Merge remote-tracking branch 'qatar/master'

[ffmpeg] / libavresample / x86 / audio_mix.asm

diff --git a/libavresample/x86/audio_mix.asm b/libavresample/x86/audio_mix.asm
index dbc79e585d8ffc28af62570ea1719c2adddd1fae..4b0434dd6de7c33aaf3f366adb3ed946c9af5e4f 100644 (file)
--- a/libavresample/x86/audio_mix.asm
+++ b/libavresample/x86/audio_mix.asm
@@ -21,6 +21,7 @@
 
 %include "x86inc.asm"
 %include "x86util.asm"
+%include "util.asm"
 
 SECTION_TEXT
 
@@ -64,3 +65,169 @@ MIX_2_TO_1_FLTP_FLT
 INIT_YMM avx
 MIX_2_TO_1_FLTP_FLT
 %endif
+
+;-----------------------------------------------------------------------------
+; void ff_mix_2_to_1_s16p_flt(int16_t **src, float **matrix, int len,
+;                             int out_ch, int in_ch);
+;-----------------------------------------------------------------------------
+
+%macro MIX_2_TO_1_S16P_FLT 0
+cglobal mix_2_to_1_s16p_flt, 3,4,6, src, matrix, len, src1
+    mov       src1q, [srcq+gprsize]
+    mov        srcq, [srcq]
+    sub       src1q, srcq
+    mov     matrixq, [matrixq  ]
+    VBROADCASTSS m4, [matrixq  ]
+    VBROADCASTSS m5, [matrixq+4]
+    ALIGN 16
+.loop:
+    mova         m0, [srcq      ]
+    mova         m2, [srcq+src1q]
+    S16_TO_S32_SX 0, 1
+    S16_TO_S32_SX 2, 3
+    cvtdq2ps     m0, m0
+    cvtdq2ps     m1, m1
+    cvtdq2ps     m2, m2
+    cvtdq2ps     m3, m3
+    mulps        m0, m4
+    mulps        m1, m4
+    mulps        m2, m5
+    mulps        m3, m5
+    addps        m0, m2
+    addps        m1, m3
+    cvtps2dq     m0, m0
+    cvtps2dq     m1, m1
+    packssdw     m0, m1
+    mova     [srcq], m0
+    add        srcq, mmsize
+    sub        lend, mmsize/2
+    jg .loop
+    REP_RET
+%endmacro
+
+INIT_XMM sse2
+MIX_2_TO_1_S16P_FLT
+INIT_XMM sse4
+MIX_2_TO_1_S16P_FLT
+
+;-----------------------------------------------------------------------------
+; void ff_mix_2_to_1_s16p_q8(int16_t **src, int16_t **matrix, int len,
+;                            int out_ch, int in_ch);
+;-----------------------------------------------------------------------------
+
+INIT_XMM sse2
+cglobal mix_2_to_1_s16p_q8, 3,4,6, src, matrix, len, src1
+    mov       src1q, [srcq+gprsize]
+    mov        srcq, [srcq]
+    sub       src1q, srcq
+    mov     matrixq, [matrixq]
+    movd         m4, [matrixq]
+    movd         m5, [matrixq]
+    SPLATW       m4, m4, 0
+    SPLATW       m5, m5, 1
+    pxor         m0, m0
+    punpcklwd    m4, m0
+    punpcklwd    m5, m0
+    ALIGN 16
+.loop:
+    mova         m0, [srcq      ]
+    mova         m2, [srcq+src1q]
+    punpckhwd    m1, m0, m0
+    punpcklwd    m0, m0
+    punpckhwd    m3, m2, m2
+    punpcklwd    m2, m2
+    pmaddwd      m0, m4
+    pmaddwd      m1, m4
+    pmaddwd      m2, m5
+    pmaddwd      m3, m5
+    paddd        m0, m2
+    paddd        m1, m3
+    psrad        m0, 8
+    psrad        m1, 8
+    packssdw     m0, m1
+    mova     [srcq], m0
+    add        srcq, mmsize
+    sub        lend, mmsize/2
+    jg .loop
+    REP_RET
+
+;-----------------------------------------------------------------------------
+; void ff_mix_1_to_2_fltp_flt(float **src, float **matrix, int len,
+;                             int out_ch, int in_ch);
+;-----------------------------------------------------------------------------
+
+%macro MIX_1_TO_2_FLTP_FLT 0
+cglobal mix_1_to_2_fltp_flt, 3,5,4, src0, matrix0, len, src1, matrix1
+    mov       src1q, [src0q+gprsize]
+    mov       src0q, [src0q]
+    sub       src1q, src0q
+    mov    matrix1q, [matrix0q+gprsize]
+    mov    matrix0q, [matrix0q]
+    VBROADCASTSS m2, [matrix0q]
+    VBROADCASTSS m3, [matrix1q]
+    ALIGN 16
+.loop:
+    mova         m0, [src0q]
+    mulps        m1, m0, m3
+    mulps        m0, m0, m2
+    mova  [src0q      ], m0
+    mova  [src0q+src1q], m1
+    add       src0q, mmsize
+    sub        lend, mmsize/4
+    jg .loop
+    REP_RET
+%endmacro
+
+INIT_XMM sse
+MIX_1_TO_2_FLTP_FLT
+%if HAVE_AVX
+INIT_YMM avx
+MIX_1_TO_2_FLTP_FLT
+%endif
+
+;-----------------------------------------------------------------------------
+; void ff_mix_1_to_2_s16p_flt(int16_t **src, float **matrix, int len,
+;                             int out_ch, int in_ch);
+;-----------------------------------------------------------------------------
+
+%macro MIX_1_TO_2_S16P_FLT 0
+cglobal mix_1_to_2_s16p_flt, 3,5,6, src0, matrix0, len, src1, matrix1
+    mov       src1q, [src0q+gprsize]
+    mov       src0q, [src0q]
+    sub       src1q, src0q
+    mov    matrix1q, [matrix0q+gprsize]
+    mov    matrix0q, [matrix0q]
+    VBROADCASTSS m4, [matrix0q]
+    VBROADCASTSS m5, [matrix1q]
+    ALIGN 16
+.loop:
+    mova         m0, [src0q]
+    S16_TO_S32_SX 0, 2
+    cvtdq2ps     m0, m0
+    cvtdq2ps     m2, m2
+    mulps        m1, m0, m5
+    mulps        m0, m0, m4
+    mulps        m3, m2, m5
+    mulps        m2, m2, m4
+    cvtps2dq     m0, m0
+    cvtps2dq     m1, m1
+    cvtps2dq     m2, m2
+    cvtps2dq     m3, m3
+    packssdw     m0, m2
+    packssdw     m1, m3
+    mova  [src0q      ], m0
+    mova  [src0q+src1q], m1
+    add       src0q, mmsize
+    sub        lend, mmsize/2
+    jg .loop
+    REP_RET
+%endmacro
+
+INIT_XMM sse2
+MIX_1_TO_2_S16P_FLT
+INIT_XMM sse4
+MIX_1_TO_2_S16P_FLT
+%if HAVE_AVX
+INIT_XMM avx
+MIX_1_TO_2_S16P_FLT
+%endif