1 | /* Function log1pf 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 | * 1+x = 2^k*(xh + xl) is computed in high-low parts; xh in [1, 2) |
23 | * Get short reciprocal approximation Rcp ~ 1/xh |
24 | * R = (Rcp*xh - 1.0) + Rcp*xl |
25 | * log1p(x) = k*log(2.0) - log(Rcp) + poly(R) |
26 | * log(Rcp) is tabulated |
27 | * |
28 | * |
29 | */ |
30 | |
31 | /* Offsets for data table __svml_slog1p_data_internal |
32 | */ |
33 | #define SgnMask 0 |
34 | #define sOne 64 |
35 | #define sPoly_1 128 |
36 | #define sPoly_2 192 |
37 | #define sPoly_3 256 |
38 | #define sPoly_4 320 |
39 | #define sPoly_5 384 |
40 | #define sPoly_6 448 |
41 | #define sPoly_7 512 |
42 | #define sPoly_8 576 |
43 | #define iHiDelta 640 |
44 | #define iLoRange 704 |
45 | #define iBrkValue 768 |
46 | #define iOffExpoMask 832 |
47 | #define sLn2 896 |
48 | |
49 | #include <sysdep.h> |
50 | |
51 | .section .text.evex512, "ax" , @progbits |
52 | ENTRY(_ZGVeN16v_log1pf_skx) |
53 | pushq %rbp |
54 | cfi_def_cfa_offset(16) |
55 | movq %rsp, %rbp |
56 | cfi_def_cfa(6, 16) |
57 | cfi_offset(6, -16) |
58 | andq $-64, %rsp |
59 | subq $192, %rsp |
60 | vmovups sOne+__svml_slog1p_data_internal(%rip), %zmm2 |
61 | |
62 | /* reduction: compute r, n */ |
63 | vmovups iBrkValue+__svml_slog1p_data_internal(%rip), %zmm12 |
64 | vmovups SgnMask+__svml_slog1p_data_internal(%rip), %zmm4 |
65 | vmovaps %zmm0, %zmm3 |
66 | |
67 | /* compute 1+x as high, low parts */ |
68 | vmaxps {sae}, %zmm3, %zmm2, %zmm5 |
69 | vminps {sae}, %zmm3, %zmm2, %zmm7 |
70 | vandnps %zmm3, %zmm4, %zmm1 |
71 | vpternlogd $255, %zmm4, %zmm4, %zmm4 |
72 | vaddps {rn-sae}, %zmm7, %zmm5, %zmm9 |
73 | vpsubd %zmm12, %zmm9, %zmm10 |
74 | vsubps {rn-sae}, %zmm9, %zmm5, %zmm6 |
75 | |
76 | /* check argument value ranges */ |
77 | vpaddd iHiDelta+__svml_slog1p_data_internal(%rip), %zmm9, %zmm8 |
78 | vpsrad $23, %zmm10, %zmm13 |
79 | vmovups sPoly_5+__svml_slog1p_data_internal(%rip), %zmm9 |
80 | vpcmpd $5, iLoRange+__svml_slog1p_data_internal(%rip), %zmm8, %k1 |
81 | vpslld $23, %zmm13, %zmm14 |
82 | vaddps {rn-sae}, %zmm7, %zmm6, %zmm15 |
83 | vcvtdq2ps {rn-sae}, %zmm13, %zmm0 |
84 | vpsubd %zmm14, %zmm2, %zmm13 |
85 | vmovups sPoly_8+__svml_slog1p_data_internal(%rip), %zmm7 |
86 | vmovups sPoly_1+__svml_slog1p_data_internal(%rip), %zmm14 |
87 | vmulps {rn-sae}, %zmm13, %zmm15, %zmm6 |
88 | vpandd iOffExpoMask+__svml_slog1p_data_internal(%rip), %zmm10, %zmm11 |
89 | vpaddd %zmm12, %zmm11, %zmm5 |
90 | vmovups sPoly_4+__svml_slog1p_data_internal(%rip), %zmm10 |
91 | vmovups sPoly_3+__svml_slog1p_data_internal(%rip), %zmm11 |
92 | vmovups sPoly_2+__svml_slog1p_data_internal(%rip), %zmm12 |
93 | |
94 | /* polynomial evaluation */ |
95 | vsubps {rn-sae}, %zmm2, %zmm5, %zmm2 |
96 | vaddps {rn-sae}, %zmm6, %zmm2, %zmm15 |
97 | vmovups sPoly_7+__svml_slog1p_data_internal(%rip), %zmm2 |
98 | vfmadd231ps {rn-sae}, %zmm15, %zmm7, %zmm2 |
99 | vpandnd %zmm8, %zmm8, %zmm4{%k1} |
100 | vmovups sPoly_6+__svml_slog1p_data_internal(%rip), %zmm8 |
101 | |
102 | /* combine and get argument value range mask */ |
103 | vptestmd %zmm4, %zmm4, %k0 |
104 | vfmadd213ps {rn-sae}, %zmm8, %zmm15, %zmm2 |
105 | kmovw %k0, %edx |
106 | vfmadd213ps {rn-sae}, %zmm9, %zmm15, %zmm2 |
107 | vfmadd213ps {rn-sae}, %zmm10, %zmm15, %zmm2 |
108 | vfmadd213ps {rn-sae}, %zmm11, %zmm15, %zmm2 |
109 | vfmadd213ps {rn-sae}, %zmm12, %zmm15, %zmm2 |
110 | vfmadd213ps {rn-sae}, %zmm14, %zmm15, %zmm2 |
111 | vmulps {rn-sae}, %zmm15, %zmm2, %zmm4 |
112 | vfmadd213ps {rn-sae}, %zmm15, %zmm15, %zmm4 |
113 | |
114 | /* final reconstruction */ |
115 | vmovups sLn2+__svml_slog1p_data_internal(%rip), %zmm15 |
116 | vfmadd213ps {rn-sae}, %zmm4, %zmm15, %zmm0 |
117 | vorps %zmm1, %zmm0, %zmm0 |
118 | testl %edx, %edx |
119 | |
120 | /* Go to special inputs processing branch */ |
121 | jne L(SPECIAL_VALUES_BRANCH) |
122 | # LOE rbx r12 r13 r14 r15 edx zmm0 zmm3 |
123 | |
124 | /* Restore registers |
125 | * and exit the function |
126 | */ |
127 | |
128 | L(EXIT): |
129 | movq %rbp, %rsp |
130 | popq %rbp |
131 | cfi_def_cfa(7, 8) |
132 | cfi_restore(6) |
133 | ret |
134 | cfi_def_cfa(6, 16) |
135 | cfi_offset(6, -16) |
136 | |
137 | /* Branch to process |
138 | * special inputs |
139 | */ |
140 | |
141 | L(SPECIAL_VALUES_BRANCH): |
142 | vmovups %zmm3, 64(%rsp) |
143 | vmovups %zmm0, 128(%rsp) |
144 | # LOE rbx r12 r13 r14 r15 edx zmm0 |
145 | |
146 | xorl %eax, %eax |
147 | # LOE rbx r12 r13 r14 r15 eax edx |
148 | |
149 | vzeroupper |
150 | movq %r12, 16(%rsp) |
151 | /* 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) */ |
152 | .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x50, 0xff, 0xff, 0xff, 0x22 |
153 | movl %eax, %r12d |
154 | movq %r13, 8(%rsp) |
155 | /* 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) */ |
156 | .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x48, 0xff, 0xff, 0xff, 0x22 |
157 | movl %edx, %r13d |
158 | movq %r14, (%rsp) |
159 | /* 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) */ |
160 | .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x40, 0xff, 0xff, 0xff, 0x22 |
161 | # LOE rbx r15 r12d r13d |
162 | |
163 | /* Range mask |
164 | * bits check |
165 | */ |
166 | |
167 | L(RANGEMASK_CHECK): |
168 | btl %r12d, %r13d |
169 | |
170 | /* Call scalar math function */ |
171 | jc L(SCALAR_MATH_CALL) |
172 | # LOE rbx r15 r12d r13d |
173 | |
174 | /* Special inputs |
175 | * processing loop |
176 | */ |
177 | |
178 | L(SPECIAL_VALUES_LOOP): |
179 | incl %r12d |
180 | cmpl $16, %r12d |
181 | |
182 | /* Check bits in range mask */ |
183 | jl L(RANGEMASK_CHECK) |
184 | # LOE rbx r15 r12d r13d |
185 | |
186 | movq 16(%rsp), %r12 |
187 | cfi_restore(12) |
188 | movq 8(%rsp), %r13 |
189 | cfi_restore(13) |
190 | movq (%rsp), %r14 |
191 | cfi_restore(14) |
192 | vmovups 128(%rsp), %zmm0 |
193 | |
194 | /* Go to exit */ |
195 | jmp L(EXIT) |
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 | /* 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) */ |
199 | .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x48, 0xff, 0xff, 0xff, 0x22 |
200 | /* 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) */ |
201 | .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x40, 0xff, 0xff, 0xff, 0x22 |
202 | # LOE rbx r12 r13 r14 r15 zmm0 |
203 | |
204 | /* Scalar math function call |
205 | * to process special input |
206 | */ |
207 | |
208 | L(SCALAR_MATH_CALL): |
209 | movl %r12d, %r14d |
210 | vmovss 64(%rsp, %r14, 4), %xmm0 |
211 | call log1pf@PLT |
212 | # LOE rbx r14 r15 r12d r13d xmm0 |
213 | |
214 | vmovss %xmm0, 128(%rsp, %r14, 4) |
215 | |
216 | /* Process special inputs in loop */ |
217 | jmp L(SPECIAL_VALUES_LOOP) |
218 | # LOE rbx r15 r12d r13d |
219 | END(_ZGVeN16v_log1pf_skx) |
220 | |
221 | .section .rodata, "a" |
222 | .align 64 |
223 | |
224 | #ifdef __svml_slog1p_data_internal_typedef |
225 | typedef unsigned int VUINT32; |
226 | typedef struct { |
227 | __declspec(align(64)) VUINT32 SgnMask[16][1]; |
228 | __declspec(align(64)) VUINT32 sOne[16][1]; |
229 | __declspec(align(64)) VUINT32 sPoly[8][16][1]; |
230 | __declspec(align(64)) VUINT32 iHiDelta[16][1]; |
231 | __declspec(align(64)) VUINT32 iLoRange[16][1]; |
232 | __declspec(align(64)) VUINT32 iBrkValue[16][1]; |
233 | __declspec(align(64)) VUINT32 iOffExpoMask[16][1]; |
234 | __declspec(align(64)) VUINT32 sLn2[16][1]; |
235 | } __svml_slog1p_data_internal; |
236 | #endif |
237 | __svml_slog1p_data_internal: |
238 | /* SgnMask */ |
239 | .long 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff |
240 | /* sOne = SP 1.0 */ |
241 | .align 64 |
242 | .long 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000 |
243 | /* sPoly[] = SP polynomial */ |
244 | .align 64 |
245 | .long 0xbf000000, 0xbf000000, 0xbf000000, 0xbf000000, 0xbf000000, 0xbf000000, 0xbf000000, 0xbf000000, 0xbf000000, 0xbf000000, 0xbf000000, 0xbf000000, 0xbf000000, 0xbf000000, 0xbf000000, 0xbf000000 /* -5.0000000000000000000000000e-01 P0 */ |
246 | .long 0x3eaaaa94, 0x3eaaaa94, 0x3eaaaa94, 0x3eaaaa94, 0x3eaaaa94, 0x3eaaaa94, 0x3eaaaa94, 0x3eaaaa94, 0x3eaaaa94, 0x3eaaaa94, 0x3eaaaa94, 0x3eaaaa94, 0x3eaaaa94, 0x3eaaaa94, 0x3eaaaa94, 0x3eaaaa94 /* 3.3333265781402587890625000e-01 P1 */ |
247 | .long 0xbe80058e, 0xbe80058e, 0xbe80058e, 0xbe80058e, 0xbe80058e, 0xbe80058e, 0xbe80058e, 0xbe80058e, 0xbe80058e, 0xbe80058e, 0xbe80058e, 0xbe80058e, 0xbe80058e, 0xbe80058e, 0xbe80058e, 0xbe80058e /* -2.5004237890243530273437500e-01 P2 */ |
248 | .long 0x3e4ce190, 0x3e4ce190, 0x3e4ce190, 0x3e4ce190, 0x3e4ce190, 0x3e4ce190, 0x3e4ce190, 0x3e4ce190, 0x3e4ce190, 0x3e4ce190, 0x3e4ce190, 0x3e4ce190, 0x3e4ce190, 0x3e4ce190, 0x3e4ce190, 0x3e4ce190 /* 2.0007920265197753906250000e-01 P3 */ |
249 | .long 0xbe28ad37, 0xbe28ad37, 0xbe28ad37, 0xbe28ad37, 0xbe28ad37, 0xbe28ad37, 0xbe28ad37, 0xbe28ad37, 0xbe28ad37, 0xbe28ad37, 0xbe28ad37, 0xbe28ad37, 0xbe28ad37, 0xbe28ad37, 0xbe28ad37, 0xbe28ad37 /* -1.6472326219081878662109375e-01 P4 */ |
250 | .long 0x3e0fcb12, 0x3e0fcb12, 0x3e0fcb12, 0x3e0fcb12, 0x3e0fcb12, 0x3e0fcb12, 0x3e0fcb12, 0x3e0fcb12, 0x3e0fcb12, 0x3e0fcb12, 0x3e0fcb12, 0x3e0fcb12, 0x3e0fcb12, 0x3e0fcb12, 0x3e0fcb12, 0x3e0fcb12 /* 1.4042308926582336425781250e-01 P5 */ |
251 | .long 0xbe1ad9e3, 0xbe1ad9e3, 0xbe1ad9e3, 0xbe1ad9e3, 0xbe1ad9e3, 0xbe1ad9e3, 0xbe1ad9e3, 0xbe1ad9e3, 0xbe1ad9e3, 0xbe1ad9e3, 0xbe1ad9e3, 0xbe1ad9e3, 0xbe1ad9e3, 0xbe1ad9e3, 0xbe1ad9e3, 0xbe1ad9e3 /* -1.5122179687023162841796875e-01 P6 */ |
252 | .long 0x3e0d84ed, 0x3e0d84ed, 0x3e0d84ed, 0x3e0d84ed, 0x3e0d84ed, 0x3e0d84ed, 0x3e0d84ed, 0x3e0d84ed, 0x3e0d84ed, 0x3e0d84ed, 0x3e0d84ed, 0x3e0d84ed, 0x3e0d84ed, 0x3e0d84ed, 0x3e0d84ed, 0x3e0d84ed /* 1.3820238411426544189453125e-01 P7 */ |
253 | /* iHiDelta = SP 80000000-7f000000 */ |
254 | .align 64 |
255 | .long 0x01000000, 0x01000000, 0x01000000, 0x01000000, 0x01000000, 0x01000000, 0x01000000, 0x01000000, 0x01000000, 0x01000000, 0x01000000, 0x01000000, 0x01000000, 0x01000000, 0x01000000, 0x01000000 |
256 | /* iLoRange = SP 00800000+iHiDelta */ |
257 | .align 64 |
258 | .long 0x01800000, 0x01800000, 0x01800000, 0x01800000, 0x01800000, 0x01800000, 0x01800000, 0x01800000, 0x01800000, 0x01800000, 0x01800000, 0x01800000, 0x01800000, 0x01800000, 0x01800000, 0x01800000 |
259 | /* iBrkValue = SP 2/3 */ |
260 | .align 64 |
261 | .long 0x3f2aaaab, 0x3f2aaaab, 0x3f2aaaab, 0x3f2aaaab, 0x3f2aaaab, 0x3f2aaaab, 0x3f2aaaab, 0x3f2aaaab, 0x3f2aaaab, 0x3f2aaaab, 0x3f2aaaab, 0x3f2aaaab, 0x3f2aaaab, 0x3f2aaaab, 0x3f2aaaab, 0x3f2aaaab |
262 | /* iOffExpoMask = SP significand mask */ |
263 | .align 64 |
264 | .long 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff |
265 | /* sLn2 = SP ln(2) */ |
266 | .align 64 |
267 | .long 0x3f317218, 0x3f317218, 0x3f317218, 0x3f317218, 0x3f317218, 0x3f317218, 0x3f317218, 0x3f317218, 0x3f317218, 0x3f317218, 0x3f317218, 0x3f317218, 0x3f317218, 0x3f317218, 0x3f317218, 0x3f317218 |
268 | .align 64 |
269 | .type __svml_slog1p_data_internal, @object |
270 | .size __svml_slog1p_data_internal, .-__svml_slog1p_data_internal |
271 | |