AuroraMiddleware/zstd
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Merge pull request #1866 from legrosbuffle/dev

Browse Source 
Optimized loop bounds to allow the compiler to unroll the loop.
This commit is contained in:
Nick Terrell 2019-11-18 16:16:30 -08:00 committed by GitHub
commit 6a7f65117e
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 22 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

32
lib/decompress/huf_decompress.c
View File

@ -181,17 +181,29 @@ size_t HUF_readDTableX1_wksp(HUF_DTable* DTable, const void* src, size_t srcSize
/* fill DTable */
{ U32 n;
for (n=0; n<nbSymbols; n++) {
U32 const w = huffWeight[n];
U32 const length = (1 << w) >> 1;
U32 u;
size_t const nEnd = nbSymbols;
for (n=0; n<nEnd; n++) {
size_t const w = huffWeight[n];
size_t const length = (1 << w) >> 1;
size_t const uStart = rankVal[w];
size_t const uEnd = uStart + length;
size_t u;
HUF_DEltX1 D;
D.byte = (BYTE)n; D.nbBits = (BYTE)(tableLog + 1 - w);
for (u = rankVal[w]; u < rankVal[w] + length; u++)
dt[u] = D;
rankVal[w] += length;
} }
D.byte = (BYTE)n;
D.nbBits = (BYTE)(tableLog + 1 - w);
rankVal[w] = uEnd;
if (length < 4) {
/* Use length in the loop bound so the compiler knows it is short. */
for (u = 0; u < length; ++u)
dt[uStart + u] = D;
} else {
/* Unroll the loop 4 times, we know it is a power of 2. */
for (u = uStart; u < uEnd; u += 4) {
dt[u + 0] = D;
dt[u + 1] = D;
dt[u + 2] = D;
dt[u + 3] = D;
} } } }
return iSize;
}