen_vx_vmul10(DisasContext *ctx, bool add_cin, bool ret_carry)
175{
176    TCGv_i64 t0;
177    TCGv_i64 t1;
178    TCGv_i64 t2;
179    TCGv_i64 avr;
180    TCGv_i64 ten, z;
181
182    if (unlikely(!ctx->altivec_enabled)) {
183        gen_exception(ctx, POWERPC_EXCP_VPU);
184        return;
185    }
186
187    t0 = tcg_temp_new_i64();
188    t1 = tcg_temp_new_i64();
189    t2 = tcg_temp_new_i64();
190    avr = tcg_temp_new_i64();
191    ten = tcg_constant_i64(10);
192    z = tcg_constant_i64(0);
193
194    if (add_cin) {
195        get_avr64(avr, rA(ctx->opcode), false);
196        tcg_gen_mulu2_i64(t0, t1, avr, ten);
197        get_avr64(avr, rB(ctx->opcode), false);
198        tcg_gen_andi_i64(t2, avr, 0xF);
199        tcg_gen_add2_i64(avr, t2, t0, t1, t2, z);
200        set_avr64(rD(ctx->opcode), avr, false);
201    } else {
202        get_avr64(avr, rA(ctx->opcode), false);
203        tcg_gen_mulu2_i64(avr, t2, avr, ten);
204        set_avr64(rD(ctx->opcode), avr, false);
205    }
206
207    if (ret_carry) {
208        get_avr64(avr, rA(ctx->opcode), true);
209        tcg_gen_mulu2_i64(t0, t1, avr, ten);
210        tcg_gen_add2_i64(t0, avr, t0, t1, t2, z);
211        set_avr64(rD(ctx->opcode), avr, false);
212        set_avr64(rD(ctx->opcode), z, true);
213    } else {
214        get_avr64(avr, rA(ctx->opcode), true);
215        tcg_gen_mul_i64(t0, avr, ten);
216        tcg_gen_add_i64(avr, t0, t2);
217        set_avr64(rD(ctx->opcode), avr, true);
218    }
219}
220
221#define GEN_VX_VMUL10(name, add_cin, ret_carry)                         \
222    static void glue(gen_, name)(DisasContext *ctx)                     \
223    { gen_vx_vmul10(ctx, add_cin, ret_carry); }
224
225GEN_VX_VMUL10(vmul10uq, 0, 0);
226GEN_VX_VMUL10(vmul10euq, 1, 0);
227GEN_VX_VMUL10(vmul10cuq, 0, 1);
228GEN_VX_VMUL10(vmul10ecuq, 1, 1);
229
230#define GEN_VXFORM_V(name, vece, tcg_op, opc2, opc3)                    \
231static void glue(gen_, name)(DisasContext *ctx)                         \
232{                                                                       \
233    if (unlikely(!ctx->altivec_enabled)) {                              \
234        gen_exception(ctx, POWERPC_EXCP_VPU);                           \
235        return;                                                         \
236    }                                                                   \
237                                                                        \
238    tcg_op(vece,                                                        \
239           avr_full_offset(rD(ctx->opcode)),                            \
240           avr_full_offset(rA(ctx->opcode)),                            \
241           avr_full_offset(rB(ctx->opcode)),                            \
242           16, 16);                                                     \
243}
244
245/* Logical operations */
246GEN_VXFORM_V(vand, MO_64, tcg_gen_gvec_and, 2, 16);
247GEN_VXFORM_V(vandc, MO_64, tcg_gen_gvec_andc, 2, 17);
248GEN_VXFORM_V(vor, MO_64, tcg_gen_gvec_or, 2, 18);
249GEN_VXFORM_V(vxor, MO_64, tcg_gen_gvec_xor, 2, 19);
250GEN_VXFORM_V(vnor, MO_64, tcg_gen_gvec_nor, 2, 20);
251GEN_VXFORM_V(veqv, MO_64, tcg_gen_gvec_eqv, 2, 26);
252GEN_VXFORM_V(vnand, MO_64, tcg_gen_gvec_nand, 2, 22);
253GEN_VXFORM_V(vorc, MO_64, tcg_gen_gvec_orc, 2, 21);
254
255#define GEN_VXFORM(name, opc2, opc3)                                    \
256static void glue(gen_, name)(DisasContext *ctx)                         \
257{                                                                       \
258    TCGv_ptr ra, rb, rd;                                                \
259    if (unlikely(!ctx->altivec_enabled)) {                              \
260        gen_exception(ctx, POWERPC_EXCP_VPU);                           \
261        return;                                                         \
262    }                                                                   \
263    ra = gen_avr_ptr(rA(ctx->opcode));                                  \
264    rb = gen_avr_ptr(rB(ctx->opcode));                                  \
265    rd = gen_avr_ptr(rD(ctx->opcode));                                  \
266    gen_helper_##name(rd, ra, rb);                                      \
267}
268
269#define GEN_VXFORM_TRANS(name, opc2, opc3)                              \
270static void glue(gen_, name)(DisasContext *ctx)                         \
271{                                                                       \
272    if (unlikely(!ctx->altivec_enabled)) {                              \
273        gen_exception(ctx, POWERPC_EXCP_VPU);                           \
274        return;                                                         \
275    }                                                                   \
276    trans_##name(ctx);                                                  \
277}
278
279#define GEN_VXFORM_ENV(name, opc2, opc3)                                \
280static void glue(gen_, name)(DisasContext *ctx)                         \
281{                                                                       \
282    TCGv_ptr ra, rb, rd;                                                \
283    if (unlikely(!ctx->altivec_enabled)) {                              \
284        gen_exception(ctx, POWERPC_EXCP_VPU);                           \
285        return;                                                         \
286    }                                                                   \
287    ra = gen_avr_ptr(rA(ctx->opcode));                                  \
288    rb = gen_avr_ptr(rB(ctx->opcode));                                  \
289    rd = gen_avr_ptr(rD(ctx->opcode));                                  \
290    gen_helper_##name(cpu_env, rd, ra, rb);                             \
291}
292
293#define GEN_VXFORM3(name, opc2, opc3)                                   \
294static void glue(gen_, name)(DisasContext *ctx)                         \
295{                                                                       \
296    TCGv_ptr ra, rb, rc, rd;                                            \
297    if (unlikely(!ctx->altivec_enabled)) {                              \
298        gen_exception(ctx, POWERPC_EXCP_VPU);                           \
299        return;                                                         \
300    }                                                                   \
301    ra = gen_avr_ptr(rA(ctx->opcode));                                  \
302    rb = gen_avr_ptr(rB(ctx->opcode));                                  \
303    rc = gen_avr_ptr(rC(ctx->opcode));                                  \
304    rd = gen_avr_ptr(rD(ctx->opcode));                                  \
305    gen_helper_##name(rd, ra, rb, rc);                                  \
306}
307
308/*
309 * Support for Altivec instruction pairs that use bit 31 (Rc) as
310 * an opcode bit.  In general, these pairs come from different
311 * versions of the ISA, so we must also support a pair of flags for
312 * each instruction.
313 */
314#define GEN_VXFORM_DUAL(name0, flg0, flg2_0, name1, flg1, flg2_1)          \
315static void glue(gen_, name0##_##name1)(DisasContext *ctx)             \
316{                                                                      \
317    if ((Rc(ctx->opcode) == 0) &&                                      \
318        ((ctx->insns_flags & flg0) || (ctx->insns_flags2 & flg2_0))) { \
319        gen_##name0(ctx);                                              \
320    } else if ((Rc(ctx->opcode) == 1) &&                               \
321        ((ctx->insns_flags & flg1) || (ctx->insns_flags2 & flg2_1))) { \
322        gen_##name1(ctx);                                              \
323    } else {                                                           \
324        gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL);            \
325    }                                                                  \
326}
327
328/*
329 * We use this macro if one instruction is realized with direct
330 * translation, and second one with helper.
331 */
332#define GEN_VXFORM_TRANS_DUAL(name0, flg0, flg2_0, name1, flg1, flg2_1)\
333static void glue(gen_, name0##_##name1)(DisasContext *ctx)             \
334{                                                                      \
335    if ((Rc(ctx->opcode) == 0) &&                                      \
336        ((ctx->insns_flags & flg0) || (ctx->insns_flags2 & flg2_0))) { \
337        if (unlikely(!ctx->altivec_enabled)) {                         \
338            gen_exception(ctx, POWERPC_EXCP_VPU);                      \
339            return;                                                    \
340        }                                                              \
341        trans_##name0(ctx);                                            \
342    } else if ((Rc(ctx->opcode) == 1) &&                               \
343        ((ctx->insns_flags & flg1) || (ctx->insns_flags2 & flg2_1))) { \
344        gen_##name1(ctx);                                              \
345    } else {                                                           \
346        gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL);            \
347    }                                                                  \
348}
349
350/* Adds support to provide invalid mask */
351#define GEN_VXFORM_DUAL_EXT(name0, flg0, flg2_0, inval0,                \
352                            name1, flg1, flg2_1, inval1)                \
353static void glue(gen_, name0##_##name1)(DisasContext *ctx)              \
354{                                                                       \
355    if ((Rc(ctx->opcode) == 0) &&                                       \
356        ((ctx->insns_flags & flg0) || (ctx->insns_flags2 & flg2_0)) &&  \
357        !(ctx->opcode & inval0)) {                                      \
358        gen_##name0(ctx);                                               \
359    } else if ((Rc(ctx->opcode) == 1) &&                                \
360               ((ctx->insns_flags & flg1) || (ctx->insns_flags2 & flg2_1)) && \
361               !(ctx->opcode & inval1)) {                               \
362        gen_##name1(ctx);                                               \
363    } else {                                                            \
364        gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL);             \
365    }                                                                   \
366}
367
368#define GEN_VXFORM_HETRO(name, opc2, opc3)                              \
369static void glue(gen_, name)(DisasContext *ctx)                         \
370{                                                                       \
371    TCGv_ptr rb;                                                        \
372    if (unlikely(!ctx->altivec_enabled)) {                              \
373        gen_exception(ctx, POWERPC_EXCP_VPU);                           \
374        return;                                                         \
375    }                                                                   \
376    rb = gen_avr_ptr(rB(ctx->opcode));                                  \
377    gen_helper_##name(cpu_gpr[rD(ctx->opcode)], cpu_gpr[rA(ctx->opcode)], rb); \
378}
379
380GEN_VXFORM_V(vaddubm, MO_8, tcg_gen_gvec_add, 0, 0);
381GEN_VXFORM_DUAL_EXT(vaddubm, PPC_ALTIVEC, PPC_NONE, 0,       \
382                    vmul10cuq, PPC_NONE, PPC2_ISA300, 0x0000F800)
383GEN_VXFORM_V(vadduhm, MO_16, tcg_gen_gvec_add, 0, 1);
384GEN_VXFORM_DUAL(vadduhm, PPC_ALTIVEC, PPC_NONE,  \
385                vmul10ecuq, PPC_NONE, PPC2_ISA300)
386GEN_VXFORM_V(vadduwm, MO_32, tcg_gen_gvec_add, 0, 2);
387GEN_VXFORM_V(vaddudm, MO_64, tcg_gen_gvec_add, 0, 3);
388GEN_VXFORM_V(vsububm, MO_8, tcg_gen_gvec_sub, 0, 16);
389GEN_VXFORM_V(vsubuhm, MO_16, tcg_gen_gvec_sub, 0, 17);
390GEN_VXFORM_V(vsubuwm, MO_32, tcg_gen_gvec_sub, 0, 18);
391GEN_VXFORM_V(vsubudm, MO_64, tcg_gen_gvec_sub, 0, 19);
392GEN_VXFORM_V(vmaxub, MO_8, tcg_gen_gvec_umax, 1, 0);
393GEN_VXFORM_V(vmaxuh, MO_16, tcg_gen_gvec_umax, 1, 1);
394GEN_VXFORM_V(vmaxuw, MO_32, tcg_gen_gvec_umax, 1, 2);
395GEN_VXFORM_V(vmaxud, MO_64, tcg_gen_gvec_umax, 1, 3);
396GEN_VXFORM_V(vmaxsb, MO_8, tcg_gen_gvec_smax, 1, 4);
397GEN_VXFORM_V(vmaxsh, MO_16, tcg_gen_gvec_smax, 1, 5);
398GEN_VXFORM_V(vmaxsw, MO_32, tcg_gen_gvec_smax, 1, 6);
399GEN_VXFORM_V(vmaxsd, MO_64, tcg_gen_gvec_smax, 1, 7);
400GEN_VXFORM_V(vminub, MO_8, tcg_gen_gvec_umin, 1, 8);
401GEN_VXFORM_V(vminuh, MO_16, tcg_gen_gvec_umin, 1, 9);
402GEN_VXFORM_V(vminuw, MO_32, tcg_gen_gvec_umin, 1, 10);
403GEN_VXFORM_V(vminud, MO_64, tcg_gen_gvec_umin, 1, 11);
404GEN_VXFORM_V(vminsb, MO_8, tcg_gen_gvec_smin, 1, 12);
405GEN_VXFORM_V(vminsh, MO_16, tcg_gen_gvec_smin, 1, 13);
406GEN_VXFORM_V(vminsw, MO_32, tcg_gen_gvec_smin, 1, 14);
407GEN_VXFORM_V(vminsd, MO_64, tcg_gen_gvec_smin, 1, 15);
408GEN_VXFORM(vmrghb, 6, 0);
409GEN_VXFORM(vmrghh, 6, 1);
410GEN_VXFORM(vmrghw, 6, 2);
411GEN_VXFORM(vmrglb, 6, 4);
412GEN_VXFORM(vmrglh, 6, 5);
413GEN_VXFORM(vmrglw, 6, 6);
414
415static void trans_vmrgew(DisasContext *ctx)
416{
417    int VT = rD(ctx->opcode);
418    int VA = rA(ctx->opcode);
419    int VB = rB(ctx->opcode);
420    TCGv_i64 tmp = tcg_temp_new_i64();
421    TCGv_i64 avr = tcg_temp_new_i64();
422
423    get_avr64(avr, VB, true);
424    tcg_gen_shri_i64(tmp, avr, 32);
425    get_avr64(avr, VA, true);
426    tcg_gen_deposit_i64(avr, avr, tmp, 0, 32);
427    set_avr64(VT, avr, true);
428
429    get_avr64(avr, VB, false);
430    tcg_gen_shri_i64(tmp, avr, 32);
431    get_avr64(avr, VA, false);
432    tcg_gen_deposit_i64(avr, avr, tmp, 0, 32);
433    set_avr64(VT, avr, false);
434}
435
436static void trans_vmrgow(DisasContext *ctx)
437{
438    int VT = rD(ctx->opcode);
439    int VA = rA(ctx->opcode);
440    int VB = rB(ctx->opcode);
441    TCGv_i64 t0 = tcg_temp_new_i64();
442    TCGv_i64 t1 = tcg_temp_new_i64();
443    TCGv_i64 avr = tcg_temp_new_i64();
444
445    get_avr64(t0, VB, true);
446    get_avr64(t1, VA, true);
447    tcg_gen_deposit_i64(avr, t0, t1, 32, 32);
448    set_avr64(VT, avr, true);
449
450    get_avr64(t0, VB, false);
451    get_avr64(t1, VA, false);
452    tcg_gen_deposit_i64(avr, t0, t1, 32, 32);
453    set_avr64(VT, avr, false);
454}
455
456/*
457 * lvsl VRT,RA,RB - Load Vector for Shift Left
458 *
459 * Let the EA be the sum (rA|0)+(rB). Let sh=EA[28–31].
460 * Let X be the 32-byte value 0x00 || 0x01 || 0x02 || ... || 0x1E || 0x1F.
461 * Bytes sh:sh+15 of X are placed into vD.
462 */
463static void trans_lvsl(DisasContext *ctx)
464{
465    int VT = rD(ctx->opcode);
466    TCGv_i64 result = tcg_temp_new_i64();
467    TCGv_i64 sh = tcg_temp_new_i64();
468    TCGv EA = tcg_temp_new();
469
470    /* Get sh(from description) by anding EA with 0xf. */
471    gen_addr_reg_index(ctx, EA);
472    tcg_gen_extu_tl_i64(sh, EA);
473    tcg_gen_andi_i64(sh, sh, 0xfULL);
474
475    /*
476     * Create bytes sh:sh+7 of X(from description) and place them in
477     * higher doubleword of vD.
478     */
479    tcg_gen_muli_i64(sh, sh, 0x0101010101010101ULL);
480    tcg_gen_addi_i64(result, sh, 0x0001020304050607ull);
481    set_avr64(VT, result, true);
482    /*
483     * Create bytes sh+8:sh+15 of X(from description) and place them in
484     * lower doubleword of vD.
485     */
486    tcg_gen_addi_i64(result, sh, 0x08090a0b0c0d0e0fULL);
487    set_avr64(VT, result, false);
488}
489
490/*
491 * lvsr VRT,RA,RB - Load Vector for Shift Right
492 *
493 * Let the EA be the sum (rA|0)+(rB). Let sh=EA[28–31].
494 * Let X be the 32-byte value 0x00 || 0x01 || 0x02 || ... || 0x1E || 0x1F.
495 * Bytes (16-sh):(31-sh) of X are placed into vD.
496 */
497static void trans_lvsr(DisasContext *ctx)
498{
499    int VT = rD(ctx->opcode);
500    TCGv_i64 result = tcg_temp_new_i64();
501    TCGv_i64 sh = tcg_temp_new_i64();
502    TCGv EA = tcg_temp_new();
503
504
505    /* Get sh(from description) by anding EA with 0xf. */
506    gen_addr_reg_index(ctx, EA);
507    tcg_gen_extu_tl_i64(sh, EA);
508    tcg_gen_andi_i64(sh, sh, 0xfULL);
509
510    /*
511     * Create bytes (16-sh):(23-sh) of X(from description) and place them in
512     * higher doubleword of vD.
513     */
514    tcg_gen_muli_i64(sh, sh, 0x0101010101010101ULL);
515    tcg_gen_subfi_i64(result, 0x1011121314151617ULL, sh);
516    set_avr64(VT, result, true);
517    /*
518     * Create bytes (24-sh):(32-sh) of X(from description) and place them in
519     * lower doubleword of vD.
520     */
521    tcg_gen_subfi_i64(result, 0x18191a1b1c1d1e1fULL, sh);
522    set_avr64(VT, result, false);
523}
524
525/*
526 * vsl VRT,VRA,VRB - Vector Shift Left
527 *
528 * Shifting left 128 bit value of vA by value specified in bits 125-127 of vB.
529 * Lowest 3 bits in each byte element of register vB must be identical or
530 * result is undefined.
531 */
532static void trans_vsl(DisasContext *ctx)
533{
534    int VT = rD(ctx->opcode);
535    int VA = rA(ctx->opcode);
536    int VB = rB(ctx->opcode);
537    TCGv_i64 avr = tcg_temp_new_i64();
538    TCGv_i64 sh = tcg_temp_new_i64();
539    TCGv_i64 carry = tcg_temp_new_i64();
540    TCGv_i64 tmp = tcg_temp_new_i64();
541
542    /* Place bits 125-127 of vB in 'sh'. */
543    get_avr64(avr, VB, false);
544    tcg_gen_andi_i64(sh, avr, 0x07ULL);
545
546    /*
547     * Save highest 'sh' bits of lower doubleword element of vA in variable
548     * 'carry' and perform shift on lower doubleword.
549     */
550    get_avr64(avr, VA, false);
551    tcg_gen_subfi_i64(tmp, 32, sh);
552    tcg_gen_shri_i64(carry, avr, 32);
553    tcg_gen_shr_i64(carry, carry, tmp);
554    tcg_gen_shl_i64(avr, avr, sh);
555    set_avr64(VT, avr, false);
556
557    /*
558     * Perform shift on higher doubleword element of vA and replace lowest
559     * 'sh' bits with 'carry'.
560     */
561    get_avr64(avr, VA, true);
562    tcg_gen_shl_i64(avr, avr, sh);
563    tcg_gen_or_i64(avr, avr, carry);
564    set_avr64(VT, avr, true);
565}
566
567/*
568 * vsr VRT,VRA,VRB - Vector Shift Right
569 *
570 * Shifting right 128 bit value of vA by value specified in bits 125-127 of vB.
571 * Lowest 3 bits in each byte element of register vB must be identical or
572 * result is undefined.
573 */
574static void trans_vsr(DisasContext *ctx)
575{
576    int VT = rD(ctx->opcode);
577    int VA = rA(ctx->opcode);
578    int VB = rB(ctx->opcode);
579    TCGv_i64 avr = tcg_temp_new_i64();
580    TCGv_i64 sh = tcg_temp_new_i64();
581    TCGv_i64 carry = tcg_temp_new_i64();
582    TCGv_i64 tmp = tcg_temp_new_i64();
583
584    /* Place bits 125-127 of vB in 'sh'. */
585    get_avr64(avr, VB, false);
586    tcg_gen_andi_i64(sh, avr, 0x07ULL);
587
588    /*
589     * Save lowest 'sh' bits of higher doubleword element of vA in variable
590     * 'carry' and perform shift on higher doubleword.
591     */
592    get_avr64(avr, VA, true);
593    tcg_gen_subfi_i64(tmp, 32, sh);
594    tcg_gen_shli_i64(carry, avr, 32);
595    tcg_gen_shl_i64(carry, carry, tmp);
596    tcg_gen_shr_i64(avr, avr, sh);
597    set_avr64(VT, avr, true);
598    /*
599     * Perform shift on lower doubleword element of vA and replace highest
600     * 'sh' bits with 'carry'.
601     */
602    get_avr64(avr, VA, false);
603    tcg_gen_shr_i64(avr, avr, sh);
604    tcg_gen_or_i64(avr, avr, carry);
605    set_avr64(VT, avr, false);
606}
607
608/*
609 * vgbbd VRT,VRB - Vector Gather Bits by Bytes by Doubleword
610 *
611 * All ith bits (i in range 1 to 8) of each byte of doubleword element in source
612 * register are concatenated and placed into ith byte of appropriate doubleword
613 * element in destination register.
614 *
615 * Following solution is done for both doubleword elements of source register
616 * in parallel, in order to reduce the number of instructions needed(that's why
617 * arrays are used):
618 * First, both doubleword elements of source register vB are placed in
619 * appropriate element of array avr. Bits are gathered in 2x8 iterations(2 for
620 * loops). In first iteration bit 1 of byte 1, bit 2 of byte 2,... bit 8 of
621 * byte 8 are in their final spots so avr[i], i={0,1} can be and-ed with
622 * tcg_mask. For every following iteration, both avr[i] and tcg_mask variables
623 * have to be shifted right for 7 and 8 places, respectively, in order to get
624 * bit 1 of byte 2, bit 2 of byte 3.. bit 7 of byte 8 in their final spots so
625 * shifted avr values(saved in tmp) can be and-ed with new value of tcg_mask...
626 * After first 8 iteration(first loop), all the first bits are in their final
627 * places, all second bits but second bit from eight byte are in their places...
628 * only 1 eight bit from eight byte is in it's place). In second loop we do all
629 * operations symmetrically, in order to get other half of bits in their final
630 * spots. Results for first and second doubleword elements are saved in
631 * result[0] and result[1] respectively. In the end those results are saved in
632 * appropriate doubleword element of destination register vD.
633 */
634static void trans_vgbbd(DisasContext *ctx)
635{
636    int VT = rD(ctx->opcode);
637    int VB = rB(ctx->opcode);
638    TCGv_i64 tmp = tcg_temp_new_i64();
639    uint64_t mask = 0x8040201008040201ULL;
640    int i, j;
641
642    TCGv_i64 result[2];
643    result[0] = tcg_temp_new_i64();
644    result[1] = tcg_temp_new_i64();
645    TCGv_i64 avr[2];
646    avr[0] = tcg_temp_new_i64();
647    avr[1] = tcg_temp_new_i64();
648    TCGv_i64 tcg_mask = tcg_temp_new_i64();
649
650    tcg_gen_movi_i64(tcg_mask, mask);
651    for (j = 0; j < 2; j++) {
652        get_avr64(avr[j], VB, j);
653        tcg_gen_and_i64(result[j], avr[j], tcg_mask);
654    }
655    for (i = 1; i < 8; i++) {
656        tcg_gen_movi_i64(tcg_mask, mask >> (i * 8));
657        for (j = 0; j < 2; j++) {
658            tcg_gen_shri_i64(tmp, avr[j], i * 7);
659            tcg_gen_and_i64(tmp, tmp, tcg_mask);
660            tcg_gen_or_i64(result[j], result[j], tmp);
661        }
662    }
663    for (i = 1; i < 8; i++) {
664        tcg_gen_movi_i64(tcg_mask, mask << (i * 8));
665        for (j = 0; j < 2; j++) {
666            tcg_gen_shli_i64(tmp, avr[j], i * 7);
667            tcg_gen_and_i64(tmp, tmp, tcg_mask);
668            tcg_gen_or_i64(result[j], result[j], tmp);
669        }
670    }
671    for (j = 0; j < 2; j++) {
672        set_avr64(VT, result[j], j);
673    }
674}
675
676/*
677 * vclzw VRT,VRB - Vector Count Leading Zeros Word
678 *
679 * Counting the number of leading zero bits of each word element in source
680 * register and placing result in appropriate word element of destination
681 * register.
682 */
683static void trans_vclzw(DisasContext *ctx)
684{
685    int VT = rD(ctx->opcode);
686    int VB = rB(ctx->opcode);
687    TCGv_i32 tmp = tcg_temp_new_i32();
688    int i;
689
690    /* Perform count for every word element using tcg_gen_clzi_i32. */
691    for (i = 0; i < 4; i++) {
692        tcg_gen_ld_i32(tmp, cpu_env,
693            offsetof(CPUPPCState, vsr[32 + VB].u64[0]) + i * 4);
694        tcg_gen_clzi_i32(tmp, tmp, 32);
695        tcg_gen_st_i32(tmp, cpu_env,
696            offsetof(CPUPPCState, vsr[32 + VT].u64[0]) + i * 4);
697    }
698}
699
700/*
701 * vclzd VRT,VRB - Vector Count Leading Zeros Doubleword
702 *
703 * Counting the number of leading zero bits of each doubleword element in source
704 * register and placing result in appropriate doubleword element of destination
705 * register.
706 */
707static void trans_vclzd(DisasContext *ctx)
708{
709    int VT = rD(ctx->opcode);
710    int VB = rB(ctx->opcode);
711    TCGv_i64 avr = tcg_temp_new_i64();
712
713    /* high doubleword */
714    get_avr64(avr, VB, true);
715    tcg_gen_clzi_i64(avr, avr, 64);
716    set_avr64(VT, avr, true);
717
718    /* low doubleword */
719    get_avr64(avr, VB, false);
720    tcg_gen_clzi_i64(avr, avr, 64);
721    set_avr64(VT, avr, false);
722}
723
724GEN_VXFORM_V(vmuluwm, MO_32, tcg_gen_gvec_mul, 4, 2);
725GEN_VXFORM(vsrv, 2, 28);
726GEN_VXFORM(vslv, 2, 29);
727GEN_VXFORM(vslo, 6, 16);
728GEN_VXFORM(vsro, 6, 17);
729
730static bool do_vector_gvec3_VX(DisasContext *ctx, arg_VX *a, int vece,
731                               void (*gen_gvec)(unsigned, uint32_t, uint32_t,
732                                                uint32_t, uint32_t, uint32_t))
733{
734    REQUIRE_VECTOR(ctx);
735
736    gen_gvec(vece, avr_full_offset(a->vrt), avr_full_offset(a->vra),
737             avr_full_offset(a->vrb), 16, 16);
738
739    return true;
740}
741
742TRANS_FLAGS(ALTIVEC, VSLB, do_vector_gvec3_VX, MO_8, tcg_gen_gvec_shlv);
743TRANS_FLAGS(ALTIVEC, VSLH, do_vector_gvec3_VX, MO_16, tcg_gen_gvec_shlv);
744TRANS_FLAGS(ALTIVEC, VSLW, do_vector_gvec3_VX, MO_32, tcg_gen_gvec_shlv);
745TRANS_FLAGS2(ALTIVEC_207, VSLD, do_vector_gvec3_VX, MO_64, tcg_gen_gvec_shlv);
746
747TRANS_FLAGS(ALTIVEC, VSRB, do_vector_gvec3_VX, MO_8, tcg_gen_gvec_shrv);
748TRANS_FLAGS(ALTIVEC, VSRH, do_vector_gvec3_VX, MO_16, tcg_gen_gvec_shrv);
749TRANS_FLAGS(ALTIVEC, VSRW, do_vector_gvec3_VX, MO_32, tcg_gen_gvec_shrv);
750TRANS_FLAGS2(ALTIVEC_207, VSRD, do_vector_gvec3_VX, MO_64, tcg_gen_gvec_shrv);
751
752TRANS_FLAGS(ALTIVEC, VSRAB, do_vector_gvec3_VX, MO_8, tcg_gen_gvec_sarv);
753TRANS_FLAGS(ALTIVEC, VSRAH, do_vector_gvec3_VX, MO_16, tcg_gen_gvec_sarv);
754TRANS_FLAGS(ALTIVEC, VSRAW, do_vector_gvec3_VX, MO_32, tcg_gen_gvec_sarv);
755TRANS_FLAGS2(ALTIVEC_207, VSRAD, do_vector_gvec3_VX, MO_64, tcg_gen_gvec_sarv);
756
757TRANS_FLAGS(ALTIVEC, VRLB, do_vector_gvec3_VX, MO_8, tcg_gen_gvec_rotlv)
758TRANS_FLAGS(ALTIVEC, VRLH, do_vector_gvec3_VX, MO_16, tcg_gen_gvec_rotlv)
759TRANS_FLAGS(ALTIVEC, VRLW, do_vector_gvec3_VX, MO_32, tcg_gen_gvec_rotlv)
760TRANS_FLAGS2(ALTIVEC_207, VRLD, do_vector_gvec3_VX, MO_64, tcg_gen_gvec_rotlv)
761
762static TCGv_vec do_vrl_mask_vec(unsigned vece, TCGv_vec vrb)
763{
764    TCGv_vec t0 = tcg_temp_new_vec_matching(vrb),
765             t1 = tcg_temp_new_vec_matching(vrb),
766             t2 = tcg_temp_new_vec_matching(vrb),
767             ones = tcg_constant_vec_matching(vrb, vece, -1);
768
769    /* Extract b and e */
770    tcg_gen_dupi_vec(vece, t2, (8 << vece) - 1);
771
772    tcg_gen_shri_vec(vece, t0, vrb, 16);
773    tcg_gen_and_vec(vece, t0, t0, t2);
774
775    tcg_gen_shri_vec(vece, t1, vrb, 8);
776    tcg_gen_and_vec(vece, t1, t1, t2);
777
778    /* Compare b and e to negate the mask where begin > end */
779    tcg_gen_cmp_vec(TCG_COND_GT, vece, t2, t0, t1);
780
781    /* Create the mask with (~0 >> b) ^ ((~0 >> e) >> 1) */
782    tcg_gen_shrv_vec(vece, t0, ones, t0);
783    tcg_gen_shrv_vec(vece, t1, ones, t1);
784    tcg_gen_shri_vec(vece, t1, t1, 1);
785    tcg_gen_xor_vec(vece, t0, t0, t1);
786
787    /* negate the mask */
788    tcg_gen_xor_vec(vece, t0, t0, t2);
789
790    return t0;
791}
792
793static void gen_vrlnm_vec(unsigned vece, TCGv_vec vrt, TCGv_vec vra,
794                          TCGv_vec vrb)
795{
796    TCGv_vec mask, n = tcg_temp_new_vec_matching(vrt);
797
798    /* Create the mask */
799    mask = do_vrl_mask_vec(vece, vrb);
800
801    /* Extract n */
802    tcg_gen_dupi_vec(vece, n, (8 << vece) - 1);
803    tcg_gen_and_vec(vece, n, vrb, n);
804
805    /* Rotate and mask */
806    tcg_gen_rotlv_vec(vece, vrt, vra, n);
807    tcg_gen_and_vec(vece, vrt, vrt, mask);
808}
809
810static bool do_vrlnm(DisasContext *ctx, arg_VX *a, int vece)
811{
812    static const TCGOpcode vecop_list[] = {
813        INDEX_op_cmp_vec, INDEX_op_rotlv_vec, INDEX_op_sari_vec,
814        INDEX_op_shli_vec, INDEX_op_shri_vec, INDEX_op_shrv_vec, 0
815    };
816    static const GVecGen3 ops[2] = {
817        {
818            .fniv = gen_vrlnm_vec,
819            .fno = gen_helper_VRLWNM,
820            .opt_opc = vecop_list,
821            .load_dest = true,
822            .vece = MO_32
823        },
824        {
825            .fniv = gen_vrlnm_vec,
826            .fno = gen_helper_VRLDNM,
827            .opt_opc = vecop_list,
828            .load_dest = true,
829            .vece = MO_64
830        }
831    };
832
833    REQUIRE_INSNS_FLAGS2(ctx, ISA300);
834    REQUIRE_VSX(ctx);
835
836    tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra),
837                   avr_full_offset(a->vrb), 16, 16, &ops[vece - 2]);
838
839    return true;
840}
841
842TRANS(VRLWNM, do_vrlnm, MO_32)
843TRANS(VRLDNM, do_vrlnm, MO_64)
844
845static void gen_vrlmi_vec(unsigned vece, TCGv_vec vrt, TCGv_vec vra,
846                          TCGv_vec vrb)
847{
848    TCGv_vec mask, n = tcg_temp_new_vec_matching(vrt),
849             tmp = tcg_temp_new_vec_matching(vrt);
850
851    /* Create the mask */
852    mask = do_vrl_mask_vec(vece, vrb);
853
854    /* Extract n */
855    tcg_gen_dupi_vec(vece, n, (8 << vece) - 1);
856    tcg_gen_and_vec(vece, n, vrb, n);
857
858    /* Rotate and insert */
859    tcg_gen_rotlv_vec(vece, tmp, vra, n);
860    tcg_gen_bitsel_vec(vece, vrt, mask, tmp, vrt);
861}
862
863static bool do_vrlmi(DisasContext *ctx, arg_VX *a, int vece)
864{
865    static const TCGOpcode vecop_list[] = {
866        INDEX_op_cmp_vec, INDEX_op_rotlv_vec, INDEX_op_sari_vec,
867        INDEX_op_shli_vec, INDEX_op_shri_vec, INDEX_op_shrv_vec, 0
868    };
869    static const GVecGen3 ops[2] = {
870        {
871            .fniv = gen_vrlmi_vec,
872            .fno = gen_helper_VRLWMI,
873            .opt_opc = vecop_list,
874            .load_dest = true,
875            .vece = MO_32
876        },
877        {
878            .fniv = gen_vrlnm_vec,
879            .fno = gen_helper_VRLDMI,
880            .opt_opc = vecop_list,
881            .load_dest = true,
882            .vece = MO_64
883        }
884    };
885
886    REQUIRE_INSNS_FLAGS2(ctx, ISA300);
887    REQUIRE_VSX(ctx);
888
889    tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra),
890                   avr_full_offset(a->vrb), 16, 16, &ops[vece - 2]);
891
892    return true;
893}
894
895TRANS(VRLWMI, do_vrlmi, MO_32)
896TRANS(VRLDMI, do_vrlmi, MO_64)
897
898static bool do_vector_shift_quad(DisasContext *ctx, arg_VX *a, bool right,
899                                 bool alg)
900{
901    TCGv_i64 hi, lo, t0, t1, n, zero = tcg_constant_i64(0);
902
903    REQUIRE_VECTOR(ctx);
904
905    n = tcg_temp_new_i64();
906    hi = tcg_temp_new_i64();
907    lo = tcg_temp_new_i64();
908    t0 = tcg_temp_new_i64();
909
910    get_avr64(lo, a->vra, false);
911    get_avr64(hi, a->vra, true);
912
913    get_avr64(n, a->vrb, true);
914
915    tcg_gen_andi_i64(t0, n, 64);
916    if (right) {
917        tcg_gen_movcond_i64(TCG_COND_NE, lo, t0, zero, hi, lo);
918        if (alg) {
919            t1 = tcg_temp_new_i64();
920            tcg_gen_sari_i64(t1, lo, 63);
921        } else {
922            t1 = zero;
923        }
924        tcg_gen_movcond_i64(TCG_COND_NE, hi, t0, zero, t1, hi);
925    } else {
926        tcg_gen_movcond_i64(TCG_COND_NE, hi, t0, zero, lo, hi);
927        tcg_gen_movcond_i64(TCG_COND_NE, lo, t0, zero, zero, lo);
928    }
929    tcg_gen_andi_i64(n, n, 0x3F);
930
931    if (right) {
932        if (alg) {
933            tcg_gen_sar_i64(t0, hi, n);
934        } else {
935            tcg_gen_shr_i64(t0, hi, n);
936        }
937    } else {
938        tcg_gen_shl_i64(t0, lo, n);
939    }
940    set_avr64(a->vrt, t0, right);
941
942    if (right) {
943        tcg_gen_shr_i64(lo, lo, n);
944    } else {
945        tcg_gen_shl_i64(hi, hi, n);
946    }
947    tcg_gen_xori_i64(n, n, 63);
948    if (right) {
949        tcg_gen_shl_i64(hi, hi, n);
950        tcg_gen_shli_i64(hi, hi, 1);
951    } else {
952        tcg_gen_shr_i64(lo, lo, n);
953        tcg_gen_shri_i64(lo, lo, 1);
954    }
955    tcg_gen_or_i64(hi, hi, lo);
956    set_avr64(a->vrt, hi, !right);
957    return true;
958}
959
960TRANS_FLAGS2(ISA310, VSLQ, do_vector_shift_quad, false, false);
961TRANS_FLAGS2(ISA310, VSRQ, do_vector_shift_quad, true, false);
962TRANS_FLAGS2(ISA310, VSRAQ, do_vector_shift_quad, true, true);
963
964static void do_vrlq_mask(TCGv_i64 mh, TCGv_i64 ml, TCGv_i64 b, TCGv_i64 e)
965{
966    TCGv_i64 th, tl, t0, t1, zero = tcg_constant_i64(0),
967             ones = tcg_constant_i64(-1);
968
969    th = tcg_temp_new_i64();
970    tl = tcg_temp_new_i64();
971    t0 = tcg_temp_new_i64();
972    t1 = tcg_temp_new_i64();
973
974    /* m = ~0 >> b */
975    tcg_gen_andi_i64(t0, b, 64);
976    tcg_gen_movcond_i64(TCG_COND_NE, t1, t0, zero, zero, ones);
977    tcg_gen_andi_i64(t0, b, 0x3F);
978    tcg_gen_shr_i64(mh, t1, t0);
979    tcg_gen_shr_i64(ml, ones, t0);
980    tcg_gen_xori_i64(t0, t0, 63);
981    tcg_gen_shl_i64(t1, t1, t0);
982    tcg_gen_shli_i64(t1, t1, 1);
983    tcg_gen_or_i64(ml, t1, ml);
984
985    /* t = ~0 >> e */
986    tcg_gen_andi_i64(t0, e, 64);
987    tcg_gen_movcond_i64(TCG_COND_NE, t1, t0, zero, zero, ones);
988    tcg_gen_andi_i64(t0, e, 0x3F);
989    tcg_gen_shr_i64(th, t1, t0);
990    tcg_gen_shr_i64(tl, ones, t0);
991    tcg_gen_xori_i64(t0, t0, 63);
992    tcg_gen_shl_i64(t1, t1, t0);
993    tcg_gen_shli_i64(t1, t1, 1);
994    tcg_gen_or_i64(tl, t1, tl);
995
996    /* t = t >> 1 */
997    tcg_gen_extract2_i64(tl, tl, th, 1);
998    tcg_gen_shri_i64(th, th, 1);
999
1000    /* m = m ^ t */
1001    tcg_gen_xor_i64(mh, mh, th);
1002    tcg_gen_xor_i64(ml, ml, tl);
1003
1004    /* Negate the mask if begin > end */
1005    tcg_gen_movcond_i64(TCG_COND_GT, t0, b, e, ones, zero);
1006
1007    tcg_gen_xor_i64(mh, mh, t0);
1008    tcg_gen_xor_i64(ml, ml, t0);
1009}
1010
1011static bool do_vector_rotl_quad(DisasContext *ctx, arg_VX *a, bool mask,
1012                                bool insert)
1013{
1014    TCGv_i64 ah, al, vrb, n, t0, t1, zero = tcg_constant_i64(0);
1015
1016    REQUIRE_VECTOR(ctx);
1017    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
1018
1019    ah = tcg_temp_new_i64();
1020    al = tcg_temp_new_i64();
1021    vrb = tcg_temp_new_i64();
1022    n = tcg_temp_new_i64();
1023    t0 = tcg_temp_new_i64();
1024    t1 = tcg_temp_new_i64();
1025
1026    get_avr64(ah, a->vra, true);
1027    get_avr64(al, a->vra, false);
1028    get_avr64(vrb, a->vrb, true);
1029
1030    tcg_gen_mov_i64(t0, ah);
1031    tcg_gen_andi_i64(t1, vrb, 64);
1032    tcg_gen_movcond_i64(TCG_COND_NE, ah, t1, zero, al, ah);
1033    tcg_gen_movcond_i64(TCG_COND_NE, al, t1, zero, t0, al);
1034    tcg_gen_andi_i64(n, vrb, 0x3F);
1035
1036    tcg_gen_shl_i64(t0, ah, n);
1037    tcg_gen_shl_i64(t1, al, n);
1038
1039    tcg_gen_xori_i64(n, n, 63);
1040
1041    tcg_gen_shr_i64(al, al, n);
1042    tcg_gen_shri_i64(al, al, 1);
1043    tcg_gen_or_i64(t0, al, t0);
1044
1045    tcg_gen_shr_i64(ah, ah, n);
1046    tcg_gen_shri_i64(ah, ah, 1);
1047    tcg_gen_or_i64(t1, ah, t1);
1048
1049    if (mask || insert) {
1050        tcg_gen_extract_i64(n, vrb, 8, 7);
1051        tcg_gen_extract_i64(vrb, vrb, 16, 7);
1052
1053        do_vrlq_mask(ah, al, vrb, n);
1054
1055        tcg_gen_and_i64(t0, t0, ah);
1056        tcg_gen_and_i64(t1, t1, al);
1057
1058        if (insert) {
1059            get_avr64(n, a->vrt, true);
1060            get_avr64(vrb, a->vrt, false);
1061            tcg_gen_andc_i64(n, n, ah);
1062            tcg_gen_andc_i64(vrb, vrb, al);
1063            tcg_gen_or_i64(t0, t0, n);
1064            tcg_gen_or_i64(t1, t1, vrb);
1065        }
1066    }
1067
1068    set_avr64(a->vrt, t0, true);
1069    set_avr64(a->vrt, t1, false);
1070    return true;
1071}
1072
1073TRANS(VRLQ, do_vector_rotl_quad, false, false)
1074TRANS(VRLQNM, do_vector_rotl_quad, true, false)
1075TRANS(VRLQMI, do_vector_rotl_quad, false, true)
1076
1077#define GEN_VXFORM_SAT(NAME, VECE, NORM, SAT, OPC2, OPC3)               \
1078static void glue(glue(gen_, NAME), _vec)(unsigned vece, TCGv_vec t,     \
1079                                         TCGv_vec sat, TCGv_vec a,      \
1080                                         TCGv_vec b)                    \
1081{                                                                       \
1082    TCGv_vec x = tcg_temp_new_vec_matching(t);                          \
1083    glue(glue(tcg_gen_, NORM), _vec)(VECE, x, a, b);                    \
1084    glue(glue(tcg_gen_, SAT), _vec)(VECE, t, a, b);                     \
1085    tcg_gen_cmp_vec(TCG_COND_NE, VECE, x, x, t);                        \
1086    tcg_gen_or_vec(VECE, sat, sat, x);                                  \
1087}                                                                       \
1088static void glue(gen_, NAME)(DisasContext *ctx)                         \
1089{                                                                       \
1090    static const TCGOpcode vecop_list[] = {                             \
1091        glue(glue(INDEX_op_, NORM), _vec),                              \
1092        glue(glue(INDEX_op_, SAT), _vec),                               \
1093        INDEX_op_cmp_vec, 0                                             \
1094    };                                                                  \
1095    static const GVecGen4 g = {                                         \
1096        .fniv = glue(glue(gen_, NAME), _vec),                           \
1097        .fno = glue(gen_helper_, NAME),                                 \
1098        .opt_opc = vecop_list,                                          \
1099        .write_aofs = true,                                             \
1100        .vece = VECE,                                                   \
1101    };                                                                  \
1102    if (unlikely(!ctx->altivec_enabled)) {                              \
1103        gen_exception(ctx, POWERPC_EXCP_VPU);                           \
1104        return;                                                         \
1105    }                                                                   \
1106    tcg_gen_gvec_4(avr_full_offset(rD(ctx->opcode)),                    \
1107                   offsetof(CPUPPCState, vscr_sat),                     \
1108                   avr_full_offset(rA(ctx->opcode)),                    \
1109                   avr_full_offset(rB(ctx->opcode)),                    \
1110                   16, 16, &g);                                         \
1111}
1112
1113GEN_VXFORM_SAT(vaddubs, MO_8, add, usadd, 0, 8);
1114GEN_VXFORM_DUAL_EXT(vaddubs, PPC_ALTIVEC, PPC_NONE, 0,       \
1115                    vmul10uq, PPC_NONE, PPC2_ISA300, 0x0000F800)
1116GEN_VXFORM_SAT(vadduhs, MO_16, add, usadd, 0, 9);
1117GEN_VXFORM_DUAL(vadduhs, PPC_ALTIVEC, PPC_NONE, \
1118                vmul10euq, PPC_NONE, PPC2_ISA300)
1119GEN_VXFORM_SAT(vadduws, MO_32, add, usadd, 0, 10);
1120GEN_VXFORM_SAT(vaddsbs, MO_8, add, ssadd, 0, 12);
1121GEN_VXFORM_SAT(vaddshs, MO_16, add, ssadd, 0, 13);
1122GEN_VXFORM_SAT(vaddsws, MO_32, add, ssadd, 0, 14);
1123GEN_VXFORM_SAT(vsububs, MO_8, sub, ussub, 0, 24);
1124GEN_VXFORM_SAT(vsubuhs, MO_16, sub, ussub, 0, 25);
1125GEN_VXFORM_SAT(vsubuws, MO_32, sub, ussub, 0, 26);
1126GEN_VXFORM_SAT(vsubsbs, MO_8, sub, sssub, 0, 28);
1127GEN_VXFORM_SAT(vsubshs, MO_16, sub, sssub, 0, 29);
1128GEN_VXFORM_SAT(vsubsws, MO_32, sub, sssub, 0, 30);
1129GEN_VXFORM_TRANS(vsl, 2, 7);
1130GEN_VXFORM_TRANS(vsr, 2, 11);
1131GEN_VXFORM_ENV(vpkuhum, 7, 0);
1132GEN_VXFORM_ENV(vpkuwum, 7, 1);
1133GEN_VXFORM_ENV(vpkudum, 7, 17);
1134GEN_VXFORM_ENV(vpkuhus, 7, 2);
1135GEN_VXFORM_ENV(vpkuwus, 7, 3);
1136GEN_VXFORM_ENV(vpkudus, 7, 19);
1137GEN_VXFORM_ENV(vpkshus, 7, 4);
1138GEN_VXFORM_ENV(vpkswus, 7, 5);
1139GEN_VXFORM_ENV(vpksdus, 7, 21);
1140GEN_VXFORM_ENV(vpkshss, 7, 6);
1141GEN_VXFORM_ENV(vpkswss, 7, 7);
1142GEN_VXFORM_ENV(vpksdss, 7, 23);
1143GEN_VXFORM(vpkpx, 7, 12);
1144GEN_VXFORM_ENV(vsum4ubs, 4, 24);
1145GEN_VXFORM_ENV(vsum4sbs, 4, 28);
1146GEN_VXFORM_ENV(vsum4shs, 4, 25);
1147GEN_VXFORM_ENV(vsum2sws, 4, 26);
1148GEN_VXFORM_ENV(vsumsws, 4, 30);
1149GEN_VXFORM_ENV(vaddfp, 5, 0);
1150GEN_VXFORM_ENV(vsubfp, 5, 1);
1151GEN_VXFORM_ENV(vmaxfp, 5, 16);
1152GEN_VXFORM_ENV(vminfp, 5, 17);
1153GEN_VXFORM_HETRO(vextublx, 6, 24)
1154GEN_VXFORM_HETRO(vextuhlx, 6, 25)
1155GEN_VXFORM_HETRO(vextuwlx, 6, 26)
1156GEN_VXFORM_TRANS_DUAL(vmrgow, PPC_NONE, PPC2_ALTIVEC_207,
1157                vextuwlx, PPC_NONE, PPC2_ISA300)
1158GEN_VXFORM_HETRO(vextubrx, 6, 28)
1159GEN_VXFORM_HETRO(vextuhrx, 6, 29)
1160GEN_VXFORM_HETRO(vextuwrx, 6, 30)
1161GEN_VXFORM_TRANS(lvsl, 6, 31)
1162GEN_VXFORM_TRANS(lvsr, 6, 32)
1163GEN_VXFORM_TRANS_DUAL(vmrgew, PPC_NONE, PPC2_ALTIVEC_207,
1164                vextuwrx, PPC_NONE, PPC2_ISA300)
1165
1166#define GEN_VXRFORM1(opname, name, str, opc2, opc3)                     \
1167static void glue(gen_, name)(DisasContext *ctx)                         \
1168    {                                                                   \
1169        TCGv_ptr ra, rb, rd;                                            \
1170        if (unlikely(!ctx->altivec_enabled)) {                          \
1171            gen_exception(ctx, POWERPC_EXCP_VPU);                       \
1172            return;                                                     \
1173        }                                                               \
1174        ra = gen_avr_ptr(rA(ctx->opcode));                              \
1175        rb = gen_avr_ptr(rB(ctx->opcode));                              \
1176        rd = gen_avr_ptr(rD(ctx->opcode));                              \
1177        gen_helper_##opname(cpu_env, rd, ra, rb);                       \
1178    }
1179
1180#define GEN_VXRFORM(name, opc2, opc3)                                \
1181    GEN_VXRFORM1(name, name, #name, opc2, opc3)                      \
1182    GEN_VXRFORM1(name##_dot, name##_, #name ".", opc2, (opc3 | (0x1 << 4)))
1183
1184/*
1185 * Support for Altivec instructions that use bit 31 (Rc) as an opcode
1186 * bit but also use bit 21 as an actual Rc bit.  In general, thse pairs
1187 * come from different versions of the ISA, so we must also support a
1188 * pair of flags for each instruction.
1189 */
1190#define GEN_VXRFORM_DUAL(name0, flg0, flg2_0, name1, flg1, flg2_1)     \
1191static void glue(gen_, name0##_##name1)(DisasContext *ctx)             \
1192{                                                                      \
1193    if ((Rc(ctx->opcode) == 0) &&                                      \
1194        ((ctx->insns_flags & flg0) || (ctx->insns_flags2 & flg2_0))) { \
1195        if (Rc21(ctx->opcode) == 0) {                                  \
1196            gen_##name0(ctx);                                          \
1197        } else {                                                       \
1198            gen_##name0##_(ctx);                                       \
1199        }                                                              \
1200    } else if ((Rc(ctx->opcode) == 1) &&                               \
1201        ((ctx->insns_flags & flg1) || (ctx->insns_flags2 & flg2_1))) { \
1202        if (Rc21(ctx->opcode) == 0) {                                  \
1203            gen_##name1(ctx);                                          \
1204        } else {                                                       \
1205            gen_##name1##_(ctx);                                       \
1206        }                                                              \
1207    } else {                                                           \
1208        gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL);            \
1209    }                                                                  \
1210}
1211
1212static void do_vcmp_rc(int vrt)
1213{
1214    TCGv_i64 tmp, set, clr;
1215
1216    tmp = tcg_temp_new_i64();
1217    set = tcg_temp_new_i64();
1218    clr = tcg_temp_new_i64();
1219
1220    get_avr64(tmp, vrt, true);
1221    tcg_gen_mov_i64(set, tmp);
1222    get_avr64(tmp, vrt, false);
1223    tcg_gen_or_i64(clr, set, tmp);
1224    tcg_gen_and_i64(set, set, tmp);
1225
1226    tcg_gen_setcondi_i64(TCG_COND_EQ, clr, clr, 0);
1227    tcg_gen_shli_i64(clr, clr, 1);
1228
1229    tcg_gen_setcondi_i64(TCG_COND_EQ, set, set, -1);
1230    tcg_gen_shli_i64(set, set, 3);
1231
1232    tcg_gen_or_i64(tmp, set, clr);
1233    tcg_gen_extrl_i64_i32(cpu_crf[6], tmp);
1234}
1235
1236static bool do_vcmp(DisasContext *ctx, arg_VC *a, TCGCond cond, int vece)
1237{
1238    REQUIRE_VECTOR(ctx);
1239
1240    tcg_gen_gvec_cmp(cond, vece, avr_full_offset(a->vrt),
1241                     avr_full_offset(a->vra), avr_full_offset(a->vrb), 16, 16);
1242
1243    if (a->rc) {
1244        do_vcmp_rc(a->vrt);
1245    }
1246
1247    return true;
1248}
1249
1250TRANS_FLAGS(ALTIVEC, VCMPEQUB, do_vcmp, TCG_COND_EQ, MO_8)
1251TRANS_FLAGS(ALTIVEC, VCMPEQUH, do_vcmp, TCG_COND_EQ, MO_16)
1252TRANS_FLAGS(ALTIVEC, VCMPEQUW, do_vcmp, TCG_COND_EQ, MO_32)
1253TRANS_FLAGS2(ALTIVEC_207, VCMPEQUD, do_vcmp, TCG_COND_EQ, MO_64)
1254
1255TRANS_FLAGS(ALTIVEC, VCMPGTSB, do_vcmp, TCG_COND_GT, MO_8)
1256TRANS_FLAGS(ALTIVEC, VCMPGTSH, do_vcmp, TCG_COND_GT, MO_16)
1257TRANS_FLAGS(ALTIVEC, VCMPGTSW, do_vcmp, TCG_COND_GT, MO_32)
1258TRANS_FLAGS2(ALTIVEC_207, VCMPGTSD, do_vcmp, TCG_COND_GT, MO_64)
1259TRANS_FLAGS(ALTIVEC, VCMPGTUB, do_vcmp, TCG_COND_GTU, MO_8)
1260TRANS_FLAGS(ALTIVEC, VCMPGTUH, do_vcmp, TCG_COND_GTU, MO_16)
1261TRANS_FLAGS(ALTIVEC, VCMPGTUW, do_vcmp, TCG_COND_GTU, MO_32)
1262TRANS_FLAGS2(ALTIVEC_207, VCMPGTUD, do_vcmp, TCG_COND_GTU, MO_64)
1263
1264TRANS_FLAGS2(ISA300, VCMPNEB, do_vcmp, TCG_COND_NE, MO_8)
1265TRANS_FLAGS2(ISA300, VCMPNEH, do_vcmp, TCG_COND_NE, MO_16)
1266TRANS_FLAGS2(ISA300, VCMPNEW, do_vcmp, TCG_COND_NE, MO_32)
1267
1268static void gen_vcmpnez_vec(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b)
1269{
1270    TCGv_vec t0, t1, zero;
1271
1272    t0 = tcg_temp_new_vec_matching(t);
1273    t1 = tcg_temp_new_vec_matching(t);
1274    zero = tcg_constant_vec_matching(t, vece, 0);
1275
1276    tcg_gen_cmp_vec(TCG_COND_EQ, vece, t0, a, zero);
1277    tcg_gen_cmp_vec(TCG_COND_EQ, vece, t1, b, zero);
1278    tcg_gen_cmp_vec(TCG_COND_NE, vece, t, a, b);
1279
1280    tcg_gen_or_vec(vece, t, t, t0);
1281    tcg_gen_or_vec(vece, t, t, t1);
1282}
1283
1284static bool do_vcmpnez(DisasContext *ctx, arg_VC *a, int vece)
1285{
1286    static const TCGOpcode vecop_list[] = {
1287        INDEX_op_cmp_vec, 0
1288    };
1289    static const GVecGen3 ops[3] = {
1290        {
1291            .fniv = gen_vcmpnez_vec,
1292            .fno = gen_helper_VCMPNEZB,
1293            .opt_opc = vecop_list,
1294            .vece = MO_8
1295        },
1296        {
1297            .fniv = gen_vcmpnez_vec,
1298            .fno = gen_helper_VCMPNEZH,
1299            .opt_opc = vecop_list,
1300            .vece = MO_16
1301        },
1302        {
1303            .fniv = gen_vcmpnez_vec,
1304            .fno = gen_helper_VCMPNEZW,
1305            .opt_opc = vecop_list,
1306            .vece = MO_32
1307        }
1308    };
1309
1310    REQUIRE_INSNS_FLAGS2(ctx, ISA300);
1311    REQUIRE_VECTOR(ctx);
1312
1313    tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra),
1314                   avr_full_offset(a->vrb), 16, 16, &ops[vece]);
1315
1316    if (a->rc) {
1317        do_vcmp_rc(a->vrt);
1318    }
1319
1320    return true;
1321}
1322
1323TRANS(VCMPNEZB, do_vcmpnez, MO_8)
1324TRANS(VCMPNEZH, do_vcmpnez, MO_16)
1325TRANS(VCMPNEZW, do_vcmpnez, MO_32)
1326
1327static bool trans_VCMPEQUQ(DisasContext *ctx, arg_VC *a)
1328{
1329    TCGv_i64 t0, t1, t2;
1330
1331    t0 = tcg_temp_new_i64();
1332    t1 = tcg_temp_new_i64();
1333    t2 = tcg_temp_new_i64();
1334
1335    get_avr64(t0, a->vra, true);
1336    get_avr64(t1, a->vrb, true);
1337    tcg_gen_xor_i64(t2, t0, t1);
1338
1339    get_avr64(t0, a->vra, false);
1340    get_avr64(t1, a->vrb, false);
1341    tcg_gen_xor_i64(t1, t0, t1);
1342
1343    tcg_gen_or_i64(t1, t1, t2);
1344    tcg_gen_negsetcond_i64(TCG_COND_EQ, t1, t1, tcg_constant_i64(0));
1345
1346    set_avr64(a->vrt, t1, true);
1347    set_avr64(a->vrt, t1, false);
1348
1349    if (a->rc) {
1350        tcg_gen_extrl_i64_i32(cpu_crf[6], t1);
1351        tcg_gen_andi_i32(cpu_crf[6], cpu_crf[6], 0xa);
1352        tcg_gen_xori_i32(cpu_crf[6], cpu_crf[6], 0x2);
1353    }
1354    return true;
1355}
1356
1357static bool do_vcmpgtq(DisasContext *ctx, arg_VC *a, bool sign)
1358{
1359    TCGv_i64 t0, t1, t2;
1360
1361    t0 = tcg_temp_new_i64();
1362    t1 = tcg_temp_new_i64();
1363    t2 = tcg_temp_new_i64();
1364
1365    get_avr64(t0, a->vra, false);
1366    get_avr64(t1, a->vrb, false);
1367    tcg_gen_negsetcond_i64(TCG_COND_GTU, t2, t0, t1);
1368
1369    get_avr64(t0, a->vra, true);
1370    get_avr64(t1, a->vrb, true);
1371    tcg_gen_movcond_i64(TCG_COND_EQ, t2, t0, t1, t2, tcg_constant_i64(0));
1372    tcg_gen_negsetcond_i64(sign ? TCG_COND_GT : TCG_COND_GTU, t1, t0, t1);
1373
1374    tcg_gen_or_i64(t1, t1, t2);
1375
1376    set_avr64(a->vrt, t1, true);
1377    set_avr64(a->vrt, t1, false);
1378
1379    if (a->rc) {
1380        tcg_gen_extrl_i64_i32(cpu_crf[6], t1);
1381        tcg_gen_andi_i32(cpu_crf[6], cpu_crf[6], 0xa);
1382        tcg_gen_xori_i32(cpu_crf[6], cpu_crf[6], 0x2);
1383    }
1384    return true;
1385}
1386
1387TRANS(VCMPGTSQ, do_vcmpgtq, true)
1388TRANS(VCMPGTUQ, do_vcmpgtq, false)
1389
1390static bool do_vcmpq(DisasContext *ctx, arg_VX_bf *a, bool sign)
1391{
1392    TCGv_i64 vra, vrb;
1393    TCGLabel *gt, *lt, *done;
1394
1395    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
1396    REQUIRE_VECTOR(ctx);
1397
1398    vra = tcg_temp_new_i64();
1399    vrb = tcg_temp_new_i64();
1400    gt = gen_new_label();
1401    lt = gen_new_label();
1402    done = gen_new_label();
1403
1404    get_avr64(vra, a->vra, true);
1405    get_avr64(vrb, a->vrb, true);
1406    tcg_gen_brcond_i64((sign ? TCG_COND_GT : TCG_COND_GTU), vra, vrb, gt);
1407    tcg_gen_brcond_i64((sign ? TCG_COND_LT : TCG_COND_LTU), vra, vrb, lt);
1408
1409    get_avr64(vra, a->vra, false);
1410    get_avr64(vrb, a->vrb, false);
1411    tcg_gen_brcond_i64(TCG_COND_GTU, vra, vrb, gt);
1412    tcg_gen_brcond_i64(TCG_COND_LTU, vra, vrb, lt);
1413
1414    tcg_gen_movi_i32(cpu_crf[a->bf], CRF_EQ);
1415    tcg_gen_br(done);
1416
1417    gen_set_label(gt);
1418    tcg_gen_movi_i32(cpu_crf[a->bf], CRF_GT);
1419    tcg_gen_br(done);
1420
1421    gen_set_label(lt);
1422    tcg_gen_movi_i32(cpu_crf[a->bf], CRF_LT);
1423    tcg_gen_br(done);
1424
1425    gen_set_label(done);
1426    return true;
1427}
1428
1429TRANS(VCMPSQ, do_vcmpq, true)
1430TRANS(VCMPUQ, do_vcmpq, false)
1431
1432GEN_VXRFORM(vcmpeqfp, 3, 3)
1433GEN_VXRFORM(vcmpgefp, 3, 7)
1434GEN_VXRFORM(vcmpgtfp, 3, 11)
1435GEN_VXRFORM(vcmpbfp, 3, 15)
1436
1437static void gen_vsplti(DisasContext *ctx, int vece)
1438{
1439    int simm;
1440
1441    if (unlikely(!ctx->altivec_enabled)) {
1442        gen_exception(ctx, POWERPC_EXCP_VPU);
1443        return;
1444    }
1445
1446    simm = SIMM5(ctx->opcode);
1447    tcg_gen_gvec_dup_imm(vece, avr_full_offset(rD(ctx->opcode)), 16, 16, simm);
1448}
1449
1450#define GEN_VXFORM_VSPLTI(name, vece, opc2, opc3) \
1451static void glue(gen_, name)(DisasContext *ctx) { gen_vsplti(ctx, vece); }
1452
1453GEN_VXFORM_VSPLTI(vspltisb, MO_8, 6, 12);
1454GEN_VXFORM_VSPLTI(vspltish, MO_16, 6, 13);
1455GEN_VXFORM_VSPLTI(vspltisw, MO_32, 6, 14);
1456
1457#define GEN_VXFORM_NOA(name, opc2, opc3)                                \
1458static void glue(gen_, name)(DisasContext *ctx)                         \
1459    {                                                                   \
1460        TCGv_ptr rb, rd;                                                \
1461        if (unlikely(!ctx->altivec_enabled)) {                          \
1462            gen_exception(ctx, POWERPC_EXCP_VPU);                       \
1463            return;                                                     \
1464        }                                                               \
1465        rb = gen_avr_ptr(rB(ctx->opcode));                              \
1466        rd = gen_avr_ptr(rD(ctx->opcode));                              \
1467        gen_helper_##name(rd, rb);                                      \
1468    }
1469
1470#define GEN_VXFORM_NOA_ENV(name, opc2, opc3)                            \
1471static void glue(gen_, name)(DisasContext *ctx)                         \
1472    {                                                                   \
1473        TCGv_ptr rb, rd;                                                \
1474                                                                        \
1475        if (unlikely(!ctx->altivec_enabled)) {                          \
1476            gen_exception(ctx, POWERPC_EXCP_VPU);                       \
1477            return;                                                     \
1478        }                                                               \
1479        rb = gen_avr_ptr(rB(ctx->opcode));                              \
1480        rd = gen_avr_ptr(rD(ctx->opcode));                              \
1481        gen_helper_##name(cpu_env, rd, rb);                             \
1482    }
1483
1484#define GEN_VXFORM_NOA_2(name, opc2, opc3, opc4)                        \
1485static void glue(gen_, name)(DisasContext *ctx)                         \
1486    {                                                                   \
1487        TCGv_ptr rb, rd;                                                \
1488        if (unlikely(!ctx->altivec_enabled)) {                          \
1489            gen_exception(ctx, POWERPC_EXCP_VPU);                       \
1490            return;                                                     \
1491        }                                                               \
1492        rb = gen_avr_ptr(rB(ctx->opcode));                              \
1493        rd = gen_avr_ptr(rD(ctx->opcode));                              \
1494        gen_helper_##name(rd, rb);                                      \
1495    }
1496
1497#define GEN_VXFORM_NOA_3(name, opc2, opc3, opc4)                        \
1498static void glue(gen_, name)(DisasContext *ctx)                         \
1499    {                                                                   \
1500        TCGv_ptr rb;                                                    \
1501        if (unlikely(!ctx->altivec_enabled)) {                          \
1502            gen_exception(ctx, POWERPC_EXCP_VPU);                       \
1503            return;                                                     \
1504        }                                                               \
1505        rb = gen_avr_ptr(rB(ctx->opcode));                              \
1506        gen_helper_##name(cpu_gpr[rD(ctx->opcode)], rb);                \
1507    }
1508GEN_VXFORM_NOA(vupkhsb, 7, 8);
1509GEN_VXFORM_NOA(vupkhsh, 7, 9);
1510GEN_VXFORM_NOA(vupkhsw, 7, 25);
1511GEN_VXFORM_NOA(vupklsb, 7, 10);
1512GEN_VXFORM_NOA(vupklsh, 7, 11);
1513GEN_VXFORM_NOA(vupklsw, 7, 27);
1514GEN_VXFORM_NOA(vupkhpx, 7, 13);
1515GEN_VXFORM_NOA(vupklpx, 7, 15);
1516GEN_VXFORM_NOA_ENV(vrefp, 5, 4);
1517GEN_VXFORM_NOA_ENV(vrsqrtefp, 5, 5);
1518GEN_VXFORM_NOA_ENV(vexptefp, 5, 6);
1519GEN_VXFORM_NOA_ENV(vlogefp, 5, 7);
1520GEN_VXFORM_NOA_ENV(vrfim, 5, 11);
1521GEN_VXFORM_NOA_ENV(vrfin, 5, 8);
1522GEN_VXFORM_NOA_ENV(vrfip, 5, 10);
1523GEN_VXFORM_NOA_ENV(vrfiz, 5, 9);
1524
1525static void gen_vprtyb_vec(unsigned vece, TCGv_vec t, TCGv_vec b)
1526{
1527    int i;
1528    TCGv_vec tmp = tcg_temp_new_vec_matching(b);
1529    /* MO_32 is 2, so 2 iterations for MO_32 and 3 for MO_64 */
1530    for (i = 0; i < vece; i++) {
1531        tcg_gen_shri_vec(vece, tmp, b, (4 << (vece - i)));
1532        tcg_gen_xor_vec(vece, b, tmp, b);
1533    }
1534    tcg_gen_and_vec(vece, t, b, tcg_constant_vec_matching(t, vece, 1));
1535}
1536
1537/* vprtybw */
1538static void gen_vprtyb_i32(TCGv_i32 t, TCGv_i32 b)
1539{
1540    tcg_gen_ctpop_i32(t, b);
1541    tcg_gen_and_i32(t, t, tcg_constant_i32(1));
1542}
1543
1544/* vprtybd */
1545static void gen_vprtyb_i64(TCGv_i64 t, TCGv_i64 b)
1546{
1547    tcg_gen_ctpop_i64(t, b);
1548    tcg_gen_and_i64(t, t, tcg_constant_i64(1));
1549}
1550
1551static bool do_vx_vprtyb(DisasContext *ctx, arg_VX_tb *a, unsigned vece)
1552{
1553    static const TCGOpcode vecop_list[] = {
1554        INDEX_op_shri_vec, 0
1555    };
1556
1557    static const GVecGen2 op[] = {
1558        {
1559            .fniv = gen_vprtyb_vec,
1560            .fni4 = gen_vprtyb_i32,
1561            .opt_opc = vecop_list,
1562            .vece = MO_32
1563        },
1564        {
1565            .fniv = gen_vprtyb_vec,
1566            .fni8 = gen_vprtyb_i64,
1567            .opt_opc = vecop_list,
1568            .vece = MO_64
1569        },
1570        {
1571            .fno = gen_helper_VPRTYBQ,
1572            .vece = MO_128
1573        },
1574    };
1575
1576    REQUIRE_INSNS_FLAGS2(ctx, ISA300);
1577    REQUIRE_VECTOR(ctx);
1578
1579    tcg_gen_gvec_2(avr_full_offset(a->vrt), avr_full_offset(a->vrb),
1580                   16, 16, &op[vece - MO_32]);
1581
1582    return true;
1583}
1584
1585TRANS(VPRTYBW, do_vx_vprtyb, MO_32)
1586TRANS(VPRTYBD, do_vx_vprtyb, MO_64)
1587TRANS(VPRTYBQ, do_vx_vprtyb, MO_128)
1588
1589static void gen_vsplt(DisasContext *ctx, int vece)
1590{
1591    int uimm, dofs, bofs;
1592
1593    if (unlikely(!ctx->altivec_enabled)) {
1594        gen_exception(ctx, POWERPC_EXCP_VPU);
1595        return;
1596    }
1597
1598    uimm = UIMM5(ctx->opcode);
1599    bofs = avr_full_offset(rB(ctx->opcode));
1600    dofs = avr_full_offset(rD(ctx->opcode));
1601
1602    /* Experimental testing shows that hardware masks the immediate.  */
1603    bofs += (uimm << vece) & 15;
1604#if !HOST_BIG_ENDIAN
1605    bofs ^= 15;
1606    bofs &= ~((1 << vece) - 1);
1607#endif
1608
1609    tcg_gen_gvec_dup_mem(vece, dofs, bofs, 16, 16);
1610}
1611
1612#define GEN_VXFORM_VSPLT(name, vece, opc2, opc3) \
1613static void glue(gen_, name)(DisasContext *ctx) { gen_vsplt(ctx, vece); }
1614
1615#define GEN_VXFORM_UIMM_ENV(name, opc2, opc3)                           \
1616static void glue(gen_, name)(DisasContext *ctx)                         \
1617    {                                                                   \
1618        TCGv_ptr rb, rd;                                                \
1619        TCGv_i32 uimm;                                                  \
1620                                                                        \
1621        if (unlikely(!ctx->altivec_enabled)) {                          \
1622            gen_exception(ctx, POWERPC_EXCP_VPU);                       \
1623            return;                                                     \
1624        }                                                               \
1625        uimm = tcg_constant_i32(UIMM5(ctx->opcode));                    \
1626        rb = gen_avr_ptr(rB(ctx->opcode));                              \
1627        rd = gen_avr_ptr(rD(ctx->opcode));                              \
1628        gen_helper_##name(cpu_env, rd, rb, uimm);                       \
1629    }
1630
1631#define GEN_VXFORM_UIMM_SPLAT(name, opc2, opc3, splat_max)              \
1632static void glue(gen_, name)(DisasContext *ctx)                         \
1633    {                                                                   \
1634        TCGv_ptr rb, rd;                                                \
1635        uint8_t uimm = UIMM4(ctx->opcode);                              \
1636        TCGv_i32 t0;                                                    \
1637        if (unlikely(!ctx->altivec_enabled)) {                          \
1638            gen_exception(ctx, POWERPC_EXCP_VPU);                       \
1639            return;                                                     \
1640        }                                                               \
1641        if (uimm > splat_max) {                                         \
1642            uimm = 0;                                                   \
1643        }                                                               \
1644        t0 = tcg_temp_new_i32();                                        \
1645        tcg_gen_movi_i32(t0, uimm);                                     \
1646        rb = gen_avr_ptr(rB(ctx->opcode));                              \
1647        rd = gen_avr_ptr(rD(ctx->opcode));                              \
1648        gen_helper_##name(rd, rb, t0);                                  \
1649    }
1650
1651GEN_VXFORM_VSPLT(vspltb, MO_8, 6, 8);
1652GEN_VXFORM_VSPLT(vsplth, MO_16, 6, 9);
1653GEN_VXFORM_VSPLT(vspltw, MO_32, 6, 10);
1654GEN_VXFORM_UIMM_SPLAT(vextractub, 6, 8, 15);
1655GEN_VXFORM_UIMM_SPLAT(vextractuh, 6, 9, 14);
1656GEN_VXFORM_UIMM_SPLAT(vextractuw, 6, 10, 12);
1657GEN_VXFORM_UIMM_SPLAT(vextractd, 6, 11, 8);
1658GEN_VXFORM_UIMM_ENV(vcfux, 5, 12);
1659GEN_VXFORM_UIMM_ENV(vcfsx, 5, 13);
1660GEN_VXFORM_UIMM_ENV(vctuxs, 5, 14);
1661GEN_VXFORM_UIMM_ENV(vctsxs, 5, 15);
1662GEN_VXFORM_DUAL(vspltb, PPC_ALTIVEC, PPC_NONE,
1663                vextractub, PPC_NONE, PPC2_ISA300);
1664GEN_VXFORM_DUAL(vsplth, PPC_ALTIVEC, PPC_NONE,
1665                vextractuh, PPC_NONE, PPC2_ISA300);
1666GEN_VXFORM_DUAL(vspltw, PPC_ALTIVEC, PPC_NONE,
1667                vextractuw, PPC_NONE, PPC2_ISA300);
1668
1669static bool trans_VGNB(DisasContext *ctx, arg_VX_n *a)
1670{
1671    /*
1672     * Similar to do_vextractm, we'll use a sequence of mask-shift-or operations
1673     * to gather the bits. The masks can be created with
1674     *
1675     * uint64_t mask(uint64_t n, uint64_t step)
1676     * {
1677     *     uint64_t p = ((1UL << (1UL << step)) - 1UL) << ((n - 1UL) << step),
1678     *                  plen = n << step, m = 0;
1679     *     for(int i = 0; i < 64/plen; i++) {
1680     *         m |= p;
1681     *         m = ror64(m, plen);
1682     *     }
1683     *     p >>= plen * DIV_ROUND_UP(64, plen) - 64;
1684     *     return m | p;
1685     * }
1686     *
1687     * But since there are few values of N, we'll use a lookup table to avoid
1688     * these calculations at runtime.
1689     */
1690    static const uint64_t mask[6][5] = {
1691        {
1692            0xAAAAAAAAAAAAAAAAULL, 0xccccccccccccccccULL, 0xf0f0f0f0f0f0f0f0ULL,
1693            0xff00ff00ff00ff00ULL, 0xffff0000ffff0000ULL
1694        },
1695        {
1696            0x9249249249249249ULL, 0xC30C30C30C30C30CULL, 0xF00F00F00F00F00FULL,
1697            0xFF0000FF0000FF00ULL, 0xFFFF00000000FFFFULL
1698        },
1699        {
1700            /* For N >= 4, some mask operations can be elided */
1701            0x8888888888888888ULL, 0, 0xf000f000f000f000ULL, 0,
1702            0xFFFF000000000000ULL
1703        },
1704        {
1705            0x8421084210842108ULL, 0, 0xF0000F0000F0000FULL, 0, 0
1706        },
1707        {
1708            0x8208208208208208ULL, 0, 0xF00000F00000F000ULL, 0, 0
1709        },
1710        {
1711            0x8102040810204081ULL, 0, 0xF000000F000000F0ULL, 0, 0
1712        }
1713    };
1714    uint64_t m;
1715    int i, sh, nbits = DIV_ROUND_UP(64, a->n);
1716    TCGv_i64 hi, lo, t0, t1;
1717
1718    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
1719    REQUIRE_VECTOR(ctx);
1720
1721    if (a->n < 2) {
1722        /*
1723         * "N can be any value between 2 and 7, inclusive." Otherwise, the
1724         * result is undefined, so we don't need to change RT. Also, N > 7 is
1725         * impossible since the immediate field is 3 bits only.
1726         */
1727        return true;
1728    }
1729
1730    hi = tcg_temp_new_i64();
1731    lo = tcg_temp_new_i64();
1732    t0 = tcg_temp_new_i64();
1733    t1 = tcg_temp_new_i64();
1734
1735    get_avr64(hi, a->vrb, true);
1736    get_avr64(lo, a->vrb, false);
1737
1738    /* Align the lower doubleword so we can use the same mask */
1739    tcg_gen_shli_i64(lo, lo, a->n * nbits - 64);
1740
1741    /*
1742     * Starting from the most significant bit, gather every Nth bit with a
1743     * sequence of mask-shift-or operation. E.g.: for N=3
1744     * AxxBxxCxxDxxExxFxxGxxHxxIxxJxxKxxLxxMxxNxxOxxPxxQxxRxxSxxTxxUxxV
1745     *     & rep(0b100)
1746     * A..B..C..D..E..F..G..H..I..J..K..L..M..N..O..P..Q..R..S..T..U..V
1747     *     << 2
1748     * .B..C..D..E..F..G..H..I..J..K..L..M..N..O..P..Q..R..S..T..U..V..
1749     *     |
1750     * AB.BC.CD.DE.EF.FG.GH.HI.IJ.JK.KL.LM.MN.NO.OP.PQ.QR.RS.ST.TU.UV.V
1751     *  & rep(0b110000)
1752     * AB....CD....EF....GH....IJ....KL....MN....OP....QR....ST....UV..
1753     *     << 4
1754     * ..CD....EF....GH....IJ....KL....MN....OP....QR....ST....UV......
1755     *     |
1756     * ABCD..CDEF..EFGH..GHIJ..IJKL..KLMN..MNOP..OPQR..QRST..STUV..UV..
1757     *     & rep(0b111100000000)
1758     * ABCD........EFGH........IJKL........MNOP........QRST........UV..
1759     *     << 8
1760     * ....EFGH........IJKL........MNOP........QRST........UV..........
1761     *     |
1762     * ABCDEFGH....EFGHIJKL....IJKLMNOP....MNOPQRST....QRSTUV......UV..
1763     *  & rep(0b111111110000000000000000)
1764     * ABCDEFGH................IJKLMNOP................QRSTUV..........
1765     *     << 16
1766     * ........IJKLMNOP................QRSTUV..........................
1767     *     |
1768     * ABCDEFGHIJKLMNOP........IJKLMNOPQRSTUV..........QRSTUV..........
1769     *     & rep(0b111111111111111100000000000000000000000000000000)
1770     * ABCDEFGHIJKLMNOP................................QRSTUV..........
1771     *     << 32
1772     * ................QRSTUV..........................................
1773     *     |
1774     * ABCDEFGHIJKLMNOPQRSTUV..........................QRSTUV..........
1775     */
1776    for (i = 0, sh = a->n - 1; i < 5; i++, sh <<= 1) {
1777        m = mask[a->n - 2][i];
1778        if (m) {
1779            tcg_gen_andi_i64(hi, hi, m);
1780            tcg_gen_andi_i64(lo, lo, m);
1781        }
1782        if (sh < 64) {
1783            tcg_gen_shli_i64(t0, hi, sh);
1784            tcg_gen_shli_i64(t1, lo, sh);
1785            tcg_gen_or_i64(hi, t0, hi);
1786            tcg_gen_or_i64(lo, t1, lo);
1787        }
1788    }
1789
1790    tcg_gen_andi_i64(hi, hi, ~(~0ULL >> nbits));
1791    tcg_gen_andi_i64(lo, lo, ~(~0ULL >> nbits));
1792    tcg_gen_shri_i64(lo, lo, nbits);
1793    tcg_gen_or_i64(hi, hi, lo);
1794    tcg_gen_trunc_i64_tl(cpu_gpr[a->rt], hi);
1795    return true;
1796}
1797
1798static bool do_vextdx(DisasContext *ctx, arg_VA *a, int size, bool right,
1799               void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv))
1800{
1801    TCGv_ptr vrt, vra, vrb;
1802    TCGv rc;
1803
1804    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
1805    REQUIRE_VECTOR(ctx);
1806
1807    vrt = gen_avr_ptr(a->vrt);
1808    vra = gen_avr_ptr(a->vra);
1809    vrb = gen_avr_ptr(a->vrb);
1810    rc = tcg_temp_new();
1811
1812    tcg_gen_andi_tl(rc, cpu_gpr[a->rc], 0x1F);
1813    if (right) {
1814        tcg_gen_subfi_tl(rc, 32 - size, rc);
1815    }
1816    gen_helper(cpu_env, vrt, vra, vrb, rc);
1817    return true;
1818}
1819
1820TRANS(VEXTDUBVLX, do_vextdx, 1, false, gen_helper_VEXTDUBVLX)
1821TRANS(VEXTDUHVLX, do_vextdx, 2, false, gen_helper_VEXTDUHVLX)
1822TRANS(VEXTDUWVLX, do_vextdx, 4, false, gen_helper_VEXTDUWVLX)
1823TRANS(VEXTDDVLX, do_vextdx, 8, false, gen_helper_VEXTDDVLX)
1824
1825TRANS(VEXTDUBVRX, do_vextdx, 1, true, gen_helper_VEXTDUBVLX)
1826TRANS(VEXTDUHVRX, do_vextdx, 2, true, gen_helper_VEXTDUHVLX)
1827TRANS(VEXTDUWVRX, do_vextdx, 4, true, gen_helper_VEXTDUWVLX)
1828TRANS(VEXTDDVRX, do_vextdx, 8, true, gen_helper_VEXTDDVLX)
1829
1830static bool do_vinsx(DisasContext *ctx, int vrt, int size, bool right, TCGv ra,
1831            TCGv_i64 rb, void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv))
1832{
1833    TCGv_ptr t;
1834    TCGv idx;
1835
1836    t = gen_avr_ptr(vrt);
1837    idx = tcg_temp_new();
1838
1839    tcg_gen_andi_tl(idx, ra, 0xF);
1840    if (right) {
1841        tcg_gen_subfi_tl(idx, 16 - size, idx);
1842    }
1843
1844    gen_helper(cpu_env, t, rb, idx);
1845    return true;
1846}
1847
1848static bool do_vinsvx(DisasContext *ctx, int vrt, int size, bool right, TCGv ra,
1849                int vrb, void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv))
1850{
1851    TCGv_i64 val;
1852
1853    val = tcg_temp_new_i64();
1854    get_avr64(val, vrb, true);
1855    return do_vinsx(ctx, vrt, size, right, ra, val, gen_helper);
1856}
1857
1858static bool do_vinsx_VX(DisasContext *ctx, arg_VX *a, int size, bool right,
1859                        void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv))
1860{
1861    TCGv_i64 val;
1862
1863    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
1864    REQUIRE_VECTOR(ctx);
1865
1866    val = tcg_temp_new_i64();
1867    tcg_gen_extu_tl_i64(val, cpu_gpr[a->vrb]);
1868
1869    return do_vinsx(ctx, a->vrt, size, right, cpu_gpr[a->vra], val, gen_helper);
1870}
1871
1872static bool do_vinsvx_VX(DisasContext *ctx, arg_VX *a, int size, bool right,
1873                        void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv))
1874{
1875    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
1876    REQUIRE_VECTOR(ctx);
1877
1878    return do_vinsvx(ctx, a->vrt, size, right, cpu_gpr[a->vra], a->vrb,
1879                     gen_helper);
1880}
1881
1882static bool do_vins_VX_uim4(DisasContext *ctx, arg_VX_uim4 *a, int size,
1883                        void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv))
1884{
1885    TCGv_i64 val;
1886
1887    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
1888    REQUIRE_VECTOR(ctx);
1889
1890    if (a->uim > (16 - size)) {
1891        /*
1892         * PowerISA v3.1 says that the resulting value is undefined in this
1893         * case, so just log a guest error and leave VRT unchanged. The
1894         * real hardware would do a partial insert, e.g. if VRT is zeroed and
1895         * RB is 0x12345678, executing "vinsw VRT,RB,14" results in
1896         * VRT = 0x0000...00001234, but we don't bother to reproduce this
1897         * behavior as software shouldn't rely on it.
1898         */
1899        qemu_log_mask(LOG_GUEST_ERROR, "Invalid index for VINS* at"
1900            " 0x" TARGET_FMT_lx ", UIM = %d > %d\n", ctx->cia, a->uim,
1901            16 - size);
1902        return true;
1903    }
1904
1905    val = tcg_temp_new_i64();
1906    tcg_gen_extu_tl_i64(val, cpu_gpr[a->vrb]);
1907
1908    return do_vinsx(ctx, a->vrt, size, false, tcg_constant_tl(a->uim), val,
1909                    gen_helper);
1910}
1911
1912static bool do_vinsert_VX_uim4(DisasContext *ctx, arg_VX_uim4 *a, int size,
1913                        void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv))
1914{
1915    REQUIRE_INSNS_FLAGS2(ctx, ISA300);
1916    REQUIRE_VECTOR(ctx);
1917
1918    if (a->uim > (16 - size)) {
1919        qemu_log_mask(LOG_GUEST_ERROR, "Invalid index for VINSERT* at"
1920            " 0x" TARGET_FMT_lx ", UIM = %d > %d\n", ctx->cia, a->uim,
1921            16 - size);
1922        return true;
1923    }
1924
1925    return do_vinsvx(ctx, a->vrt, size, false, tcg_constant_tl(a->uim), a->vrb,
1926                     gen_helper);
1927}
1928
1929TRANS(VINSBLX, do_vinsx_VX, 1, false, gen_helper_VINSBLX)
1930TRANS(VINSHLX, do_vinsx_VX, 2, false, gen_helper_VINSHLX)
1931TRANS(VINSWLX, do_vinsx_VX, 4, false, gen_helper_VINSWLX)
1932TRANS(VINSDLX, do_vinsx_VX, 8, false, gen_helper_VINSDLX)
1933
1934TRANS(VINSBRX, do_vinsx_VX, 1, true, gen_helper_VINSBLX)
1935TRANS(VINSHRX, do_vinsx_VX, 2, true, gen_helper_VINSHLX)
1936TRANS(VINSWRX, do_vinsx_VX, 4, true, gen_helper_VINSWLX)
1937TRANS(VINSDRX, do_vinsx_VX, 8, true, gen_helper_VINSDLX)
1938
1939TRANS(VINSW, do_vins_VX_uim4, 4, gen_helper_VINSWLX)
1940TRANS(VINSD, do_vins_VX_uim4, 8, gen_helper_VINSDLX)
1941
1942TRANS(VINSBVLX, do_vinsvx_VX, 1, false, gen_helper_VINSBLX)
1943TRANS(VINSHVLX, do_vinsvx_VX, 2, false, gen_helper_VINSHLX)
1944TRANS(VINSWVLX, do_vinsvx_VX, 4, false, gen_helper_VINSWLX)
1945
1946TRANS(VINSBVRX, do_vinsvx_VX, 1, true, gen_helper_VINSBLX)
1947TRANS(VINSHVRX, do_vinsvx_VX, 2, true, gen_helper_VINSHLX)
1948TRANS(VINSWVRX, do_vinsvx_VX, 4, true, gen_helper_VINSWLX)
1949
1950TRANS(VINSERTB, do_vinsert_VX_uim4, 1, gen_helper_VINSBLX)
1951TRANS(VINSERTH, do_vinsert_VX_uim4, 2, gen_helper_VINSHLX)
1952TRANS(VINSERTW, do_vinsert_VX_uim4, 4, gen_helper_VINSWLX)
1953TRANS(VINSERTD, do_vinsert_VX_uim4, 8, gen_helper_VINSDLX)
1954
1955static void gen_vsldoi(DisasContext *ctx)
1956{
1957    TCGv_ptr ra, rb, rd;
1958    TCGv_i32 sh;
1959    if (unlikely(!ctx->altivec_enabled)) {
1960        gen_exception(ctx, POWERPC_EXCP_VPU);
1961        return;
1962    }
1963    ra = gen_avr_ptr(rA(ctx->opcode));
1964    rb = gen_avr_ptr(rB(ctx->opcode));
1965    rd = gen_avr_ptr(rD(ctx->opcode));
1966    sh = tcg_constant_i32(VSH(ctx->opcode));
1967    gen_helper_vsldoi(rd, ra, rb, sh);
1968}
1969
1970static bool trans_VSLDBI(DisasContext *ctx, arg_VN *a)
1971{
1972    TCGv_i64 t0, t1, t2;
1973
1974    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
1975    REQUIRE_VECTOR(ctx);
1976
1977    t0 = tcg_temp_new_i64();
1978    t1 = tcg_temp_new_i64();
1979
1980    get_avr64(t0, a->vra, true);
1981    get_avr64(t1, a->vra, false);
1982
1983    if (a->sh != 0) {
1984        t2 = tcg_temp_new_i64();
1985
1986        get_avr64(t2, a->vrb, true);
1987
1988        tcg_gen_extract2_i64(t0, t1, t0, 64 - a->sh);
1989        tcg_gen_extract2_i64(t1, t2, t1, 64 - a->sh);
1990    }
1991
1992    set_avr64(a->vrt, t0, true);
1993    set_avr64(a->vrt, t1, false);
1994    return true;
1995}
1996
1997static bool trans_VSRDBI(DisasContext *ctx, arg_VN *a)
1998{
1999    TCGv_i64 t2, t1, t0;
2000
2001    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2002    REQUIRE_VECTOR(ctx);
2003
2004    t0 = tcg_temp_new_i64();
2005    t1 = tcg_temp_new_i64();
2006
2007    get_avr64(t0, a->vrb, false);
2008    get_avr64(t1, a->vrb, true);
2009
2010    if (a->sh != 0) {
2011        t2 = tcg_temp_new_i64();
2012
2013        get_avr64(t2, a->vra, false);
2014
2015        tcg_gen_extract2_i64(t0, t0, t1, a->sh);
2016        tcg_gen_extract2_i64(t1, t1, t2, a->sh);
2017    }
2018
2019    set_avr64(a->vrt, t0, false);
2020    set_avr64(a->vrt, t1, true);
2021    return true;
2022}
2023
2024static bool do_vexpand(DisasContext *ctx, arg_VX_tb *a, unsigned vece)
2025{
2026    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2027    REQUIRE_VECTOR(ctx);
2028
2029    tcg_gen_gvec_sari(vece, avr_full_offset(a->vrt), avr_full_offset(a->vrb),
2030                      (8 << vece) - 1, 16, 16);
2031
2032    return true;
2033}
2034
2035TRANS(VEXPANDBM, do_vexpand, MO_8)
2036TRANS(VEXPANDHM, do_vexpand, MO_16)
2037TRANS(VEXPANDWM, do_vexpand, MO_32)
2038TRANS(VEXPANDDM, do_vexpand, MO_64)
2039
2040static bool trans_VEXPANDQM(DisasContext *ctx, arg_VX_tb *a)
2041{
2042    TCGv_i64 tmp;
2043
2044    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2045    REQUIRE_VECTOR(ctx);
2046
2047    tmp = tcg_temp_new_i64();
2048
2049    get_avr64(tmp, a->vrb, true);
2050    tcg_gen_sari_i64(tmp, tmp, 63);
2051    set_avr64(a->vrt, tmp, false);
2052    set_avr64(a->vrt, tmp, true);
2053    return true;
2054}
2055
2056static bool do_vextractm(DisasContext *ctx, arg_VX_tb *a, unsigned vece)
2057{
2058    const uint64_t elem_width = 8 << vece, elem_count_half = 8 >> vece,
2059                   mask = dup_const(vece, 1ULL << (elem_width - 1));
2060    uint64_t i, j;
2061    TCGv_i64 lo, hi, t0, t1;
2062
2063    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2064    REQUIRE_VECTOR(ctx);
2065
2066    hi = tcg_temp_new_i64();
2067    lo = tcg_temp_new_i64();
2068    t0 = tcg_temp_new_i64();
2069    t1 = tcg_temp_new_i64();
2070
2071    get_avr64(lo, a->vrb, false);
2072    get_avr64(hi, a->vrb, true);
2073
2074    tcg_gen_andi_i64(lo, lo, mask);
2075    tcg_gen_andi_i64(hi, hi, mask);
2076
2077    /*
2078     * Gather the most significant bit of each element in the highest element
2079     * element. E.g. for bytes:
2080     * aXXXXXXXbXXXXXXXcXXXXXXXdXXXXXXXeXXXXXXXfXXXXXXXgXXXXXXXhXXXXXXX
2081     *     & dup(1 << (elem_width - 1))
2082     * a0000000b0000000c0000000d0000000e0000000f0000000g0000000h0000000
2083     *     << 32 - 4
2084     * 0000e0000000f0000000g0000000h00000000000000000000000000000000000
2085     *     |
2086     * a000e000b000f000c000g000d000h000e0000000f0000000g0000000h0000000
2087     *     << 16 - 2
2088     * 00c000g000d000h000e0000000f0000000g0000000h000000000000000000000
2089     *     |
2090     * a0c0e0g0b0d0f0h0c0e0g000d0f0h000e0g00000f0h00000g0000000h0000000
2091     *     << 8 - 1
2092     * 0b0d0f0h0c0e0g000d0f0h000e0g00000f0h00000g0000000h00000000000000
2093     *     |
2094     * abcdefghbcdefgh0cdefgh00defgh000efgh0000fgh00000gh000000h0000000
2095     */
2096    for (i = elem_count_half / 2, j = 32; i > 0; i >>= 1, j >>= 1) {
2097        tcg_gen_shli_i64(t0, hi, j - i);
2098        tcg_gen_shli_i64(t1, lo, j - i);
2099        tcg_gen_or_i64(hi, hi, t0);
2100        tcg_gen_or_i64(lo, lo, t1);
2101    }
2102
2103    tcg_gen_shri_i64(hi, hi, 64 - elem_count_half);
2104    tcg_gen_extract2_i64(lo, lo, hi, 64 - elem_count_half);
2105    tcg_gen_trunc_i64_tl(cpu_gpr[a->vrt], lo);
2106    return true;
2107}
2108
2109TRANS(VEXTRACTBM, do_vextractm, MO_8)
2110TRANS(VEXTRACTHM, do_vextractm, MO_16)
2111TRANS(VEXTRACTWM, do_vextractm, MO_32)
2112TRANS(VEXTRACTDM, do_vextractm, MO_64)
2113
2114static bool trans_VEXTRACTQM(DisasContext *ctx, arg_VX_tb *a)
2115{
2116    TCGv_i64 tmp;
2117
2118    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2119    REQUIRE_VECTOR(ctx);
2120
2121    tmp = tcg_temp_new_i64();
2122
2123    get_avr64(tmp, a->vrb, true);
2124    tcg_gen_shri_i64(tmp, tmp, 63);
2125    tcg_gen_trunc_i64_tl(cpu_gpr[a->vrt], tmp);
2126    return true;
2127}
2128
2129static bool do_mtvsrm(DisasContext *ctx, arg_VX_tb *a, unsigned vece)
2130{
2131    const uint64_t elem_width = 8 << vece, elem_count_half = 8 >> vece;
2132    uint64_t c;
2133    int i, j;
2134    TCGv_i64 hi, lo, t0, t1;
2135
2136    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2137    REQUIRE_VECTOR(ctx);
2138
2139    hi = tcg_temp_new_i64();
2140    lo = tcg_temp_new_i64();
2141    t0 = tcg_temp_new_i64();
2142    t1 = tcg_temp_new_i64();
2143
2144    tcg_gen_extu_tl_i64(t0, cpu_gpr[a->vrb]);
2145    tcg_gen_extract_i64(hi, t0, elem_count_half, elem_count_half);
2146    tcg_gen_extract_i64(lo, t0, 0, elem_count_half);
2147
2148    /*
2149     * Spread the bits into their respective elements.
2150     * E.g. for bytes:
2151     * 00000000000000000000000000000000000000000000000000000000abcdefgh
2152     *   << 32 - 4
2153     * 0000000000000000000000000000abcdefgh0000000000000000000000000000
2154     *   |
2155     * 0000000000000000000000000000abcdefgh00000000000000000000abcdefgh
2156     *   << 16 - 2
2157     * 00000000000000abcdefgh00000000000000000000abcdefgh00000000000000
2158     *   |
2159     * 00000000000000abcdefgh000000abcdefgh000000abcdefgh000000abcdefgh
2160     *   << 8 - 1
2161     * 0000000abcdefgh000000abcdefgh000000abcdefgh000000abcdefgh0000000
2162     *   |
2163     * 0000000abcdefgXbcdefgXbcdefgXbcdefgXbcdefgXbcdefgXbcdefgXbcdefgh
2164     *   & dup(1)
2165     * 0000000a0000000b0000000c0000000d0000000e0000000f0000000g0000000h
2166     *   * 0xff
2167     * aaaaaaaabbbbbbbbccccccccddddddddeeeeeeeeffffffffgggggggghhhhhhhh
2168     */
2169    for (i = elem_count_half / 2, j = 32; i > 0; i >>= 1, j >>= 1) {
2170        tcg_gen_shli_i64(t0, hi, j - i);
2171        tcg_gen_shli_i64(t1, lo, j - i);
2172        tcg_gen_or_i64(hi, hi, t0);
2173        tcg_gen_or_i64(lo, lo, t1);
2174    }
2175
2176    c = dup_const(vece, 1);
2177    tcg_gen_andi_i64(hi, hi, c);
2178    tcg_gen_andi_i64(lo, lo, c);
2179
2180    c = MAKE_64BIT_MASK(0, elem_width);
2181    tcg_gen_muli_i64(hi, hi, c);
2182    tcg_gen_muli_i64(lo, lo, c);
2183
2184    set_avr64(a->vrt, lo, false);
2185    set_avr64(a->vrt, hi, true);
2186    return true;
2187}
2188
2189TRANS(MTVSRBM, do_mtvsrm, MO_8)
2190TRANS(MTVSRHM, do_mtvsrm, MO_16)
2191TRANS(MTVSRWM, do_mtvsrm, MO_32)
2192TRANS(MTVSRDM, do_mtvsrm, MO_64)
2193
2194static bool trans_MTVSRQM(DisasContext *ctx, arg_VX_tb *a)
2195{
2196    TCGv_i64 tmp;
2197
2198    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2199    REQUIRE_VECTOR(ctx);
2200
2201    tmp = tcg_temp_new_i64();
2202
2203    tcg_gen_ext_tl_i64(tmp, cpu_gpr[a->vrb]);
2204    tcg_gen_sextract_i64(tmp, tmp, 0, 1);
2205    set_avr64(a->vrt, tmp, false);
2206    set_avr64(a->vrt, tmp, true);
2207    return true;
2208}
2209
2210static bool trans_MTVSRBMI(DisasContext *ctx, arg_DX_b *a)
2211{
2212    const uint64_t mask = dup_const(MO_8, 1);
2213    uint64_t hi, lo;
2214
2215    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2216    REQUIRE_VECTOR(ctx);
2217
2218    hi = extract16(a->b, 8, 8);
2219    lo = extract16(a->b, 0, 8);
2220
2221    for (int i = 4, j = 32; i > 0; i >>= 1, j >>= 1) {
2222        hi |= hi << (j - i);
2223        lo |= lo << (j - i);
2224    }
2225
2226    hi = (hi & mask) * 0xFF;
2227    lo = (lo & mask) * 0xFF;
2228
2229    set_avr64(a->vrt, tcg_constant_i64(hi), true);
2230    set_avr64(a->vrt, tcg_constant_i64(lo), false);
2231
2232    return true;
2233}
2234
2235static bool do_vcntmb(DisasContext *ctx, arg_VX_mp *a, int vece)
2236{
2237    TCGv_i64 r[2], mask;
2238
2239    r[0] = tcg_temp_new_i64();
2240    r[1] = tcg_temp_new_i64();
2241    mask = tcg_constant_i64(dup_const(vece, 1ULL << ((8 << vece) - 1)));
2242
2243    for (int i = 0; i < 2; i++) {
2244        get_avr64(r[i], a->vrb, i);
2245        if (a->mp) {
2246            tcg_gen_and_i64(r[i], mask, r[i]);
2247        } else {
2248            tcg_gen_andc_i64(r[i], mask, r[i]);
2249        }
2250        tcg_gen_ctpop_i64(r[i], r[i]);
2251    }
2252
2253    tcg_gen_add_i64(r[0], r[0], r[1]);
2254    tcg_gen_shli_i64(r[0], r[0], TARGET_LONG_BITS - 8 + vece);
2255    tcg_gen_trunc_i64_tl(cpu_gpr[a->rt], r[0]);
2256    return true;
2257}
2258
2259TRANS(VCNTMBB, do_vcntmb, MO_8)
2260TRANS(VCNTMBH, do_vcntmb, MO_16)
2261TRANS(VCNTMBW, do_vcntmb, MO_32)
2262TRANS(VCNTMBD, do_vcntmb, MO_64)
2263
2264static bool do_vstri(DisasContext *ctx, arg_VX_tb_rc *a,
2265                     void (*gen_helper)(TCGv_i32, TCGv_ptr, TCGv_ptr))
2266{
2267    TCGv_ptr vrt, vrb;
2268
2269    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2270    REQUIRE_VECTOR(ctx);
2271
2272    vrt = gen_avr_ptr(a->vrt);
2273    vrb = gen_avr_ptr(a->vrb);
2274
2275    if (a->rc) {
2276        gen_helper(cpu_crf[6], vrt, vrb);
2277    } else {
2278        TCGv_i32 discard = tcg_temp_new_i32();
2279        gen_helper(discard, vrt, vrb);
2280    }
2281    return true;
2282}
2283
2284TRANS(VSTRIBL, do_vstri, gen_helper_VSTRIBL)
2285TRANS(VSTRIBR, do_vstri, gen_helper_VSTRIBR)
2286TRANS(VSTRIHL, do_vstri, gen_helper_VSTRIHL)
2287TRANS(VSTRIHR, do_vstri, gen_helper_VSTRIHR)
2288
2289static bool do_vclrb(DisasContext *ctx, arg_VX *a, bool right)
2290{
2291    TCGv_i64 rb, mh, ml, tmp,
2292             ones = tcg_constant_i64(-1),
2293             zero = tcg_constant_i64(0);
2294
2295    rb = tcg_temp_new_i64();
2296    mh = tcg_temp_new_i64();
2297    ml = tcg_temp_new_i64();
2298    tmp = tcg_temp_new_i64();
2299
2300    tcg_gen_extu_tl_i64(rb, cpu_gpr[a->vrb]);
2301    tcg_gen_andi_i64(tmp, rb, 7);
2302    tcg_gen_shli_i64(tmp, tmp, 3);
2303    if (right) {
2304        tcg_gen_shr_i64(tmp, ones, tmp);
2305    } else {
2306        tcg_gen_shl_i64(tmp, ones, tmp);
2307    }
2308    tcg_gen_not_i64(tmp, tmp);
2309
2310    if (right) {
2311        tcg_gen_movcond_i64(TCG_COND_LTU, mh, rb, tcg_constant_i64(8),
2312                            tmp, ones);
2313        tcg_gen_movcond_i64(TCG_COND_LTU, ml, rb, tcg_constant_i64(8),
2314                            zero, tmp);
2315        tcg_gen_movcond_i64(TCG_COND_LTU, ml, rb, tcg_constant_i64(16),
2316                            ml, ones);
2317    } else {
2318        tcg_gen_movcond_i64(TCG_COND_LTU, ml, rb, tcg_constant_i64(8),
2319                            tmp, ones);
2320        tcg_gen_movcond_i64(TCG_COND_LTU, mh, rb, tcg_constant_i64(8),
2321                            zero, tmp);
2322        tcg_gen_movcond_i64(TCG_COND_LTU, mh, rb, tcg_constant_i64(16),
2323                            mh, ones);
2324    }
2325
2326    get_avr64(tmp, a->vra, true);
2327    tcg_gen_and_i64(tmp, tmp, mh);
2328    set_avr64(a->vrt, tmp, true);
2329
2330    get_avr64(tmp, a->vra, false);
2331    tcg_gen_and_i64(tmp, tmp, ml);
2332    set_avr64(a->vrt, tmp, false);
2333    return true;
2334}
2335
2336TRANS(VCLRLB, do_vclrb, false)
2337TRANS(VCLRRB, do_vclrb, true)
2338
2339#define GEN_VAFORM_PAIRED(name0, name1, opc2)                           \
2340static void glue(gen_, name0##_##name1)(DisasContext *ctx)              \
2341    {                                                                   \
2342        TCGv_ptr ra, rb, rc, rd;                                        \
2343        if (unlikely(!ctx->altivec_enabled)) {                          \
2344            gen_exception(ctx, POWERPC_EXCP_VPU);                       \
2345            return;                                                     \
2346        }                                                               \
2347        ra = gen_avr_ptr(rA(ctx->opcode));                              \
2348        rb = gen_avr_ptr(rB(ctx->opcode));                              \
2349        rc = gen_avr_ptr(rC(ctx->opcode));                              \
2350        rd = gen_avr_ptr(rD(ctx->opcode));                              \
2351        if (Rc(ctx->opcode)) {                                          \
2352            gen_helper_##name1(cpu_env, rd, ra, rb, rc);                \
2353        } else {                                                        \
2354            gen_helper_##name0(cpu_env, rd, ra, rb, rc);                \
2355        }                                                               \
2356    }
2357
2358GEN_VAFORM_PAIRED(vmaddfp, vnmsubfp, 23)
2359
2360static bool do_va_helper(DisasContext *ctx, arg_VA *a,
2361    void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_ptr))
2362{
2363    TCGv_ptr vrt, vra, vrb, vrc;
2364    REQUIRE_VECTOR(ctx);
2365
2366    vrt = gen_avr_ptr(a->vrt);
2367    vra = gen_avr_ptr(a->vra);
2368    vrb = gen_avr_ptr(a->vrb);
2369    vrc = gen_avr_ptr(a->rc);
2370    gen_helper(vrt, vra, vrb, vrc);
2371    return true;
2372}
2373
2374TRANS_FLAGS2(ALTIVEC_207, VADDECUQ, do_va_helper, gen_helper_VADDECUQ)
2375TRANS_FLAGS2(ALTIVEC_207, VADDEUQM, do_va_helper, gen_helper_VADDEUQM)
2376
2377TRANS_FLAGS2(ALTIVEC_207, VSUBEUQM, do_va_helper, gen_helper_VSUBEUQM)
2378TRANS_FLAGS2(ALTIVEC_207, VSUBECUQ, do_va_helper, gen_helper_VSUBECUQ)
2379
2380TRANS_FLAGS(ALTIVEC, VPERM, do_va_helper, gen_helper_VPERM)
2381TRANS_FLAGS2(ISA300, VPERMR, do_va_helper, gen_helper_VPERMR)
2382
2383static void gen_vmladduhm_vec(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b,
2384                              TCGv_vec c)
2385{
2386    tcg_gen_mul_vec(vece, t, a, b);
2387    tcg_gen_add_vec(vece, t, t, c);
2388}
2389
2390static bool trans_VMLADDUHM(DisasContext *ctx, arg_VA *a)
2391{
2392    static const TCGOpcode vecop_list[] = {
2393        INDEX_op_add_vec, INDEX_op_mul_vec, 0
2394    };
2395
2396    static const GVecGen4 op = {
2397        .fno = gen_helper_VMLADDUHM,
2398        .fniv = gen_vmladduhm_vec,
2399        .opt_opc = vecop_list,
2400        .vece = MO_16
2401    };
2402
2403    REQUIRE_INSNS_FLAGS(ctx, ALTIVEC);
2404    REQUIRE_VECTOR(ctx);
2405
2406    tcg_gen_gvec_4(avr_full_offset(a->vrt), avr_full_offset(a->vra),
2407                   avr_full_offset(a->vrb), avr_full_offset(a->rc),
2408                   16, 16, &op);
2409
2410    return true;
2411}
2412
2413static bool trans_VSEL(DisasContext *ctx, arg_VA *a)
2414{
2415    REQUIRE_INSNS_FLAGS(ctx, ALTIVEC);
2416    REQUIRE_VECTOR(ctx);
2417
2418    tcg_gen_gvec_bitsel(MO_64, avr_full_offset(a->vrt), avr_full_offset(a->rc),
2419                        avr_full_offset(a->vrb), avr_full_offset(a->vra),
2420                        16, 16);
2421
2422    return true;
2423}
2424
2425TRANS_FLAGS(ALTIVEC, VMSUMUBM, do_va_helper, gen_helper_VMSUMUBM)
2426TRANS_FLAGS(ALTIVEC, VMSUMMBM, do_va_helper, gen_helper_VMSUMMBM)
2427TRANS_FLAGS(ALTIVEC, VMSUMSHM, do_va_helper, gen_helper_VMSUMSHM)
2428TRANS_FLAGS(ALTIVEC, VMSUMUHM, do_va_helper, gen_helper_VMSUMUHM)
2429
2430static bool do_va_env_helper(DisasContext *ctx, arg_VA *a,
2431    void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_ptr))
2432{
2433    TCGv_ptr vrt, vra, vrb, vrc;
2434    REQUIRE_VECTOR(ctx);
2435
2436    vrt = gen_avr_ptr(a->vrt);
2437    vra = gen_avr_ptr(a->vra);
2438    vrb = gen_avr_ptr(a->vrb);
2439    vrc = gen_avr_ptr(a->rc);
2440    gen_helper(cpu_env, vrt, vra, vrb, vrc);
2441    return true;
2442}
2443
2444TRANS_FLAGS(ALTIVEC, VMSUMUHS, do_va_env_helper, gen_helper_VMSUMUHS)
2445TRANS_FLAGS(ALTIVEC, VMSUMSHS, do_va_env_helper, gen_helper_VMSUMSHS)
2446
2447TRANS_FLAGS(ALTIVEC, VMHADDSHS, do_va_env_helper, gen_helper_VMHADDSHS)
2448TRANS_FLAGS(ALTIVEC, VMHRADDSHS, do_va_env_helper, gen_helper_VMHRADDSHS)
2449
2450GEN_VXFORM_NOA(vclzb, 1, 28)
2451GEN_VXFORM_NOA(vclzh, 1, 29)
2452GEN_VXFORM_TRANS(vclzw, 1, 30)
2453GEN_VXFORM_TRANS(vclzd, 1, 31)
2454
2455static bool do_vneg(DisasContext *ctx, arg_VX_tb *a, unsigned vece)
2456{
2457    REQUIRE_INSNS_FLAGS2(ctx, ISA300);
2458    REQUIRE_VECTOR(ctx);
2459
2460    tcg_gen_gvec_neg(vece, avr_full_offset(a->vrt), avr_full_offset(a->vrb),
2461                     16, 16);
2462    return true;
2463}
2464
2465TRANS(VNEGW, do_vneg, MO_32)
2466TRANS(VNEGD, do_vneg, MO_64)
2467
2468static void gen_vexts_i64(TCGv_i64 t, TCGv_i64 b, int64_t s)
2469{
2470    tcg_gen_sextract_i64(t, b, 0, 64 - s);
2471}
2472
2473static void gen_vexts_i32(TCGv_i32 t, TCGv_i32 b, int32_t s)
2474{
2475    tcg_gen_sextract_i32(t, b, 0, 32 - s);
2476}
2477
2478static void gen_vexts_vec(unsigned vece, TCGv_vec t, TCGv_vec b, int64_t s)
2479{
2480    tcg_gen_shli_vec(vece, t, b, s);
2481    tcg_gen_sari_vec(vece, t, t, s);
2482}
2483
2484static bool do_vexts(DisasContext *ctx, arg_VX_tb *a, unsigned vece, int64_t s)
2485{
2486    static const TCGOpcode vecop_list[] = {
2487        INDEX_op_shli_vec, INDEX_op_sari_vec, 0
2488    };
2489
2490    static const GVecGen2i op[2] = {
2491        {
2492            .fni4 = gen_vexts_i32,
2493            .fniv = gen_vexts_vec,
2494            .opt_opc = vecop_list,
2495            .vece = MO_32
2496        },
2497        {
2498            .fni8 = gen_vexts_i64,
2499            .fniv = gen_vexts_vec,
2500            .opt_opc = vecop_list,
2501            .vece = MO_64
2502        },
2503    };
2504
2505    REQUIRE_INSNS_FLAGS2(ctx, ISA300);
2506    REQUIRE_VECTOR(ctx);
2507
2508    tcg_gen_gvec_2i(avr_full_offset(a->vrt), avr_full_offset(a->vrb),
2509                    16, 16, s, &op[vece - MO_32]);
2510
2511    return true;
2512}
2513
2514TRANS(VEXTSB2W, do_vexts, MO_32, 24);
2515TRANS(VEXTSH2W, do_vexts, MO_32, 16);
2516TRANS(VEXTSB2D, do_vexts, MO_64, 56);
2517TRANS(VEXTSH2D, do_vexts, MO_64, 48);
2518TRANS(VEXTSW2D, do_vexts, MO_64, 32);
2519
2520static bool trans_VEXTSD2Q(DisasContext *ctx, arg_VX_tb *a)
2521{
2522    TCGv_i64 tmp;
2523
2524    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2525    REQUIRE_VECTOR(ctx);
2526
2527    tmp = tcg_temp_new_i64();
2528
2529    get_avr64(tmp, a->vrb, false);
2530    set_avr64(a->vrt, tmp, false);
2531    tcg_gen_sari_i64(tmp, tmp, 63);
2532    set_avr64(a->vrt, tmp, true);
2533    return true;
2534}
2535
2536GEN_VXFORM_NOA_2(vctzb, 1, 24, 28)
2537GEN_VXFORM_NOA_2(vctzh, 1, 24, 29)
2538GEN_VXFORM_NOA_2(vctzw, 1, 24, 30)
2539GEN_VXFORM_NOA_2(vctzd, 1, 24, 31)
2540GEN_VXFORM_NOA_3(vclzlsbb, 1, 24, 0)
2541GEN_VXFORM_NOA_3(vctzlsbb, 1, 24, 1)
2542GEN_VXFORM_NOA(vpopcntb, 1, 28)
2543GEN_VXFORM_NOA(vpopcnth, 1, 29)
2544GEN_VXFORM_NOA(vpopcntw, 1, 30)
2545GEN_VXFORM_NOA(vpopcntd, 1, 31)
2546GEN_VXFORM_DUAL(vclzb, PPC_NONE, PPC2_ALTIVEC_207, \
2547                vpopcntb, PPC_NONE, PPC2_ALTIVEC_207)
2548GEN_VXFORM_DUAL(vclzh, PPC_NONE, PPC2_ALTIVEC_207, \
2549                vpopcnth, PPC_NONE, PPC2_ALTIVEC_207)
2550GEN_VXFORM_DUAL(vclzw, PPC_NONE, PPC2_ALTIVEC_207, \
2551                vpopcntw, PPC_NONE, PPC2_ALTIVEC_207)
2552GEN_VXFORM_DUAL(vclzd, PPC_NONE, PPC2_ALTIVEC_207, \
2553                vpopcntd, PPC_NONE, PPC2_ALTIVEC_207)
2554GEN_VXFORM(vbpermd, 6, 23);
2555GEN_VXFORM(vbpermq, 6, 21);
2556GEN_VXFORM_TRANS(vgbbd, 6, 20);
2557GEN_VXFORM(vpmsumb, 4, 16)
2558GEN_VXFORM(vpmsumh, 4, 17)
2559GEN_VXFORM(vpmsumw, 4, 18)
2560
2561#define GEN_BCD(op)                                 \
2562static void gen_##op(DisasContext *ctx)             \
2563{                                                   \
2564    TCGv_ptr ra, rb, rd;                            \
2565    TCGv_i32 ps;                                    \
2566                                                    \
2567    if (unlikely(!ctx->altivec_enabled)) {          \
2568        gen_exception(ctx, POWERPC_EXCP_VPU);       \
2569        return;                                     \
2570    }                                               \
2571                                                    \
2572    ra = gen_avr_ptr(rA(ctx->opcode));              \
2573    rb = gen_avr_ptr(rB(ctx->opcode));              \
2574    rd = gen_avr_ptr(rD(ctx->opcode));              \
2575                                                    \
2576    ps = tcg_constant_i32((ctx->opcode & 0x200) != 0); \
2577                                                    \
2578    gen_helper_##op(cpu_crf[6], rd, ra, rb, ps);    \
2579}
2580
2581#define GEN_BCD2(op)                                \
2582static void gen_##op(DisasContext *ctx)             \
2583{                                                   \
2584    TCGv_ptr rd, rb;                                \
2585    TCGv_i32 ps;                                    \
2586                                                    \
2587    if (unlikely(!ctx->altivec_enabled)) {          \
2588        gen_exception(ctx, POWERPC_EXCP_VPU);       \
2589        return;                                     \
2590    }                                               \
2591                                                    \
2592    rb = gen_avr_ptr(rB(ctx->opcode));              \
2593    rd = gen_avr_ptr(rD(ctx->opcode));              \
2594                                                    \
2595    ps = tcg_constant_i32((ctx->opcode & 0x200) != 0); \
2596                                                    \
2597    gen_helper_##op(cpu_crf[6], rd, rb, ps);        \
2598}
2599
2600GEN_BCD(bcdadd)
2601GEN_BCD(bcdsub)
2602GEN_BCD2(bcdcfn)
2603GEN_BCD2(bcdctn)
2604GEN_BCD2(bcdcfz)
2605GEN_BCD2(bcdctz)
2606GEN_BCD2(bcdcfsq)
2607GEN_BCD2(bcdctsq)
2608GEN_BCD2(bcdsetsgn)
2609GEN_BCD(bcdcpsgn);
2610GEN_BCD(bcds);
2611GEN_BCD(bcdus);
2612GEN_BCD(bcdsr);
2613GEN_BCD(bcdtrunc);
2614GEN_BCD(bcdutrunc);
2615
2616static void gen_xpnd04_1(DisasContext *ctx)
2617{
2618    switch (opc4(ctx->opcode)) {
2619    case 0:
2620        gen_bcdctsq(ctx);
2621        break;
2622    case 2:
2623        gen_bcdcfsq(ctx);
2624        break;
2625    case 4:
2626        gen_bcdctz(ctx);
2627        break;
2628    case 5:
2629        gen_bcdctn(ctx);
2630        break;
2631    case 6:
2632        gen_bcdcfz(ctx);
2633        break;
2634    case 7:
2635        gen_bcdcfn(ctx);
2636        break;
2637    case 31:
2638        gen_bcdsetsgn(ctx);
2639        break;
2640    default:
2641        gen_invalid(ctx);
2642        break;
2643    }
2644}
2645
2646static void gen_xpnd04_2(DisasContext *ctx)
2647{
2648    switch (opc4(ctx->opcode)) {
2649    case 0:
2650        gen_bcdctsq(ctx);
2651        break;
2652    case 2:
2653        gen_bcdcfsq(ctx);
2654        break;
2655    case 4:
2656        gen_bcdctz(ctx);
2657        break;
2658    case 6:
2659        gen_bcdcfz(ctx);
2660        break;
2661    case 7:
2662        gen_bcdcfn(ctx);
2663        break;
2664    case 31:
2665        gen_bcdsetsgn(ctx);
2666        break;
2667    default:
2668        gen_invalid(ctx);
2669        break;
2670    }
2671}
2672
2673
2674GEN_VXFORM_DUAL(vsubsws, PPC_ALTIVEC, PPC_NONE, \
2675                xpnd04_2, PPC_NONE, PPC2_ISA300)
2676
2677GEN_VXFORM_DUAL(vsububm, PPC_ALTIVEC, PPC_NONE, \
2678                bcdadd, PPC_NONE, PPC2_ALTIVEC_207)
2679GEN_VXFORM_DUAL(vsububs, PPC_ALTIVEC, PPC_NONE, \
2680                bcdadd, PPC_NONE, PPC2_ALTIVEC_207)
2681GEN_VXFORM_DUAL(vsubuhm, PPC_ALTIVEC, PPC_NONE, \
2682                bcdsub, PPC_NONE, PPC2_ALTIVEC_207)
2683GEN_VXFORM_DUAL(vsubuhs, PPC_ALTIVEC, PPC_NONE, \
2684                bcdsub, PPC_NONE, PPC2_ALTIVEC_207)
2685GEN_VXFORM_DUAL(vaddshs, PPC_ALTIVEC, PPC_NONE, \
2686                bcdcpsgn, PPC_NONE, PPC2_ISA300)
2687GEN_VXFORM_DUAL(vsubudm, PPC2_ALTIVEC_207, PPC_NONE, \
2688                bcds, PPC_NONE, PPC2_ISA300)
2689GEN_VXFORM_DUAL(vsubuwm, PPC_ALTIVEC, PPC_NONE, \
2690                bcdus, PPC_NONE, PPC2_ISA300)
2691GEN_VXFORM_DUAL(vsubsbs, PPC_ALTIVEC, PPC_NONE, \
2692                bcdtrunc, PPC_NONE, PPC2_ISA300)
2693
2694static void gen_vsbox(DisasContext *ctx)
2695{
2696    TCGv_ptr ra, rd;
2697    if (unlikely(!ctx->altivec_enabled)) {
2698        gen_exception(ctx, POWERPC_EXCP_VPU);
2699        return;
2700    }
2701    ra = gen_avr_ptr(rA(ctx->opcode));
2702    rd = gen_avr_ptr(rD(ctx->opcode));
2703    gen_helper_vsbox(rd, ra);
2704}
2705
2706GEN_VXFORM(vcipher, 4, 20)
2707GEN_VXFORM(vcipherlast, 4, 20)
2708GEN_VXFORM(vncipher, 4, 21)
2709GEN_VXFORM(vncipherlast, 4, 21)
2710
2711GEN_VXFORM_DUAL(vcipher, PPC_NONE, PPC2_ALTIVEC_207,
2712                vcipherlast, PPC_NONE, PPC2_ALTIVEC_207)
2713GEN_VXFORM_DUAL(vncipher, PPC_NONE, PPC2_ALTIVEC_207,
2714                vncipherlast, PPC_NONE, PPC2_ALTIVEC_207)
2715
2716#define VSHASIGMA(op)                         \
2717static void gen_##op(DisasContext *ctx)       \
2718{                                             \
2719    TCGv_ptr ra, rd;                          \
2720    TCGv_i32 st_six;                          \
2721    if (unlikely(!ctx->altivec_enabled)) {    \
2722        gen_exception(ctx, POWERPC_EXCP_VPU); \
2723        return;                               \
2724    }                                         \
2725    ra = gen_avr_ptr(rA(ctx->opcode));        \
2726    rd = gen_avr_ptr(rD(ctx->opcode));        \
2727    st_six = tcg_constant_i32(rB(ctx->opcode));  \
2728    gen_helper_##op(rd, ra, st_six);          \
2729}
2730
2731VSHASIGMA(vshasigmaw)
2732VSHASIGMA(vshasigmad)
2733
2734GEN_VXFORM3(vpermxor, 22, 0xFF)
2735GEN_VXFORM_DUAL(vsldoi, PPC_ALTIVEC, PPC_NONE,
2736                vpermxor, PPC_NONE, PPC2_ALTIVEC_207)
2737
2738static bool trans_VCFUGED(DisasContext *ctx, arg_VX *a)
2739{
2740    static const GVecGen3 g = {
2741        .fni8 = gen_helper_CFUGED,
2742        .vece = MO_64,
2743    };
2744
2745    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2746    REQUIRE_VECTOR(ctx);
2747
2748    tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra),
2749                   avr_full_offset(a->vrb), 16, 16, &g);
2750
2751    return true;
2752}
2753
2754static bool trans_VCLZDM(DisasContext *ctx, arg_VX *a)
2755{
2756    static const GVecGen3i g = {
2757        .fni8 = do_cntzdm,
2758        .vece = MO_64,
2759    };
2760
2761    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2762    REQUIRE_VECTOR(ctx);
2763
2764    tcg_gen_gvec_3i(avr_full_offset(a->vrt), avr_full_offset(a->vra),
2765                    avr_full_offset(a->vrb), 16, 16, false, &g);
2766
2767    return true;
2768}
2769
2770static bool trans_VCTZDM(DisasContext *ctx, arg_VX *a)
2771{
2772    static const GVecGen3i g = {
2773        .fni8 = do_cntzdm,
2774        .vece = MO_64,
2775    };
2776
2777    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2778    REQUIRE_VECTOR(ctx);
2779
2780    tcg_gen_gvec_3i(avr_full_offset(a->vrt), avr_full_offset(a->vra),
2781                    avr_full_offset(a->vrb), 16, 16, true, &g);
2782
2783    return true;
2784}
2785
2786static bool trans_VPDEPD(DisasContext *ctx, arg_VX *a)
2787{
2788    static const GVecGen3 g = {
2789        .fni8 = gen_helper_PDEPD,
2790        .vece = MO_64,
2791    };
2792
2793    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2794    REQUIRE_VECTOR(ctx);
2795
2796    tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra),
2797                   avr_full_offset(a->vrb), 16, 16, &g);
2798
2799    return true;
2800}
2801
2802static bool trans_VPEXTD(DisasContext *ctx, arg_VX *a)
2803{
2804    static const GVecGen3 g = {
2805        .fni8 = gen_helper_PEXTD,
2806        .vece = MO_64,
2807    };
2808
2809    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2810    REQUIRE_VECTOR(ctx);
2811
2812    tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra),
2813                   avr_full_offset(a->vrb), 16, 16, &g);
2814
2815    return true;
2816}
2817
2818static bool trans_VMSUMUDM(DisasContext *ctx, arg_VA *a)
2819{
2820    TCGv_i64 rl, rh, src1, src2;
2821    int dw;
2822
2823    REQUIRE_INSNS_FLAGS2(ctx, ISA300);
2824    REQUIRE_VECTOR(ctx);
2825
2826    rh = tcg_temp_new_i64();
2827    rl = tcg_temp_new_i64();
2828    src1 = tcg_temp_new_i64();
2829    src2 = tcg_temp_new_i64();
2830
2831    get_avr64(rl, a->rc, false);
2832    get_avr64(rh, a->rc, true);
2833
2834    for (dw = 0; dw < 2; dw++) {
2835        get_avr64(src1, a->vra, dw);
2836        get_avr64(src2, a->vrb, dw);
2837        tcg_gen_mulu2_i64(src1, src2, src1, src2);
2838        tcg_gen_add2_i64(rl, rh, rl, rh, src1, src2);
2839    }
2840
2841    set_avr64(a->vrt, rl, false);
2842    set_avr64(a->vrt, rh, true);
2843    return true;
2844}
2845
2846static bool trans_VMSUMCUD(DisasContext *ctx, arg_VA *a)
2847{
2848    TCGv_i64 tmp0, tmp1, prod1h, prod1l, prod0h, prod0l, zero;
2849
2850    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2851    REQUIRE_VECTOR(ctx);
2852
2853    tmp0 = tcg_temp_new_i64();
2854    tmp1 = tcg_temp_new_i64();
2855    prod1h = tcg_temp_new_i64();
2856    prod1l = tcg_temp_new_i64();
2857    prod0h = tcg_temp_new_i64();
2858    prod0l = tcg_temp_new_i64();
2859    zero = tcg_constant_i64(0);
2860
2861    /* prod1 = vsr[vra+32].dw[1] * vsr[vrb+32].dw[1] */
2862    get_avr64(tmp0, a->vra, false);
2863    get_avr64(tmp1, a->vrb, false);
2864    tcg_gen_mulu2_i64(prod1l, prod1h, tmp0, tmp1);
2865
2866    /* prod0 = vsr[vra+32].dw[0] * vsr[vrb+32].dw[0] */
2867    get_avr64(tmp0, a->vra, true);
2868    get_avr64(tmp1, a->vrb, true);
2869    tcg_gen_mulu2_i64(prod0l, prod0h, tmp0, tmp1);
2870
2871    /* Sum lower 64-bits elements */
2872    get_avr64(tmp1, a->rc, false);
2873    tcg_gen_add2_i64(tmp1, tmp0, tmp1, zero, prod1l, zero);
2874    tcg_gen_add2_i64(tmp1, tmp0, tmp1, tmp0, prod0l, zero);
2875
2876    /*
2877     * Discard lower 64-bits, leaving the carry into bit 64.
2878     * Then sum the higher 64-bit elements.
2879     */
2880    get_avr64(tmp1, a->rc, true);
2881    tcg_gen_add2_i64(tmp1, tmp0, tmp0, zero, tmp1, zero);
2882    tcg_gen_add2_i64(tmp1, tmp0, tmp1, tmp0, prod1h, zero);
2883    tcg_gen_add2_i64(tmp1, tmp0, tmp1, tmp0, prod0h, zero);
2884
2885    /* Discard 64 more bits to complete the CHOP128(temp >> 128) */
2886    set_avr64(a->vrt, tmp0, false);
2887    set_avr64(a->vrt, zero, true);
2888    return true;
2889}
2890
2891static bool do_vx_helper(DisasContext *ctx, arg_VX *a,
2892                         void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_ptr))
2893{
2894    TCGv_ptr ra, rb, rd;
2895    REQUIRE_VECTOR(ctx);
2896
2897    ra = gen_avr_ptr(a->vra);
2898    rb = gen_avr_ptr(a->vrb);
2899    rd = gen_avr_ptr(a->vrt);
2900    gen_helper(rd, ra, rb);
2901    return true;
2902}
2903
2904TRANS_FLAGS2(ALTIVEC_207, VADDCUQ, do_vx_helper, gen_helper_VADDCUQ)
2905TRANS_FLAGS2(ALTIVEC_207, VADDUQM, do_vx_helper, gen_helper_VADDUQM)
2906
2907TRANS_FLAGS2(ALTIVEC_207, VPMSUMD, do_vx_helper, gen_helper_VPMSUMD)
2908
2909TRANS_FLAGS2(ALTIVEC_207, VSUBCUQ, do_vx_helper, gen_helper_VSUBCUQ)
2910TRANS_FLAGS2(ALTIVEC_207, VSUBUQM, do_vx_helper, gen_helper_VSUBUQM)
2911
2912static void gen_VADDCUW_vec(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b)
2913{
2914    tcg_gen_not_vec(vece, a, a);
2915    tcg_gen_cmp_vec(TCG_COND_LTU, vece, t, a, b);
2916    tcg_gen_and_vec(vece, t, t, tcg_constant_vec_matching(t, vece, 1));
2917}
2918
2919static void gen_VADDCUW_i32(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b)
2920{
2921    tcg_gen_not_i32(a, a);
2922    tcg_gen_setcond_i32(TCG_COND_LTU, t, a, b);
2923}
2924
2925static void gen_VSUBCUW_vec(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b)
2926{
2927    tcg_gen_cmp_vec(TCG_COND_GEU, vece, t, a, b);
2928    tcg_gen_and_vec(vece, t, t, tcg_constant_vec_matching(t, vece, 1));
2929}
2930
2931static void gen_VSUBCUW_i32(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b)
2932{
2933    tcg_gen_setcond_i32(TCG_COND_GEU, t, a, b);
2934}
2935
2936static bool do_vx_vaddsubcuw(DisasContext *ctx, arg_VX *a, int add)
2937{
2938    static const TCGOpcode vecop_list[] = {
2939        INDEX_op_cmp_vec, 0
2940    };
2941
2942    static const GVecGen3 op[] = {
2943        {
2944            .fniv = gen_VSUBCUW_vec,
2945            .fni4 = gen_VSUBCUW_i32,
2946            .opt_opc = vecop_list,
2947            .vece = MO_32
2948        },
2949        {
2950            .fniv = gen_VADDCUW_vec,
2951            .fni4 = gen_VADDCUW_i32,
2952            .opt_opc = vecop_list,
2953            .vece = MO_32
2954        },
2955    };
2956
2957    REQUIRE_INSNS_FLAGS(ctx, ALTIVEC);
2958    REQUIRE_VECTOR(ctx);
2959
2960    tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra),
2961                   avr_full_offset(a->vrb), 16, 16, &op[add]);
2962
2963    return true;
2964}
2965
2966TRANS(VSUBCUW, do_vx_vaddsubcuw, 0)
2967TRANS(VADDCUW, do_vx_vaddsubcuw, 1)
2968
2969static bool do_vx_vmuleo(DisasContext *ctx, arg_VX *a, bool even,
2970                         void (*gen_mul)(TCGv_i64, TCGv_i64, TCGv_i64, TCGv_i64))
2971{
2972    TCGv_i64 vra, vrb, vrt0, vrt1;
2973    REQUIRE_VECTOR(ctx);
2974
2975    vra = tcg_temp_new_i64();
2976    vrb = tcg_temp_new_i64();
2977    vrt0 = tcg_temp_new_i64();
2978    vrt1 = tcg_temp_new_i64();
2979
2980    get_avr64(vra, a->vra, even);
2981    get_avr64(vrb, a->vrb, even);
2982    gen_mul(vrt0, vrt1, vra, vrb);
2983    set_avr64(a->vrt, vrt0, false);
2984    set_avr64(a->vrt, vrt1, true);
2985    return true;
2986}
2987
2988static bool trans_VMULLD(DisasContext *ctx, arg_VX *a)
2989{
2990    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2991    REQUIRE_VECTOR(ctx);
2992
2993    tcg_gen_gvec_mul(MO_64, avr_full_offset(a->vrt), avr_full_offset(a->vra),
2994                     avr_full_offset(a->vrb), 16, 16);
2995
2996    return true;
2997}
2998
2999TRANS_FLAGS(ALTIVEC, VMULESB, do_vx_helper, gen_helper_VMULESB)
3000TRANS_FLAGS(ALTIVEC, VMULOSB, do_vx_helper, gen_helper_VMULOSB)
3001TRANS_FLAGS(ALTIVEC, VMULEUB, do_vx_helper, gen_helper_VMULEUB)
3002TRANS_FLAGS(ALTIVEC, VMULOUB, do_vx_helper, gen_helper_VMULOUB)
3003TRANS_FLAGS(ALTIVEC, VMULESH, do_vx_helper, gen_helper_VMULESH)
3004TRANS_FLAGS(ALTIVEC, VMULOSH, do_vx_helper, gen_helper_VMULOSH)
3005TRANS_FLAGS(ALTIVEC, VMULEUH, do_vx_helper, gen_helper_VMULEUH)
3006TRANS_FLAGS(ALTIVEC, VMULOUH, do_vx_helper, gen_helper_VMULOUH)
3007TRANS_FLAGS2(ALTIVEC_207, VMULESW, do_vx_helper, gen_helper_VMULESW)
3008TRANS_FLAGS2(ALTIVEC_207, VMULOSW, do_vx_helper, gen_helper_VMULOSW)
3009TRANS_FLAGS2(ALTIVEC_207, VMULEUW, do_vx_helper, gen_helper_VMULEUW)
3010TRANS_FLAGS2(ALTIVEC_207, VMULOUW, do_vx_helper, gen_helper_VMULOUW)
3011TRANS_FLAGS2(ISA310, VMULESD, do_vx_vmuleo, true , tcg_gen_muls2_i64)
3012TRANS_FLAGS2(ISA310, VMULOSD, do_vx_vmuleo, false, tcg_gen_muls2_i64)
3013TRANS_FLAGS2(ISA310, VMULEUD, do_vx_vmuleo, true , tcg_gen_mulu2_i64)
3014TRANS_FLAGS2(ISA310, VMULOUD, do_vx_vmuleo, false, tcg_gen_mulu2_i64)
3015
3016static void do_vx_vmulhw_i64(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b, bool sign)
3017{
3018    TCGv_i64 hh, lh, temp;
3019
3020    hh = tcg_temp_new_i64();
3021    lh = tcg_temp_new_i64();
3022    temp = tcg_temp_new_i64();
3023
3024    if (sign) {
3025        tcg_gen_ext32s_i64(lh, a);
3026        tcg_gen_ext32s_i64(temp, b);
3027    } else {
3028        tcg_gen_ext32u_i64(lh, a);
3029        tcg_gen_ext32u_i64(temp, b);
3030    }
3031    tcg_gen_mul_i64(lh, lh, temp);
3032
3033    if (sign) {
3034        tcg_gen_sari_i64(hh, a, 32);
3035        tcg_gen_sari_i64(temp, b, 32);
3036    } else {
3037        tcg_gen_shri_i64(hh, a, 32);
3038        tcg_gen_shri_i64(temp, b, 32);
3039    }
3040    tcg_gen_mul_i64(hh, hh, temp);
3041
3042    tcg_gen_shri_i64(lh, lh, 32);
3043    tcg_gen_deposit_i64(t, hh, lh, 0, 32);
3044}
3045
3046static void do_vx_vmulhd_i64(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b, bool sign)
3047{
3048    TCGv_i64 tlow;
3049
3050    tlow  = tcg_temp_new_i64();
3051    if (sign) {
3052        tcg_gen_muls2_i64(tlow, t, a, b);
3053    } else {
3054        tcg_gen_mulu2_i64(tlow, t, a, b);
3055    }
3056}
3057
3058static bool do_vx_mulh(DisasContext *ctx, arg_VX *a, bool sign,
3059                       void (*func)(TCGv_i64, TCGv_i64, TCGv_i64, bool))
3060{
3061    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
3062    REQUIRE_VECTOR(ctx);
3063
3064    TCGv_i64 vra, vrb, vrt;
3065    int i;
3066
3067    vra = tcg_temp_new_i64();
3068    vrb = tcg_temp_new_i64();
3069    vrt = tcg_temp_new_i64();
3070
3071    for (i = 0; i < 2; i++) {
3072        get_avr64(vra, a->vra, i);
3073        get_avr64(vrb, a->vrb, i);
3074        get_avr64(vrt, a->vrt, i);
3075
3076        func(vrt, vra, vrb, sign);
3077
3078        set_avr64(a->vrt, vrt, i);
3079    }
3080    return true;
3081}
3082
3083TRANS(VMULHSW, do_vx_mulh, true , do_vx_vmulhw_i64)
3084TRANS(VMULHSD, do_vx_mulh, true , do_vx_vmulhd_i64)
3085TRANS(VMULHUW, do_vx_mulh, false, do_vx_vmulhw_i64)
3086TRANS(VMULHUD, do_vx_mulh, false, do_vx_vmulhd_i64)
3087
3088static void do_vavg(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b,
3089                    void (*gen_shr_vec)(unsigned, TCGv_vec, TCGv_vec, int64_t))
3090{
3091    TCGv_vec tmp = tcg_temp_new_vec_matching(t);
3092    tcg_gen_or_vec(vece, tmp, a, b);
3093    tcg_gen_and_vec(vece, tmp, tmp, tcg_constant_vec_matching(t, vece, 1));
3094    gen_shr_vec(vece, a, a, 1);
3095    gen_shr_vec(vece, b, b, 1);
3096    tcg_gen_add_vec(vece, t, a, b);
3097    tcg_gen_add_vec(vece, t, t, tmp);
3098}
3099
3100QEMU_FLATTEN
3101static void gen_vavgu(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b)
3102{
3103    do_vavg(vece, t, a, b, tcg_gen_shri_vec);
3104}
3105
3106QEMU_FLATTEN
3107static void gen_vavgs(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b)
3108{
3109    do_vavg(vece, t, a, b, tcg_gen_sari_vec);
3110}
3111
3112static bool do_vx_vavg(DisasContext *ctx, arg_VX *a, int sign, int vece)
3113{
3114    static const TCGOpcode vecop_list_s[] = {
3115        INDEX_op_add_vec, INDEX_op_sari_vec, 0
3116    };
3117    static const TCGOpcode vecop_list_u[] = {
3118        INDEX_op_add_vec, INDEX_op_shri_vec, 0
3119    };
3120
3121    static const GVecGen3 op[2][3] = {
3122        {
3123            {
3124                .fniv = gen_vavgu,
3125                .fno = gen_helper_VAVGUB,
3126                .opt_opc = vecop_list_u,
3127                .vece = MO_8
3128            },
3129            {
3130                .fniv = gen_vavgu,
3131                .fno = gen_helper_VAVGUH,
3132                .opt_opc = vecop_list_u,
3133                .vece = MO_16
3134            },
3135            {
3136                .fniv = gen_vavgu,
3137                .fno = gen_helper_VAVGUW,
3138                .opt_opc = vecop_list_u,
3139                .vece = MO_32
3140            },
3141        },
3142        {
3143            {
3144                .fniv = gen_vavgs,
3145                .fno = gen_helper_VAVGSB,
3146                .opt_opc = vecop_list_s,
3147                .vece = MO_8
3148            },
3149            {
3150                .fniv = gen_vavgs,
3151                .fno = gen_helper_VAVGSH,
3152                .opt_opc = vecop_list_s,
3153                .vece = MO_16
3154            },
3155            {
3156                .fniv = gen_vavgs,
3157                .fno = gen_helper_VAVGSW,
3158                .opt_opc = vecop_list_s,
3159                .vece = MO_32
3160            },
3161        },
3162    };
3163
3164    REQUIRE_VECTOR(ctx);
3165
3166    tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra),
3167                   avr_full_offset(a->vrb), 16, 16, &op[sign][vece]);
3168
3169
3170    return true;
3171}
3172
3173
3174TRANS_FLAGS(ALTIVEC, VAVGSB, do_vx_vavg, 1, MO_8)
3175TRANS_FLAGS(ALTIVEC, VAVGSH, do_vx_vavg, 1, MO_16)
3176TRANS_FLAGS(ALTIVEC, VAVGSW, do_vx_vavg, 1, MO_32)
3177TRANS_FLAGS(ALTIVEC, VAVGUB, do_vx_vavg, 0, MO_8)
3178TRANS_FLAGS(ALTIVEC, VAVGUH, do_vx_vavg, 0, MO_16)
3179TRANS_FLAGS(ALTIVEC, VAVGUW, do_vx_vavg, 0, MO_32)
3180
3181static void gen_vabsdu(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b)
3182{
3183    tcg_gen_umax_vec(vece, t, a, b);
3184    tcg_gen_umin_vec(vece, a, a, b);
3185    tcg_gen_sub_vec(vece, t, t, a);
3186}
3187
3188static bool do_vabsdu(DisasContext *ctx, arg_VX *a, const int vece)
3189{
3190    static const TCGOpcode vecop_list[] = {
3191        INDEX_op_umax_vec, INDEX_op_umin_vec, INDEX_op_sub_vec, 0
3192    };
3193
3194    static const GVecGen3 op[] = {
3195        {
3196            .fniv = gen_vabsdu,
3197            .fno = gen_helper_VABSDUB,
3198            .opt_opc = vecop_list,
3199            .vece = MO_8
3200        },
3201        {
3202            .fniv = gen_vabsdu,
3203            .fno = gen_helper_VABSDUH,
3204            .opt_opc = vecop_list,
3205            .vece = MO_16
3206        },
3207        {
3208            .fniv = gen_vabsdu,
3209            .fno = gen_helper_VABSDUW,
3210            .opt_opc = vecop_list,
3211            .vece = MO_32
3212        },
3213    };
3214
3215    REQUIRE_VECTOR(ctx);
3216
3217    tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra),
3218                   avr_full_offset(a->vrb), 16, 16, &op[vece]);
3219
3220    return true;
3221}
3222
3223TRANS_FLAGS2(ISA300, VABSDUB, do_vabsdu, MO_8)
3224TRANS_FLAGS2(ISA300, VABSDUH, do_vabsdu, MO_16)
3225TRANS_FLAGS2(ISA300, VABSDUW, do_vabsdu, MO_32)
3226
3227static bool do_vdiv_vmod(DisasContext *ctx, arg_VX *a, const int vece,
3228                         void (*func_32)(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b),
3229                         void (*func_64)(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b))
3230{
3231    const GVecGen3 op = {
3232        .fni4 = func_32,
3233        .fni8 = func_64,
3234        .vece = vece
3235    };
3236
3237    REQUIRE_VECTOR(ctx);
3238
3239    tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra),
3240                   avr_full_offset(a->vrb), 16, 16, &op);
3241
3242    return true;
3243}
3244
3245#define DIVU32(NAME, DIV)                                               \
3246static void NAME(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b)                    \
3247{                                                                       \
3248    TCGv_i32 zero = tcg_constant_i32(0);                                \
3249    TCGv_i32 one = tcg_constant_i32(1);                                 \
3250    tcg_gen_movcond_i32(TCG_COND_EQ, b, b, zero, one, b);               \
3251    DIV(t, a, b);                                                       \
3252}
3253
3254#define DIVS32(NAME, DIV)                                               \
3255static void NAME(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b)                    \
3256{                                                                       \
3257    TCGv_i32 t0 = tcg_temp_new_i32();                                   \
3258    TCGv_i32 t1 = tcg_temp_new_i32();                                   \
3259    tcg_gen_setcondi_i32(TCG_COND_EQ, t0, a, INT32_MIN);                \
3260    tcg_gen_setcondi_i32(TCG_COND_EQ, t1, b, -1);                       \
3261    tcg_gen_and_i32(t0, t0, t1);                                        \
3262    tcg_gen_setcondi_i32(TCG_COND_EQ, t1, b, 0);                        \
3263    tcg_gen_or_i32(t0, t0, t1);                                         \
3264    tcg_gen_movi_i32(t1, 0);                                            \
3265    tcg_gen_movcond_i32(TCG_COND_NE, b, t0, t1, t0, b);                 \
3266    DIV(t, a, b);                                                       \
3267}
3268
3269#define DIVU64(NAME, DIV)                                               \
3270static void NAME(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b)                    \
3271{                                                                       \
3272    TCGv_i64 zero = tcg_constant_i64(0);                                \
3273    TCGv_i64 one = tcg_constant_i64(1);                                 \
3274    tcg_gen_movcond_i64(TCG_COND_EQ, b, b, zero, one, b);               \
3275    DIV(t, a, b);                                                       \
3276}
3277
3278#define DIVS64(NAME, DIV)                                               \
3279static void NAME(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b)                    \
3280{                                                                       \
3281    TCGv_i64 t0 = tcg_temp_new_i64();                                   \
3282    TCGv_i64 t1 = tcg_temp_new_i64();                                   \
3283    tcg_gen_setcondi_i64(TCG_COND_EQ, t0, a, INT64_MIN);                \
3284    tcg_gen_setcondi_i64(TCG_COND_EQ, t1, b, -1);                       \
3285    tcg_gen_and_i64(t0, t0, t1);                                        \
3286    tcg_gen_setcondi_i64(TCG_COND_EQ, t1, b, 0);                        \
3287    tcg_gen_or_i64(t0, t0, t1);                                         \
3288    tcg_gen_movi_i64(t1, 0);                                            \
3289    tcg_gen_movcond_i64(TCG_COND_NE, b, t0, t1, t0, b);                 \
3290    DIV(t, a, b);                                                       \
3291}
3292
3293DIVS32(do_divsw, tcg_gen_div_i32)
3294DIVU32(do_divuw, tcg_gen_divu_i32)
3295DIVS64(do_divsd, tcg_gen_div_i64)
3296DIVU64(do_divud, tcg_gen_divu_i64)
3297
3298TRANS_FLAGS2(ISA310, VDIVSW, do_vdiv_vmod, MO_32, do_divsw, NULL)
3299TRANS_FLAGS2(ISA310, VDIVUW, do_vdiv_vmod, MO_32, do_divuw, NULL)
3300TRANS_FLAGS2(ISA310, VDIVSD, do_vdiv_vmod, MO_64, NULL, do_divsd)
3301TRANS_FLAGS2(ISA310, VDIVUD, do_vdiv_vmod, MO_64, NULL, do_divud)
3302TRANS_FLAGS2(ISA310, VDIVSQ, do_vx_helper, gen_helper_VDIVSQ)
3303TRANS_FLAGS2(ISA310, VDIVUQ, do_vx_helper, gen_helper_VDIVUQ)
3304
3305static void do_dives_i32(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b)
3306{
3307    TCGv_i64 val1, val2;
3308
3309    val1 = tcg_temp_new_i64();
3310    val2 = tcg_temp_new_i64();
3311
3312    tcg_gen_ext_i32_i64(val1, a);
3313    tcg_gen_ext_i32_i64(val2, b);
3314
3315    /* (a << 32)/b */
3316    tcg_gen_shli_i64(val1, val1, 32);
3317    tcg_gen_div_i64(val1, val1, val2);
3318
3319    /* if quotient doesn't fit in 32 bits the result is undefined */
3320    tcg_gen_extrl_i64_i32(t, val1);
3321}
3322
3323static void do_diveu_i32(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b)
3324{
3325    TCGv_i64 val1, val2;
3326
3327    val1 = tcg_temp_new_i64();
3328    val2 = tcg_temp_new_i64();
3329
3330    tcg_gen_extu_i32_i64(val1, a);
3331    tcg_gen_extu_i32_i64(val2, b);
3332
3333    /* (a << 32)/b */
3334    tcg_gen_shli_i64(val1, val1, 32);
3335    tcg_gen_divu_i64(val1, val1, val2);
3336
3337    /* if quotient doesn't fit in 32 bits the result is undefined */
3338    tcg_gen_extrl_i64_i32(t, val1);
3339}
3340
3341DIVS32(do_divesw, do_dives_i32)
3342DIVU32(do_diveuw, do_diveu_i32)
3343
3344DIVS32(do_modsw, tcg_gen_rem_i32)
3345DIVU32(do_moduw, tcg_gen_remu_i32)
3346DIVS64(do_modsd, tcg_gen_rem_i64)
3347DIVU64(do_modud, tcg_gen_remu_i64)
3348
3349TRANS_FLAGS2(ISA310, VDIVESW, do_vdiv_vmod, MO_32, do_divesw, NULL)
3350TRANS_FLAGS2(ISA310, VDIVEUW, do_vdiv_vmod, MO_32, do_diveuw, NULL)
3351TRANS_FLAGS2(ISA310, VDIVESD, do_vx_helper, gen_helper_VDIVESD)
3352TRANS_FLAGS2(ISA310, VDIVEUD, do_vx_helper, gen_helper_VDIVEUD)
3353TRANS_FLAGS2(ISA310, VDIVESQ, do_vx_helper, gen_helper_VDIVESQ)
3354TRANS_FLAGS2(ISA310, VDIVEUQ, do_vx_helper, gen_helper_VDIVEUQ)
3355
3356TRANS_FLAGS2(ISA310, VMODSW, do_vdiv_vmod, MO_32, do_modsw , NULL)
3357TRANS_FLAGS2(ISA310, VMODUW, do_vdiv_vmod, MO_32, do_moduw, NULL)
3358TRANS_FLAGS2(ISA310, VMODSD, do_vdiv_vmod, MO_64, NULL, do_modsd)
3359TRANS_FLAGS2(ISA310, VMODUD, do_vdiv_vmod, MO_64, NULL, do_modud)
3360TRANS_FLAGS2(ISA310, VMODSQ, do_vx_helper, gen_helper_VMODSQ)
3361TRANS_FLAGS2(ISA310, VMODUQ, do_vx_helper, gen_helper_VMODUQ)
3362
3363#undef DIVS32
3364#undef DIVU32
3365#undef DIVS64
3366#undef DIVU64
3367
3368#undef GEN_VR_LDX
3369#undef GEN_VR_STX
3370#undef GEN_VR_LVE
3371#undef GEN_VR_STVE
3372
3373#undef GEN_VX_LOGICAL
3374#undef GEN_VX_LOGICAL_207
3375#undef GEN_VXFORM
3376#undef GEN_VXFORM_207
3377#undef GEN_VXFORM_DUAL
3378#undef GEN_VXRFORM_DUAL
3379#undef GEN_VXRFORM1
3380#undef GEN_VXRFORM
3381#undef GEN_VXFORM_VSPLTI
3382#undef GEN_VXFORM_NOA
3383#undef GEN_VXFORM_UIMM
3384#undef GEN_VAFORM_PAIRED
3385
3386#undef GEN_BCD2
3387