/*
* Copyright (C) 2025 Linux Studio Plugins Project
* (C) 2025 Vladimir Sadovnikov
*
* This file is part of lsp-dsp-lib
* Created on: 10 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_SIN_H_
#define PRIVATE_DSP_ARCH_X86_SSE2_PMATH_SIN_H_
#ifndef PRIVATE_DSP_ARCH_X86_SSE2_IMPL
#error "This header should not be included directly"
#endif /* PRIVATE_DSP_ARCH_X86_SSE2_IMPL */
namespace lsp
{
namespace sse2
{
/*
Time-optimized sinf implementation:
float sinf(float x)
{
// Limit x to [-pi/2 .. pi/2]
x += PI/2;
int32_t period = (x / 2*PI) - (x < 0.0f);
x = x - period * 2*PI;
x = (x < PI) ? x - PI/2 : 3*PI/2 - x;
const float x2 = x*x;
return x * (1.0f + x2*(-0.166666666667f + x2*(0.00833333333333f + x2 * (-0.000198412698413f + x2 * (2.7557319224e-06f + x2 * -2.50521083854e-08f)))));
}
*/
IF_ARCH_X86(
static const float kp_gen_const[] __lsp_aligned16 =
{
0.0f, 1.0f, 2.0f, 3.0f, // +0x00: Initial values 0..3
LSP_DSP_VEC4(4.0f) // +0x10: Step
};
)
IF_ARCH_X86(
static const uint32_t sinf_const[] __lsp_aligned16 =
{
LSP_DSP_VEC4(0x3fc90fdb), // +0x00: PI/2
LSP_DSP_VEC4(0x40490fdb), // +0x10: PI
LSP_DSP_VEC4(0x3e22f983), // +0x20: 1/(2*PI)
LSP_DSP_VEC4(0x40c90fdb), // +0x30: 2*PI
LSP_DSP_VEC4(0x4096cbe4), // +0x40: 3*PI/2
LSP_DSP_VEC4(0xb2d7322b), // +0x50: C0 = -1/11! = -2.50521083854e-08
LSP_DSP_VEC4(0x3638ef1d), // +0x60: C1 = 1/9! = 2.7557319224e-06
LSP_DSP_VEC4(0xb9500d01), // +0x70: C2 = -1/7! = -0.000198412698413
LSP_DSP_VEC4(0x3c088889), // +0x80: C3 = 1/5! = 0.00833333333333
LSP_DSP_VEC4(0xbe2aaaab), // +0x90: C4 = -1/3! = -0.166666666667
LSP_DSP_VEC4(0x3f800000), // +0xa0: 1.0
};
)
#define SINF_X_PLUS_PI_2_CORE_X8 \
/* xmm0 = X1 = x1 + PI/2 */ \
/* xmm4 = X2 = x2 + PI/2 */ \
__ASM_EMIT("movaps %%xmm0, %%xmm1") /* xmm1 = X */ \
__ASM_EMIT("movaps %%xmm4, %%xmm5") \
__ASM_EMIT("movaps %%xmm0, %%xmm3") /* xmm3 = X */ \
__ASM_EMIT("movaps %%xmm4, %%xmm7") \
__ASM_EMIT("mulps 0x20 + %[S2C], %%xmm1") /* xmm1 = X / (2*PI) */ \
__ASM_EMIT("mulps 0x20 + %[S2C], %%xmm5") \
__ASM_EMIT("psrad $31, %%xmm3") /* xmm3 = [ X < 0.0 ] ? -1 : 0 */ \
__ASM_EMIT("psrad $31, %%xmm7") \
__ASM_EMIT("cvttps2dq %%xmm1, %%xmm1") /* xmm1 = int(X / (2*PI)) */ \
__ASM_EMIT("cvttps2dq %%xmm5, %%xmm5") \
__ASM_EMIT("paddd %%xmm3, %%xmm1") /* xmm1 = int(X / (2*PI)) + [ X < 0.0 ] ? -1 : 0 */ \
__ASM_EMIT("paddd %%xmm7, %%xmm5") \
__ASM_EMIT("cvtdq2ps %%xmm1, %%xmm1") /* xmm1 = period = int(X / (2*PI)) [ X < 0.0 ] ? -1 : 0 */ \
__ASM_EMIT("cvtdq2ps %%xmm5, %%xmm5") \
__ASM_EMIT("mulps 0x30 + %[S2C], %%xmm1") /* xmm1 = period * 2 * PI */ \
__ASM_EMIT("mulps 0x30 + %[S2C], %%xmm5") \
__ASM_EMIT("subps %%xmm1, %%xmm0") /* xmm0 = Y = X - period * 2 * PI */ \
__ASM_EMIT("subps %%xmm5, %%xmm4") \
__ASM_EMIT("movaps %%xmm0, %%xmm2") /* xmm2 = Y */ \
__ASM_EMIT("movaps %%xmm4, %%xmm6") \
__ASM_EMIT("movaps 0x40 + %[S2C], %%xmm1") /* xmm1 = 3*PI/2 */ \
__ASM_EMIT("movaps 0x40 + %[S2C], %%xmm5") \
__ASM_EMIT("cmpps $1, 0x10 + %[S2C], %%xmm2") /* xmm2 = [ Y < PI ] */ \
__ASM_EMIT("cmpps $1, 0x10 + %[S2C], %%xmm6") \
__ASM_EMIT("subps %%xmm0, %%xmm1") /* xmm1 = 3*PI/2 - Y */ \
__ASM_EMIT("subps %%xmm4, %%xmm5") \
__ASM_EMIT("subps 0x00 + %[S2C], %%xmm0") /* xmm0 = Y - PI/2 */ \
__ASM_EMIT("subps 0x00 + %[S2C], %%xmm4") \
__ASM_EMIT("andps %%xmm2, %%xmm0") /* xmm0 = (Y - PI/2) & [ Y < PI ] */ \
__ASM_EMIT("andps %%xmm6, %%xmm4") \
__ASM_EMIT("andnps %%xmm1, %%xmm2") /* xmm2 = (3*PI/2 - Y) & [ Y >= PI ] */ \
__ASM_EMIT("andnps %%xmm5, %%xmm6") \
__ASM_EMIT("orps %%xmm2, %%xmm0") /* xmm0 = XX = [ Y < PI ] ? (Y - PI/2) : (3*PI/2 - Y) */ \
__ASM_EMIT("orps %%xmm6, %%xmm4") \
/* xmm0 = XX1 */ \
__ASM_EMIT("movaps %%xmm0, %%xmm1") /* xmm1 = XX */ \
__ASM_EMIT("movaps %%xmm4, %%xmm5") \
__ASM_EMIT("mulps %%xmm0, %%xmm0") /* xmm0 = X2 = XX*XX */ \
__ASM_EMIT("mulps %%xmm4, %%xmm4") \
__ASM_EMIT("movaps %%xmm0, %%xmm2") /* xmm2 = X2 */ \
__ASM_EMIT("movaps %%xmm4, %%xmm6") \
__ASM_EMIT("mulps 0x50 + %[S2C], %%xmm0") /* xmm0 = X2*C0 */ \
__ASM_EMIT("mulps 0x50 + %[S2C], %%xmm4") \
__ASM_EMIT("addps 0x60 + %[S2C], %%xmm0") /* xmm0 = C1 + X2*C0 */ \
__ASM_EMIT("addps 0x60 + %[S2C], %%xmm4") \
__ASM_EMIT("mulps %%xmm2, %%xmm0") /* xmm0 = X2*(C1 + X2*C0) */ \
__ASM_EMIT("mulps %%xmm6, %%xmm4") \
__ASM_EMIT("addps 0x70 + %[S2C], %%xmm0") /* xmm0 = C2 + X2*(C1 + X2*C0) */ \
__ASM_EMIT("addps 0x70 + %[S2C], %%xmm4") \
__ASM_EMIT("mulps %%xmm2, %%xmm0") /* xmm0 = X2*(C2 + X2*(C1 + X2*C0)) */ \
__ASM_EMIT("mulps %%xmm6, %%xmm4") \
__ASM_EMIT("addps 0x80 + %[S2C], %%xmm0") /* xmm0 = C3 + X2*(C2 + X2*(C1 + X2*C0)) */ \
__ASM_EMIT("addps 0x80 + %[S2C], %%xmm4") \
__ASM_EMIT("mulps %%xmm2, %%xmm0") /* xmm0 = X2*(C3 + X2*(C2 + X2*(C1 + X2*C0))) */ \
__ASM_EMIT("mulps %%xmm6, %%xmm4") \
__ASM_EMIT("addps 0x90 + %[S2C], %%xmm0") /* xmm0 = C4 + X2*(C3 + X2*(C2 + X2*(C1 + X2*C0))) */ \
__ASM_EMIT("addps 0x90 + %[S2C], %%xmm4") \
__ASM_EMIT("mulps %%xmm2, %%xmm0") /* xmm0 = X2*(C4 + X2*(C3 + X2*(C2 + X2*(C1 + X2*C0)))) */ \
__ASM_EMIT("mulps %%xmm6, %%xmm4") \
__ASM_EMIT("addps 0xa0 + %[S2C], %%xmm0") /* xmm0 = 1.0 + X2*(C4 + X2*(C3 + X2*(C2 + X2*(C1 + X2*C0)))) */ \
__ASM_EMIT("addps 0xa0 + %[S2C], %%xmm4") \
__ASM_EMIT("mulps %%xmm1, %%xmm0") /* xmm0 = sinf(x) = XX * (1.0 + X2*(C4 + X2*(C3 + X2*(C2 + X2*(C1 + X2*C0))))) */ \
__ASM_EMIT("mulps %%xmm5, %%xmm4")
#define SINF_X_PLUS_PI_2_CORE_X4 \
/* xmm0 = X = x + PI/2 */ \
__ASM_EMIT("movaps %%xmm0, %%xmm1") /* xmm1 = X */ \
__ASM_EMIT("movaps %%xmm0, %%xmm3") /* xmm3 = X */ \
__ASM_EMIT("mulps 0x20 + %[S2C], %%xmm1") /* xmm1 = X / (2*PI) */ \
__ASM_EMIT("psrad $31, %%xmm3") /* xmm3 = [ X < 0.0 ] ? -1 : 0 */ \
__ASM_EMIT("cvttps2dq %%xmm1, %%xmm1") /* xmm1 = int(X / (2*PI)) */ \
__ASM_EMIT("paddd %%xmm3, %%xmm1") /* xmm1 = int(X / (2*PI)) + [ X < 0.0 ] ? -1 : 0 */ \
__ASM_EMIT("cvtdq2ps %%xmm1, %%xmm1") /* xmm1 = period = int(X / (2*PI)) [ X < 0.0 ] ? -1 : 0 */ \
__ASM_EMIT("mulps 0x30 + %[S2C], %%xmm1") /* xmm1 = period * 2 * PI */ \
__ASM_EMIT("subps %%xmm1, %%xmm0") /* xmm0 = Y = X - period * 2 * PI */ \
__ASM_EMIT("movaps %%xmm0, %%xmm2") /* xmm2 = Y */ \
__ASM_EMIT("movaps 0x40 + %[S2C], %%xmm1") /* xmm1 = 3*PI/2 */ \
__ASM_EMIT("cmpps $1, 0x10 + %[S2C], %%xmm2") /* xmm2 = [ Y < PI ] */ \
__ASM_EMIT("subps %%xmm0, %%xmm1") /* xmm1 = 3*PI/2 - Y */ \
__ASM_EMIT("subps 0x00 + %[S2C], %%xmm0") /* xmm0 = Y - PI/2 */ \
__ASM_EMIT("andps %%xmm2, %%xmm0") /* xmm0 = (Y - PI/2) & [ Y < PI ] */ \
__ASM_EMIT("andnps %%xmm1, %%xmm2") /* xmm2 = (3*PI/2 - Y) & [ Y >= PI ] */ \
__ASM_EMIT("orps %%xmm2, %%xmm0") /* xmm0 = XX = [ Y < PI ] ? (Y - PI/2) : (3*PI/2 - Y) */ \
/* xmm0 = XX1 */ \
__ASM_EMIT("movaps %%xmm0, %%xmm1") /* xmm1 = XX */ \
__ASM_EMIT("mulps %%xmm0, %%xmm0") /* xmm0 = X2 = XX*XX */ \
__ASM_EMIT("movaps %%xmm0, %%xmm2") /* xmm2 = X2 */ \
__ASM_EMIT("mulps 0x50 + %[S2C], %%xmm0") /* xmm0 = X2*C0 */ \
__ASM_EMIT("addps 0x60 + %[S2C], %%xmm0") /* xmm0 = C1 + X2*C0 */ \
__ASM_EMIT("mulps %%xmm2, %%xmm0") /* xmm0 = X2*(C1 + X2*C0) */ \
__ASM_EMIT("addps 0x70 + %[S2C], %%xmm0") /* xmm0 = C2 + X2*(C1 + X2*C0) */ \
__ASM_EMIT("mulps %%xmm2, %%xmm0") /* xmm0 = X2*(C2 + X2*(C1 + X2*C0)) */ \
__ASM_EMIT("addps 0x80 + %[S2C], %%xmm0") /* xmm0 = C3 + X2*(C2 + X2*(C1 + X2*C0)) */ \
__ASM_EMIT("mulps %%xmm2, %%xmm0") /* xmm0 = X2*(C3 + X2*(C2 + X2*(C1 + X2*C0))) */ \
__ASM_EMIT("addps 0x90 + %[S2C], %%xmm0") /* xmm0 = C4 + X2*(C3 + X2*(C2 + X2*(C1 + X2*C0))) */ \
__ASM_EMIT("mulps %%xmm2, %%xmm0") /* xmm0 = X2*(C4 + X2*(C3 + X2*(C2 + X2*(C1 + X2*C0)))) */ \
__ASM_EMIT("addps 0xa0 + %[S2C], %%xmm0") /* xmm0 = 1.0 + X2*(C4 + X2*(C3 + X2*(C2 + X2*(C1 + X2*C0)))) */ \
__ASM_EMIT("mulps %%xmm1, %%xmm0") /* xmm0 = sinf(x) = XX * (1.0 + X2*(C4 + X2*(C3 + X2*(C2 + X2*(C1 + X2*C0))))) */
void sinf1(float *dst, size_t count)
{
IF_ARCH_X86(float *src);
ARCH_X86_ASM(
// x8 blocks
__ASM_EMIT("sub $8, %[count]")
__ASM_EMIT("jb 2f")
__ASM_EMIT("1:")
__ASM_EMIT("movaps 0x00 + %[S2C], %%xmm1") // xmm1 = PI/2
__ASM_EMIT("movups 0x00(%[dst]), %%xmm0") // xmm0 = x1
__ASM_EMIT("movups 0x10(%[dst]), %%xmm4") // xmm4 = x2
__ASM_EMIT("addps %%xmm1, %%xmm0") // xmm0 = x1 + PI/2
__ASM_EMIT("addps %%xmm1, %%xmm4") // xmm4 = x2 + PI/2
SINF_X_PLUS_PI_2_CORE_X8
__ASM_EMIT("movups %%xmm0, 0x00(%[dst])")
__ASM_EMIT("movups %%xmm4, 0x10(%[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("movups 0x00(%[dst]), %%xmm0") // xmm0 = x
__ASM_EMIT("addps 0x00 + %[S2C], %%xmm0") // xmm0 = x + PI/2
SINF_X_PLUS_PI_2_CORE_X4
__ASM_EMIT("movups %%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 12f")
__ASM_EMIT("mov %[dst], %[src]")
__ASM_EMIT("test $1, %[count]")
__ASM_EMIT("jz 6f")
__ASM_EMIT("movss 0x00(%[src]), %%xmm0")
__ASM_EMIT("add $4, %[src]")
__ASM_EMIT("6:")
__ASM_EMIT("test $2, %[count]")
__ASM_EMIT("jz 8f")
__ASM_EMIT("movhps 0x00(%[src]), %%xmm0")
__ASM_EMIT("8:")
__ASM_EMIT("addps 0x00 + %[S2C], %%xmm0") // xmm0 = x + PI/2
SINF_X_PLUS_PI_2_CORE_X4
__ASM_EMIT("test $1, %[count]")
__ASM_EMIT("jz 10f")
__ASM_EMIT("movss %%xmm0, 0x00(%[dst])")
__ASM_EMIT("add $4, %[dst]")
__ASM_EMIT("10:")
__ASM_EMIT("test $2, %[count]")
__ASM_EMIT("jz 12f")
__ASM_EMIT("movhps %%xmm0, 0x00(%[dst])")
// End
__ASM_EMIT("12:")
: [dst] "+r" (dst), [src] "=&r" (src), [count] "+r" (count)
: [S2C] "o" (sinf_const)
: "cc", "memory",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7"
);
}
void sinf2(float *dst, const float *src, size_t count)
{
ARCH_X86_ASM(
// x8 blocks
__ASM_EMIT("sub $8, %[count]")
__ASM_EMIT("jb 2f")
__ASM_EMIT("1:")
__ASM_EMIT("movaps 0x00 + %[S2C], %%xmm1") // xmm1 = PI/2
__ASM_EMIT("movups 0x00(%[src]), %%xmm0") // xmm0 = x1
__ASM_EMIT("movups 0x10(%[src]), %%xmm4") // xmm4 = x2
__ASM_EMIT("addps %%xmm1, %%xmm0") // xmm0 = x1 + PI/2
__ASM_EMIT("addps %%xmm1, %%xmm4") // xmm4 = x2 + PI/2
SINF_X_PLUS_PI_2_CORE_X8
__ASM_EMIT("movups %%xmm0, 0x00(%[dst])")
__ASM_EMIT("movups %%xmm4, 0x10(%[dst])")
__ASM_EMIT("add $0x20, %[src]")
__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("movups 0x00(%[src]), %%xmm0") // xmm0 = x
__ASM_EMIT("addps 0x00 + %[S2C], %%xmm0") // xmm0 = x + PI/2
SINF_X_PLUS_PI_2_CORE_X4
__ASM_EMIT("movups %%xmm0, 0x00(%[dst])")
__ASM_EMIT("add $0x10, %[src]")
__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 12f")
__ASM_EMIT("test $1, %[count]")
__ASM_EMIT("jz 6f")
__ASM_EMIT("movss 0x00(%[src]), %%xmm0")
__ASM_EMIT("add $4, %[src]")
__ASM_EMIT("6:")
__ASM_EMIT("test $2, %[count]")
__ASM_EMIT("jz 8f")
__ASM_EMIT("movhps 0x00(%[src]), %%xmm0")
__ASM_EMIT("8:")
__ASM_EMIT("addps 0x00 + %[S2C], %%xmm0") // xmm0 = x + PI/2
SINF_X_PLUS_PI_2_CORE_X4
__ASM_EMIT("test $1, %[count]")
__ASM_EMIT("jz 10f")
__ASM_EMIT("movss %%xmm0, 0x00(%[dst])")
__ASM_EMIT("add $4, %[dst]")
__ASM_EMIT("10:")
__ASM_EMIT("test $2, %[count]")
__ASM_EMIT("jz 12f")
__ASM_EMIT("movhps %%xmm0, 0x00(%[dst])")
// End
__ASM_EMIT("12:")
: [dst] "+r" (dst), [src] "+r" (src), [count] "+r" (count)
: [S2C] "o" (sinf_const)
: "cc", "memory",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7"
);
}
void sinf_kp1(float *dst, float k, float p, size_t count)
{
ARCH_X86_ASM(
// Prepare
__ASM_EMIT("movss %[k], %%xmm6") // xmm6 = k
__ASM_EMIT("movss %[p], %%xmm7") // xmm7 = p
__ASM_EMIT("shufps $0x00, %%xmm6, %%xmm6") // xmm6 = k k k k
__ASM_EMIT("shufps $0x00, %%xmm7, %%xmm7") // xmm7 = p p p p
__ASM_EMIT("movaps 0x00 + %[S2KP], %%xmm4") // xmm4 = i = 0 1 2 3
__ASM_EMIT("movaps 0x10 + %[S2KP], %%xmm5") // xmm5 = step = 4 4 4 4
__ASM_EMIT("addps 0x00 + %[S2C], %%xmm7") // xmm7 = p+PI/2 p+PI/2 p+PI/2 p+PI/2
// x4 blocks
__ASM_EMIT("sub $4, %[count]")
__ASM_EMIT("jb 2f")
__ASM_EMIT("1:")
__ASM_EMIT("movaps %%xmm4, %%xmm0") // xmm0 = i
__ASM_EMIT("mulps %%xmm6, %%xmm0") // xmm0 = k*i
__ASM_EMIT("addps %%xmm5, %%xmm4") // xmm4 = i + step
__ASM_EMIT("addps %%xmm7, %%xmm0") // xmm0 = k*i + p + PI/2
SINF_X_PLUS_PI_2_CORE_X4
__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 %%xmm4, %%xmm0") // xmm0 = i
__ASM_EMIT("mulps %%xmm6, %%xmm0") // xmm0 = k*i
__ASM_EMIT("addps %%xmm7, %%xmm0") // xmm0 = k*i + p + PI/2
SINF_X_PLUS_PI_2_CORE_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)
: [S2C] "o" (sinf_const),
[S2KP] "o" (kp_gen_const),
[k] "m" (k),
[p] "m" (p)
: "cc", "memory",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7"
);
}
} /* namespace sse2 */
} /* namespace lsp */
#endif /* PRIVATE_DSP_ARCH_X86_SSE2_PMATH_SIN_H_ */