swscale/aarch64/range_convert: saturate output instead of limiting input

aarch64 A55:
chrRangeFromJpeg8_1920_c:    28836.2 (1.00x)
chrRangeFromJpeg8_1920_neon:  5312.6 (5.43x)  5313.9 (5.43x)
chrRangeToJpeg8_1920_c:      44196.2 (1.00x)
chrRangeToJpeg8_1920_neon:    6034.6 (7.32x)  5551.3 (7.96x)
lumRangeFromJpeg8_1920_c:    15388.5 (1.00x)
lumRangeFromJpeg8_1920_neon:  3150.7 (4.88x)  3152.3 (4.88x)
lumRangeToJpeg8_1920_c:      23069.7 (1.00x)
lumRangeToJpeg8_1920_neon:    3873.2 (5.96x)  3628.7 (6.36x)

aarch64 A76:
chrRangeFromJpeg8_1920_c:     6334.7 (1.00x)
chrRangeFromJpeg8_1920_neon:  2264.5 (2.80x)  2344.5 (2.70x)
chrRangeToJpeg8_1920_c:      11474.5 (1.00x)
chrRangeToJpeg8_1920_neon:    2646.5 (4.34x)  2824.2 (4.06x)
lumRangeFromJpeg8_1920_c:     4453.2 (1.00x)
lumRangeFromJpeg8_1920_neon:  1104.8 (4.03x)  1104.5 (4.03x)
lumRangeToJpeg8_1920_c:       6645.0 (1.00x)
lumRangeToJpeg8_1920_neon:    1310.5 (5.07x)  1329.8 (5.00x)

2 files changed, 18 insertions, 26 deletions

diff --git a/libswscale/aarch64/range_convert_neon.S b/libswscale/aarch64/range_convert_neon.S
index 30991ab2a6..2f418adb24 100644
--- a/libswscale/aarch64/range_convert_neon.S
+++ b/libswscale/aarch64/range_convert_neon.S
@@ -20,12 +20,8 @@
 
 #include "libavutil/aarch64/asm.S"
 
-.macro lumConvertRange name, max, mult, offset, shift
+.macro lumConvertRange name, fromto, mult, offset, shift
 function ff_\name, export=1
-.if \max != 0
-        mov             w3, #\max
-        dup             v24.8h, w3
-.endif
         mov             w3, #\mult
         dup             v25.4s, w3
         movz            w3, #(\offset & 0xffff)
@@ -33,17 +29,19 @@ function ff_\name, export=1
         dup             v26.4s, w3
 1:
         ld1             {v0.8h}, [x0]
-.if \max != 0
-        smin            v0.8h, v0.8h, v24.8h
-.endif
         mov             v16.16b, v26.16b
         mov             v18.16b, v26.16b
         sxtl            v20.4s, v0.4h
         sxtl2           v22.4s, v0.8h
         mla             v16.4s, v20.4s, v25.4s
         mla             v18.4s, v22.4s, v25.4s
+.ifc \fromto, To
+        sqshrn          v0.4h, v16.4s, #\shift
+        sqshrn2         v0.8h, v18.4s, #\shift
+.else
         shrn            v0.4h, v16.4s, #\shift
         shrn2           v0.8h, v18.4s, #\shift
+.endif
         subs            w1, w1, #8
         st1             {v0.8h}, [x0], #16
         b.gt            1b
@@ -51,12 +49,8 @@ function ff_\name, export=1
 endfunc
 .endm
 
-.macro chrConvertRange name, max, mult, offset, shift
+.macro chrConvertRange name, fromto, mult, offset, shift
 function ff_\name, export=1
-.if \max != 0
-        mov             w3, #\max
-        dup             v24.8h, w3
-.endif
         mov             w3, #\mult
         dup             v25.4s, w3
         movz            w3, #(\offset & 0xffff)
@@ -65,10 +59,6 @@ function ff_\name, export=1
 1:
         ld1             {v0.8h}, [x0]
         ld1             {v1.8h}, [x1]
-.if \max != 0
-        smin            v0.8h, v0.8h, v24.8h
-        smin            v1.8h, v1.8h, v24.8h
-.endif
         mov             v16.16b, v26.16b
         mov             v17.16b, v26.16b
         mov             v18.16b, v26.16b
@@ -81,10 +71,17 @@ function ff_\name, export=1
         mla             v17.4s, v21.4s, v25.4s
         mla             v18.4s, v22.4s, v25.4s
         mla             v19.4s, v23.4s, v25.4s
+.ifc \fromto, To
+        sqshrn          v0.4h, v16.4s, #\shift
+        sqshrn          v1.4h, v17.4s, #\shift
+        sqshrn2         v0.8h, v18.4s, #\shift
+        sqshrn2         v1.8h, v19.4s, #\shift
+.else
         shrn            v0.4h, v16.4s, #\shift
         shrn            v1.4h, v17.4s, #\shift
         shrn2           v0.8h, v18.4s, #\shift
         shrn2           v1.8h, v19.4s, #\shift
+.endif
         subs            w2, w2, #8
         st1             {v0.8h}, [x0], #16
         st1             {v1.8h}, [x1], #16
@@ -93,7 +90,7 @@ function ff_\name, export=1
 endfunc
 .endm
 
-lumConvertRange lumRangeToJpeg_neon,   30189, 19077, -39057361, 14
-chrConvertRange chrRangeToJpeg_neon,   30775,  4663,  -9289992, 12
-lumConvertRange lumRangeFromJpeg_neon,     0, 14071,  33561947, 14
-chrConvertRange chrRangeFromJpeg_neon,     0,  1799,   4081085, 11
+lumConvertRange lumRangeToJpeg_neon,   To,   19077, -39057361, 14
+chrConvertRange chrRangeToJpeg_neon,   To,    4663,  -9289992, 12
+lumConvertRange lumRangeFromJpeg_neon, From, 14071,  33561947, 14
+chrConvertRange chrRangeFromJpeg_neon, From,  1799,   4081085, 11
diff --git a/libswscale/aarch64/swscale.c b/libswscale/aarch64/swscale.c
index 1fce77df26..5173359e09 100644
--- a/libswscale/aarch64/swscale.c
+++ b/libswscale/aarch64/swscale.c
@@ -225,10 +225,6 @@ void ff_chrRangeToJpeg_neon(int16_t *dstU, int16_t *dstV, int width);
 
 av_cold void ff_sws_init_range_convert_aarch64(SwsInternal *c)
 {
-    /* This code is currently disabled because of changes in the base
-     * implementation of these functions. This code should be enabled
-     * again once those changes are ported to this architecture. */
-#if 0
     int cpu_flags = av_get_cpu_flags();
 
     if (have_neon(cpu_flags)) {
@@ -242,7 +238,6 @@ av_cold void ff_sws_init_range_convert_aarch64(SwsInternal *c)
             }
         }
     }
-#endif
 }
 
 av_cold void ff_sws_init_swscale_aarch64(SwsInternal *c)