]> git.ipfire.org Git - thirdparty/qemu.git/blob - target/ppc/translate/vmx-impl.c.inc
projects / thirdparty / qemu.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge remote-tracking branch 'remotes/dgibson/tags/ppc-for-5.2-20200818' into staging
[thirdparty/qemu.git] / target / ppc / translate / vmx-impl.c.inc
1 /*
2 * translate/vmx-impl.c
3 *
4 * Altivec/VMX translation
5 */
6
7 /*** Altivec vector extension ***/
8 /* Altivec registers moves */
9
10 static 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) \
18 static 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 tcg_temp_free(EA); \
49 tcg_temp_free_i64(avr); \
50 }
51
52 #define GEN_VR_STX(name, opc2, opc3) \
53 static void gen_st##name(DisasContext *ctx) \
54 { \
55 TCGv EA; \
56 TCGv_i64 avr; \
57 if (unlikely(!ctx->altivec_enabled)) { \
58 gen_exception(ctx, POWERPC_EXCP_VPU); \
59 return; \
60 } \
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); \
66 /* \
67 * We only need to swap high and low halves. gen_qemu_st64_i64 \
68 * does necessary 64-bit byteswap already. \
69 */ \
70 if (ctx->le_mode) { \
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); \
76 } else { \
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); \
82 } \
83 tcg_temp_free(EA); \
84 tcg_temp_free_i64(avr); \
85 }
86
87 #define GEN_VR_LVE(name, opc2, opc3, size) \
88 static void gen_lve##name(DisasContext *ctx) \
89 { \
90 TCGv EA; \
91 TCGv_ptr rs; \
92 if (unlikely(!ctx->altivec_enabled)) { \
93 gen_exception(ctx, POWERPC_EXCP_VPU); \
94 return; \
95 } \
96 gen_set_access_type(ctx, ACCESS_INT); \
97 EA = tcg_temp_new(); \
98 gen_addr_reg_index(ctx, EA); \
99 if (size > 1) { \
100 tcg_gen_andi_tl(EA, EA, ~(size - 1)); \
101 } \
102 rs = gen_avr_ptr(rS(ctx->opcode)); \
103 gen_helper_lve##name(cpu_env, rs, EA); \
104 tcg_temp_free(EA); \
105 tcg_temp_free_ptr(rs); \
106 }
107
108 #define GEN_VR_STVE(name, opc2, opc3, size) \
109 static void gen_stve##name(DisasContext *ctx) \
110 { \
111 TCGv EA; \
112 TCGv_ptr rs; \
113 if (unlikely(!ctx->altivec_enabled)) { \
114 gen_exception(ctx, POWERPC_EXCP_VPU); \
115 return; \
116 } \
117 gen_set_access_type(ctx, ACCESS_INT); \
118 EA = tcg_temp_new(); \
119 gen_addr_reg_index(ctx, EA); \
120 if (size > 1) { \
121 tcg_gen_andi_tl(EA, EA, ~(size - 1)); \
122 } \
123 rs = gen_avr_ptr(rS(ctx->opcode)); \
124 gen_helper_stve##name(cpu_env, rs, EA); \
125 tcg_temp_free(EA); \
126 tcg_temp_free_ptr(rs); \
127 }
128
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);
132
133 GEN_VR_LVE(bx, 0x07, 0x00, 1);
134 GEN_VR_LVE(hx, 0x07, 0x01, 2);
135 GEN_VR_LVE(wx, 0x07, 0x02, 4);
136
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);
140
141 GEN_VR_STVE(bx, 0x07, 0x04, 1);
142 GEN_VR_STVE(hx, 0x07, 0x05, 2);
143 GEN_VR_STVE(wx, 0x07, 0x06, 4);
144
145 static void gen_mfvscr(DisasContext *ctx)
146 {
147 TCGv_i32 t;
148 TCGv_i64 avr;
149 if (unlikely(!ctx->altivec_enabled)) {
150 gen_exception(ctx, POWERPC_EXCP_VPU);
151 return;
152 }
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);
162 }
163
164 static void gen_mtvscr(DisasContext *ctx)
165 {
166 TCGv_i32 val;
167 int bofs;
168
169 if (unlikely(!ctx->altivec_enabled)) {
170 gen_exception(ctx, POWERPC_EXCP_VPU);
171 return;
172 }
173
174 val = tcg_temp_new_i32();
175 bofs = avr_full_offset(rB(ctx->opcode));
176 #ifdef HOST_WORDS_BIGENDIAN
177 bofs += 3 * 4;
178 #endif
179
180 tcg_gen_ld_i32(val, cpu_env, bofs);
181 gen_helper_mtvscr(cpu_env, val);
182 tcg_temp_free_i32(val);
183 }
184
185 #define GEN_VX_VMUL10(name, add_cin, ret_carry) \
186 static void glue(gen_, name)(DisasContext *ctx) \
187 { \
188 TCGv_i64 t0; \
189 TCGv_i64 t1; \
190 TCGv_i64 t2; \
191 TCGv_i64 avr; \
192 TCGv_i64 ten, z; \
193 \
194 if (unlikely(!ctx->altivec_enabled)) { \
195 gen_exception(ctx, POWERPC_EXCP_VPU); \
196 return; \
197 } \
198 \
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); \
205 \
206 if (add_cin) { \
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); \
213 } else { \
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); \
217 } \
218 \
219 if (ret_carry) { \
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); \
225 } else { \
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); \
230 } \
231 \
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); \
238 } \
239
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);
244
245 #define GEN_VXFORM_V(name, vece, tcg_op, opc2, opc3) \
246 static void glue(gen_, name)(DisasContext *ctx) \
247 { \
248 if (unlikely(!ctx->altivec_enabled)) { \
249 gen_exception(ctx, POWERPC_EXCP_VPU); \
250 return; \
251 } \
252 \
253 tcg_op(vece, \
254 avr_full_offset(rD(ctx->opcode)), \
255 avr_full_offset(rA(ctx->opcode)), \
256 avr_full_offset(rB(ctx->opcode)), \
257 16, 16); \
258 }
259
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);
269
270 #define GEN_VXFORM(name, opc2, opc3) \
271 static void glue(gen_, name)(DisasContext *ctx) \
272 { \
273 TCGv_ptr ra, rb, rd; \
274 if (unlikely(!ctx->altivec_enabled)) { \
275 gen_exception(ctx, POWERPC_EXCP_VPU); \
276 return; \
277 } \
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); \
285 }
286
287 #define GEN_VXFORM_TRANS(name, opc2, opc3) \
288 static void glue(gen_, name)(DisasContext *ctx) \
289 { \
290 if (unlikely(!ctx->altivec_enabled)) { \
291 gen_exception(ctx, POWERPC_EXCP_VPU); \
292 return; \
293 } \
294 trans_##name(ctx); \
295 }
296
297 #define GEN_VXFORM_ENV(name, opc2, opc3) \
298 static void glue(gen_, name)(DisasContext *ctx) \
299 { \
300 TCGv_ptr ra, rb, rd; \
301 if (unlikely(!ctx->altivec_enabled)) { \
302 gen_exception(ctx, POWERPC_EXCP_VPU); \
303 return; \
304 } \
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); \
312 }
313
314 #define GEN_VXFORM3(name, opc2, opc3) \
315 static void glue(gen_, name)(DisasContext *ctx) \
316 { \
317 TCGv_ptr ra, rb, rc, rd; \
318 if (unlikely(!ctx->altivec_enabled)) { \
319 gen_exception(ctx, POWERPC_EXCP_VPU); \
320 return; \
321 } \
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); \
331 }
332
333 /*
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
337 * each instruction.
338 */
339 #define GEN_VXFORM_DUAL(name0, flg0, flg2_0, name1, flg1, flg2_1) \
340 static void glue(gen_, name0##_##name1)(DisasContext *ctx) \
341 { \
342 if ((Rc(ctx->opcode) == 0) && \
343 ((ctx->insns_flags & flg0) || (ctx->insns_flags2 & flg2_0))) { \
344 gen_##name0(ctx); \
345 } else if ((Rc(ctx->opcode) == 1) && \
346 ((ctx->insns_flags & flg1) || (ctx->insns_flags2 & flg2_1))) { \
347 gen_##name1(ctx); \
348 } else { \
349 gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL); \
350 } \
351 }
352
353 /*
354 * We use this macro if one instruction is realized with direct
355 * translation, and second one with helper.
356 */
357 #define GEN_VXFORM_TRANS_DUAL(name0, flg0, flg2_0, name1, flg1, flg2_1)\
358 static void glue(gen_, name0##_##name1)(DisasContext *ctx) \
359 { \
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); \
364 return; \
365 } \
366 trans_##name0(ctx); \
367 } else if ((Rc(ctx->opcode) == 1) && \
368 ((ctx->insns_flags & flg1) || (ctx->insns_flags2 & flg2_1))) { \
369 gen_##name1(ctx); \
370 } else { \
371 gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL); \
372 } \
373 }
374
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) \
379 { \
380 if ((Rc(ctx->opcode) == 0) && \
381 ((ctx->insns_flags & flg0) || (ctx->insns_flags2 & flg2_0)) && \
382 !(ctx->opcode & inval0)) { \
383 gen_##name0(ctx); \
384 } else if ((Rc(ctx->opcode) == 1) && \
385 ((ctx->insns_flags & flg1) || (ctx->insns_flags2 & flg2_1)) && \
386 !(ctx->opcode & inval1)) { \
387 gen_##name1(ctx); \
388 } else { \
389 gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL); \
390 } \
391 }
392
393 #define GEN_VXFORM_HETRO(name, opc2, opc3) \
394 static void glue(gen_, name)(DisasContext *ctx) \
395 { \
396 TCGv_ptr rb; \
397 if (unlikely(!ctx->altivec_enabled)) { \
398 gen_exception(ctx, POWERPC_EXCP_VPU); \
399 return; \
400 } \
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); \
404 }
405
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);
455
456 static void trans_vmrgew(DisasContext *ctx)
457 {
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();
463
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);
469
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);
475
476 tcg_temp_free_i64(tmp);
477 tcg_temp_free_i64(avr);
478 }
479
480 static void trans_vmrgow(DisasContext *ctx)
481 {
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();
488
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);
493
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);
498
499 tcg_temp_free_i64(t0);
500 tcg_temp_free_i64(t1);
501 tcg_temp_free_i64(avr);
502 }
503
504 /*
505 * lvsl VRT,RA,RB - Load Vector for Shift Left
506 *
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.
510 */
511 static void trans_lvsl(DisasContext *ctx)
512 {
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();
517
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);
522
523 /*
524 * Create bytes sh:sh+7 of X(from description) and place them in
525 * higher doubleword of vD.
526 */
527 tcg_gen_muli_i64(sh, sh, 0x0101010101010101ULL);
528 tcg_gen_addi_i64(result, sh, 0x0001020304050607ull);
529 set_avr64(VT, result, true);
530 /*
531 * Create bytes sh+8:sh+15 of X(from description) and place them in
532 * lower doubleword of vD.
533 */
534 tcg_gen_addi_i64(result, sh, 0x08090a0b0c0d0e0fULL);
535 set_avr64(VT, result, false);
536
537 tcg_temp_free_i64(result);
538 tcg_temp_free_i64(sh);
539 tcg_temp_free(EA);
540 }
541
542 /*
543 * lvsr VRT,RA,RB - Load Vector for Shift Right
544 *
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.
548 */
549 static void trans_lvsr(DisasContext *ctx)
550 {
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();
555
556
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);
561
562 /*
563 * Create bytes (16-sh):(23-sh) of X(from description) and place them in
564 * higher doubleword of vD.
565 */
566 tcg_gen_muli_i64(sh, sh, 0x0101010101010101ULL);
567 tcg_gen_subfi_i64(result, 0x1011121314151617ULL, sh);
568 set_avr64(VT, result, true);
569 /*
570 * Create bytes (24-sh):(32-sh) of X(from description) and place them in
571 * lower doubleword of vD.
572 */
573 tcg_gen_subfi_i64(result, 0x18191a1b1c1d1e1fULL, sh);
574 set_avr64(VT, result, false);
575
576 tcg_temp_free_i64(result);
577 tcg_temp_free_i64(sh);
578 tcg_temp_free(EA);
579 }
580
581 /*
582 * vsl VRT,VRA,VRB - Vector Shift Left
583 *
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.
587 */
588 static void trans_vsl(DisasContext *ctx)
589 {
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();
597
598 /* Place bits 125-127 of vB in 'sh'. */
599 get_avr64(avr, VB, false);
600 tcg_gen_andi_i64(sh, avr, 0x07ULL);
601
602 /*
603 * Save highest 'sh' bits of lower doubleword element of vA in variable
604 * 'carry' and perform shift on lower doubleword.
605 */
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);
612
613 /*
614 * Perform shift on higher doubleword element of vA and replace lowest
615 * 'sh' bits with 'carry'.
616 */
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);
621
622 tcg_temp_free_i64(avr);
623 tcg_temp_free_i64(sh);
624 tcg_temp_free_i64(carry);
625 tcg_temp_free_i64(tmp);
626 }
627
628 /*
629 * vsr VRT,VRA,VRB - Vector Shift Right
630 *
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.
634 */
635 static void trans_vsr(DisasContext *ctx)
636 {
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();
644
645 /* Place bits 125-127 of vB in 'sh'. */
646 get_avr64(avr, VB, false);
647 tcg_gen_andi_i64(sh, avr, 0x07ULL);
648
649 /*
650 * Save lowest 'sh' bits of higher doubleword element of vA in variable
651 * 'carry' and perform shift on higher doubleword.
652 */
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);
659 /*
660 * Perform shift on lower doubleword element of vA and replace highest
661 * 'sh' bits with 'carry'.
662 */
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);
667
668 tcg_temp_free_i64(avr);
669 tcg_temp_free_i64(sh);
670 tcg_temp_free_i64(carry);
671 tcg_temp_free_i64(tmp);
672 }
673
674 /*
675 * vgbbd VRT,VRB - Vector Gather Bits by Bytes by Doubleword
676 *
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.
680 *
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
683 * arrays are used):
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.
699 */
700 static void trans_vgbbd(DisasContext *ctx)
701 {
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;
706 int i, j;
707
708 TCGv_i64 result[2];
709 result[0] = tcg_temp_new_i64();
710 result[1] = tcg_temp_new_i64();
711 TCGv_i64 avr[2];
712 avr[0] = tcg_temp_new_i64();
713 avr[1] = tcg_temp_new_i64();
714 TCGv_i64 tcg_mask = tcg_temp_new_i64();
715
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);
720 }
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);
727 }
728 }
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);
735 }
736 }
737 for (j = 0; j < 2; j++) {
738 set_avr64(VT, result[j], j);
739 }
740
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]);
747 }
748
749 /*
750 * vclzw VRT,VRB - Vector Count Leading Zeros Word
751 *
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
754 * register.
755 */
756 static void trans_vclzw(DisasContext *ctx)
757 {
758 int VT = rD(ctx->opcode);
759 int VB = rB(ctx->opcode);
760 TCGv_i32 tmp = tcg_temp_new_i32();
761 int i;
762
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);
770 }
771
772 tcg_temp_free_i32(tmp);
773 }
774
775 /*
776 * vclzd VRT,VRB - Vector Count Leading Zeros Doubleword
777 *
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
780 * register.
781 */
782 static void trans_vclzd(DisasContext *ctx)
783 {
784 int VT = rD(ctx->opcode);
785 int VB = rB(ctx->opcode);
786 TCGv_i64 avr = tcg_temp_new_i64();
787
788 /* high doubleword */
789 get_avr64(avr, VB, true);
790 tcg_gen_clzi_i64(avr, avr, 64);
791 set_avr64(VT, avr, true);
792
793 /* low doubleword */
794 get_avr64(avr, VB, false);
795 tcg_gen_clzi_i64(avr, avr, 64);
796 set_avr64(VT, avr, false);
797
798 tcg_temp_free_i64(avr);
799 }
800
801 GEN_VXFORM(vmuloub, 4, 0);
802 GEN_VXFORM(vmulouh, 4, 1);
803 GEN_VXFORM(vmulouw, 4, 2);
804 GEN_VXFORM_V(vmuluwm, MO_32, tcg_gen_gvec_mul, 4, 2);
805 GEN_VXFORM_DUAL(vmulouw, PPC_ALTIVEC, PPC_NONE,
806 vmuluwm, PPC_NONE, PPC2_ALTIVEC_207)
807 GEN_VXFORM(vmulosb, 4, 4);
808 GEN_VXFORM(vmulosh, 4, 5);
809 GEN_VXFORM(vmulosw, 4, 6);
810 GEN_VXFORM_V(vmulld, MO_64, tcg_gen_gvec_mul, 4, 7);
811 GEN_VXFORM(vmuleub, 4, 8);
812 GEN_VXFORM(vmuleuh, 4, 9);
813 GEN_VXFORM(vmuleuw, 4, 10);
814 GEN_VXFORM(vmulhuw, 4, 10);
815 GEN_VXFORM(vmulhud, 4, 11);
816 GEN_VXFORM_DUAL(vmuleuw, PPC_ALTIVEC, PPC_NONE,
817 vmulhuw, PPC_NONE, PPC2_ISA310);
818 GEN_VXFORM(vmulesb, 4, 12);
819 GEN_VXFORM(vmulesh, 4, 13);
820 GEN_VXFORM(vmulesw, 4, 14);
821 GEN_VXFORM(vmulhsw, 4, 14);
822 GEN_VXFORM_DUAL(vmulesw, PPC_ALTIVEC, PPC_NONE,
823 vmulhsw, PPC_NONE, PPC2_ISA310);
824 GEN_VXFORM(vmulhsd, 4, 15);
825 GEN_VXFORM_V(vslb, MO_8, tcg_gen_gvec_shlv, 2, 4);
826 GEN_VXFORM_V(vslh, MO_16, tcg_gen_gvec_shlv, 2, 5);
827 GEN_VXFORM_V(vslw, MO_32, tcg_gen_gvec_shlv, 2, 6);
828 GEN_VXFORM(vrlwnm, 2, 6);
829 GEN_VXFORM_DUAL(vslw, PPC_ALTIVEC, PPC_NONE, \
830 vrlwnm, PPC_NONE, PPC2_ISA300)
831 GEN_VXFORM_V(vsld, MO_64, tcg_gen_gvec_shlv, 2, 23);
832 GEN_VXFORM_V(vsrb, MO_8, tcg_gen_gvec_shrv, 2, 8);
833 GEN_VXFORM_V(vsrh, MO_16, tcg_gen_gvec_shrv, 2, 9);
834 GEN_VXFORM_V(vsrw, MO_32, tcg_gen_gvec_shrv, 2, 10);
835 GEN_VXFORM_V(vsrd, MO_64, tcg_gen_gvec_shrv, 2, 27);
836 GEN_VXFORM_V(vsrab, MO_8, tcg_gen_gvec_sarv, 2, 12);
837 GEN_VXFORM_V(vsrah, MO_16, tcg_gen_gvec_sarv, 2, 13);
838 GEN_VXFORM_V(vsraw, MO_32, tcg_gen_gvec_sarv, 2, 14);
839 GEN_VXFORM_V(vsrad, MO_64, tcg_gen_gvec_sarv, 2, 15);
840 GEN_VXFORM(vsrv, 2, 28);
841 GEN_VXFORM(vslv, 2, 29);
842 GEN_VXFORM(vslo, 6, 16);
843 GEN_VXFORM(vsro, 6, 17);
844 GEN_VXFORM(vaddcuw, 0, 6);
845 GEN_VXFORM(vsubcuw, 0, 22);
846
847 #define GEN_VXFORM_SAT(NAME, VECE, NORM, SAT, OPC2, OPC3) \
848 static void glue(glue(gen_, NAME), _vec)(unsigned vece, TCGv_vec t, \
849 TCGv_vec sat, TCGv_vec a, \
850 TCGv_vec b) \
851 { \
852 TCGv_vec x = tcg_temp_new_vec_matching(t); \
853 glue(glue(tcg_gen_, NORM), _vec)(VECE, x, a, b); \
854 glue(glue(tcg_gen_, SAT), _vec)(VECE, t, a, b); \
855 tcg_gen_cmp_vec(TCG_COND_NE, VECE, x, x, t); \
856 tcg_gen_or_vec(VECE, sat, sat, x); \
857 tcg_temp_free_vec(x); \
858 } \
859 static void glue(gen_, NAME)(DisasContext *ctx) \
860 { \
861 static const TCGOpcode vecop_list[] = { \
862 glue(glue(INDEX_op_, NORM), _vec), \
863 glue(glue(INDEX_op_, SAT), _vec), \
864 INDEX_op_cmp_vec, 0 \
865 }; \
866 static const GVecGen4 g = { \
867 .fniv = glue(glue(gen_, NAME), _vec), \
868 .fno = glue(gen_helper_, NAME), \
869 .opt_opc = vecop_list, \
870 .write_aofs = true, \
871 .vece = VECE, \
872 }; \
873 if (unlikely(!ctx->altivec_enabled)) { \
874 gen_exception(ctx, POWERPC_EXCP_VPU); \
875 return; \
876 } \
877 tcg_gen_gvec_4(avr_full_offset(rD(ctx->opcode)), \
878 offsetof(CPUPPCState, vscr_sat), \
879 avr_full_offset(rA(ctx->opcode)), \
880 avr_full_offset(rB(ctx->opcode)), \
881 16, 16, &g); \
882 }
883
884 GEN_VXFORM_SAT(vaddubs, MO_8, add, usadd, 0, 8);
885 GEN_VXFORM_DUAL_EXT(vaddubs, PPC_ALTIVEC, PPC_NONE, 0, \
886 vmul10uq, PPC_NONE, PPC2_ISA300, 0x0000F800)
887 GEN_VXFORM_SAT(vadduhs, MO_16, add, usadd, 0, 9);
888 GEN_VXFORM_DUAL(vadduhs, PPC_ALTIVEC, PPC_NONE, \
889 vmul10euq, PPC_NONE, PPC2_ISA300)
890 GEN_VXFORM_SAT(vadduws, MO_32, add, usadd, 0, 10);
891 GEN_VXFORM_SAT(vaddsbs, MO_8, add, ssadd, 0, 12);
892 GEN_VXFORM_SAT(vaddshs, MO_16, add, ssadd, 0, 13);
893 GEN_VXFORM_SAT(vaddsws, MO_32, add, ssadd, 0, 14);
894 GEN_VXFORM_SAT(vsububs, MO_8, sub, ussub, 0, 24);
895 GEN_VXFORM_SAT(vsubuhs, MO_16, sub, ussub, 0, 25);
896 GEN_VXFORM_SAT(vsubuws, MO_32, sub, ussub, 0, 26);
897 GEN_VXFORM_SAT(vsubsbs, MO_8, sub, sssub, 0, 28);
898 GEN_VXFORM_SAT(vsubshs, MO_16, sub, sssub, 0, 29);
899 GEN_VXFORM_SAT(vsubsws, MO_32, sub, sssub, 0, 30);
900 GEN_VXFORM(vadduqm, 0, 4);
901 GEN_VXFORM(vaddcuq, 0, 5);
902 GEN_VXFORM3(vaddeuqm, 30, 0);
903 GEN_VXFORM3(vaddecuq, 30, 0);
904 GEN_VXFORM_DUAL(vaddeuqm, PPC_NONE, PPC2_ALTIVEC_207, \
905 vaddecuq, PPC_NONE, PPC2_ALTIVEC_207)
906 GEN_VXFORM(vsubuqm, 0, 20);
907 GEN_VXFORM(vsubcuq, 0, 21);
908 GEN_VXFORM3(vsubeuqm, 31, 0);
909 GEN_VXFORM3(vsubecuq, 31, 0);
910 GEN_VXFORM_DUAL(vsubeuqm, PPC_NONE, PPC2_ALTIVEC_207, \
911 vsubecuq, PPC_NONE, PPC2_ALTIVEC_207)
912 GEN_VXFORM_V(vrlb, MO_8, tcg_gen_gvec_rotlv, 2, 0);
913 GEN_VXFORM_V(vrlh, MO_16, tcg_gen_gvec_rotlv, 2, 1);
914 GEN_VXFORM_V(vrlw, MO_32, tcg_gen_gvec_rotlv, 2, 2);
915 GEN_VXFORM(vrlwmi, 2, 2);
916 GEN_VXFORM_DUAL(vrlw, PPC_ALTIVEC, PPC_NONE, \
917 vrlwmi, PPC_NONE, PPC2_ISA300)
918 GEN_VXFORM_V(vrld, MO_64, tcg_gen_gvec_rotlv, 2, 3);
919 GEN_VXFORM(vrldmi, 2, 3);
920 GEN_VXFORM_DUAL(vrld, PPC_NONE, PPC2_ALTIVEC_207, \
921 vrldmi, PPC_NONE, PPC2_ISA300)
922 GEN_VXFORM_TRANS(vsl, 2, 7);
923 GEN_VXFORM(vrldnm, 2, 7);
924 GEN_VXFORM_DUAL(vsl, PPC_ALTIVEC, PPC_NONE, \
925 vrldnm, PPC_NONE, PPC2_ISA300)
926 GEN_VXFORM_TRANS(vsr, 2, 11);
927 GEN_VXFORM_ENV(vpkuhum, 7, 0);
928 GEN_VXFORM_ENV(vpkuwum, 7, 1);
929 GEN_VXFORM_ENV(vpkudum, 7, 17);
930 GEN_VXFORM_ENV(vpkuhus, 7, 2);
931 GEN_VXFORM_ENV(vpkuwus, 7, 3);
932 GEN_VXFORM_ENV(vpkudus, 7, 19);
933 GEN_VXFORM_ENV(vpkshus, 7, 4);
934 GEN_VXFORM_ENV(vpkswus, 7, 5);
935 GEN_VXFORM_ENV(vpksdus, 7, 21);
936 GEN_VXFORM_ENV(vpkshss, 7, 6);
937 GEN_VXFORM_ENV(vpkswss, 7, 7);
938 GEN_VXFORM_ENV(vpksdss, 7, 23);
939 GEN_VXFORM(vpkpx, 7, 12);
940 GEN_VXFORM_ENV(vsum4ubs, 4, 24);
941 GEN_VXFORM_ENV(vsum4sbs, 4, 28);
942 GEN_VXFORM_ENV(vsum4shs, 4, 25);
943 GEN_VXFORM_ENV(vsum2sws, 4, 26);
944 GEN_VXFORM_ENV(vsumsws, 4, 30);
945 GEN_VXFORM_ENV(vaddfp, 5, 0);
946 GEN_VXFORM_ENV(vsubfp, 5, 1);
947 GEN_VXFORM_ENV(vmaxfp, 5, 16);
948 GEN_VXFORM_ENV(vminfp, 5, 17);
949 GEN_VXFORM_HETRO(vextublx, 6, 24)
950 GEN_VXFORM_HETRO(vextuhlx, 6, 25)
951 GEN_VXFORM_HETRO(vextuwlx, 6, 26)
952 GEN_VXFORM_TRANS_DUAL(vmrgow, PPC_NONE, PPC2_ALTIVEC_207,
953 vextuwlx, PPC_NONE, PPC2_ISA300)
954 GEN_VXFORM_HETRO(vextubrx, 6, 28)
955 GEN_VXFORM_HETRO(vextuhrx, 6, 29)
956 GEN_VXFORM_HETRO(vextuwrx, 6, 30)
957 GEN_VXFORM_TRANS(lvsl, 6, 31)
958 GEN_VXFORM_TRANS(lvsr, 6, 32)
959 GEN_VXFORM_TRANS_DUAL(vmrgew, PPC_NONE, PPC2_ALTIVEC_207,
960 vextuwrx, PPC_NONE, PPC2_ISA300)
961
962 #define GEN_VXRFORM1(opname, name, str, opc2, opc3) \
963 static void glue(gen_, name)(DisasContext *ctx) \
964 { \
965 TCGv_ptr ra, rb, rd; \
966 if (unlikely(!ctx->altivec_enabled)) { \
967 gen_exception(ctx, POWERPC_EXCP_VPU); \
968 return; \
969 } \
970 ra = gen_avr_ptr(rA(ctx->opcode)); \
971 rb = gen_avr_ptr(rB(ctx->opcode)); \
972 rd = gen_avr_ptr(rD(ctx->opcode)); \
973 gen_helper_##opname(cpu_env, rd, ra, rb); \
974 tcg_temp_free_ptr(ra); \
975 tcg_temp_free_ptr(rb); \
976 tcg_temp_free_ptr(rd); \
977 }
978
979 #define GEN_VXRFORM(name, opc2, opc3) \
980 GEN_VXRFORM1(name, name, #name, opc2, opc3) \
981 GEN_VXRFORM1(name##_dot, name##_, #name ".", opc2, (opc3 | (0x1 << 4)))
982
983 /*
984 * Support for Altivec instructions that use bit 31 (Rc) as an opcode
985 * bit but also use bit 21 as an actual Rc bit. In general, thse pairs
986 * come from different versions of the ISA, so we must also support a
987 * pair of flags for each instruction.
988 */
989 #define GEN_VXRFORM_DUAL(name0, flg0, flg2_0, name1, flg1, flg2_1) \
990 static void glue(gen_, name0##_##name1)(DisasContext *ctx) \
991 { \
992 if ((Rc(ctx->opcode) == 0) && \
993 ((ctx->insns_flags & flg0) || (ctx->insns_flags2 & flg2_0))) { \
994 if (Rc21(ctx->opcode) == 0) { \
995 gen_##name0(ctx); \
996 } else { \
997 gen_##name0##_(ctx); \
998 } \
999 } else if ((Rc(ctx->opcode) == 1) && \
1000 ((ctx->insns_flags & flg1) || (ctx->insns_flags2 & flg2_1))) { \
1001 if (Rc21(ctx->opcode) == 0) { \
1002 gen_##name1(ctx); \
1003 } else { \
1004 gen_##name1##_(ctx); \
1005 } \
1006 } else { \
1007 gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL); \
1008 } \
1009 }
1010
1011 GEN_VXRFORM(vcmpequb, 3, 0)
1012 GEN_VXRFORM(vcmpequh, 3, 1)
1013 GEN_VXRFORM(vcmpequw, 3, 2)
1014 GEN_VXRFORM(vcmpequd, 3, 3)
1015 GEN_VXRFORM(vcmpnezb, 3, 4)
1016 GEN_VXRFORM(vcmpnezh, 3, 5)
1017 GEN_VXRFORM(vcmpnezw, 3, 6)
1018 GEN_VXRFORM(vcmpgtsb, 3, 12)
1019 GEN_VXRFORM(vcmpgtsh, 3, 13)
1020 GEN_VXRFORM(vcmpgtsw, 3, 14)
1021 GEN_VXRFORM(vcmpgtsd, 3, 15)
1022 GEN_VXRFORM(vcmpgtub, 3, 8)
1023 GEN_VXRFORM(vcmpgtuh, 3, 9)
1024 GEN_VXRFORM(vcmpgtuw, 3, 10)
1025 GEN_VXRFORM(vcmpgtud, 3, 11)
1026 GEN_VXRFORM(vcmpeqfp, 3, 3)
1027 GEN_VXRFORM(vcmpgefp, 3, 7)
1028 GEN_VXRFORM(vcmpgtfp, 3, 11)
1029 GEN_VXRFORM(vcmpbfp, 3, 15)
1030 GEN_VXRFORM(vcmpneb, 3, 0)
1031 GEN_VXRFORM(vcmpneh, 3, 1)
1032 GEN_VXRFORM(vcmpnew, 3, 2)
1033
1034 GEN_VXRFORM_DUAL(vcmpequb, PPC_ALTIVEC, PPC_NONE, \
1035 vcmpneb, PPC_NONE, PPC2_ISA300)
1036 GEN_VXRFORM_DUAL(vcmpequh, PPC_ALTIVEC, PPC_NONE, \
1037 vcmpneh, PPC_NONE, PPC2_ISA300)
1038 GEN_VXRFORM_DUAL(vcmpequw, PPC_ALTIVEC, PPC_NONE, \
1039 vcmpnew, PPC_NONE, PPC2_ISA300)
1040 GEN_VXRFORM_DUAL(vcmpeqfp, PPC_ALTIVEC, PPC_NONE, \
1041 vcmpequd, PPC_NONE, PPC2_ALTIVEC_207)
1042 GEN_VXRFORM_DUAL(vcmpbfp, PPC_ALTIVEC, PPC_NONE, \
1043 vcmpgtsd, PPC_NONE, PPC2_ALTIVEC_207)
1044 GEN_VXRFORM_DUAL(vcmpgtfp, PPC_ALTIVEC, PPC_NONE, \
1045 vcmpgtud, PPC_NONE, PPC2_ALTIVEC_207)
1046
1047 static void gen_vsplti(DisasContext *ctx, int vece)
1048 {
1049 int simm;
1050
1051 if (unlikely(!ctx->altivec_enabled)) {
1052 gen_exception(ctx, POWERPC_EXCP_VPU);
1053 return;
1054 }
1055
1056 simm = SIMM5(ctx->opcode);
1057 tcg_gen_gvec_dup_imm(vece, avr_full_offset(rD(ctx->opcode)), 16, 16, simm);
1058 }
1059
1060 #define GEN_VXFORM_VSPLTI(name, vece, opc2, opc3) \
1061 static void glue(gen_, name)(DisasContext *ctx) { gen_vsplti(ctx, vece); }
1062
1063 GEN_VXFORM_VSPLTI(vspltisb, MO_8, 6, 12);
1064 GEN_VXFORM_VSPLTI(vspltish, MO_16, 6, 13);
1065 GEN_VXFORM_VSPLTI(vspltisw, MO_32, 6, 14);
1066
1067 #define GEN_VXFORM_NOA(name, opc2, opc3) \
1068 static void glue(gen_, name)(DisasContext *ctx) \
1069 { \
1070 TCGv_ptr rb, rd; \
1071 if (unlikely(!ctx->altivec_enabled)) { \
1072 gen_exception(ctx, POWERPC_EXCP_VPU); \
1073 return; \
1074 } \
1075 rb = gen_avr_ptr(rB(ctx->opcode)); \
1076 rd = gen_avr_ptr(rD(ctx->opcode)); \
1077 gen_helper_##name(rd, rb); \
1078 tcg_temp_free_ptr(rb); \
1079 tcg_temp_free_ptr(rd); \
1080 }
1081
1082 #define GEN_VXFORM_NOA_ENV(name, opc2, opc3) \
1083 static void glue(gen_, name)(DisasContext *ctx) \
1084 { \
1085 TCGv_ptr rb, rd; \
1086 \
1087 if (unlikely(!ctx->altivec_enabled)) { \
1088 gen_exception(ctx, POWERPC_EXCP_VPU); \
1089 return; \
1090 } \
1091 rb = gen_avr_ptr(rB(ctx->opcode)); \
1092 rd = gen_avr_ptr(rD(ctx->opcode)); \
1093 gen_helper_##name(cpu_env, rd, rb); \
1094 tcg_temp_free_ptr(rb); \
1095 tcg_temp_free_ptr(rd); \
1096 }
1097
1098 #define GEN_VXFORM_NOA_2(name, opc2, opc3, opc4) \
1099 static void glue(gen_, name)(DisasContext *ctx) \
1100 { \
1101 TCGv_ptr rb, rd; \
1102 if (unlikely(!ctx->altivec_enabled)) { \
1103 gen_exception(ctx, POWERPC_EXCP_VPU); \
1104 return; \
1105 } \
1106 rb = gen_avr_ptr(rB(ctx->opcode)); \
1107 rd = gen_avr_ptr(rD(ctx->opcode)); \
1108 gen_helper_##name(rd, rb); \
1109 tcg_temp_free_ptr(rb); \
1110 tcg_temp_free_ptr(rd); \
1111 }
1112
1113 #define GEN_VXFORM_NOA_3(name, opc2, opc3, opc4) \
1114 static void glue(gen_, name)(DisasContext *ctx) \
1115 { \
1116 TCGv_ptr rb; \
1117 if (unlikely(!ctx->altivec_enabled)) { \
1118 gen_exception(ctx, POWERPC_EXCP_VPU); \
1119 return; \
1120 } \
1121 rb = gen_avr_ptr(rB(ctx->opcode)); \
1122 gen_helper_##name(cpu_gpr[rD(ctx->opcode)], rb); \
1123 tcg_temp_free_ptr(rb); \
1124 }
1125 GEN_VXFORM_NOA(vupkhsb, 7, 8);
1126 GEN_VXFORM_NOA(vupkhsh, 7, 9);
1127 GEN_VXFORM_NOA(vupkhsw, 7, 25);
1128 GEN_VXFORM_NOA(vupklsb, 7, 10);
1129 GEN_VXFORM_NOA(vupklsh, 7, 11);
1130 GEN_VXFORM_NOA(vupklsw, 7, 27);
1131 GEN_VXFORM_NOA(vupkhpx, 7, 13);
1132 GEN_VXFORM_NOA(vupklpx, 7, 15);
1133 GEN_VXFORM_NOA_ENV(vrefp, 5, 4);
1134 GEN_VXFORM_NOA_ENV(vrsqrtefp, 5, 5);
1135 GEN_VXFORM_NOA_ENV(vexptefp, 5, 6);
1136 GEN_VXFORM_NOA_ENV(vlogefp, 5, 7);
1137 GEN_VXFORM_NOA_ENV(vrfim, 5, 11);
1138 GEN_VXFORM_NOA_ENV(vrfin, 5, 8);
1139 GEN_VXFORM_NOA_ENV(vrfip, 5, 10);
1140 GEN_VXFORM_NOA_ENV(vrfiz, 5, 9);
1141 GEN_VXFORM_NOA(vprtybw, 1, 24);
1142 GEN_VXFORM_NOA(vprtybd, 1, 24);
1143 GEN_VXFORM_NOA(vprtybq, 1, 24);
1144
1145 static void gen_vsplt(DisasContext *ctx, int vece)
1146 {
1147 int uimm, dofs, bofs;
1148
1149 if (unlikely(!ctx->altivec_enabled)) {
1150 gen_exception(ctx, POWERPC_EXCP_VPU);
1151 return;
1152 }
1153
1154 uimm = UIMM5(ctx->opcode);
1155 bofs = avr_full_offset(rB(ctx->opcode));
1156 dofs = avr_full_offset(rD(ctx->opcode));
1157
1158 /* Experimental testing shows that hardware masks the immediate. */
1159 bofs += (uimm << vece) & 15;
1160 #ifndef HOST_WORDS_BIGENDIAN
1161 bofs ^= 15;
1162 bofs &= ~((1 << vece) - 1);
1163 #endif
1164
1165 tcg_gen_gvec_dup_mem(vece, dofs, bofs, 16, 16);
1166 }
1167
1168 #define GEN_VXFORM_VSPLT(name, vece, opc2, opc3) \
1169 static void glue(gen_, name)(DisasContext *ctx) { gen_vsplt(ctx, vece); }
1170
1171 #define GEN_VXFORM_UIMM_ENV(name, opc2, opc3) \
1172 static void glue(gen_, name)(DisasContext *ctx) \
1173 { \
1174 TCGv_ptr rb, rd; \
1175 TCGv_i32 uimm; \
1176 \
1177 if (unlikely(!ctx->altivec_enabled)) { \
1178 gen_exception(ctx, POWERPC_EXCP_VPU); \
1179 return; \
1180 } \
1181 uimm = tcg_const_i32(UIMM5(ctx->opcode)); \
1182 rb = gen_avr_ptr(rB(ctx->opcode)); \
1183 rd = gen_avr_ptr(rD(ctx->opcode)); \
1184 gen_helper_##name(cpu_env, rd, rb, uimm); \
1185 tcg_temp_free_i32(uimm); \
1186 tcg_temp_free_ptr(rb); \
1187 tcg_temp_free_ptr(rd); \
1188 }
1189
1190 #define GEN_VXFORM_UIMM_SPLAT(name, opc2, opc3, splat_max) \
1191 static void glue(gen_, name)(DisasContext *ctx) \
1192 { \
1193 TCGv_ptr rb, rd; \
1194 uint8_t uimm = UIMM4(ctx->opcode); \
1195 TCGv_i32 t0; \
1196 if (unlikely(!ctx->altivec_enabled)) { \
1197 gen_exception(ctx, POWERPC_EXCP_VPU); \
1198 return; \
1199 } \
1200 if (uimm > splat_max) { \
1201 uimm = 0; \
1202 } \
1203 t0 = tcg_temp_new_i32(); \
1204 tcg_gen_movi_i32(t0, uimm); \
1205 rb = gen_avr_ptr(rB(ctx->opcode)); \
1206 rd = gen_avr_ptr(rD(ctx->opcode)); \
1207 gen_helper_##name(rd, rb, t0); \
1208 tcg_temp_free_i32(t0); \
1209 tcg_temp_free_ptr(rb); \
1210 tcg_temp_free_ptr(rd); \
1211 }
1212
1213 GEN_VXFORM_VSPLT(vspltb, MO_8, 6, 8);
1214 GEN_VXFORM_VSPLT(vsplth, MO_16, 6, 9);
1215 GEN_VXFORM_VSPLT(vspltw, MO_32, 6, 10);
1216 GEN_VXFORM_UIMM_SPLAT(vextractub, 6, 8, 15);
1217 GEN_VXFORM_UIMM_SPLAT(vextractuh, 6, 9, 14);
1218 GEN_VXFORM_UIMM_SPLAT(vextractuw, 6, 10, 12);
1219 GEN_VXFORM_UIMM_SPLAT(vextractd, 6, 11, 8);
1220 GEN_VXFORM_UIMM_SPLAT(vinsertb, 6, 12, 15);
1221 GEN_VXFORM_UIMM_SPLAT(vinserth, 6, 13, 14);
1222 GEN_VXFORM_UIMM_SPLAT(vinsertw, 6, 14, 12);
1223 GEN_VXFORM_UIMM_SPLAT(vinsertd, 6, 15, 8);
1224 GEN_VXFORM_UIMM_ENV(vcfux, 5, 12);
1225 GEN_VXFORM_UIMM_ENV(vcfsx, 5, 13);
1226 GEN_VXFORM_UIMM_ENV(vctuxs, 5, 14);
1227 GEN_VXFORM_UIMM_ENV(vctsxs, 5, 15);
1228 GEN_VXFORM_DUAL(vspltb, PPC_ALTIVEC, PPC_NONE,
1229 vextractub, PPC_NONE, PPC2_ISA300);
1230 GEN_VXFORM_DUAL(vsplth, PPC_ALTIVEC, PPC_NONE,
1231 vextractuh, PPC_NONE, PPC2_ISA300);
1232 GEN_VXFORM_DUAL(vspltw, PPC_ALTIVEC, PPC_NONE,
1233 vextractuw, PPC_NONE, PPC2_ISA300);
1234 GEN_VXFORM_DUAL(vspltisb, PPC_ALTIVEC, PPC_NONE,
1235 vinsertb, PPC_NONE, PPC2_ISA300);
1236 GEN_VXFORM_DUAL(vspltish, PPC_ALTIVEC, PPC_NONE,
1237 vinserth, PPC_NONE, PPC2_ISA300);
1238 GEN_VXFORM_DUAL(vspltisw, PPC_ALTIVEC, PPC_NONE,
1239 vinsertw, PPC_NONE, PPC2_ISA300);
1240
1241 static void gen_vsldoi(DisasContext *ctx)
1242 {
1243 TCGv_ptr ra, rb, rd;
1244 TCGv_i32 sh;
1245 if (unlikely(!ctx->altivec_enabled)) {
1246 gen_exception(ctx, POWERPC_EXCP_VPU);
1247 return;
1248 }
1249 ra = gen_avr_ptr(rA(ctx->opcode));
1250 rb = gen_avr_ptr(rB(ctx->opcode));
1251 rd = gen_avr_ptr(rD(ctx->opcode));
1252 sh = tcg_const_i32(VSH(ctx->opcode));
1253 gen_helper_vsldoi(rd, ra, rb, sh);
1254 tcg_temp_free_ptr(ra);
1255 tcg_temp_free_ptr(rb);
1256 tcg_temp_free_ptr(rd);
1257 tcg_temp_free_i32(sh);
1258 }
1259
1260 #define GEN_VAFORM_PAIRED(name0, name1, opc2) \
1261 static void glue(gen_, name0##_##name1)(DisasContext *ctx) \
1262 { \
1263 TCGv_ptr ra, rb, rc, rd; \
1264 if (unlikely(!ctx->altivec_enabled)) { \
1265 gen_exception(ctx, POWERPC_EXCP_VPU); \
1266 return; \
1267 } \
1268 ra = gen_avr_ptr(rA(ctx->opcode)); \
1269 rb = gen_avr_ptr(rB(ctx->opcode)); \
1270 rc = gen_avr_ptr(rC(ctx->opcode)); \
1271 rd = gen_avr_ptr(rD(ctx->opcode)); \
1272 if (Rc(ctx->opcode)) { \
1273 gen_helper_##name1(cpu_env, rd, ra, rb, rc); \
1274 } else { \
1275 gen_helper_##name0(cpu_env, rd, ra, rb, rc); \
1276 } \
1277 tcg_temp_free_ptr(ra); \
1278 tcg_temp_free_ptr(rb); \
1279 tcg_temp_free_ptr(rc); \
1280 tcg_temp_free_ptr(rd); \
1281 }
1282
1283 GEN_VAFORM_PAIRED(vmhaddshs, vmhraddshs, 16)
1284
1285 static void gen_vmladduhm(DisasContext *ctx)
1286 {
1287 TCGv_ptr ra, rb, rc, rd;
1288 if (unlikely(!ctx->altivec_enabled)) {
1289 gen_exception(ctx, POWERPC_EXCP_VPU);
1290 return;
1291 }
1292 ra = gen_avr_ptr(rA(ctx->opcode));
1293 rb = gen_avr_ptr(rB(ctx->opcode));
1294 rc = gen_avr_ptr(rC(ctx->opcode));
1295 rd = gen_avr_ptr(rD(ctx->opcode));
1296 gen_helper_vmladduhm(rd, ra, rb, rc);
1297 tcg_temp_free_ptr(ra);
1298 tcg_temp_free_ptr(rb);
1299 tcg_temp_free_ptr(rc);
1300 tcg_temp_free_ptr(rd);
1301 }
1302
1303 static void gen_vpermr(DisasContext *ctx)
1304 {
1305 TCGv_ptr ra, rb, rc, rd;
1306 if (unlikely(!ctx->altivec_enabled)) {
1307 gen_exception(ctx, POWERPC_EXCP_VPU);
1308 return;
1309 }
1310 ra = gen_avr_ptr(rA(ctx->opcode));
1311 rb = gen_avr_ptr(rB(ctx->opcode));
1312 rc = gen_avr_ptr(rC(ctx->opcode));
1313 rd = gen_avr_ptr(rD(ctx->opcode));
1314 gen_helper_vpermr(cpu_env, rd, ra, rb, rc);
1315 tcg_temp_free_ptr(ra);
1316 tcg_temp_free_ptr(rb);
1317 tcg_temp_free_ptr(rc);
1318 tcg_temp_free_ptr(rd);
1319 }
1320
1321 GEN_VAFORM_PAIRED(vmsumubm, vmsummbm, 18)
1322 GEN_VAFORM_PAIRED(vmsumuhm, vmsumuhs, 19)
1323 GEN_VAFORM_PAIRED(vmsumshm, vmsumshs, 20)
1324 GEN_VAFORM_PAIRED(vsel, vperm, 21)
1325 GEN_VAFORM_PAIRED(vmaddfp, vnmsubfp, 23)
1326
1327 GEN_VXFORM_NOA(vclzb, 1, 28)
1328 GEN_VXFORM_NOA(vclzh, 1, 29)
1329 GEN_VXFORM_TRANS(vclzw, 1, 30)
1330 GEN_VXFORM_TRANS(vclzd, 1, 31)
1331 GEN_VXFORM_NOA_2(vnegw, 1, 24, 6)
1332 GEN_VXFORM_NOA_2(vnegd, 1, 24, 7)
1333 GEN_VXFORM_NOA_2(vextsb2w, 1, 24, 16)
1334 GEN_VXFORM_NOA_2(vextsh2w, 1, 24, 17)
1335 GEN_VXFORM_NOA_2(vextsb2d, 1, 24, 24)
1336 GEN_VXFORM_NOA_2(vextsh2d, 1, 24, 25)
1337 GEN_VXFORM_NOA_2(vextsw2d, 1, 24, 26)
1338 GEN_VXFORM_NOA_2(vctzb, 1, 24, 28)
1339 GEN_VXFORM_NOA_2(vctzh, 1, 24, 29)
1340 GEN_VXFORM_NOA_2(vctzw, 1, 24, 30)
1341 GEN_VXFORM_NOA_2(vctzd, 1, 24, 31)
1342 GEN_VXFORM_NOA_3(vclzlsbb, 1, 24, 0)
1343 GEN_VXFORM_NOA_3(vctzlsbb, 1, 24, 1)
1344 GEN_VXFORM_NOA(vpopcntb, 1, 28)
1345 GEN_VXFORM_NOA(vpopcnth, 1, 29)
1346 GEN_VXFORM_NOA(vpopcntw, 1, 30)
1347 GEN_VXFORM_NOA(vpopcntd, 1, 31)
1348 GEN_VXFORM_DUAL(vclzb, PPC_NONE, PPC2_ALTIVEC_207, \
1349 vpopcntb, PPC_NONE, PPC2_ALTIVEC_207)
1350 GEN_VXFORM_DUAL(vclzh, PPC_NONE, PPC2_ALTIVEC_207, \
1351 vpopcnth, PPC_NONE, PPC2_ALTIVEC_207)
1352 GEN_VXFORM_DUAL(vclzw, PPC_NONE, PPC2_ALTIVEC_207, \
1353 vpopcntw, PPC_NONE, PPC2_ALTIVEC_207)
1354 GEN_VXFORM_DUAL(vclzd, PPC_NONE, PPC2_ALTIVEC_207, \
1355 vpopcntd, PPC_NONE, PPC2_ALTIVEC_207)
1356 GEN_VXFORM(vbpermd, 6, 23);
1357 GEN_VXFORM(vbpermq, 6, 21);
1358 GEN_VXFORM_TRANS(vgbbd, 6, 20);
1359 GEN_VXFORM(vpmsumb, 4, 16)
1360 GEN_VXFORM(vpmsumh, 4, 17)
1361 GEN_VXFORM(vpmsumw, 4, 18)
1362 GEN_VXFORM(vpmsumd, 4, 19)
1363
1364 #define GEN_BCD(op) \
1365 static void gen_##op(DisasContext *ctx) \
1366 { \
1367 TCGv_ptr ra, rb, rd; \
1368 TCGv_i32 ps; \
1369 \
1370 if (unlikely(!ctx->altivec_enabled)) { \
1371 gen_exception(ctx, POWERPC_EXCP_VPU); \
1372 return; \
1373 } \
1374 \
1375 ra = gen_avr_ptr(rA(ctx->opcode)); \
1376 rb = gen_avr_ptr(rB(ctx->opcode)); \
1377 rd = gen_avr_ptr(rD(ctx->opcode)); \
1378 \
1379 ps = tcg_const_i32((ctx->opcode & 0x200) != 0); \
1380 \
1381 gen_helper_##op(cpu_crf[6], rd, ra, rb, ps); \
1382 \
1383 tcg_temp_free_ptr(ra); \
1384 tcg_temp_free_ptr(rb); \
1385 tcg_temp_free_ptr(rd); \
1386 tcg_temp_free_i32(ps); \
1387 }
1388
1389 #define GEN_BCD2(op) \
1390 static void gen_##op(DisasContext *ctx) \
1391 { \
1392 TCGv_ptr rd, rb; \
1393 TCGv_i32 ps; \
1394 \
1395 if (unlikely(!ctx->altivec_enabled)) { \
1396 gen_exception(ctx, POWERPC_EXCP_VPU); \
1397 return; \
1398 } \
1399 \
1400 rb = gen_avr_ptr(rB(ctx->opcode)); \
1401 rd = gen_avr_ptr(rD(ctx->opcode)); \
1402 \
1403 ps = tcg_const_i32((ctx->opcode & 0x200) != 0); \
1404 \
1405 gen_helper_##op(cpu_crf[6], rd, rb, ps); \
1406 \
1407 tcg_temp_free_ptr(rb); \
1408 tcg_temp_free_ptr(rd); \
1409 tcg_temp_free_i32(ps); \
1410 }
1411
1412 GEN_BCD(bcdadd)
1413 GEN_BCD(bcdsub)
1414 GEN_BCD2(bcdcfn)
1415 GEN_BCD2(bcdctn)
1416 GEN_BCD2(bcdcfz)
1417 GEN_BCD2(bcdctz)
1418 GEN_BCD2(bcdcfsq)
1419 GEN_BCD2(bcdctsq)
1420 GEN_BCD2(bcdsetsgn)
1421 GEN_BCD(bcdcpsgn);
1422 GEN_BCD(bcds);
1423 GEN_BCD(bcdus);
1424 GEN_BCD(bcdsr);
1425 GEN_BCD(bcdtrunc);
1426 GEN_BCD(bcdutrunc);
1427
1428 static void gen_xpnd04_1(DisasContext *ctx)
1429 {
1430 switch (opc4(ctx->opcode)) {
1431 case 0:
1432 gen_bcdctsq(ctx);
1433 break;
1434 case 2:
1435 gen_bcdcfsq(ctx);
1436 break;
1437 case 4:
1438 gen_bcdctz(ctx);
1439 break;
1440 case 5:
1441 gen_bcdctn(ctx);
1442 break;
1443 case 6:
1444 gen_bcdcfz(ctx);
1445 break;
1446 case 7:
1447 gen_bcdcfn(ctx);
1448 break;
1449 case 31:
1450 gen_bcdsetsgn(ctx);
1451 break;
1452 default:
1453 gen_invalid(ctx);
1454 break;
1455 }
1456 }
1457
1458 static void gen_xpnd04_2(DisasContext *ctx)
1459 {
1460 switch (opc4(ctx->opcode)) {
1461 case 0:
1462 gen_bcdctsq(ctx);
1463 break;
1464 case 2:
1465 gen_bcdcfsq(ctx);
1466 break;
1467 case 4:
1468 gen_bcdctz(ctx);
1469 break;
1470 case 6:
1471 gen_bcdcfz(ctx);
1472 break;
1473 case 7:
1474 gen_bcdcfn(ctx);
1475 break;
1476 case 31:
1477 gen_bcdsetsgn(ctx);
1478 break;
1479 default:
1480 gen_invalid(ctx);
1481 break;
1482 }
1483 }
1484
1485
1486 GEN_VXFORM_DUAL(vsubcuw, PPC_ALTIVEC, PPC_NONE, \
1487 xpnd04_1, PPC_NONE, PPC2_ISA300)
1488 GEN_VXFORM_DUAL(vsubsws, PPC_ALTIVEC, PPC_NONE, \
1489 xpnd04_2, PPC_NONE, PPC2_ISA300)
1490
1491 GEN_VXFORM_DUAL(vsububm, PPC_ALTIVEC, PPC_NONE, \
1492 bcdadd, PPC_NONE, PPC2_ALTIVEC_207)
1493 GEN_VXFORM_DUAL(vsububs, PPC_ALTIVEC, PPC_NONE, \
1494 bcdadd, PPC_NONE, PPC2_ALTIVEC_207)
1495 GEN_VXFORM_DUAL(vsubuhm, PPC_ALTIVEC, PPC_NONE, \
1496 bcdsub, PPC_NONE, PPC2_ALTIVEC_207)
1497 GEN_VXFORM_DUAL(vsubuhs, PPC_ALTIVEC, PPC_NONE, \
1498 bcdsub, PPC_NONE, PPC2_ALTIVEC_207)
1499 GEN_VXFORM_DUAL(vaddshs, PPC_ALTIVEC, PPC_NONE, \
1500 bcdcpsgn, PPC_NONE, PPC2_ISA300)
1501 GEN_VXFORM_DUAL(vsubudm, PPC2_ALTIVEC_207, PPC_NONE, \
1502 bcds, PPC_NONE, PPC2_ISA300)
1503 GEN_VXFORM_DUAL(vsubuwm, PPC_ALTIVEC, PPC_NONE, \
1504 bcdus, PPC_NONE, PPC2_ISA300)
1505 GEN_VXFORM_DUAL(vsubsbs, PPC_ALTIVEC, PPC_NONE, \
1506 bcdtrunc, PPC_NONE, PPC2_ISA300)
1507 GEN_VXFORM_DUAL(vsubuqm, PPC2_ALTIVEC_207, PPC_NONE, \
1508 bcdtrunc, PPC_NONE, PPC2_ISA300)
1509 GEN_VXFORM_DUAL(vsubcuq, PPC2_ALTIVEC_207, PPC_NONE, \
1510 bcdutrunc, PPC_NONE, PPC2_ISA300)
1511
1512
1513 static void gen_vsbox(DisasContext *ctx)
1514 {
1515 TCGv_ptr ra, rd;
1516 if (unlikely(!ctx->altivec_enabled)) {
1517 gen_exception(ctx, POWERPC_EXCP_VPU);
1518 return;
1519 }
1520 ra = gen_avr_ptr(rA(ctx->opcode));
1521 rd = gen_avr_ptr(rD(ctx->opcode));
1522 gen_helper_vsbox(rd, ra);
1523 tcg_temp_free_ptr(ra);
1524 tcg_temp_free_ptr(rd);
1525 }
1526
1527 GEN_VXFORM(vcipher, 4, 20)
1528 GEN_VXFORM(vcipherlast, 4, 20)
1529 GEN_VXFORM(vncipher, 4, 21)
1530 GEN_VXFORM(vncipherlast, 4, 21)
1531
1532 GEN_VXFORM_DUAL(vcipher, PPC_NONE, PPC2_ALTIVEC_207,
1533 vcipherlast, PPC_NONE, PPC2_ALTIVEC_207)
1534 GEN_VXFORM_DUAL(vncipher, PPC_NONE, PPC2_ALTIVEC_207,
1535 vncipherlast, PPC_NONE, PPC2_ALTIVEC_207)
1536
1537 #define VSHASIGMA(op) \
1538 static void gen_##op(DisasContext *ctx) \
1539 { \
1540 TCGv_ptr ra, rd; \
1541 TCGv_i32 st_six; \
1542 if (unlikely(!ctx->altivec_enabled)) { \
1543 gen_exception(ctx, POWERPC_EXCP_VPU); \
1544 return; \
1545 } \
1546 ra = gen_avr_ptr(rA(ctx->opcode)); \
1547 rd = gen_avr_ptr(rD(ctx->opcode)); \
1548 st_six = tcg_const_i32(rB(ctx->opcode)); \
1549 gen_helper_##op(rd, ra, st_six); \
1550 tcg_temp_free_ptr(ra); \
1551 tcg_temp_free_ptr(rd); \
1552 tcg_temp_free_i32(st_six); \
1553 }
1554
1555 VSHASIGMA(vshasigmaw)
1556 VSHASIGMA(vshasigmad)
1557
1558 GEN_VXFORM3(vpermxor, 22, 0xFF)
1559 GEN_VXFORM_DUAL(vsldoi, PPC_ALTIVEC, PPC_NONE,
1560 vpermxor, PPC_NONE, PPC2_ALTIVEC_207)
1561
1562 #undef GEN_VR_LDX
1563 #undef GEN_VR_STX
1564 #undef GEN_VR_LVE
1565 #undef GEN_VR_STVE
1566
1567 #undef GEN_VX_LOGICAL
1568 #undef GEN_VX_LOGICAL_207
1569 #undef GEN_VXFORM
1570 #undef GEN_VXFORM_207
1571 #undef GEN_VXFORM_DUAL
1572 #undef GEN_VXRFORM_DUAL
1573 #undef GEN_VXRFORM1
1574 #undef GEN_VXRFORM
1575 #undef GEN_VXFORM_VSPLTI
1576 #undef GEN_VXFORM_NOA
1577 #undef GEN_VXFORM_UIMM
1578 #undef GEN_VAFORM_PAIRED
1579
1580 #undef GEN_BCD2
Mirror of https://git.qemu.org/git/qemu.git