Subtile decoding: memory use reduction and perf improvements#1010

Merged

rouault merged 29 commits intouclouvain:masterfrom rouault:subtile_decoding_stage3

Sep 5, 2017

+2,305-776

Commits on Sep 1, 2017

opj_t1_allocate_buffers(): remove useless overflow checks
rouault
committed
opj_j2k_setup_encoder(): validate code block width/height
rouault
committed
opj_compress: reorder checks related to code block dimensions, to avoid potential int overflow
rouault
committed
Sub-tile decoding: only allocate tile component buffer of the needed dimension
rouault
committed
opj_dwt_decode_partial_tile(): avoid undefined behaviour in lifting operation by properly initializing working buffer
rouault
committed
opj_tcd_init_tile(): fix typo on overflow detection condition (introduced in previous commit)
rouault
committed
Remove limitation that prevents from opening images bigger than 4 billion pixels
rouault
committed
Fix compiler warning in release mode
rouault
committed
TCD: allow tile buffer to be greater than 4GB on 64 bit hosts (but number of pixels must remain under 4 billion)
rouault
committed
Various changes to allow tile buffers of more than 4giga pixels
rouault
committed
opj_j2k_decode_tiles(): apply whole single tile image decoding optimization to reading at reduced resolution as well
rouault
committed
Allow several repeated calls to opj_set_decode_area() and opj_decode() for single-tiled images
rouault
committed
Optimize reading/write into sparse array
rouault
committed
opj_j2k_update_image_data(): avoid zero-ing the buffer if not needed
rouault
committed
Optimize opj_dwt_decode_partial_1() when cas == 0
rouault
committed
test_decode_area: accept user bounds in -strip_height mode
rouault
committed
Tiny perf improvement in T1 stage for subtile decoding
rouault
committed
Sub-tile decoding: speed up vertical pass in IDWT5x3 by processing 4 cols at a time
rouault
committed
opj_dwt_decode_partial_1_parallel(): add SSE2 optimization
rouault
committed
opj_tcd_dc_level_shift_decode(): optimize lossy case
rouault
committed
opj_dwt_decode_partial_97(): simplify/more efficient use of sparse arrays in vertical pass
rouault
committed
opj_v4dwt_decode_step2_sse(): loop unroll
rouault
committed
sparse_array: optimizations for lossy case
rouault
committed
opj_t1_decode_cblk(): move some code to codeblock processor for (theoretical) better multi-threading in subtile decoding
rouault
committed
Replace error message 'Not enough memory for tile data' by 'Size of tile data exceeds system limits' (refs uclouvain#730 (comment))
rouault
committed
opj_t1_clbl_decode_processor(): use SSE2 in subtile decoding code path, for irreversible
rouault
committed
opj_j2k_update_image_data(): avoid allocating image buffer if we can just reuse the tile buffer one
rouault
committed
opj_v4dwt_decode_step1_sse(): rework a bit to improve code generation
rouault
committed

Commits on Sep 4, 2017

Replace uses of size_t by OPJ_SIZE_T
rouault
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Subtile decoding: memory use reduction and perf improvements#1010

Commits on Sep 1, 2017

opj_t1_allocate_buffers(): remove useless overflow checks

opj_j2k_setup_encoder(): validate code block width/height

opj_compress: reorder checks related to code block dimensions, to avoid potential int overflow

Sub-tile decoding: only allocate tile component buffer of the needed dimension

opj_dwt_decode_partial_tile(): avoid undefined behaviour in lifting operation by properly initializing working buffer

opj_tcd_init_tile(): fix typo on overflow detection condition (introduced in previous commit)

Remove limitation that prevents from opening images bigger than 4 billion pixels

Fix compiler warning in release mode

TCD: allow tile buffer to be greater than 4GB on 64 bit hosts (but number of pixels must remain under 4 billion)

Various changes to allow tile buffers of more than 4giga pixels

opj_j2k_decode_tiles(): apply whole single tile image decoding optimization to reading at reduced resolution as well

Allow several repeated calls to opj_set_decode_area() and opj_decode() for single-tiled images

Optimize reading/write into sparse array

opj_j2k_update_image_data(): avoid zero-ing the buffer if not needed

Optimize opj_dwt_decode_partial_1() when cas == 0

test_decode_area: accept user bounds in -strip_height mode

Tiny perf improvement in T1 stage for subtile decoding

Sub-tile decoding: speed up vertical pass in IDWT5x3 by processing 4 cols at a time

opj_dwt_decode_partial_1_parallel(): add SSE2 optimization

opj_tcd_dc_level_shift_decode(): optimize lossy case

opj_dwt_decode_partial_97(): simplify/more efficient use of sparse arrays in vertical pass

opj_v4dwt_decode_step2_sse(): loop unroll

sparse_array: optimizations for lossy case

opj_t1_decode_cblk(): move some code to codeblock processor for (theoretical) better multi-threading in subtile decoding

Replace error message 'Not enough memory for tile data' by 'Size of tile data exceeds system limits' (refs uclouvain#730 (comment))

opj_t1_clbl_decode_processor(): use SSE2 in subtile decoding code path, for irreversible

opj_j2k_update_image_data(): avoid allocating image buffer if we can just reuse the tile buffer one

opj_v4dwt_decode_step1_sse(): rework a bit to improve code generation

Commits on Sep 4, 2017

Replace uses of size_t by OPJ_SIZE_T