📄 dct64_sse.s
字号:
/* dct64_sse: MMX/SSE optimized dct64 copyright 2006-2007 by Zuxy Meng <zuxy.meng@gmail.com> / the mpg123 project - free software under the terms of the LGPL 2.1 see COPYING and AUTHORS files in distribution or http://mpg123.org initially written by the mysterious higway for MMX (apparently) then developed into SSE opt by Zuxy Meng, also building on Romain Dolbeau's AltiVec Both have agreed to distribution under LGPL 2.1 . Transformed back into standalone asm, with help of gcc -S -DHAVE_CONFIG_H -I. -march=pentium3 -O3 -Wall -pedantic -fno-strict-aliasing -DREAL_IS_FLOAT -c -o dct64_sse.{S,c} Original comment from MPlayer source follows:*//* * Discrete Cosine Tansform (DCT) for SSE * based upon code from mp3lib/dct64.c, mp3lib/dct64_altivec.c * and mp3lib/dct64_MMX.c */#include "mangle.h"#ifndef __APPLE__ .section .rodata#else .data#endif ALIGN16 /* .type nnnn, @object .size nnnn, 16 */nnnn: .long -2147483648 .long -2147483648 .long -2147483648 .long -2147483648 ALIGN16 /* .type ppnn, @object .size ppnn, 16 */ppnn: .long 0 .long 0 .long -2147483648 .long -2147483648 ALIGN16 /* .type pnpn, @object .size pnpn, 16 */pnpn: .long 0 .long -2147483648 .long 0 .long -2147483648 ALIGN4 /* .type one.4748, @object .size one.4748, 4 */one.4748: .long 1065353216 /* no .data ? */ /* .local b2.4747 */ ALIGN16 COMM(b2.4747,128,16) /* .local b1.4746 */ ALIGN16 COMM(b1.4746,128,16) .text ALIGN16,,15.globl ASM_NAME(dct64_sse) /* .type ASM_NAME(dct64_sse), @function */ASM_NAME(dct64_sse): pushl %ebp movl %esp, %ebp movl 16(%ebp), %eax pushl %ebx movl 8(%ebp), %ecx#APP/* for (i = 0; i < 0x20 / 2; i += 4) cycle 1 */ movaps ASM_NAME(costab_mmxsse), %xmm3 shufps $27, %xmm3, %xmm3 MOVUAPS (%eax), %xmm1 movaps %xmm1, %xmm4 MOVUAPS 112(%eax), %xmm2 shufps $27, %xmm4, %xmm4 movaps %xmm2, %xmm0 shufps $27, %xmm0, %xmm0 addps %xmm0, %xmm1 movaps %xmm1, b1.4746 subps %xmm2, %xmm4 mulps %xmm3, %xmm4 movaps %xmm4, b1.4746+112 #NO_APP movl 12(%ebp), %ebx#APP/* for (i = 0; i < 0x20 / 2; i += 4) cycle 2 */ movaps ASM_NAME(costab_mmxsse)+16, %xmm3 shufps $27, %xmm3, %xmm3 MOVUAPS 16(%eax), %xmm1 movaps %xmm1, %xmm4 MOVUAPS 96(%eax), %xmm2 shufps $27, %xmm4, %xmm4 movaps %xmm2, %xmm0 shufps $27, %xmm0, %xmm0 addps %xmm0, %xmm1 movaps %xmm1, b1.4746+16 subps %xmm2, %xmm4 mulps %xmm3, %xmm4 movaps %xmm4, b1.4746+96 /* for (i = 0; i < 0x20 / 2; i += 4) cycle 3 */ movaps ASM_NAME(costab_mmxsse)+32, %xmm3 shufps $27, %xmm3, %xmm3 MOVUAPS 32(%eax), %xmm1 movaps %xmm1, %xmm4 MOVUAPS 80(%eax), %xmm2 shufps $27, %xmm4, %xmm4 movaps %xmm2, %xmm0 shufps $27, %xmm0, %xmm0 addps %xmm0, %xmm1 movaps %xmm1, b1.4746+32 subps %xmm2, %xmm4 mulps %xmm3, %xmm4 movaps %xmm4, b1.4746+80 /* for (i = 0; i < 0x20 / 2; i += 4) cycle 4 */ movaps ASM_NAME(costab_mmxsse)+48, %xmm3 shufps $27, %xmm3, %xmm3 MOVUAPS 48(%eax), %xmm1 movaps %xmm1, %xmm4 MOVUAPS 64(%eax), %xmm2 shufps $27, %xmm4, %xmm4 movaps %xmm2, %xmm0 shufps $27, %xmm0, %xmm0 addps %xmm0, %xmm1 movaps %xmm1, b1.4746+48 subps %xmm2, %xmm4 mulps %xmm3, %xmm4 movaps %xmm4, b1.4746+64 movaps b1.4746, %xmm1 movaps b1.4746+16, %xmm3 movaps b1.4746+32, %xmm4 movaps b1.4746+48, %xmm6 movaps %xmm1, %xmm7 shufps $27, %xmm7, %xmm7 movaps %xmm3, %xmm5 shufps $27, %xmm5, %xmm5 movaps %xmm4, %xmm2 shufps $27, %xmm2, %xmm2 movaps %xmm6, %xmm0 shufps $27, %xmm0, %xmm0 addps %xmm0, %xmm1 movaps %xmm1, b2.4747 addps %xmm2, %xmm3 movaps %xmm3, b2.4747+16 subps %xmm4, %xmm5 movaps %xmm5, b2.4747+32 subps %xmm6, %xmm7 movaps %xmm7, b2.4747+48 movaps b1.4746+64, %xmm1 movaps b1.4746+80, %xmm3 movaps b1.4746+96, %xmm4 movaps b1.4746+112, %xmm6 movaps %xmm1, %xmm7 shufps $27, %xmm7, %xmm7 movaps %xmm3, %xmm5 shufps $27, %xmm5, %xmm5 movaps %xmm4, %xmm2 shufps $27, %xmm2, %xmm2 movaps %xmm6, %xmm0 shufps $27, %xmm0, %xmm0 addps %xmm0, %xmm1 movaps %xmm1, b2.4747+64 addps %xmm2, %xmm3 movaps %xmm3, b2.4747+80 subps %xmm4, %xmm5 movaps %xmm5, b2.4747+96 subps %xmm6, %xmm7 movaps %xmm7, b2.4747+112 movaps b2.4747+32, %xmm0 movaps b2.4747+48, %xmm1 movaps ASM_NAME(costab_mmxsse)+64, %xmm4 xorps %xmm6, %xmm6 shufps $27, %xmm4, %xmm4 mulps %xmm4, %xmm1 movaps ASM_NAME(costab_mmxsse)+80, %xmm2 xorps %xmm7, %xmm7 shufps $27, %xmm2, %xmm2 mulps %xmm2, %xmm0 movaps %xmm0, b2.4747+32 movaps %xmm1, b2.4747+48 movaps b2.4747+96, %xmm3 mulps %xmm2, %xmm3 subps %xmm3, %xmm6 movaps %xmm6, b2.4747+96 movaps b2.4747+112, %xmm5 mulps %xmm4, %xmm5 subps %xmm5, %xmm7 movaps %xmm7, b2.4747+112 movaps ASM_NAME(costab_mmxsse)+96, %xmm0 shufps $27, %xmm0, %xmm0 movaps nnnn, %xmm5 movaps %xmm5, %xmm6 movaps b2.4747, %xmm2 movaps b2.4747+16, %xmm3 movaps %xmm2, %xmm4 xorps %xmm5, %xmm6 shufps $27, %xmm4, %xmm4 movaps %xmm3, %xmm1 shufps $27, %xmm1, %xmm1 addps %xmm1, %xmm2 movaps %xmm2, b1.4746 subps %xmm3, %xmm4 xorps %xmm6, %xmm4 mulps %xmm0, %xmm4 movaps %xmm4, b1.4746+16 movaps b2.4747+32, %xmm2 movaps b2.4747+48, %xmm3 movaps %xmm2, %xmm4 xorps %xmm5, %xmm6 shufps $27, %xmm4, %xmm4 movaps %xmm3, %xmm1 shufps $27, %xmm1, %xmm1 addps %xmm1, %xmm2 movaps %xmm2, b1.4746+32 subps %xmm3, %xmm4 xorps %xmm6, %xmm4 mulps %xmm0, %xmm4 movaps %xmm4, b1.4746+48 movaps b2.4747+64, %xmm2 movaps b2.4747+80, %xmm3 movaps %xmm2, %xmm4 xorps %xmm5, %xmm6 shufps $27, %xmm4, %xmm4 movaps %xmm3, %xmm1 shufps $27, %xmm1, %xmm1 addps %xmm1, %xmm2 movaps %xmm2, b1.4746+64 subps %xmm3, %xmm4 xorps %xmm6, %xmm4 mulps %xmm0, %xmm4 movaps %xmm4, b1.4746+80 movaps b2.4747+96, %xmm2 movaps b2.4747+112, %xmm3 movaps %xmm2, %xmm4 xorps %xmm5, %xmm6 shufps $27, %xmm4, %xmm4 movaps %xmm3, %xmm1 shufps $27, %xmm1, %xmm1 addps %xmm1, %xmm2 movaps %xmm2, b1.4746+96 subps %xmm3, %xmm4 xorps %xmm6, %xmm4 mulps %xmm0, %xmm4 movaps %xmm4, b1.4746+112 movss one.4748, %xmm1 movss ASM_NAME(costab_mmxsse)+112, %xmm0 movaps %xmm1, %xmm3 unpcklps %xmm0, %xmm3 movss ASM_NAME(costab_mmxsse)+116, %xmm2 movaps %xmm1, %xmm0 unpcklps %xmm2, %xmm0 unpcklps %xmm3, %xmm0 movaps ppnn, %xmm2 movaps b1.4746, %xmm3 movaps %xmm3, %xmm4 shufps $20, %xmm4, %xmm4 shufps $235, %xmm3, %xmm3 xorps %xmm2, %xmm3 addps %xmm3, %xmm4 mulps %xmm0, %xmm4 movaps %xmm4, b2.4747 movaps b1.4746+16, %xmm6 movaps %xmm6, %xmm5 shufps $27, %xmm5, %xmm5 xorps %xmm2, %xmm5 addps %xmm5, %xmm6 mulps %xmm0, %xmm6 movaps %xmm6, b2.4747+16 movaps b1.4746+32, %xmm3 movaps %xmm3, %xmm4 shufps $20, %xmm4, %xmm4 shufps $235, %xmm3, %xmm3 xorps %xmm2, %xmm3 addps %xmm3, %xmm4 mulps %xmm0, %xmm4 movaps %xmm4, b2.4747+32 movaps b1.4746+48, %xmm6 movaps %xmm6, %xmm5 shufps $27, %xmm5, %xmm5 xorps %xmm2, %xmm5 addps %xmm5, %xmm6 mulps %xmm0, %xmm6 movaps %xmm6, b2.4747+48 movaps b1.4746+64, %xmm3 movaps %xmm3, %xmm4 shufps $20, %xmm4, %xmm4 shufps $235, %xmm3, %xmm3 xorps %xmm2, %xmm3 addps %xmm3, %xmm4 mulps %xmm0, %xmm4 movaps %xmm4, b2.4747+64 movaps b1.4746+80, %xmm6 movaps %xmm6, %xmm5 shufps $27, %xmm5, %xmm5 xorps %xmm2, %xmm5 addps %xmm5, %xmm6 mulps %xmm0, %xmm6 movaps %xmm6, b2.4747+80 movaps b1.4746+96, %xmm3 movaps %xmm3, %xmm4 shufps $20, %xmm4, %xmm4 shufps $235, %xmm3, %xmm3 xorps %xmm2, %xmm3 addps %xmm3, %xmm4 mulps %xmm0, %xmm4 movaps %xmm4, b2.4747+96 movaps b1.4746+112, %xmm6 movaps %xmm6, %xmm5 shufps $27, %xmm5, %xmm5 xorps %xmm2, %xmm5 addps %xmm5, %xmm6 mulps %xmm0, %xmm6 movaps %xmm6, b2.4747+112 movss ASM_NAME(costab_mmxsse)+120, %xmm0 movaps %xmm1, %xmm2 movaps %xmm0, %xmm7 unpcklps %xmm1, %xmm2 unpcklps %xmm0, %xmm7 movaps pnpn, %xmm0 unpcklps %xmm7, %xmm2 movaps b2.4747+32, %xmm1 movaps %xmm1, %xmm3 shufps $224, %xmm3, %xmm3 shufps $181, %xmm1, %xmm1 xorps %xmm0, %xmm1 addps %xmm1, %xmm3 mulps %xmm2, %xmm3 movaps %xmm3, b1.4746+32 movaps b2.4747+48, %xmm4 movaps %xmm4, %xmm5 shufps $224, %xmm5, %xmm5 shufps $181, %xmm4, %xmm4 xorps %xmm0, %xmm4 addps %xmm4, %xmm5 mulps %xmm2, %xmm5 movaps %xmm5, b1.4746+48 movaps b2.4747+64, %xmm1 movaps %xmm1, %xmm3 shufps $224, %xmm3, %xmm3 shufps $181, %xmm1, %xmm1 xorps %xmm0, %xmm1 addps %xmm1, %xmm3 mulps %xmm2, %xmm3 movaps %xmm3, b1.4746+64 movaps b2.4747+80, %xmm4 movaps %xmm4, %xmm5 shufps $224, %xmm5, %xmm5 shufps $181, %xmm4, %xmm4 xorps %xmm0, %xmm4 addps %xmm4, %xmm5 mulps %xmm2, %xmm5 movaps %xmm5, b1.4746+80 movaps b2.4747+96, %xmm1 movaps %xmm1, %xmm3 shufps $224, %xmm3, %xmm3 shufps $181, %xmm1, %xmm1 xorps %xmm0, %xmm1 addps %xmm1, %xmm3 mulps %xmm2, %xmm3 movaps %xmm3, b1.4746+96 movaps b2.4747+112, %xmm4 movaps %xmm4, %xmm5 shufps $224, %xmm5, %xmm5 shufps $181, %xmm4, %xmm4 xorps %xmm0, %xmm4 addps %xmm4, %xmm5 mulps %xmm2, %xmm5 movaps %xmm5, b1.4746+112 #NO_APP flds b1.4746+40 movl $b1.4746, %edx movl $b2.4747, %eax fadds b1.4746+44 fstps b1.4746+40 flds b1.4746+56 fadds b1.4746+60 flds b1.4746+48 fadd %st(1), %st fstps b1.4746+48 fadds b1.4746+52 fstps b1.4746+56 flds b1.4746+52 fadds b1.4746+60 fstps b1.4746+52 flds b1.4746+72 fadds b1.4746+76 fstps b1.4746+72 flds b1.4746+88 fadds b1.4746+92 flds b1.4746+80 fadd %st(1), %st fstps b1.4746+80 fadds b1.4746+84 fstps b1.4746+88 flds b1.4746+84 fadds b1.4746+92 fstps b1.4746+84 flds b1.4746+104 fadds b1.4746+108 fstps b1.4746+104 flds b1.4746+120 fadds b1.4746+124 flds b1.4746+112 fadd %st(1), %st fstps b1.4746+112 fadds b1.4746+116 fstps b1.4746+120 flds b1.4746+116 fadds b1.4746+124 fstps b1.4746+116#APP flds ASM_NAME(costab_mmxsse)+120 flds (%eax) fadds 4(%eax) fistp 512(%ecx) flds (%eax) fsubs 4(%eax) fmul %st(1) fistp (%ecx) flds 12(%eax) fsubs 8(%eax) fmul %st(1) fist 256(%ebx) fadds 12(%eax) fadds 8(%eax) fistp 256(%ecx) flds 16(%eax) fsubs 20(%eax) fmul %st(1) flds 28(%eax) fsubs 24(%eax) fmul %st(2) fist 384(%ebx) fld %st(0) fadds 24(%eax) fadds 28(%eax) fld %st(0) fadds 16(%eax) fadds 20(%eax) fistp 384(%ecx) fadd %st(2) fistp 128(%ecx) faddp %st(1) fistp 128(%ebx) flds 32(%edx) fadds 48(%edx) fistp 448(%ecx) flds 48(%edx) fadds 40(%edx) fistp 320(%ecx) flds 40(%edx) fadds 56(%edx) fistp 192(%ecx) flds 56(%edx) fadds 36(%edx) fistp 64(%ecx) flds 36(%edx) fadds 52(%edx) fistp 64(%ebx) flds 52(%edx) fadds 44(%edx) fistp 192(%ebx) flds 60(%edx) fist 448(%ebx) fadds 44(%edx) fistp 320(%ebx) flds 96(%edx) fadds 112(%edx) fld %st(0) fadds 64(%edx) fistp 480(%ecx) fadds 80(%edx) fistp 416(%ecx) flds 112(%edx) fadds 104(%edx) fld %st(0) fadds 80(%edx) fistp 352(%ecx) fadds 72(%edx) fistp 288(%ecx) flds 104(%edx) fadds 120(%edx) fld %st(0) fadds 72(%edx) fistp 224(%ecx) fadds 88(%edx) fistp 160(%ecx) flds 120(%edx) fadds 100(%edx) fld %st(0) fadds 88(%edx) fistp 96(%ecx) fadds 68(%edx) fistp 32(%ecx) flds 100(%edx) fadds 116(%edx) fld %st(0) fadds 68(%edx) fistp 32(%ebx) fadds 84(%edx) fistp 96(%ebx) flds 116(%edx) fadds 108(%edx) fld %st(0) fadds 84(%edx) fistp 160(%ebx) fadds 76(%edx) fistp 224(%ebx) flds 108(%edx) fadds 124(%edx) fld %st(0) fadds 76(%edx) fistp 288(%ebx) fadds 92(%edx) fistp 352(%ebx) flds 124(%edx) fist 480(%ebx) fadds 92(%edx) fistp 416(%ebx) ffreep %st(0) #NO_APP movzwl (%ecx), %eax movw %ax, (%ebx) popl %ebx popl %ebp ret /* .size ASM_NAME(dct64_sse), .-ASM_NAME(dct64_sse) */⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -