x86/swr: add ff_float_to_int32_a_avx2

13797 decicycles in ff_float_to_int32_a_sse2, 32768 runs, 0 skips
8603 decicycles in ff_float_to_int32_a_avx2, 32766 runs, 2 skips

Reviewed-by: Christophe Gisquet <christophe.gisquet@gmail.com>
Signed-off-by: James Almer <jamrial@gmail.com>

libswresample/x86/audio_convert.asm

@@ -342,10 +342,10 @@ pack_6ch_%2_to_%1_u_int %+ SUFFIX
     mulps %1, %5
     mulps %2, %5
     cvtps2dq  %6, %1
-    cmpnltps %1, %5
+    cmpps %1, %1, %5, 5
     paddd %1, %6
     cvtps2dq  %6, %2
-    cmpnltps %2, %5
+    cmpps %2, %2, %5, 5
     paddd %2, %6
 %endmacro
 
@@ -474,3 +474,9 @@ INIT_YMM avx
 CONV float, int32, u, 2, 2, INT32_TO_FLOAT_N, INT32_TO_FLOAT_INIT
 CONV float, int32, a, 2, 2, INT32_TO_FLOAT_N, INT32_TO_FLOAT_INIT
 %endif
+
+%if HAVE_AVX2_EXTERNAL
+INIT_YMM avx2
+CONV int32, float, u, 2, 2, FLOAT_TO_INT32_N, FLOAT_TO_INT32_INIT
+CONV int32, float, a, 2, 2, FLOAT_TO_INT32_N, FLOAT_TO_INT32_INIT
+%endif

libswresample/x86/audio_convert_init.c

@@ -25,7 +25,7 @@
 #define PROTO(pre, in, out, cap) void ff ## pre ## in## _to_ ##out## _a_ ##cap(uint8_t **dst, const uint8_t **src, int len);
 #define PROTO2(pre, out, cap) PROTO(pre, int16, out, cap) PROTO(pre, int32, out, cap) PROTO(pre, float, out, cap)
 #define PROTO3(pre, cap) PROTO2(pre, int16, cap) PROTO2(pre, int32, cap) PROTO2(pre, float, cap)
-#define PROTO4(pre) PROTO3(pre, mmx) PROTO3(pre, sse) PROTO3(pre, sse2) PROTO3(pre, ssse3) PROTO3(pre, sse4) PROTO3(pre, avx)
+#define PROTO4(pre) PROTO3(pre, mmx) PROTO3(pre, sse) PROTO3(pre, sse2) PROTO3(pre, ssse3) PROTO3(pre, sse4) PROTO3(pre, avx) PROTO3(pre, avx2)
 PROTO4(_)
 PROTO4(_pack_2ch_)
 PROTO4(_pack_6ch_)
@@ -139,4 +139,8 @@ MULTI_CAPS_FUNC(SSE2, sse2)
                 ac->simd_f =  ff_pack_6ch_float_to_int32_a_avx;
         }
     }
+    if(EXTERNAL_AVX2(mm_flags)) {
+        if(   out_fmt == AV_SAMPLE_FMT_S32  && in_fmt == AV_SAMPLE_FMT_FLT || out_fmt == AV_SAMPLE_FMT_S32P && in_fmt == AV_SAMPLE_FMT_FLTP)
+            ac->simd_f =  ff_float_to_int32_a_avx2;
+    }
 }