xref: /openbmc/qemu/tests/tcg/hexagon/usr.c (revision 5242ef88)
1 /*
2  *  Copyright(c) 2022 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 /*
19  * Test instructions that might set bits in user status register (USR)
20  */
21 
22 #include <stdio.h>
23 #include <stdint.h>
24 
25 int err;
26 
27 static void __check(int line, uint32_t val, uint32_t expect)
28 {
29     if (val != expect) {
30         printf("ERROR at line %d: %d != %d\n", line, val, expect);
31         err++;
32     }
33 }
34 
35 #define check(RES, EXP) __check(__LINE__, RES, EXP)
36 
37 static void __check32(int line, uint32_t val, uint32_t expect)
38 {
39     if (val != expect) {
40         printf("ERROR at line %d: 0x%08x != 0x%08x\n", line, val, expect);
41         err++;
42     }
43 }
44 
45 #define check32(RES, EXP) __check32(__LINE__, RES, EXP)
46 
47 static void __check64(int line, uint64_t val, uint64_t expect)
48 {
49     if (val != expect) {
50         printf("ERROR at line %d: 0x%016llx != 0x%016llx\n", line, val, expect);
51         err++;
52     }
53 }
54 
55 #define check64(RES, EXP) __check64(__LINE__, RES, EXP)
56 
57 /*
58  * Some of the instructions tested are only available on certain versions
59  * of the Hexagon core
60  */
61 #define CORE_HAS_AUDIO    (__HEXAGON_ARCH__ >= 67 && defined(__HEXAGON_AUDIO__))
62 #define CORE_IS_V67       (__HEXAGON_ARCH__ >= 67)
63 
64 /* Define the bits in Hexagon USR register */
65 #define USR_OVF_BIT          0        /* Sticky saturation overflow */
66 #define USR_FPINVF_BIT       1        /* IEEE FP invalid sticky flag */
67 #define USR_FPDBZF_BIT       2        /* IEEE FP divide-by-zero sticky flag */
68 #define USR_FPOVFF_BIT       3        /* IEEE FP overflow sticky flag */
69 #define USR_FPUNFF_BIT       4        /* IEEE FP underflow sticky flag */
70 #define USR_FPINPF_BIT       5        /* IEEE FP inexact sticky flag */
71 
72 /* Corresponding values in USR */
73 #define USR_CLEAR            0
74 #define USR_OVF              (1 << USR_OVF_BIT)
75 #define USR_FPINVF           (1 << USR_FPINVF_BIT)
76 #define USR_FPDBZF           (1 << USR_FPDBZF_BIT)
77 #define USR_FPOVFF           (1 << USR_FPOVFF_BIT)
78 #define USR_FPUNFF           (1 << USR_FPUNFF_BIT)
79 #define USR_FPINPF           (1 << USR_FPINPF_BIT)
80 
81 /* Some useful floating point values */
82 const uint32_t SF_INF =              0x7f800000;
83 const uint32_t SF_QNaN =             0x7fc00000;
84 const uint32_t SF_SNaN =             0x7fb00000;
85 const uint32_t SF_QNaN_neg =         0xffc00000;
86 const uint32_t SF_SNaN_neg =         0xffb00000;
87 const uint32_t SF_HEX_NaN =          0xffffffff;
88 const uint32_t SF_zero =             0x00000000;
89 const uint32_t SF_one =              0x3f800000;
90 const uint32_t SF_one_recip =        0x3f7f0001;         /* 0.9960...  */
91 const uint32_t SF_one_invsqrta =     0x3f7f0000;         /* 0.99609375 */
92 const uint32_t SF_two =              0x40000000;
93 const uint32_t SF_four =             0x40800000;
94 const uint32_t SF_small_neg =        0xab98fba8;
95 const uint32_t SF_large_pos =        0x5afa572e;
96 
97 const uint64_t DF_QNaN =             0x7ff8000000000000ULL;
98 const uint64_t DF_SNaN =             0x7ff7000000000000ULL;
99 const uint64_t DF_QNaN_neg =         0xfff8000000000000ULL;
100 const uint64_t DF_SNaN_neg =         0xfff7000000000000ULL;
101 const uint64_t DF_HEX_NaN =          0xffffffffffffffffULL;
102 const uint64_t DF_zero =             0x0000000000000000ULL;
103 const uint64_t DF_any =              0x3f80000000000000ULL;
104 const uint64_t DF_one =              0x3ff0000000000000ULL;
105 const uint64_t DF_one_hh =           0x3ff001ff80000000ULL;     /* 1.00048... */
106 const uint64_t DF_small_neg =        0xbd731f7500000000ULL;
107 const uint64_t DF_large_pos =        0x7f80000000000001ULL;
108 
109 /*
110  * Templates for functions to execute an instruction
111  *
112  * The templates vary by the number of arguments and the types of the args
113  * and result.  We use one letter in the macro name for the result and each
114  * argument:
115  *     x             unknown (specified in a subsequent template) or don't care
116  *     R             register (32 bits)
117  *     P             pair (64 bits)
118  *     p             predicate
119  *     I             immediate
120  *     Xx            read/write
121  */
122 
123 /* Clear bits 0-5 in USR */
124 #define CLEAR_USRBITS \
125     "r2 = usr\n\t" \
126     "r2 = and(r2, #0xffffffc0)\n\t" \
127     "usr = r2\n\t"
128 
129 /* Template for instructions with one register operand */
130 #define FUNC_x_OP_x(RESTYPE, SRCTYPE, NAME, INSN) \
131 static RESTYPE NAME(SRCTYPE src, uint32_t *usr_result) \
132 { \
133     RESTYPE result; \
134     uint32_t usr; \
135     asm(CLEAR_USRBITS \
136         INSN  "\n\t" \
137         "%1 = usr\n\t" \
138         : "=r"(result), "=r"(usr) \
139         : "r"(src) \
140         : "r2", "usr"); \
141       *usr_result = usr & 0x3f; \
142       return result; \
143 }
144 
145 #define FUNC_R_OP_R(NAME, INSN) \
146 FUNC_x_OP_x(uint32_t, uint32_t, NAME, INSN)
147 
148 #define FUNC_R_OP_P(NAME, INSN) \
149 FUNC_x_OP_x(uint32_t, uint64_t, NAME, INSN)
150 
151 #define FUNC_P_OP_P(NAME, INSN) \
152 FUNC_x_OP_x(uint64_t, uint64_t, NAME, INSN)
153 
154 #define FUNC_P_OP_R(NAME, INSN) \
155 FUNC_x_OP_x(uint64_t, uint32_t, NAME, INSN)
156 
157 /*
158  * Template for instructions with a register and predicate result
159  * and one register operand
160  */
161 #define FUNC_xp_OP_x(RESTYPE, SRCTYPE, NAME, INSN) \
162 static RESTYPE NAME(SRCTYPE src, uint8_t *pred_result, uint32_t *usr_result) \
163 { \
164     RESTYPE result; \
165     uint8_t pred; \
166     uint32_t usr; \
167     asm(CLEAR_USRBITS \
168         INSN  "\n\t" \
169         "%1 = p2\n\t" \
170         "%2 = usr\n\t" \
171         : "=r"(result), "=r"(pred), "=r"(usr) \
172         : "r"(src) \
173         : "r2", "p2", "usr"); \
174     *pred_result = pred; \
175     *usr_result = usr & 0x3f; \
176     return result; \
177 }
178 
179 #define FUNC_Rp_OP_R(NAME, INSN) \
180 FUNC_xp_OP_x(uint32_t, uint32_t, NAME, INSN)
181 
182 /* Template for instructions with two register operands */
183 #define FUNC_x_OP_xx(RESTYPE, SRC1TYPE, SRC2TYPE, NAME, INSN) \
184 static RESTYPE NAME(SRC1TYPE src1, SRC2TYPE src2, uint32_t *usr_result) \
185 { \
186     RESTYPE result; \
187     uint32_t usr; \
188     asm(CLEAR_USRBITS \
189         INSN "\n\t" \
190         "%1 = usr\n\t" \
191         : "=r"(result), "=r"(usr) \
192         : "r"(src1), "r"(src2) \
193         : "r2", "usr"); \
194     *usr_result = usr & 0x3f; \
195     return result; \
196 }
197 
198 #define FUNC_P_OP_PP(NAME, INSN) \
199 FUNC_x_OP_xx(uint64_t, uint64_t, uint64_t, NAME, INSN)
200 
201 #define FUNC_R_OP_PP(NAME, INSN) \
202 FUNC_x_OP_xx(uint32_t, uint64_t, uint64_t, NAME, INSN)
203 
204 #define FUNC_P_OP_RR(NAME, INSN) \
205 FUNC_x_OP_xx(uint64_t, uint32_t, uint32_t, NAME, INSN)
206 
207 #define FUNC_R_OP_RR(NAME, INSN) \
208 FUNC_x_OP_xx(uint32_t, uint32_t, uint32_t, NAME, INSN)
209 
210 #define FUNC_R_OP_PR(NAME, INSN) \
211 FUNC_x_OP_xx(uint32_t, uint64_t, uint32_t, NAME, INSN)
212 
213 #define FUNC_P_OP_PR(NAME, INSN) \
214 FUNC_x_OP_xx(uint64_t, uint64_t, uint32_t, NAME, INSN)
215 
216 /*
217  * Template for instructions with a register and predicate result
218  * and two register operands
219  */
220 #define FUNC_xp_OP_xx(RESTYPE, SRC1TYPE, SRC2TYPE, NAME, INSN) \
221 static RESTYPE NAME(SRC1TYPE src1, SRC2TYPE src2, \
222                     uint8_t *pred_result, uint32_t *usr_result) \
223 { \
224     RESTYPE result; \
225     uint8_t pred; \
226     uint32_t usr; \
227     asm(CLEAR_USRBITS \
228         INSN  "\n\t" \
229         "%1 = p2\n\t" \
230         "%2 = usr\n\t" \
231         : "=r"(result), "=r"(pred), "=r"(usr) \
232         : "r"(src1), "r"(src2) \
233         : "r2", "p2", "usr"); \
234     *pred_result = pred; \
235     *usr_result = usr & 0x3f; \
236     return result; \
237 }
238 
239 #define FUNC_Rp_OP_RR(NAME, INSN) \
240 FUNC_xp_OP_xx(uint32_t, uint32_t, uint32_t, NAME, INSN)
241 
242 /* Template for instructions with one register and one immediate */
243 #define FUNC_x_OP_xI(RESTYPE, SRC1TYPE, NAME, INSN) \
244 static RESTYPE NAME(SRC1TYPE src1, int32_t src2, uint32_t *usr_result) \
245 { \
246     RESTYPE result; \
247     uint32_t usr; \
248     asm(CLEAR_USRBITS \
249         INSN "\n\t" \
250         "%1 = usr\n\t" \
251         : "=r"(result), "=r"(usr) \
252         : "r"(src1), "i"(src2) \
253         : "r2", "usr"); \
254     *usr_result = usr & 0x3f; \
255     return result; \
256 }
257 
258 #define FUNC_R_OP_RI(NAME, INSN) \
259 FUNC_x_OP_xI(uint32_t, uint32_t, NAME, INSN)
260 
261 #define FUNC_R_OP_PI(NAME, INSN) \
262 FUNC_x_OP_xI(uint32_t, uint64_t, NAME, INSN)
263 
264 /*
265  * Template for instructions with a read/write result
266  * and two register operands
267  */
268 #define FUNC_Xx_OP_xx(RESTYPE, SRC1TYPE, SRC2TYPE, NAME, INSN) \
269 static RESTYPE NAME(RESTYPE result, SRC1TYPE src1, SRC2TYPE src2, \
270                     uint32_t *usr_result) \
271 { \
272     uint32_t usr; \
273     asm(CLEAR_USRBITS \
274         INSN "\n\t" \
275         "%1 = usr\n\t" \
276         : "+r"(result), "=r"(usr) \
277         : "r"(src1), "r"(src2) \
278         : "r2", "usr"); \
279     *usr_result = usr & 0x3f; \
280     return result; \
281 }
282 
283 #define FUNC_XR_OP_RR(NAME, INSN) \
284 FUNC_Xx_OP_xx(uint32_t, uint32_t, uint32_t, NAME, INSN)
285 
286 #define FUNC_XP_OP_PP(NAME, INSN) \
287 FUNC_Xx_OP_xx(uint64_t, uint64_t, uint64_t, NAME, INSN)
288 
289 #define FUNC_XP_OP_RR(NAME, INSN) \
290 FUNC_Xx_OP_xx(uint64_t, uint32_t, uint32_t, NAME, INSN)
291 
292 /*
293  * Template for instructions with a read/write result
294  * and two register operands
295  */
296 #define FUNC_Xxp_OP_xx(RESTYPE, SRC1TYPE, SRC2TYPE, NAME, INSN) \
297 static RESTYPE NAME(RESTYPE result, SRC1TYPE src1, SRC2TYPE src2, \
298                     uint8_t *pred_result, uint32_t *usr_result) \
299 { \
300     uint32_t usr; \
301     uint8_t pred; \
302     asm(CLEAR_USRBITS \
303         INSN "\n\t" \
304         "%1 = p2\n\t" \
305         "%2 = usr\n\t" \
306         : "+r"(result), "=r"(pred), "=r"(usr) \
307         : "r"(src1), "r"(src2) \
308         : "r2", "usr"); \
309     *pred_result = pred; \
310     *usr_result = usr & 0x3f; \
311     return result; \
312 }
313 
314 #define FUNC_XPp_OP_PP(NAME, INSN) \
315 FUNC_Xxp_OP_xx(uint64_t, uint64_t, uint64_t, NAME, INSN)
316 
317 /*
318  * Template for instructions with a read/write result and
319  * two register and one predicate operands
320  */
321 #define FUNC_Xx_OP_xxp(RESTYPE, SRC1TYPE, SRC2TYPE, NAME, INSN) \
322 static RESTYPE NAME(RESTYPE result, SRC1TYPE src1, SRC2TYPE src2, uint8_t pred,\
323                     uint32_t *usr_result) \
324 { \
325     uint32_t usr; \
326     asm(CLEAR_USRBITS \
327         "p2 = %4\n\t" \
328         INSN "\n\t" \
329         "%1 = usr\n\t" \
330         : "+r"(result), "=r"(usr) \
331         : "r"(src1), "r"(src2), "r"(pred) \
332         : "r2", "p2", "usr"); \
333     *usr_result = usr & 0x3f; \
334     return result; \
335 }
336 
337 #define FUNC_XR_OP_RRp(NAME, INSN) \
338 FUNC_Xx_OP_xxp(uint32_t, uint32_t, uint32_t, NAME, INSN)
339 
340 /* Template for compare instructions with two register operands */
341 #define FUNC_CMP_xx(SRC1TYPE, SRC2TYPE, NAME, INSN) \
342 static uint32_t NAME(SRC1TYPE src1, SRC2TYPE src2, uint32_t *usr_result) \
343 { \
344     uint32_t result; \
345     uint32_t usr; \
346     asm(CLEAR_USRBITS \
347         INSN "\n\t" \
348         "%0 = p1\n\t" \
349         "%1 = usr\n\t" \
350         : "=r"(result), "=r"(usr) \
351         : "r"(src1), "r"(src2) \
352         : "p1", "r2", "usr"); \
353     *usr_result = usr & 0x3f; \
354     return result; \
355 }
356 
357 #define FUNC_CMP_RR(NAME, INSN) \
358 FUNC_CMP_xx(uint32_t, uint32_t, NAME, INSN)
359 
360 #define FUNC_CMP_PP(NAME, INSN) \
361 FUNC_CMP_xx(uint64_t, uint64_t, NAME, INSN)
362 
363 /*
364  * Function declarations using the templates
365  */
366 FUNC_R_OP_R(satub,              "%0 = satub(%2)")
367 FUNC_P_OP_PP(vaddubs,           "%0 = vaddub(%2, %3):sat")
368 FUNC_P_OP_PP(vadduhs,           "%0 = vadduh(%2, %3):sat")
369 FUNC_P_OP_PP(vsububs,           "%0 = vsubub(%2, %3):sat")
370 FUNC_P_OP_PP(vsubuhs,           "%0 = vsubuh(%2, %3):sat")
371 
372 /* Add vector of half integers with saturation and pack to unsigned bytes */
373 FUNC_R_OP_PP(vaddhubs,          "%0 = vaddhub(%2, %3):sat")
374 
375 /* Vector saturate half to unsigned byte */
376 FUNC_R_OP_P(vsathub,            "%0 = vsathub(%2)")
377 
378 /* Similar to above but takes a 32-bit argument */
379 FUNC_R_OP_R(svsathub,           "%0 = vsathub(%2)")
380 
381 /* Vector saturate word to unsigned half */
382 FUNC_P_OP_P(vsatwuh_nopack,     "%0 = vsatwuh(%2)")
383 
384 /* Similar to above but returns a 32-bit result */
385 FUNC_R_OP_P(vsatwuh,            "%0 = vsatwuh(%2)")
386 
387 /* Vector arithmetic shift halfwords with saturate and pack */
388 FUNC_R_OP_PI(asrhub_sat,        "%0 = vasrhub(%2, #%3):sat")
389 
390 /* Vector arithmetic shift halfwords with round, saturate and pack */
391 FUNC_R_OP_PI(asrhub_rnd_sat,    "%0 = vasrhub(%2, #%3):raw")
392 
393 FUNC_R_OP_RR(addsat,            "%0 = add(%2, %3):sat")
394 /* Similar to above but with register pairs */
395 FUNC_P_OP_PP(addpsat,           "%0 = add(%2, %3):sat")
396 
397 FUNC_XR_OP_RR(mpy_acc_sat_hh_s0, "%0 += mpy(%2.H, %3.H):sat")
398 FUNC_R_OP_RR(mpy_sat_hh_s1,     "%0 = mpy(%2.H, %3.H):<<1:sat")
399 FUNC_R_OP_RR(mpy_sat_rnd_hh_s1, "%0 = mpy(%2.H, %3.H):<<1:rnd:sat")
400 FUNC_R_OP_RR(mpy_up_s1_sat,     "%0 = mpy(%2, %3):<<1:sat")
401 FUNC_P_OP_RR(vmpy2s_s1,         "%0 = vmpyh(%2, %3):<<1:sat")
402 FUNC_P_OP_RR(vmpy2su_s1,        "%0 = vmpyhsu(%2, %3):<<1:sat")
403 FUNC_R_OP_RR(vmpy2s_s1pack,     "%0 = vmpyh(%2, %3):<<1:rnd:sat")
404 FUNC_P_OP_PP(vmpy2es_s1,        "%0 = vmpyeh(%2, %3):<<1:sat")
405 FUNC_R_OP_PP(vdmpyrs_s1,        "%0 = vdmpy(%2, %3):<<1:rnd:sat")
406 FUNC_XP_OP_PP(vdmacs_s0,        "%0 += vdmpy(%2, %3):sat")
407 FUNC_R_OP_RR(cmpyrs_s0,         "%0 = cmpy(%2, %3):rnd:sat")
408 FUNC_XP_OP_RR(cmacs_s0,         "%0 += cmpy(%2, %3):sat")
409 FUNC_XP_OP_RR(cnacs_s0,         "%0 -= cmpy(%2, %3):sat")
410 FUNC_P_OP_PP(vrcmpys_s1_h,      "%0 = vrcmpys(%2, %3):<<1:sat:raw:hi")
411 FUNC_XP_OP_PP(mmacls_s0,        "%0 += vmpyweh(%2, %3):sat")
412 FUNC_R_OP_RR(hmmpyl_rs1,        "%0 = mpy(%2, %3.L):<<1:rnd:sat")
413 FUNC_XP_OP_PP(mmaculs_s0,       "%0 += vmpyweuh(%2, %3):sat")
414 FUNC_R_OP_PR(cmpyi_wh,          "%0 = cmpyiwh(%2, %3):<<1:rnd:sat")
415 FUNC_P_OP_PP(vcmpy_s0_sat_i,    "%0 = vcmpyi(%2, %3):sat")
416 FUNC_P_OP_PR(vcrotate,          "%0 = vcrotate(%2, %3)")
417 FUNC_P_OP_PR(vcnegh,            "%0 = vcnegh(%2, %3)")
418 
419 #if CORE_HAS_AUDIO
420 FUNC_R_OP_PP(wcmpyrw,           "%0 = cmpyrw(%2, %3):<<1:sat")
421 #endif
422 
423 FUNC_R_OP_RR(addh_l16_sat_ll,   "%0 = add(%2.L, %3.L):sat")
424 FUNC_P_OP_P(vconj,              "%0 = vconj(%2):sat")
425 FUNC_P_OP_PP(vxaddsubw,         "%0 = vxaddsubw(%2, %3):sat")
426 FUNC_P_OP_P(vabshsat,           "%0 = vabsh(%2):sat")
427 FUNC_P_OP_PP(vnavgwr,           "%0 = vnavgw(%2, %3):rnd:sat")
428 FUNC_R_OP_RI(round_ri_sat,      "%0 = round(%2, #%3):sat")
429 FUNC_R_OP_RR(asr_r_r_sat,       "%0 = asr(%2, %3):sat")
430 
431 FUNC_XPp_OP_PP(ACS,             "%0, p2 = vacsh(%3, %4)")
432 
433 /* Floating point */
434 FUNC_R_OP_RR(sfmin,             "%0 = sfmin(%2, %3)")
435 FUNC_R_OP_RR(sfmax,             "%0 = sfmax(%2, %3)")
436 FUNC_R_OP_RR(sfadd,             "%0 = sfadd(%2, %3)")
437 FUNC_R_OP_RR(sfsub,             "%0 = sfsub(%2, %3)")
438 FUNC_R_OP_RR(sfmpy,             "%0 = sfmpy(%2, %3)")
439 FUNC_XR_OP_RR(sffma,            "%0 += sfmpy(%2, %3)")
440 FUNC_XR_OP_RR(sffms,            "%0 -= sfmpy(%2, %3)")
441 FUNC_CMP_RR(sfcmpuo,            "p1 = sfcmp.uo(%2, %3)")
442 FUNC_CMP_RR(sfcmpeq,            "p1 = sfcmp.eq(%2, %3)")
443 FUNC_CMP_RR(sfcmpgt,            "p1 = sfcmp.gt(%2, %3)")
444 FUNC_CMP_RR(sfcmpge,            "p1 = sfcmp.ge(%2, %3)")
445 
446 FUNC_P_OP_PP(dfadd,             "%0 = dfadd(%2, %3)")
447 FUNC_P_OP_PP(dfsub,             "%0 = dfsub(%2, %3)")
448 
449 #if CORE_IS_V67
450 FUNC_P_OP_PP(dfmin,             "%0 = dfmin(%2, %3)")
451 FUNC_P_OP_PP(dfmax,             "%0 = dfmax(%2, %3)")
452 FUNC_XP_OP_PP(dfmpyhh,          "%0 += dfmpyhh(%2, %3)")
453 #endif
454 
455 FUNC_CMP_PP(dfcmpuo,            "p1 = dfcmp.uo(%2, %3)")
456 FUNC_CMP_PP(dfcmpeq,            "p1 = dfcmp.eq(%2, %3)")
457 FUNC_CMP_PP(dfcmpgt,            "p1 = dfcmp.gt(%2, %3)")
458 FUNC_CMP_PP(dfcmpge,            "p1 = dfcmp.ge(%2, %3)")
459 
460 /* Conversions from sf */
461 FUNC_P_OP_R(conv_sf2df,         "%0 = convert_sf2df(%2)")
462 FUNC_R_OP_R(conv_sf2uw,         "%0 = convert_sf2uw(%2)")
463 FUNC_R_OP_R(conv_sf2w,          "%0 = convert_sf2w(%2)")
464 FUNC_P_OP_R(conv_sf2ud,         "%0 = convert_sf2ud(%2)")
465 FUNC_P_OP_R(conv_sf2d,          "%0 = convert_sf2d(%2)")
466 FUNC_R_OP_R(conv_sf2uw_chop,    "%0 = convert_sf2uw(%2):chop")
467 FUNC_R_OP_R(conv_sf2w_chop,     "%0 = convert_sf2w(%2):chop")
468 FUNC_P_OP_R(conv_sf2ud_chop,    "%0 = convert_sf2ud(%2):chop")
469 FUNC_P_OP_R(conv_sf2d_chop,     "%0 = convert_sf2d(%2):chop")
470 
471 /* Conversions from df */
472 FUNC_R_OP_P(conv_df2sf,         "%0 = convert_df2sf(%2)")
473 FUNC_R_OP_P(conv_df2uw,         "%0 = convert_df2uw(%2)")
474 FUNC_R_OP_P(conv_df2w,          "%0 = convert_df2w(%2)")
475 FUNC_P_OP_P(conv_df2ud,         "%0 = convert_df2ud(%2)")
476 FUNC_P_OP_P(conv_df2d,          "%0 = convert_df2d(%2)")
477 FUNC_R_OP_P(conv_df2uw_chop,    "%0 = convert_df2uw(%2):chop")
478 FUNC_R_OP_P(conv_df2w_chop,     "%0 = convert_df2w(%2):chop")
479 FUNC_P_OP_P(conv_df2ud_chop,    "%0 = convert_df2ud(%2):chop")
480 FUNC_P_OP_P(conv_df2d_chop,     "%0 = convert_df2d(%2):chop")
481 
482 /* Integer to float conversions */
483 FUNC_R_OP_R(conv_uw2sf,         "%0 = convert_uw2sf(%2)")
484 FUNC_R_OP_R(conv_w2sf,          "%0 = convert_w2sf(%2)")
485 FUNC_R_OP_P(conv_ud2sf,         "%0 = convert_ud2sf(%2)")
486 FUNC_R_OP_P(conv_d2sf,          "%0 = convert_d2sf(%2)")
487 
488 /* Special purpose floating point instructions */
489 FUNC_XR_OP_RRp(sffma_sc,        "%0 += sfmpy(%2, %3, p2):scale")
490 FUNC_Rp_OP_RR(sfrecipa,         "%0, p2 = sfrecipa(%3, %4)")
491 FUNC_R_OP_RR(sffixupn,          "%0 = sffixupn(%2, %3)")
492 FUNC_R_OP_RR(sffixupd,          "%0 = sffixupd(%2, %3)")
493 FUNC_R_OP_R(sffixupr,           "%0 = sffixupr(%2)")
494 FUNC_Rp_OP_R(sfinvsqrta,        "%0, p2 = sfinvsqrta(%3)")
495 
496 /*
497  * Templates for test cases
498  *
499  * Same naming convention as the function templates
500  */
501 #define TEST_x_OP_x(RESTYPE, CHECKFN, SRCTYPE, FUNC, SRC, RES, USR_RES) \
502     do { \
503         RESTYPE result; \
504         SRCTYPE src = SRC; \
505         uint32_t usr_result; \
506         result = FUNC(src, &usr_result); \
507         CHECKFN(result, RES); \
508         check(usr_result, USR_RES); \
509     } while (0)
510 
511 #define TEST_R_OP_R(FUNC, SRC, RES, USR_RES) \
512 TEST_x_OP_x(uint32_t, check32, uint32_t, FUNC, SRC, RES, USR_RES)
513 
514 #define TEST_R_OP_P(FUNC, SRC, RES, USR_RES) \
515 TEST_x_OP_x(uint32_t, check32, uint64_t, FUNC, SRC, RES, USR_RES)
516 
517 #define TEST_P_OP_P(FUNC, SRC, RES, USR_RES) \
518 TEST_x_OP_x(uint64_t, check64, uint64_t, FUNC, SRC, RES, USR_RES)
519 
520 #define TEST_P_OP_R(FUNC, SRC, RES, USR_RES) \
521 TEST_x_OP_x(uint64_t, check64, uint32_t, FUNC, SRC, RES, USR_RES)
522 
523 #define TEST_xp_OP_x(RESTYPE, CHECKFN, SRCTYPE, FUNC, SRC, \
524                      RES, PRED_RES, USR_RES) \
525     do { \
526         RESTYPE result; \
527         SRCTYPE src = SRC; \
528         uint8_t pred_result; \
529         uint32_t usr_result; \
530         result = FUNC(src, &pred_result, &usr_result); \
531         CHECKFN(result, RES); \
532         check(pred_result, PRED_RES); \
533         check(usr_result, USR_RES); \
534     } while (0)
535 
536 #define TEST_Rp_OP_R(FUNC, SRC, RES, PRED_RES, USR_RES) \
537 TEST_xp_OP_x(uint32_t, check32, uint32_t, FUNC, SRC, RES, PRED_RES, USR_RES)
538 
539 #define TEST_x_OP_xx(RESTYPE, CHECKFN, SRC1TYPE, SRC2TYPE, \
540                      FUNC, SRC1, SRC2, RES, USR_RES) \
541     do { \
542         RESTYPE result; \
543         SRC1TYPE src1 = SRC1; \
544         SRC2TYPE src2 = SRC2; \
545         uint32_t usr_result; \
546         result = FUNC(src1, src2, &usr_result); \
547         CHECKFN(result, RES); \
548         check(usr_result, USR_RES); \
549     } while (0)
550 
551 #define TEST_P_OP_PP(FUNC, SRC1, SRC2, RES, USR_RES) \
552 TEST_x_OP_xx(uint64_t, check64, uint64_t, uint64_t, \
553              FUNC, SRC1, SRC2, RES, USR_RES)
554 
555 #define TEST_R_OP_PP(FUNC, SRC1, SRC2, RES, USR_RES) \
556 TEST_x_OP_xx(uint32_t, check32, uint64_t, uint64_t, \
557              FUNC, SRC1, SRC2, RES, USR_RES)
558 
559 #define TEST_P_OP_RR(FUNC, SRC1, SRC2, RES, USR_RES) \
560 TEST_x_OP_xx(uint64_t, check64, uint32_t, uint32_t, \
561              FUNC, SRC1, SRC2, RES, USR_RES)
562 
563 #define TEST_R_OP_RR(FUNC, SRC1, SRC2, RES, USR_RES) \
564 TEST_x_OP_xx(uint32_t, check32, uint32_t, uint32_t, \
565              FUNC, SRC1, SRC2, RES, USR_RES)
566 
567 #define TEST_R_OP_PR(FUNC, SRC1, SRC2, RES, USR_RES) \
568 TEST_x_OP_xx(uint32_t, check32, uint64_t, uint32_t, \
569              FUNC, SRC1, SRC2, RES, USR_RES)
570 
571 #define TEST_P_OP_PR(FUNC, SRC1, SRC2, RES, USR_RES) \
572 TEST_x_OP_xx(uint64_t, check64, uint64_t, uint32_t, \
573              FUNC, SRC1, SRC2, RES, USR_RES)
574 
575 #define TEST_xp_OP_xx(RESTYPE, CHECKFN, SRC1TYPE, SRC2TYPE, FUNC, SRC1, SRC2, \
576                       RES, PRED_RES, USR_RES) \
577     do { \
578         RESTYPE result; \
579         SRC1TYPE src1 = SRC1; \
580         SRC2TYPE src2 = SRC2; \
581         uint8_t pred_result; \
582         uint32_t usr_result; \
583         result = FUNC(src1, src2, &pred_result, &usr_result); \
584         CHECKFN(result, RES); \
585         check(pred_result, PRED_RES); \
586         check(usr_result, USR_RES); \
587     } while (0)
588 
589 #define TEST_Rp_OP_RR(FUNC, SRC1, SRC2, RES, PRED_RES, USR_RES) \
590 TEST_xp_OP_xx(uint32_t, check32, uint32_t, uint32_t, FUNC, SRC1, SRC2, \
591               RES, PRED_RES, USR_RES)
592 
593 #define TEST_x_OP_xI(RESTYPE, CHECKFN, SRC1TYPE, \
594                      FUNC, SRC1, SRC2, RES, USR_RES) \
595     do { \
596         RESTYPE result; \
597         SRC1TYPE src1 = SRC1; \
598         uint32_t src2 = SRC2; \
599         uint32_t usr_result; \
600         result = FUNC(src1, src2, &usr_result); \
601         CHECKFN(result, RES); \
602         check(usr_result, USR_RES); \
603     } while (0)
604 
605 #define TEST_R_OP_RI(FUNC, SRC1, SRC2, RES, USR_RES) \
606 TEST_x_OP_xI(uint32_t, check32, uint32_t, \
607              FUNC, SRC1, SRC2, RES, USR_RES)
608 
609 #define TEST_R_OP_PI(FUNC, SRC1, SRC2, RES, USR_RES) \
610 TEST_x_OP_xI(uint32_t, check64, uint64_t, \
611              FUNC, SRC1, SRC2, RES, USR_RES)
612 
613 #define TEST_Xx_OP_xx(RESTYPE, CHECKFN, SRC1TYPE, SRC2TYPE, \
614                       FUNC, RESIN, SRC1, SRC2, RES, USR_RES) \
615     do { \
616         RESTYPE result = RESIN; \
617         SRC1TYPE src1 = SRC1; \
618         SRC2TYPE src2 = SRC2; \
619         uint32_t usr_result; \
620         result = FUNC(result, src1, src2, &usr_result); \
621         CHECKFN(result, RES); \
622         check(usr_result, USR_RES); \
623     } while (0)
624 
625 #define TEST_XR_OP_RR(FUNC, RESIN, SRC1, SRC2, RES, USR_RES) \
626 TEST_Xx_OP_xx(uint32_t, check32, uint32_t, uint32_t, \
627               FUNC, RESIN, SRC1, SRC2, RES, USR_RES)
628 
629 #define TEST_XP_OP_PP(FUNC, RESIN, SRC1, SRC2, RES, USR_RES) \
630 TEST_Xx_OP_xx(uint64_t, check64, uint64_t, uint64_t, \
631               FUNC, RESIN, SRC1, SRC2, RES, USR_RES)
632 
633 #define TEST_XP_OP_RR(FUNC, RESIN, SRC1, SRC2, RES, USR_RES) \
634 TEST_Xx_OP_xx(uint64_t, check64, uint32_t, uint32_t, \
635               FUNC, RESIN, SRC1, SRC2, RES, USR_RES)
636 
637 #define TEST_Xxp_OP_xx(RESTYPE, CHECKFN, SRC1TYPE, SRC2TYPE, \
638                        FUNC, RESIN, SRC1, SRC2, RES, PRED_RES, USR_RES) \
639     do { \
640         RESTYPE result = RESIN; \
641         SRC1TYPE src1 = SRC1; \
642         SRC2TYPE src2 = SRC2; \
643         uint8_t pred_res; \
644         uint32_t usr_result; \
645         result = FUNC(result, src1, src2, &pred_res, &usr_result); \
646         CHECKFN(result, RES); \
647         check(usr_result, USR_RES); \
648     } while (0)
649 
650 #define TEST_XPp_OP_PP(FUNC, RESIN, SRC1, SRC2, RES, PRED_RES, USR_RES) \
651 TEST_Xxp_OP_xx(uint64_t, check64, uint64_t, uint64_t, FUNC, RESIN, SRC1, SRC2, \
652                RES, PRED_RES, USR_RES)
653 
654 #define TEST_Xx_OP_xxp(RESTYPE, CHECKFN, SRC1TYPE, SRC2TYPE, \
655                       FUNC, RESIN, SRC1, SRC2, PRED, RES, USR_RES) \
656     do { \
657         RESTYPE result = RESIN; \
658         SRC1TYPE src1 = SRC1; \
659         SRC2TYPE src2 = SRC2; \
660         uint8_t pred = PRED; \
661         uint32_t usr_result; \
662         result = FUNC(result, src1, src2, pred, &usr_result); \
663         CHECKFN(result, RES); \
664         check(usr_result, USR_RES); \
665     } while (0)
666 
667 #define TEST_XR_OP_RRp(FUNC, RESIN, SRC1, SRC2, PRED, RES, USR_RES) \
668 TEST_Xx_OP_xxp(uint32_t, check32, uint32_t, uint32_t, \
669               FUNC, RESIN, SRC1, SRC2, PRED, RES, USR_RES)
670 
671 #define TEST_CMP_xx(SRC1TYPE, SRC2TYPE, \
672                     FUNC, SRC1, SRC2, RES, USR_RES) \
673     do { \
674         uint32_t result; \
675         SRC1TYPE src1 = SRC1; \
676         SRC2TYPE src2 = SRC2; \
677         uint32_t usr_result; \
678         result = FUNC(src1, src2, &usr_result); \
679         check(result, RES); \
680         check(usr_result, USR_RES); \
681     } while (0)
682 
683 #define TEST_CMP_RR(FUNC, SRC1, SRC2, RES, USR_RES) \
684 TEST_CMP_xx(uint32_t, uint32_t, FUNC, SRC1, SRC2, RES, USR_RES)
685 
686 #define TEST_CMP_PP(FUNC, SRC1, SRC2, RES, USR_RES) \
687 TEST_CMP_xx(uint64_t, uint64_t, FUNC, SRC1, SRC2, RES, USR_RES)
688 
689 int main()
690 {
691     TEST_R_OP_R(satub,       0,         0,         USR_CLEAR);
692     TEST_R_OP_R(satub,       0xff,      0xff,      USR_CLEAR);
693     TEST_R_OP_R(satub,       0xfff,     0xff,      USR_OVF);
694     TEST_R_OP_R(satub,       -1,        0,         USR_OVF);
695 
696     TEST_P_OP_PP(vaddubs,    0xfeLL,    0x01LL,    0xffLL,    USR_CLEAR);
697     TEST_P_OP_PP(vaddubs,    0xffLL,    0xffLL,    0xffLL,    USR_OVF);
698 
699     TEST_P_OP_PP(vadduhs,    0xfffeLL,  0x1LL,     0xffffLL,  USR_CLEAR);
700     TEST_P_OP_PP(vadduhs,    0xffffLL,  0x1LL,     0xffffLL,  USR_OVF);
701 
702     TEST_P_OP_PP(vsububs, 0x0807060504030201LL, 0x0101010101010101LL,
703                  0x0706050403020100LL, USR_CLEAR);
704     TEST_P_OP_PP(vsububs, 0x0807060504030201LL, 0x0202020202020202LL,
705                  0x0605040302010000LL, USR_OVF);
706 
707     TEST_P_OP_PP(vsubuhs, 0x0004000300020001LL, 0x0001000100010001LL,
708                  0x0003000200010000LL, USR_CLEAR);
709     TEST_P_OP_PP(vsubuhs, 0x0004000300020001LL, 0x0002000200020002LL,
710                  0x0002000100000000LL, USR_OVF);
711 
712     TEST_R_OP_PP(vaddhubs, 0x0004000300020001LL, 0x0001000100010001LL,
713                  0x05040302, USR_CLEAR);
714     TEST_R_OP_PP(vaddhubs, 0x7fff000300020001LL, 0x0002000200020002LL,
715                  0xff050403, USR_OVF);
716 
717     TEST_R_OP_P(vsathub,         0x0001000300020001LL, 0x01030201, USR_CLEAR);
718     TEST_R_OP_P(vsathub,         0x010000700080ffffLL, 0xff708000, USR_OVF);
719 
720     TEST_R_OP_P(vsatwuh,         0x0000ffff00000001LL, 0xffff0001, USR_CLEAR);
721     TEST_R_OP_P(vsatwuh,         0x800000000000ffffLL, 0x0000ffff, USR_OVF);
722 
723     TEST_P_OP_P(vsatwuh_nopack,  0x0000ffff00000001LL, 0x0000ffff00000001LL,
724                 USR_CLEAR);
725     TEST_P_OP_P(vsatwuh_nopack,  0x800000000000ffffLL, 0x000000000000ffffLL,
726                 USR_OVF);
727 
728     TEST_R_OP_R(svsathub,        0x00020001,           0x0201,     USR_CLEAR);
729     TEST_R_OP_R(svsathub,        0x0080ffff,           0x8000,     USR_OVF);
730 
731     TEST_R_OP_PI(asrhub_sat,     0x004f003f002f001fLL, 3,    0x09070503,
732                  USR_CLEAR);
733     TEST_R_OP_PI(asrhub_sat,     0x004fffff8fff001fLL, 3,    0x09000003,
734                  USR_OVF);
735 
736     TEST_R_OP_PI(asrhub_rnd_sat, 0x004f003f002f001fLL, 2,    0x0a080604,
737                  USR_CLEAR);
738     TEST_R_OP_PI(asrhub_rnd_sat, 0x004fffff8fff001fLL, 2,    0x0a000004,
739                  USR_OVF);
740 
741     TEST_R_OP_RR(addsat,        1,              2,              3,
742                  USR_CLEAR);
743     TEST_R_OP_RR(addsat,        0x7fffffff,     0x00000010,     0x7fffffff,
744                  USR_OVF);
745     TEST_R_OP_RR(addsat,        0x80000000,     0x80000006,     0x80000000,
746                  USR_OVF);
747 
748     TEST_P_OP_PP(addpsat, 1LL, 2LL, 3LL, USR_CLEAR);
749     /* overflow to max positive */
750     TEST_P_OP_PP(addpsat, 0x7ffffffffffffff0LL, 0x0000000000000010LL,
751                  0x7fffffffffffffffLL, USR_OVF);
752     /* overflow to min negative */
753     TEST_P_OP_PP(addpsat, 0x8000000000000003LL, 0x8000000000000006LL,
754                  0x8000000000000000LL, USR_OVF);
755 
756     TEST_XR_OP_RR(mpy_acc_sat_hh_s0, 0x7fffffff, 0xffff0000, 0x11110000,
757                   0x7fffeeee, USR_CLEAR);
758     TEST_XR_OP_RR(mpy_acc_sat_hh_s0, 0x7fffffff, 0x7fff0000, 0x7fff0000,
759                   0x7fffffff, USR_OVF);
760 
761     TEST_R_OP_RR(mpy_sat_hh_s1,        0xffff0000, 0x11110000, 0xffffddde,
762                  USR_CLEAR);
763     TEST_R_OP_RR(mpy_sat_hh_s1,        0x7fff0000, 0x7fff0000, 0x7ffe0002,
764                  USR_CLEAR);
765     TEST_R_OP_RR(mpy_sat_hh_s1,        0x80000000, 0x80000000, 0x7fffffff,
766                  USR_OVF);
767 
768     TEST_R_OP_RR(mpy_sat_rnd_hh_s1,    0xffff0000, 0x11110000, 0x00005dde,
769                  USR_CLEAR);
770     TEST_R_OP_RR(mpy_sat_rnd_hh_s1,    0x7fff0000, 0x7fff0000, 0x7ffe8002,
771                  USR_CLEAR);
772     TEST_R_OP_RR(mpy_sat_rnd_hh_s1,    0x80000000, 0x80000000, 0x7fffffff,
773                  USR_OVF);
774 
775     TEST_R_OP_RR(mpy_up_s1_sat,        0xffff0000, 0x11110000, 0xffffddde,
776                  USR_CLEAR);
777     TEST_R_OP_RR(mpy_up_s1_sat,        0x7fff0000, 0x7fff0000, 0x7ffe0002,
778                  USR_CLEAR);
779     TEST_R_OP_RR(mpy_up_s1_sat,        0x80000000, 0x80000000, 0x7fffffff,
780                  USR_OVF);
781 
782     TEST_P_OP_RR(vmpy2s_s1,  0x7fff0000, 0x7fff0000, 0x7ffe000200000000LL,
783                  USR_CLEAR);
784     TEST_P_OP_RR(vmpy2s_s1,  0x80000000, 0x80000000, 0x7fffffff00000000LL,
785                  USR_OVF);
786 
787     TEST_P_OP_RR(vmpy2su_s1, 0x7fff0000, 0x7fff0000, 0x7ffe000200000000LL,
788                  USR_CLEAR);
789     TEST_P_OP_RR(vmpy2su_s1, 0xffffbd97, 0xffffffff, 0xfffe000280000000LL,
790                  USR_OVF);
791 
792     TEST_R_OP_RR(vmpy2s_s1pack,        0x7fff0000, 0x7fff0000, 0x7ffe0000,
793                  USR_CLEAR);
794     TEST_R_OP_RR(vmpy2s_s1pack,        0x80008000, 0x80008000, 0x7fff7fff,
795                  USR_OVF);
796 
797     TEST_P_OP_PP(vmpy2es_s1, 0x7fff7fff7fff7fffLL, 0x1fff1fff1fff1fffLL,
798                  0x1ffec0021ffec002LL, USR_CLEAR);
799     TEST_P_OP_PP(vmpy2es_s1, 0x8000800080008000LL, 0x8000800080008000LL,
800                  0x7fffffff7fffffffLL, USR_OVF);
801 
802     TEST_R_OP_PP(vdmpyrs_s1, 0x7fff7fff7fff7fffLL, 0x1fff1fff1fff1fffLL,
803                  0x3ffe3ffe, USR_CLEAR);
804     TEST_R_OP_PP(vdmpyrs_s1, 0x8000800080008000LL, 0x8000800080008000LL,
805                  0x7fff7fffLL, USR_OVF);
806 
807     TEST_XP_OP_PP(vdmacs_s0, 0x0fffffffULL, 0x00ff00ff00ff00ffLL,
808                   0x00ff00ff00ff00ffLL, 0x0001fc021001fc01LL, USR_CLEAR);
809     TEST_XP_OP_PP(vdmacs_s0, 0x01111111ULL, 0x8000800080001000LL,
810                   0x8000800080008000LL, 0x7fffffff39111111LL, USR_OVF);
811 
812     TEST_R_OP_RR(cmpyrs_s0,            0x7fff0000, 0x7fff0000, 0x0000c001,
813                  USR_CLEAR);
814     TEST_R_OP_RR(cmpyrs_s0,            0x80008000, 0x80008000, 0x7fff0000,
815                  USR_OVF);
816 
817     TEST_XP_OP_RR(cmacs_s0, 0x0fffffff, 0x7fff0000, 0x7fff0000,
818                   0x00000000d000fffeLL, USR_CLEAR);
819     TEST_XP_OP_RR(cmacs_s0, 0x0fff1111, 0x80008000, 0x80008000,
820                   0x7fffffff0fff1111LL, USR_OVF);
821 
822     TEST_XP_OP_RR(cnacs_s0, 0x000000108fffffffULL, 0x7fff0000, 0x7fff0000,
823                   0x00000010cfff0000ULL, USR_CLEAR);
824     TEST_XP_OP_RR(cnacs_s0, 0x000000108ff1111fULL, 0x00002001, 0x00007ffd,
825                   0x0000001080000000ULL, USR_OVF);
826 
827     TEST_P_OP_PP(vrcmpys_s1_h, 0x00ff00ff00ff00ffLL, 0x00ff00ff00ff00ffLL,
828                  0x0003f8040003f804LL, USR_CLEAR);
829     TEST_P_OP_PP(vrcmpys_s1_h, 0x8000800080008000LL, 0x8000800080008000LL,
830                  0x7fffffff7fffffffLL, USR_OVF);
831 
832     TEST_XP_OP_PP(mmacls_s0, 0x6fffffff, 0x00ff00ff00ff00ffLL,
833                   0x00ff00ff00ff00ffLL, 0x0000fe017000fe00LL, USR_CLEAR);
834     TEST_XP_OP_PP(mmacls_s0, 0x6f1111ff, 0x8000800080008000LL,
835                   0x1000100080008000LL, 0xf80008007fffffffLL, USR_OVF);
836 
837     TEST_R_OP_RR(hmmpyl_rs1,           0x7fff0000, 0x7fff0001, 0x0000fffe,
838                  USR_CLEAR);
839     TEST_R_OP_RR(hmmpyl_rs1,           0x80000000, 0x80008000, 0x7fffffff,
840                  USR_OVF);
841 
842     TEST_XP_OP_PP(mmaculs_s0, 0x000000007fffffffULL, 0xffff800080008000LL,
843                   0xffff800080008000LL, 0xffffc00040003fffLL, USR_CLEAR);
844     TEST_XP_OP_PP(mmaculs_s0, 0x000011107fffffffULL, 0x00ff00ff00ff00ffLL,
845                   0x00ff00ff001100ffLL, 0x00010f117fffffffLL, USR_OVF);
846 
847     TEST_R_OP_PR(cmpyi_wh, 0x7fff000000000000LL, 0x7fff0001, 0x0000fffe,
848                  USR_CLEAR);
849     TEST_R_OP_PR(cmpyi_wh, 0x8000000000000000LL, 0x80008000, 0x7fffffff,
850                  USR_OVF);
851 
852     TEST_P_OP_PP(vcmpy_s0_sat_i, 0x00ff00ff00ff00ffLL, 0x00ff00ff00ff00ffLL,
853                  0x0001fc020001fc02LL, USR_CLEAR);
854     TEST_P_OP_PP(vcmpy_s0_sat_i, 0x8000800080008000LL, 0x8000800080008000LL,
855                  0x7fffffff7fffffffLL, USR_OVF);
856 
857     TEST_P_OP_PR(vcrotate, 0x8000000000000000LL, 0x00000002,
858                  0x8000000000000000LL, USR_CLEAR);
859     TEST_P_OP_PR(vcrotate, 0x7fff80007fff8000LL, 0x00000001,
860                  0x7fff80007fff7fffLL, USR_OVF);
861 
862     TEST_P_OP_PR(vcnegh, 0x8000000000000000LL, 0x00000002,
863                  0x8000000000000000LL, USR_CLEAR);
864     TEST_P_OP_PR(vcnegh, 0x7fff80007fff8000LL, 0x00000001,
865                  0x7fff80007fff7fffLL, USR_OVF);
866 
867 #if CORE_HAS_AUDIO
868     TEST_R_OP_PP(wcmpyrw, 0x8765432101234567LL, 0x00000002ffffffffLL,
869                  0x00000001, USR_CLEAR);
870     TEST_R_OP_PP(wcmpyrw, 0x800000007fffffffLL, 0x000000ff7fffffffLL,
871                  0x7fffffff, USR_OVF);
872     TEST_R_OP_PP(wcmpyrw, 0x7fffffff80000000LL, 0x7fffffff000000ffLL,
873                  0x80000000, USR_OVF);
874 #else
875     printf("Audio instructions skipped\n");
876 #endif
877 
878     TEST_R_OP_RR(addh_l16_sat_ll,      0x0000ffff, 0x00000002, 0x00000001,
879                  USR_CLEAR);
880     TEST_R_OP_RR(addh_l16_sat_ll,      0x00007fff, 0x00000005, 0x00007fff,
881                  USR_OVF);
882     TEST_R_OP_RR(addh_l16_sat_ll,      0x00008000, 0x00008000, 0xffff8000,
883                  USR_OVF);
884 
885     TEST_P_OP_P(vconj, 0x0000ffff00000001LL, 0x0000ffff00000001LL, USR_CLEAR);
886     TEST_P_OP_P(vconj, 0x800000000000ffffLL, 0x7fff00000000ffffLL, USR_OVF);
887 
888     TEST_P_OP_PP(vxaddsubw, 0x8765432101234567LL, 0x00000002ffffffffLL,
889                  0x8765432201234569LL, USR_CLEAR);
890     TEST_P_OP_PP(vxaddsubw, 0x7fffffff7fffffffLL, 0xffffffffffffffffLL,
891                  0x7fffffff7ffffffeLL, USR_OVF);
892     TEST_P_OP_PP(vxaddsubw, 0x800000000fffffffLL, 0x0000000a00000008LL,
893                  0x8000000010000009LL, USR_OVF);
894 
895     TEST_P_OP_P(vabshsat, 0x0001000afffff800LL, 0x0001000a00010800LL,
896                 USR_CLEAR);
897     TEST_P_OP_P(vabshsat, 0x8000000b000c000aLL, 0x7fff000b000c000aLL,
898              USR_OVF);
899 
900     TEST_P_OP_PP(vnavgwr, 0x8765432101234567LL, 0x00000002ffffffffLL,
901                  0xc3b2a1900091a2b4LL, USR_CLEAR);
902     TEST_P_OP_PP(vnavgwr, 0x7fffffff8000000aLL, 0x80000000ffffffffLL,
903                  0x7fffffffc0000006LL, USR_OVF);
904 
905     TEST_R_OP_RI(round_ri_sat,         0x0000ffff, 2, 0x00004000, USR_CLEAR);
906     TEST_R_OP_RI(round_ri_sat,         0x7fffffff, 2, 0x1fffffff, USR_OVF);
907 
908     TEST_R_OP_RR(asr_r_r_sat,          0x0000ffff, 0x00000002, 0x00003fff,
909                  USR_CLEAR);
910     TEST_R_OP_RR(asr_r_r_sat,          0x00ffffff, 0xfffffff5, 0x7fffffff,
911                  USR_OVF);
912     TEST_R_OP_RR(asr_r_r_sat,          0x80000000, 0xfffffff5, 0x80000000,
913                  USR_OVF);
914 
915     TEST_XPp_OP_PP(ACS, 0x0004000300020001ULL, 0x0001000200030004ULL,
916                    0x0000000000000000ULL, 0x0004000300030004ULL, 0xf0,
917                    USR_CLEAR);
918     TEST_XPp_OP_PP(ACS, 0x0004000300020001ULL, 0x0001000200030004ULL,
919                    0x000affff000d0000ULL, 0x000e0003000f0004ULL, 0xcc,
920                    USR_CLEAR);
921     TEST_XPp_OP_PP(ACS, 0x00047fff00020001ULL, 0x00017fff00030004ULL,
922                   0x000a0fff000d0000ULL, 0x000e7fff000f0004ULL, 0xfc,
923                   USR_OVF);
924     TEST_XPp_OP_PP(ACS, 0x00047fff00020001ULL, 0x00017fff00030004ULL,
925                    0x000a0fff000d0000ULL, 0x000e7fff000f0004ULL, 0xf0,
926                    USR_OVF);
927 
928     /* Floating point */
929     TEST_R_OP_RR(sfmin,  SF_one,      SF_small_neg,   SF_small_neg, USR_CLEAR);
930     TEST_R_OP_RR(sfmin,  SF_one,      SF_SNaN,        SF_one,       USR_FPINVF);
931     TEST_R_OP_RR(sfmin,  SF_SNaN,     SF_one,         SF_one,       USR_FPINVF);
932     TEST_R_OP_RR(sfmin,  SF_one,      SF_QNaN,        SF_one,       USR_CLEAR);
933     TEST_R_OP_RR(sfmin,  SF_QNaN,     SF_one,         SF_one,       USR_CLEAR);
934     TEST_R_OP_RR(sfmin,  SF_SNaN,     SF_QNaN,        SF_HEX_NaN,   USR_FPINVF);
935     TEST_R_OP_RR(sfmin,  SF_QNaN,     SF_SNaN,        SF_HEX_NaN,   USR_FPINVF);
936 
937     TEST_R_OP_RR(sfmax,  SF_one,      SF_small_neg,   SF_one,       USR_CLEAR);
938     TEST_R_OP_RR(sfmax,  SF_one,      SF_SNaN,        SF_one,       USR_FPINVF);
939     TEST_R_OP_RR(sfmax,  SF_SNaN,     SF_one,         SF_one,       USR_FPINVF);
940     TEST_R_OP_RR(sfmax,  SF_one,      SF_QNaN,        SF_one,       USR_CLEAR);
941     TEST_R_OP_RR(sfmax,  SF_QNaN,     SF_one,         SF_one,       USR_CLEAR);
942     TEST_R_OP_RR(sfmax,  SF_SNaN,     SF_QNaN,        SF_HEX_NaN,   USR_FPINVF);
943     TEST_R_OP_RR(sfmax,  SF_QNaN,     SF_SNaN,        SF_HEX_NaN,   USR_FPINVF);
944 
945     TEST_R_OP_RR(sfadd,  SF_one,      SF_QNaN,        SF_HEX_NaN,   USR_CLEAR);
946     TEST_R_OP_RR(sfadd,  SF_one,      SF_SNaN,        SF_HEX_NaN,   USR_FPINVF);
947     TEST_R_OP_RR(sfadd,  SF_QNaN,     SF_SNaN,        SF_HEX_NaN,   USR_FPINVF);
948     TEST_R_OP_RR(sfadd,  SF_SNaN,     SF_QNaN,        SF_HEX_NaN,   USR_FPINVF);
949 
950     TEST_R_OP_RR(sfsub,  SF_one,      SF_QNaN,        SF_HEX_NaN,   USR_CLEAR);
951     TEST_R_OP_RR(sfsub,  SF_one,      SF_SNaN,        SF_HEX_NaN,   USR_FPINVF);
952     TEST_R_OP_RR(sfsub,  SF_QNaN,     SF_SNaN,        SF_HEX_NaN,   USR_FPINVF);
953     TEST_R_OP_RR(sfsub,  SF_SNaN,     SF_QNaN,        SF_HEX_NaN,   USR_FPINVF);
954 
955     TEST_R_OP_RR(sfmpy,  SF_one,      SF_QNaN,        SF_HEX_NaN,   USR_CLEAR);
956     TEST_R_OP_RR(sfmpy,  SF_one,      SF_SNaN,        SF_HEX_NaN,   USR_FPINVF);
957     TEST_R_OP_RR(sfmpy,  SF_QNaN,     SF_SNaN,        SF_HEX_NaN,   USR_FPINVF);
958     TEST_R_OP_RR(sfmpy,  SF_SNaN,     SF_QNaN,        SF_HEX_NaN,   USR_FPINVF);
959 
960     TEST_XR_OP_RR(sffma, SF_one,   SF_one,    SF_one,   SF_two,     USR_CLEAR);
961     TEST_XR_OP_RR(sffma, SF_zero,  SF_one,    SF_QNaN,  SF_HEX_NaN, USR_CLEAR);
962     TEST_XR_OP_RR(sffma, SF_zero,  SF_one,    SF_SNaN,  SF_HEX_NaN, USR_FPINVF);
963     TEST_XR_OP_RR(sffma, SF_zero,  SF_QNaN,   SF_SNaN,  SF_HEX_NaN, USR_FPINVF);
964     TEST_XR_OP_RR(sffma, SF_zero,  SF_SNaN,   SF_QNaN,  SF_HEX_NaN, USR_FPINVF);
965 
966     TEST_XR_OP_RR(sffms, SF_one,   SF_one,    SF_one,   SF_zero,    USR_CLEAR);
967     TEST_XR_OP_RR(sffms, SF_zero,  SF_one,    SF_QNaN,  SF_HEX_NaN, USR_CLEAR);
968     TEST_XR_OP_RR(sffms, SF_zero,  SF_one,    SF_SNaN,  SF_HEX_NaN, USR_FPINVF);
969     TEST_XR_OP_RR(sffms, SF_zero,  SF_QNaN,   SF_SNaN,  SF_HEX_NaN, USR_FPINVF);
970     TEST_XR_OP_RR(sffms, SF_zero,  SF_SNaN,   SF_QNaN,  SF_HEX_NaN, USR_FPINVF);
971 
972     TEST_CMP_RR(sfcmpuo, SF_one,      SF_large_pos,    0x00,    USR_CLEAR);
973     TEST_CMP_RR(sfcmpuo, SF_INF,      SF_large_pos,    0x00,    USR_CLEAR);
974     TEST_CMP_RR(sfcmpuo, SF_QNaN,     SF_large_pos,    0xff,    USR_CLEAR);
975     TEST_CMP_RR(sfcmpuo, SF_QNaN_neg, SF_large_pos,    0xff,    USR_CLEAR);
976     TEST_CMP_RR(sfcmpuo, SF_SNaN,     SF_large_pos,    0xff,    USR_FPINVF);
977     TEST_CMP_RR(sfcmpuo, SF_SNaN_neg, SF_large_pos,    0xff,    USR_FPINVF);
978     TEST_CMP_RR(sfcmpuo, SF_QNaN,     SF_QNaN,         0xff,    USR_CLEAR);
979     TEST_CMP_RR(sfcmpuo, SF_QNaN,     SF_SNaN,         0xff,    USR_FPINVF);
980 
981     TEST_CMP_RR(sfcmpeq, SF_one,      SF_QNaN,         0x00,    USR_CLEAR);
982     TEST_CMP_RR(sfcmpeq, SF_one,      SF_SNaN,         0x00,    USR_FPINVF);
983     TEST_CMP_RR(sfcmpgt, SF_one,      SF_QNaN,         0x00,    USR_CLEAR);
984     TEST_CMP_RR(sfcmpgt, SF_one,      SF_SNaN,         0x00,    USR_FPINVF);
985     TEST_CMP_RR(sfcmpge, SF_one,      SF_QNaN,         0x00,    USR_CLEAR);
986     TEST_CMP_RR(sfcmpge, SF_one,      SF_SNaN,         0x00,    USR_FPINVF);
987 
988     TEST_P_OP_PP(dfadd,  DF_any,    DF_QNaN,         DF_HEX_NaN,    USR_CLEAR);
989     TEST_P_OP_PP(dfadd,  DF_any,    DF_SNaN,         DF_HEX_NaN,    USR_FPINVF);
990     TEST_P_OP_PP(dfadd,  DF_QNaN,   DF_SNaN,         DF_HEX_NaN,    USR_FPINVF);
991     TEST_P_OP_PP(dfadd,  DF_SNaN,   DF_QNaN,         DF_HEX_NaN,    USR_FPINVF);
992 
993     TEST_P_OP_PP(dfsub,  DF_any,    DF_QNaN,         DF_HEX_NaN,    USR_CLEAR);
994     TEST_P_OP_PP(dfsub,  DF_any,    DF_SNaN,         DF_HEX_NaN,    USR_FPINVF);
995     TEST_P_OP_PP(dfsub,  DF_QNaN,   DF_SNaN,         DF_HEX_NaN,    USR_FPINVF);
996     TEST_P_OP_PP(dfsub,  DF_SNaN,   DF_QNaN,         DF_HEX_NaN,    USR_FPINVF);
997 
998 #if CORE_IS_V67
999     TEST_P_OP_PP(dfmin,  DF_any,    DF_small_neg,    DF_small_neg,  USR_CLEAR);
1000     TEST_P_OP_PP(dfmin,  DF_any,    DF_SNaN,         DF_any,        USR_FPINVF);
1001     TEST_P_OP_PP(dfmin,  DF_SNaN,   DF_any,          DF_any,        USR_FPINVF);
1002     TEST_P_OP_PP(dfmin,  DF_any,    DF_QNaN,         DF_any,        USR_CLEAR);
1003     TEST_P_OP_PP(dfmin,  DF_QNaN,   DF_any,          DF_any,        USR_CLEAR);
1004     TEST_P_OP_PP(dfmin,  DF_SNaN,   DF_QNaN,         DF_HEX_NaN,    USR_FPINVF);
1005     TEST_P_OP_PP(dfmin,  DF_QNaN,   DF_SNaN,         DF_HEX_NaN,    USR_FPINVF);
1006 
1007     TEST_P_OP_PP(dfmax,  DF_any,    DF_small_neg,    DF_any,        USR_CLEAR);
1008     TEST_P_OP_PP(dfmax,  DF_any,    DF_SNaN,         DF_any,        USR_FPINVF);
1009     TEST_P_OP_PP(dfmax,  DF_SNaN,   DF_any,          DF_any,        USR_FPINVF);
1010     TEST_P_OP_PP(dfmax,  DF_any,    DF_QNaN,         DF_any,        USR_CLEAR);
1011     TEST_P_OP_PP(dfmax,  DF_QNaN,   DF_any,          DF_any,        USR_CLEAR);
1012     TEST_P_OP_PP(dfmax,  DF_SNaN,   DF_QNaN,         DF_HEX_NaN,    USR_FPINVF);
1013     TEST_P_OP_PP(dfmax,  DF_QNaN,   DF_SNaN,         DF_HEX_NaN,    USR_FPINVF);
1014 
1015     TEST_XP_OP_PP(dfmpyhh, DF_one,   DF_one,  DF_one,   DF_one_hh,  USR_CLEAR);
1016     TEST_XP_OP_PP(dfmpyhh, DF_zero,  DF_any,  DF_QNaN,  DF_HEX_NaN, USR_CLEAR);
1017     TEST_XP_OP_PP(dfmpyhh, DF_zero,  DF_any,  DF_SNaN,  DF_HEX_NaN, USR_FPINVF);
1018     TEST_XP_OP_PP(dfmpyhh, DF_zero,  DF_QNaN, DF_SNaN,  DF_HEX_NaN, USR_FPINVF);
1019     TEST_XP_OP_PP(dfmpyhh, DF_zero,  DF_SNaN, DF_QNaN,  DF_HEX_NaN, USR_FPINVF);
1020 #else
1021     printf("v67 instructions skipped\n");
1022 #endif
1023 
1024     TEST_CMP_PP(dfcmpuo, DF_small_neg, DF_any,          0x00,    USR_CLEAR);
1025     TEST_CMP_PP(dfcmpuo, DF_large_pos, DF_any,          0x00,    USR_CLEAR);
1026     TEST_CMP_PP(dfcmpuo, DF_QNaN,      DF_any,          0xff,    USR_CLEAR);
1027     TEST_CMP_PP(dfcmpuo, DF_QNaN_neg,  DF_any,          0xff,    USR_CLEAR);
1028     TEST_CMP_PP(dfcmpuo, DF_SNaN,      DF_any,          0xff,    USR_FPINVF);
1029     TEST_CMP_PP(dfcmpuo, DF_SNaN_neg,  DF_any,          0xff,    USR_FPINVF);
1030     TEST_CMP_PP(dfcmpuo, DF_QNaN,      DF_QNaN,         0xff,    USR_CLEAR);
1031     TEST_CMP_PP(dfcmpuo, DF_QNaN,      DF_SNaN,         0xff,    USR_FPINVF);
1032 
1033     TEST_CMP_PP(dfcmpeq, DF_any,       DF_QNaN,         0x00,    USR_CLEAR);
1034     TEST_CMP_PP(dfcmpeq, DF_any,       DF_SNaN,         0x00,    USR_FPINVF);
1035     TEST_CMP_PP(dfcmpgt, DF_any,       DF_QNaN,         0x00,    USR_CLEAR);
1036     TEST_CMP_PP(dfcmpgt, DF_any,       DF_SNaN,         0x00,    USR_FPINVF);
1037     TEST_CMP_PP(dfcmpge, DF_any,       DF_QNaN,         0x00,    USR_CLEAR);
1038     TEST_CMP_PP(dfcmpge, DF_any,       DF_SNaN,         0x00,    USR_FPINVF);
1039 
1040     TEST_P_OP_R(conv_sf2df,       SF_QNaN,  DF_HEX_NaN,             USR_CLEAR);
1041     TEST_P_OP_R(conv_sf2df,       SF_SNaN,  DF_HEX_NaN,             USR_FPINVF);
1042     TEST_R_OP_R(conv_sf2uw,       SF_QNaN,  0xffffffff,             USR_FPINVF);
1043     TEST_R_OP_R(conv_sf2uw,       SF_SNaN,  0xffffffff,             USR_FPINVF);
1044     TEST_R_OP_R(conv_sf2w,        SF_QNaN,  0xffffffff,             USR_FPINVF);
1045     TEST_R_OP_R(conv_sf2w,        SF_SNaN,  0xffffffff,             USR_FPINVF);
1046     TEST_P_OP_R(conv_sf2ud,       SF_QNaN,  0xffffffffffffffffULL,  USR_FPINVF);
1047     TEST_P_OP_R(conv_sf2ud,       SF_SNaN,  0xffffffffffffffffULL,  USR_FPINVF);
1048     TEST_P_OP_R(conv_sf2d,        SF_QNaN,  0xffffffffffffffffULL,  USR_FPINVF);
1049     TEST_P_OP_R(conv_sf2d,        SF_SNaN,  0xffffffffffffffffULL,  USR_FPINVF);
1050     TEST_R_OP_R(conv_sf2uw_chop,  SF_QNaN,  0xffffffff,             USR_FPINVF);
1051     TEST_R_OP_R(conv_sf2uw_chop,  SF_SNaN,  0xffffffff,             USR_FPINVF);
1052     TEST_R_OP_R(conv_sf2w_chop,   SF_QNaN,  0xffffffff,             USR_FPINVF);
1053     TEST_R_OP_R(conv_sf2w_chop,   SF_SNaN,  0xffffffff,             USR_FPINVF);
1054     TEST_P_OP_R(conv_sf2ud_chop,  SF_QNaN,  0xffffffffffffffffULL,  USR_FPINVF);
1055     TEST_P_OP_R(conv_sf2ud_chop,  SF_SNaN,  0xffffffffffffffffULL,  USR_FPINVF);
1056     TEST_P_OP_R(conv_sf2d_chop,   SF_QNaN,  0xffffffffffffffffULL,  USR_FPINVF);
1057     TEST_P_OP_R(conv_sf2d_chop,   SF_SNaN,  0xffffffffffffffffULL,  USR_FPINVF);
1058 
1059     TEST_R_OP_P(conv_df2sf,       DF_QNaN,  SF_HEX_NaN,             USR_CLEAR);
1060     TEST_R_OP_P(conv_df2sf,       DF_SNaN,  SF_HEX_NaN,             USR_FPINVF);
1061     TEST_R_OP_P(conv_df2uw,       DF_QNaN,  0xffffffff,             USR_FPINVF);
1062     TEST_R_OP_P(conv_df2uw,       DF_SNaN,  0xffffffff,             USR_FPINVF);
1063     TEST_R_OP_P(conv_df2w,        DF_QNaN,  0xffffffff,             USR_FPINVF);
1064     TEST_R_OP_P(conv_df2w,        DF_SNaN,  0xffffffff,             USR_FPINVF);
1065     TEST_P_OP_P(conv_df2ud,       DF_QNaN,  0xffffffffffffffffULL,  USR_FPINVF);
1066     TEST_P_OP_P(conv_df2ud,       DF_SNaN,  0xffffffffffffffffULL,  USR_FPINVF);
1067     TEST_P_OP_P(conv_df2d,        DF_QNaN,  0xffffffffffffffffULL,  USR_FPINVF);
1068     TEST_P_OP_P(conv_df2d,        DF_SNaN,  0xffffffffffffffffULL,  USR_FPINVF);
1069     TEST_R_OP_P(conv_df2uw_chop,  DF_QNaN,  0xffffffff,             USR_FPINVF);
1070     TEST_R_OP_P(conv_df2uw_chop,  DF_SNaN,  0xffffffff,             USR_FPINVF);
1071 
1072     /* Test for typo in HELPER(conv_df2uw_chop) */
1073     TEST_R_OP_P(conv_df2uw_chop, 0xffffff7f00000001ULL, 0xffffffff, USR_FPINVF);
1074 
1075     TEST_R_OP_P(conv_df2w_chop,   DF_QNaN,  0xffffffff,             USR_FPINVF);
1076     TEST_R_OP_P(conv_df2w_chop,   DF_SNaN,  0xffffffff,             USR_FPINVF);
1077     TEST_P_OP_P(conv_df2ud_chop,  DF_QNaN,  0xffffffffffffffffULL,  USR_FPINVF);
1078     TEST_P_OP_P(conv_df2ud_chop,  DF_SNaN,  0xffffffffffffffffULL,  USR_FPINVF);
1079     TEST_P_OP_P(conv_df2d_chop,   DF_QNaN,  0xffffffffffffffffULL,  USR_FPINVF);
1080     TEST_P_OP_P(conv_df2d_chop,   DF_SNaN,  0xffffffffffffffffULL,  USR_FPINVF);
1081 
1082     TEST_R_OP_R(conv_uw2sf,    0x00000001,             SF_one,      USR_CLEAR);
1083     TEST_R_OP_R(conv_uw2sf,    0x010020a5,             0x4b801052,  USR_FPINPF);
1084     TEST_R_OP_R(conv_w2sf,     0x00000001,             SF_one,      USR_CLEAR);
1085     TEST_R_OP_R(conv_w2sf,     0x010020a5,             0x4b801052,  USR_FPINPF);
1086     TEST_R_OP_P(conv_ud2sf,    0x0000000000000001ULL,  SF_one,      USR_CLEAR);
1087     TEST_R_OP_P(conv_ud2sf,    0x00000000010020a5ULL,  0x4b801052,  USR_FPINPF);
1088     TEST_R_OP_P(conv_d2sf,     0x0000000000000001ULL,  SF_one,      USR_CLEAR);
1089     TEST_R_OP_P(conv_d2sf,     0x00000000010020a5ULL,  0x4b801052,  USR_FPINPF);
1090 
1091     TEST_XR_OP_RRp(sffma_sc, SF_one,   SF_one,    SF_one,   1, SF_four,
1092                    USR_CLEAR);
1093     TEST_XR_OP_RRp(sffma_sc, SF_QNaN,  SF_one,    SF_one,   1, SF_HEX_NaN,
1094                    USR_CLEAR);
1095     TEST_XR_OP_RRp(sffma_sc, SF_one,   SF_QNaN,   SF_one,   1, SF_HEX_NaN,
1096                    USR_CLEAR);
1097     TEST_XR_OP_RRp(sffma_sc, SF_one,   SF_one,    SF_QNaN,  1, SF_HEX_NaN,
1098                    USR_CLEAR);
1099     TEST_XR_OP_RRp(sffma_sc, SF_SNaN,  SF_one,    SF_one,   1, SF_HEX_NaN,
1100                    USR_FPINVF);
1101     TEST_XR_OP_RRp(sffma_sc, SF_one,   SF_SNaN,   SF_one,   1, SF_HEX_NaN,
1102                    USR_FPINVF);
1103     TEST_XR_OP_RRp(sffma_sc, SF_one,   SF_one,    SF_SNaN,  1, SF_HEX_NaN,
1104                    USR_FPINVF);
1105 
1106     TEST_Rp_OP_RR(sfrecipa, SF_one,    SF_one,    SF_one_recip,   0x00,
1107                   USR_CLEAR);
1108     TEST_Rp_OP_RR(sfrecipa, SF_QNaN,   SF_one,    SF_HEX_NaN,     0x00,
1109                   USR_CLEAR);
1110     TEST_Rp_OP_RR(sfrecipa, SF_one,    SF_QNaN,   SF_HEX_NaN,     0x00,
1111                   USR_CLEAR);
1112     TEST_Rp_OP_RR(sfrecipa, SF_one,    SF_SNaN,   SF_HEX_NaN,     0x00,
1113                   USR_FPINVF);
1114     TEST_Rp_OP_RR(sfrecipa, SF_SNaN,   SF_one,    SF_HEX_NaN,     0x00,
1115                   USR_FPINVF);
1116 
1117     TEST_R_OP_RR(sffixupn, SF_one,     SF_one,    SF_one,       USR_CLEAR);
1118     TEST_R_OP_RR(sffixupn, SF_QNaN,    SF_one,    SF_HEX_NaN,   USR_CLEAR);
1119     TEST_R_OP_RR(sffixupn, SF_one,     SF_QNaN,   SF_HEX_NaN,   USR_CLEAR);
1120     TEST_R_OP_RR(sffixupn, SF_SNaN,    SF_one,    SF_HEX_NaN,   USR_FPINVF);
1121     TEST_R_OP_RR(sffixupn, SF_one,     SF_SNaN,   SF_HEX_NaN,   USR_FPINVF);
1122 
1123     TEST_R_OP_RR(sffixupd, SF_one,     SF_one,    SF_one,       USR_CLEAR);
1124     TEST_R_OP_RR(sffixupd, SF_QNaN,    SF_one,    SF_HEX_NaN,   USR_CLEAR);
1125     TEST_R_OP_RR(sffixupd, SF_one,     SF_QNaN,   SF_HEX_NaN,   USR_CLEAR);
1126     TEST_R_OP_RR(sffixupd, SF_SNaN,    SF_one,    SF_HEX_NaN,   USR_FPINVF);
1127     TEST_R_OP_RR(sffixupd, SF_one,     SF_SNaN,   SF_HEX_NaN,   USR_FPINVF);
1128 
1129     TEST_R_OP_R(sffixupr, SF_one,             SF_one,           USR_CLEAR);
1130     TEST_R_OP_R(sffixupr, SF_QNaN,            SF_HEX_NaN,       USR_CLEAR);
1131     TEST_R_OP_R(sffixupr, SF_SNaN,            SF_HEX_NaN,       USR_FPINVF);
1132 
1133     TEST_Rp_OP_R(sfinvsqrta, SF_one,        SF_one_invsqrta,  0x00, USR_CLEAR);
1134     TEST_Rp_OP_R(sfinvsqrta, SF_zero,       SF_one,           0x00, USR_CLEAR);
1135     TEST_Rp_OP_R(sfinvsqrta, SF_QNaN,       SF_HEX_NaN,       0x00, USR_CLEAR);
1136     TEST_Rp_OP_R(sfinvsqrta, SF_small_neg,  SF_HEX_NaN,       0x00, USR_FPINVF);
1137     TEST_Rp_OP_R(sfinvsqrta, SF_SNaN,       SF_HEX_NaN,       0x00, USR_FPINVF);
1138 
1139     puts(err ? "FAIL" : "PASS");
1140     return err;
1141 }
1142