daiduo/TurboPFor-Integer-Compression
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix buffer overruns in transpose.c when ESIZE is 2

Browse Source 
MSVS reports such warnings/errors otherwise:

TurboPFor\transpose.c(1120): error C4789: buffer 'ov' of size 32 bytes will be overrun; 16 bytes will be written starting at offset 48
TurboPFor\transpose.c(1112): error C4789: buffer 'ov' of size 32 bytes will be overrun; 16 bytes will be written starting at offset 32
This commit is contained in:
Pavel P
2023-01-03 00:48:00 +03:00
parent 1ccb7ea88b
commit 10372a5cd0
1 changed files with 4 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
transpose.c
View File

@ -596,7 +596,7 @@ void TEMPLATE2(TPENC256V, ESIZE)(unsigned char *in, unsigned n, unsigned char *o
for(ip = in,op = out; ip != in+v; ip += ESIZE*32, op += ESIZE*32/STRIDE) { for(ip = in,op = out; ip != in+v; ip += ESIZE*32, op += ESIZE*32/STRIDE) {
unsigned char *p = op; PREFETCH(ip+ESIZE*192,0); unsigned char *p = op; PREFETCH(ip+ESIZE*192,0);
__m256i iv[ESIZE],ov[ESIZE]; __m256i iv[ESIZE],ov[ESIZE == 2 ? ESIZE + 2 : ESIZE];
#if ESIZE == 2 #if ESIZE == 2
ov[0] = _mm256_shuffle_epi8(LD256((__m256i *) ip ), sv0); ov[0] = _mm256_shuffle_epi8(LD256((__m256i *) ip ), sv0);
ov[1] = _mm256_shuffle_epi8(LD256((__m256i *)(ip+32)), sv1); ov[1] = _mm256_shuffle_epi8(LD256((__m256i *)(ip+32)), sv1);
@ -724,7 +724,7 @@ void TEMPLATE2(TPDEC256V, ESIZE)(unsigned char *in, unsigned n, unsigned char *o
#endif #endif
for(op = out,ip = in; op != out+v; ip += ESIZE*32/STRIDE, op += ESIZE*32) { unsigned char *p = ip; PREFETCH(ip+ESIZE*192,0); for(op = out,ip = in; op != out+v; ip += ESIZE*32/STRIDE, op += ESIZE*32) { unsigned char *p = ip; PREFETCH(ip+ESIZE*192,0);
__m256i iv[ESIZE], ov[ESIZE]; __m256i iv[ESIZE], ov[ESIZE == 2 ? ESIZE + 2 : ESIZE];
#if STRIDE > ESIZE #if STRIDE > ESIZE
NBL0(0,1); NBL( 2,3); NB(0,1,iv[0]); NB(2,3,iv[1]); NBL0(0,1); NBL( 2,3); NB(0,1,iv[0]); NB(2,3,iv[1]);
@ -842,7 +842,7 @@ void TEMPLATE2(TPENC128V, ESIZE)(unsigned char *in, unsigned n, unsigned char *o
#endif #endif
for(ip = in, op = out; ip != in+v; ip+=ESIZE*16,op += ESIZE*16/STRIDE) { unsigned char *p = op; PREFETCH(ip+(ESIZE*16)*ESIZE,0); for(ip = in, op = out; ip != in+v; ip+=ESIZE*16,op += ESIZE*16/STRIDE) { unsigned char *p = op; PREFETCH(ip+(ESIZE*16)*ESIZE,0);
__m128i iv[ESIZE],ov[ESIZE]; __m128i iv[ESIZE],ov[ESIZE == 2 ? ESIZE + 2 : ESIZE];
#if defined(__SSSE3__) || defined(__ARM_NEON) #if defined(__SSSE3__) || defined(__ARM_NEON)
#if ESIZE == 2 #if ESIZE == 2
#ifdef __ARM_NEON #ifdef __ARM_NEON
@ -1100,7 +1100,7 @@ void TEMPLATE2(TPDEC128V, ESIZE)(unsigned char *in, unsigned n, unsigned char *o
for(op = out,ip = in; op != out+v; op+=ESIZE*16,ip += ESIZE*16/STRIDE) { for(op = out,ip = in; op != out+v; op+=ESIZE*16,ip += ESIZE*16/STRIDE) {
unsigned char *p=ip; PREFETCH(ip+(ESIZE*16/STRIDE)*ESIZE,0); unsigned char *p=ip; PREFETCH(ip+(ESIZE*16/STRIDE)*ESIZE,0);
__m128i iv[ESIZE], ov[ESIZE]; __m128i iv[ESIZE], ov[ESIZE == 2 ? ESIZE + 2 : ESIZE];
#if STRIDE > ESIZE //------------ Nibble transpose ------------------- #if STRIDE > ESIZE //------------ Nibble transpose -------------------
ov[0] = _mm_loadl_epi64((__m128i *) p ); ov[0] = _mm_loadl_epi64((__m128i *) p );