Debian Codegraph @

modules/lsp-dsp-lib/include/private/dsp/arch/arm/neon-d32/pmath/sqr.h

raw
1 /*
2 * Copyright (C) 2023 Linux Studio Plugins Project <https://lsp-plug.in/>
3 * (C) 2023 Vladimir Sadovnikov <[email protected]>
4 *
5 * This file is part of lsp-dsp-lib
6 * Created on: 20 сент. 2023 г.
7 *
8 * lsp-dsp-lib is free software: you can redistribute it and/or modify
9 * it under the terms of the GNU Lesser General Public License as published by
10 * the Free Software Foundation, either version 3 of the License, or
11 * any later version.
12 *
13 * lsp-dsp-lib is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU Lesser General Public License for more details.
17 *
18 * You should have received a copy of the GNU Lesser General Public License
19 * along with lsp-dsp-lib. If not, see <https://www.gnu.org/licenses/>.
20 */
21
22 #ifndef PRIVATE_DSP_ARCH_ARM_NEON_D32_PMATH_SQR_H_
23 #define PRIVATE_DSP_ARCH_ARM_NEON_D32_PMATH_SQR_H_
24
25 #ifndef PRIVATE_DSP_ARCH_ARM_NEON_D32_IMPL
26 #error "This header should not be included directly"
27 #endif /* PRIVATE_DSP_ARCH_ARM_NEON_D32_IMPL */
28
29 namespace lsp
30 {
31 namespace neon_d32
32 {
33 #define SQR_CORE(DST, SRC, INC) \
34 /* 32x blocks */ \
35 __ASM_EMIT("subs %[count], #32") \
36 __ASM_EMIT("blo 2f") \
37 __ASM_EMIT("1:") \
38 __ASM_EMIT("vldm %[" SRC "]" INC ", {q0-q7}") \
39 __ASM_EMIT("vmul.f32 q0, q0, q0") \
40 __ASM_EMIT("vmul.f32 q1, q1, q1") \
41 __ASM_EMIT("vmul.f32 q2, q2, q2") \
42 __ASM_EMIT("vmul.f32 q3, q3, q3") \
43 __ASM_EMIT("vmul.f32 q4, q4, q4") \
44 __ASM_EMIT("vmul.f32 q5, q5, q5") \
45 __ASM_EMIT("vmul.f32 q6, q6, q6") \
46 __ASM_EMIT("vmul.f32 q7, q7, q7") \
47 __ASM_EMIT("subs %[count], #32") \
48 __ASM_EMIT("vstm %[" DST "]!, {q0-q7}") \
49 __ASM_EMIT("bhs 1b") \
50 __ASM_EMIT("2:") \
51 /* 16x block */ \
52 __ASM_EMIT("adds %[count], #16") \
53 __ASM_EMIT("blt 4f") \
54 __ASM_EMIT("vldm %[" SRC "]" INC ", {q0-q3}") \
55 __ASM_EMIT("vmul.f32 q0, q0, q0") \
56 __ASM_EMIT("vmul.f32 q1, q1, q1") \
57 __ASM_EMIT("vmul.f32 q2, q2, q2") \
58 __ASM_EMIT("vmul.f32 q3, q3, q3") \
59 __ASM_EMIT("subs %[count], #16") \
60 __ASM_EMIT("vstm %[" DST "]!, {q0-q3}") \
61 __ASM_EMIT("4:") \
62 /* 8x block */ \
63 __ASM_EMIT("adds %[count], #8") \
64 __ASM_EMIT("blt 6f") \
65 __ASM_EMIT("vldm %[" SRC "]" INC ", {q0-q1}") \
66 __ASM_EMIT("vmul.f32 q0, q0, q0") \
67 __ASM_EMIT("vmul.f32 q1, q1, q1") \
68 __ASM_EMIT("subs %[count], #8") \
69 __ASM_EMIT("vstm %[" DST "]!, {q0-q1}") \
70 __ASM_EMIT("6:") \
71 /* 4x block */ \
72 __ASM_EMIT("adds %[count], #4") \
73 __ASM_EMIT("blt 8f") \
74 __ASM_EMIT("vldm %[" SRC "]" INC ", {q0}") \
75 __ASM_EMIT("vmul.f32 q0, q0, q0") \
76 __ASM_EMIT("subs %[count], #4") \
77 __ASM_EMIT("vstm %[" DST "]!, {q0}") \
78 __ASM_EMIT("8:") \
79 /* 1x block */ \
80 __ASM_EMIT("adds %[count], #3") /* 4 - 3 */ \
81 __ASM_EMIT("blt 10f") \
82 __ASM_EMIT("9:") \
83 __ASM_EMIT("vld1.32 {d0[], d1[]}, [%[" SRC "]]" INC) \
84 __ASM_EMIT("vmul.f32 q0, q0, q0") \
85 __ASM_EMIT("subs %[count], #1") \
86 __ASM_EMIT("vst1.32 {d0[0]}, [%[" DST "]]!") \
87 __ASM_EMIT("bge 9b") \
88 __ASM_EMIT("10:")
89
90 void sqr1(float *dst, size_t count)
91 {
92 ARCH_ARM_ASM(
93 SQR_CORE("dst", "dst", "")
94 : [dst] "+r" (dst), [count] "+r" (count)
95 :
96 : "cc", "memory",
97 "q0", "q1", "q2", "q3",
98 "q4", "q5", "q6", "q7"
99 );
100 }
101
102 void sqr2(float *dst, const float *src, size_t count)
103 {
104 ARCH_ARM_ASM(
105 SQR_CORE("dst", "src", "!")
106 : [dst] "+r" (dst), [src] "+r" (src), [count] "+r" (count)
107 :
108 : "cc", "memory",
109 "q0", "q1", "q2", "q3",
110 "q4", "q5", "q6", "q7"
111 );
112 }
113
114 #undef SQR_CORE
115
116 } /* namespace neon_d32 */
117 } /* namespace lsp */
118
119
120 #endif /* PRIVATE_DSP_ARCH_ARM_NEON_D32_PMATH_SQR_H_ */