[FFmpeg-devel] [PATCH] Fix ff_imdct_calc_sse() on gcc-4.6.

[Måns Rullgård]

Alex Converse <alex.converse at gmail.com> writes:

> From 72140b8760c0014e6e05484e17440795a7a6e651 Mon Sep 17 00:00:00 2001
> From: Alex Converse <alex.converse at gmail.com>
> Date: Sun, 30 Jan 2011 01:04:41 -0800
> Subject: [PATCH] Fix ff_imdct_calc_sse() on gcc-4.6.
>
> Gcc 4.6 only preserves the first value when using a vector with an "m"
> constraint.
> ---
>  libavcodec/x86/fft_sse.c |    7 +++----
>  1 files changed, 3 insertions(+), 4 deletions(-)
>
> diff --git a/libavcodec/x86/fft_sse.c b/libavcodec/x86/fft_sse.c
> index 9f02816..0d151bc 100644
> --- a/libavcodec/x86/fft_sse.c
> +++ b/libavcodec/x86/fft_sse.c
> @@ -23,7 +23,7 @@
>  #include "libavcodec/dsputil.h"
>  #include "fft.h"
>  
> -DECLARE_ALIGNED(16, static const int, m1m1m1m1)[4] =
> +DECLARE_ASM_CONST(16, int, ff_m1m1m1m1)[4] =
>      { 1 << 31, 1 << 31, 1 << 31, 1 << 31 };
>  
>  void ff_fft_dispatch_sse(FFTComplex *z, int nbits);
> @@ -82,7 +82,7 @@ void ff_imdct_calc_sse(FFTContext *s, FFTSample *output, const FFTSample *input)
>      j = -n;
>      k = n-16;
>      __asm__ volatile(
> -        "movaps %4, %%xmm7 \n"
> +        "movaps "MANGLE(ff_m1m1m1m1)", %%xmm7 \n"
>          "1: \n"
>          "movaps       (%2,%1), %%xmm0 \n"
>          "movaps       (%3,%0), %%xmm1 \n"
> @@ -95,8 +95,7 @@ void ff_imdct_calc_sse(FFTContext *s, FFTSample *output, const FFTSample *input)
>          "add $16, %0 \n"
>          "jl 1b \n"
>          :"+r"(j), "+r"(k)
> -        :"r"(output+n4), "r"(output+n4*3),
> -         "m"(*m1m1m1m1)
> +        :"r"(output+n4), "r"(output+n4*3)
>          XMM_CLOBBERS_ONLY("%xmm0", "%xmm1", "%xmm7")
>      );
>  }

This looks correct as far as I can tell.

-- 
M?ns Rullg?rd
mans at mansr.com