/* * Copyright (C) 2025 Linux Studio Plugins Project * (C) 2025 Vladimir Sadovnikov * * This file is part of lsp-dsp-lib * Created on: 22 апр. 2025 г. * * lsp-dsp-lib is free software: you can redistribute it and/or modify * it under the terms of the GNU Lesser General Public License as published by * the Free Software Foundation, either version 3 of the License, or * any later version. * * lsp-dsp-lib is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public License * along with lsp-dsp-lib. If not, see . */ #ifndef PRIVATE_DSP_ARCH_AARCH64_ASIMD_PMATH_LANCZOS_H_ #define PRIVATE_DSP_ARCH_AARCH64_ASIMD_PMATH_LANCZOS_H_ #ifndef PRIVATE_DSP_ARCH_AARCH64_ASIMD_IMPL #error "This header should not be included directly" #endif /* PRIVATE_DSP_ARCH_AARCH64_ASIMD_IMPL */ #include namespace lsp { namespace asimd { IF_ARCH_AARCH64( static const uint32_t lanczos_const[] __lsp_aligned16 = { LSP_DSP_VEC4(0x38d1b717), // +0x00: Sinc threshold = 1e-4 LSP_DSP_VEC4(0x3f800000), // +0x10: 1.0 }; ) typedef struct lanczos_gen_t { float x1[8]; // +0x00: Computed X1 float n[8]; // +0x20: Numerator } lanczos_gen_t; #define LANCZOS_GEN_FUNC_X8 \ /* v0 = x1[0] */ \ /* v1 = x1[1] */ \ __ASM_EMIT("fmul v2.4s, v0.4s, v27.4s") /* v2 = x2 = x1*a */ \ __ASM_EMIT("fmul v3.4s, v1.4s, v27.4s") \ __ASM_EMIT("fabs v4.4s, v0.4s") /* v4 = fabsf(x1) */ \ __ASM_EMIT("fabs v5.4s, v1.4s") \ __ASM_EMIT("fmul v8.4s, v0.4s, v2.4s") /* v8 = x1*x2 */ \ __ASM_EMIT("fmul v9.4s, v1.4s, v3.4s") \ __ASM_EMIT("frecpe v10.4s, v8.4s") /* v10 = s2 */ \ __ASM_EMIT("frecpe v11.4s, v9.4s") \ __ASM_EMIT("frecps v12.4s, v10.4s, v8.4s") /* v12 = (2 - R*s2) */ \ __ASM_EMIT("frecps v13.4s, v11.4s, v9.4s") \ __ASM_EMIT("fmul v10.4s, v12.4s, v10.4s") /* v10 = s2' = s2 * (2 - R*s2) */ \ __ASM_EMIT("fmul v11.4s, v13.4s, v11.4s") \ __ASM_EMIT("frecps v12.4s, v10.4s, v8.4s") /* v12 = (2 - R*s2') */ \ __ASM_EMIT("frecps v13.4s, v11.4s, v9.4s") \ __ASM_EMIT("fmul v8.4s, v12.4s, v10.4s") /* v8 = n = 1/(x1*x2) */ \ __ASM_EMIT("fmul v9.4s, v13.4s, v11.4s") \ __ASM_EMIT("fadd v0.4s, v0.4s, v16.4s") /* v0 = x1 + PI/2 */ \ __ASM_EMIT("fadd v1.4s, v1.4s, v16.4s") \ __ASM_EMIT("fadd v2.4s, v2.4s, v16.4s") /* v2 = x2 + PI/2 */ \ __ASM_EMIT("fadd v3.4s, v3.4s, v16.4s") \ __ASM_EMIT("stp q4, q5, [%[state], 0x00]") /* save fabfs(x1) */ \ __ASM_EMIT("stp q8, q9, [%[state], 0x20]") /* save n */ \ SINF_X_PLUS_PI_2_CORE_X16 \ /* v0 = sinf(x1[0]), v1 = sinf(x1[1]) */ \ /* v2 = sinf(x2[0]), v3 = sinf(x2[1]) */ \ __ASM_EMIT("ldp q4, q5, [%[state], 0x00]") /* v4 = fabsf(x1) */ \ __ASM_EMIT("ldp q6, q7, [%[state], 0x20]") /* v6 = n = 1/(x1*x2) */ \ __ASM_EMIT("ldp q8, q9, [%[LC]]") /* v8 = 1e-4, v9 = 1.0 */ \ __ASM_EMIT("fmul v0.4s, v0.4s, v2.4s") /* v0 = sinf(x1)*sinf(x2) */ \ __ASM_EMIT("fmul v1.4s, v1.4s, v3.4s") \ __ASM_EMIT("fcmge v2.4s, v4.4s, v8.4s") /* v2 = [ fabsf(x1) >= 1e-4 ] */ \ __ASM_EMIT("fcmge v3.4s, v5.4s, v8.4s") \ __ASM_EMIT("fmul v0.4s, v0.4s, v6.4s") /* v6 = sinf(x1)*sinf(x2)/(x1*x2) */ \ __ASM_EMIT("fmul v1.4s, v1.4s, v7.4s") \ __ASM_EMIT("fcmgt v4.4s, v26.4s, v4.4s") /* v4 = [ fabsf(x1) < t ] */ \ __ASM_EMIT("fcmgt v5.4s, v26.4s, v5.4s") \ __ASM_EMIT("bif v0.16b, v9.16b, v2.16b") /* v0 = [ fabsf(x1) >= 1e-4 ] ? f : 1.0 */ \ __ASM_EMIT("bif v1.16b, v9.16b, v3.16b") \ __ASM_EMIT("and v0.16b, v0.16b, v4.16b") /* v0 = [ fabsf(x1) < t ] ? ([ fabsf(x1) >= 1e-4 ] ? f : 1.0) : 0.0 */ \ __ASM_EMIT("and v1.16b, v1.16b, v5.16b") #define LANCZOS_GEN_FUNC_X4 \ /* v0 = x1 */ \ __ASM_EMIT("fmul v1.4s, v0.4s, v27.4s") /* v1 = x2 = x1*a */ \ __ASM_EMIT("fabs v2.4s, v0.4s") /* v2 = fabsf(x1) */ \ __ASM_EMIT("fmul v3.4s, v0.4s, v1.4s") /* v3 = x1*x2 */ \ __ASM_EMIT("frecpe v10.4s, v3.4s") /* v10 = s2 */ \ __ASM_EMIT("frecps v12.4s, v10.4s, v3.4s") /* v12 = (2 - R*s2) */ \ __ASM_EMIT("fmul v10.4s, v12.4s, v10.4s") /* v10 = s2' = s2 * (2 - R*s2) */ \ __ASM_EMIT("frecps v12.4s, v10.4s, v3.4s") /* v12 = (2 - R*s2') */ \ __ASM_EMIT("fmul v3.4s, v12.4s, v10.4s") /* v3 = n = 1/(x1*x2) */ \ __ASM_EMIT("fadd v0.4s, v0.4s, v16.4s") /* v0 = x1 + PI/2 */ \ __ASM_EMIT("fadd v1.4s, v1.4s, v16.4s") /* v1 = x2 + PI/2 */ \ SINF_X_PLUS_PI_2_CORE_X8 \ /* v0 = sinf(x1), v1 = sinf(x2) */ \ __ASM_EMIT("ldp q8, q9, [%[LC]]") /* v8 = 1e-4, v9 = 1.0 */ \ __ASM_EMIT("fmul v0.4s, v0.4s, v1.4s") /* v0 = sinf(x1)*sinf(x2) */ \ __ASM_EMIT("fcmge v4.4s, v2.4s, v8.4s") /* v4 = [ fabsf(x1) >= 1e-4 ] */ \ __ASM_EMIT("fmul v0.4s, v0.4s, v3.4s") /* v0 = sinf(x1)*sinf(x2)/(x1*x2) */ \ __ASM_EMIT("fcmgt v5.4s, v26.4s, v2.4s") /* v5 = [ fabsf(x1) < t ] */ \ __ASM_EMIT("bif v0.16b, v9.16b, v4.16b") /* v0 = [ fabsf(x1) >= 1e-4 ] ? f : 1.0 */ \ __ASM_EMIT("and v0.16b, v0.16b, v5.16b") /* v0 = [ fabsf(x1) < t ] ? ([ fabsf(x1) >= 1e-4 ] ? f : 1.0) : 0.0 */ void lanczos1(float *dst, float k, float p, float t, float a, size_t count) { IF_ARCH_AARCH64( lanczos_gen_t state; ); ARCH_AARCH64_ASM( __ASM_EMIT("ldp q16, q17, [%[S2C], #0x00]") /* v16 = PI/2, v17 = PI */ __ASM_EMIT("ldp q18, q19, [%[S2C], #0x20]") /* v18 = 1/(2*PI), v19 = C0 */ __ASM_EMIT("ldp q20, q21, [%[S2C], #0x40]") /* v20 = C1, v21 = C2 */ __ASM_EMIT("ldp q22, q23, [%[S2C], #0x60]") /* v22 = C3, v23 = C4 */ __ASM_EMIT("ld1r {v24.4s}, [%[k]]") /* v24 = k */ __ASM_EMIT("ld1r {v25.4s}, [%[p]]") /* v25 = p */ __ASM_EMIT("ld1r {v26.4s}, [%[t]]") /* v26 = t */ __ASM_EMIT("ld1r {v27.4s}, [%[a]]") /* v27 = a */ __ASM_EMIT("ldp q28, q29, [%[S2KP], #0x00]") /* v28 = i0 = 0 1 2 3, v29 = i1 = 4 5 6 7 */ __ASM_EMIT("ldp q30, q31, [%[S2KP], #0x20]") /* v30 = s0 = 8 8 8 8, v31 = s1 = 4 4 4 4 */ // x8 blocks __ASM_EMIT("subs %[count], %[count], #8") __ASM_EMIT("b.lo 2f") __ASM_EMIT("1:") __ASM_EMIT("fmul v0.4s, v28.4s, v24.4s") // v0 = k*i0 __ASM_EMIT("fmul v1.4s, v29.4s, v24.4s") // v1 = k*i1 __ASM_EMIT("fadd v28.4s, v28.4s, v30.4s") // v28 = i0 + step __ASM_EMIT("fadd v29.4s, v29.4s, v30.4s") // v29 = i1 + step __ASM_EMIT("fsub v0.4s, v0.4s, v25.4s") // v0 = k*i0 - p __ASM_EMIT("fsub v1.4s, v1.4s, v25.4s") // v1 = k*i1 - p LANCZOS_GEN_FUNC_X8 __ASM_EMIT("subs %[count], %[count], #8") __ASM_EMIT("stp q0, q1, [%[dst], #0x00]") __ASM_EMIT("add %[dst], %[dst], #0x20") __ASM_EMIT("b.hs 1b") __ASM_EMIT("2:") // x4 block __ASM_EMIT("adds %[count], %[count], #4") __ASM_EMIT("b.lt 4f") __ASM_EMIT("fmul v0.4s, v28.4s, v24.4s") // v0 = k*i0 __ASM_EMIT("fadd v28.4s, v28.4s, v31.4s") // v28 = i0 + step __ASM_EMIT("fsub v0.4s, v0.4s, v25.4s") // v0 = k*i0 - p LANCZOS_GEN_FUNC_X4 __ASM_EMIT("sub %[count], %[count], #4") __ASM_EMIT("str q0, [%[dst], #0x00]") __ASM_EMIT("add %[dst], %[dst], #0x10") __ASM_EMIT("4:") // Tail: 1x-3x block __ASM_EMIT("adds %[count], %[count], #4") __ASM_EMIT("b.ls 8f") __ASM_EMIT("fmul v0.4s, v28.4s, v24.4s") // v0 = k*i0 __ASM_EMIT("fsub v0.4s, v0.4s, v25.4s") // v0 = k*i0 - p LANCZOS_GEN_FUNC_X4 __ASM_EMIT("tst %[count], #1") __ASM_EMIT("b.eq 6f") __ASM_EMIT("st1 {v0.s}[0], [%[dst]]") __ASM_EMIT("add %[dst], %[dst], #0x04") __ASM_EMIT("ext v0.16b, v0.16b, v0.16b, #4") // v0 = S1 S2 S3 S0 __ASM_EMIT("6:") __ASM_EMIT("tst %[count], #2") __ASM_EMIT("b.eq 8f") __ASM_EMIT("st1 {v0.d}[0], [%[dst]]") // End __ASM_EMIT("8:") : [dst] "+r" (dst), [count] "+r" (count) : [k] "r" (&k), [p] "r" (&p), [t] "r" (&t), [a] "r" (&a), [S2C] "r" (&sinf_const[0]), [S2KP] "r" (&kp_gen_const[0]), [LC] "r" (&lanczos_const[0]), [state] "r" (&state) : "cc", "memory", "v0", "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v8", "v9", "v10", "v11", "v12", "v13", "v14", "v15", "v16", "v17", "v18", "v19", "v20", "v21", "v22", "v23", "v24", "v25", "v26", "v27", "v28", "v29", "v30", "v31" ); } } /* namespace asimd */ } /* namespace lsp */ #endif /* PRIVATE_DSP_ARCH_AARCH64_ASIMD_PMATH_LANCZOS_H_ */