aarch64: vp9itxfm: Make the larger core transforms standalone functions

[ffmpeg] / libavcodec / huffyuvencdsp.c

diff --git a/libavcodec/huffyuvencdsp.c b/libavcodec/huffyuvencdsp.c
index fdcd0b06aa703b8474e089883c38578b870a184a..427d118be45189ec75cd548ac96957800bd9c574 100644 (file)
--- a/libavcodec/huffyuvencdsp.c
+++ b/libavcodec/huffyuvencdsp.c
@@ -21,64 +21,58 @@
 #include "huffyuvencdsp.h"
 #include "mathops.h"
 
-// 0x7f7f7f7f or 0x7f7f7f7f7f7f7f7f or whatever, depending on the cpu's native arithmetic size
-#define pb_7f (~0UL / 255 * 0x7f)
-#define pb_80 (~0UL / 255 * 0x80)
+// 0x00010001 or 0x0001000100010001 or whatever, depending on the cpu's native arithmetic size
+#define pw_1 (ULONG_MAX / UINT16_MAX)
 
-static void diff_bytes_c(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, intptr_t w)
-{
+static void diff_int16_c(uint16_t *dst, const uint16_t *src1, const uint16_t *src2, unsigned mask, int w){
     long i;
-
 #if !HAVE_FAST_UNALIGNED
-    if (((long)src1 | (long)src2) & (sizeof(long) - 1)) {
-        for (i = 0; i + 7 < w; i += 8) {
-            dst[i + 0] = src1[i + 0] - src2[i + 0];
-            dst[i + 1] = src1[i + 1] - src2[i + 1];
-            dst[i + 2] = src1[i + 2] - src2[i + 2];
-            dst[i + 3] = src1[i + 3] - src2[i + 3];
-            dst[i + 4] = src1[i + 4] - src2[i + 4];
-            dst[i + 5] = src1[i + 5] - src2[i + 5];
-            dst[i + 6] = src1[i + 6] - src2[i + 6];
-            dst[i + 7] = src1[i + 7] - src2[i + 7];
+    if((long)src2 & (sizeof(long)-1)){
+        for(i=0; i+3<w; i+=4){
+            dst[i+0] = (src1[i+0]-src2[i+0]) & mask;
+            dst[i+1] = (src1[i+1]-src2[i+1]) & mask;
+            dst[i+2] = (src1[i+2]-src2[i+2]) & mask;
+            dst[i+3] = (src1[i+3]-src2[i+3]) & mask;
         }
-    } else
+    }else
 #endif
-    for (i = 0; i <= w - (int) sizeof(long); i += sizeof(long)) {
-        long a = *(long *) (src1 + i);
-        long b = *(long *) (src2 + i);
-        *(long *) (dst + i) = ((a | pb_80) - (b & pb_7f)) ^
-                              ((a ^ b ^ pb_80) & pb_80);
+    {
+        unsigned long pw_lsb = (mask >> 1) * pw_1;
+        unsigned long pw_msb = pw_lsb +  pw_1;
+
+        for (i = 0; i <= w - (int)sizeof(long)/2; i += sizeof(long)/2) {
+            long a = *(long*)(src1+i);
+            long b = *(long*)(src2+i);
+            *(long*)(dst+i) = ((a|pw_msb) - (b&pw_lsb)) ^ ((a^b^pw_msb)&pw_msb);
+        }
     }
-    for (; i < w; i++)
-        dst[i + 0] = src1[i + 0] - src2[i + 0];
+    for (; i<w; i++)
+        dst[i] = (src1[i] - src2[i]) & mask;
 }
 
-static void sub_hfyu_median_pred_c(uint8_t *dst, const uint8_t *src1,
-                                   const uint8_t *src2, intptr_t w,
-                                   int *left, int *left_top)
-{
+static void sub_hfyu_median_pred_int16_c(uint16_t *dst, const uint16_t *src1, const uint16_t *src2, unsigned mask, int w, int *left, int *left_top){
     int i;
-    uint8_t l, lt;
+    uint16_t l, lt;
 
     l  = *left;
     lt = *left_top;
 
-    for (i = 0; i < w; i++) {
-        const int pred = mid_pred(l, src1[i], (l + src1[i] - lt) & 0xFF);
-        lt     = src1[i];
-        l      = src2[i];
-        dst[i] = l - pred;
+    for(i=0; i<w; i++){
+        const int pred = mid_pred(l, src1[i], (l + src1[i] - lt) & mask);
+        lt = src1[i];
+        l  = src2[i];
+        dst[i] = (l - pred) & mask;
     }
 
     *left     = l;
     *left_top = lt;
 }
 
-av_cold void ff_huffyuvencdsp_init(HuffYUVEncDSPContext *c)
+av_cold void ff_huffyuvencdsp_init(HuffYUVEncDSPContext *c, AVCodecContext *avctx)
 {
-    c->diff_bytes           = diff_bytes_c;
-    c->sub_hfyu_median_pred = sub_hfyu_median_pred_c;
+    c->diff_int16           = diff_int16_c;
+    c->sub_hfyu_median_pred_int16 = sub_hfyu_median_pred_int16_c;
 
     if (ARCH_X86)
-        ff_huffyuvencdsp_init_x86(c);
+        ff_huffyuvencdsp_init_x86(c, avctx);
 }