/* * 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_ARM_NEON_D32_FFT_SCRAMBLE_H_ #define PRIVATE_DSP_ARCH_ARM_NEON_D32_FFT_SCRAMBLE_H_ #ifndef PRIVATE_DSP_ARCH_ARM_NEON_D32_IMPL #error "This header should not be included directly" #endif /* PRIVATE_DSP_ARCH_ARM_NEON_D32_IMPL */ #include namespace lsp { namespace neon_d32 { void scramble_direct(float *dst_re, float *dst_im, const float *src_re, const float *src_im, size_t rank) { if ((dst_re == src_re) || (dst_im == src_im)) { IF_ARCH_ARM( size_t count = 1 << rank; size_t i, j, rrank = rank; float *d_re, *d_im; ); // Self algorithm ARCH_ARM_ASM( // Do bit-reverse shuffle __ASM_EMIT("rsb %[rrank], %[rrank], #32") // rrank = 32 - rank __ASM_EMIT("mov %[i], #1") // i = 1 __ASM_EMIT("1:") #if defined(ARCH_ARM6) __ASM_EMIT("push {%[dst_re]}") __ASM_EMIT("push {%[dst_im]}") LSP_ARMV6_MV_RBIT32("%[j]", "%[i]", "%[dst_re]", "%[dst_im]", "%[masks]") // j = reverse_bits(i) __ASM_EMIT("pop {%[dst_im]}") __ASM_EMIT("pop {%[dst_re]}") #else __ASM_EMIT("rbit %[j], %[i]") // j = reverse_bits(i) #endif __ASM_EMIT("add %[src_re], #4") __ASM_EMIT("lsr %[j], %[rrank]") // j = reverse_bits(i) >> rank __ASM_EMIT("add %[src_im], #4") __ASM_EMIT("cmp %[i], %[j]") // i <=> j __ASM_EMIT("bhs 2f") // if (i >= j) continue __ASM_EMIT("add %[d_re], %[dst_re], %[j], LSL #2") // d_re = &dst_re[j] __ASM_EMIT("vldm %[src_re], {s0}") // s0 = *src_re __ASM_EMIT("add %[d_im], %[dst_im], %[j], LSL #2") // d_im = &dst_im[j] __ASM_EMIT("vldm %[src_im], {s1}") // s1 = *src_im __ASM_EMIT("vldm %[d_re], {s2}") // s2 = *td_re __ASM_EMIT("vldm %[d_im], {s3}") // s3 = *td_im __ASM_EMIT("vstm %[src_re], {s2}") // *(src_re++) = s2 __ASM_EMIT("vstm %[src_im], {s3}") // *(src_im++) = s3 __ASM_EMIT("vstm %[d_re], {s0}") // *d_re = s0 __ASM_EMIT("vstm %[d_im], {s1}") // *d_im = s1 __ASM_EMIT("2:") __ASM_EMIT("add %[i], #1") // i++ __ASM_EMIT("cmp %[i], %[count]") // i <=> count __ASM_EMIT("blo 1b") // Perform x8 butterflies __ASM_EMIT("3:") __ASM_EMIT("vld2.32 {q0-q1}, [%[dst_re]]") // q0 = r0 r2 r4 r6, q1 = r1 r3 r5 r7 __ASM_EMIT("vld2.32 {q2-q3}, [%[dst_im]]") // q2 = i0 i2 i4 i6, q3 = i1 i3 i5 i7 __ASM_EMIT("vadd.f32 q4, q0, q1") // q4 = r0+r1 r2+r3 r4+r5 r6+r7 = r0' r2' r4' r6' __ASM_EMIT("vadd.f32 q5, q2, q3") // q5 = i0+i1 i2+i3 i4+i5 i6+i7 = i0' i2' i4' i6' __ASM_EMIT("vsub.f32 q0, q0, q1") // q0 = r0-r1 r2-r3 r4-r5 r6-r7 = r1' r3' r5' r7' __ASM_EMIT("vsub.f32 q1, q2, q3") // q1 = i0-i1 i2-i3 i4-i5 i6-i7 = i1' i3' i5' i7' // q4 = r0' r2' r4' r6' // q0 = r1' r3' r5' r7' // q5 = i0' i2' i4' i6' // q1 = i1' i3' i5' i7' __ASM_EMIT("vuzp.32 q4, q0") // q4 = r0' r4' r1' r5', q0 = r2' r6' r3' r7' __ASM_EMIT("vuzp.32 q5, q1") // q5 = i0' i4' i1' i5', q1 = i2' i6' i3' i7' __ASM_EMIT("vswp d1, d3") // q0 = r2' r6' i3' i7', q1 = i2' i6' r3' r7' __ASM_EMIT("vadd.f32 q2, q4, q0") // q2 = r0'+r2' r4'+r6' r1'+i3' r5'+i7' = r0" r4" r1" r5" __ASM_EMIT("vsub.f32 q3, q4, q0") // q3 = r0'-r2' r4'-r6' r1'-i3' r5'-i7' = r2" r6" r3" r7" __ASM_EMIT("vadd.f32 q0, q5, q1") // q0 = i0'+i2' i4'+i6' i1'+r3' i5'+r7' = i0" i4" i3" i7" __ASM_EMIT("vsub.f32 q1, q5, q1") // q1 = i0'-i2' i4'-i6' i1'-r3' i5'-r7' = i2" i6" i1" i5" // q0 = i0" i4" i3" i7" // q1 = i2" i6" i1" i5" // q2 = r0" r4" r1" r5" // q3 = r2" r6" r3" r7" __ASM_EMIT("vswp d1, d3") // q0 = i0" i4" i1" i5", q1 = i2" i6" i3" i7" __ASM_EMIT("vuzp.32 q2, q3") // q2 = r0" r1" r2" r3", q3 = r4" r5" r6" r7" __ASM_EMIT("vuzp.32 q0, q1") // q0 = i0" i1" i2" i3", q1 = i4" i5" i6" i7" __ASM_EMIT("vst1.32 {q2-q3}, [%[dst_re]]!") __ASM_EMIT("subs %[count], #8") // i <=> count __ASM_EMIT("vst1.32 {q0-q1}, [%[dst_im]]!") __ASM_EMIT("bne 3b") : [src_re] "+r" (src_re), [src_im] "+r" (src_im), [dst_re] "+r" (dst_re), [dst_im] "+r" (dst_im), [d_re] "=&r" (d_re), [d_im] "=&r" (d_im), [rrank] "+r" (rrank), [i] "=&r" (i), [j] "=&r" (j), [count] "+r" (count) : IF_ARCH_ARM6([masks] "r" (lsp_rb_masks)) : "cc", "memory", "q0", "q1", "q2", "q3", "q4", "q5" ); } else { IF_ARCH_ARM( size_t i, j, rrank = rank - 3; size_t regs = 1 << rrank; float *s_re, *s_im; ); ARCH_ARM_ASM( __ASM_EMIT("eor %[i], %[i]") // i = 0 __ASM_EMIT("rsb %[rrank], %[rrank], #32") // rrank = 32 - rank __ASM_EMIT("1:") #if defined(ARCH_ARM6) __ASM_EMIT("push {%[dst_re]}") __ASM_EMIT("push {%[dst_im]}") LSP_ARMV6_MV_RBIT32("%[j]", "%[i]", "%[dst_re]", "%[dst_im]", "%[masks]") // j = reverse_bits(i) __ASM_EMIT("pop {%[dst_im]}") __ASM_EMIT("pop {%[dst_re]}") #else __ASM_EMIT("rbit %[j], %[i]") // j = reverse_bits(i) #endif __ASM_EMIT("lsr %[j], %[rrank]") // j = reverse_bits(i) >> rank __ASM_EMIT("add %[s_re], %[src_re], %[j], LSL #2") // s_re = &src_re[i] __ASM_EMIT("add %[s_im], %[src_im], %[j], LSL #2") // s_re = &src_im[i] __ASM_EMIT("vldm %[s_re], {s0}") // q0 = r0 ? ? ? __ASM_EMIT("vldm %[s_im], {s8}") // q2 = i0 ? ? ? __ASM_EMIT("add %[s_re], %[s_re], %[regs], LSL #2") // s_re = &src_re[i + regs] __ASM_EMIT("add %[s_im], %[s_im], %[regs], LSL #2") // s_re = &src_im[i + regs] __ASM_EMIT("vldm %[s_re], {s2}") // q0 = r0 ? r4 ? __ASM_EMIT("vldm %[s_im], {s10}") // q2 = i0 ? i4 ? __ASM_EMIT("add %[s_re], %[s_re], %[regs], LSL #2") // s_re = &src_re[i + regs*2] __ASM_EMIT("add %[s_im], %[s_im], %[regs], LSL #2") // s_re = &src_im[i + regs*2] __ASM_EMIT("vldm %[s_re], {s1}") // q0 = r0 r2 r4 ? __ASM_EMIT("vldm %[s_im], {s9}") // q2 = i0 i2 i4 ? __ASM_EMIT("add %[s_re], %[s_re], %[regs], LSL #2") // s_re = &src_re[i + regs*3] __ASM_EMIT("add %[s_im], %[s_im], %[regs], LSL #2") // s_re = &src_im[i + regs*3] __ASM_EMIT("vldm %[s_re], {s3}") // q0 = r0 r2 r4 r6 __ASM_EMIT("vldm %[s_im], {s11}") // q2 = i0 i2 i4 i6 __ASM_EMIT("add %[s_re], %[s_re], %[regs], LSL #2") // s_re = &src_re[i + regs*4] __ASM_EMIT("add %[s_im], %[s_im], %[regs], LSL #2") // s_re = &src_im[i + regs*4] __ASM_EMIT("vldm %[s_re], {s4}") // q1 = r1 ? ? ? __ASM_EMIT("vldm %[s_im], {s12}") // q3 = i1 ? ? ? __ASM_EMIT("add %[s_re], %[s_re], %[regs], LSL #2") // s_re = &src_re[i + regs*5] __ASM_EMIT("add %[s_im], %[s_im], %[regs], LSL #2") // s_re = &src_im[i + regs*5] __ASM_EMIT("vldm %[s_re], {s6}") // q1 = r1 ? r5 ? __ASM_EMIT("vldm %[s_im], {s14}") // q3 = i1 ? i5 ? __ASM_EMIT("add %[s_re], %[s_re], %[regs], LSL #2") // s_re = &src_re[i + regs*6] __ASM_EMIT("add %[s_im], %[s_im], %[regs], LSL #2") // s_re = &src_im[i + regs*6] __ASM_EMIT("vldm %[s_re], {s5}") // q1 = r1 r3 r5 ? __ASM_EMIT("vldm %[s_im], {s13}") // q3 = i1 i3 i5 ? __ASM_EMIT("add %[s_re], %[s_re], %[regs], LSL #2") // s_re = &src_re[i + regs*7] __ASM_EMIT("add %[s_im], %[s_im], %[regs], LSL #2") // s_re = &src_im[i + regs*7] __ASM_EMIT("vldm %[s_re], {s7}") // q1 = r1 r3 r5 r7 __ASM_EMIT("vldm %[s_im], {s15}") // q3 = i1 i3 i5 i7 // q0 = r0 r2 r4 r6 // q1 = r1 r3 r5 r7 // q2 = i0 i2 i4 i6 // q3 = i1 i3 i5 i7 __ASM_EMIT("add %[i], #1") // Perform x8 butterflies __ASM_EMIT("vadd.f32 q4, q0, q1") // q4 = r0+r1 r2+r3 r4+r5 r6+r7 = r0' r2' r4' r6' __ASM_EMIT("vadd.f32 q5, q2, q3") // q5 = i0+i1 i2+i3 i4+i5 i6+i7 = i0' i2' i4' i6' __ASM_EMIT("vsub.f32 q0, q0, q1") // q0 = r0-r1 r2-r3 r4-r5 r6-r7 = r1' r3' r5' r7' __ASM_EMIT("vsub.f32 q1, q2, q3") // q1 = i0-i1 i2-i3 i4-i5 i6-i7 = i1' i3' i5' i7' // q4 = r0' r2' r4' r6' // q0 = r1' r3' r5' r7' // q5 = i0' i2' i4' i6' // q1 = i1' i3' i5' i7' __ASM_EMIT("vuzp.32 q4, q0") // q4 = r0' r4' r1' r5', q0 = r2' r6' r3' r7' __ASM_EMIT("vuzp.32 q5, q1") // q5 = i0' i4' i1' i5', q1 = i2' i6' i3' i7' __ASM_EMIT("vswp d1, d3") // q0 = r2' r6' i3' i7', q1 = i2' i6' r3' r7' __ASM_EMIT("vadd.f32 q2, q4, q0") // q2 = r0'+r2' r4'+r6' r1'+i3' r5'+i7' = r0" r4" r1" r5" __ASM_EMIT("vsub.f32 q3, q4, q0") // q3 = r0'-r2' r4'-r6' r1'-i3' r5'-i7' = r2" r6" r3" r7" __ASM_EMIT("vadd.f32 q0, q5, q1") // q0 = i0'+i2' i4'+i6' i1'+r3' i5'+r7' = i0" i4" i3" i7" __ASM_EMIT("vsub.f32 q1, q5, q1") // q1 = i0'-i2' i4'-i6' i1'-r3' i5'-r7' = i2" i6" i1" i5" // q0 = i0" i4" i3" i7" // q1 = i2" i6" i1" i5" // q2 = r0" r4" r1" r5" // q3 = r2" r6" r3" r7" __ASM_EMIT("vswp d1, d3") // q0 = i0" i4" i1" i5", q1 = i2" i6" i3" i7" __ASM_EMIT("vuzp.32 q2, q3") // q2 = r0" r1" r2" r3", q3 = r4" r5" r6" r7" __ASM_EMIT("vuzp.32 q0, q1") // q0 = i0" i1" i2" i3", q1 = i4" i5" i6" i7" __ASM_EMIT("vst1.32 {q2-q3}, [%[dst_re]]!") __ASM_EMIT("cmp %[i], %[regs]") __ASM_EMIT("vst1.32 {q0-q1}, [%[dst_im]]!") __ASM_EMIT("blo 1b") : [dst_re] "+r" (dst_re), [dst_im] "+r" (dst_im), [s_re] "=&r" (s_re), [s_im] "=&r" (s_im), [rrank] "+r" (rrank), [i] "=&r" (i), [j] "=&r" (j) : [src_re] "r" (src_re), [src_im] "r" (src_im), [regs] "r" (regs) IF_ARCH_ARM6(, [masks] "r" (lsp_rb_masks)) : "cc", "memory", "q0", "q1", "q2", "q3", "q4", "q5" ); } } void scramble_reverse(float *dst_re, float *dst_im, const float *src_re, const float *src_im, size_t rank) { if ((dst_re == src_re) || (dst_im == src_im)) { IF_ARCH_ARM( size_t count = 1 << rank; size_t i, j, rrank = rank; float *d_re, *d_im; ); // Self algorithm ARCH_ARM_ASM( // Do bit-reverse shuffle __ASM_EMIT("rsb %[rrank], %[rrank], #32") // rrank = 32 - rank __ASM_EMIT("mov %[i], #1") // i = 1 __ASM_EMIT("1:") #if defined(ARCH_ARM6) __ASM_EMIT("push {%[dst_re]}") __ASM_EMIT("push {%[dst_im]}") LSP_ARMV6_MV_RBIT32("%[j]", "%[i]", "%[dst_re]", "%[dst_im]", "%[masks]") // j = reverse_bits(i) __ASM_EMIT("pop {%[dst_im]}") __ASM_EMIT("pop {%[dst_re]}") #else __ASM_EMIT("rbit %[j], %[i]") // j = reverse_bits(i) #endif __ASM_EMIT("add %[src_re], #4") __ASM_EMIT("lsr %[j], %[rrank]") // j = reverse_bits(i) >> rank __ASM_EMIT("add %[src_im], #4") __ASM_EMIT("cmp %[i], %[j]") // i <=> j __ASM_EMIT("bhs 2f") // if (i >= j) continue __ASM_EMIT("add %[d_re], %[dst_re], %[j], LSL #2") // d_re = &dst_re[j] __ASM_EMIT("vldm %[src_re], {s0}") // s0 = *src_re __ASM_EMIT("add %[d_im], %[dst_im], %[j], LSL #2") // d_im = &dst_im[j] __ASM_EMIT("vldm %[src_im], {s1}") // s1 = *src_im __ASM_EMIT("vldm %[d_re], {s2}") // s2 = *td_re __ASM_EMIT("vldm %[d_im], {s3}") // s3 = *td_im __ASM_EMIT("vstm %[src_re], {s2}") // *(src_re++) = s2 __ASM_EMIT("vstm %[src_im], {s3}") // *(src_im++) = s3 __ASM_EMIT("vstm %[d_re], {s0}") // *d_re = s0 __ASM_EMIT("vstm %[d_im], {s1}") // *d_im = s1 __ASM_EMIT("2:") __ASM_EMIT("add %[i], #1") // i++ __ASM_EMIT("cmp %[i], %[count]") // i <=> count __ASM_EMIT("blo 1b") __ASM_EMIT("eor %[i], %[i]") // Perform x8 butterflies __ASM_EMIT("3:") __ASM_EMIT("vld2.32 {q0-q1}, [%[dst_re]]") // q0 = r0 r2 r4 r6, q1 = r1 r3 r5 r7 __ASM_EMIT("vld2.32 {q2-q3}, [%[dst_im]]") // q2 = i0 i2 i4 i6, q3 = i1 i3 i5 i7 __ASM_EMIT("vadd.f32 q4, q0, q1") // q4 = r0+r1 r2+r3 r4+r5 r6+r7 = r0' r2' r4' r6' __ASM_EMIT("vadd.f32 q5, q2, q3") // q5 = i0+i1 i2+i3 i4+i5 i6+i7 = i0' i2' i4' i6' __ASM_EMIT("vsub.f32 q0, q0, q1") // q0 = r0-r1 r2-r3 r4-r5 r6-r7 = r1' r3' r5' r7' __ASM_EMIT("vsub.f32 q1, q2, q3") // q1 = i0-i1 i2-i3 i4-i5 i6-i7 = i1' i3' i5' i7' // q4 = r0' r2' r4' r6' // q0 = r1' r3' r5' r7' // q5 = i0' i2' i4' i6' // q1 = i1' i3' i5' i7' __ASM_EMIT("vuzp.32 q4, q0") // q4 = r0' r4' r1' r5', q0 = r2' r6' r3' r7' __ASM_EMIT("vuzp.32 q5, q1") // q5 = i0' i4' i1' i5', q1 = i2' i6' i3' i7' __ASM_EMIT("vswp d1, d3") // q0 = r2' r6' i3' i7', q1 = i2' i6' r3' r7' __ASM_EMIT("vadd.f32 q2, q4, q0") // q2 = r0'+r2' r4'+r6' r1'+i3' r5'+i7' = r0" r4" r3" r7" __ASM_EMIT("vsub.f32 q3, q4, q0") // q3 = r0'-r2' r4'-r6' r1'-i3' r5'-i7' = r2" r6" r1" r5" __ASM_EMIT("vadd.f32 q0, q5, q1") // q0 = i0'+i2' i4'+i6' i1'+r3' i5'+r7' = i0" i4" i1" i5" __ASM_EMIT("vsub.f32 q1, q5, q1") // q1 = i0'-i2' i4'-i6' i1'-r3' i5'-r7' = i2" i6" i3" i7" // q0 = i0" i4" i1" i5" // q1 = i2" i6" i3" i7" // q2 = r0" r4" r3" r7" // q3 = r2" r6" r1" r5" __ASM_EMIT("vswp d5, d7") // q2 = r0" r4" r1" r5", q3 = r2" r6" r3" r7" __ASM_EMIT("vuzp.32 q2, q3") // q2 = r0" r1" r2" r3", q3 = r4" r5" r6" r7" __ASM_EMIT("vuzp.32 q0, q1") // q0 = i0" i1" i2" i3", q1 = i4" i5" i6" i7" __ASM_EMIT("vst1.32 {q2-q3}, [%[dst_re]]!") __ASM_EMIT("subs %[count], #8") // i <=> count __ASM_EMIT("vst1.32 {q0-q1}, [%[dst_im]]!") __ASM_EMIT("bne 3b") : [src_re] "+r" (src_re), [src_im] "+r" (src_im), [dst_re] "+r" (dst_re), [dst_im] "+r" (dst_im), [d_re] "=&r" (d_re), [d_im] "=&r" (d_im), [rrank] "+r" (rrank), [i] "=&r" (i), [j] "=&r" (j), [count] "+r" (count) : IF_ARCH_ARM6([masks] "r" (lsp_rb_masks)) : "cc", "memory", "q0", "q1", "q2", "q3", "q4", "q5" ); } else { IF_ARCH_ARM( size_t i, j, rrank = rank - 3; size_t regs = 1 << rrank; float *s_re, *s_im; ); ARCH_ARM_ASM( __ASM_EMIT("eor %[i], %[i]") // i = 0 __ASM_EMIT("rsb %[rrank], %[rrank], #32") // rrank = 32 - rank __ASM_EMIT("1:") #if defined(ARCH_ARM6) __ASM_EMIT("push {%[dst_re]}") __ASM_EMIT("push {%[dst_im]}") LSP_ARMV6_MV_RBIT32("%[j]", "%[i]", "%[dst_re]", "%[dst_im]", "%[masks]") // j = reverse_bits(i) __ASM_EMIT("pop {%[dst_im]}") __ASM_EMIT("pop {%[dst_re]}") #else __ASM_EMIT("rbit %[j], %[i]") // j = reverse_bits(i) #endif __ASM_EMIT("lsr %[j], %[rrank]") // j = reverse_bits(i) >> rank __ASM_EMIT("add %[s_re], %[src_re], %[j], LSL #2") // s_re = &src_re[i] __ASM_EMIT("add %[s_im], %[src_im], %[j], LSL #2") // s_re = &src_im[i] __ASM_EMIT("vldm %[s_re], {s0}") // q0 = r0 ? ? ? __ASM_EMIT("vldm %[s_im], {s8}") // q2 = i0 ? ? ? __ASM_EMIT("add %[s_re], %[s_re], %[regs], LSL #2") // s_re = &src_re[i + regs] __ASM_EMIT("add %[s_im], %[s_im], %[regs], LSL #2") // s_re = &src_im[i + regs] __ASM_EMIT("vldm %[s_re], {s2}") // q0 = r0 ? r4 ? __ASM_EMIT("vldm %[s_im], {s10}") // q2 = i0 ? i4 ? __ASM_EMIT("add %[s_re], %[s_re], %[regs], LSL #2") // s_re = &src_re[i + regs*2] __ASM_EMIT("add %[s_im], %[s_im], %[regs], LSL #2") // s_re = &src_im[i + regs*2] __ASM_EMIT("vldm %[s_re], {s1}") // q0 = r0 r2 r4 ? __ASM_EMIT("vldm %[s_im], {s9}") // q2 = i0 i2 i4 ? __ASM_EMIT("add %[s_re], %[s_re], %[regs], LSL #2") // s_re = &src_re[i + regs*3] __ASM_EMIT("add %[s_im], %[s_im], %[regs], LSL #2") // s_re = &src_im[i + regs*3] __ASM_EMIT("vldm %[s_re], {s3}") // q0 = r0 r2 r4 r6 __ASM_EMIT("vldm %[s_im], {s11}") // q2 = i0 i2 i4 i6 __ASM_EMIT("add %[s_re], %[s_re], %[regs], LSL #2") // s_re = &src_re[i + regs*4] __ASM_EMIT("add %[s_im], %[s_im], %[regs], LSL #2") // s_re = &src_im[i + regs*4] __ASM_EMIT("vldm %[s_re], {s4}") // q1 = r1 ? ? ? __ASM_EMIT("vldm %[s_im], {s12}") // q3 = i1 ? ? ? __ASM_EMIT("add %[s_re], %[s_re], %[regs], LSL #2") // s_re = &src_re[i + regs*5] __ASM_EMIT("add %[s_im], %[s_im], %[regs], LSL #2") // s_re = &src_im[i + regs*5] __ASM_EMIT("vldm %[s_re], {s6}") // q1 = r1 ? r5 ? __ASM_EMIT("vldm %[s_im], {s14}") // q3 = i1 ? i5 ? __ASM_EMIT("add %[s_re], %[s_re], %[regs], LSL #2") // s_re = &src_re[i + regs*6] __ASM_EMIT("add %[s_im], %[s_im], %[regs], LSL #2") // s_re = &src_im[i + regs*6] __ASM_EMIT("vldm %[s_re], {s5}") // q1 = r1 r3 r5 ? __ASM_EMIT("vldm %[s_im], {s13}") // q3 = i1 i3 i5 ? __ASM_EMIT("add %[s_re], %[s_re], %[regs], LSL #2") // s_re = &src_re[i + regs*7] __ASM_EMIT("add %[s_im], %[s_im], %[regs], LSL #2") // s_re = &src_im[i + regs*7] __ASM_EMIT("vldm %[s_re], {s7}") // q1 = r1 r3 r5 r7 __ASM_EMIT("vldm %[s_im], {s15}") // q3 = i1 i3 i5 i7 // q0 = r0 r2 r4 r6 // q1 = r1 r3 r5 r7 // q2 = i0 i2 i4 i6 // q3 = i1 i3 i5 i7 __ASM_EMIT("add %[i], #1") // Perform x8 butterflies __ASM_EMIT("vadd.f32 q4, q0, q1") // q4 = r0+r1 r2+r3 r4+r5 r6+r7 = r0' r2' r4' r6' __ASM_EMIT("vadd.f32 q5, q2, q3") // q5 = i0+i1 i2+i3 i4+i5 i6+i7 = i0' i2' i4' i6' __ASM_EMIT("vsub.f32 q0, q0, q1") // q0 = r0-r1 r2-r3 r4-r5 r6-r7 = r1' r3' r5' r7' __ASM_EMIT("vsub.f32 q1, q2, q3") // q1 = i0-i1 i2-i3 i4-i5 i6-i7 = i1' i3' i5' i7' // q4 = r0' r2' r4' r6' // q0 = r1' r3' r5' r7' // q5 = i0' i2' i4' i6' // q1 = i1' i3' i5' i7' __ASM_EMIT("vuzp.32 q4, q0") // q4 = r0' r4' r1' r5', q0 = r2' r6' r3' r7' __ASM_EMIT("vuzp.32 q5, q1") // q5 = i0' i4' i1' i5', q1 = i2' i6' i3' i7' __ASM_EMIT("vswp d1, d3") // q0 = r2' r6' i3' i7', q1 = i2' i6' r3' r7' __ASM_EMIT("vadd.f32 q2, q4, q0") // q2 = r0'+r2' r4'+r6' r1'+i3' r5'+i7' = r0" r4" r3" r7" __ASM_EMIT("vsub.f32 q3, q4, q0") // q3 = r0'-r2' r4'-r6' r1'-i3' r5'-i7' = r2" r6" r1" r5" __ASM_EMIT("vadd.f32 q0, q5, q1") // q0 = i0'+i2' i4'+i6' i1'+r3' i5'+r7' = i0" i4" i1" i5" __ASM_EMIT("vsub.f32 q1, q5, q1") // q1 = i0'-i2' i4'-i6' i1'-r3' i5'-r7' = i2" i6" i3" i7" // q0 = i0" i4" i1" i5" // q1 = i2" i6" i3" i7" // q2 = r0" r4" r3" r7" // q3 = r2" r6" r1" r5" __ASM_EMIT("vswp d5, d7") // q2 = r0" r4" r1" r5", q3 = r2" r6" r3" r7" __ASM_EMIT("vuzp.32 q2, q3") // q2 = r0" r1" r2" r3", q3 = r4" r5" r6" r7" __ASM_EMIT("vuzp.32 q0, q1") // q0 = i0" i1" i2" i3", q1 = i4" i5" i6" i7" __ASM_EMIT("vst1.32 {q2-q3}, [%[dst_re]]!") __ASM_EMIT("cmp %[i], %[regs]") __ASM_EMIT("vst1.32 {q0-q1}, [%[dst_im]]!") __ASM_EMIT("blo 1b") : [dst_re] "+r" (dst_re), [dst_im] "+r" (dst_im), [s_re] "=&r" (s_re), [s_im] "=&r" (s_im), [rrank] "+r" (rrank), [i] "=&r" (i), [j] "=&r" (j) : [src_re] "r" (src_re), [src_im] "r" (src_im), [regs] "r" (regs) IF_ARCH_ARM6(, [masks] "r" (lsp_rb_masks)) : "cc", "memory", "q0", "q1", "q2", "q3", "q4", "q5" ); } } } } #endif /* PRIVATE_DSP_ARCH_ARM_NEON_D32_FFT_SCRAMBLE_H_ */