dany/libwebp
1
0
Fork 0
mirror of https://github.com/webmproject/libwebp.git synced 2024-12-27 06:08:21 +01:00
Code Issues Actions Packages Projects Releases Wiki Activity

fix 16b overflow in SSE2

Browse Source 
the 'accum' variable can be larger than 15b for large
rescale values.

Assert triggered:
 src/dsp/rescaler_sse2.c:249: RescalerExportRowExpand_SSE2: Assertion `v >= 0 && v <= 255' failed.
 src/dsp/rescaler_sse2.c:350: RescalerExportRowShrink_SSE2: Assertion `v >= 0 && v <= 255' failed.

-> fall back to C implementation in this case for now

Change-Id: I7ea1cb72301cafc1459be403f6a6f4e3cbc89bb1
This commit is contained in:
Pascal Massimino 2018-04-11 15:17:14 +02:00 committed by James Zern
parent e577feb7c2
commit c1cb86af5f
1 changed files with 9 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
src/dsp/rescaler_sse2.c
View File

@ -36,7 +36,7 @@ static void LoadTwoPixels_SSE2(const uint8_t* const src, __m128i* out) {
} }
// input: 8 bytes ABCDEFGH -> output: A0B0C0D0E0F0G0H0 // input: 8 bytes ABCDEFGH -> output: A0B0C0D0E0F0G0H0
static void LoadHeightPixels_SSE2(const uint8_t* const src, __m128i* out) { static void LoadEightPixels_SSE2(const uint8_t* const src, __m128i* out) {
const __m128i zero = _mm_setzero_si128(); const __m128i zero = _mm_setzero_si128();
const __m128i A = _mm_loadl_epi64((const __m128i*)(src)); // ABCDEFGH const __m128i A = _mm_loadl_epi64((const __m128i*)(src)); // ABCDEFGH
*out = _mm_unpacklo_epi8(A, zero); *out = _mm_unpacklo_epi8(A, zero);
@ -50,13 +50,15 @@ static void RescalerImportRowExpand_SSE2(WebPRescaler* const wrk,
int accum = x_add; int accum = x_add;
__m128i cur_pixels; __m128i cur_pixels;
assert(!WebPRescalerInputDone(wrk)); // SSE2 implementation only works with 16b signed arithmetic at max.
assert(wrk->x_expand); if (wrk->src_width < 8 || accum >= (1 << 15)) {
if (wrk->num_channels == 4) {
if (wrk->src_width < 2) {
WebPRescalerImportRowExpand_C(wrk, src); WebPRescalerImportRowExpand_C(wrk, src);
return; return;
} }
assert(!WebPRescalerInputDone(wrk));
assert(wrk->x_expand);
if (wrk->num_channels == 4) {
LoadTwoPixels_SSE2(src, &cur_pixels); LoadTwoPixels_SSE2(src, &cur_pixels);
src += 4; src += 4;
while (1) { while (1) {
@ -75,11 +77,7 @@ static void RescalerImportRowExpand_SSE2(WebPRescaler* const wrk,
} else { } else {
int left; int left;
const uint8_t* const src_limit = src + wrk->src_width - 8; const uint8_t* const src_limit = src + wrk->src_width - 8;
if (wrk->src_width < 8) { LoadEightPixels_SSE2(src, &cur_pixels);
WebPRescalerImportRowExpand_C(wrk, src);
return;
}
LoadHeightPixels_SSE2(src, &cur_pixels);
src += 7; src += 7;
left = 7; left = 7;
while (1) { while (1) {
@ -94,7 +92,7 @@ static void RescalerImportRowExpand_SSE2(WebPRescaler* const wrk,
if (--left) { if (--left) {
cur_pixels = _mm_srli_si128(cur_pixels, 2); cur_pixels = _mm_srli_si128(cur_pixels, 2);
} else if (src <= src_limit) { } else if (src <= src_limit) {
LoadHeightPixels_SSE2(src, &cur_pixels); LoadEightPixels_SSE2(src, &cur_pixels);
src += 7; src += 7;
left = 7; left = 7;
} else { // tail } else { // tail