xref: /openbmc/qemu/target/hexagon/gen_tcg.h (revision 30b6852c)
1 /*
2  *  Copyright(c) 2019-2021 Qualcomm Innovation Center, Inc. All Rights Reserved.
3  *
4  *  This program is free software; you can redistribute it and/or modify
5  *  it under the terms of the GNU General Public License as published by
6  *  the Free Software Foundation; either version 2 of the License, or
7  *  (at your option) any later version.
8  *
9  *  This program is distributed in the hope that it will be useful,
10  *  but WITHOUT ANY WARRANTY; without even the implied warranty of
11  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12  *  GNU General Public License for more details.
13  *
14  *  You should have received a copy of the GNU General Public License
15  *  along with this program; if not, see <http://www.gnu.org/licenses/>.
16  */
17 
18 #ifndef HEXAGON_GEN_TCG_H
19 #define HEXAGON_GEN_TCG_H
20 
21 /*
22  * Here is a primer to understand the tag names for load/store instructions
23  *
24  * Data types
25  *      b        signed byte                       r0 = memb(r2+#0)
26  *     ub        unsigned byte                     r0 = memub(r2+#0)
27  *      h        signed half word (16 bits)        r0 = memh(r2+#0)
28  *     uh        unsigned half word                r0 = memuh(r2+#0)
29  *      i        integer (32 bits)                 r0 = memw(r2+#0)
30  *      d        double word (64 bits)             r1:0 = memd(r2+#0)
31  *
32  * Addressing modes
33  *     _io       indirect with offset              r0 = memw(r1+#4)
34  *     _ur       absolute with register offset     r0 = memw(r1<<#4+##variable)
35  *     _rr       indirect with register offset     r0 = memw(r1+r4<<#2)
36  *     gp        global pointer relative           r0 = memw(gp+#200)
37  *     _sp       stack pointer relative            r0 = memw(r29+#12)
38  *     _ap       absolute set                      r0 = memw(r1=##variable)
39  *     _pr       post increment register           r0 = memw(r1++m1)
40  *     _pbr      post increment bit reverse        r0 = memw(r1++m1:brev)
41  *     _pi       post increment immediate          r0 = memb(r1++#1)
42  *     _pci      post increment circular immediate r0 = memw(r1++#4:circ(m0))
43  *     _pcr      post increment circular register  r0 = memw(r1++I:circ(m0))
44  */
45 
46 /* Macros for complex addressing modes */
47 #define GET_EA_ap \
48     do { \
49         fEA_IMM(UiV); \
50         tcg_gen_movi_tl(ReV, UiV); \
51     } while (0)
52 #define GET_EA_pr \
53     do { \
54         fEA_REG(RxV); \
55         fPM_M(RxV, MuV); \
56     } while (0)
57 #define GET_EA_pbr \
58     do { \
59         gen_helper_fbrev(EA, RxV); \
60         tcg_gen_add_tl(RxV, RxV, MuV); \
61     } while (0)
62 #define GET_EA_pi \
63     do { \
64         fEA_REG(RxV); \
65         fPM_I(RxV, siV); \
66     } while (0)
67 #define GET_EA_pci \
68     do { \
69         TCGv tcgv_siV = tcg_const_tl(siV); \
70         tcg_gen_mov_tl(EA, RxV); \
71         gen_helper_fcircadd(RxV, RxV, tcgv_siV, MuV, \
72                             hex_gpr[HEX_REG_CS0 + MuN]); \
73         tcg_temp_free(tcgv_siV); \
74     } while (0)
75 #define GET_EA_pcr(SHIFT) \
76     do { \
77         TCGv ireg = tcg_temp_new(); \
78         tcg_gen_mov_tl(EA, RxV); \
79         gen_read_ireg(ireg, MuV, (SHIFT)); \
80         gen_helper_fcircadd(RxV, RxV, ireg, MuV, hex_gpr[HEX_REG_CS0 + MuN]); \
81         tcg_temp_free(ireg); \
82     } while (0)
83 
84 /* Instructions with multiple definitions */
85 #define fGEN_TCG_LOAD_AP(RES, SIZE, SIGN) \
86     do { \
87         fMUST_IMMEXT(UiV); \
88         fEA_IMM(UiV); \
89         fLOAD(1, SIZE, SIGN, EA, RES); \
90         tcg_gen_movi_tl(ReV, UiV); \
91     } while (0)
92 
93 #define fGEN_TCG_L4_loadrub_ap(SHORTCODE) \
94     fGEN_TCG_LOAD_AP(RdV, 1, u)
95 #define fGEN_TCG_L4_loadrb_ap(SHORTCODE) \
96     fGEN_TCG_LOAD_AP(RdV, 1, s)
97 #define fGEN_TCG_L4_loadruh_ap(SHORTCODE) \
98     fGEN_TCG_LOAD_AP(RdV, 2, u)
99 #define fGEN_TCG_L4_loadrh_ap(SHORTCODE) \
100     fGEN_TCG_LOAD_AP(RdV, 2, s)
101 #define fGEN_TCG_L4_loadri_ap(SHORTCODE) \
102     fGEN_TCG_LOAD_AP(RdV, 4, u)
103 #define fGEN_TCG_L4_loadrd_ap(SHORTCODE) \
104     fGEN_TCG_LOAD_AP(RddV, 8, u)
105 
106 #define fGEN_TCG_L2_loadrub_pci(SHORTCODE)    SHORTCODE
107 #define fGEN_TCG_L2_loadrb_pci(SHORTCODE)     SHORTCODE
108 #define fGEN_TCG_L2_loadruh_pci(SHORTCODE)    SHORTCODE
109 #define fGEN_TCG_L2_loadrh_pci(SHORTCODE)     SHORTCODE
110 #define fGEN_TCG_L2_loadri_pci(SHORTCODE)     SHORTCODE
111 #define fGEN_TCG_L2_loadrd_pci(SHORTCODE)     SHORTCODE
112 
113 #define fGEN_TCG_LOAD_pcr(SHIFT, LOAD) \
114     do { \
115         TCGv ireg = tcg_temp_new(); \
116         tcg_gen_mov_tl(EA, RxV); \
117         gen_read_ireg(ireg, MuV, SHIFT); \
118         gen_helper_fcircadd(RxV, RxV, ireg, MuV, hex_gpr[HEX_REG_CS0 + MuN]); \
119         LOAD; \
120         tcg_temp_free(ireg); \
121     } while (0)
122 
123 #define fGEN_TCG_L2_loadrub_pcr(SHORTCODE) \
124       fGEN_TCG_LOAD_pcr(0, fLOAD(1, 1, u, EA, RdV))
125 #define fGEN_TCG_L2_loadrb_pcr(SHORTCODE) \
126       fGEN_TCG_LOAD_pcr(0, fLOAD(1, 1, s, EA, RdV))
127 #define fGEN_TCG_L2_loadruh_pcr(SHORTCODE) \
128       fGEN_TCG_LOAD_pcr(1, fLOAD(1, 2, u, EA, RdV))
129 #define fGEN_TCG_L2_loadrh_pcr(SHORTCODE) \
130       fGEN_TCG_LOAD_pcr(1, fLOAD(1, 2, s, EA, RdV))
131 #define fGEN_TCG_L2_loadri_pcr(SHORTCODE) \
132       fGEN_TCG_LOAD_pcr(2, fLOAD(1, 4, u, EA, RdV))
133 #define fGEN_TCG_L2_loadrd_pcr(SHORTCODE) \
134       fGEN_TCG_LOAD_pcr(3, fLOAD(1, 8, u, EA, RddV))
135 
136 #define fGEN_TCG_L2_loadrub_pr(SHORTCODE)      SHORTCODE
137 #define fGEN_TCG_L2_loadrub_pbr(SHORTCODE)     SHORTCODE
138 #define fGEN_TCG_L2_loadrub_pi(SHORTCODE)      SHORTCODE
139 #define fGEN_TCG_L2_loadrb_pr(SHORTCODE)       SHORTCODE
140 #define fGEN_TCG_L2_loadrb_pbr(SHORTCODE)      SHORTCODE
141 #define fGEN_TCG_L2_loadrb_pi(SHORTCODE)       SHORTCODE
142 #define fGEN_TCG_L2_loadruh_pr(SHORTCODE)      SHORTCODE
143 #define fGEN_TCG_L2_loadruh_pbr(SHORTCODE)     SHORTCODE
144 #define fGEN_TCG_L2_loadruh_pi(SHORTCODE)      SHORTCODE
145 #define fGEN_TCG_L2_loadrh_pr(SHORTCODE)       SHORTCODE
146 #define fGEN_TCG_L2_loadrh_pbr(SHORTCODE)      SHORTCODE
147 #define fGEN_TCG_L2_loadrh_pi(SHORTCODE)       SHORTCODE
148 #define fGEN_TCG_L2_loadri_pr(SHORTCODE)       SHORTCODE
149 #define fGEN_TCG_L2_loadri_pbr(SHORTCODE)      SHORTCODE
150 #define fGEN_TCG_L2_loadri_pi(SHORTCODE)       SHORTCODE
151 #define fGEN_TCG_L2_loadrd_pr(SHORTCODE)       SHORTCODE
152 #define fGEN_TCG_L2_loadrd_pbr(SHORTCODE)      SHORTCODE
153 #define fGEN_TCG_L2_loadrd_pi(SHORTCODE)       SHORTCODE
154 
155 /*
156  * These instructions load 2 bytes and places them in
157  * two halves of the destination register.
158  * The GET_EA macro determines the addressing mode.
159  * The SIGN argument determines whether to zero-extend or
160  * sign-extend.
161  */
162 #define fGEN_TCG_loadbXw2(GET_EA, SIGN) \
163     do { \
164         TCGv tmp = tcg_temp_new(); \
165         TCGv byte = tcg_temp_new(); \
166         GET_EA; \
167         fLOAD(1, 2, u, EA, tmp); \
168         tcg_gen_movi_tl(RdV, 0); \
169         for (int i = 0; i < 2; i++) { \
170             gen_set_half(i, RdV, gen_get_byte(byte, i, tmp, (SIGN))); \
171         } \
172         tcg_temp_free(tmp); \
173         tcg_temp_free(byte); \
174     } while (0)
175 
176 #define fGEN_TCG_L2_loadbzw2_io(SHORTCODE) \
177     fGEN_TCG_loadbXw2(fEA_RI(RsV, siV), false)
178 #define fGEN_TCG_L4_loadbzw2_ur(SHORTCODE) \
179     fGEN_TCG_loadbXw2(fEA_IRs(UiV, RtV, uiV), false)
180 #define fGEN_TCG_L2_loadbsw2_io(SHORTCODE) \
181     fGEN_TCG_loadbXw2(fEA_RI(RsV, siV), true)
182 #define fGEN_TCG_L4_loadbsw2_ur(SHORTCODE) \
183     fGEN_TCG_loadbXw2(fEA_IRs(UiV, RtV, uiV), true)
184 #define fGEN_TCG_L4_loadbzw2_ap(SHORTCODE) \
185     fGEN_TCG_loadbXw2(GET_EA_ap, false)
186 #define fGEN_TCG_L2_loadbzw2_pr(SHORTCODE) \
187     fGEN_TCG_loadbXw2(GET_EA_pr, false)
188 #define fGEN_TCG_L2_loadbzw2_pbr(SHORTCODE) \
189     fGEN_TCG_loadbXw2(GET_EA_pbr, false)
190 #define fGEN_TCG_L2_loadbzw2_pi(SHORTCODE) \
191     fGEN_TCG_loadbXw2(GET_EA_pi, false)
192 #define fGEN_TCG_L4_loadbsw2_ap(SHORTCODE) \
193     fGEN_TCG_loadbXw2(GET_EA_ap, true)
194 #define fGEN_TCG_L2_loadbsw2_pr(SHORTCODE) \
195     fGEN_TCG_loadbXw2(GET_EA_pr, true)
196 #define fGEN_TCG_L2_loadbsw2_pbr(SHORTCODE) \
197     fGEN_TCG_loadbXw2(GET_EA_pbr, true)
198 #define fGEN_TCG_L2_loadbsw2_pi(SHORTCODE) \
199     fGEN_TCG_loadbXw2(GET_EA_pi, true)
200 #define fGEN_TCG_L2_loadbzw2_pci(SHORTCODE) \
201     fGEN_TCG_loadbXw2(GET_EA_pci, false)
202 #define fGEN_TCG_L2_loadbsw2_pci(SHORTCODE) \
203     fGEN_TCG_loadbXw2(GET_EA_pci, true)
204 #define fGEN_TCG_L2_loadbzw2_pcr(SHORTCODE) \
205     fGEN_TCG_loadbXw2(GET_EA_pcr(1), false)
206 #define fGEN_TCG_L2_loadbsw2_pcr(SHORTCODE) \
207     fGEN_TCG_loadbXw2(GET_EA_pcr(1), true)
208 
209 /*
210  * These instructions load 4 bytes and places them in
211  * four halves of the destination register pair.
212  * The GET_EA macro determines the addressing mode.
213  * The SIGN argument determines whether to zero-extend or
214  * sign-extend.
215  */
216 #define fGEN_TCG_loadbXw4(GET_EA, SIGN) \
217     do { \
218         TCGv tmp = tcg_temp_new(); \
219         TCGv byte = tcg_temp_new(); \
220         GET_EA; \
221         fLOAD(1, 4, u, EA, tmp);  \
222         tcg_gen_movi_i64(RddV, 0); \
223         for (int i = 0; i < 4; i++) { \
224             gen_set_half_i64(i, RddV, gen_get_byte(byte, i, tmp, (SIGN)));  \
225         }  \
226         tcg_temp_free(tmp); \
227         tcg_temp_free(byte); \
228     } while (0)
229 
230 #define fGEN_TCG_L2_loadbzw4_io(SHORTCODE) \
231     fGEN_TCG_loadbXw4(fEA_RI(RsV, siV), false)
232 #define fGEN_TCG_L4_loadbzw4_ur(SHORTCODE) \
233     fGEN_TCG_loadbXw4(fEA_IRs(UiV, RtV, uiV), false)
234 #define fGEN_TCG_L2_loadbsw4_io(SHORTCODE) \
235     fGEN_TCG_loadbXw4(fEA_RI(RsV, siV), true)
236 #define fGEN_TCG_L4_loadbsw4_ur(SHORTCODE) \
237     fGEN_TCG_loadbXw4(fEA_IRs(UiV, RtV, uiV), true)
238 #define fGEN_TCG_L2_loadbzw4_pci(SHORTCODE) \
239     fGEN_TCG_loadbXw4(GET_EA_pci, false)
240 #define fGEN_TCG_L2_loadbsw4_pci(SHORTCODE) \
241     fGEN_TCG_loadbXw4(GET_EA_pci, true)
242 #define fGEN_TCG_L2_loadbzw4_pcr(SHORTCODE) \
243     fGEN_TCG_loadbXw4(GET_EA_pcr(2), false)
244 #define fGEN_TCG_L2_loadbsw4_pcr(SHORTCODE) \
245     fGEN_TCG_loadbXw4(GET_EA_pcr(2), true)
246 #define fGEN_TCG_L4_loadbzw4_ap(SHORTCODE) \
247     fGEN_TCG_loadbXw4(GET_EA_ap, false)
248 #define fGEN_TCG_L2_loadbzw4_pr(SHORTCODE) \
249     fGEN_TCG_loadbXw4(GET_EA_pr, false)
250 #define fGEN_TCG_L2_loadbzw4_pbr(SHORTCODE) \
251     fGEN_TCG_loadbXw4(GET_EA_pbr, false)
252 #define fGEN_TCG_L2_loadbzw4_pi(SHORTCODE) \
253     fGEN_TCG_loadbXw4(GET_EA_pi, false)
254 #define fGEN_TCG_L4_loadbsw4_ap(SHORTCODE) \
255     fGEN_TCG_loadbXw4(GET_EA_ap, true)
256 #define fGEN_TCG_L2_loadbsw4_pr(SHORTCODE) \
257     fGEN_TCG_loadbXw4(GET_EA_pr, true)
258 #define fGEN_TCG_L2_loadbsw4_pbr(SHORTCODE) \
259     fGEN_TCG_loadbXw4(GET_EA_pbr, true)
260 #define fGEN_TCG_L2_loadbsw4_pi(SHORTCODE) \
261     fGEN_TCG_loadbXw4(GET_EA_pi, true)
262 
263 /*
264  * These instructions load a half word, shift the destination right by 16 bits
265  * and place the loaded value in the high half word of the destination pair.
266  * The GET_EA macro determines the addressing mode.
267  */
268 #define fGEN_TCG_loadalignh(GET_EA) \
269     do { \
270         TCGv tmp = tcg_temp_new(); \
271         TCGv_i64 tmp_i64 = tcg_temp_new_i64(); \
272         GET_EA;  \
273         fLOAD(1, 2, u, EA, tmp);  \
274         tcg_gen_extu_i32_i64(tmp_i64, tmp); \
275         tcg_gen_shri_i64(RyyV, RyyV, 16); \
276         tcg_gen_deposit_i64(RyyV, RyyV, tmp_i64, 48, 16); \
277         tcg_temp_free(tmp); \
278         tcg_temp_free_i64(tmp_i64); \
279     } while (0)
280 
281 #define fGEN_TCG_L4_loadalignh_ur(SHORTCODE) \
282     fGEN_TCG_loadalignh(fEA_IRs(UiV, RtV, uiV))
283 #define fGEN_TCG_L2_loadalignh_io(SHORTCODE) \
284     fGEN_TCG_loadalignh(fEA_RI(RsV, siV))
285 #define fGEN_TCG_L2_loadalignh_pci(SHORTCODE) \
286     fGEN_TCG_loadalignh(GET_EA_pci)
287 #define fGEN_TCG_L2_loadalignh_pcr(SHORTCODE) \
288     fGEN_TCG_loadalignh(GET_EA_pcr(1))
289 #define fGEN_TCG_L4_loadalignh_ap(SHORTCODE) \
290     fGEN_TCG_loadalignh(GET_EA_ap)
291 #define fGEN_TCG_L2_loadalignh_pr(SHORTCODE) \
292     fGEN_TCG_loadalignh(GET_EA_pr)
293 #define fGEN_TCG_L2_loadalignh_pbr(SHORTCODE) \
294     fGEN_TCG_loadalignh(GET_EA_pbr)
295 #define fGEN_TCG_L2_loadalignh_pi(SHORTCODE) \
296     fGEN_TCG_loadalignh(GET_EA_pi)
297 
298 /* Same as above, but loads a byte instead of half word */
299 #define fGEN_TCG_loadalignb(GET_EA) \
300     do { \
301         TCGv tmp = tcg_temp_new(); \
302         TCGv_i64 tmp_i64 = tcg_temp_new_i64(); \
303         GET_EA;  \
304         fLOAD(1, 1, u, EA, tmp);  \
305         tcg_gen_extu_i32_i64(tmp_i64, tmp); \
306         tcg_gen_shri_i64(RyyV, RyyV, 8); \
307         tcg_gen_deposit_i64(RyyV, RyyV, tmp_i64, 56, 8); \
308         tcg_temp_free(tmp); \
309         tcg_temp_free_i64(tmp_i64); \
310     } while (0)
311 
312 #define fGEN_TCG_L2_loadalignb_io(SHORTCODE) \
313     fGEN_TCG_loadalignb(fEA_RI(RsV, siV))
314 #define fGEN_TCG_L4_loadalignb_ur(SHORTCODE) \
315     fGEN_TCG_loadalignb(fEA_IRs(UiV, RtV, uiV))
316 #define fGEN_TCG_L2_loadalignb_pci(SHORTCODE) \
317     fGEN_TCG_loadalignb(GET_EA_pci)
318 #define fGEN_TCG_L2_loadalignb_pcr(SHORTCODE) \
319     fGEN_TCG_loadalignb(GET_EA_pcr(0))
320 #define fGEN_TCG_L4_loadalignb_ap(SHORTCODE) \
321     fGEN_TCG_loadalignb(GET_EA_ap)
322 #define fGEN_TCG_L2_loadalignb_pr(SHORTCODE) \
323     fGEN_TCG_loadalignb(GET_EA_pr)
324 #define fGEN_TCG_L2_loadalignb_pbr(SHORTCODE) \
325     fGEN_TCG_loadalignb(GET_EA_pbr)
326 #define fGEN_TCG_L2_loadalignb_pi(SHORTCODE) \
327     fGEN_TCG_loadalignb(GET_EA_pi)
328 
329 /*
330  * Predicated loads
331  * Here is a primer to understand the tag names
332  *
333  * Predicate used
334  *      t        true "old" value                  if (p0) r0 = memb(r2+#0)
335  *      f        false "old" value                 if (!p0) r0 = memb(r2+#0)
336  *      tnew     true "new" value                  if (p0.new) r0 = memb(r2+#0)
337  *      fnew     false "new" value                 if (!p0.new) r0 = memb(r2+#0)
338  */
339 #define fGEN_TCG_PRED_LOAD(GET_EA, PRED, SIZE, SIGN) \
340     do { \
341         TCGv LSB = tcg_temp_local_new(); \
342         TCGLabel *label = gen_new_label(); \
343         GET_EA; \
344         PRED;  \
345         PRED_LOAD_CANCEL(LSB, EA); \
346         tcg_gen_movi_tl(RdV, 0); \
347         tcg_gen_brcondi_tl(TCG_COND_EQ, LSB, 0, label); \
348             fLOAD(1, SIZE, SIGN, EA, RdV); \
349         gen_set_label(label); \
350         tcg_temp_free(LSB); \
351     } while (0)
352 
353 #define fGEN_TCG_L2_ploadrubt_pi(SHORTCODE) \
354     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBOLD(PtV), 1, u)
355 #define fGEN_TCG_L2_ploadrubf_pi(SHORTCODE) \
356     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBOLDNOT(PtV), 1, u)
357 #define fGEN_TCG_L2_ploadrubtnew_pi(SHORTCODE) \
358     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBNEW(PtN), 1, u)
359 #define fGEN_TCG_L2_ploadrubfnew_pi(SHORTCODE) \
360     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBNEWNOT(PtN), 1, u)
361 #define fGEN_TCG_L2_ploadrbt_pi(SHORTCODE) \
362     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBOLD(PtV), 1, s)
363 #define fGEN_TCG_L2_ploadrbf_pi(SHORTCODE) \
364     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBOLDNOT(PtV), 1, s)
365 #define fGEN_TCG_L2_ploadrbtnew_pi(SHORTCODE) \
366     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBNEW(PtN), 1, s)
367 #define fGEN_TCG_L2_ploadrbfnew_pi(SHORTCODE) \
368     fGEN_TCG_PRED_LOAD({ fEA_REG(RxV); fPM_I(RxV, siV); }, \
369                        fLSBNEWNOT(PtN), 1, s)
370 
371 #define fGEN_TCG_L2_ploadruht_pi(SHORTCODE) \
372     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBOLD(PtV), 2, u)
373 #define fGEN_TCG_L2_ploadruhf_pi(SHORTCODE) \
374     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBOLDNOT(PtV), 2, u)
375 #define fGEN_TCG_L2_ploadruhtnew_pi(SHORTCODE) \
376     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBNEW(PtN), 2, u)
377 #define fGEN_TCG_L2_ploadruhfnew_pi(SHORTCODE) \
378     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBNEWNOT(PtN), 2, u)
379 #define fGEN_TCG_L2_ploadrht_pi(SHORTCODE) \
380     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBOLD(PtV), 2, s)
381 #define fGEN_TCG_L2_ploadrhf_pi(SHORTCODE) \
382     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBOLDNOT(PtV), 2, s)
383 #define fGEN_TCG_L2_ploadrhtnew_pi(SHORTCODE) \
384     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBNEW(PtN), 2, s)
385 #define fGEN_TCG_L2_ploadrhfnew_pi(SHORTCODE) \
386     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBNEWNOT(PtN), 2, s)
387 
388 #define fGEN_TCG_L2_ploadrit_pi(SHORTCODE) \
389     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBOLD(PtV), 4, u)
390 #define fGEN_TCG_L2_ploadrif_pi(SHORTCODE) \
391     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBOLDNOT(PtV), 4, u)
392 #define fGEN_TCG_L2_ploadritnew_pi(SHORTCODE) \
393     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBNEW(PtN), 4, u)
394 #define fGEN_TCG_L2_ploadrifnew_pi(SHORTCODE) \
395     fGEN_TCG_PRED_LOAD(GET_EA_pi, fLSBNEWNOT(PtN), 4, u)
396 
397 /* Predicated loads into a register pair */
398 #define fGEN_TCG_PRED_LOAD_PAIR(GET_EA, PRED) \
399     do { \
400         TCGv LSB = tcg_temp_local_new(); \
401         TCGLabel *label = gen_new_label(); \
402         GET_EA; \
403         PRED;  \
404         PRED_LOAD_CANCEL(LSB, EA); \
405         tcg_gen_movi_i64(RddV, 0); \
406         tcg_gen_brcondi_tl(TCG_COND_EQ, LSB, 0, label); \
407             fLOAD(1, 8, u, EA, RddV); \
408         gen_set_label(label); \
409         tcg_temp_free(LSB); \
410     } while (0)
411 
412 #define fGEN_TCG_L2_ploadrdt_pi(SHORTCODE) \
413     fGEN_TCG_PRED_LOAD_PAIR(GET_EA_pi, fLSBOLD(PtV))
414 #define fGEN_TCG_L2_ploadrdf_pi(SHORTCODE) \
415     fGEN_TCG_PRED_LOAD_PAIR(GET_EA_pi, fLSBOLDNOT(PtV))
416 #define fGEN_TCG_L2_ploadrdtnew_pi(SHORTCODE) \
417     fGEN_TCG_PRED_LOAD_PAIR(GET_EA_pi, fLSBNEW(PtN))
418 #define fGEN_TCG_L2_ploadrdfnew_pi(SHORTCODE) \
419     fGEN_TCG_PRED_LOAD_PAIR(GET_EA_pi, fLSBNEWNOT(PtN))
420 
421 /* load-locked and store-locked */
422 #define fGEN_TCG_L2_loadw_locked(SHORTCODE) \
423     SHORTCODE
424 #define fGEN_TCG_L4_loadd_locked(SHORTCODE) \
425     SHORTCODE
426 #define fGEN_TCG_S2_storew_locked(SHORTCODE) \
427     SHORTCODE
428 #define fGEN_TCG_S4_stored_locked(SHORTCODE) \
429     SHORTCODE
430 
431 #define fGEN_TCG_STORE(SHORTCODE) \
432     do { \
433         TCGv HALF = tcg_temp_new(); \
434         TCGv BYTE = tcg_temp_new(); \
435         SHORTCODE; \
436         tcg_temp_free(HALF); \
437         tcg_temp_free(BYTE); \
438     } while (0)
439 
440 #define fGEN_TCG_STORE_pcr(SHIFT, STORE) \
441     do { \
442         TCGv ireg = tcg_temp_new(); \
443         TCGv HALF = tcg_temp_new(); \
444         TCGv BYTE = tcg_temp_new(); \
445         tcg_gen_mov_tl(EA, RxV); \
446         gen_read_ireg(ireg, MuV, SHIFT); \
447         gen_helper_fcircadd(RxV, RxV, ireg, MuV, hex_gpr[HEX_REG_CS0 + MuN]); \
448         STORE; \
449         tcg_temp_free(ireg); \
450         tcg_temp_free(HALF); \
451         tcg_temp_free(BYTE); \
452     } while (0)
453 
454 #define fGEN_TCG_S2_storerb_pbr(SHORTCODE) \
455     fGEN_TCG_STORE(SHORTCODE)
456 #define fGEN_TCG_S2_storerb_pci(SHORTCODE) \
457     fGEN_TCG_STORE(SHORTCODE)
458 #define fGEN_TCG_S2_storerb_pcr(SHORTCODE) \
459     fGEN_TCG_STORE_pcr(0, fSTORE(1, 1, EA, fGETBYTE(0, RtV)))
460 
461 #define fGEN_TCG_S2_storerh_pbr(SHORTCODE) \
462     fGEN_TCG_STORE(SHORTCODE)
463 #define fGEN_TCG_S2_storerh_pci(SHORTCODE) \
464     fGEN_TCG_STORE(SHORTCODE)
465 #define fGEN_TCG_S2_storerh_pcr(SHORTCODE) \
466     fGEN_TCG_STORE_pcr(1, fSTORE(1, 2, EA, fGETHALF(0, RtV)))
467 
468 #define fGEN_TCG_S2_storerf_pbr(SHORTCODE) \
469     fGEN_TCG_STORE(SHORTCODE)
470 #define fGEN_TCG_S2_storerf_pci(SHORTCODE) \
471     fGEN_TCG_STORE(SHORTCODE)
472 #define fGEN_TCG_S2_storerf_pcr(SHORTCODE) \
473     fGEN_TCG_STORE_pcr(1, fSTORE(1, 2, EA, fGETHALF(1, RtV)))
474 
475 #define fGEN_TCG_S2_storeri_pbr(SHORTCODE) \
476     fGEN_TCG_STORE(SHORTCODE)
477 #define fGEN_TCG_S2_storeri_pci(SHORTCODE) \
478     fGEN_TCG_STORE(SHORTCODE)
479 #define fGEN_TCG_S2_storeri_pcr(SHORTCODE) \
480     fGEN_TCG_STORE_pcr(2, fSTORE(1, 4, EA, RtV))
481 
482 #define fGEN_TCG_S2_storerd_pbr(SHORTCODE) \
483     fGEN_TCG_STORE(SHORTCODE)
484 #define fGEN_TCG_S2_storerd_pci(SHORTCODE) \
485     fGEN_TCG_STORE(SHORTCODE)
486 #define fGEN_TCG_S2_storerd_pcr(SHORTCODE) \
487     fGEN_TCG_STORE_pcr(3, fSTORE(1, 8, EA, RttV))
488 
489 #define fGEN_TCG_S2_storerbnew_pbr(SHORTCODE) \
490     fGEN_TCG_STORE(SHORTCODE)
491 #define fGEN_TCG_S2_storerbnew_pci(SHORTCODE) \
492     fGEN_TCG_STORE(SHORTCODE)
493 #define fGEN_TCG_S2_storerbnew_pcr(SHORTCODE) \
494     fGEN_TCG_STORE_pcr(0, fSTORE(1, 1, EA, fGETBYTE(0, NtN)))
495 
496 #define fGEN_TCG_S2_storerhnew_pbr(SHORTCODE) \
497     fGEN_TCG_STORE(SHORTCODE)
498 #define fGEN_TCG_S2_storerhnew_pci(SHORTCODE) \
499     fGEN_TCG_STORE(SHORTCODE)
500 #define fGEN_TCG_S2_storerhnew_pcr(SHORTCODE) \
501     fGEN_TCG_STORE_pcr(1, fSTORE(1, 2, EA, fGETHALF(0, NtN)))
502 
503 #define fGEN_TCG_S2_storerinew_pbr(SHORTCODE) \
504     fGEN_TCG_STORE(SHORTCODE)
505 #define fGEN_TCG_S2_storerinew_pci(SHORTCODE) \
506     fGEN_TCG_STORE(SHORTCODE)
507 #define fGEN_TCG_S2_storerinew_pcr(SHORTCODE) \
508     fGEN_TCG_STORE_pcr(2, fSTORE(1, 4, EA, NtN))
509 
510 /*
511  * Mathematical operations with more than one definition require
512  * special handling
513  */
514 #define fGEN_TCG_A5_ACS(SHORTCODE) \
515     do { \
516         gen_helper_vacsh_pred(PeV, cpu_env, RxxV, RssV, RttV); \
517         gen_helper_vacsh_val(RxxV, cpu_env, RxxV, RssV, RttV); \
518     } while (0)
519 
520 /*
521  * Approximate reciprocal
522  * r3,p1 = sfrecipa(r0, r1)
523  *
524  * The helper packs the 2 32-bit results into a 64-bit value,
525  * so unpack them into the proper results.
526  */
527 #define fGEN_TCG_F2_sfrecipa(SHORTCODE) \
528     do { \
529         TCGv_i64 tmp = tcg_temp_new_i64(); \
530         gen_helper_sfrecipa(tmp, cpu_env, RsV, RtV);  \
531         tcg_gen_extrh_i64_i32(RdV, tmp); \
532         tcg_gen_extrl_i64_i32(PeV, tmp); \
533         tcg_temp_free_i64(tmp); \
534     } while (0)
535 
536 /*
537  * Approximation of the reciprocal square root
538  * r1,p0 = sfinvsqrta(r0)
539  *
540  * The helper packs the 2 32-bit results into a 64-bit value,
541  * so unpack them into the proper results.
542  */
543 #define fGEN_TCG_F2_sfinvsqrta(SHORTCODE) \
544     do { \
545         TCGv_i64 tmp = tcg_temp_new_i64(); \
546         gen_helper_sfinvsqrta(tmp, cpu_env, RsV); \
547         tcg_gen_extrh_i64_i32(RdV, tmp); \
548         tcg_gen_extrl_i64_i32(PeV, tmp); \
549         tcg_temp_free_i64(tmp); \
550     } while (0)
551 
552 /*
553  * Add or subtract with carry.
554  * Predicate register is used as an extra input and output.
555  * r5:4 = add(r1:0, r3:2, p1):carry
556  */
557 #define fGEN_TCG_A4_addp_c(SHORTCODE) \
558     do { \
559         TCGv_i64 carry = tcg_temp_new_i64(); \
560         TCGv_i64 zero = tcg_const_i64(0); \
561         tcg_gen_extu_i32_i64(carry, PxV); \
562         tcg_gen_andi_i64(carry, carry, 1); \
563         tcg_gen_add2_i64(RddV, carry, RssV, zero, carry, zero); \
564         tcg_gen_add2_i64(RddV, carry, RddV, carry, RttV, zero); \
565         tcg_gen_extrl_i64_i32(PxV, carry); \
566         gen_8bitsof(PxV, PxV); \
567         tcg_temp_free_i64(carry); \
568         tcg_temp_free_i64(zero); \
569     } while (0)
570 
571 /* r5:4 = sub(r1:0, r3:2, p1):carry */
572 #define fGEN_TCG_A4_subp_c(SHORTCODE) \
573     do { \
574         TCGv_i64 carry = tcg_temp_new_i64(); \
575         TCGv_i64 zero = tcg_const_i64(0); \
576         TCGv_i64 not_RttV = tcg_temp_new_i64(); \
577         tcg_gen_extu_i32_i64(carry, PxV); \
578         tcg_gen_andi_i64(carry, carry, 1); \
579         tcg_gen_not_i64(not_RttV, RttV); \
580         tcg_gen_add2_i64(RddV, carry, RssV, zero, carry, zero); \
581         tcg_gen_add2_i64(RddV, carry, RddV, carry, not_RttV, zero); \
582         tcg_gen_extrl_i64_i32(PxV, carry); \
583         gen_8bitsof(PxV, PxV); \
584         tcg_temp_free_i64(carry); \
585         tcg_temp_free_i64(zero); \
586         tcg_temp_free_i64(not_RttV); \
587     } while (0)
588 
589 /*
590  * Compare each of the 8 unsigned bytes
591  * The minimum is placed in each byte of the destination.
592  * Each bit of the predicate is set true if the bit from the first operand
593  * is greater than the bit from the second operand.
594  * r5:4,p1 = vminub(r1:0, r3:2)
595  */
596 #define fGEN_TCG_A6_vminub_RdP(SHORTCODE) \
597     do { \
598         TCGv left = tcg_temp_new(); \
599         TCGv right = tcg_temp_new(); \
600         TCGv tmp = tcg_temp_new(); \
601         tcg_gen_movi_tl(PeV, 0); \
602         tcg_gen_movi_i64(RddV, 0); \
603         for (int i = 0; i < 8; i++) { \
604             gen_get_byte_i64(left, i, RttV, false); \
605             gen_get_byte_i64(right, i, RssV, false); \
606             tcg_gen_setcond_tl(TCG_COND_GT, tmp, left, right); \
607             tcg_gen_deposit_tl(PeV, PeV, tmp, i, 1); \
608             tcg_gen_umin_tl(tmp, left, right); \
609             gen_set_byte_i64(i, RddV, tmp); \
610         } \
611         tcg_temp_free(left); \
612         tcg_temp_free(right); \
613         tcg_temp_free(tmp); \
614     } while (0)
615 
616 /* Floating point */
617 #define fGEN_TCG_F2_conv_sf2df(SHORTCODE) \
618     gen_helper_conv_sf2df(RddV, cpu_env, RsV)
619 #define fGEN_TCG_F2_conv_df2sf(SHORTCODE) \
620     gen_helper_conv_df2sf(RdV, cpu_env, RssV)
621 #define fGEN_TCG_F2_conv_uw2sf(SHORTCODE) \
622     gen_helper_conv_uw2sf(RdV, cpu_env, RsV)
623 #define fGEN_TCG_F2_conv_uw2df(SHORTCODE) \
624     gen_helper_conv_uw2df(RddV, cpu_env, RsV)
625 #define fGEN_TCG_F2_conv_w2sf(SHORTCODE) \
626     gen_helper_conv_w2sf(RdV, cpu_env, RsV)
627 #define fGEN_TCG_F2_conv_w2df(SHORTCODE) \
628     gen_helper_conv_w2df(RddV, cpu_env, RsV)
629 #define fGEN_TCG_F2_conv_ud2sf(SHORTCODE) \
630     gen_helper_conv_ud2sf(RdV, cpu_env, RssV)
631 #define fGEN_TCG_F2_conv_ud2df(SHORTCODE) \
632     gen_helper_conv_ud2df(RddV, cpu_env, RssV)
633 #define fGEN_TCG_F2_conv_d2sf(SHORTCODE) \
634     gen_helper_conv_d2sf(RdV, cpu_env, RssV)
635 #define fGEN_TCG_F2_conv_d2df(SHORTCODE) \
636     gen_helper_conv_d2df(RddV, cpu_env, RssV)
637 #define fGEN_TCG_F2_conv_sf2uw(SHORTCODE) \
638     gen_helper_conv_sf2uw(RdV, cpu_env, RsV)
639 #define fGEN_TCG_F2_conv_sf2w(SHORTCODE) \
640     gen_helper_conv_sf2w(RdV, cpu_env, RsV)
641 #define fGEN_TCG_F2_conv_sf2ud(SHORTCODE) \
642     gen_helper_conv_sf2ud(RddV, cpu_env, RsV)
643 #define fGEN_TCG_F2_conv_sf2d(SHORTCODE) \
644     gen_helper_conv_sf2d(RddV, cpu_env, RsV)
645 #define fGEN_TCG_F2_conv_df2uw(SHORTCODE) \
646     gen_helper_conv_df2uw(RdV, cpu_env, RssV)
647 #define fGEN_TCG_F2_conv_df2w(SHORTCODE) \
648     gen_helper_conv_df2w(RdV, cpu_env, RssV)
649 #define fGEN_TCG_F2_conv_df2ud(SHORTCODE) \
650     gen_helper_conv_df2ud(RddV, cpu_env, RssV)
651 #define fGEN_TCG_F2_conv_df2d(SHORTCODE) \
652     gen_helper_conv_df2d(RddV, cpu_env, RssV)
653 #define fGEN_TCG_F2_conv_sf2uw_chop(SHORTCODE) \
654     gen_helper_conv_sf2uw_chop(RdV, cpu_env, RsV)
655 #define fGEN_TCG_F2_conv_sf2w_chop(SHORTCODE) \
656     gen_helper_conv_sf2w_chop(RdV, cpu_env, RsV)
657 #define fGEN_TCG_F2_conv_sf2ud_chop(SHORTCODE) \
658     gen_helper_conv_sf2ud_chop(RddV, cpu_env, RsV)
659 #define fGEN_TCG_F2_conv_sf2d_chop(SHORTCODE) \
660     gen_helper_conv_sf2d_chop(RddV, cpu_env, RsV)
661 #define fGEN_TCG_F2_conv_df2uw_chop(SHORTCODE) \
662     gen_helper_conv_df2uw_chop(RdV, cpu_env, RssV)
663 #define fGEN_TCG_F2_conv_df2w_chop(SHORTCODE) \
664     gen_helper_conv_df2w_chop(RdV, cpu_env, RssV)
665 #define fGEN_TCG_F2_conv_df2ud_chop(SHORTCODE) \
666     gen_helper_conv_df2ud_chop(RddV, cpu_env, RssV)
667 #define fGEN_TCG_F2_conv_df2d_chop(SHORTCODE) \
668     gen_helper_conv_df2d_chop(RddV, cpu_env, RssV)
669 #define fGEN_TCG_F2_sfadd(SHORTCODE) \
670     gen_helper_sfadd(RdV, cpu_env, RsV, RtV)
671 #define fGEN_TCG_F2_sfsub(SHORTCODE) \
672     gen_helper_sfsub(RdV, cpu_env, RsV, RtV)
673 #define fGEN_TCG_F2_sfcmpeq(SHORTCODE) \
674     gen_helper_sfcmpeq(PdV, cpu_env, RsV, RtV)
675 #define fGEN_TCG_F2_sfcmpgt(SHORTCODE) \
676     gen_helper_sfcmpgt(PdV, cpu_env, RsV, RtV)
677 #define fGEN_TCG_F2_sfcmpge(SHORTCODE) \
678     gen_helper_sfcmpge(PdV, cpu_env, RsV, RtV)
679 #define fGEN_TCG_F2_sfcmpuo(SHORTCODE) \
680     gen_helper_sfcmpuo(PdV, cpu_env, RsV, RtV)
681 #define fGEN_TCG_F2_sfmax(SHORTCODE) \
682     gen_helper_sfmax(RdV, cpu_env, RsV, RtV)
683 #define fGEN_TCG_F2_sfmin(SHORTCODE) \
684     gen_helper_sfmin(RdV, cpu_env, RsV, RtV)
685 #define fGEN_TCG_F2_sfclass(SHORTCODE) \
686     do { \
687         TCGv imm = tcg_constant_tl(uiV); \
688         gen_helper_sfclass(PdV, cpu_env, RsV, imm); \
689     } while (0)
690 #define fGEN_TCG_F2_sffixupn(SHORTCODE) \
691     gen_helper_sffixupn(RdV, cpu_env, RsV, RtV)
692 #define fGEN_TCG_F2_sffixupd(SHORTCODE) \
693     gen_helper_sffixupd(RdV, cpu_env, RsV, RtV)
694 #define fGEN_TCG_F2_sffixupr(SHORTCODE) \
695     gen_helper_sffixupr(RdV, cpu_env, RsV)
696 #define fGEN_TCG_F2_dfadd(SHORTCODE) \
697     gen_helper_dfadd(RddV, cpu_env, RssV, RttV)
698 #define fGEN_TCG_F2_dfsub(SHORTCODE) \
699     gen_helper_dfsub(RddV, cpu_env, RssV, RttV)
700 #define fGEN_TCG_F2_dfmax(SHORTCODE) \
701     gen_helper_dfmax(RddV, cpu_env, RssV, RttV)
702 #define fGEN_TCG_F2_dfmin(SHORTCODE) \
703     gen_helper_dfmin(RddV, cpu_env, RssV, RttV)
704 #define fGEN_TCG_F2_dfcmpeq(SHORTCODE) \
705     gen_helper_dfcmpeq(PdV, cpu_env, RssV, RttV)
706 #define fGEN_TCG_F2_dfcmpgt(SHORTCODE) \
707     gen_helper_dfcmpgt(PdV, cpu_env, RssV, RttV)
708 #define fGEN_TCG_F2_dfcmpge(SHORTCODE) \
709     gen_helper_dfcmpge(PdV, cpu_env, RssV, RttV)
710 #define fGEN_TCG_F2_dfcmpuo(SHORTCODE) \
711     gen_helper_dfcmpuo(PdV, cpu_env, RssV, RttV)
712 #define fGEN_TCG_F2_dfclass(SHORTCODE) \
713     do { \
714         TCGv imm = tcg_constant_tl(uiV); \
715         gen_helper_dfclass(PdV, cpu_env, RssV, imm); \
716     } while (0)
717 #define fGEN_TCG_F2_sfmpy(SHORTCODE) \
718     gen_helper_sfmpy(RdV, cpu_env, RsV, RtV)
719 #define fGEN_TCG_F2_sffma(SHORTCODE) \
720     gen_helper_sffma(RxV, cpu_env, RxV, RsV, RtV)
721 #define fGEN_TCG_F2_sffma_sc(SHORTCODE) \
722     gen_helper_sffma_sc(RxV, cpu_env, RxV, RsV, RtV, PuV)
723 #define fGEN_TCG_F2_sffms(SHORTCODE) \
724     gen_helper_sffms(RxV, cpu_env, RxV, RsV, RtV)
725 #define fGEN_TCG_F2_sffma_lib(SHORTCODE) \
726     gen_helper_sffma_lib(RxV, cpu_env, RxV, RsV, RtV)
727 #define fGEN_TCG_F2_sffms_lib(SHORTCODE) \
728     gen_helper_sffms_lib(RxV, cpu_env, RxV, RsV, RtV)
729 
730 #define fGEN_TCG_F2_dfmpyfix(SHORTCODE) \
731     gen_helper_dfmpyfix(RddV, cpu_env, RssV, RttV)
732 #define fGEN_TCG_F2_dfmpyhh(SHORTCODE) \
733     gen_helper_dfmpyhh(RxxV, cpu_env, RxxV, RssV, RttV)
734 
735 /* Nothing to do for these in qemu, need to suppress compiler warnings */
736 #define fGEN_TCG_Y4_l2fetch(SHORTCODE) \
737     do { \
738         RsV = RsV; \
739         RtV = RtV; \
740     } while (0)
741 #define fGEN_TCG_Y5_l2fetch(SHORTCODE) \
742     do { \
743         RsV = RsV; \
744     } while (0)
745 
746 #endif
747