Evas: Neon: More .fpu neon flags

Also clean up some ugly code.

SVN revision: 55314

legacy/evas/src/lib/engines/common/evas_blit_main.c

@@ -136,6 +136,7 @@ evas_common_copy_pixels_rev_neon(DATA32 *src, DATA32 *dst, int len)
    uint32_t *tmp = (void *)37;
 #define AP	"evas_common_copy_rev_pixels_neon_"
    asm volatile (
+		".fpu neon				\n\t"
 		// Can we do 32 byte?
 		"andS		%[tmp],	%[d], $0x1f	\n\t"
 		"beq		"AP"quadstart		\n\t"
@@ -334,6 +335,7 @@ evas_common_copy_pixels_neon(DATA32 *src, DATA32 *dst, int len){
    e = dst + len;
 #define AP	"evas_common_copy_pixels_neon_"
    asm volatile (
+		".fpu neon				\n\t"
 		// Can we do 32 byte?
 		"andS		%[tmp],	%[d], $0x1f	\n\t"
 		"beq		"AP"quadstart		\n\t"

legacy/evas/src/lib/engines/common/evas_convert_rgb_32.c

@@ -102,6 +102,7 @@ evas_common_convert_rgba_to_32bpp_rgb_8888_rot_90 (DATA32 *src, DATA8 *dst, int
    } else {
 #define AP	"convert_rgba32_rot_90_"
 	asm volatile (
+	".fpu neon						\n\t"
 	"	mov		%[s1],	%[src]			\n\t"
 	"	add		%[s1],	%[h],lsl #2		\n\t"
 	"	sub		%[s1],	#8			\n\t"

legacy/evas/src/lib/engines/common/evas_op_blend/op_blend_pixel_neon.c

@@ -151,10 +151,6 @@ _op_blend_p_dp_neon(DATA32 *s, DATA8 *m, DATA32 c, DATA32 *d, int l) {
                 "bhi	 "AP"octloopint\n\t"
 
 	AP "loopout:					\n\t"
-//"sub %[tmp], %[d], #4\n\t"
-//"vmov.i16	d0, $0xff00 \n\t"
-//"vst1.32	d0[0],	[%[tmp]]		\n\t"
-
 		"cmp 		%[d], %[e]		\n\t"
                 "beq 		"AP"done		\n\t"
 		"sub		%[tmp],%[e], %[d]	\n\t"

legacy/evas/src/lib/engines/common/evas_op_copy/op_copy_pixel_neon.c

@@ -6,7 +6,7 @@ _op_copy_p_dp_neon(DATA32 *s, DATA8 *m __UNUSED__, DATA32 c __UNUSED__, DATA32 *
    uint32_t *e;
    e = d + l;
 //#ifdef NEON_INSTRINSICS_OK
-#if 1
+#if 0
    // odd this is faster than the below asm... :(
    e -= 15;
    uint32x4_t col1, col2, col3, col4;