Integral projection crashed when external resize is used

[GoogleCodeExporter]

Integral projection crashed when external resize is used. The integral 
projection code seems to assume scale factors are 1. 

Original issue reported on code.google.com by aconve...@google.com on 5 May 2015 at 4:49

[GoogleCodeExporter]

Original comment by ya...@google.com on 7 May 2015 at 10:14

• Changed state: Assigned

[GoogleCodeExporter]

Is there anything test vector that reproduces the issue?

Original comment by jingn...@google.com on 7 May 2015 at 10:26

[GoogleCodeExporter]

Apply the attached patch then run:
./test_libvpx --gtest_filter=VP9/ResizeTest.TestExternalResizeWorks/0 
--single-process-tests

Starting program: /usr/local/google/home/aconverse/src/vpx/build/test_libvpx 
--gtest_filter=VP9/ResizeTest.TestExternalResizeWorks/0 --single-process-tests
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/usr/grte/v4/lib64/libthread_db.so.1".
Loading v17 libstdc++ pretty-printers.
Note: Google Test filter = VP9/ResizeTest.TestExternalResizeWorks/0-:AVX2/*
[==========] Running 1 test from 1 test case.
[----------] Global test environment set-up.
[----------] 1 test from VP9/ResizeTest
[ RUN      ] VP9/ResizeTest.TestExternalResizeWorks/0
[New Thread 0x7ffff67d3700 (LWP 1198)]

Program received signal SIGSEGV, Segmentation fault.
vp9_int_pro_col_sse2 (ref=0x7ffff67e10d2 "", width=width@entry=64)
    at ../libvpx/vp9/encoder/x86/vp9_avg_intrin_sse2.c:362
362   __m128i s0 = _mm_sad_epu8(src_line, zero);
(gdb) bt
#0  vp9_int_pro_col_sse2 (ref=0x7ffff67e10d2 "", width=width@entry=64)
    at ../libvpx/vp9/encoder/x86/vp9_avg_intrin_sse2.c:362
#1  0x00000000006247c9 in vp9_int_pro_motion_estimation (
    cpi=cpi@entry=0x7ffff7f0c020, x=x@entry=0x7ffff7f18020, 
    bsize=bsize@entry=BLOCK_64X64) at ../libvpx/vp9/encoder/vp9_mcomp.c:1826
#2  0x00000000006e31c6 in choose_partitioning (cpi=cpi@entry=0x7ffff7f0c020, 
    tile=tile@entry=0x11ab0d0, x=x@entry=0x7ffff7f18020, 
    mi_row=mi_row@entry=0, mi_col=mi_col@entry=8)
    at ../libvpx/vp9/encoder/vp9_encodeframe.c:726
#3  0x00000000006e5f49 in encode_nonrd_sb_row (tp=0x7fffffffcf00, 
    mi_row=<optimized out>, tile_data=<optimized out>, td=0x7ffff7f18020, 
    cpi=0x7ffff7f0c020) at ../libvpx/vp9/encoder/vp9_encodeframe.c:3633
#4  vp9_encode_tile (cpi=cpi@entry=0x7ffff7f0c020, td=td@entry=0x7ffff7f18020, 
    tile_row=tile_row@entry=0, tile_col=tile_col@entry=0)
    at ../libvpx/vp9/encoder/vp9_encodeframe.c:3824
#5  0x00000000006e7583 in encode_tiles (cpi=0x7ffff7f0c020)
    at ../libvpx/vp9/encoder/vp9_encodeframe.c:3844
#6  encode_frame_internal (cpi=cpi@entry=0x7ffff7f0c020)
    at ../libvpx/vp9/encoder/vp9_encodeframe.c:3958
#7  0x00000000006e7a06 in vp9_encode_frame (cpi=cpi@entry=0x7ffff7f0c020)
    at ../libvpx/vp9/encoder/vp9_encodeframe.c:4119
#8  0x000000000062d251 in encode_without_recode_loop (cpi=0x7ffff7f0c020)
    at ../libvpx/vp9/encoder/vp9_encoder.c:3046

Original comment by aconve...@google.com on 12 May 2015 at 5:33

Attachments:

• resize-at-speed5.diff

[GoogleCodeExporter]

https://chromium-review.googlesource.com/#/c/272921/

Original comment by jingn...@google.com on 22 May 2015 at 7:07

• Changed state: Started

[GoogleCodeExporter]

Original comment by jingn...@google.com on 28 May 2015 at 6:40

• Changed state: Fixed