/* * Copyright (C) 2025 Linux Studio Plugins Project * (C) 2025 Vladimir Sadovnikov * * This file is part of lsp-dsp-lib * Created on: 16 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_X86_SSE2_PMATH_LANCZOS_H_ #define PRIVATE_DSP_ARCH_X86_SSE2_PMATH_LANCZOS_H_ #ifndef PRIVATE_DSP_ARCH_X86_SSE2_IMPL #error "This header should not be included directly" #endif /* PRIVATE_DSP_ARCH_X86_SSE2_IMPL */ #include namespace lsp { namespace sse2 { IF_ARCH_X86( static const uint32_t lanczos_const[] __lsp_aligned16 = { LSP_DSP_VEC4(0x7fffffff), // +0x00: Mask for fabsf LSP_DSP_VEC4(0x38d1b717), // +0x10: Sinc threshold = 1e-4 LSP_DSP_VEC4(0x3f800000), // +0x20: 1.0 }; ) typedef struct lanczos_gen_t { float k[4]; // +0x00: Multiplier float p[4]; // +0x10: Initial phase float a[4]; // +0x20: Number of lobes float i[4]; // +0x30: Indices float x1[4]; // +0x40: Computed X1 float d[4]; // +0x50: Denominator float t[4]; // +0x60: Time range } lanczos_gen_t; #define LANCZOS_GEN_FUNC_X4 \ /* xmm0 = x1 */ \ __ASM_EMIT("movaps %%xmm0, %%xmm4") /* xmm4 = x1 */ \ __ASM_EMIT("movaps %%xmm0, %%xmm2") /* xmm2 = x1 */ \ __ASM_EMIT("mulps 0x20(%[state]), %%xmm4") /* xmm4 = x2 = x1*a */ \ __ASM_EMIT("movaps %%xmm0, 0x40(%[state])") /* *x = x1 */ \ __ASM_EMIT("mulps %%xmm4, %%xmm2") /* xmm2 = d = x1 * x2 */ \ __ASM_EMIT("addps 0x00 + %[S2C], %%xmm0") /* xmm0 = x1 + PI/2 */ \ __ASM_EMIT("addps 0x00 + %[S2C], %%xmm4") /* xmm4 = x2 + PI/2 */ \ __ASM_EMIT("movaps %%xmm2, 0x50(%[state])") /* *d = d */ \ SINF_X_PLUS_PI_2_CORE_X8 /* xmm0 = sinf(x1), xmm4 = sinf(x2) */ \ __ASM_EMIT("movaps 0x40(%[state]), %%xmm1") /* xmm1 = x */ \ __ASM_EMIT("mulps %%xmm4, %%xmm0") /* xmm0 = sinf(x1)*sinf(x2) */ \ __ASM_EMIT("andps 0x00 + %[LC], %%xmm1") /* xmm1 = fabsf(x) */ \ __ASM_EMIT("divps 0x50(%[state]), %%xmm0") /* xmm0 = F = (sinf(x1)*sinf(x2)) / (x1 * x2) */ \ __ASM_EMIT("movaps %%xmm1, %%xmm2") /* xmm2 = fabsf(x) */ \ __ASM_EMIT("cmpps $5, 0x10 + %[LC], %%xmm1") /* xmm1 = [ fabsf(x) >= 1e-4 ] */ \ __ASM_EMIT("andps %%xmm1, %%xmm0") /* xmm0 = [ fabsf(x) >= 1e-4 ] & f */ \ __ASM_EMIT("cmpps $1, 0x60(%[state]), %%xmm2") /* xmm2 = [ fabsf(x) < t ] */ \ __ASM_EMIT("andnps 0x20 + %[LC], %%xmm1") /* xmm1 = [ fabsf(x) < 1e-4 ] & 1.0 */ \ __ASM_EMIT("orps %%xmm1, %%xmm0") /* xmm0 = [ fabsf(x) >= 1e-4 ] ? f : 1.0 */ \ __ASM_EMIT("andps %%xmm2, %%xmm0") /* xmm0 = [ fabsf(x) < t ] ? ([ fabsf(x) >= 1e-4 ] ? f : 1.0) : 0.0 */ void lanczos1(float *dst, float k, float p, float t, float a, size_t count) { IF_ARCH_X86( lanczos_gen_t state __lsp_aligned16; float stub[4] __lsp_aligned16; ); ARCH_X86_ASM( // Prepare __ASM_EMIT("movss %[k], %%xmm4") // xmm4 = k __ASM_EMIT("movss %[p], %%xmm5") // xmm5 = p __ASM_EMIT("movss %[a], %%xmm6") // xmm6 = a __ASM_EMIT("movss %[t], %%xmm7") // xmm7 = t __ASM_EMIT("shufps $0x00, %%xmm4, %%xmm4") // xmm4 = k k k k __ASM_EMIT("shufps $0x00, %%xmm5, %%xmm5") // xmm5 = p p p p __ASM_EMIT("shufps $0x00, %%xmm6, %%xmm6") // xmm6 = a a a a __ASM_EMIT("shufps $0x00, %%xmm7, %%xmm7") // xmm7 = t t t t __ASM_EMIT("movaps %%xmm4, 0x00(%[state])") __ASM_EMIT("movaps %%xmm5, 0x10(%[state])") __ASM_EMIT("movaps %%xmm6, 0x20(%[state])") __ASM_EMIT("movaps %%xmm7, 0x60(%[state])") __ASM_EMIT("movaps 0x00 + %[LGEN], %%xmm1") // xmm1 = i = 0 1 2 3 // x4 blocks __ASM_EMIT("sub $4, %[count]") __ASM_EMIT("jb 2f") __ASM_EMIT("1:") __ASM_EMIT("movaps %%xmm1, %%xmm0") // xmm0 = i __ASM_EMIT("mulps 0x00(%[state]), %%xmm0") // xmm0 = k*i __ASM_EMIT("addps 0x10 + %[LGEN], %%xmm1") // xmm1 = i' = i + step __ASM_EMIT("subps 0x10(%[state]), %%xmm0") // xmm0 = x1 = k*i - p __ASM_EMIT("movaps %%xmm1, 0x30(%[state])") // *i = i' LANCZOS_GEN_FUNC_X4 __ASM_EMIT("movaps 0x30(%[state]), %%xmm1") // xmm1 = i' __ASM_EMIT("movups %%xmm0, 0x00(%[dst])") __ASM_EMIT("add $0x10, %[dst]") __ASM_EMIT("sub $4, %[count]") __ASM_EMIT("jae 1b") __ASM_EMIT("2:") // Tail: 1x-3x block __ASM_EMIT("add $4, %[count]") __ASM_EMIT("jle 6f") __ASM_EMIT("movaps %%xmm1, %%xmm0") // xmm0 = i __ASM_EMIT("mulps 0x00(%[state]), %%xmm0") // xmm0 = k*i __ASM_EMIT("subps 0x10(%[state]), %%xmm0") // xmm0 = x1 = k*i - p LANCZOS_GEN_FUNC_X4 __ASM_EMIT("test $1, %[count]") __ASM_EMIT("jz 4f") __ASM_EMIT("movss %%xmm0, 0x00(%[dst])") __ASM_EMIT("shufps $0x39, %%xmm0, %%xmm0") // xmm0 = S1 S2 S3 S0 __ASM_EMIT("add $4, %[dst]") __ASM_EMIT("4:") __ASM_EMIT("test $2, %[count]") __ASM_EMIT("jz 6f") __ASM_EMIT("movlps %%xmm0, 0x00(%[dst])") // End __ASM_EMIT("6:") : [dst] "+r" (dst), [count] "+r" (count) : [state] "r" (&state), [S2C] "o" (sinf_const), [LGEN] "o" (kp_gen_const), [LC] "o" (lanczos_const), [stub] "o" (stub), [k] "m" (k), [p] "m" (p), [a] "m" (a), [t] "m" (t) : "cc", "memory", "%xmm0", "%xmm1", "%xmm2", "%xmm3", "%xmm4", "%xmm5", "%xmm6", "%xmm7" ); } } /* namespace sse2 */ } /* namespace lsp */ #endif /* PRIVATE_DSP_ARCH_X86_SSE2_PMATH_LANCZOS_H_ */