From 6f67d9833b5a7d78d27c283d7e4238e6d3e5fd98 Mon Sep 17 00:00:00 2001 From: Justin Ruggles Date: Sun, 13 May 2012 19:09:01 +0200 Subject: libswresample: Implement MMX, SSE4 and AVX 6ch float and int32 packing function. Signed-off-by: Michael Niedermayer --- libswresample/x86/audio_convert.asm | 78 +++++++++++++++++++++++++++++++++++++ 1 file changed, 78 insertions(+) (limited to 'libswresample/x86/audio_convert.asm') diff --git a/libswresample/x86/audio_convert.asm b/libswresample/x86/audio_convert.asm index 5768405733..1fa519aa84 100644 --- a/libswresample/x86/audio_convert.asm +++ b/libswresample/x86/audio_convert.asm @@ -199,6 +199,84 @@ cglobal %2_to_%1_%3, 3, 3, 6, dst, src, len REP_RET %endmacro +%macro CONV_FLTP_TO_FLT_6CH 3 +cglobal pack_6ch_%2_to_%1_%3, 2,8,7, dst, src, src1, src2, src3, src4, src5, len +%if ARCH_X86_64 + mov lend, r2d +%else + %define lend dword r2m +%endif + mov src1q, [srcq+1*gprsize] + mov src2q, [srcq+2*gprsize] + mov src3q, [srcq+3*gprsize] + mov src4q, [srcq+4*gprsize] + mov src5q, [srcq+5*gprsize] + mov srcq, [srcq] + mov dstq, [dstq] + sub src1q, srcq + sub src2q, srcq + sub src3q, srcq + sub src4q, srcq + sub src5q, srcq +.loop: + mova m0, [srcq ] + mova m1, [srcq+src1q] + mova m2, [srcq+src2q] + mova m3, [srcq+src3q] + mova m4, [srcq+src4q] + mova m5, [srcq+src5q] +%if cpuflag(sse4) + SBUTTERFLYPS 0, 1, 6 + SBUTTERFLYPS 2, 3, 6 + SBUTTERFLYPS 4, 5, 6 + + blendps m6, m4, m0, 1100b + movlhps m0, m2 + movhlps m4, m2 + blendps m2, m5, m1, 1100b + movlhps m1, m3 + movhlps m5, m3 + + movaps [dstq ], m0 + movaps [dstq+16], m6 + movaps [dstq+32], m4 + movaps [dstq+48], m1 + movaps [dstq+64], m2 + movaps [dstq+80], m5 +%else ; mmx + SBUTTERFLY dq, 0, 1, 6 + SBUTTERFLY dq, 2, 3, 6 + SBUTTERFLY dq, 4, 5, 6 + + movq [dstq ], m0 + movq [dstq+ 8], m2 + movq [dstq+16], m4 + movq [dstq+24], m1 + movq [dstq+32], m3 + movq [dstq+40], m5 +%endif + add srcq, mmsize + add dstq, mmsize*6 + sub lend, mmsize/4 + jg .loop +%if mmsize == 8 + emms + RET +%else + REP_RET +%endif +%endmacro + +INIT_MMX mmx +CONV_FLTP_TO_FLT_6CH float,float,a +INIT_XMM sse4 +CONV_FLTP_TO_FLT_6CH float,float,a +%if HAVE_AVX +INIT_XMM avx +CONV_FLTP_TO_FLT_6CH float,float,a +%endif + + %macro INT16_TO_INT32_N 0 pxor m2, m2 pxor m3, m3 -- cgit v1.2.3