/* * Copyright (C) 2026 Linux Studio Plugins Project * (C) 2026 Vladimir Sadovnikov * * This file is part of lsp-dsp-lib * Created on: 17 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_AVX2_PMATH_LANCZOS_H_ #define PRIVATE_DSP_ARCH_X86_AVX2_PMATH_LANCZOS_H_ #ifndef PRIVATE_DSP_ARCH_X86_AVX2_IMPL #error "This header should not be included directly" #endif /* PRIVATE_DSP_ARCH_X86_AVX2_IMPL */ #include namespace lsp { namespace avx2 { IF_ARCH_X86( static const uint32_t lanczos_const[] __lsp_aligned32 = { LSP_DSP_VEC8(0x7fffffff), // +0x00: Mask for fabsf LSP_DSP_VEC8(0x38d1b717), // +0x20: Sinc threshold = 1e-4 LSP_DSP_VEC8(0x3f800000), // +0x40: 1.0 }; ) typedef struct lanczos_gen_t { float k[8]; // +0x00: Multiplier float p[8]; // +0x20: Initial phase float a[8]; // +0x40: Number of lobes float i[8]; // +0x60: Indices float x1[8]; // +0x80: Computed X1 float d[8]; // +0xa0: Denominator float t[8]; // +0xc0: Time range } lanczos_gen_t; #define LANCZOS_GEN_FUNC_X8 \ /* ymm0 = x1 */ \ __ASM_EMIT("vmulps 0x40(%[state]), %%ymm0, %%ymm4") /* ymm4 = x2 = x1*a */ \ __ASM_EMIT("vmovaps %%ymm0, 0x80(%[state])") /* *x1 = x1 */ \ __ASM_EMIT("vmulps %%ymm0, %%ymm4, %%ymm2") /* ymm2 = d = x1 * x2 */ \ __ASM_EMIT("vaddps 0x00 + %[S2C], %%ymm0, %%ymm0") /* ymm0 = x1 + PI/2 */ \ __ASM_EMIT("vaddps 0x00 + %[S2C], %%ymm4, %%ymm4") /* ymm4 = x2 + PI/2 */ \ __ASM_EMIT("vmovaps %%ymm2, 0xa0(%[state])") /* *d = d */ \ SINF_X_PLUS_PI_2_CORE_X16 /* ymm0 = sinf(x1), ymm4 = sinf(x2) */ \ __ASM_EMIT("vmovaps 0x80(%[state]), %%ymm1") /* ymm1 = x1 */ \ __ASM_EMIT("vmulps %%ymm4, %%ymm0, %%ymm0") /* ymm0 = sinf(x1)*sinf(x2) */ \ __ASM_EMIT("vandps 0x00 + %[LC], %%ymm1, %%ymm1") /* ymm1 = fabsf(x) */ \ __ASM_EMIT("vdivps 0xa0(%[state]), %%ymm0, %%ymm0") /* ymm0 = F = (sinf(x1)*sinf(x2)) / (x1 * x2) */ \ __ASM_EMIT("vcmpps $1, 0x20 + %[LC], %%ymm1, %%ymm2") /* ymm2 = [ fabsf(x) < 1e-4 ] */ \ __ASM_EMIT("vcmpps $1, 0xc0(%[state]), %%ymm1, %%ymm1") /* ymm1 = [ fabsf(x) < t ] */ \ __ASM_EMIT("vblendvps %%ymm2, 0x40 + %[LC], %%ymm0, %%ymm0") /* ymm0 = [ fabsf(x) >= 1e-4 ] ? f : 1.0 */ \ __ASM_EMIT("vandps %%ymm1, %%ymm0, %%ymm0") /* ymm0 = [ fabsf(x) < t ] ? ([ fabsf(x) >= 1e-4 ] ? f : 1.0) : 0.0 */ #define LANCZOS_GEN_FUNC_X4 \ /* xmm0 = x1 */ \ __ASM_EMIT("vmulps 0x40(%[state]), %%xmm0, %%xmm4" ) /* xmm4 = x2 = x1*a */ \ __ASM_EMIT("vmovaps %%xmm0, 0x80(%[state])") /* *x1 = x1 */ \ __ASM_EMIT("vmulps %%xmm0, %%xmm4, %%xmm2") /* xmm2 = d = x1 * x2 */ \ __ASM_EMIT("vinsertf128 $1, %%xmm4, %%ymm0, %%ymm0") /* ymm0 = x1 x1 x1 x1 x2 x2 x2 x2 */ \ __ASM_EMIT("vmovaps %%xmm2, 0xa0(%[state])") /* *d = d */ \ __ASM_EMIT("vaddps 0x00 + %[S2C], %%ymm0, %%ymm0") /* xmm0 = x1 + PI/2 ... x2 + PI/2 ... */ \ SINF_X_PLUS_PI_2_CORE_X8 /* ymm0 = sinf(x1) ... sinf(x2) ... */ \ __ASM_EMIT("vextractf128 $1, %%ymm0, %%xmm4") /* xmm4 = sif(x2) */ \ __ASM_EMIT("vmovaps 0x80(%[state]), %%xmm1") /* xmm1 = x1 */ \ __ASM_EMIT("vmulps %%xmm4, %%xmm0, %%xmm0") /* xmm0 = sinf(x1)*sinf(x2) */ \ __ASM_EMIT("vandps 0x00 + %[LC], %%xmm1, %%xmm1") /* xmm1 = fabsf(x) */ \ __ASM_EMIT("vdivps 0xa0(%[state]), %%xmm0, %%xmm0") /* xmm0 = F = (sinf(x1)*sinf(x2)) / (x1 * x2) */ \ __ASM_EMIT("vcmpps $1, 0x20 + %[LC], %%xmm1, %%xmm2") /* xmm2 = [ fabsf(x) < 1e-4 ] */ \ __ASM_EMIT("vcmpps $1, 0xc0(%[state]), %%xmm1, %%xmm1") /* xmm1 = [ fabsf(x) < t ] */ \ __ASM_EMIT("vblendvps %%xmm2, 0x40 + %[LC], %%xmm0, %%xmm0") /* xmm0 = [ fabsf(x) >= 1e-4 ] ? f : 1.0 */ \ __ASM_EMIT("vandps %%xmm1, %%xmm0, %%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_aligned32; float stub[8] __lsp_aligned16; ); ARCH_X86_ASM( // Prepare __ASM_EMIT("vbroadcastss %[k], %%ymm4") // ymm4 = k __ASM_EMIT("vbroadcastss %[p], %%ymm5") // ymm5 = p __ASM_EMIT("vbroadcastss %[a], %%ymm6") // ymm6 = a __ASM_EMIT("vbroadcastss %[t], %%ymm7") // ymm7 = t __ASM_EMIT("vmovaps %%ymm4, 0x00(%[state])") __ASM_EMIT("vmovaps %%ymm5, 0x20(%[state])") __ASM_EMIT("vmovaps %%ymm6, 0x40(%[state])") __ASM_EMIT("vmovaps %%ymm7, 0xc0(%[state])") __ASM_EMIT("vmovaps 0x00 + %[LGEN], %%ymm1") // ymm1 = i = 0 1 2 3 4 5 6 7 // x8 blocks __ASM_EMIT("sub $8, %[count]") __ASM_EMIT("jb 2f") __ASM_EMIT("1:") __ASM_EMIT("vmulps 0x00(%[state]), %%ymm1, %%ymm0") // ymm0 = k*i __ASM_EMIT("vaddps 0x60 + %[LGEN], %%ymm1, %%ymm1") // ymm1 = i' = i + step __ASM_EMIT("vsubps 0x20(%[state]), %%ymm0, %%ymm0") // ymm0 = x1 = k*i - p __ASM_EMIT("vmovaps %%ymm1, 0x60(%[state])") // *i = i' LANCZOS_GEN_FUNC_X8 __ASM_EMIT("vmovaps 0x60(%[state]), %%ymm1") // ymm1 = i' __ASM_EMIT("vmovups %%ymm0, 0x00(%[dst])") __ASM_EMIT("add $0x20, %[dst]") __ASM_EMIT("sub $8, %[count]") __ASM_EMIT("jae 1b") __ASM_EMIT("2:") // x4 block __ASM_EMIT("add $4, %[count]") __ASM_EMIT("jl 4f") __ASM_EMIT("vmulps 0x00(%[state]), %%xmm1, %%xmm0") // xmm0 = k*i __ASM_EMIT("vaddps 0x80 + %[LGEN], %%xmm1, %%xmm1") // xmm1 = i' = i + step __ASM_EMIT("vsubps 0x20(%[state]), %%xmm0, %%xmm0") // xmm0 = x1 = k*i - p __ASM_EMIT("vmovaps %%xmm1, 0x60(%[state])") // *i = i' LANCZOS_GEN_FUNC_X4 __ASM_EMIT("vmovaps 0x60(%[state]), %%xmm1") // xmm1 = i' __ASM_EMIT("vmovups %%xmm0, 0x00(%[dst])") __ASM_EMIT("add $0x10, %[dst]") __ASM_EMIT("sub $4, %[count]") __ASM_EMIT("4:") // Tail: 1x-3x block __ASM_EMIT("add $4, %[count]") __ASM_EMIT("jle 8f") __ASM_EMIT("vmulps 0x00(%[state]), %%xmm1, %%xmm0") // xmm0 = k*i __ASM_EMIT("vsubps 0x20(%[state]), %%xmm0, %%xmm0") // xmm0 = x1 = k*i - p LANCZOS_GEN_FUNC_X4 __ASM_EMIT("test $1, %[count]") __ASM_EMIT("jz 6f") __ASM_EMIT("vmovss %%xmm0, 0x00(%[dst])") __ASM_EMIT("vshufps $0x39, %%xmm0, %%xmm0, %%xmm0") __ASM_EMIT("add $4, %[dst]") __ASM_EMIT("6:") __ASM_EMIT("test $2, %[count]") __ASM_EMIT("jz 8f") __ASM_EMIT("vmovlps %%xmm0, 0x00(%[dst])") // End __ASM_EMIT("8:") : [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" ); } #define LANCZOS_GEN_X64_FUNC_X8 \ /* ymm8 = k */ \ /* ymm9 = p */ \ /* ymm10 = t */ \ /* ymm11 = a */ \ /* ymm12 = x1 */ \ /* ymm13 = d */ \ /* ymm14 = i */ \ /* ymm15 = step */ \ __ASM_EMIT("vmulps %%ymm11, %%ymm12, %%ymm4") /* ymm12 = x2 = x1*a */ \ __ASM_EMIT("vmulps %%ymm12, %%ymm4, %%ymm13") /* ymm13 = d = x1 * x2 */ \ __ASM_EMIT("vaddps 0x00 + %[S2C], %%ymm12, %%ymm0") /* ymm0 = x1 + PI/2 */ \ __ASM_EMIT("vaddps 0x00 + %[S2C], %%ymm4, %%ymm4") /* ymm4 = x2 + PI/2 */ \ SINF_X_PLUS_PI_2_CORE_X16 /* ymm0 = sinf(x1), ymm4 = sinf(x2) */ \ __ASM_EMIT("vmulps %%ymm4, %%ymm0, %%ymm0") /* ymm0 = sinf(x1)*sinf(x2) */ \ __ASM_EMIT("vandps 0x00 + %[LC], %%ymm12, %%ymm12") /* ymm12 = fabsf(x1) */ \ __ASM_EMIT("vdivps %%ymm13, %%ymm0, %%ymm0") /* ymm0 = F = (sinf(x1)*sinf(x2)) / (x1 * x2) */ \ __ASM_EMIT("vcmpps $1, 0x20 + %[LC], %%ymm12, %%ymm2") /* ymm2 = [ fabsf(x1) < 1e-4 ] */ \ __ASM_EMIT("vcmpps $1, %%ymm10, %%ymm12, %%ymm1") /* ymm1 = [ fabsf(x1) < t ] */ \ __ASM_EMIT("vblendvps %%ymm2, 0x40 + %[LC], %%ymm0, %%ymm0") /* ymm0 = [ fabsf(x1) >= 1e-4 ] ? f : 1.0 */ \ __ASM_EMIT("vandps %%ymm1, %%ymm0, %%ymm0") /* ymm0 = [ fabsf(x1) < t ] ? ([ fabsf(x1) >= 1e-4 ] ? f : 1.0) : 0.0 */ #define LANCZOS_GEN_X64_FUNC_X4 \ /* xmm0 = x1 */ \ __ASM_EMIT("vmulps %%xmm11, %%xmm12, %%xmm4") /* xmm12 = x2 = x1*a */ \ __ASM_EMIT("vmulps %%xmm12, %%xmm4, %%xmm13") /* xmm13 = d = x1 * x2 */ \ __ASM_EMIT("vinsertf128 $1, %%xmm4, %%ymm12, %%ymm0") /* ymm0 = x1 x1 x1 x1 x2 x2 x2 x2 */ \ __ASM_EMIT("vaddps 0x00 + %[S2C], %%ymm0, %%ymm0") /* ymm0 = x1 + PI/2 ... x2 + PI/2 */ \ SINF_X_PLUS_PI_2_CORE_X8 /* ymm0 = sinf(x1) ... sinf(x2) */ \ __ASM_EMIT("vextractf128 $1, %%ymm0, %%xmm4") /* xmm4 = sif(x2) */ \ __ASM_EMIT("vmulps %%ymm4, %%ymm0, %%ymm0") /* ymm0 = sinf(x1)*sinf(x2) */ \ __ASM_EMIT("vandps 0x00 + %[LC], %%ymm12, %%ymm12") /* ymm12 = fabsf(x1) */ \ __ASM_EMIT("vdivps %%ymm13, %%ymm0, %%ymm0") /* ymm0 = F = (sinf(x1)*sinf(x2)) / (x1 * x2) */ \ __ASM_EMIT("vcmpps $1, 0x20 + %[LC], %%ymm12, %%ymm2") /* ymm2 = [ fabsf(x1) < 1e-4 ] */ \ __ASM_EMIT("vcmpps $1, %%ymm10, %%ymm12, %%ymm1") /* ymm1 = [ fabsf(x1) < t ] */ \ __ASM_EMIT("vblendvps %%ymm2, 0x40 + %[LC], %%ymm0, %%ymm0") /* ymm0 = [ fabsf(x1) >= 1e-4 ] ? f : 1.0 */ \ __ASM_EMIT("vandps %%ymm1, %%ymm0, %%ymm0") /* ymm0 = [ fabsf(x1) < t ] ? ([ fabsf(x1) >= 1e-4 ] ? f : 1.0) : 0.0 */ void x64_lanczos1(float *dst, float k, float p, float t, float a, size_t count) { ARCH_X86_64_ASM( // Prepare __ASM_EMIT("vbroadcastss %[k], %%ymm8") // ymm8 = k __ASM_EMIT("vbroadcastss %[p], %%ymm9") // ymm9 = p __ASM_EMIT("vbroadcastss %[t], %%ymm10") // ymm10 = t __ASM_EMIT("vbroadcastss %[a], %%ymm11") // ymm11 = a __ASM_EMIT("vmovaps 0x00 + %[LGEN], %%ymm14") // ymm14 = i = 0 1 2 3 4 5 6 7 // x8 blocks __ASM_EMIT("sub $8, %[count]") __ASM_EMIT("jb 2f") __ASM_EMIT("vmovaps 0x60 + %[LGEN], %%ymm15") // ymm15 = step __ASM_EMIT("1:") __ASM_EMIT("vmulps %%ymm8, %%ymm14, %%ymm12") // ymm12 = k*i __ASM_EMIT("vaddps %%ymm15, %%ymm14, %%ymm14") // ymm14 = i' = i + step __ASM_EMIT("vsubps %%ymm9, %%ymm12, %%ymm12") // ymm12 = x1 = k*i - p LANCZOS_GEN_X64_FUNC_X8 __ASM_EMIT("vmovups %%ymm0, 0x00(%[dst])") __ASM_EMIT("add $0x20, %[dst]") __ASM_EMIT("sub $8, %[count]") __ASM_EMIT("jae 1b") __ASM_EMIT("2:") // x4 block __ASM_EMIT("add $4, %[count]") __ASM_EMIT("jl 4f") __ASM_EMIT("vmulps %%xmm8, %%xmm14, %%xmm12") // xmm12 = k*i __ASM_EMIT("vaddps 0x80 + %[LGEN], %%xmm14, %%xmm14") // xmm14 = i' = i + step __ASM_EMIT("vsubps %%xmm9, %%xmm12, %%xmm12") // xmm12 = x1 = k*i - p LANCZOS_GEN_X64_FUNC_X4 __ASM_EMIT("vmovups %%xmm0, 0x00(%[dst])") __ASM_EMIT("add $0x10, %[dst]") __ASM_EMIT("sub $4, %[count]") __ASM_EMIT("4:") // Tail: 1x-3x block __ASM_EMIT("add $4, %[count]") __ASM_EMIT("jle 8f") __ASM_EMIT("vmulps %%xmm8, %%xmm14, %%xmm12") // xmm12 = k*i __ASM_EMIT("vsubps %%xmm9, %%xmm12, %%xmm12") // xmm12 = x1 = k*i - p LANCZOS_GEN_X64_FUNC_X4 __ASM_EMIT("test $1, %[count]") __ASM_EMIT("jz 6f") __ASM_EMIT("vmovss %%xmm0, 0x00(%[dst])") __ASM_EMIT("vshufps $0x39, %%xmm0, %%xmm0, %%xmm0") __ASM_EMIT("add $4, %[dst]") __ASM_EMIT("6:") __ASM_EMIT("test $2, %[count]") __ASM_EMIT("jz 8f") __ASM_EMIT("vmovlps %%xmm0, 0x00(%[dst])") // End __ASM_EMIT("8:") : [dst] "+r" (dst), [count] "+r" (count) : [S2C] "o" (sinf_const), [LGEN] "o" (kp_gen_const), [LC] "o" (lanczos_const), [k] "m" (k), [p] "m" (p), [a] "m" (a), [t] "m" (t) : "cc", "memory", "%xmm0", "%xmm1", "%xmm2", "%xmm3", "%xmm4", "%xmm5", "%xmm6", "%xmm7", "%xmm8", "%xmm9", "%xmm10", "%xmm11", "%xmm12", "%xmm13", "%xmm14", "%xmm15" ); } #define LANCZOS_GEN_FUNC_X8_FMA3 \ /* ymm0 = x1 */ \ __ASM_EMIT("vmulps 0x40(%[state]), %%ymm0, %%ymm4") /* ymm4 = x2 = x1*a */ \ __ASM_EMIT("vmovaps %%ymm0, 0x80(%[state])") /* *x1 = x1 */ \ __ASM_EMIT("vmulps %%ymm0, %%ymm4, %%ymm2") /* ymm2 = d = x1 * x2 */ \ __ASM_EMIT("vaddps 0x00 + %[S2C], %%ymm0, %%ymm0") /* ymm0 = x1 + PI/2 */ \ __ASM_EMIT("vaddps 0x00 + %[S2C], %%ymm4, %%ymm4") /* ymm4 = x2 + PI/2 */ \ __ASM_EMIT("vmovaps %%ymm2, 0xa0(%[state])") /* *d = d */ \ SINF_X_PLUS_PI_2_CORE_X16_FMA3 /* ymm0 = sinf(x1), ymm4 = sinf(x2) */ \ __ASM_EMIT("vmovaps 0x80(%[state]), %%ymm1") /* ymm1 = x1 */ \ __ASM_EMIT("vmulps %%ymm4, %%ymm0, %%ymm0") /* ymm0 = sinf(x1)*sinf(x2) */ \ __ASM_EMIT("vandps 0x00 + %[LC], %%ymm1, %%ymm1") /* ymm1 = fabsf(x) */ \ __ASM_EMIT("vdivps 0xa0(%[state]), %%ymm0, %%ymm0") /* ymm0 = F = (sinf(x1)*sinf(x2)) / (x1 * x2) */ \ __ASM_EMIT("vcmpps $1, 0x20 + %[LC], %%ymm1, %%ymm2") /* ymm2 = [ fabsf(x) < 1e-4 ] */ \ __ASM_EMIT("vcmpps $1, 0xc0(%[state]), %%ymm1, %%ymm1") /* ymm1 = [ fabsf(x) < t ] */ \ __ASM_EMIT("vblendvps %%ymm2, 0x40 + %[LC], %%ymm0, %%ymm0") /* ymm0 = [ fabsf(x) >= 1e-4 ] ? f : 1.0 */ \ __ASM_EMIT("vandps %%ymm1, %%ymm0, %%ymm0") /* ymm0 = [ fabsf(x) < t ] ? ([ fabsf(x) >= 1e-4 ] ? f : 1.0) : 0.0 */ #define LANCZOS_GEN_FUNC_X4_FMA3 \ /* xmm0 = x1 */ \ __ASM_EMIT("vmulps 0x40(%[state]), %%xmm0, %%xmm4") /* xmm4 = x2 = x1*a */ \ __ASM_EMIT("vmovaps %%xmm0, 0x80(%[state])") /* *x1 = x1 */ \ __ASM_EMIT("vmulps %%xmm0, %%xmm4, %%xmm2") /* xmm2 = d = x1 * x2 */ \ __ASM_EMIT("vinsertf128 $1, %%xmm4, %%ymm0, %%ymm0") /* ymm0 = x1 x1 x1 x1 x2 x2 x2 x2 */ \ __ASM_EMIT("vmovaps %%xmm2, 0xa0(%[state])") /* *d = d */ \ __ASM_EMIT("vaddps 0x00 + %[S2C], %%ymm0, %%ymm0") /* xmm0 = x1 + PI/2 ... x2 + PI/2 ... */ \ SINF_X_PLUS_PI_2_CORE_X8_FMA3 /* ymm0 = sinf(x1) ... sinf(x2) */ \ __ASM_EMIT("vextractf128 $1, %%ymm0, %%xmm4") /* xmm4 = sif(x2) */ \ __ASM_EMIT("vmovaps 0x80(%[state]), %%xmm1") /* xmm1 = x1 */ \ __ASM_EMIT("vmulps %%xmm4, %%xmm0, %%xmm0") /* xmm0 = sinf(x1)*sinf(x2) */ \ __ASM_EMIT("vandps 0x00 + %[LC], %%xmm1, %%xmm1") /* xmm1 = fabsf(x) */ \ __ASM_EMIT("vdivps 0xa0(%[state]), %%xmm0, %%xmm0") /* xmm0 = F = (sinf(x1)*sinf(x2)) / (x1 * x2) */ \ __ASM_EMIT("vcmpps $1, 0x20 + %[LC], %%xmm1, %%xmm2") /* xmm2 = [ fabsf(x) < 1e-4 ] */ \ __ASM_EMIT("vcmpps $1,0xc0(%[state]), %%xmm1, %%xmm1") /* xmm1 = [ fabsf(x) < t ] */ \ __ASM_EMIT("vblendvps %%xmm2, 0x40 + %[LC], %%xmm0, %%xmm0") /* xmm0 = [ fabsf(x) >= 1e-4 ] ? f : 1.0 */ \ __ASM_EMIT("vandps %%xmm1, %%xmm0, %%xmm0") /* xmm0 = [ fabsf(x) < t ] ? ([ fabsf(x) >= 1e-4 ] ? f : 1.0) : 0.0 */ void lanczos1_fma3(float *dst, float k, float p, float t, float a, size_t count) { IF_ARCH_X86( lanczos_gen_t state __lsp_aligned32; float stub[8] __lsp_aligned16; ); ARCH_X86_ASM( // Prepare __ASM_EMIT("vbroadcastss %[k], %%ymm4") // ymm4 = k __ASM_EMIT("vbroadcastss %[p], %%ymm5") // ymm5 = p __ASM_EMIT("vbroadcastss %[a], %%ymm6") // ymm6 = a __ASM_EMIT("vbroadcastss %[t], %%ymm7") // ymm7 = t __ASM_EMIT("vmovaps %%ymm4, 0x00(%[state])") __ASM_EMIT("vmovaps %%ymm5, 0x20(%[state])") __ASM_EMIT("vmovaps %%ymm6, 0x40(%[state])") __ASM_EMIT("vmovaps %%ymm7, 0xc0(%[state])") __ASM_EMIT("vmovaps 0x00 + %[LGEN], %%ymm1") // ymm1 = i = 0 1 2 3 4 5 6 7 // x8 blocks __ASM_EMIT("sub $8, %[count]") __ASM_EMIT("jb 2f") __ASM_EMIT("1:") __ASM_EMIT("vmulps 0x00(%[state]), %%ymm1, %%ymm0") // ymm0 = k*i __ASM_EMIT("vaddps 0x60 + %[LGEN], %%ymm1, %%ymm1") // ymm1 = i' = i + step __ASM_EMIT("vsubps 0x20(%[state]), %%ymm0, %%ymm0") // ymm0 = x1 = k*i - p __ASM_EMIT("vmovaps %%ymm1, 0x60(%[state])") // *i = i' LANCZOS_GEN_FUNC_X8_FMA3 __ASM_EMIT("vmovaps 0x60(%[state]), %%ymm1") // ymm1 = i' __ASM_EMIT("vmovups %%ymm0, 0x00(%[dst])") __ASM_EMIT("add $0x20, %[dst]") __ASM_EMIT("sub $8, %[count]") __ASM_EMIT("jae 1b") __ASM_EMIT("2:") // x4 block __ASM_EMIT("add $4, %[count]") __ASM_EMIT("jl 4f") __ASM_EMIT("vmulps 0x00(%[state]), %%xmm1, %%xmm0") // xmm0 = k*i __ASM_EMIT("vaddps 0x80 + %[LGEN], %%xmm1, %%xmm1") // xmm1 = i' = i + step __ASM_EMIT("vsubps 0x20(%[state]), %%xmm0, %%xmm0") // xmm0 = x1 = k*i - p __ASM_EMIT("vmovaps %%xmm1, 0x60(%[state])") // *i = i' LANCZOS_GEN_FUNC_X4_FMA3 __ASM_EMIT("vmovaps 0x60(%[state]), %%xmm1") // xmm1 = i' __ASM_EMIT("vmovups %%xmm0, 0x00(%[dst])") __ASM_EMIT("add $0x10, %[dst]") __ASM_EMIT("sub $4, %[count]") __ASM_EMIT("4:") // Tail: 1x-3x block __ASM_EMIT("add $4, %[count]") __ASM_EMIT("jle 8f") __ASM_EMIT("vmulps 0x00(%[state]), %%xmm1, %%xmm0") // xmm0 = k*i __ASM_EMIT("vsubps 0x20(%[state]), %%xmm0, %%xmm0") // xmm0 = x1 = k*i - p LANCZOS_GEN_FUNC_X4_FMA3 __ASM_EMIT("test $1, %[count]") __ASM_EMIT("jz 6f") __ASM_EMIT("vmovss %%xmm0, 0x00(%[dst])") __ASM_EMIT("vshufps $0x39, %%xmm0, %%xmm0, %%xmm0") __ASM_EMIT("add $4, %[dst]") __ASM_EMIT("6:") __ASM_EMIT("test $2, %[count]") __ASM_EMIT("jz 8f") __ASM_EMIT("vmovlps %%xmm0, 0x00(%[dst])") // End __ASM_EMIT("8:") : [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" ); } #define LANCZOS_GEN_X64_FUNC_X8_FMA3 \ /* ymm8 = k */ \ /* ymm9 = p */ \ /* ymm10 = t */ \ /* ymm11 = a */ \ /* ymm12 = x1 */ \ /* ymm13 = d */ \ /* ymm14 = i */ \ /* ymm15 = step */ \ __ASM_EMIT("vmulps %%ymm11, %%ymm12, %%ymm4") /* ymm12 = x2 = x1*a */ \ __ASM_EMIT("vmulps %%ymm12, %%ymm4, %%ymm13") /* ymm13 = d = x1 * x2 */ \ __ASM_EMIT("vaddps 0x00 + %[S2C], %%ymm12, %%ymm0") /* ymm0 = x1 + PI/2 */ \ __ASM_EMIT("vaddps 0x00 + %[S2C], %%ymm4, %%ymm4") /* ymm4 = x2 + PI/2 */ \ SINF_X_PLUS_PI_2_CORE_X16_FMA3 /* ymm0 = sinf(x1), ymm4 = sinf(x2) */ \ __ASM_EMIT("vmulps %%ymm4, %%ymm0, %%ymm0") /* ymm0 = sinf(x1)*sinf(x2) */ \ __ASM_EMIT("vandps 0x00 + %[LC], %%ymm12, %%ymm12") /* ymm12 = fabsf(x1) */ \ __ASM_EMIT("vdivps %%ymm13, %%ymm0, %%ymm0") /* ymm0 = F = (sinf(x1)*sinf(x2)) / (x1 * x2) */ \ __ASM_EMIT("vcmpps $1, 0x20 + %[LC], %%ymm12, %%ymm2") /* ymm2 = [ fabsf(x1) < 1e-4 ] */ \ __ASM_EMIT("vcmpps $1, %%ymm10, %%ymm12, %%ymm1") /* ymm1 = [ fabsf(x1) < t ] */ \ __ASM_EMIT("vblendvps %%ymm2, 0x40 + %[LC], %%ymm0, %%ymm0") /* ymm0 = [ fabsf(x1) >= 1e-4 ] ? f : 1.0 */ \ __ASM_EMIT("vandps %%ymm1, %%ymm0, %%ymm0") /* ymm0 = [ fabsf(x1) < t ] ? ([ fabsf(x1) >= 1e-4 ] ? f : 1.0) : 0.0 */ #define LANCZOS_GEN_X64_FUNC_X4_FMA3 \ /* xmm0 = x1 */ \ __ASM_EMIT("vmulps %%xmm11, %%xmm12, %%xmm4") /* xmm12 = x2 = x1*a */ \ __ASM_EMIT("vmulps %%xmm12, %%xmm4, %%xmm13") /* xmm13 = d = x1 * x2 */ \ __ASM_EMIT("vinsertf128 $1, %%xmm4, %%ymm12, %%ymm0") /* ymm0 = x1 x1 x1 x1 x2 x2 x2 x2 */ \ __ASM_EMIT("vaddps 0x00 + %[S2C], %%ymm0, %%ymm0") /* ymm0 = x1 + PI/2 ... x2 + PI/2 */ \ SINF_X_PLUS_PI_2_CORE_X8_FMA3 /* ymm0 = sinf(x1) ... sinf(x2) */ \ __ASM_EMIT("vextractf128 $1, %%ymm0, %%xmm4") /* xmm4 = sif(x2) */ \ __ASM_EMIT("vmulps %%ymm4, %%ymm0, %%ymm0") /* ymm0 = sinf(x1)*sinf(x2) */ \ __ASM_EMIT("vandps 0x00 + %[LC], %%ymm12, %%ymm12") /* ymm12 = fabsf(x1) */ \ __ASM_EMIT("vdivps %%ymm13, %%ymm0, %%ymm0") /* ymm0 = F = (sinf(x1)*sinf(x2)) / (x1 * x2) */ \ __ASM_EMIT("vcmpps $1, 0x20 + %[LC], %%ymm12, %%ymm2") /* ymm2 = [ fabsf(x1) < 1e-4 ] */ \ __ASM_EMIT("vcmpps $1, %%ymm10, %%ymm12, %%ymm1") /* ymm1 = [ fabsf(x1) < t ] */ \ __ASM_EMIT("vblendvps %%ymm2, 0x40 + %[LC], %%ymm0, %%ymm0") /* ymm0 = [ fabsf(x1) >= 1e-4 ] ? f : 1.0 */ \ __ASM_EMIT("vandps %%ymm1, %%ymm0, %%ymm0") /* ymm0 = [ fabsf(x1) < t ] ? ([ fabsf(x1) >= 1e-4 ] ? f : 1.0) : 0.0 */ void x64_lanczos1_fma3(float *dst, float k, float p, float t, float a, size_t count) { ARCH_X86_64_ASM( // Prepare __ASM_EMIT("vbroadcastss %[k], %%ymm8") // ymm8 = k __ASM_EMIT("vbroadcastss %[p], %%ymm9") // ymm9 = p __ASM_EMIT("vbroadcastss %[t], %%ymm10") // ymm10 = t __ASM_EMIT("vbroadcastss %[a], %%ymm11") // ymm11 = a __ASM_EMIT("vmovaps 0x00 + %[LGEN], %%ymm14") // ymm14 = i = 0 1 2 3 4 5 6 7 // x8 blocks __ASM_EMIT("sub $8, %[count]") __ASM_EMIT("jb 2f") __ASM_EMIT("vmovaps 0x60 + %[LGEN], %%ymm15") // ymm15 = step __ASM_EMIT("1:") __ASM_EMIT("vmulps %%ymm8, %%ymm14, %%ymm12") // ymm12 = k*i __ASM_EMIT("vaddps %%ymm15, %%ymm14, %%ymm14") // ymm14 = i' = i + step __ASM_EMIT("vsubps %%ymm9, %%ymm12, %%ymm12") // ymm12 = x1 = k*i - p LANCZOS_GEN_X64_FUNC_X8_FMA3 __ASM_EMIT("vmovups %%ymm0, 0x00(%[dst])") __ASM_EMIT("add $0x20, %[dst]") __ASM_EMIT("sub $8, %[count]") __ASM_EMIT("jae 1b") __ASM_EMIT("2:") // x4 block __ASM_EMIT("add $4, %[count]") __ASM_EMIT("jl 4f") __ASM_EMIT("vmulps %%xmm8, %%xmm14, %%xmm12") // xmm12 = k*i __ASM_EMIT("vaddps 0x80 + %[LGEN], %%xmm14, %%xmm14") // xmm14 = i' = i + step __ASM_EMIT("vsubps %%xmm9, %%xmm12, %%xmm12") // xmm12 = x1 = k*i - p LANCZOS_GEN_X64_FUNC_X4_FMA3 __ASM_EMIT("vmovups %%xmm0, 0x00(%[dst])") __ASM_EMIT("add $0x10, %[dst]") __ASM_EMIT("sub $4, %[count]") __ASM_EMIT("4:") // Tail: 1x-3x block __ASM_EMIT("add $4, %[count]") __ASM_EMIT("jle 8f") __ASM_EMIT("vmulps %%xmm8, %%xmm14, %%xmm12") // xmm12 = k*i __ASM_EMIT("vsubps %%xmm9, %%xmm12, %%xmm12") // xmm12 = x1 = k*i - p LANCZOS_GEN_X64_FUNC_X4_FMA3 __ASM_EMIT("test $1, %[count]") __ASM_EMIT("jz 6f") __ASM_EMIT("vmovss %%xmm0, 0x00(%[dst])") __ASM_EMIT("vshufps $0x39, %%xmm0, %%xmm0, %%xmm0") __ASM_EMIT("add $4, %[dst]") __ASM_EMIT("6:") __ASM_EMIT("test $2, %[count]") __ASM_EMIT("jz 8f") __ASM_EMIT("vmovlps %%xmm0, 0x00(%[dst])") // End __ASM_EMIT("8:") : [dst] "+r" (dst), [count] "+r" (count) : [S2C] "o" (sinf_const), [LGEN] "o" (kp_gen_const), [LC] "o" (lanczos_const), [k] "m" (k), [p] "m" (p), [a] "m" (a), [t] "m" (t) : "cc", "memory", "%xmm0", "%xmm1", "%xmm2", "%xmm3", "%xmm4", "%xmm5", "%xmm6", "%xmm7", "%xmm8", "%xmm9", "%xmm10", "%xmm11", "%xmm12", "%xmm13", "%xmm14", "%xmm15" ); } } /* namespace avx2 */ } /* namespace lsp */ #endif /* PRIVATE_DSP_ARCH_X86_AVX2_PMATH_LANCZOS_H_ */