4 * Altivec/VMX translation
7 /*** Altivec vector extension ***/
8 /* Altivec registers moves */
10 static inline TCGv_ptr gen_avr_ptr(int reg)
12 TCGv_ptr r = tcg_temp_new_ptr();
13 tcg_gen_addi_ptr(r, cpu_env, avr_full_offset(reg));
17 #define GEN_VR_LDX(name, opc2, opc3) \
18 static void glue(gen_, name)(DisasContext *ctx) \
22 if (unlikely(!ctx->altivec_enabled)) { \
23 gen_exception(ctx, POWERPC_EXCP_VPU); \
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); \
32 * We only need to swap high and low halves. gen_qemu_ld64_i64 \
33 * does necessary 64-bit byteswap already. \
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); \
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); \
49 tcg_temp_free_i64(avr); \
52 #define GEN_VR_STX(name, opc2, opc3) \
53 static void gen_st##name(DisasContext *ctx) \
57 if (unlikely(!ctx->altivec_enabled)) { \
58 gen_exception(ctx, POWERPC_EXCP_VPU); \
61 gen_set_access_type(ctx, ACCESS_INT); \
62 avr = tcg_temp_new_i64(); \
63 EA = tcg_temp_new(); \
64 gen_addr_reg_index(ctx, EA); \
65 tcg_gen_andi_tl(EA, EA, ~0xf); \
67 * We only need to swap high and low halves. gen_qemu_st64_i64 \
68 * does necessary 64-bit byteswap already. \
71 get_avr64(avr, rD(ctx->opcode), false); \
72 gen_qemu_st64_i64(ctx, avr, EA); \
73 tcg_gen_addi_tl(EA, EA, 8); \
74 get_avr64(avr, rD(ctx->opcode), true); \
75 gen_qemu_st64_i64(ctx, avr, EA); \
77 get_avr64(avr, rD(ctx->opcode), true); \
78 gen_qemu_st64_i64(ctx, avr, EA); \
79 tcg_gen_addi_tl(EA, EA, 8); \
80 get_avr64(avr, rD(ctx->opcode), false); \
81 gen_qemu_st64_i64(ctx, avr, EA); \
84 tcg_temp_free_i64(avr); \
87 #define GEN_VR_LVE(name, opc2, opc3, size) \
88 static void gen_lve##name(DisasContext *ctx) \
92 if (unlikely(!ctx->altivec_enabled)) { \
93 gen_exception(ctx, POWERPC_EXCP_VPU); \
96 gen_set_access_type(ctx, ACCESS_INT); \
97 EA = tcg_temp_new(); \
98 gen_addr_reg_index(ctx, EA); \
100 tcg_gen_andi_tl(EA, EA, ~(size - 1)); \
102 rs = gen_avr_ptr(rS(ctx->opcode)); \
103 gen_helper_lve##name(cpu_env, rs, EA); \
105 tcg_temp_free_ptr(rs); \
108 #define GEN_VR_STVE(name, opc2, opc3, size) \
109 static void gen_stve##name(DisasContext *ctx) \
113 if (unlikely(!ctx->altivec_enabled)) { \
114 gen_exception(ctx, POWERPC_EXCP_VPU); \
117 gen_set_access_type(ctx, ACCESS_INT); \
118 EA = tcg_temp_new(); \
119 gen_addr_reg_index(ctx, EA); \
121 tcg_gen_andi_tl(EA, EA, ~(size - 1)); \
123 rs = gen_avr_ptr(rS(ctx->opcode)); \
124 gen_helper_stve##name(cpu_env, rs, EA); \
126 tcg_temp_free_ptr(rs); \
129 GEN_VR_LDX(lvx, 0x07, 0x03);
130 /* As we don't emulate the cache, lvxl is stricly equivalent to lvx */
131 GEN_VR_LDX(lvxl, 0x07, 0x0B);
133 GEN_VR_LVE(bx, 0x07, 0x00, 1);
134 GEN_VR_LVE(hx, 0x07, 0x01, 2);
135 GEN_VR_LVE(wx, 0x07, 0x02, 4);
137 GEN_VR_STX(svx, 0x07, 0x07);
138 /* As we don't emulate the cache, stvxl is stricly equivalent to stvx */
139 GEN_VR_STX(svxl, 0x07, 0x0F);
141 GEN_VR_STVE(bx, 0x07, 0x04, 1);
142 GEN_VR_STVE(hx, 0x07, 0x05, 2);
143 GEN_VR_STVE(wx, 0x07, 0x06, 4);
145 static void gen_mfvscr(DisasContext *ctx)
149 if (unlikely(!ctx->altivec_enabled)) {
150 gen_exception(ctx, POWERPC_EXCP_VPU);
153 avr = tcg_temp_new_i64();
154 tcg_gen_movi_i64(avr, 0);
155 set_avr64(rD(ctx->opcode), avr, true);
156 t = tcg_temp_new_i32();
157 gen_helper_mfvscr(t, cpu_env);
158 tcg_gen_extu_i32_i64(avr, t);
159 set_avr64(rD(ctx->opcode), avr, false);
160 tcg_temp_free_i32(t);
161 tcg_temp_free_i64(avr);
164 static void gen_mtvscr(DisasContext *ctx)
169 if (unlikely(!ctx->altivec_enabled)) {
170 gen_exception(ctx, POWERPC_EXCP_VPU);
174 val = tcg_temp_new_i32();
175 bofs = avr_full_offset(rB(ctx->opcode));
180 tcg_gen_ld_i32(val, cpu_env, bofs);
181 gen_helper_mtvscr(cpu_env, val);
182 tcg_temp_free_i32(val);
185 #define GEN_VX_VMUL10(name, add_cin, ret_carry) \
186 static void glue(gen_, name)(DisasContext *ctx) \
194 if (unlikely(!ctx->altivec_enabled)) { \
195 gen_exception(ctx, POWERPC_EXCP_VPU); \
199 t0 = tcg_temp_new_i64(); \
200 t1 = tcg_temp_new_i64(); \
201 t2 = tcg_temp_new_i64(); \
202 avr = tcg_temp_new_i64(); \
203 ten = tcg_const_i64(10); \
204 z = tcg_const_i64(0); \
207 get_avr64(avr, rA(ctx->opcode), false); \
208 tcg_gen_mulu2_i64(t0, t1, avr, ten); \
209 get_avr64(avr, rB(ctx->opcode), false); \
210 tcg_gen_andi_i64(t2, avr, 0xF); \
211 tcg_gen_add2_i64(avr, t2, t0, t1, t2, z); \
212 set_avr64(rD(ctx->opcode), avr, false); \
214 get_avr64(avr, rA(ctx->opcode), false); \
215 tcg_gen_mulu2_i64(avr, t2, avr, ten); \
216 set_avr64(rD(ctx->opcode), avr, false); \
220 get_avr64(avr, rA(ctx->opcode), true); \
221 tcg_gen_mulu2_i64(t0, t1, avr, ten); \
222 tcg_gen_add2_i64(t0, avr, t0, t1, t2, z); \
223 set_avr64(rD(ctx->opcode), avr, false); \
224 set_avr64(rD(ctx->opcode), z, true); \
226 get_avr64(avr, rA(ctx->opcode), true); \
227 tcg_gen_mul_i64(t0, avr, ten); \
228 tcg_gen_add_i64(avr, t0, t2); \
229 set_avr64(rD(ctx->opcode), avr, true); \
232 tcg_temp_free_i64(t0); \
233 tcg_temp_free_i64(t1); \
234 tcg_temp_free_i64(t2); \
235 tcg_temp_free_i64(avr); \
236 tcg_temp_free_i64(ten); \
237 tcg_temp_free_i64(z); \
240 GEN_VX_VMUL10(vmul10uq, 0, 0);
241 GEN_VX_VMUL10(vmul10euq, 1, 0);
242 GEN_VX_VMUL10(vmul10cuq, 0, 1);
243 GEN_VX_VMUL10(vmul10ecuq, 1, 1);
245 #define GEN_VXFORM_V(name, vece, tcg_op, opc2, opc3) \
246 static void glue(gen_, name)(DisasContext *ctx) \
248 if (unlikely(!ctx->altivec_enabled)) { \
249 gen_exception(ctx, POWERPC_EXCP_VPU); \
254 avr_full_offset(rD(ctx->opcode)), \
255 avr_full_offset(rA(ctx->opcode)), \
256 avr_full_offset(rB(ctx->opcode)), \
260 /* Logical operations */
261 GEN_VXFORM_V(vand, MO_64, tcg_gen_gvec_and, 2, 16);
262 GEN_VXFORM_V(vandc, MO_64, tcg_gen_gvec_andc, 2, 17);
263 GEN_VXFORM_V(vor, MO_64, tcg_gen_gvec_or, 2, 18);
264 GEN_VXFORM_V(vxor, MO_64, tcg_gen_gvec_xor, 2, 19);
265 GEN_VXFORM_V(vnor, MO_64, tcg_gen_gvec_nor, 2, 20);
266 GEN_VXFORM_V(veqv, MO_64, tcg_gen_gvec_eqv, 2, 26);
267 GEN_VXFORM_V(vnand, MO_64, tcg_gen_gvec_nand, 2, 22);
268 GEN_VXFORM_V(vorc, MO_64, tcg_gen_gvec_orc, 2, 21);
270 #define GEN_VXFORM(name, opc2, opc3) \
271 static void glue(gen_, name)(DisasContext *ctx) \
273 TCGv_ptr ra, rb, rd; \
274 if (unlikely(!ctx->altivec_enabled)) { \
275 gen_exception(ctx, POWERPC_EXCP_VPU); \
278 ra = gen_avr_ptr(rA(ctx->opcode)); \
279 rb = gen_avr_ptr(rB(ctx->opcode)); \
280 rd = gen_avr_ptr(rD(ctx->opcode)); \
281 gen_helper_##name(rd, ra, rb); \
282 tcg_temp_free_ptr(ra); \
283 tcg_temp_free_ptr(rb); \
284 tcg_temp_free_ptr(rd); \
287 #define GEN_VXFORM_TRANS(name, opc2, opc3) \
288 static void glue(gen_, name)(DisasContext *ctx) \
290 if (unlikely(!ctx->altivec_enabled)) { \
291 gen_exception(ctx, POWERPC_EXCP_VPU); \
297 #define GEN_VXFORM_ENV(name, opc2, opc3) \
298 static void glue(gen_, name)(DisasContext *ctx) \
300 TCGv_ptr ra, rb, rd; \
301 if (unlikely(!ctx->altivec_enabled)) { \
302 gen_exception(ctx, POWERPC_EXCP_VPU); \
305 ra = gen_avr_ptr(rA(ctx->opcode)); \
306 rb = gen_avr_ptr(rB(ctx->opcode)); \
307 rd = gen_avr_ptr(rD(ctx->opcode)); \
308 gen_helper_##name(cpu_env, rd, ra, rb); \
309 tcg_temp_free_ptr(ra); \
310 tcg_temp_free_ptr(rb); \
311 tcg_temp_free_ptr(rd); \
314 #define GEN_VXFORM3(name, opc2, opc3) \
315 static void glue(gen_, name)(DisasContext *ctx) \
317 TCGv_ptr ra, rb, rc, rd; \
318 if (unlikely(!ctx->altivec_enabled)) { \
319 gen_exception(ctx, POWERPC_EXCP_VPU); \
322 ra = gen_avr_ptr(rA(ctx->opcode)); \
323 rb = gen_avr_ptr(rB(ctx->opcode)); \
324 rc = gen_avr_ptr(rC(ctx->opcode)); \
325 rd = gen_avr_ptr(rD(ctx->opcode)); \
326 gen_helper_##name(rd, ra, rb, rc); \
327 tcg_temp_free_ptr(ra); \
328 tcg_temp_free_ptr(rb); \
329 tcg_temp_free_ptr(rc); \
330 tcg_temp_free_ptr(rd); \
334 * Support for Altivec instruction pairs that use bit 31 (Rc) as
335 * an opcode bit. In general, these pairs come from different
336 * versions of the ISA, so we must also support a pair of flags for
339 #define GEN_VXFORM_DUAL(name0, flg0, flg2_0, name1, flg1, flg2_1) \
340 static void glue(gen_, name0##_##name1)(DisasContext *ctx) \
342 if ((Rc(ctx->opcode) == 0) && \
343 ((ctx->insns_flags & flg0) || (ctx->insns_flags2 & flg2_0))) { \
345 } else if ((Rc(ctx->opcode) == 1) && \
346 ((ctx->insns_flags & flg1) || (ctx->insns_flags2 & flg2_1))) { \
349 gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL); \
354 * We use this macro if one instruction is realized with direct
355 * translation, and second one with helper.
357 #define GEN_VXFORM_TRANS_DUAL(name0, flg0, flg2_0, name1, flg1, flg2_1)\
358 static void glue(gen_, name0##_##name1)(DisasContext *ctx) \
360 if ((Rc(ctx->opcode) == 0) && \
361 ((ctx->insns_flags & flg0) || (ctx->insns_flags2 & flg2_0))) { \
362 if (unlikely(!ctx->altivec_enabled)) { \
363 gen_exception(ctx, POWERPC_EXCP_VPU); \
366 trans_##name0(ctx); \
367 } else if ((Rc(ctx->opcode) == 1) && \
368 ((ctx->insns_flags & flg1) || (ctx->insns_flags2 & flg2_1))) { \
371 gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL); \
375 /* Adds support to provide invalid mask */
376 #define GEN_VXFORM_DUAL_EXT(name0, flg0, flg2_0, inval0, \
377 name1, flg1, flg2_1, inval1) \
378 static void glue(gen_, name0##_##name1)(DisasContext *ctx) \
380 if ((Rc(ctx->opcode) == 0) && \
381 ((ctx->insns_flags & flg0) || (ctx->insns_flags2 & flg2_0)) && \
382 !(ctx->opcode & inval0)) { \
384 } else if ((Rc(ctx->opcode) == 1) && \
385 ((ctx->insns_flags & flg1) || (ctx->insns_flags2 & flg2_1)) && \
386 !(ctx->opcode & inval1)) { \
389 gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL); \
393 #define GEN_VXFORM_HETRO(name, opc2, opc3) \
394 static void glue(gen_, name)(DisasContext *ctx) \
397 if (unlikely(!ctx->altivec_enabled)) { \
398 gen_exception(ctx, POWERPC_EXCP_VPU); \
401 rb = gen_avr_ptr(rB(ctx->opcode)); \
402 gen_helper_##name(cpu_gpr[rD(ctx->opcode)], cpu_gpr[rA(ctx->opcode)], rb); \
403 tcg_temp_free_ptr(rb); \
406 GEN_VXFORM_V(vaddubm, MO_8, tcg_gen_gvec_add, 0, 0);
407 GEN_VXFORM_DUAL_EXT(vaddubm, PPC_ALTIVEC, PPC_NONE, 0, \
408 vmul10cuq, PPC_NONE, PPC2_ISA300, 0x0000F800)
409 GEN_VXFORM_V(vadduhm, MO_16, tcg_gen_gvec_add, 0, 1);
410 GEN_VXFORM_DUAL(vadduhm, PPC_ALTIVEC, PPC_NONE, \
411 vmul10ecuq, PPC_NONE, PPC2_ISA300)
412 GEN_VXFORM_V(vadduwm, MO_32, tcg_gen_gvec_add, 0, 2);
413 GEN_VXFORM_V(vaddudm, MO_64, tcg_gen_gvec_add, 0, 3);
414 GEN_VXFORM_V(vsububm, MO_8, tcg_gen_gvec_sub, 0, 16);
415 GEN_VXFORM_V(vsubuhm, MO_16, tcg_gen_gvec_sub, 0, 17);
416 GEN_VXFORM_V(vsubuwm, MO_32, tcg_gen_gvec_sub, 0, 18);
417 GEN_VXFORM_V(vsubudm, MO_64, tcg_gen_gvec_sub, 0, 19);
418 GEN_VXFORM_V(vmaxub, MO_8, tcg_gen_gvec_umax, 1, 0);
419 GEN_VXFORM_V(vmaxuh, MO_16, tcg_gen_gvec_umax, 1, 1);
420 GEN_VXFORM_V(vmaxuw, MO_32, tcg_gen_gvec_umax, 1, 2);
421 GEN_VXFORM_V(vmaxud, MO_64, tcg_gen_gvec_umax, 1, 3);
422 GEN_VXFORM_V(vmaxsb, MO_8, tcg_gen_gvec_smax, 1, 4);
423 GEN_VXFORM_V(vmaxsh, MO_16, tcg_gen_gvec_smax, 1, 5);
424 GEN_VXFORM_V(vmaxsw, MO_32, tcg_gen_gvec_smax, 1, 6);
425 GEN_VXFORM_V(vmaxsd, MO_64, tcg_gen_gvec_smax, 1, 7);
426 GEN_VXFORM_V(vminub, MO_8, tcg_gen_gvec_umin, 1, 8);
427 GEN_VXFORM_V(vminuh, MO_16, tcg_gen_gvec_umin, 1, 9);
428 GEN_VXFORM_V(vminuw, MO_32, tcg_gen_gvec_umin, 1, 10);
429 GEN_VXFORM_V(vminud, MO_64, tcg_gen_gvec_umin, 1, 11);
430 GEN_VXFORM_V(vminsb, MO_8, tcg_gen_gvec_smin, 1, 12);
431 GEN_VXFORM_V(vminsh, MO_16, tcg_gen_gvec_smin, 1, 13);
432 GEN_VXFORM_V(vminsw, MO_32, tcg_gen_gvec_smin, 1, 14);
433 GEN_VXFORM_V(vminsd, MO_64, tcg_gen_gvec_smin, 1, 15);
434 GEN_VXFORM(vavgub, 1, 16);
435 GEN_VXFORM(vabsdub, 1, 16);
436 GEN_VXFORM_DUAL(vavgub, PPC_ALTIVEC, PPC_NONE, \
437 vabsdub, PPC_NONE, PPC2_ISA300)
438 GEN_VXFORM(vavguh, 1, 17);
439 GEN_VXFORM(vabsduh, 1, 17);
440 GEN_VXFORM_DUAL(vavguh, PPC_ALTIVEC, PPC_NONE, \
441 vabsduh, PPC_NONE, PPC2_ISA300)
442 GEN_VXFORM(vavguw, 1, 18);
443 GEN_VXFORM(vabsduw, 1, 18);
444 GEN_VXFORM_DUAL(vavguw, PPC_ALTIVEC, PPC_NONE, \
445 vabsduw, PPC_NONE, PPC2_ISA300)
446 GEN_VXFORM(vavgsb, 1, 20);
447 GEN_VXFORM(vavgsh, 1, 21);
448 GEN_VXFORM(vavgsw, 1, 22);
449 GEN_VXFORM(vmrghb, 6, 0);
450 GEN_VXFORM(vmrghh, 6, 1);
451 GEN_VXFORM(vmrghw, 6, 2);
452 GEN_VXFORM(vmrglb, 6, 4);
453 GEN_VXFORM(vmrglh, 6, 5);
454 GEN_VXFORM(vmrglw, 6, 6);
456 static void trans_vmrgew(DisasContext *ctx)
458 int VT = rD(ctx->opcode);
459 int VA = rA(ctx->opcode);
460 int VB = rB(ctx->opcode);
461 TCGv_i64 tmp = tcg_temp_new_i64();
462 TCGv_i64 avr = tcg_temp_new_i64();
464 get_avr64(avr, VB, true);
465 tcg_gen_shri_i64(tmp, avr, 32);
466 get_avr64(avr, VA, true);
467 tcg_gen_deposit_i64(avr, avr, tmp, 0, 32);
468 set_avr64(VT, avr, true);
470 get_avr64(avr, VB, false);
471 tcg_gen_shri_i64(tmp, avr, 32);
472 get_avr64(avr, VA, false);
473 tcg_gen_deposit_i64(avr, avr, tmp, 0, 32);
474 set_avr64(VT, avr, false);
476 tcg_temp_free_i64(tmp);
477 tcg_temp_free_i64(avr);
480 static void trans_vmrgow(DisasContext *ctx)
482 int VT = rD(ctx->opcode);
483 int VA = rA(ctx->opcode);
484 int VB = rB(ctx->opcode);
485 TCGv_i64 t0 = tcg_temp_new_i64();
486 TCGv_i64 t1 = tcg_temp_new_i64();
487 TCGv_i64 avr = tcg_temp_new_i64();
489 get_avr64(t0, VB, true);
490 get_avr64(t1, VA, true);
491 tcg_gen_deposit_i64(avr, t0, t1, 32, 32);
492 set_avr64(VT, avr, true);
494 get_avr64(t0, VB, false);
495 get_avr64(t1, VA, false);
496 tcg_gen_deposit_i64(avr, t0, t1, 32, 32);
497 set_avr64(VT, avr, false);
499 tcg_temp_free_i64(t0);
500 tcg_temp_free_i64(t1);
501 tcg_temp_free_i64(avr);
505 * lvsl VRT,RA,RB - Load Vector for Shift Left
507 * Let the EA be the sum (rA|0)+(rB). Let sh=EA[28–31].
508 * Let X be the 32-byte value 0x00 || 0x01 || 0x02 || ... || 0x1E || 0x1F.
509 * Bytes sh:sh+15 of X are placed into vD.
511 static void trans_lvsl(DisasContext *ctx)
513 int VT = rD(ctx->opcode);
514 TCGv_i64 result = tcg_temp_new_i64();
515 TCGv_i64 sh = tcg_temp_new_i64();
516 TCGv EA = tcg_temp_new();
518 /* Get sh(from description) by anding EA with 0xf. */
519 gen_addr_reg_index(ctx, EA);
520 tcg_gen_extu_tl_i64(sh, EA);
521 tcg_gen_andi_i64(sh, sh, 0xfULL);
524 * Create bytes sh:sh+7 of X(from description) and place them in
525 * higher doubleword of vD.
527 tcg_gen_muli_i64(sh, sh, 0x0101010101010101ULL);
528 tcg_gen_addi_i64(result, sh, 0x0001020304050607ull);
529 set_avr64(VT, result, true);
531 * Create bytes sh+8:sh+15 of X(from description) and place them in
532 * lower doubleword of vD.
534 tcg_gen_addi_i64(result, sh, 0x08090a0b0c0d0e0fULL);
535 set_avr64(VT, result, false);
537 tcg_temp_free_i64(result);
538 tcg_temp_free_i64(sh);
543 * lvsr VRT,RA,RB - Load Vector for Shift Right
545 * Let the EA be the sum (rA|0)+(rB). Let sh=EA[28–31].
546 * Let X be the 32-byte value 0x00 || 0x01 || 0x02 || ... || 0x1E || 0x1F.
547 * Bytes (16-sh):(31-sh) of X are placed into vD.
549 static void trans_lvsr(DisasContext *ctx)
551 int VT = rD(ctx->opcode);
552 TCGv_i64 result = tcg_temp_new_i64();
553 TCGv_i64 sh = tcg_temp_new_i64();
554 TCGv EA = tcg_temp_new();
557 /* Get sh(from description) by anding EA with 0xf. */
558 gen_addr_reg_index(ctx, EA);
559 tcg_gen_extu_tl_i64(sh, EA);
560 tcg_gen_andi_i64(sh, sh, 0xfULL);
563 * Create bytes (16-sh):(23-sh) of X(from description) and place them in
564 * higher doubleword of vD.
566 tcg_gen_muli_i64(sh, sh, 0x0101010101010101ULL);
567 tcg_gen_subfi_i64(result, 0x1011121314151617ULL, sh);
568 set_avr64(VT, result, true);
570 * Create bytes (24-sh):(32-sh) of X(from description) and place them in
571 * lower doubleword of vD.
573 tcg_gen_subfi_i64(result, 0x18191a1b1c1d1e1fULL, sh);
574 set_avr64(VT, result, false);
576 tcg_temp_free_i64(result);
577 tcg_temp_free_i64(sh);
582 * vsl VRT,VRA,VRB - Vector Shift Left
584 * Shifting left 128 bit value of vA by value specified in bits 125-127 of vB.
585 * Lowest 3 bits in each byte element of register vB must be identical or
586 * result is undefined.
588 static void trans_vsl(DisasContext *ctx)
590 int VT = rD(ctx->opcode);
591 int VA = rA(ctx->opcode);
592 int VB = rB(ctx->opcode);
593 TCGv_i64 avr = tcg_temp_new_i64();
594 TCGv_i64 sh = tcg_temp_new_i64();
595 TCGv_i64 carry = tcg_temp_new_i64();
596 TCGv_i64 tmp = tcg_temp_new_i64();
598 /* Place bits 125-127 of vB in 'sh'. */
599 get_avr64(avr, VB, false);
600 tcg_gen_andi_i64(sh, avr, 0x07ULL);
603 * Save highest 'sh' bits of lower doubleword element of vA in variable
604 * 'carry' and perform shift on lower doubleword.
606 get_avr64(avr, VA, false);
607 tcg_gen_subfi_i64(tmp, 32, sh);
608 tcg_gen_shri_i64(carry, avr, 32);
609 tcg_gen_shr_i64(carry, carry, tmp);
610 tcg_gen_shl_i64(avr, avr, sh);
611 set_avr64(VT, avr, false);
614 * Perform shift on higher doubleword element of vA and replace lowest
615 * 'sh' bits with 'carry'.
617 get_avr64(avr, VA, true);
618 tcg_gen_shl_i64(avr, avr, sh);
619 tcg_gen_or_i64(avr, avr, carry);
620 set_avr64(VT, avr, true);
622 tcg_temp_free_i64(avr);
623 tcg_temp_free_i64(sh);
624 tcg_temp_free_i64(carry);
625 tcg_temp_free_i64(tmp);
629 * vsr VRT,VRA,VRB - Vector Shift Right
631 * Shifting right 128 bit value of vA by value specified in bits 125-127 of vB.
632 * Lowest 3 bits in each byte element of register vB must be identical or
633 * result is undefined.
635 static void trans_vsr(DisasContext *ctx)
637 int VT = rD(ctx->opcode);
638 int VA = rA(ctx->opcode);
639 int VB = rB(ctx->opcode);
640 TCGv_i64 avr = tcg_temp_new_i64();
641 TCGv_i64 sh = tcg_temp_new_i64();
642 TCGv_i64 carry = tcg_temp_new_i64();
643 TCGv_i64 tmp = tcg_temp_new_i64();
645 /* Place bits 125-127 of vB in 'sh'. */
646 get_avr64(avr, VB, false);
647 tcg_gen_andi_i64(sh, avr, 0x07ULL);
650 * Save lowest 'sh' bits of higher doubleword element of vA in variable
651 * 'carry' and perform shift on higher doubleword.
653 get_avr64(avr, VA, true);
654 tcg_gen_subfi_i64(tmp, 32, sh);
655 tcg_gen_shli_i64(carry, avr, 32);
656 tcg_gen_shl_i64(carry, carry, tmp);
657 tcg_gen_shr_i64(avr, avr, sh);
658 set_avr64(VT, avr, true);
660 * Perform shift on lower doubleword element of vA and replace highest
661 * 'sh' bits with 'carry'.
663 get_avr64(avr, VA, false);
664 tcg_gen_shr_i64(avr, avr, sh);
665 tcg_gen_or_i64(avr, avr, carry);
666 set_avr64(VT, avr, false);
668 tcg_temp_free_i64(avr);
669 tcg_temp_free_i64(sh);
670 tcg_temp_free_i64(carry);
671 tcg_temp_free_i64(tmp);
675 * vgbbd VRT,VRB - Vector Gather Bits by Bytes by Doubleword
677 * All ith bits (i in range 1 to 8) of each byte of doubleword element in source
678 * register are concatenated and placed into ith byte of appropriate doubleword
679 * element in destination register.
681 * Following solution is done for both doubleword elements of source register
682 * in parallel, in order to reduce the number of instructions needed(that's why
684 * First, both doubleword elements of source register vB are placed in
685 * appropriate element of array avr. Bits are gathered in 2x8 iterations(2 for
686 * loops). In first iteration bit 1 of byte 1, bit 2 of byte 2,... bit 8 of
687 * byte 8 are in their final spots so avr[i], i={0,1} can be and-ed with
688 * tcg_mask. For every following iteration, both avr[i] and tcg_mask variables
689 * have to be shifted right for 7 and 8 places, respectively, in order to get
690 * bit 1 of byte 2, bit 2 of byte 3.. bit 7 of byte 8 in their final spots so
691 * shifted avr values(saved in tmp) can be and-ed with new value of tcg_mask...
692 * After first 8 iteration(first loop), all the first bits are in their final
693 * places, all second bits but second bit from eight byte are in their places...
694 * only 1 eight bit from eight byte is in it's place). In second loop we do all
695 * operations symmetrically, in order to get other half of bits in their final
696 * spots. Results for first and second doubleword elements are saved in
697 * result[0] and result[1] respectively. In the end those results are saved in
698 * appropriate doubleword element of destination register vD.
700 static void trans_vgbbd(DisasContext *ctx)
702 int VT = rD(ctx->opcode);
703 int VB = rB(ctx->opcode);
704 TCGv_i64 tmp = tcg_temp_new_i64();
705 uint64_t mask = 0x8040201008040201ULL;
709 result[0] = tcg_temp_new_i64();
710 result[1] = tcg_temp_new_i64();
712 avr[0] = tcg_temp_new_i64();
713 avr[1] = tcg_temp_new_i64();
714 TCGv_i64 tcg_mask = tcg_temp_new_i64();
716 tcg_gen_movi_i64(tcg_mask, mask);
717 for (j = 0; j < 2; j++) {
718 get_avr64(avr[j], VB, j);
719 tcg_gen_and_i64(result[j], avr[j], tcg_mask);
721 for (i = 1; i < 8; i++) {
722 tcg_gen_movi_i64(tcg_mask, mask >> (i * 8));
723 for (j = 0; j < 2; j++) {
724 tcg_gen_shri_i64(tmp, avr[j], i * 7);
725 tcg_gen_and_i64(tmp, tmp, tcg_mask);
726 tcg_gen_or_i64(result[j], result[j], tmp);
729 for (i = 1; i < 8; i++) {
730 tcg_gen_movi_i64(tcg_mask, mask << (i * 8));
731 for (j = 0; j < 2; j++) {
732 tcg_gen_shli_i64(tmp, avr[j], i * 7);
733 tcg_gen_and_i64(tmp, tmp, tcg_mask);
734 tcg_gen_or_i64(result[j], result[j], tmp);
737 for (j = 0; j < 2; j++) {
738 set_avr64(VT, result[j], j);
741 tcg_temp_free_i64(tmp);
742 tcg_temp_free_i64(tcg_mask);
743 tcg_temp_free_i64(result[0]);
744 tcg_temp_free_i64(result[1]);
745 tcg_temp_free_i64(avr[0]);
746 tcg_temp_free_i64(avr[1]);
750 * vclzw VRT,VRB - Vector Count Leading Zeros Word
752 * Counting the number of leading zero bits of each word element in source
753 * register and placing result in appropriate word element of destination
756 static void trans_vclzw(DisasContext *ctx)
758 int VT = rD(ctx->opcode);
759 int VB = rB(ctx->opcode);
760 TCGv_i32 tmp = tcg_temp_new_i32();
763 /* Perform count for every word element using tcg_gen_clzi_i32. */
764 for (i = 0; i < 4; i++) {
765 tcg_gen_ld_i32(tmp, cpu_env,
766 offsetof(CPUPPCState, vsr[32 + VB].u64[0]) + i * 4);
767 tcg_gen_clzi_i32(tmp, tmp, 32);
768 tcg_gen_st_i32(tmp, cpu_env,
769 offsetof(CPUPPCState, vsr[32 + VT].u64[0]) + i * 4);
772 tcg_temp_free_i32(tmp);
776 * vclzd VRT,VRB - Vector Count Leading Zeros Doubleword
778 * Counting the number of leading zero bits of each doubleword element in source
779 * register and placing result in appropriate doubleword element of destination
782 static void trans_vclzd(DisasContext *ctx)
784 int VT = rD(ctx->opcode);
785 int VB = rB(ctx->opcode);
786 TCGv_i64 avr = tcg_temp_new_i64();
788 /* high doubleword */
789 get_avr64(avr, VB, true);
790 tcg_gen_clzi_i64(avr, avr, 64);
791 set_avr64(VT, avr, true);
794 get_avr64(avr, VB, false);
795 tcg_gen_clzi_i64(avr, avr, 64);
796 set_avr64(VT, avr, false);
798 tcg_temp_free_i64(avr);
801 GEN_VXFORM_V(vmuluwm, MO_32, tcg_gen_gvec_mul, 4, 2);
802 GEN_VXFORM(vsrv, 2, 28);
803 GEN_VXFORM(vslv, 2, 29);
804 GEN_VXFORM(vslo, 6, 16);
805 GEN_VXFORM(vsro, 6, 17);
807 static bool do_vector_gvec3_VX(DisasContext *ctx, arg_VX *a, int vece,
808 void (*gen_gvec)(unsigned, uint32_t, uint32_t,
809 uint32_t, uint32_t, uint32_t))
813 gen_gvec(vece, avr_full_offset(a->vrt), avr_full_offset(a->vra),
814 avr_full_offset(a->vrb), 16, 16);
819 TRANS_FLAGS(ALTIVEC, VSLB, do_vector_gvec3_VX, MO_8, tcg_gen_gvec_shlv);
820 TRANS_FLAGS(ALTIVEC, VSLH, do_vector_gvec3_VX, MO_16, tcg_gen_gvec_shlv);
821 TRANS_FLAGS(ALTIVEC, VSLW, do_vector_gvec3_VX, MO_32, tcg_gen_gvec_shlv);
822 TRANS_FLAGS2(ALTIVEC_207, VSLD, do_vector_gvec3_VX, MO_64, tcg_gen_gvec_shlv);
824 TRANS_FLAGS(ALTIVEC, VSRB, do_vector_gvec3_VX, MO_8, tcg_gen_gvec_shrv);
825 TRANS_FLAGS(ALTIVEC, VSRH, do_vector_gvec3_VX, MO_16, tcg_gen_gvec_shrv);
826 TRANS_FLAGS(ALTIVEC, VSRW, do_vector_gvec3_VX, MO_32, tcg_gen_gvec_shrv);
827 TRANS_FLAGS2(ALTIVEC_207, VSRD, do_vector_gvec3_VX, MO_64, tcg_gen_gvec_shrv);
829 TRANS_FLAGS(ALTIVEC, VSRAB, do_vector_gvec3_VX, MO_8, tcg_gen_gvec_sarv);
830 TRANS_FLAGS(ALTIVEC, VSRAH, do_vector_gvec3_VX, MO_16, tcg_gen_gvec_sarv);
831 TRANS_FLAGS(ALTIVEC, VSRAW, do_vector_gvec3_VX, MO_32, tcg_gen_gvec_sarv);
832 TRANS_FLAGS2(ALTIVEC_207, VSRAD, do_vector_gvec3_VX, MO_64, tcg_gen_gvec_sarv);
834 TRANS_FLAGS(ALTIVEC, VRLB, do_vector_gvec3_VX, MO_8, tcg_gen_gvec_rotlv)
835 TRANS_FLAGS(ALTIVEC, VRLH, do_vector_gvec3_VX, MO_16, tcg_gen_gvec_rotlv)
836 TRANS_FLAGS(ALTIVEC, VRLW, do_vector_gvec3_VX, MO_32, tcg_gen_gvec_rotlv)
837 TRANS_FLAGS2(ALTIVEC_207, VRLD, do_vector_gvec3_VX, MO_64, tcg_gen_gvec_rotlv)
839 static TCGv_vec do_vrl_mask_vec(unsigned vece, TCGv_vec vrb)
841 TCGv_vec t0 = tcg_temp_new_vec_matching(vrb),
842 t1 = tcg_temp_new_vec_matching(vrb),
843 t2 = tcg_temp_new_vec_matching(vrb),
844 ones = tcg_constant_vec_matching(vrb, vece, -1);
846 /* Extract b and e */
847 tcg_gen_dupi_vec(vece, t2, (8 << vece) - 1);
849 tcg_gen_shri_vec(vece, t0, vrb, 16);
850 tcg_gen_and_vec(vece, t0, t0, t2);
852 tcg_gen_shri_vec(vece, t1, vrb, 8);
853 tcg_gen_and_vec(vece, t1, t1, t2);
855 /* Compare b and e to negate the mask where begin > end */
856 tcg_gen_cmp_vec(TCG_COND_GT, vece, t2, t0, t1);
858 /* Create the mask with (~0 >> b) ^ ((~0 >> e) >> 1) */
859 tcg_gen_shrv_vec(vece, t0, ones, t0);
860 tcg_gen_shrv_vec(vece, t1, ones, t1);
861 tcg_gen_shri_vec(vece, t1, t1, 1);
862 tcg_gen_xor_vec(vece, t0, t0, t1);
864 /* negate the mask */
865 tcg_gen_xor_vec(vece, t0, t0, t2);
867 tcg_temp_free_vec(t1);
868 tcg_temp_free_vec(t2);
873 static void gen_vrlnm_vec(unsigned vece, TCGv_vec vrt, TCGv_vec vra,
876 TCGv_vec mask, n = tcg_temp_new_vec_matching(vrt);
878 /* Create the mask */
879 mask = do_vrl_mask_vec(vece, vrb);
882 tcg_gen_dupi_vec(vece, n, (8 << vece) - 1);
883 tcg_gen_and_vec(vece, n, vrb, n);
885 /* Rotate and mask */
886 tcg_gen_rotlv_vec(vece, vrt, vra, n);
887 tcg_gen_and_vec(vece, vrt, vrt, mask);
889 tcg_temp_free_vec(n);
890 tcg_temp_free_vec(mask);
893 static bool do_vrlnm(DisasContext *ctx, arg_VX *a, int vece)
895 static const TCGOpcode vecop_list[] = {
896 INDEX_op_cmp_vec, INDEX_op_rotlv_vec, INDEX_op_sari_vec,
897 INDEX_op_shli_vec, INDEX_op_shri_vec, INDEX_op_shrv_vec, 0
899 static const GVecGen3 ops[2] = {
901 .fniv = gen_vrlnm_vec,
902 .fno = gen_helper_VRLWNM,
903 .opt_opc = vecop_list,
908 .fniv = gen_vrlnm_vec,
909 .fno = gen_helper_VRLDNM,
910 .opt_opc = vecop_list,
916 REQUIRE_INSNS_FLAGS2(ctx, ISA300);
919 tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra),
920 avr_full_offset(a->vrb), 16, 16, &ops[vece - 2]);
925 TRANS(VRLWNM, do_vrlnm, MO_32)
926 TRANS(VRLDNM, do_vrlnm, MO_64)
928 static void gen_vrlmi_vec(unsigned vece, TCGv_vec vrt, TCGv_vec vra,
931 TCGv_vec mask, n = tcg_temp_new_vec_matching(vrt),
932 tmp = tcg_temp_new_vec_matching(vrt);
934 /* Create the mask */
935 mask = do_vrl_mask_vec(vece, vrb);
938 tcg_gen_dupi_vec(vece, n, (8 << vece) - 1);
939 tcg_gen_and_vec(vece, n, vrb, n);
941 /* Rotate and insert */
942 tcg_gen_rotlv_vec(vece, tmp, vra, n);
943 tcg_gen_bitsel_vec(vece, vrt, mask, tmp, vrt);
945 tcg_temp_free_vec(n);
946 tcg_temp_free_vec(tmp);
947 tcg_temp_free_vec(mask);
950 static bool do_vrlmi(DisasContext *ctx, arg_VX *a, int vece)
952 static const TCGOpcode vecop_list[] = {
953 INDEX_op_cmp_vec, INDEX_op_rotlv_vec, INDEX_op_sari_vec,
954 INDEX_op_shli_vec, INDEX_op_shri_vec, INDEX_op_shrv_vec, 0
956 static const GVecGen3 ops[2] = {
958 .fniv = gen_vrlmi_vec,
959 .fno = gen_helper_VRLWMI,
960 .opt_opc = vecop_list,
965 .fniv = gen_vrlnm_vec,
966 .fno = gen_helper_VRLDMI,
967 .opt_opc = vecop_list,
973 REQUIRE_INSNS_FLAGS2(ctx, ISA300);
976 tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra),
977 avr_full_offset(a->vrb), 16, 16, &ops[vece - 2]);
982 TRANS(VRLWMI, do_vrlmi, MO_32)
983 TRANS(VRLDMI, do_vrlmi, MO_64)
985 static bool do_vector_shift_quad(DisasContext *ctx, arg_VX *a, bool right,
988 TCGv_i64 hi, lo, t0, t1, n, zero = tcg_constant_i64(0);
992 n = tcg_temp_new_i64();
993 hi = tcg_temp_new_i64();
994 lo = tcg_temp_new_i64();
995 t0 = tcg_temp_new_i64();
996 t1 = tcg_const_i64(0);
998 get_avr64(lo, a->vra, false);
999 get_avr64(hi, a->vra, true);
1001 get_avr64(n, a->vrb, true);
1003 tcg_gen_andi_i64(t0, n, 64);
1005 tcg_gen_movcond_i64(TCG_COND_NE, lo, t0, zero, hi, lo);
1007 tcg_gen_sari_i64(t1, lo, 63);
1009 tcg_gen_movcond_i64(TCG_COND_NE, hi, t0, zero, t1, hi);
1011 tcg_gen_movcond_i64(TCG_COND_NE, hi, t0, zero, lo, hi);
1012 tcg_gen_movcond_i64(TCG_COND_NE, lo, t0, zero, zero, lo);
1014 tcg_gen_andi_i64(n, n, 0x3F);
1018 tcg_gen_sar_i64(t0, hi, n);
1020 tcg_gen_shr_i64(t0, hi, n);
1023 tcg_gen_shl_i64(t0, lo, n);
1025 set_avr64(a->vrt, t0, right);
1028 tcg_gen_shr_i64(lo, lo, n);
1030 tcg_gen_shl_i64(hi, hi, n);
1032 tcg_gen_xori_i64(n, n, 63);
1034 tcg_gen_shl_i64(hi, hi, n);
1035 tcg_gen_shli_i64(hi, hi, 1);
1037 tcg_gen_shr_i64(lo, lo, n);
1038 tcg_gen_shri_i64(lo, lo, 1);
1040 tcg_gen_or_i64(hi, hi, lo);
1041 set_avr64(a->vrt, hi, !right);
1043 tcg_temp_free_i64(hi);
1044 tcg_temp_free_i64(lo);
1045 tcg_temp_free_i64(t0);
1046 tcg_temp_free_i64(t1);
1047 tcg_temp_free_i64(n);
1052 TRANS_FLAGS2(ISA310, VSLQ, do_vector_shift_quad, false, false);
1053 TRANS_FLAGS2(ISA310, VSRQ, do_vector_shift_quad, true, false);
1054 TRANS_FLAGS2(ISA310, VSRAQ, do_vector_shift_quad, true, true);
1056 static void do_vrlq_mask(TCGv_i64 mh, TCGv_i64 ml, TCGv_i64 b, TCGv_i64 e)
1058 TCGv_i64 th, tl, t0, t1, zero = tcg_constant_i64(0),
1059 ones = tcg_constant_i64(-1);
1061 th = tcg_temp_new_i64();
1062 tl = tcg_temp_new_i64();
1063 t0 = tcg_temp_new_i64();
1064 t1 = tcg_temp_new_i64();
1067 tcg_gen_andi_i64(t0, b, 64);
1068 tcg_gen_movcond_i64(TCG_COND_NE, t1, t0, zero, zero, ones);
1069 tcg_gen_andi_i64(t0, b, 0x3F);
1070 tcg_gen_shr_i64(mh, t1, t0);
1071 tcg_gen_shr_i64(ml, ones, t0);
1072 tcg_gen_xori_i64(t0, t0, 63);
1073 tcg_gen_shl_i64(t1, t1, t0);
1074 tcg_gen_shli_i64(t1, t1, 1);
1075 tcg_gen_or_i64(ml, t1, ml);
1078 tcg_gen_andi_i64(t0, e, 64);
1079 tcg_gen_movcond_i64(TCG_COND_NE, t1, t0, zero, zero, ones);
1080 tcg_gen_andi_i64(t0, e, 0x3F);
1081 tcg_gen_shr_i64(th, t1, t0);
1082 tcg_gen_shr_i64(tl, ones, t0);
1083 tcg_gen_xori_i64(t0, t0, 63);
1084 tcg_gen_shl_i64(t1, t1, t0);
1085 tcg_gen_shli_i64(t1, t1, 1);
1086 tcg_gen_or_i64(tl, t1, tl);
1089 tcg_gen_extract2_i64(tl, tl, th, 1);
1090 tcg_gen_shri_i64(th, th, 1);
1093 tcg_gen_xor_i64(mh, mh, th);
1094 tcg_gen_xor_i64(ml, ml, tl);
1096 /* Negate the mask if begin > end */
1097 tcg_gen_movcond_i64(TCG_COND_GT, t0, b, e, ones, zero);
1099 tcg_gen_xor_i64(mh, mh, t0);
1100 tcg_gen_xor_i64(ml, ml, t0);
1102 tcg_temp_free_i64(th);
1103 tcg_temp_free_i64(tl);
1104 tcg_temp_free_i64(t0);
1105 tcg_temp_free_i64(t1);
1108 static bool do_vector_rotl_quad(DisasContext *ctx, arg_VX *a, bool mask,
1111 TCGv_i64 ah, al, vrb, n, t0, t1, zero = tcg_constant_i64(0);
1113 REQUIRE_VECTOR(ctx);
1114 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
1116 ah = tcg_temp_new_i64();
1117 al = tcg_temp_new_i64();
1118 vrb = tcg_temp_new_i64();
1119 n = tcg_temp_new_i64();
1120 t0 = tcg_temp_new_i64();
1121 t1 = tcg_temp_new_i64();
1123 get_avr64(ah, a->vra, true);
1124 get_avr64(al, a->vra, false);
1125 get_avr64(vrb, a->vrb, true);
1127 tcg_gen_mov_i64(t0, ah);
1128 tcg_gen_andi_i64(t1, vrb, 64);
1129 tcg_gen_movcond_i64(TCG_COND_NE, ah, t1, zero, al, ah);
1130 tcg_gen_movcond_i64(TCG_COND_NE, al, t1, zero, t0, al);
1131 tcg_gen_andi_i64(n, vrb, 0x3F);
1133 tcg_gen_shl_i64(t0, ah, n);
1134 tcg_gen_shl_i64(t1, al, n);
1136 tcg_gen_xori_i64(n, n, 63);
1138 tcg_gen_shr_i64(al, al, n);
1139 tcg_gen_shri_i64(al, al, 1);
1140 tcg_gen_or_i64(t0, al, t0);
1142 tcg_gen_shr_i64(ah, ah, n);
1143 tcg_gen_shri_i64(ah, ah, 1);
1144 tcg_gen_or_i64(t1, ah, t1);
1146 if (mask || insert) {
1147 tcg_gen_extract_i64(n, vrb, 8, 7);
1148 tcg_gen_extract_i64(vrb, vrb, 16, 7);
1150 do_vrlq_mask(ah, al, vrb, n);
1152 tcg_gen_and_i64(t0, t0, ah);
1153 tcg_gen_and_i64(t1, t1, al);
1156 get_avr64(n, a->vrt, true);
1157 get_avr64(vrb, a->vrt, false);
1158 tcg_gen_andc_i64(n, n, ah);
1159 tcg_gen_andc_i64(vrb, vrb, al);
1160 tcg_gen_or_i64(t0, t0, n);
1161 tcg_gen_or_i64(t1, t1, vrb);
1165 set_avr64(a->vrt, t0, true);
1166 set_avr64(a->vrt, t1, false);
1168 tcg_temp_free_i64(ah);
1169 tcg_temp_free_i64(al);
1170 tcg_temp_free_i64(vrb);
1171 tcg_temp_free_i64(n);
1172 tcg_temp_free_i64(t0);
1173 tcg_temp_free_i64(t1);
1178 TRANS(VRLQ, do_vector_rotl_quad, false, false)
1179 TRANS(VRLQNM, do_vector_rotl_quad, true, false)
1180 TRANS(VRLQMI, do_vector_rotl_quad, false, true)
1182 #define GEN_VXFORM_SAT(NAME, VECE, NORM, SAT, OPC2, OPC3) \
1183 static void glue(glue(gen_, NAME), _vec)(unsigned vece, TCGv_vec t, \
1184 TCGv_vec sat, TCGv_vec a, \
1187 TCGv_vec x = tcg_temp_new_vec_matching(t); \
1188 glue(glue(tcg_gen_, NORM), _vec)(VECE, x, a, b); \
1189 glue(glue(tcg_gen_, SAT), _vec)(VECE, t, a, b); \
1190 tcg_gen_cmp_vec(TCG_COND_NE, VECE, x, x, t); \
1191 tcg_gen_or_vec(VECE, sat, sat, x); \
1192 tcg_temp_free_vec(x); \
1194 static void glue(gen_, NAME)(DisasContext *ctx) \
1196 static const TCGOpcode vecop_list[] = { \
1197 glue(glue(INDEX_op_, NORM), _vec), \
1198 glue(glue(INDEX_op_, SAT), _vec), \
1199 INDEX_op_cmp_vec, 0 \
1201 static const GVecGen4 g = { \
1202 .fniv = glue(glue(gen_, NAME), _vec), \
1203 .fno = glue(gen_helper_, NAME), \
1204 .opt_opc = vecop_list, \
1205 .write_aofs = true, \
1208 if (unlikely(!ctx->altivec_enabled)) { \
1209 gen_exception(ctx, POWERPC_EXCP_VPU); \
1212 tcg_gen_gvec_4(avr_full_offset(rD(ctx->opcode)), \
1213 offsetof(CPUPPCState, vscr_sat), \
1214 avr_full_offset(rA(ctx->opcode)), \
1215 avr_full_offset(rB(ctx->opcode)), \
1219 GEN_VXFORM_SAT(vaddubs, MO_8, add, usadd, 0, 8);
1220 GEN_VXFORM_DUAL_EXT(vaddubs, PPC_ALTIVEC, PPC_NONE, 0, \
1221 vmul10uq, PPC_NONE, PPC2_ISA300, 0x0000F800)
1222 GEN_VXFORM_SAT(vadduhs, MO_16, add, usadd, 0, 9);
1223 GEN_VXFORM_DUAL(vadduhs, PPC_ALTIVEC, PPC_NONE, \
1224 vmul10euq, PPC_NONE, PPC2_ISA300)
1225 GEN_VXFORM_SAT(vadduws, MO_32, add, usadd, 0, 10);
1226 GEN_VXFORM_SAT(vaddsbs, MO_8, add, ssadd, 0, 12);
1227 GEN_VXFORM_SAT(vaddshs, MO_16, add, ssadd, 0, 13);
1228 GEN_VXFORM_SAT(vaddsws, MO_32, add, ssadd, 0, 14);
1229 GEN_VXFORM_SAT(vsububs, MO_8, sub, ussub, 0, 24);
1230 GEN_VXFORM_SAT(vsubuhs, MO_16, sub, ussub, 0, 25);
1231 GEN_VXFORM_SAT(vsubuws, MO_32, sub, ussub, 0, 26);
1232 GEN_VXFORM_SAT(vsubsbs, MO_8, sub, sssub, 0, 28);
1233 GEN_VXFORM_SAT(vsubshs, MO_16, sub, sssub, 0, 29);
1234 GEN_VXFORM_SAT(vsubsws, MO_32, sub, sssub, 0, 30);
1235 GEN_VXFORM_TRANS(vsl, 2, 7);
1236 GEN_VXFORM_TRANS(vsr, 2, 11);
1237 GEN_VXFORM_ENV(vpkuhum, 7, 0);
1238 GEN_VXFORM_ENV(vpkuwum, 7, 1);
1239 GEN_VXFORM_ENV(vpkudum, 7, 17);
1240 GEN_VXFORM_ENV(vpkuhus, 7, 2);
1241 GEN_VXFORM_ENV(vpkuwus, 7, 3);
1242 GEN_VXFORM_ENV(vpkudus, 7, 19);
1243 GEN_VXFORM_ENV(vpkshus, 7, 4);
1244 GEN_VXFORM_ENV(vpkswus, 7, 5);
1245 GEN_VXFORM_ENV(vpksdus, 7, 21);
1246 GEN_VXFORM_ENV(vpkshss, 7, 6);
1247 GEN_VXFORM_ENV(vpkswss, 7, 7);
1248 GEN_VXFORM_ENV(vpksdss, 7, 23);
1249 GEN_VXFORM(vpkpx, 7, 12);
1250 GEN_VXFORM_ENV(vsum4ubs, 4, 24);
1251 GEN_VXFORM_ENV(vsum4sbs, 4, 28);
1252 GEN_VXFORM_ENV(vsum4shs, 4, 25);
1253 GEN_VXFORM_ENV(vsum2sws, 4, 26);
1254 GEN_VXFORM_ENV(vsumsws, 4, 30);
1255 GEN_VXFORM_ENV(vaddfp, 5, 0);
1256 GEN_VXFORM_ENV(vsubfp, 5, 1);
1257 GEN_VXFORM_ENV(vmaxfp, 5, 16);
1258 GEN_VXFORM_ENV(vminfp, 5, 17);
1259 GEN_VXFORM_HETRO(vextublx, 6, 24)
1260 GEN_VXFORM_HETRO(vextuhlx, 6, 25)
1261 GEN_VXFORM_HETRO(vextuwlx, 6, 26)
1262 GEN_VXFORM_TRANS_DUAL(vmrgow, PPC_NONE, PPC2_ALTIVEC_207,
1263 vextuwlx, PPC_NONE, PPC2_ISA300)
1264 GEN_VXFORM_HETRO(vextubrx, 6, 28)
1265 GEN_VXFORM_HETRO(vextuhrx, 6, 29)
1266 GEN_VXFORM_HETRO(vextuwrx, 6, 30)
1267 GEN_VXFORM_TRANS(lvsl, 6, 31)
1268 GEN_VXFORM_TRANS(lvsr, 6, 32)
1269 GEN_VXFORM_TRANS_DUAL(vmrgew, PPC_NONE, PPC2_ALTIVEC_207,
1270 vextuwrx, PPC_NONE, PPC2_ISA300)
1272 #define GEN_VXRFORM1(opname, name, str, opc2, opc3) \
1273 static void glue(gen_, name)(DisasContext *ctx) \
1275 TCGv_ptr ra, rb, rd; \
1276 if (unlikely(!ctx->altivec_enabled)) { \
1277 gen_exception(ctx, POWERPC_EXCP_VPU); \
1280 ra = gen_avr_ptr(rA(ctx->opcode)); \
1281 rb = gen_avr_ptr(rB(ctx->opcode)); \
1282 rd = gen_avr_ptr(rD(ctx->opcode)); \
1283 gen_helper_##opname(cpu_env, rd, ra, rb); \
1284 tcg_temp_free_ptr(ra); \
1285 tcg_temp_free_ptr(rb); \
1286 tcg_temp_free_ptr(rd); \
1289 #define GEN_VXRFORM(name, opc2, opc3) \
1290 GEN_VXRFORM1(name, name, #name, opc2, opc3) \
1291 GEN_VXRFORM1(name##_dot, name##_, #name ".", opc2, (opc3 | (0x1 << 4)))
1294 * Support for Altivec instructions that use bit 31 (Rc) as an opcode
1295 * bit but also use bit 21 as an actual Rc bit. In general, thse pairs
1296 * come from different versions of the ISA, so we must also support a
1297 * pair of flags for each instruction.
1299 #define GEN_VXRFORM_DUAL(name0, flg0, flg2_0, name1, flg1, flg2_1) \
1300 static void glue(gen_, name0##_##name1)(DisasContext *ctx) \
1302 if ((Rc(ctx->opcode) == 0) && \
1303 ((ctx->insns_flags & flg0) || (ctx->insns_flags2 & flg2_0))) { \
1304 if (Rc21(ctx->opcode) == 0) { \
1307 gen_##name0##_(ctx); \
1309 } else if ((Rc(ctx->opcode) == 1) && \
1310 ((ctx->insns_flags & flg1) || (ctx->insns_flags2 & flg2_1))) { \
1311 if (Rc21(ctx->opcode) == 0) { \
1314 gen_##name1##_(ctx); \
1317 gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL); \
1321 static void do_vcmp_rc(int vrt)
1323 TCGv_i64 tmp, set, clr;
1325 tmp = tcg_temp_new_i64();
1326 set = tcg_temp_new_i64();
1327 clr = tcg_temp_new_i64();
1329 get_avr64(tmp, vrt, true);
1330 tcg_gen_mov_i64(set, tmp);
1331 get_avr64(tmp, vrt, false);
1332 tcg_gen_or_i64(clr, set, tmp);
1333 tcg_gen_and_i64(set, set, tmp);
1335 tcg_gen_setcondi_i64(TCG_COND_EQ, clr, clr, 0);
1336 tcg_gen_shli_i64(clr, clr, 1);
1338 tcg_gen_setcondi_i64(TCG_COND_EQ, set, set, -1);
1339 tcg_gen_shli_i64(set, set, 3);
1341 tcg_gen_or_i64(tmp, set, clr);
1342 tcg_gen_extrl_i64_i32(cpu_crf[6], tmp);
1344 tcg_temp_free_i64(tmp);
1345 tcg_temp_free_i64(set);
1346 tcg_temp_free_i64(clr);
1349 static bool do_vcmp(DisasContext *ctx, arg_VC *a, TCGCond cond, int vece)
1351 REQUIRE_VECTOR(ctx);
1353 tcg_gen_gvec_cmp(cond, vece, avr_full_offset(a->vrt),
1354 avr_full_offset(a->vra), avr_full_offset(a->vrb), 16, 16);
1363 TRANS_FLAGS(ALTIVEC, VCMPEQUB, do_vcmp, TCG_COND_EQ, MO_8)
1364 TRANS_FLAGS(ALTIVEC, VCMPEQUH, do_vcmp, TCG_COND_EQ, MO_16)
1365 TRANS_FLAGS(ALTIVEC, VCMPEQUW, do_vcmp, TCG_COND_EQ, MO_32)
1366 TRANS_FLAGS2(ALTIVEC_207, VCMPEQUD, do_vcmp, TCG_COND_EQ, MO_64)
1368 TRANS_FLAGS(ALTIVEC, VCMPGTSB, do_vcmp, TCG_COND_GT, MO_8)
1369 TRANS_FLAGS(ALTIVEC, VCMPGTSH, do_vcmp, TCG_COND_GT, MO_16)
1370 TRANS_FLAGS(ALTIVEC, VCMPGTSW, do_vcmp, TCG_COND_GT, MO_32)
1371 TRANS_FLAGS2(ALTIVEC_207, VCMPGTSD, do_vcmp, TCG_COND_GT, MO_64)
1372 TRANS_FLAGS(ALTIVEC, VCMPGTUB, do_vcmp, TCG_COND_GTU, MO_8)
1373 TRANS_FLAGS(ALTIVEC, VCMPGTUH, do_vcmp, TCG_COND_GTU, MO_16)
1374 TRANS_FLAGS(ALTIVEC, VCMPGTUW, do_vcmp, TCG_COND_GTU, MO_32)
1375 TRANS_FLAGS2(ALTIVEC_207, VCMPGTUD, do_vcmp, TCG_COND_GTU, MO_64)
1377 TRANS_FLAGS2(ISA300, VCMPNEB, do_vcmp, TCG_COND_NE, MO_8)
1378 TRANS_FLAGS2(ISA300, VCMPNEH, do_vcmp, TCG_COND_NE, MO_16)
1379 TRANS_FLAGS2(ISA300, VCMPNEW, do_vcmp, TCG_COND_NE, MO_32)
1381 static void gen_vcmpnez_vec(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b)
1383 TCGv_vec t0, t1, zero;
1385 t0 = tcg_temp_new_vec_matching(t);
1386 t1 = tcg_temp_new_vec_matching(t);
1387 zero = tcg_constant_vec_matching(t, vece, 0);
1389 tcg_gen_cmp_vec(TCG_COND_EQ, vece, t0, a, zero);
1390 tcg_gen_cmp_vec(TCG_COND_EQ, vece, t1, b, zero);
1391 tcg_gen_cmp_vec(TCG_COND_NE, vece, t, a, b);
1393 tcg_gen_or_vec(vece, t, t, t0);
1394 tcg_gen_or_vec(vece, t, t, t1);
1396 tcg_temp_free_vec(t0);
1397 tcg_temp_free_vec(t1);
1400 static bool do_vcmpnez(DisasContext *ctx, arg_VC *a, int vece)
1402 static const TCGOpcode vecop_list[] = {
1405 static const GVecGen3 ops[3] = {
1407 .fniv = gen_vcmpnez_vec,
1408 .fno = gen_helper_VCMPNEZB,
1409 .opt_opc = vecop_list,
1413 .fniv = gen_vcmpnez_vec,
1414 .fno = gen_helper_VCMPNEZH,
1415 .opt_opc = vecop_list,
1419 .fniv = gen_vcmpnez_vec,
1420 .fno = gen_helper_VCMPNEZW,
1421 .opt_opc = vecop_list,
1426 REQUIRE_INSNS_FLAGS2(ctx, ISA300);
1427 REQUIRE_VECTOR(ctx);
1429 tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra),
1430 avr_full_offset(a->vrb), 16, 16, &ops[vece]);
1439 TRANS(VCMPNEZB, do_vcmpnez, MO_8)
1440 TRANS(VCMPNEZH, do_vcmpnez, MO_16)
1441 TRANS(VCMPNEZW, do_vcmpnez, MO_32)
1443 static bool trans_VCMPEQUQ(DisasContext *ctx, arg_VC *a)
1445 TCGv_i64 t0, t1, t2;
1447 t0 = tcg_temp_new_i64();
1448 t1 = tcg_temp_new_i64();
1449 t2 = tcg_temp_new_i64();
1451 get_avr64(t0, a->vra, true);
1452 get_avr64(t1, a->vrb, true);
1453 tcg_gen_xor_i64(t2, t0, t1);
1455 get_avr64(t0, a->vra, false);
1456 get_avr64(t1, a->vrb, false);
1457 tcg_gen_xor_i64(t1, t0, t1);
1459 tcg_gen_or_i64(t1, t1, t2);
1460 tcg_gen_setcondi_i64(TCG_COND_EQ, t1, t1, 0);
1461 tcg_gen_neg_i64(t1, t1);
1463 set_avr64(a->vrt, t1, true);
1464 set_avr64(a->vrt, t1, false);
1467 tcg_gen_extrl_i64_i32(cpu_crf[6], t1);
1468 tcg_gen_andi_i32(cpu_crf[6], cpu_crf[6], 0xa);
1469 tcg_gen_xori_i32(cpu_crf[6], cpu_crf[6], 0x2);
1472 tcg_temp_free_i64(t0);
1473 tcg_temp_free_i64(t1);
1474 tcg_temp_free_i64(t2);
1479 static bool do_vcmpgtq(DisasContext *ctx, arg_VC *a, bool sign)
1481 TCGv_i64 t0, t1, t2;
1483 t0 = tcg_temp_new_i64();
1484 t1 = tcg_temp_new_i64();
1485 t2 = tcg_temp_new_i64();
1487 get_avr64(t0, a->vra, false);
1488 get_avr64(t1, a->vrb, false);
1489 tcg_gen_setcond_i64(TCG_COND_GTU, t2, t0, t1);
1491 get_avr64(t0, a->vra, true);
1492 get_avr64(t1, a->vrb, true);
1493 tcg_gen_movcond_i64(TCG_COND_EQ, t2, t0, t1, t2, tcg_constant_i64(0));
1494 tcg_gen_setcond_i64(sign ? TCG_COND_GT : TCG_COND_GTU, t1, t0, t1);
1496 tcg_gen_or_i64(t1, t1, t2);
1497 tcg_gen_neg_i64(t1, t1);
1499 set_avr64(a->vrt, t1, true);
1500 set_avr64(a->vrt, t1, false);
1503 tcg_gen_extrl_i64_i32(cpu_crf[6], t1);
1504 tcg_gen_andi_i32(cpu_crf[6], cpu_crf[6], 0xa);
1505 tcg_gen_xori_i32(cpu_crf[6], cpu_crf[6], 0x2);
1508 tcg_temp_free_i64(t0);
1509 tcg_temp_free_i64(t1);
1510 tcg_temp_free_i64(t2);
1515 TRANS(VCMPGTSQ, do_vcmpgtq, true)
1516 TRANS(VCMPGTUQ, do_vcmpgtq, false)
1518 static bool do_vcmpq(DisasContext *ctx, arg_VX_bf *a, bool sign)
1521 TCGLabel *gt, *lt, *done;
1523 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
1524 REQUIRE_VECTOR(ctx);
1526 vra = tcg_temp_local_new_i64();
1527 vrb = tcg_temp_local_new_i64();
1528 gt = gen_new_label();
1529 lt = gen_new_label();
1530 done = gen_new_label();
1532 get_avr64(vra, a->vra, true);
1533 get_avr64(vrb, a->vrb, true);
1534 tcg_gen_brcond_i64((sign ? TCG_COND_GT : TCG_COND_GTU), vra, vrb, gt);
1535 tcg_gen_brcond_i64((sign ? TCG_COND_LT : TCG_COND_LTU), vra, vrb, lt);
1537 get_avr64(vra, a->vra, false);
1538 get_avr64(vrb, a->vrb, false);
1539 tcg_gen_brcond_i64(TCG_COND_GTU, vra, vrb, gt);
1540 tcg_gen_brcond_i64(TCG_COND_LTU, vra, vrb, lt);
1542 tcg_gen_movi_i32(cpu_crf[a->bf], CRF_EQ);
1546 tcg_gen_movi_i32(cpu_crf[a->bf], CRF_GT);
1550 tcg_gen_movi_i32(cpu_crf[a->bf], CRF_LT);
1553 gen_set_label(done);
1554 tcg_temp_free_i64(vra);
1555 tcg_temp_free_i64(vrb);
1560 TRANS(VCMPSQ, do_vcmpq, true)
1561 TRANS(VCMPUQ, do_vcmpq, false)
1563 GEN_VXRFORM(vcmpeqfp, 3, 3)
1564 GEN_VXRFORM(vcmpgefp, 3, 7)
1565 GEN_VXRFORM(vcmpgtfp, 3, 11)
1566 GEN_VXRFORM(vcmpbfp, 3, 15)
1568 static void gen_vsplti(DisasContext *ctx, int vece)
1572 if (unlikely(!ctx->altivec_enabled)) {
1573 gen_exception(ctx, POWERPC_EXCP_VPU);
1577 simm = SIMM5(ctx->opcode);
1578 tcg_gen_gvec_dup_imm(vece, avr_full_offset(rD(ctx->opcode)), 16, 16, simm);
1581 #define GEN_VXFORM_VSPLTI(name, vece, opc2, opc3) \
1582 static void glue(gen_, name)(DisasContext *ctx) { gen_vsplti(ctx, vece); }
1584 GEN_VXFORM_VSPLTI(vspltisb, MO_8, 6, 12);
1585 GEN_VXFORM_VSPLTI(vspltish, MO_16, 6, 13);
1586 GEN_VXFORM_VSPLTI(vspltisw, MO_32, 6, 14);
1588 #define GEN_VXFORM_NOA(name, opc2, opc3) \
1589 static void glue(gen_, name)(DisasContext *ctx) \
1592 if (unlikely(!ctx->altivec_enabled)) { \
1593 gen_exception(ctx, POWERPC_EXCP_VPU); \
1596 rb = gen_avr_ptr(rB(ctx->opcode)); \
1597 rd = gen_avr_ptr(rD(ctx->opcode)); \
1598 gen_helper_##name(rd, rb); \
1599 tcg_temp_free_ptr(rb); \
1600 tcg_temp_free_ptr(rd); \
1603 #define GEN_VXFORM_NOA_ENV(name, opc2, opc3) \
1604 static void glue(gen_, name)(DisasContext *ctx) \
1608 if (unlikely(!ctx->altivec_enabled)) { \
1609 gen_exception(ctx, POWERPC_EXCP_VPU); \
1612 rb = gen_avr_ptr(rB(ctx->opcode)); \
1613 rd = gen_avr_ptr(rD(ctx->opcode)); \
1614 gen_helper_##name(cpu_env, rd, rb); \
1615 tcg_temp_free_ptr(rb); \
1616 tcg_temp_free_ptr(rd); \
1619 #define GEN_VXFORM_NOA_2(name, opc2, opc3, opc4) \
1620 static void glue(gen_, name)(DisasContext *ctx) \
1623 if (unlikely(!ctx->altivec_enabled)) { \
1624 gen_exception(ctx, POWERPC_EXCP_VPU); \
1627 rb = gen_avr_ptr(rB(ctx->opcode)); \
1628 rd = gen_avr_ptr(rD(ctx->opcode)); \
1629 gen_helper_##name(rd, rb); \
1630 tcg_temp_free_ptr(rb); \
1631 tcg_temp_free_ptr(rd); \
1634 #define GEN_VXFORM_NOA_3(name, opc2, opc3, opc4) \
1635 static void glue(gen_, name)(DisasContext *ctx) \
1638 if (unlikely(!ctx->altivec_enabled)) { \
1639 gen_exception(ctx, POWERPC_EXCP_VPU); \
1642 rb = gen_avr_ptr(rB(ctx->opcode)); \
1643 gen_helper_##name(cpu_gpr[rD(ctx->opcode)], rb); \
1644 tcg_temp_free_ptr(rb); \
1646 GEN_VXFORM_NOA(vupkhsb, 7, 8);
1647 GEN_VXFORM_NOA(vupkhsh, 7, 9);
1648 GEN_VXFORM_NOA(vupkhsw, 7, 25);
1649 GEN_VXFORM_NOA(vupklsb, 7, 10);
1650 GEN_VXFORM_NOA(vupklsh, 7, 11);
1651 GEN_VXFORM_NOA(vupklsw, 7, 27);
1652 GEN_VXFORM_NOA(vupkhpx, 7, 13);
1653 GEN_VXFORM_NOA(vupklpx, 7, 15);
1654 GEN_VXFORM_NOA_ENV(vrefp, 5, 4);
1655 GEN_VXFORM_NOA_ENV(vrsqrtefp, 5, 5);
1656 GEN_VXFORM_NOA_ENV(vexptefp, 5, 6);
1657 GEN_VXFORM_NOA_ENV(vlogefp, 5, 7);
1658 GEN_VXFORM_NOA_ENV(vrfim, 5, 11);
1659 GEN_VXFORM_NOA_ENV(vrfin, 5, 8);
1660 GEN_VXFORM_NOA_ENV(vrfip, 5, 10);
1661 GEN_VXFORM_NOA_ENV(vrfiz, 5, 9);
1662 GEN_VXFORM_NOA(vprtybw, 1, 24);
1663 GEN_VXFORM_NOA(vprtybd, 1, 24);
1664 GEN_VXFORM_NOA(vprtybq, 1, 24);
1666 static void gen_vsplt(DisasContext *ctx, int vece)
1668 int uimm, dofs, bofs;
1670 if (unlikely(!ctx->altivec_enabled)) {
1671 gen_exception(ctx, POWERPC_EXCP_VPU);
1675 uimm = UIMM5(ctx->opcode);
1676 bofs = avr_full_offset(rB(ctx->opcode));
1677 dofs = avr_full_offset(rD(ctx->opcode));
1679 /* Experimental testing shows that hardware masks the immediate. */
1680 bofs += (uimm << vece) & 15;
1681 #if !HOST_BIG_ENDIAN
1683 bofs &= ~((1 << vece) - 1);
1686 tcg_gen_gvec_dup_mem(vece, dofs, bofs, 16, 16);
1689 #define GEN_VXFORM_VSPLT(name, vece, opc2, opc3) \
1690 static void glue(gen_, name)(DisasContext *ctx) { gen_vsplt(ctx, vece); }
1692 #define GEN_VXFORM_UIMM_ENV(name, opc2, opc3) \
1693 static void glue(gen_, name)(DisasContext *ctx) \
1698 if (unlikely(!ctx->altivec_enabled)) { \
1699 gen_exception(ctx, POWERPC_EXCP_VPU); \
1702 uimm = tcg_const_i32(UIMM5(ctx->opcode)); \
1703 rb = gen_avr_ptr(rB(ctx->opcode)); \
1704 rd = gen_avr_ptr(rD(ctx->opcode)); \
1705 gen_helper_##name(cpu_env, rd, rb, uimm); \
1706 tcg_temp_free_i32(uimm); \
1707 tcg_temp_free_ptr(rb); \
1708 tcg_temp_free_ptr(rd); \
1711 #define GEN_VXFORM_UIMM_SPLAT(name, opc2, opc3, splat_max) \
1712 static void glue(gen_, name)(DisasContext *ctx) \
1715 uint8_t uimm = UIMM4(ctx->opcode); \
1717 if (unlikely(!ctx->altivec_enabled)) { \
1718 gen_exception(ctx, POWERPC_EXCP_VPU); \
1721 if (uimm > splat_max) { \
1724 t0 = tcg_temp_new_i32(); \
1725 tcg_gen_movi_i32(t0, uimm); \
1726 rb = gen_avr_ptr(rB(ctx->opcode)); \
1727 rd = gen_avr_ptr(rD(ctx->opcode)); \
1728 gen_helper_##name(rd, rb, t0); \
1729 tcg_temp_free_i32(t0); \
1730 tcg_temp_free_ptr(rb); \
1731 tcg_temp_free_ptr(rd); \
1734 GEN_VXFORM_VSPLT(vspltb, MO_8, 6, 8);
1735 GEN_VXFORM_VSPLT(vsplth, MO_16, 6, 9);
1736 GEN_VXFORM_VSPLT(vspltw, MO_32, 6, 10);
1737 GEN_VXFORM_UIMM_SPLAT(vextractub, 6, 8, 15);
1738 GEN_VXFORM_UIMM_SPLAT(vextractuh, 6, 9, 14);
1739 GEN_VXFORM_UIMM_SPLAT(vextractuw, 6, 10, 12);
1740 GEN_VXFORM_UIMM_SPLAT(vextractd, 6, 11, 8);
1741 GEN_VXFORM_UIMM_ENV(vcfux, 5, 12);
1742 GEN_VXFORM_UIMM_ENV(vcfsx, 5, 13);
1743 GEN_VXFORM_UIMM_ENV(vctuxs, 5, 14);
1744 GEN_VXFORM_UIMM_ENV(vctsxs, 5, 15);
1745 GEN_VXFORM_DUAL(vspltb, PPC_ALTIVEC, PPC_NONE,
1746 vextractub, PPC_NONE, PPC2_ISA300);
1747 GEN_VXFORM_DUAL(vsplth, PPC_ALTIVEC, PPC_NONE,
1748 vextractuh, PPC_NONE, PPC2_ISA300);
1749 GEN_VXFORM_DUAL(vspltw, PPC_ALTIVEC, PPC_NONE,
1750 vextractuw, PPC_NONE, PPC2_ISA300);
1752 static bool trans_VGNB(DisasContext *ctx, arg_VX_n *a)
1755 * Similar to do_vextractm, we'll use a sequence of mask-shift-or operations
1756 * to gather the bits. The masks can be created with
1758 * uint64_t mask(uint64_t n, uint64_t step)
1760 * uint64_t p = ((1UL << (1UL << step)) - 1UL) << ((n - 1UL) << step),
1761 * plen = n << step, m = 0;
1762 * for(int i = 0; i < 64/plen; i++) {
1764 * m = ror64(m, plen);
1766 * p >>= plen * DIV_ROUND_UP(64, plen) - 64;
1770 * But since there are few values of N, we'll use a lookup table to avoid
1771 * these calculations at runtime.
1773 static const uint64_t mask[6][5] = {
1775 0xAAAAAAAAAAAAAAAAULL, 0xccccccccccccccccULL, 0xf0f0f0f0f0f0f0f0ULL,
1776 0xff00ff00ff00ff00ULL, 0xffff0000ffff0000ULL
1779 0x9249249249249249ULL, 0xC30C30C30C30C30CULL, 0xF00F00F00F00F00FULL,
1780 0xFF0000FF0000FF00ULL, 0xFFFF00000000FFFFULL
1783 /* For N >= 4, some mask operations can be elided */
1784 0x8888888888888888ULL, 0, 0xf000f000f000f000ULL, 0,
1785 0xFFFF000000000000ULL
1788 0x8421084210842108ULL, 0, 0xF0000F0000F0000FULL, 0, 0
1791 0x8208208208208208ULL, 0, 0xF00000F00000F000ULL, 0, 0
1794 0x8102040810204081ULL, 0, 0xF000000F000000F0ULL, 0, 0
1798 int i, sh, nbits = DIV_ROUND_UP(64, a->n);
1799 TCGv_i64 hi, lo, t0, t1;
1801 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
1802 REQUIRE_VECTOR(ctx);
1806 * "N can be any value between 2 and 7, inclusive." Otherwise, the
1807 * result is undefined, so we don't need to change RT. Also, N > 7 is
1808 * impossible since the immediate field is 3 bits only.
1813 hi = tcg_temp_new_i64();
1814 lo = tcg_temp_new_i64();
1815 t0 = tcg_temp_new_i64();
1816 t1 = tcg_temp_new_i64();
1818 get_avr64(hi, a->vrb, true);
1819 get_avr64(lo, a->vrb, false);
1821 /* Align the lower doubleword so we can use the same mask */
1822 tcg_gen_shli_i64(lo, lo, a->n * nbits - 64);
1825 * Starting from the most significant bit, gather every Nth bit with a
1826 * sequence of mask-shift-or operation. E.g.: for N=3
1827 * AxxBxxCxxDxxExxFxxGxxHxxIxxJxxKxxLxxMxxNxxOxxPxxQxxRxxSxxTxxUxxV
1829 * A..B..C..D..E..F..G..H..I..J..K..L..M..N..O..P..Q..R..S..T..U..V
1831 * .B..C..D..E..F..G..H..I..J..K..L..M..N..O..P..Q..R..S..T..U..V..
1833 * AB.BC.CD.DE.EF.FG.GH.HI.IJ.JK.KL.LM.MN.NO.OP.PQ.QR.RS.ST.TU.UV.V
1835 * AB....CD....EF....GH....IJ....KL....MN....OP....QR....ST....UV..
1837 * ..CD....EF....GH....IJ....KL....MN....OP....QR....ST....UV......
1839 * ABCD..CDEF..EFGH..GHIJ..IJKL..KLMN..MNOP..OPQR..QRST..STUV..UV..
1840 * & rep(0b111100000000)
1841 * ABCD........EFGH........IJKL........MNOP........QRST........UV..
1843 * ....EFGH........IJKL........MNOP........QRST........UV..........
1845 * ABCDEFGH....EFGHIJKL....IJKLMNOP....MNOPQRST....QRSTUV......UV..
1846 * & rep(0b111111110000000000000000)
1847 * ABCDEFGH................IJKLMNOP................QRSTUV..........
1849 * ........IJKLMNOP................QRSTUV..........................
1851 * ABCDEFGHIJKLMNOP........IJKLMNOPQRSTUV..........QRSTUV..........
1852 * & rep(0b111111111111111100000000000000000000000000000000)
1853 * ABCDEFGHIJKLMNOP................................QRSTUV..........
1855 * ................QRSTUV..........................................
1857 * ABCDEFGHIJKLMNOPQRSTUV..........................QRSTUV..........
1859 for (i = 0, sh = a->n - 1; i < 5; i++, sh <<= 1) {
1860 m = mask[a->n - 2][i];
1862 tcg_gen_andi_i64(hi, hi, m);
1863 tcg_gen_andi_i64(lo, lo, m);
1866 tcg_gen_shli_i64(t0, hi, sh);
1867 tcg_gen_shli_i64(t1, lo, sh);
1868 tcg_gen_or_i64(hi, t0, hi);
1869 tcg_gen_or_i64(lo, t1, lo);
1873 tcg_gen_andi_i64(hi, hi, ~(~0ULL >> nbits));
1874 tcg_gen_andi_i64(lo, lo, ~(~0ULL >> nbits));
1875 tcg_gen_shri_i64(lo, lo, nbits);
1876 tcg_gen_or_i64(hi, hi, lo);
1877 tcg_gen_trunc_i64_tl(cpu_gpr[a->rt], hi);
1879 tcg_temp_free_i64(hi);
1880 tcg_temp_free_i64(lo);
1881 tcg_temp_free_i64(t0);
1882 tcg_temp_free_i64(t1);
1887 static bool do_vextdx(DisasContext *ctx, arg_VA *a, int size, bool right,
1888 void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv))
1890 TCGv_ptr vrt, vra, vrb;
1893 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
1894 REQUIRE_VECTOR(ctx);
1896 vrt = gen_avr_ptr(a->vrt);
1897 vra = gen_avr_ptr(a->vra);
1898 vrb = gen_avr_ptr(a->vrb);
1899 rc = tcg_temp_new();
1901 tcg_gen_andi_tl(rc, cpu_gpr[a->rc], 0x1F);
1903 tcg_gen_subfi_tl(rc, 32 - size, rc);
1905 gen_helper(cpu_env, vrt, vra, vrb, rc);
1907 tcg_temp_free_ptr(vrt);
1908 tcg_temp_free_ptr(vra);
1909 tcg_temp_free_ptr(vrb);
1914 TRANS(VEXTDUBVLX, do_vextdx, 1, false, gen_helper_VEXTDUBVLX)
1915 TRANS(VEXTDUHVLX, do_vextdx, 2, false, gen_helper_VEXTDUHVLX)
1916 TRANS(VEXTDUWVLX, do_vextdx, 4, false, gen_helper_VEXTDUWVLX)
1917 TRANS(VEXTDDVLX, do_vextdx, 8, false, gen_helper_VEXTDDVLX)
1919 TRANS(VEXTDUBVRX, do_vextdx, 1, true, gen_helper_VEXTDUBVLX)
1920 TRANS(VEXTDUHVRX, do_vextdx, 2, true, gen_helper_VEXTDUHVLX)
1921 TRANS(VEXTDUWVRX, do_vextdx, 4, true, gen_helper_VEXTDUWVLX)
1922 TRANS(VEXTDDVRX, do_vextdx, 8, true, gen_helper_VEXTDDVLX)
1924 static bool do_vinsx(DisasContext *ctx, int vrt, int size, bool right, TCGv ra,
1925 TCGv_i64 rb, void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv))
1930 t = gen_avr_ptr(vrt);
1931 idx = tcg_temp_new();
1933 tcg_gen_andi_tl(idx, ra, 0xF);
1935 tcg_gen_subfi_tl(idx, 16 - size, idx);
1938 gen_helper(cpu_env, t, rb, idx);
1940 tcg_temp_free_ptr(t);
1946 static bool do_vinsvx(DisasContext *ctx, int vrt, int size, bool right, TCGv ra,
1947 int vrb, void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv))
1952 val = tcg_temp_new_i64();
1953 get_avr64(val, vrb, true);
1954 ok = do_vinsx(ctx, vrt, size, right, ra, val, gen_helper);
1956 tcg_temp_free_i64(val);
1960 static bool do_vinsx_VX(DisasContext *ctx, arg_VX *a, int size, bool right,
1961 void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv))
1966 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
1967 REQUIRE_VECTOR(ctx);
1969 val = tcg_temp_new_i64();
1970 tcg_gen_extu_tl_i64(val, cpu_gpr[a->vrb]);
1972 ok = do_vinsx(ctx, a->vrt, size, right, cpu_gpr[a->vra], val, gen_helper);
1974 tcg_temp_free_i64(val);
1978 static bool do_vinsvx_VX(DisasContext *ctx, arg_VX *a, int size, bool right,
1979 void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv))
1981 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
1982 REQUIRE_VECTOR(ctx);
1984 return do_vinsvx(ctx, a->vrt, size, right, cpu_gpr[a->vra], a->vrb,
1988 static bool do_vins_VX_uim4(DisasContext *ctx, arg_VX_uim4 *a, int size,
1989 void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv))
1994 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
1995 REQUIRE_VECTOR(ctx);
1997 if (a->uim > (16 - size)) {
1999 * PowerISA v3.1 says that the resulting value is undefined in this
2000 * case, so just log a guest error and leave VRT unchanged. The
2001 * real hardware would do a partial insert, e.g. if VRT is zeroed and
2002 * RB is 0x12345678, executing "vinsw VRT,RB,14" results in
2003 * VRT = 0x0000...00001234, but we don't bother to reproduce this
2004 * behavior as software shouldn't rely on it.
2006 qemu_log_mask(LOG_GUEST_ERROR, "Invalid index for VINS* at"
2007 " 0x" TARGET_FMT_lx ", UIM = %d > %d\n", ctx->cia, a->uim,
2012 val = tcg_temp_new_i64();
2013 tcg_gen_extu_tl_i64(val, cpu_gpr[a->vrb]);
2015 ok = do_vinsx(ctx, a->vrt, size, false, tcg_constant_tl(a->uim), val,
2018 tcg_temp_free_i64(val);
2022 static bool do_vinsert_VX_uim4(DisasContext *ctx, arg_VX_uim4 *a, int size,
2023 void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv))
2025 REQUIRE_INSNS_FLAGS2(ctx, ISA300);
2026 REQUIRE_VECTOR(ctx);
2028 if (a->uim > (16 - size)) {
2029 qemu_log_mask(LOG_GUEST_ERROR, "Invalid index for VINSERT* at"
2030 " 0x" TARGET_FMT_lx ", UIM = %d > %d\n", ctx->cia, a->uim,
2035 return do_vinsvx(ctx, a->vrt, size, false, tcg_constant_tl(a->uim), a->vrb,
2039 TRANS(VINSBLX, do_vinsx_VX, 1, false, gen_helper_VINSBLX)
2040 TRANS(VINSHLX, do_vinsx_VX, 2, false, gen_helper_VINSHLX)
2041 TRANS(VINSWLX, do_vinsx_VX, 4, false, gen_helper_VINSWLX)
2042 TRANS(VINSDLX, do_vinsx_VX, 8, false, gen_helper_VINSDLX)
2044 TRANS(VINSBRX, do_vinsx_VX, 1, true, gen_helper_VINSBLX)
2045 TRANS(VINSHRX, do_vinsx_VX, 2, true, gen_helper_VINSHLX)
2046 TRANS(VINSWRX, do_vinsx_VX, 4, true, gen_helper_VINSWLX)
2047 TRANS(VINSDRX, do_vinsx_VX, 8, true, gen_helper_VINSDLX)
2049 TRANS(VINSW, do_vins_VX_uim4, 4, gen_helper_VINSWLX)
2050 TRANS(VINSD, do_vins_VX_uim4, 8, gen_helper_VINSDLX)
2052 TRANS(VINSBVLX, do_vinsvx_VX, 1, false, gen_helper_VINSBLX)
2053 TRANS(VINSHVLX, do_vinsvx_VX, 2, false, gen_helper_VINSHLX)
2054 TRANS(VINSWVLX, do_vinsvx_VX, 4, false, gen_helper_VINSWLX)
2056 TRANS(VINSBVRX, do_vinsvx_VX, 1, true, gen_helper_VINSBLX)
2057 TRANS(VINSHVRX, do_vinsvx_VX, 2, true, gen_helper_VINSHLX)
2058 TRANS(VINSWVRX, do_vinsvx_VX, 4, true, gen_helper_VINSWLX)
2060 TRANS(VINSERTB, do_vinsert_VX_uim4, 1, gen_helper_VINSBLX)
2061 TRANS(VINSERTH, do_vinsert_VX_uim4, 2, gen_helper_VINSHLX)
2062 TRANS(VINSERTW, do_vinsert_VX_uim4, 4, gen_helper_VINSWLX)
2063 TRANS(VINSERTD, do_vinsert_VX_uim4, 8, gen_helper_VINSDLX)
2065 static void gen_vsldoi(DisasContext *ctx)
2067 TCGv_ptr ra, rb, rd;
2069 if (unlikely(!ctx->altivec_enabled)) {
2070 gen_exception(ctx, POWERPC_EXCP_VPU);
2073 ra = gen_avr_ptr(rA(ctx->opcode));
2074 rb = gen_avr_ptr(rB(ctx->opcode));
2075 rd = gen_avr_ptr(rD(ctx->opcode));
2076 sh = tcg_const_i32(VSH(ctx->opcode));
2077 gen_helper_vsldoi(rd, ra, rb, sh);
2078 tcg_temp_free_ptr(ra);
2079 tcg_temp_free_ptr(rb);
2080 tcg_temp_free_ptr(rd);
2081 tcg_temp_free_i32(sh);
2084 static bool trans_VSLDBI(DisasContext *ctx, arg_VN *a)
2086 TCGv_i64 t0, t1, t2;
2088 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2089 REQUIRE_VECTOR(ctx);
2091 t0 = tcg_temp_new_i64();
2092 t1 = tcg_temp_new_i64();
2094 get_avr64(t0, a->vra, true);
2095 get_avr64(t1, a->vra, false);
2098 t2 = tcg_temp_new_i64();
2100 get_avr64(t2, a->vrb, true);
2102 tcg_gen_extract2_i64(t0, t1, t0, 64 - a->sh);
2103 tcg_gen_extract2_i64(t1, t2, t1, 64 - a->sh);
2105 tcg_temp_free_i64(t2);
2108 set_avr64(a->vrt, t0, true);
2109 set_avr64(a->vrt, t1, false);
2111 tcg_temp_free_i64(t0);
2112 tcg_temp_free_i64(t1);
2117 static bool trans_VSRDBI(DisasContext *ctx, arg_VN *a)
2119 TCGv_i64 t2, t1, t0;
2121 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2122 REQUIRE_VECTOR(ctx);
2124 t0 = tcg_temp_new_i64();
2125 t1 = tcg_temp_new_i64();
2127 get_avr64(t0, a->vrb, false);
2128 get_avr64(t1, a->vrb, true);
2131 t2 = tcg_temp_new_i64();
2133 get_avr64(t2, a->vra, false);
2135 tcg_gen_extract2_i64(t0, t0, t1, a->sh);
2136 tcg_gen_extract2_i64(t1, t1, t2, a->sh);
2138 tcg_temp_free_i64(t2);
2141 set_avr64(a->vrt, t0, false);
2142 set_avr64(a->vrt, t1, true);
2144 tcg_temp_free_i64(t0);
2145 tcg_temp_free_i64(t1);
2150 static bool do_vexpand(DisasContext *ctx, arg_VX_tb *a, unsigned vece)
2152 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2153 REQUIRE_VECTOR(ctx);
2155 tcg_gen_gvec_sari(vece, avr_full_offset(a->vrt), avr_full_offset(a->vrb),
2156 (8 << vece) - 1, 16, 16);
2161 TRANS(VEXPANDBM, do_vexpand, MO_8)
2162 TRANS(VEXPANDHM, do_vexpand, MO_16)
2163 TRANS(VEXPANDWM, do_vexpand, MO_32)
2164 TRANS(VEXPANDDM, do_vexpand, MO_64)
2166 static bool trans_VEXPANDQM(DisasContext *ctx, arg_VX_tb *a)
2170 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2171 REQUIRE_VECTOR(ctx);
2173 tmp = tcg_temp_new_i64();
2175 get_avr64(tmp, a->vrb, true);
2176 tcg_gen_sari_i64(tmp, tmp, 63);
2177 set_avr64(a->vrt, tmp, false);
2178 set_avr64(a->vrt, tmp, true);
2180 tcg_temp_free_i64(tmp);
2184 static bool do_vextractm(DisasContext *ctx, arg_VX_tb *a, unsigned vece)
2186 const uint64_t elem_width = 8 << vece, elem_count_half = 8 >> vece,
2187 mask = dup_const(vece, 1 << (elem_width - 1));
2189 TCGv_i64 lo, hi, t0, t1;
2191 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2192 REQUIRE_VECTOR(ctx);
2194 hi = tcg_temp_new_i64();
2195 lo = tcg_temp_new_i64();
2196 t0 = tcg_temp_new_i64();
2197 t1 = tcg_temp_new_i64();
2199 get_avr64(lo, a->vrb, false);
2200 get_avr64(hi, a->vrb, true);
2202 tcg_gen_andi_i64(lo, lo, mask);
2203 tcg_gen_andi_i64(hi, hi, mask);
2206 * Gather the most significant bit of each element in the highest element
2207 * element. E.g. for bytes:
2208 * aXXXXXXXbXXXXXXXcXXXXXXXdXXXXXXXeXXXXXXXfXXXXXXXgXXXXXXXhXXXXXXX
2209 * & dup(1 << (elem_width - 1))
2210 * a0000000b0000000c0000000d0000000e0000000f0000000g0000000h0000000
2212 * 0000e0000000f0000000g0000000h00000000000000000000000000000000000
2214 * a000e000b000f000c000g000d000h000e0000000f0000000g0000000h0000000
2216 * 00c000g000d000h000e0000000f0000000g0000000h000000000000000000000
2218 * a0c0e0g0b0d0f0h0c0e0g000d0f0h000e0g00000f0h00000g0000000h0000000
2220 * 0b0d0f0h0c0e0g000d0f0h000e0g00000f0h00000g0000000h00000000000000
2222 * abcdefghbcdefgh0cdefgh00defgh000efgh0000fgh00000gh000000h0000000
2224 for (i = elem_count_half / 2, j = 32; i > 0; i >>= 1, j >>= 1) {
2225 tcg_gen_shli_i64(t0, hi, j - i);
2226 tcg_gen_shli_i64(t1, lo, j - i);
2227 tcg_gen_or_i64(hi, hi, t0);
2228 tcg_gen_or_i64(lo, lo, t1);
2231 tcg_gen_shri_i64(hi, hi, 64 - elem_count_half);
2232 tcg_gen_extract2_i64(lo, lo, hi, 64 - elem_count_half);
2233 tcg_gen_trunc_i64_tl(cpu_gpr[a->vrt], lo);
2235 tcg_temp_free_i64(hi);
2236 tcg_temp_free_i64(lo);
2237 tcg_temp_free_i64(t0);
2238 tcg_temp_free_i64(t1);
2243 TRANS(VEXTRACTBM, do_vextractm, MO_8)
2244 TRANS(VEXTRACTHM, do_vextractm, MO_16)
2245 TRANS(VEXTRACTWM, do_vextractm, MO_32)
2246 TRANS(VEXTRACTDM, do_vextractm, MO_64)
2248 static bool trans_VEXTRACTQM(DisasContext *ctx, arg_VX_tb *a)
2252 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2253 REQUIRE_VECTOR(ctx);
2255 tmp = tcg_temp_new_i64();
2257 get_avr64(tmp, a->vrb, true);
2258 tcg_gen_shri_i64(tmp, tmp, 63);
2259 tcg_gen_trunc_i64_tl(cpu_gpr[a->vrt], tmp);
2261 tcg_temp_free_i64(tmp);
2266 static bool do_mtvsrm(DisasContext *ctx, arg_VX_tb *a, unsigned vece)
2268 const uint64_t elem_width = 8 << vece, elem_count_half = 8 >> vece;
2271 TCGv_i64 hi, lo, t0, t1;
2273 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2274 REQUIRE_VECTOR(ctx);
2276 hi = tcg_temp_new_i64();
2277 lo = tcg_temp_new_i64();
2278 t0 = tcg_temp_new_i64();
2279 t1 = tcg_temp_new_i64();
2281 tcg_gen_extu_tl_i64(t0, cpu_gpr[a->vrb]);
2282 tcg_gen_extract_i64(hi, t0, elem_count_half, elem_count_half);
2283 tcg_gen_extract_i64(lo, t0, 0, elem_count_half);
2286 * Spread the bits into their respective elements.
2288 * 00000000000000000000000000000000000000000000000000000000abcdefgh
2290 * 0000000000000000000000000000abcdefgh0000000000000000000000000000
2292 * 0000000000000000000000000000abcdefgh00000000000000000000abcdefgh
2294 * 00000000000000abcdefgh00000000000000000000abcdefgh00000000000000
2296 * 00000000000000abcdefgh000000abcdefgh000000abcdefgh000000abcdefgh
2298 * 0000000abcdefgh000000abcdefgh000000abcdefgh000000abcdefgh0000000
2300 * 0000000abcdefgXbcdefgXbcdefgXbcdefgXbcdefgXbcdefgXbcdefgXbcdefgh
2302 * 0000000a0000000b0000000c0000000d0000000e0000000f0000000g0000000h
2304 * aaaaaaaabbbbbbbbccccccccddddddddeeeeeeeeffffffffgggggggghhhhhhhh
2306 for (i = elem_count_half / 2, j = 32; i > 0; i >>= 1, j >>= 1) {
2307 tcg_gen_shli_i64(t0, hi, j - i);
2308 tcg_gen_shli_i64(t1, lo, j - i);
2309 tcg_gen_or_i64(hi, hi, t0);
2310 tcg_gen_or_i64(lo, lo, t1);
2313 c = dup_const(vece, 1);
2314 tcg_gen_andi_i64(hi, hi, c);
2315 tcg_gen_andi_i64(lo, lo, c);
2317 c = MAKE_64BIT_MASK(0, elem_width);
2318 tcg_gen_muli_i64(hi, hi, c);
2319 tcg_gen_muli_i64(lo, lo, c);
2321 set_avr64(a->vrt, lo, false);
2322 set_avr64(a->vrt, hi, true);
2324 tcg_temp_free_i64(hi);
2325 tcg_temp_free_i64(lo);
2326 tcg_temp_free_i64(t0);
2327 tcg_temp_free_i64(t1);
2332 TRANS(MTVSRBM, do_mtvsrm, MO_8)
2333 TRANS(MTVSRHM, do_mtvsrm, MO_16)
2334 TRANS(MTVSRWM, do_mtvsrm, MO_32)
2335 TRANS(MTVSRDM, do_mtvsrm, MO_64)
2337 static bool trans_MTVSRQM(DisasContext *ctx, arg_VX_tb *a)
2341 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2342 REQUIRE_VECTOR(ctx);
2344 tmp = tcg_temp_new_i64();
2346 tcg_gen_ext_tl_i64(tmp, cpu_gpr[a->vrb]);
2347 tcg_gen_sextract_i64(tmp, tmp, 0, 1);
2348 set_avr64(a->vrt, tmp, false);
2349 set_avr64(a->vrt, tmp, true);
2351 tcg_temp_free_i64(tmp);
2356 static bool trans_MTVSRBMI(DisasContext *ctx, arg_DX_b *a)
2358 const uint64_t mask = dup_const(MO_8, 1);
2361 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2362 REQUIRE_VECTOR(ctx);
2364 hi = extract16(a->b, 8, 8);
2365 lo = extract16(a->b, 0, 8);
2367 for (int i = 4, j = 32; i > 0; i >>= 1, j >>= 1) {
2368 hi |= hi << (j - i);
2369 lo |= lo << (j - i);
2372 hi = (hi & mask) * 0xFF;
2373 lo = (lo & mask) * 0xFF;
2375 set_avr64(a->vrt, tcg_constant_i64(hi), true);
2376 set_avr64(a->vrt, tcg_constant_i64(lo), false);
2381 static bool do_vcntmb(DisasContext *ctx, arg_VX_mp *a, int vece)
2383 TCGv_i64 rt, vrb, mask;
2384 rt = tcg_const_i64(0);
2385 vrb = tcg_temp_new_i64();
2386 mask = tcg_constant_i64(dup_const(vece, 1ULL << ((8 << vece) - 1)));
2388 for (int i = 0; i < 2; i++) {
2389 get_avr64(vrb, a->vrb, i);
2391 tcg_gen_and_i64(vrb, mask, vrb);
2393 tcg_gen_andc_i64(vrb, mask, vrb);
2395 tcg_gen_ctpop_i64(vrb, vrb);
2396 tcg_gen_add_i64(rt, rt, vrb);
2399 tcg_gen_shli_i64(rt, rt, TARGET_LONG_BITS - 8 + vece);
2400 tcg_gen_trunc_i64_tl(cpu_gpr[a->rt], rt);
2402 tcg_temp_free_i64(vrb);
2403 tcg_temp_free_i64(rt);
2408 TRANS(VCNTMBB, do_vcntmb, MO_8)
2409 TRANS(VCNTMBH, do_vcntmb, MO_16)
2410 TRANS(VCNTMBW, do_vcntmb, MO_32)
2411 TRANS(VCNTMBD, do_vcntmb, MO_64)
2413 static bool do_vstri(DisasContext *ctx, arg_VX_tb_rc *a,
2414 void (*gen_helper)(TCGv_i32, TCGv_ptr, TCGv_ptr))
2418 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2419 REQUIRE_VECTOR(ctx);
2421 vrt = gen_avr_ptr(a->vrt);
2422 vrb = gen_avr_ptr(a->vrb);
2425 gen_helper(cpu_crf[6], vrt, vrb);
2427 TCGv_i32 discard = tcg_temp_new_i32();
2428 gen_helper(discard, vrt, vrb);
2429 tcg_temp_free_i32(discard);
2432 tcg_temp_free_ptr(vrt);
2433 tcg_temp_free_ptr(vrb);
2438 TRANS(VSTRIBL, do_vstri, gen_helper_VSTRIBL)
2439 TRANS(VSTRIBR, do_vstri, gen_helper_VSTRIBR)
2440 TRANS(VSTRIHL, do_vstri, gen_helper_VSTRIHL)
2441 TRANS(VSTRIHR, do_vstri, gen_helper_VSTRIHR)
2443 static bool do_vclrb(DisasContext *ctx, arg_VX *a, bool right)
2445 TCGv_i64 rb, mh, ml, tmp,
2446 ones = tcg_constant_i64(-1),
2447 zero = tcg_constant_i64(0);
2449 rb = tcg_temp_new_i64();
2450 mh = tcg_temp_new_i64();
2451 ml = tcg_temp_new_i64();
2452 tmp = tcg_temp_new_i64();
2454 tcg_gen_extu_tl_i64(rb, cpu_gpr[a->vrb]);
2455 tcg_gen_andi_i64(tmp, rb, 7);
2456 tcg_gen_shli_i64(tmp, tmp, 3);
2458 tcg_gen_shr_i64(tmp, ones, tmp);
2460 tcg_gen_shl_i64(tmp, ones, tmp);
2462 tcg_gen_not_i64(tmp, tmp);
2465 tcg_gen_movcond_i64(TCG_COND_LTU, mh, rb, tcg_constant_i64(8),
2467 tcg_gen_movcond_i64(TCG_COND_LTU, ml, rb, tcg_constant_i64(8),
2469 tcg_gen_movcond_i64(TCG_COND_LTU, ml, rb, tcg_constant_i64(16),
2472 tcg_gen_movcond_i64(TCG_COND_LTU, ml, rb, tcg_constant_i64(8),
2474 tcg_gen_movcond_i64(TCG_COND_LTU, mh, rb, tcg_constant_i64(8),
2476 tcg_gen_movcond_i64(TCG_COND_LTU, mh, rb, tcg_constant_i64(16),
2480 get_avr64(tmp, a->vra, true);
2481 tcg_gen_and_i64(tmp, tmp, mh);
2482 set_avr64(a->vrt, tmp, true);
2484 get_avr64(tmp, a->vra, false);
2485 tcg_gen_and_i64(tmp, tmp, ml);
2486 set_avr64(a->vrt, tmp, false);
2488 tcg_temp_free_i64(rb);
2489 tcg_temp_free_i64(mh);
2490 tcg_temp_free_i64(ml);
2491 tcg_temp_free_i64(tmp);
2496 TRANS(VCLRLB, do_vclrb, false)
2497 TRANS(VCLRRB, do_vclrb, true)
2499 #define GEN_VAFORM_PAIRED(name0, name1, opc2) \
2500 static void glue(gen_, name0##_##name1)(DisasContext *ctx) \
2502 TCGv_ptr ra, rb, rc, rd; \
2503 if (unlikely(!ctx->altivec_enabled)) { \
2504 gen_exception(ctx, POWERPC_EXCP_VPU); \
2507 ra = gen_avr_ptr(rA(ctx->opcode)); \
2508 rb = gen_avr_ptr(rB(ctx->opcode)); \
2509 rc = gen_avr_ptr(rC(ctx->opcode)); \
2510 rd = gen_avr_ptr(rD(ctx->opcode)); \
2511 if (Rc(ctx->opcode)) { \
2512 gen_helper_##name1(cpu_env, rd, ra, rb, rc); \
2514 gen_helper_##name0(cpu_env, rd, ra, rb, rc); \
2516 tcg_temp_free_ptr(ra); \
2517 tcg_temp_free_ptr(rb); \
2518 tcg_temp_free_ptr(rc); \
2519 tcg_temp_free_ptr(rd); \
2522 GEN_VAFORM_PAIRED(vmaddfp, vnmsubfp, 23)
2524 static bool do_va_helper(DisasContext *ctx, arg_VA *a,
2525 void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_ptr))
2527 TCGv_ptr vrt, vra, vrb, vrc;
2528 REQUIRE_VECTOR(ctx);
2530 vrt = gen_avr_ptr(a->vrt);
2531 vra = gen_avr_ptr(a->vra);
2532 vrb = gen_avr_ptr(a->vrb);
2533 vrc = gen_avr_ptr(a->rc);
2534 gen_helper(vrt, vra, vrb, vrc);
2535 tcg_temp_free_ptr(vrt);
2536 tcg_temp_free_ptr(vra);
2537 tcg_temp_free_ptr(vrb);
2538 tcg_temp_free_ptr(vrc);
2543 TRANS_FLAGS2(ALTIVEC_207, VADDECUQ, do_va_helper, gen_helper_VADDECUQ)
2544 TRANS_FLAGS2(ALTIVEC_207, VADDEUQM, do_va_helper, gen_helper_VADDEUQM)
2546 TRANS_FLAGS2(ALTIVEC_207, VSUBEUQM, do_va_helper, gen_helper_VSUBEUQM)
2547 TRANS_FLAGS2(ALTIVEC_207, VSUBECUQ, do_va_helper, gen_helper_VSUBECUQ)
2549 TRANS_FLAGS(ALTIVEC, VPERM, do_va_helper, gen_helper_VPERM)
2550 TRANS_FLAGS2(ISA300, VPERMR, do_va_helper, gen_helper_VPERMR)
2552 static void gen_vmladduhm_vec(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b,
2555 tcg_gen_mul_vec(vece, t, a, b);
2556 tcg_gen_add_vec(vece, t, t, c);
2559 static bool trans_VMLADDUHM(DisasContext *ctx, arg_VA *a)
2561 static const TCGOpcode vecop_list[] = {
2562 INDEX_op_add_vec, INDEX_op_mul_vec, 0
2565 static const GVecGen4 op = {
2566 .fno = gen_helper_VMLADDUHM,
2567 .fniv = gen_vmladduhm_vec,
2568 .opt_opc = vecop_list,
2572 REQUIRE_INSNS_FLAGS(ctx, ALTIVEC);
2573 REQUIRE_VECTOR(ctx);
2575 tcg_gen_gvec_4(avr_full_offset(a->vrt), avr_full_offset(a->vra),
2576 avr_full_offset(a->vrb), avr_full_offset(a->rc),
2582 static bool trans_VSEL(DisasContext *ctx, arg_VA *a)
2584 REQUIRE_INSNS_FLAGS(ctx, ALTIVEC);
2585 REQUIRE_VECTOR(ctx);
2587 tcg_gen_gvec_bitsel(MO_64, avr_full_offset(a->vrt), avr_full_offset(a->rc),
2588 avr_full_offset(a->vrb), avr_full_offset(a->vra),
2594 TRANS_FLAGS(ALTIVEC, VMSUMUBM, do_va_helper, gen_helper_VMSUMUBM)
2595 TRANS_FLAGS(ALTIVEC, VMSUMMBM, do_va_helper, gen_helper_VMSUMMBM)
2596 TRANS_FLAGS(ALTIVEC, VMSUMSHM, do_va_helper, gen_helper_VMSUMSHM)
2597 TRANS_FLAGS(ALTIVEC, VMSUMUHM, do_va_helper, gen_helper_VMSUMUHM)
2599 static bool do_va_env_helper(DisasContext *ctx, arg_VA *a,
2600 void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_ptr))
2602 TCGv_ptr vrt, vra, vrb, vrc;
2603 REQUIRE_VECTOR(ctx);
2605 vrt = gen_avr_ptr(a->vrt);
2606 vra = gen_avr_ptr(a->vra);
2607 vrb = gen_avr_ptr(a->vrb);
2608 vrc = gen_avr_ptr(a->rc);
2609 gen_helper(cpu_env, vrt, vra, vrb, vrc);
2610 tcg_temp_free_ptr(vrt);
2611 tcg_temp_free_ptr(vra);
2612 tcg_temp_free_ptr(vrb);
2613 tcg_temp_free_ptr(vrc);
2618 TRANS_FLAGS(ALTIVEC, VMSUMUHS, do_va_env_helper, gen_helper_VMSUMUHS)
2619 TRANS_FLAGS(ALTIVEC, VMSUMSHS, do_va_env_helper, gen_helper_VMSUMSHS)
2621 TRANS_FLAGS(ALTIVEC, VMHADDSHS, do_va_env_helper, gen_helper_VMHADDSHS)
2622 TRANS_FLAGS(ALTIVEC, VMHRADDSHS, do_va_env_helper, gen_helper_VMHRADDSHS)
2624 GEN_VXFORM_NOA(vclzb, 1, 28)
2625 GEN_VXFORM_NOA(vclzh, 1, 29)
2626 GEN_VXFORM_TRANS(vclzw, 1, 30)
2627 GEN_VXFORM_TRANS(vclzd, 1, 31)
2628 GEN_VXFORM_NOA_2(vnegw, 1, 24, 6)
2629 GEN_VXFORM_NOA_2(vnegd, 1, 24, 7)
2631 static void gen_vexts_i64(TCGv_i64 t, TCGv_i64 b, int64_t s)
2633 tcg_gen_sextract_i64(t, b, 0, 64 - s);
2636 static void gen_vexts_i32(TCGv_i32 t, TCGv_i32 b, int32_t s)
2638 tcg_gen_sextract_i32(t, b, 0, 32 - s);
2641 static void gen_vexts_vec(unsigned vece, TCGv_vec t, TCGv_vec b, int64_t s)
2643 tcg_gen_shli_vec(vece, t, b, s);
2644 tcg_gen_sari_vec(vece, t, t, s);
2647 static bool do_vexts(DisasContext *ctx, arg_VX_tb *a, unsigned vece, int64_t s)
2649 static const TCGOpcode vecop_list[] = {
2650 INDEX_op_shli_vec, INDEX_op_sari_vec, 0
2653 static const GVecGen2i op[2] = {
2655 .fni4 = gen_vexts_i32,
2656 .fniv = gen_vexts_vec,
2657 .opt_opc = vecop_list,
2661 .fni8 = gen_vexts_i64,
2662 .fniv = gen_vexts_vec,
2663 .opt_opc = vecop_list,
2668 REQUIRE_INSNS_FLAGS2(ctx, ISA300);
2669 REQUIRE_VECTOR(ctx);
2671 tcg_gen_gvec_2i(avr_full_offset(a->vrt), avr_full_offset(a->vrb),
2672 16, 16, s, &op[vece - MO_32]);
2677 TRANS(VEXTSB2W, do_vexts, MO_32, 24);
2678 TRANS(VEXTSH2W, do_vexts, MO_32, 16);
2679 TRANS(VEXTSB2D, do_vexts, MO_64, 56);
2680 TRANS(VEXTSH2D, do_vexts, MO_64, 48);
2681 TRANS(VEXTSW2D, do_vexts, MO_64, 32);
2683 static bool trans_VEXTSD2Q(DisasContext *ctx, arg_VX_tb *a)
2687 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2688 REQUIRE_VECTOR(ctx);
2690 tmp = tcg_temp_new_i64();
2692 get_avr64(tmp, a->vrb, false);
2693 set_avr64(a->vrt, tmp, false);
2694 tcg_gen_sari_i64(tmp, tmp, 63);
2695 set_avr64(a->vrt, tmp, true);
2697 tcg_temp_free_i64(tmp);
2701 GEN_VXFORM_NOA_2(vctzb, 1, 24, 28)
2702 GEN_VXFORM_NOA_2(vctzh, 1, 24, 29)
2703 GEN_VXFORM_NOA_2(vctzw, 1, 24, 30)
2704 GEN_VXFORM_NOA_2(vctzd, 1, 24, 31)
2705 GEN_VXFORM_NOA_3(vclzlsbb, 1, 24, 0)
2706 GEN_VXFORM_NOA_3(vctzlsbb, 1, 24, 1)
2707 GEN_VXFORM_NOA(vpopcntb, 1, 28)
2708 GEN_VXFORM_NOA(vpopcnth, 1, 29)
2709 GEN_VXFORM_NOA(vpopcntw, 1, 30)
2710 GEN_VXFORM_NOA(vpopcntd, 1, 31)
2711 GEN_VXFORM_DUAL(vclzb, PPC_NONE, PPC2_ALTIVEC_207, \
2712 vpopcntb, PPC_NONE, PPC2_ALTIVEC_207)
2713 GEN_VXFORM_DUAL(vclzh, PPC_NONE, PPC2_ALTIVEC_207, \
2714 vpopcnth, PPC_NONE, PPC2_ALTIVEC_207)
2715 GEN_VXFORM_DUAL(vclzw, PPC_NONE, PPC2_ALTIVEC_207, \
2716 vpopcntw, PPC_NONE, PPC2_ALTIVEC_207)
2717 GEN_VXFORM_DUAL(vclzd, PPC_NONE, PPC2_ALTIVEC_207, \
2718 vpopcntd, PPC_NONE, PPC2_ALTIVEC_207)
2719 GEN_VXFORM(vbpermd, 6, 23);
2720 GEN_VXFORM(vbpermq, 6, 21);
2721 GEN_VXFORM_TRANS(vgbbd, 6, 20);
2722 GEN_VXFORM(vpmsumb, 4, 16)
2723 GEN_VXFORM(vpmsumh, 4, 17)
2724 GEN_VXFORM(vpmsumw, 4, 18)
2726 #define GEN_BCD(op) \
2727 static void gen_##op(DisasContext *ctx) \
2729 TCGv_ptr ra, rb, rd; \
2732 if (unlikely(!ctx->altivec_enabled)) { \
2733 gen_exception(ctx, POWERPC_EXCP_VPU); \
2737 ra = gen_avr_ptr(rA(ctx->opcode)); \
2738 rb = gen_avr_ptr(rB(ctx->opcode)); \
2739 rd = gen_avr_ptr(rD(ctx->opcode)); \
2741 ps = tcg_const_i32((ctx->opcode & 0x200) != 0); \
2743 gen_helper_##op(cpu_crf[6], rd, ra, rb, ps); \
2745 tcg_temp_free_ptr(ra); \
2746 tcg_temp_free_ptr(rb); \
2747 tcg_temp_free_ptr(rd); \
2748 tcg_temp_free_i32(ps); \
2751 #define GEN_BCD2(op) \
2752 static void gen_##op(DisasContext *ctx) \
2757 if (unlikely(!ctx->altivec_enabled)) { \
2758 gen_exception(ctx, POWERPC_EXCP_VPU); \
2762 rb = gen_avr_ptr(rB(ctx->opcode)); \
2763 rd = gen_avr_ptr(rD(ctx->opcode)); \
2765 ps = tcg_const_i32((ctx->opcode & 0x200) != 0); \
2767 gen_helper_##op(cpu_crf[6], rd, rb, ps); \
2769 tcg_temp_free_ptr(rb); \
2770 tcg_temp_free_ptr(rd); \
2771 tcg_temp_free_i32(ps); \
2790 static void gen_xpnd04_1(DisasContext *ctx)
2792 switch (opc4(ctx->opcode)) {
2820 static void gen_xpnd04_2(DisasContext *ctx)
2822 switch (opc4(ctx->opcode)) {
2848 GEN_VXFORM_DUAL(vsubsws, PPC_ALTIVEC, PPC_NONE, \
2849 xpnd04_2, PPC_NONE, PPC2_ISA300)
2851 GEN_VXFORM_DUAL(vsububm, PPC_ALTIVEC, PPC_NONE, \
2852 bcdadd, PPC_NONE, PPC2_ALTIVEC_207)
2853 GEN_VXFORM_DUAL(vsububs, PPC_ALTIVEC, PPC_NONE, \
2854 bcdadd, PPC_NONE, PPC2_ALTIVEC_207)
2855 GEN_VXFORM_DUAL(vsubuhm, PPC_ALTIVEC, PPC_NONE, \
2856 bcdsub, PPC_NONE, PPC2_ALTIVEC_207)
2857 GEN_VXFORM_DUAL(vsubuhs, PPC_ALTIVEC, PPC_NONE, \
2858 bcdsub, PPC_NONE, PPC2_ALTIVEC_207)
2859 GEN_VXFORM_DUAL(vaddshs, PPC_ALTIVEC, PPC_NONE, \
2860 bcdcpsgn, PPC_NONE, PPC2_ISA300)
2861 GEN_VXFORM_DUAL(vsubudm, PPC2_ALTIVEC_207, PPC_NONE, \
2862 bcds, PPC_NONE, PPC2_ISA300)
2863 GEN_VXFORM_DUAL(vsubuwm, PPC_ALTIVEC, PPC_NONE, \
2864 bcdus, PPC_NONE, PPC2_ISA300)
2865 GEN_VXFORM_DUAL(vsubsbs, PPC_ALTIVEC, PPC_NONE, \
2866 bcdtrunc, PPC_NONE, PPC2_ISA300)
2868 static void gen_vsbox(DisasContext *ctx)
2871 if (unlikely(!ctx->altivec_enabled)) {
2872 gen_exception(ctx, POWERPC_EXCP_VPU);
2875 ra = gen_avr_ptr(rA(ctx->opcode));
2876 rd = gen_avr_ptr(rD(ctx->opcode));
2877 gen_helper_vsbox(rd, ra);
2878 tcg_temp_free_ptr(ra);
2879 tcg_temp_free_ptr(rd);
2882 GEN_VXFORM(vcipher, 4, 20)
2883 GEN_VXFORM(vcipherlast, 4, 20)
2884 GEN_VXFORM(vncipher, 4, 21)
2885 GEN_VXFORM(vncipherlast, 4, 21)
2887 GEN_VXFORM_DUAL(vcipher, PPC_NONE, PPC2_ALTIVEC_207,
2888 vcipherlast, PPC_NONE, PPC2_ALTIVEC_207)
2889 GEN_VXFORM_DUAL(vncipher, PPC_NONE, PPC2_ALTIVEC_207,
2890 vncipherlast, PPC_NONE, PPC2_ALTIVEC_207)
2892 #define VSHASIGMA(op) \
2893 static void gen_##op(DisasContext *ctx) \
2897 if (unlikely(!ctx->altivec_enabled)) { \
2898 gen_exception(ctx, POWERPC_EXCP_VPU); \
2901 ra = gen_avr_ptr(rA(ctx->opcode)); \
2902 rd = gen_avr_ptr(rD(ctx->opcode)); \
2903 st_six = tcg_const_i32(rB(ctx->opcode)); \
2904 gen_helper_##op(rd, ra, st_six); \
2905 tcg_temp_free_ptr(ra); \
2906 tcg_temp_free_ptr(rd); \
2907 tcg_temp_free_i32(st_six); \
2910 VSHASIGMA(vshasigmaw)
2911 VSHASIGMA(vshasigmad)
2913 GEN_VXFORM3(vpermxor, 22, 0xFF)
2914 GEN_VXFORM_DUAL(vsldoi, PPC_ALTIVEC, PPC_NONE,
2915 vpermxor, PPC_NONE, PPC2_ALTIVEC_207)
2917 static bool trans_VCFUGED(DisasContext *ctx, arg_VX *a)
2919 static const GVecGen3 g = {
2920 .fni8 = gen_helper_CFUGED,
2924 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2925 REQUIRE_VECTOR(ctx);
2927 tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra),
2928 avr_full_offset(a->vrb), 16, 16, &g);
2933 static bool trans_VCLZDM(DisasContext *ctx, arg_VX *a)
2935 static const GVecGen3i g = {
2940 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2941 REQUIRE_VECTOR(ctx);
2943 tcg_gen_gvec_3i(avr_full_offset(a->vrt), avr_full_offset(a->vra),
2944 avr_full_offset(a->vrb), 16, 16, false, &g);
2949 static bool trans_VCTZDM(DisasContext *ctx, arg_VX *a)
2951 static const GVecGen3i g = {
2956 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2957 REQUIRE_VECTOR(ctx);
2959 tcg_gen_gvec_3i(avr_full_offset(a->vrt), avr_full_offset(a->vra),
2960 avr_full_offset(a->vrb), 16, 16, true, &g);
2965 static bool trans_VPDEPD(DisasContext *ctx, arg_VX *a)
2967 static const GVecGen3 g = {
2968 .fni8 = gen_helper_PDEPD,
2972 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2973 REQUIRE_VECTOR(ctx);
2975 tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra),
2976 avr_full_offset(a->vrb), 16, 16, &g);
2981 static bool trans_VPEXTD(DisasContext *ctx, arg_VX *a)
2983 static const GVecGen3 g = {
2984 .fni8 = gen_helper_PEXTD,
2988 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
2989 REQUIRE_VECTOR(ctx);
2991 tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra),
2992 avr_full_offset(a->vrb), 16, 16, &g);
2997 static bool trans_VMSUMUDM(DisasContext *ctx, arg_VA *a)
2999 TCGv_i64 rl, rh, src1, src2;
3002 REQUIRE_INSNS_FLAGS2(ctx, ISA300);
3003 REQUIRE_VECTOR(ctx);
3005 rh = tcg_temp_new_i64();
3006 rl = tcg_temp_new_i64();
3007 src1 = tcg_temp_new_i64();
3008 src2 = tcg_temp_new_i64();
3010 get_avr64(rl, a->rc, false);
3011 get_avr64(rh, a->rc, true);
3013 for (dw = 0; dw < 2; dw++) {
3014 get_avr64(src1, a->vra, dw);
3015 get_avr64(src2, a->vrb, dw);
3016 tcg_gen_mulu2_i64(src1, src2, src1, src2);
3017 tcg_gen_add2_i64(rl, rh, rl, rh, src1, src2);
3020 set_avr64(a->vrt, rl, false);
3021 set_avr64(a->vrt, rh, true);
3023 tcg_temp_free_i64(rl);
3024 tcg_temp_free_i64(rh);
3025 tcg_temp_free_i64(src1);
3026 tcg_temp_free_i64(src2);
3031 static bool trans_VMSUMCUD(DisasContext *ctx, arg_VA *a)
3033 TCGv_i64 tmp0, tmp1, prod1h, prod1l, prod0h, prod0l, zero;
3035 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
3036 REQUIRE_VECTOR(ctx);
3038 tmp0 = tcg_temp_new_i64();
3039 tmp1 = tcg_temp_new_i64();
3040 prod1h = tcg_temp_new_i64();
3041 prod1l = tcg_temp_new_i64();
3042 prod0h = tcg_temp_new_i64();
3043 prod0l = tcg_temp_new_i64();
3044 zero = tcg_constant_i64(0);
3046 /* prod1 = vsr[vra+32].dw[1] * vsr[vrb+32].dw[1] */
3047 get_avr64(tmp0, a->vra, false);
3048 get_avr64(tmp1, a->vrb, false);
3049 tcg_gen_mulu2_i64(prod1l, prod1h, tmp0, tmp1);
3051 /* prod0 = vsr[vra+32].dw[0] * vsr[vrb+32].dw[0] */
3052 get_avr64(tmp0, a->vra, true);
3053 get_avr64(tmp1, a->vrb, true);
3054 tcg_gen_mulu2_i64(prod0l, prod0h, tmp0, tmp1);
3056 /* Sum lower 64-bits elements */
3057 get_avr64(tmp1, a->rc, false);
3058 tcg_gen_add2_i64(tmp1, tmp0, tmp1, zero, prod1l, zero);
3059 tcg_gen_add2_i64(tmp1, tmp0, tmp1, tmp0, prod0l, zero);
3062 * Discard lower 64-bits, leaving the carry into bit 64.
3063 * Then sum the higher 64-bit elements.
3065 get_avr64(tmp1, a->rc, true);
3066 tcg_gen_add2_i64(tmp1, tmp0, tmp0, zero, tmp1, zero);
3067 tcg_gen_add2_i64(tmp1, tmp0, tmp1, tmp0, prod1h, zero);
3068 tcg_gen_add2_i64(tmp1, tmp0, tmp1, tmp0, prod0h, zero);
3070 /* Discard 64 more bits to complete the CHOP128(temp >> 128) */
3071 set_avr64(a->vrt, tmp0, false);
3072 set_avr64(a->vrt, zero, true);
3074 tcg_temp_free_i64(tmp0);
3075 tcg_temp_free_i64(tmp1);
3076 tcg_temp_free_i64(prod1h);
3077 tcg_temp_free_i64(prod1l);
3078 tcg_temp_free_i64(prod0h);
3079 tcg_temp_free_i64(prod0l);
3084 static bool do_vx_helper(DisasContext *ctx, arg_VX *a,
3085 void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_ptr))
3087 TCGv_ptr ra, rb, rd;
3088 REQUIRE_VECTOR(ctx);
3090 ra = gen_avr_ptr(a->vra);
3091 rb = gen_avr_ptr(a->vrb);
3092 rd = gen_avr_ptr(a->vrt);
3093 gen_helper(rd, ra, rb);
3094 tcg_temp_free_ptr(ra);
3095 tcg_temp_free_ptr(rb);
3096 tcg_temp_free_ptr(rd);
3101 TRANS_FLAGS2(ALTIVEC_207, VADDCUQ, do_vx_helper, gen_helper_VADDCUQ)
3102 TRANS_FLAGS2(ALTIVEC_207, VADDUQM, do_vx_helper, gen_helper_VADDUQM)
3104 TRANS_FLAGS2(ALTIVEC_207, VPMSUMD, do_vx_helper, gen_helper_VPMSUMD)
3106 TRANS_FLAGS2(ALTIVEC_207, VSUBCUQ, do_vx_helper, gen_helper_VSUBCUQ)
3107 TRANS_FLAGS2(ALTIVEC_207, VSUBUQM, do_vx_helper, gen_helper_VSUBUQM)
3109 static void gen_VADDCUW_vec(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b)
3111 tcg_gen_not_vec(vece, a, a);
3112 tcg_gen_cmp_vec(TCG_COND_LTU, vece, t, a, b);
3113 tcg_gen_and_vec(vece, t, t, tcg_constant_vec_matching(t, vece, 1));
3116 static void gen_VADDCUW_i32(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b)
3118 tcg_gen_not_i32(a, a);
3119 tcg_gen_setcond_i32(TCG_COND_LTU, t, a, b);
3122 static void gen_VSUBCUW_vec(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b)
3124 tcg_gen_cmp_vec(TCG_COND_GEU, vece, t, a, b);
3125 tcg_gen_and_vec(vece, t, t, tcg_constant_vec_matching(t, vece, 1));
3128 static void gen_VSUBCUW_i32(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b)
3130 tcg_gen_setcond_i32(TCG_COND_GEU, t, a, b);
3133 static bool do_vx_vaddsubcuw(DisasContext *ctx, arg_VX *a, int add)
3135 static const TCGOpcode vecop_list[] = {
3139 static const GVecGen3 op[] = {
3141 .fniv = gen_VSUBCUW_vec,
3142 .fni4 = gen_VSUBCUW_i32,
3143 .opt_opc = vecop_list,
3147 .fniv = gen_VADDCUW_vec,
3148 .fni4 = gen_VADDCUW_i32,
3149 .opt_opc = vecop_list,
3154 REQUIRE_INSNS_FLAGS(ctx, ALTIVEC);
3155 REQUIRE_VECTOR(ctx);
3157 tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra),
3158 avr_full_offset(a->vrb), 16, 16, &op[add]);
3163 TRANS(VSUBCUW, do_vx_vaddsubcuw, 0)
3164 TRANS(VADDCUW, do_vx_vaddsubcuw, 1)
3166 static bool do_vx_vmuleo(DisasContext *ctx, arg_VX *a, bool even,
3167 void (*gen_mul)(TCGv_i64, TCGv_i64, TCGv_i64, TCGv_i64))
3169 TCGv_i64 vra, vrb, vrt0, vrt1;
3170 REQUIRE_VECTOR(ctx);
3172 vra = tcg_temp_new_i64();
3173 vrb = tcg_temp_new_i64();
3174 vrt0 = tcg_temp_new_i64();
3175 vrt1 = tcg_temp_new_i64();
3177 get_avr64(vra, a->vra, even);
3178 get_avr64(vrb, a->vrb, even);
3179 gen_mul(vrt0, vrt1, vra, vrb);
3180 set_avr64(a->vrt, vrt0, false);
3181 set_avr64(a->vrt, vrt1, true);
3183 tcg_temp_free_i64(vra);
3184 tcg_temp_free_i64(vrb);
3185 tcg_temp_free_i64(vrt0);
3186 tcg_temp_free_i64(vrt1);
3191 static bool trans_VMULLD(DisasContext *ctx, arg_VX *a)
3193 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
3194 REQUIRE_VECTOR(ctx);
3196 tcg_gen_gvec_mul(MO_64, avr_full_offset(a->vrt), avr_full_offset(a->vra),
3197 avr_full_offset(a->vrb), 16, 16);
3202 TRANS_FLAGS(ALTIVEC, VMULESB, do_vx_helper, gen_helper_VMULESB)
3203 TRANS_FLAGS(ALTIVEC, VMULOSB, do_vx_helper, gen_helper_VMULOSB)
3204 TRANS_FLAGS(ALTIVEC, VMULEUB, do_vx_helper, gen_helper_VMULEUB)
3205 TRANS_FLAGS(ALTIVEC, VMULOUB, do_vx_helper, gen_helper_VMULOUB)
3206 TRANS_FLAGS(ALTIVEC, VMULESH, do_vx_helper, gen_helper_VMULESH)
3207 TRANS_FLAGS(ALTIVEC, VMULOSH, do_vx_helper, gen_helper_VMULOSH)
3208 TRANS_FLAGS(ALTIVEC, VMULEUH, do_vx_helper, gen_helper_VMULEUH)
3209 TRANS_FLAGS(ALTIVEC, VMULOUH, do_vx_helper, gen_helper_VMULOUH)
3210 TRANS_FLAGS2(ALTIVEC_207, VMULESW, do_vx_helper, gen_helper_VMULESW)
3211 TRANS_FLAGS2(ALTIVEC_207, VMULOSW, do_vx_helper, gen_helper_VMULOSW)
3212 TRANS_FLAGS2(ALTIVEC_207, VMULEUW, do_vx_helper, gen_helper_VMULEUW)
3213 TRANS_FLAGS2(ALTIVEC_207, VMULOUW, do_vx_helper, gen_helper_VMULOUW)
3214 TRANS_FLAGS2(ISA310, VMULESD, do_vx_vmuleo, true , tcg_gen_muls2_i64)
3215 TRANS_FLAGS2(ISA310, VMULOSD, do_vx_vmuleo, false, tcg_gen_muls2_i64)
3216 TRANS_FLAGS2(ISA310, VMULEUD, do_vx_vmuleo, true , tcg_gen_mulu2_i64)
3217 TRANS_FLAGS2(ISA310, VMULOUD, do_vx_vmuleo, false, tcg_gen_mulu2_i64)
3219 static void do_vx_vmulhw_i64(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b, bool sign)
3221 TCGv_i64 hh, lh, temp;
3223 hh = tcg_temp_new_i64();
3224 lh = tcg_temp_new_i64();
3225 temp = tcg_temp_new_i64();
3228 tcg_gen_ext32s_i64(lh, a);
3229 tcg_gen_ext32s_i64(temp, b);
3231 tcg_gen_ext32u_i64(lh, a);
3232 tcg_gen_ext32u_i64(temp, b);
3234 tcg_gen_mul_i64(lh, lh, temp);
3237 tcg_gen_sari_i64(hh, a, 32);
3238 tcg_gen_sari_i64(temp, b, 32);
3240 tcg_gen_shri_i64(hh, a, 32);
3241 tcg_gen_shri_i64(temp, b, 32);
3243 tcg_gen_mul_i64(hh, hh, temp);
3245 tcg_gen_shri_i64(lh, lh, 32);
3246 tcg_gen_deposit_i64(t, hh, lh, 0, 32);
3248 tcg_temp_free_i64(hh);
3249 tcg_temp_free_i64(lh);
3250 tcg_temp_free_i64(temp);
3253 static void do_vx_vmulhd_i64(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b, bool sign)
3257 tlow = tcg_temp_new_i64();
3259 tcg_gen_muls2_i64(tlow, t, a, b);
3261 tcg_gen_mulu2_i64(tlow, t, a, b);
3264 tcg_temp_free_i64(tlow);
3267 static bool do_vx_mulh(DisasContext *ctx, arg_VX *a, bool sign,
3268 void (*func)(TCGv_i64, TCGv_i64, TCGv_i64, bool))
3270 REQUIRE_INSNS_FLAGS2(ctx, ISA310);
3271 REQUIRE_VECTOR(ctx);
3273 TCGv_i64 vra, vrb, vrt;
3276 vra = tcg_temp_new_i64();
3277 vrb = tcg_temp_new_i64();
3278 vrt = tcg_temp_new_i64();
3280 for (i = 0; i < 2; i++) {
3281 get_avr64(vra, a->vra, i);
3282 get_avr64(vrb, a->vrb, i);
3283 get_avr64(vrt, a->vrt, i);
3285 func(vrt, vra, vrb, sign);
3287 set_avr64(a->vrt, vrt, i);
3290 tcg_temp_free_i64(vra);
3291 tcg_temp_free_i64(vrb);
3292 tcg_temp_free_i64(vrt);
3298 TRANS(VMULHSW, do_vx_mulh, true , do_vx_vmulhw_i64)
3299 TRANS(VMULHSD, do_vx_mulh, true , do_vx_vmulhd_i64)
3300 TRANS(VMULHUW, do_vx_mulh, false, do_vx_vmulhw_i64)
3301 TRANS(VMULHUD, do_vx_mulh, false, do_vx_vmulhd_i64)
3303 static bool do_vdiv_vmod(DisasContext *ctx, arg_VX *a, const int vece,
3304 void (*func_32)(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b),
3305 void (*func_64)(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b))
3307 const GVecGen3 op = {
3313 REQUIRE_VECTOR(ctx);
3315 tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra),
3316 avr_full_offset(a->vrb), 16, 16, &op);
3321 #define DIVU32(NAME, DIV) \
3322 static void NAME(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b) \
3324 TCGv_i32 zero = tcg_constant_i32(0); \
3325 TCGv_i32 one = tcg_constant_i32(1); \
3326 tcg_gen_movcond_i32(TCG_COND_EQ, b, b, zero, one, b); \
3330 #define DIVS32(NAME, DIV) \
3331 static void NAME(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b) \
3333 TCGv_i32 t0 = tcg_temp_new_i32(); \
3334 TCGv_i32 t1 = tcg_temp_new_i32(); \
3335 tcg_gen_setcondi_i32(TCG_COND_EQ, t0, a, INT32_MIN); \
3336 tcg_gen_setcondi_i32(TCG_COND_EQ, t1, b, -1); \
3337 tcg_gen_and_i32(t0, t0, t1); \
3338 tcg_gen_setcondi_i32(TCG_COND_EQ, t1, b, 0); \
3339 tcg_gen_or_i32(t0, t0, t1); \
3340 tcg_gen_movi_i32(t1, 0); \
3341 tcg_gen_movcond_i32(TCG_COND_NE, b, t0, t1, t0, b); \
3343 tcg_temp_free_i32(t0); \
3344 tcg_temp_free_i32(t1); \
3347 #define DIVU64(NAME, DIV) \
3348 static void NAME(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b) \
3350 TCGv_i64 zero = tcg_constant_i64(0); \
3351 TCGv_i64 one = tcg_constant_i64(1); \
3352 tcg_gen_movcond_i64(TCG_COND_EQ, b, b, zero, one, b); \
3356 #define DIVS64(NAME, DIV) \
3357 static void NAME(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b) \
3359 TCGv_i64 t0 = tcg_temp_new_i64(); \
3360 TCGv_i64 t1 = tcg_temp_new_i64(); \
3361 tcg_gen_setcondi_i64(TCG_COND_EQ, t0, a, INT64_MIN); \
3362 tcg_gen_setcondi_i64(TCG_COND_EQ, t1, b, -1); \
3363 tcg_gen_and_i64(t0, t0, t1); \
3364 tcg_gen_setcondi_i64(TCG_COND_EQ, t1, b, 0); \
3365 tcg_gen_or_i64(t0, t0, t1); \
3366 tcg_gen_movi_i64(t1, 0); \
3367 tcg_gen_movcond_i64(TCG_COND_NE, b, t0, t1, t0, b); \
3369 tcg_temp_free_i64(t0); \
3370 tcg_temp_free_i64(t1); \
3373 DIVS32(do_divsw, tcg_gen_div_i32)
3374 DIVU32(do_divuw, tcg_gen_divu_i32)
3375 DIVS64(do_divsd, tcg_gen_div_i64)
3376 DIVU64(do_divud, tcg_gen_divu_i64)
3378 TRANS_FLAGS2(ISA310, VDIVSW, do_vdiv_vmod, MO_32, do_divsw, NULL)
3379 TRANS_FLAGS2(ISA310, VDIVUW, do_vdiv_vmod, MO_32, do_divuw, NULL)
3380 TRANS_FLAGS2(ISA310, VDIVSD, do_vdiv_vmod, MO_64, NULL, do_divsd)
3381 TRANS_FLAGS2(ISA310, VDIVUD, do_vdiv_vmod, MO_64, NULL, do_divud)
3382 TRANS_FLAGS2(ISA310, VDIVSQ, do_vx_helper, gen_helper_VDIVSQ)
3383 TRANS_FLAGS2(ISA310, VDIVUQ, do_vx_helper, gen_helper_VDIVUQ)
3385 static void do_dives_i32(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b)
3387 TCGv_i64 val1, val2;
3389 val1 = tcg_temp_new_i64();
3390 val2 = tcg_temp_new_i64();
3392 tcg_gen_ext_i32_i64(val1, a);
3393 tcg_gen_ext_i32_i64(val2, b);
3396 tcg_gen_shli_i64(val1, val1, 32);
3397 tcg_gen_div_i64(val1, val1, val2);
3399 /* if quotient doesn't fit in 32 bits the result is undefined */
3400 tcg_gen_extrl_i64_i32(t, val1);
3402 tcg_temp_free_i64(val1);
3403 tcg_temp_free_i64(val2);
3406 static void do_diveu_i32(TCGv_i32 t, TCGv_i32 a, TCGv_i32 b)
3408 TCGv_i64 val1, val2;
3410 val1 = tcg_temp_new_i64();
3411 val2 = tcg_temp_new_i64();
3413 tcg_gen_extu_i32_i64(val1, a);
3414 tcg_gen_extu_i32_i64(val2, b);
3417 tcg_gen_shli_i64(val1, val1, 32);
3418 tcg_gen_divu_i64(val1, val1, val2);
3420 /* if quotient doesn't fit in 32 bits the result is undefined */
3421 tcg_gen_extrl_i64_i32(t, val1);
3423 tcg_temp_free_i64(val1);
3424 tcg_temp_free_i64(val2);
3427 DIVS32(do_divesw, do_dives_i32)
3428 DIVU32(do_diveuw, do_diveu_i32)
3430 DIVS32(do_modsw, tcg_gen_rem_i32)
3431 DIVU32(do_moduw, tcg_gen_remu_i32)
3432 DIVS64(do_modsd, tcg_gen_rem_i64)
3433 DIVU64(do_modud, tcg_gen_remu_i64)
3435 TRANS_FLAGS2(ISA310, VDIVESW, do_vdiv_vmod, MO_32, do_divesw, NULL)
3436 TRANS_FLAGS2(ISA310, VDIVEUW, do_vdiv_vmod, MO_32, do_diveuw, NULL)
3437 TRANS_FLAGS2(ISA310, VDIVESD, do_vx_helper, gen_helper_VDIVESD)
3438 TRANS_FLAGS2(ISA310, VDIVEUD, do_vx_helper, gen_helper_VDIVEUD)
3439 TRANS_FLAGS2(ISA310, VDIVESQ, do_vx_helper, gen_helper_VDIVESQ)
3440 TRANS_FLAGS2(ISA310, VDIVEUQ, do_vx_helper, gen_helper_VDIVEUQ)
3442 TRANS_FLAGS2(ISA310, VMODSW, do_vdiv_vmod, MO_32, do_modsw , NULL)
3443 TRANS_FLAGS2(ISA310, VMODUW, do_vdiv_vmod, MO_32, do_moduw, NULL)
3444 TRANS_FLAGS2(ISA310, VMODSD, do_vdiv_vmod, MO_64, NULL, do_modsd)
3445 TRANS_FLAGS2(ISA310, VMODUD, do_vdiv_vmod, MO_64, NULL, do_modud)
3446 TRANS_FLAGS2(ISA310, VMODSQ, do_vx_helper, gen_helper_VMODSQ)
3447 TRANS_FLAGS2(ISA310, VMODUQ, do_vx_helper, gen_helper_VMODUQ)
3459 #undef GEN_VX_LOGICAL
3460 #undef GEN_VX_LOGICAL_207
3462 #undef GEN_VXFORM_207
3463 #undef GEN_VXFORM_DUAL
3464 #undef GEN_VXRFORM_DUAL
3467 #undef GEN_VXFORM_VSPLTI
3468 #undef GEN_VXFORM_NOA
3469 #undef GEN_VXFORM_UIMM
3470 #undef GEN_VAFORM_PAIRED