1 | /* Function sinhf vectorized with AVX-512. |
2 | Copyright (C) 2021-2024 Free Software Foundation, Inc. |
3 | This file is part of the GNU C Library. |
4 | |
5 | The GNU C Library is free software; you can redistribute it and/or |
6 | modify it under the terms of the GNU Lesser General Public |
7 | License as published by the Free Software Foundation; either |
8 | version 2.1 of the License, or (at your option) any later version. |
9 | |
10 | The GNU C Library is distributed in the hope that it will be useful, |
11 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
12 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
13 | Lesser General Public License for more details. |
14 | |
15 | You should have received a copy of the GNU Lesser General Public |
16 | License along with the GNU C Library; if not, see |
17 | https://www.gnu.org/licenses/. */ |
18 | |
19 | /* |
20 | * ALGORITHM DESCRIPTION: |
21 | * |
22 | * Compute sinh(x) as (exp(x)-exp(-x))/2, |
23 | * where exp is calculated as |
24 | * exp(M*ln2 + ln2*(j/2^k) + r) = 2^M * 2^(j/2^k) * exp(r) |
25 | * |
26 | * Special cases: |
27 | * |
28 | * sinh(NaN) = quiet NaN, and raise invalid exception |
29 | * sinh(INF) = that INF |
30 | * sinh(x) = x for subnormals |
31 | * sinh(x) overflows for big x and returns MAXLOG+log(2) |
32 | * |
33 | */ |
34 | |
35 | /* Offsets for data table __svml_ssinh_data_internal |
36 | */ |
37 | #define _sInvLn2 0 |
38 | #define _sLn2hi 64 |
39 | #define _sLn2lo 128 |
40 | #define _sSign 192 |
41 | #define _sShifter 256 |
42 | #define _iDomainRange 320 |
43 | #define _sPC1 384 |
44 | #define _sPC2 448 |
45 | #define _sPC3 512 |
46 | #define _sPC4 576 |
47 | #define _sPC5 640 |
48 | #define _sPC6 704 |
49 | #define _iHalf 768 |
50 | |
51 | #include <sysdep.h> |
52 | |
53 | .section .text.evex512, "ax" , @progbits |
54 | ENTRY(_ZGVeN16v_sinhf_skx) |
55 | pushq %rbp |
56 | cfi_def_cfa_offset(16) |
57 | movq %rsp, %rbp |
58 | cfi_def_cfa(6, 16) |
59 | cfi_offset(6, -16) |
60 | andq $-64, %rsp |
61 | subq $192, %rsp |
62 | vmovaps %zmm0, %zmm5 |
63 | |
64 | /* |
65 | * Implementation |
66 | * Abs argument |
67 | */ |
68 | vandps _sSign+__svml_ssinh_data_internal(%rip), %zmm5, %zmm4 |
69 | |
70 | /* |
71 | * Check for overflow\underflow |
72 | * MORE faster than GE? |
73 | */ |
74 | vpternlogd $255, %zmm6, %zmm6, %zmm6 |
75 | vmovups _sShifter+__svml_ssinh_data_internal(%rip), %zmm7 |
76 | |
77 | /* |
78 | * Load argument |
79 | * dM = x/log(2) + RShifter |
80 | */ |
81 | vmovups _sInvLn2+__svml_ssinh_data_internal(%rip), %zmm11 |
82 | vmovups _sLn2hi+__svml_ssinh_data_internal(%rip), %zmm8 |
83 | vmovups _sLn2lo+__svml_ssinh_data_internal(%rip), %zmm10 |
84 | vmovups _iHalf+__svml_ssinh_data_internal(%rip), %zmm12 |
85 | vmovups _sPC5+__svml_ssinh_data_internal(%rip), %zmm0 |
86 | vmovups _sPC6+__svml_ssinh_data_internal(%rip), %zmm3 |
87 | |
88 | /* x^2 */ |
89 | vmovups _sPC2+__svml_ssinh_data_internal(%rip), %zmm2 |
90 | vxorps %zmm5, %zmm4, %zmm1 |
91 | vfmadd213ps {rn-sae}, %zmm7, %zmm1, %zmm11 |
92 | vpcmpd $2, _iDomainRange+__svml_ssinh_data_internal(%rip), %zmm1, %k1 |
93 | |
94 | /* |
95 | * G1, G2 2^N, 2^(-N) |
96 | * iM now is an EXP(2^N) |
97 | */ |
98 | vpslld $23, %zmm11, %zmm13 |
99 | |
100 | /* |
101 | * R |
102 | * sN = sM - RShifter |
103 | */ |
104 | vsubps {rn-sae}, %zmm7, %zmm11, %zmm9 |
105 | vpaddd %zmm13, %zmm12, %zmm14 |
106 | vpsubd %zmm13, %zmm12, %zmm15 |
107 | |
108 | /* sG1 = 2^(N-1)+2^(-N-1) */ |
109 | vaddps {rn-sae}, %zmm15, %zmm14, %zmm7 |
110 | vpandnd %zmm1, %zmm1, %zmm6{%k1} |
111 | |
112 | /* sR = sX - sN*Log2_hi */ |
113 | vfnmadd231ps {rn-sae}, %zmm8, %zmm9, %zmm1 |
114 | vptestmd %zmm6, %zmm6, %k0 |
115 | |
116 | /* sG2 = 2^(N-1)-2^(-N-1) */ |
117 | vsubps {rn-sae}, %zmm15, %zmm14, %zmm8 |
118 | |
119 | /* sR = (sX - sN*Log2_hi) - sN*Log2_lo */ |
120 | vfnmadd231ps {rn-sae}, %zmm10, %zmm9, %zmm1 |
121 | |
122 | /* |
123 | * sinh(r) = r*((a1=1)+r^2*(a3+r^2*(a5+{v1 r^2*a7})))) = r + r*(r^2*(a3+r^2*(a5+r^2*a7))) .... |
124 | * sSinh_r = (a3+r^2*a5) |
125 | */ |
126 | vmovups _sPC3+__svml_ssinh_data_internal(%rip), %zmm14 |
127 | kmovw %k0, %edx |
128 | |
129 | /* sR2 = sR^2 */ |
130 | vmulps {rn-sae}, %zmm1, %zmm1, %zmm6 |
131 | vfmadd231ps {rn-sae}, %zmm6, %zmm0, %zmm14 |
132 | |
133 | /* sSinh_r = r^2*(a3+r^2*a5) */ |
134 | vmulps {rn-sae}, %zmm6, %zmm14, %zmm0 |
135 | |
136 | /* sSinh_r = r + r*(r^2*(a3+r^2*a5)) */ |
137 | vfmadd213ps {rn-sae}, %zmm1, %zmm1, %zmm0 |
138 | |
139 | /* |
140 | * sinh(X) = sG2 + sG1*sinh(dR) + sG2*sR2*(a2+sR2*(a4+a6*sR2) |
141 | * sOut = (a4 +a6*sR2) |
142 | */ |
143 | vmovups _sPC4+__svml_ssinh_data_internal(%rip), %zmm1 |
144 | vfmadd231ps {rn-sae}, %zmm6, %zmm3, %zmm1 |
145 | |
146 | /* sOut = a2+sR2*(a4+a6*sR2) */ |
147 | vfmadd213ps {rn-sae}, %zmm2, %zmm6, %zmm1 |
148 | |
149 | /* sOut = sR2*(a2+sR2*(a4+a6*sR2) */ |
150 | vmulps {rn-sae}, %zmm6, %zmm1, %zmm2 |
151 | |
152 | /* sOut = sG2*sR2*(a2+sR2*(a4+a6*sR2) */ |
153 | vmulps {rn-sae}, %zmm8, %zmm2, %zmm3 |
154 | |
155 | /* sOut = sG1*sinh(dR)+sG2*sR2*(a2+sR2*(a4+a6*sR2) */ |
156 | vfmadd213ps {rn-sae}, %zmm3, %zmm0, %zmm7 |
157 | |
158 | /* sOut = sG2 + sG1*sinh(dR) + sG2*sR2*(a2+sR2*(a4+a6*sR2) */ |
159 | vaddps {rn-sae}, %zmm8, %zmm7, %zmm9 |
160 | |
161 | /* Ret H */ |
162 | vorps %zmm9, %zmm4, %zmm0 |
163 | testl %edx, %edx |
164 | |
165 | /* Go to special inputs processing branch */ |
166 | jne L(SPECIAL_VALUES_BRANCH) |
167 | # LOE rbx r12 r13 r14 r15 edx zmm0 zmm5 |
168 | |
169 | /* Restore registers |
170 | * and exit the function |
171 | */ |
172 | |
173 | L(EXIT): |
174 | movq %rbp, %rsp |
175 | popq %rbp |
176 | cfi_def_cfa(7, 8) |
177 | cfi_restore(6) |
178 | ret |
179 | cfi_def_cfa(6, 16) |
180 | cfi_offset(6, -16) |
181 | |
182 | /* Branch to process |
183 | * special inputs |
184 | */ |
185 | |
186 | L(SPECIAL_VALUES_BRANCH): |
187 | vmovups %zmm5, 64(%rsp) |
188 | vmovups %zmm0, 128(%rsp) |
189 | # LOE rbx r12 r13 r14 r15 edx zmm0 |
190 | |
191 | xorl %eax, %eax |
192 | # LOE rbx r12 r13 r14 r15 eax edx |
193 | |
194 | vzeroupper |
195 | movq %r12, 16(%rsp) |
196 | /* DW_CFA_expression: r12 (r12) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -176; DW_OP_plus) */ |
197 | .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x50, 0xff, 0xff, 0xff, 0x22 |
198 | movl %eax, %r12d |
199 | movq %r13, 8(%rsp) |
200 | /* DW_CFA_expression: r13 (r13) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -184; DW_OP_plus) */ |
201 | .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x48, 0xff, 0xff, 0xff, 0x22 |
202 | movl %edx, %r13d |
203 | movq %r14, (%rsp) |
204 | /* DW_CFA_expression: r14 (r14) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -192; DW_OP_plus) */ |
205 | .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x40, 0xff, 0xff, 0xff, 0x22 |
206 | # LOE rbx r15 r12d r13d |
207 | |
208 | /* Range mask |
209 | * bits check |
210 | */ |
211 | |
212 | L(RANGEMASK_CHECK): |
213 | btl %r12d, %r13d |
214 | |
215 | /* Call scalar math function */ |
216 | jc L(SCALAR_MATH_CALL) |
217 | # LOE rbx r15 r12d r13d |
218 | |
219 | /* Special inputs |
220 | * processing loop |
221 | */ |
222 | |
223 | L(SPECIAL_VALUES_LOOP): |
224 | incl %r12d |
225 | cmpl $16, %r12d |
226 | |
227 | /* Check bits in range mask */ |
228 | jl L(RANGEMASK_CHECK) |
229 | # LOE rbx r15 r12d r13d |
230 | |
231 | movq 16(%rsp), %r12 |
232 | cfi_restore(12) |
233 | movq 8(%rsp), %r13 |
234 | cfi_restore(13) |
235 | movq (%rsp), %r14 |
236 | cfi_restore(14) |
237 | vmovups 128(%rsp), %zmm0 |
238 | |
239 | /* Go to exit */ |
240 | jmp L(EXIT) |
241 | /* DW_CFA_expression: r12 (r12) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -176; DW_OP_plus) */ |
242 | .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x50, 0xff, 0xff, 0xff, 0x22 |
243 | /* DW_CFA_expression: r13 (r13) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -184; DW_OP_plus) */ |
244 | .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x48, 0xff, 0xff, 0xff, 0x22 |
245 | /* DW_CFA_expression: r14 (r14) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -192; DW_OP_plus) */ |
246 | .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x40, 0xff, 0xff, 0xff, 0x22 |
247 | # LOE rbx r12 r13 r14 r15 zmm0 |
248 | |
249 | /* Scalar math function call |
250 | * to process special input |
251 | */ |
252 | |
253 | L(SCALAR_MATH_CALL): |
254 | movl %r12d, %r14d |
255 | vmovss 64(%rsp, %r14, 4), %xmm0 |
256 | call sinhf@PLT |
257 | # LOE rbx r14 r15 r12d r13d xmm0 |
258 | |
259 | vmovss %xmm0, 128(%rsp, %r14, 4) |
260 | |
261 | /* Process special inputs in loop */ |
262 | jmp L(SPECIAL_VALUES_LOOP) |
263 | # LOE rbx r15 r12d r13d |
264 | END(_ZGVeN16v_sinhf_skx) |
265 | |
266 | .section .rodata, "a" |
267 | .align 64 |
268 | |
269 | #ifdef __svml_ssinh_data_internal_typedef |
270 | typedef unsigned int VUINT32; |
271 | typedef struct { |
272 | __declspec(align(64)) VUINT32 _sInvLn2[16][1]; |
273 | __declspec(align(64)) VUINT32 _sLn2hi[16][1]; |
274 | __declspec(align(64)) VUINT32 _sLn2lo[16][1]; |
275 | __declspec(align(64)) VUINT32 _sSign[16][1]; |
276 | __declspec(align(64)) VUINT32 _sShifter[16][1]; |
277 | __declspec(align(64)) VUINT32 _iDomainRange[16][1]; |
278 | __declspec(align(64)) VUINT32 _sPC1[16][1]; |
279 | __declspec(align(64)) VUINT32 _sPC2[16][1]; |
280 | __declspec(align(64)) VUINT32 _sPC3[16][1]; |
281 | __declspec(align(64)) VUINT32 _sPC4[16][1]; |
282 | __declspec(align(64)) VUINT32 _sPC5[16][1]; |
283 | __declspec(align(64)) VUINT32 _sPC6[16][1]; |
284 | __declspec(align(64)) VUINT32 _iHalf[16][1]; |
285 | } __svml_ssinh_data_internal; |
286 | #endif |
287 | __svml_ssinh_data_internal: |
288 | .long 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B, 0x3FB8AA3B /* _sInvLn2 */ // k=0 |
289 | .align 64 |
290 | .long 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000, 0x3F317000 /* _sLn2hi */ |
291 | .align 64 |
292 | .long 0x3805FDF4, 0x3805FDF4, 0x3805FDF4, 0x3805FDF4, 0x3805FDF4, 0x3805FDF4, 0x3805FDF4, 0x3805FDF4, 0x3805FDF4, 0x3805FDF4, 0x3805FDF4, 0x3805FDF4, 0x3805FDF4, 0x3805FDF4, 0x3805FDF4, 0x3805FDF4 /* _sLn2lo */ |
293 | .align 64 |
294 | .long 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000 /* _sSign */ |
295 | .align 64 |
296 | .long 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000 /* _sShifter */ |
297 | .align 64 |
298 | .long 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E, 0x42AEAC4E /* _iDomainRange */ |
299 | .align 64 |
300 | .long 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000, 0x3F800000 /* _sPC1=1 */ |
301 | .align 64 |
302 | .long 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000 /* _sPC2 */ |
303 | .align 64 |
304 | .long 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57, 0x3e2aaa57 /* _sPC3 */ |
305 | .align 64 |
306 | .long 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72, 0x3d2aaa72 /* _sPC4 */ |
307 | .align 64 |
308 | .long 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461, 0x3c091461 /* _sPC5 */ |
309 | .align 64 |
310 | .long 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3, 0x3ab6a8a3 /* _sPC6 */ |
311 | // Integer constants |
312 | .align 64 |
313 | .long 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000 /* _iHalf */ |
314 | .align 64 |
315 | .type __svml_ssinh_data_internal, @object |
316 | .size __svml_ssinh_data_internal, .-__svml_ssinh_data_internal |
317 | |