From: Dave Watson <davejwatson@fb.com>
Date: Fri, 25 Jan 2019 22:40:27 +0000 (-0800)
Subject: decompress_generic: Optimize literal copies
X-Git-Tag: upstream/1.9.3~5^2~24^2~6
X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=59332a3026eaff40e4c6d3c627bdb51bfd26fea1;p=platform%2Fupstream%2Flz4.git

decompress_generic: Optimize literal copies

Use LZ4_wildCopy16 for variable-length literals.  For literal counts that
fit in the flag byte, copy directly.  We can also omit oend checks for
roughly the same reason as the previous shortcut:  We check once that both
match length and literal length fit in FASTLOOP_SAFE_DISTANCE, including
wildcopy distance.
---

diff --git a/lib/lz4.c b/lib/lz4.c
index 1e938d0..978a957 100644
--- a/lib/lz4.c
+++ b/lib/lz4.c
@@ -1545,22 +1545,31 @@ LZ4_decompress_generic(
 
             /* decode literal length */
             if (length == RUN_MASK) {
-              variable_length_error error = ok;
-              length += read_variable_length(&ip, iend-RUN_MASK, endOnInput, endOnInput, &error);
-              if (error == initial_error) goto _output_error;
+                variable_length_error error = ok;
+                length += read_variable_length(&ip, iend-RUN_MASK, endOnInput, endOnInput, &error);
+                if (error == initial_error) goto _output_error;
                 if ((safeDecode) && unlikely((uptrval)(op)+length<(uptrval)(op))) goto _output_error;   /* overflow detection */
                 if ((safeDecode) && unlikely((uptrval)(ip)+length<(uptrval)(ip))) goto _output_error;   /* overflow detection */
-            }
 
-            /* copy literals */
-            cpy = op+length;
-            LZ4_STATIC_ASSERT(MFLIMIT >= WILDCOPYLENGTH);
-            if ( ((endOnInput) && ((cpy>oend-FASTLOOP_SAFE_DISTANCE) || (ip+length>iend-(2+1+LASTLITERALS))) )
-              || ((!endOnInput) && (cpy>oend-FASTLOOP_SAFE_DISTANCE)) )
-            {
-                goto safe_literal_copy;
+                /* copy literals */
+                cpy = op+length;
+                LZ4_STATIC_ASSERT(MFLIMIT >= WILDCOPYLENGTH);
+                if ( ((endOnInput) && ((cpy>oend-FASTLOOP_SAFE_DISTANCE) || (ip+length>iend-(2+1+LASTLITERALS))) )
+                     || ((!endOnInput) && (cpy>oend-FASTLOOP_SAFE_DISTANCE)) )
+                    {
+                        goto safe_literal_copy;
+                    }
+                LZ4_wildCopy16(op, ip, cpy);
+                ip += length; op = cpy;
             } else {
-                LZ4_wildCopy(op, ip, cpy);   /* may overwrite up to WILDCOPYLENGTH beyond cpy */
+                cpy = op+length;
+                /* We don't need to check oend, since we check it once for each loop below */
+                if ( ((endOnInput) && (ip+16>iend-(2+1+LASTLITERALS))))
+                    {
+                        goto safe_literal_copy;
+                    }
+                /* Literals can only be 14, but hope compilers optimize if we copy by a register size */
+                memcpy(op, ip, 16);
                 ip += length; op = cpy;
             }