[FFmpeg-devel] [PATCH 1/4] libswscale: Re-factor ff_shuffle_filter_coefficients.

* [FFmpeg-devel] [PATCH 1/4] libswscale: Re-factor ff_shuffle_filter_coefficients.
@ 2022-01-10 14:58 Alan Kelly
  2022-01-10 14:58 ` [FFmpeg-devel] [PATCH 2/4] libswscale: Avx2 hscale can process any input of size which is a multiple of 4 Alan Kelly
                   ` (4 more replies)
  0 siblings, 5 replies; 7+ messages in thread
From: Alan Kelly @ 2022-01-10 14:58 UTC (permalink / raw)
  To: ffmpeg-devel; +Cc: Alan Kelly

Make the code more readable, follow the style guide and propagate memory
allocation errors.
---
 libswscale/swscale_internal.h |  2 +-
 libswscale/utils.c            | 68 ++++++++++++++++++++---------------
 2 files changed, 40 insertions(+), 30 deletions(-)

diff --git a/libswscale/swscale_internal.h b/libswscale/swscale_internal.h
index 3a78d95ba6..26d28d42e6 100644
--- a/libswscale/swscale_internal.h
+++ b/libswscale/swscale_internal.h
@@ -1144,5 +1144,5 @@ void ff_sws_slice_worker(void *priv, int jobnr, int threadnr,
 #define MAX_LINES_AHEAD 4
 
 //shuffle filter and filterPos for hyScale and hcScale filters in avx2
-void ff_shuffle_filter_coefficients(SwsContext *c, int* filterPos, int filterSize, int16_t *filter, int dstW);
+int ff_shuffle_filter_coefficients(SwsContext *c, int* filterPos, int filterSize, int16_t *filter, int dstW);
 #endif /* SWSCALE_SWSCALE_INTERNAL_H */
diff --git a/libswscale/utils.c b/libswscale/utils.c
index c5ea8853d5..52f07e1661 100644
--- a/libswscale/utils.c
+++ b/libswscale/utils.c
@@ -278,39 +278,47 @@ static const FormatEntry format_entries[] = {
     [AV_PIX_FMT_P416LE]      = { 1, 1 },
 };
 
-void ff_shuffle_filter_coefficients(SwsContext *c, int *filterPos, int filterSize, int16_t *filter, int dstW){
+int ff_shuffle_filter_coefficients(SwsContext *c, int *filterPos,
+                                   int filterSize, int16_t *filter,
+                                   int dstW)
+{
 #if ARCH_X86_64
-    int i, j, k, l;
+    int i = 0, j = 0, k = 0;
     int cpu_flags = av_get_cpu_flags();
+    if (!filter || dstW % 16 != 0) return 0;
     if (EXTERNAL_AVX2_FAST(cpu_flags) && !(cpu_flags & AV_CPU_FLAG_SLOW_GATHER)) {
-        if ((c->srcBpc == 8) && (c->dstBpc <= 14)){
-            if (dstW % 16 == 0){
-                if (filter != NULL){
-                    for (i = 0; i < dstW; i += 8){
-                        FFSWAP(int, filterPos[i + 2], filterPos[i+4]);
-                        FFSWAP(int, filterPos[i + 3], filterPos[i+5]);
-                    }
-                    if (filterSize > 4){
-                        int16_t *tmp2 = av_malloc(dstW * filterSize * 2);
-                        memcpy(tmp2, filter, dstW * filterSize * 2);
-                        for (i = 0; i < dstW; i += 16){//pixel
-                            for (k = 0; k < filterSize / 4; ++k){//fcoeff
-                                for (j = 0; j < 16; ++j){//inner pixel
-                                    for (l = 0; l < 4; ++l){//coeff
-                                        int from = i * filterSize + j * filterSize + k * 4 + l;
-                                        int to = (i) * filterSize + j * 4 + l + k * 64;
-                                        filter[to] = tmp2[from];
-                                    }
-                                }
-                            }
-                        }
-                        av_free(tmp2);
-                    }
-                }
-            }
+        if ((c->srcBpc == 8) && (c->dstBpc <= 14)) {
+           int16_t *filterCopy = NULL;
+           if (filterSize > 4) {
+               if (!FF_ALLOC_TYPED_ARRAY(filterCopy, dstW * filterSize))
+                   return AVERROR(ENOMEM);
+               memcpy(filterCopy, filter, dstW * filterSize * sizeof(int16_t));
+           }
+           // Do not swap filterPos for pixels which won't be processed by
+           // the main loop.
+           for (i = 0; i + 8 <= dstW; i += 8) {
+               FFSWAP(int, filterPos[i + 2], filterPos[i + 4]);
+               FFSWAP(int, filterPos[i + 3], filterPos[i + 5]);
+           }
+           if (filterSize > 4) {
+               // 16 pixels are processed at a time.
+               for (i = 0; i + 16 <= dstW; i += 16) {
+                   // 4 filter coeffs are processed at a time.
+                   for (k = 0; k + 4 <= filterSize; k += 4) {
+                       for (j = 0; j < 16; ++j) {
+                           int from = (i + j) * filterSize + k;
+                           int to = i * filterSize + j * 4 + k * 16;
+                           memcpy(&filter[to], &filterCopy[from], 4 * sizeof(int16_t));
+                       }
+                   }
+               }
+           }
+           if (filterCopy)
+               av_free(filterCopy);
         }
     }
 #endif
+    return 0;
 }
 
 int sws_isSupportedInput(enum AVPixelFormat pix_fmt)
@@ -1836,7 +1844,8 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
                            get_local_pos(c, 0, 0, 0),
                            get_local_pos(c, 0, 0, 0))) < 0)
                 goto fail;
-            ff_shuffle_filter_coefficients(c, c->hLumFilterPos, c->hLumFilterSize, c->hLumFilter, dstW);
+            if ((ret = ff_shuffle_filter_coefficients(c, c->hLumFilterPos, c->hLumFilterSize, c->hLumFilter, dstW)) != 0)
+                goto nomem;
             if ((ret = initFilter(&c->hChrFilter, &c->hChrFilterPos,
                            &c->hChrFilterSize, c->chrXInc,
                            c->chrSrcW, c->chrDstW, filterAlign, 1 << 14,
@@ -1846,7 +1855,8 @@ av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
                            get_local_pos(c, c->chrSrcHSubSample, c->src_h_chr_pos, 0),
                            get_local_pos(c, c->chrDstHSubSample, c->dst_h_chr_pos, 0))) < 0)
                 goto fail;
-            ff_shuffle_filter_coefficients(c, c->hChrFilterPos, c->hChrFilterSize, c->hChrFilter, c->chrDstW);
+            if ((ret = ff_shuffle_filter_coefficients(c, c->hChrFilterPos, c->hChrFilterSize, c->hChrFilter, c->chrDstW)) != 0)
+                goto nomem;
         }
     } // initialize horizontal stuff
 
-- 
2.34.1.575.g55b058a8bb-goog

_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".

^ permalink raw reply	[flat|nested] 7+ messages in thread