x86/dct32: don't build ff_dct32_float_sse on x86_64

There's an SSE2 version already, and technically the SSE version
on x86_64 was wrong (using pshufd and pshuflw, SSE2 instructions).

Signed-off-by: James Almer <jamrial@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

libavcodec/x86/dct32.asm

@@ -484,7 +484,9 @@ cglobal dct32_float, 2, 3, 16, out, in, tmp
 %endif
 %endmacro
 
+%if ARCH_X86_32
 INIT_XMM sse
 DCT32_FUNC
+%endif
 INIT_XMM sse2
 DCT32_FUNC

libavcodec/x86/dct_init.c

@@ -30,7 +30,7 @@ av_cold void ff_dct_init_x86(DCTContext *s)
 {
     int cpu_flags = av_get_cpu_flags();
 
-    if (EXTERNAL_SSE(cpu_flags))
+    if (EXTERNAL_SSE(cpu_flags) && ARCH_X86_32)
         s->dct32 = ff_dct32_float_sse;
     if (EXTERNAL_SSE2(cpu_flags))
         s->dct32 = ff_dct32_float_sse2;