/* * Copyright (C) 2020 Linux Studio Plugins Project * (C) 2020 Vladimir Sadovnikov * * This file is part of lsp-dsp-lib * Created on: 31 мар. 2020 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_HMATH_HSUM_H_ #define PRIVATE_DSP_ARCH_AARCH64_ASIMD_HMATH_HSUM_H_ #ifndef PRIVATE_DSP_ARCH_AARCH64_ASIMD_IMPL #error "This header should not be included directly" #endif /* PRIVATE_DSP_ARCH_AARCH64_ASIMD_IMPL */ namespace lsp { namespace asimd { float h_sum(const float *src, size_t count) { IF_ARCH_AARCH64(float res); ARCH_AARCH64_ASM ( __ASM_EMIT("eor v0.16b, v0.16b, v0.16b") __ASM_EMIT("eor v1.16b, v1.16b, v1.16b") __ASM_EMIT("subs %[count], %[count], #24") __ASM_EMIT("b.lt 2f") /* 24x block */ __ASM_EMIT("1:") __ASM_EMIT("ldp q2, q3, [%[src], #0x00]") __ASM_EMIT("ldp q4, q5, [%[src], #0x20]") __ASM_EMIT("ldp q6, q7, [%[src], #0x40]") __ASM_EMIT("fadd v0.4s, v0.4s, v2.4s") __ASM_EMIT("fadd v1.4s, v1.4s, v3.4s") __ASM_EMIT("fadd v0.4s, v0.4s, v4.4s") __ASM_EMIT("fadd v1.4s, v1.4s, v5.4s") __ASM_EMIT("fadd v0.4s, v0.4s, v6.4s") __ASM_EMIT("fadd v1.4s, v1.4s, v7.4s") __ASM_EMIT("subs %[count], %[count], #24") __ASM_EMIT("add %[src], %[src], #0x60") __ASM_EMIT("b.hs 1b") /* 16x block */ __ASM_EMIT("2:") __ASM_EMIT("adds %[count], %[count], #8") __ASM_EMIT("b.lt 4f") __ASM_EMIT("ldp q2, q3, [%[src], #0x00]") __ASM_EMIT("ldp q4, q5, [%[src], #0x20]") __ASM_EMIT("fadd v0.4s, v0.4s, v2.4s") __ASM_EMIT("fadd v1.4s, v1.4s, v3.4s") __ASM_EMIT("fadd v0.4s, v0.4s, v4.4s") __ASM_EMIT("fadd v1.4s, v1.4s, v5.4s") __ASM_EMIT("sub %[count], %[count], #16") __ASM_EMIT("add %[src], %[src], #0x40") /* 8x block */ __ASM_EMIT("4:") __ASM_EMIT("adds %[count], %[count], #8") __ASM_EMIT("b.lt 6f") __ASM_EMIT("ldp q2, q3, [%[src], #0x00]") __ASM_EMIT("fadd v0.4s, v0.4s, v2.4s") __ASM_EMIT("fadd v1.4s, v1.4s, v3.4s") __ASM_EMIT("sub %[count], %[count], #8") __ASM_EMIT("add %[src], %[src], #0x20") /* 4x block */ __ASM_EMIT("6:") __ASM_EMIT("adds %[count], %[count], #4") __ASM_EMIT("fadd v0.4s, v0.4s, v1.4s") __ASM_EMIT("b.lt 8f") __ASM_EMIT("ldr q2, [%[src], #0x00]") __ASM_EMIT("fadd v0.4s, v0.4s, v2.4s") __ASM_EMIT("sub %[count], %[count], #4") __ASM_EMIT("add %[src], %[src], #0x10") /* 1x block */ __ASM_EMIT("8:") __ASM_EMIT("adds %[count], %[count], #3") __ASM_EMIT("b.lt 10f") __ASM_EMIT("eor v2.16b, v2.16b, v2.16b") __ASM_EMIT("9:") __ASM_EMIT("ld1 {v2.s}[0], [%[src]]") __ASM_EMIT("fadd v0.4s, v0.4s, v2.4s") __ASM_EMIT("subs %[count], %[count], #1") __ASM_EMIT("add %[src], %[src], #0x04") __ASM_EMIT("b.ge 9b") /* end of sum */ __ASM_EMIT("10:") __ASM_EMIT("ext v1.16b, v0.16b, v0.16b, #8") /* v0 = a0 a1 a2 a3, v1 = a2 a3 a0 a1 */ __ASM_EMIT("fadd v0.4s, v0.4s, v1.4s") /* v0 = a0+a2 a1+a3 a0+a2 a1+a3 */ __ASM_EMIT("ext v1.16b, v0.16b, v0.16b, #4") /* v1 = a1+a3 a0+a2 a1+a3 a0+a2 */ __ASM_EMIT("fadd %[res].4s, v0.4s, v1.4s") /* v0 = a0+a1+a2+a3 */ : [res] "=w" (res), [src] "+r" (src), [count] "+r" (count) : : "cc", "memory", "v1", "v2", "v3", "v4", "v5", "v6", "v7" ); return res; } float h_sqr_sum(const float *src, size_t count) { IF_ARCH_AARCH64(float res); ARCH_AARCH64_ASM ( __ASM_EMIT("eor v0.16b, v0.16b, v0.16b") __ASM_EMIT("eor v1.16b, v1.16b, v1.16b") __ASM_EMIT("subs %[count], %[count], #24") __ASM_EMIT("b.lt 2f") /* 24x block */ __ASM_EMIT("1:") __ASM_EMIT("ldp q2, q3, [%[src], #0x00]") __ASM_EMIT("ldp q4, q5, [%[src], #0x20]") __ASM_EMIT("ldp q6, q7, [%[src], #0x40]") __ASM_EMIT("fmla v0.4s, v2.4s, v2.4s") __ASM_EMIT("fmla v1.4s, v3.4s, v3.4s") __ASM_EMIT("fmla v0.4s, v4.4s, v4.4s") __ASM_EMIT("fmla v1.4s, v5.4s, v5.4s") __ASM_EMIT("fmla v0.4s, v6.4s, v6.4s") __ASM_EMIT("fmla v1.4s, v7.4s, v7.4s") __ASM_EMIT("subs %[count], %[count], #24") __ASM_EMIT("add %[src], %[src], #0x60") __ASM_EMIT("b.hs 1b") /* 16x block */ __ASM_EMIT("2:") __ASM_EMIT("adds %[count], %[count], #8") __ASM_EMIT("b.lt 4f") __ASM_EMIT("ldp q2, q3, [%[src], #0x00]") __ASM_EMIT("ldp q4, q5, [%[src], #0x20]") __ASM_EMIT("fmla v0.4s, v2.4s, v2.4s") __ASM_EMIT("fmla v1.4s, v3.4s, v3.4s") __ASM_EMIT("fmla v0.4s, v4.4s, v4.4s") __ASM_EMIT("fmla v1.4s, v5.4s, v5.4s") __ASM_EMIT("sub %[count], %[count], #16") __ASM_EMIT("add %[src], %[src], #0x40") /* 8x block */ __ASM_EMIT("4:") __ASM_EMIT("adds %[count], %[count], #8") __ASM_EMIT("b.lt 6f") __ASM_EMIT("ldp q2, q3, [%[src], #0x00]") __ASM_EMIT("fmla v0.4s, v2.4s, v2.4s") __ASM_EMIT("fmla v1.4s, v3.4s, v3.4s") __ASM_EMIT("sub %[count], %[count], #8") __ASM_EMIT("add %[src], %[src], #0x20") /* 4x block */ __ASM_EMIT("6:") __ASM_EMIT("adds %[count], %[count], #4") __ASM_EMIT("fadd v0.4s, v0.4s, v1.4s") __ASM_EMIT("b.lt 8f") __ASM_EMIT("ldr q2, [%[src], #0x00]") __ASM_EMIT("fmla v0.4s, v2.4s, v2.4s") __ASM_EMIT("sub %[count], %[count], #4") __ASM_EMIT("add %[src], %[src], #0x10") /* 1x block */ __ASM_EMIT("8:") __ASM_EMIT("adds %[count], %[count], #3") __ASM_EMIT("b.lt 10f") __ASM_EMIT("eor v2.16b, v2.16b, v2.16b") __ASM_EMIT("9:") __ASM_EMIT("ld1 {v2.s}[0], [%[src]]") __ASM_EMIT("fmla v0.4s, v2.4s, v2.4s") __ASM_EMIT("subs %[count], %[count], #1") __ASM_EMIT("add %[src], %[src], #0x04") __ASM_EMIT("b.ge 9b") /* end of sum */ __ASM_EMIT("10:") __ASM_EMIT("ext v1.16b, v0.16b, v0.16b, #8") /* v0 = a0 a1 a2 a3, v1 = a2 a3 a0 a1 */ __ASM_EMIT("fadd v0.4s, v0.4s, v1.4s") /* v0 = a0+a2 a1+a3 a0+a2 a1+a3 */ __ASM_EMIT("ext v1.16b, v0.16b, v0.16b, #4") /* v1 = a1+a3 a0+a2 a1+a3 a0+a2 */ __ASM_EMIT("fadd %[res].4s, v0.4s, v1.4s") /* v0 = a0+a1+a2+a3 */ : [res] "=w" (res), [src] "+r" (src), [count] "+r" (count) : : "cc", "memory", "v1", "v2", "v3", "v4", "v5", "v6", "v7" ); return res; } float h_abs_sum(const float *src, size_t count) { IF_ARCH_AARCH64(float res); ARCH_AARCH64_ASM ( __ASM_EMIT("eor v0.16b, v0.16b, v0.16b") __ASM_EMIT("eor v1.16b, v1.16b, v1.16b") __ASM_EMIT("subs %[count], %[count], #24") __ASM_EMIT("b.lt 2f") /* 24x block */ __ASM_EMIT("1:") __ASM_EMIT("ldp q2, q3, [%[src], #0x00]") __ASM_EMIT("ldp q4, q5, [%[src], #0x20]") __ASM_EMIT("ldp q6, q7, [%[src], #0x40]") __ASM_EMIT("fabs v2.4s, v2.4s") __ASM_EMIT("fabs v3.4s, v3.4s") __ASM_EMIT("fabs v4.4s, v4.4s") __ASM_EMIT("fabs v5.4s, v5.4s") __ASM_EMIT("fabs v6.4s, v6.4s") __ASM_EMIT("fabs v7.4s, v7.4s") __ASM_EMIT("fadd v0.4s, v0.4s, v2.4s") __ASM_EMIT("fadd v1.4s, v1.4s, v3.4s") __ASM_EMIT("fadd v0.4s, v0.4s, v4.4s") __ASM_EMIT("fadd v1.4s, v1.4s, v5.4s") __ASM_EMIT("fadd v0.4s, v0.4s, v6.4s") __ASM_EMIT("fadd v1.4s, v1.4s, v7.4s") __ASM_EMIT("subs %[count], %[count], #24") __ASM_EMIT("add %[src], %[src], #0x60") __ASM_EMIT("b.hs 1b") /* 16x block */ __ASM_EMIT("2:") __ASM_EMIT("adds %[count], %[count], #8") __ASM_EMIT("b.lt 4f") __ASM_EMIT("ldp q2, q3, [%[src], #0x00]") __ASM_EMIT("ldp q4, q5, [%[src], #0x20]") __ASM_EMIT("fabs v2.4s, v2.4s") __ASM_EMIT("fabs v3.4s, v3.4s") __ASM_EMIT("fabs v4.4s, v4.4s") __ASM_EMIT("fabs v5.4s, v5.4s") __ASM_EMIT("fadd v0.4s, v0.4s, v2.4s") __ASM_EMIT("fadd v1.4s, v1.4s, v3.4s") __ASM_EMIT("fadd v0.4s, v0.4s, v4.4s") __ASM_EMIT("fadd v1.4s, v1.4s, v5.4s") __ASM_EMIT("sub %[count], %[count], #16") __ASM_EMIT("add %[src], %[src], #0x40") /* 8x block */ __ASM_EMIT("4:") __ASM_EMIT("adds %[count], %[count], #8") __ASM_EMIT("b.lt 6f") __ASM_EMIT("ldp q2, q3, [%[src], #0x00]") __ASM_EMIT("fabs v2.4s, v2.4s") __ASM_EMIT("fabs v3.4s, v3.4s") __ASM_EMIT("fadd v0.4s, v0.4s, v2.4s") __ASM_EMIT("fadd v1.4s, v1.4s, v3.4s") __ASM_EMIT("sub %[count], %[count], #8") __ASM_EMIT("add %[src], %[src], #0x20") /* 4x block */ __ASM_EMIT("6:") __ASM_EMIT("adds %[count], %[count], #4") __ASM_EMIT("fadd v0.4s, v0.4s, v1.4s") __ASM_EMIT("b.lt 8f") __ASM_EMIT("ldr q2, [%[src], #0x00]") __ASM_EMIT("fabs v2.4s, v2.4s") __ASM_EMIT("fadd v0.4s, v0.4s, v2.4s") __ASM_EMIT("sub %[count], %[count], #4") __ASM_EMIT("add %[src], %[src], #0x10") /* 1x block */ __ASM_EMIT("8:") __ASM_EMIT("adds %[count], %[count], #3") __ASM_EMIT("b.lt 10f") __ASM_EMIT("eor v2.16b, v2.16b, v2.16b") __ASM_EMIT("9:") __ASM_EMIT("ld1 {v2.s}[0], [%[src]]") __ASM_EMIT("fabs v2.4s, v2.4s") __ASM_EMIT("fadd v0.4s, v0.4s, v2.4s") __ASM_EMIT("subs %[count], %[count], #1") __ASM_EMIT("add %[src], %[src], #0x04") __ASM_EMIT("b.ge 9b") /* end of sum */ __ASM_EMIT("10:") __ASM_EMIT("ext v1.16b, v0.16b, v0.16b, #8") /* v0 = a0 a1 a2 a3, v1 = a2 a3 a0 a1 */ __ASM_EMIT("fadd v0.4s, v0.4s, v1.4s") /* v0 = a0+a2 a1+a3 a0+a2 a1+a3 */ __ASM_EMIT("ext v1.16b, v0.16b, v0.16b, #4") /* v1 = a1+a3 a0+a2 a1+a3 a0+a2 */ __ASM_EMIT("fadd %[res].4s, v0.4s, v1.4s") /* v0 = a0+a1+a2+a3 */ : [res] "=w" (res), [src] "+r" (src), [count] "+r" (count) : : "cc", "memory", "v1", "v2", "v3", "v4", "v5", "v6", "v7" ); return res; } } } #endif /* PRIVATE_DSP_ARCH_AARCH64_ASIMD_HMATH_HSUM_H_ */