search

dinhminhquoi / webm

Automatically exported from code.google.com/p/webm
0 stars 0 forks source link

[NEON] fdct4x4_test failure #715

Closed GoogleCodeExporter closed 9 years ago

GoogleCodeExporter commented 9 years ago 
v1.3.0-1442-g0eca2cd

Trans4x4HT / vp9_iht4x4_16_add_neon

Tests are available:
https://gerrit.chromium.org/gerrit/#/c/68969/

---

[ RUN      ] DISABLED_NEON/Trans4x4HT.AccuracyCheck/0
test/fdct4x4_test.cc:91: Failure
Expected: (1u) >= (max_error), actual: 1 vs 5625
Error: 4x4 FHT/IHT has an individual round trip error > 1
test/fdct4x4_test.cc:94: Failure
Expected: (count_test_block) >= (total_error), actual: 10000 vs 44593958
Error: 4x4 FHT/IHT has average round trip error > 1 per block
...
[ RUN      ] DISABLED_NEON/Trans4x4HT.InvAccuracyCheck/0
test/fdct4x4_test.cc:176: Failure
Expected: (1u) >= (error), actual: 1 vs 4
Error: 16x16 IDCT has error 4 at index 0
test/fdct4x4_test.cc:176: Failure
Expected: (1u) >= (error), actual: 1 vs 49
Error: 16x16 IDCT has error 49 at index 1
test/fdct4x4_test.cc:176: Failure
...

Original issue reported on code.google.com by jz...@google.com on 26 Feb 2014 at 8:14

GoogleCodeExporter commented 9 years ago 
Consider taking this patch if perf is similar:
https://gerrit.chromium.org/gerrit/#/c/68969/

Original comment by johannkoenig@google.com on 27 Feb 2014 at 3:21

GoogleCodeExporter commented 9 years ago 
Sorry, meant this:
https://gerrit.chromium.org/gerrit/#/c/68921

Original comment by johannkoenig@google.com on 27 Feb 2014 at 3:23

GoogleCodeExporter commented 9 years ago 
I could reproduce this issue now with building the test_vpx on Android. 

Will check the performance of that patch tomorrow.

Original comment by hku...@google.com on 27 Feb 2014 at 4:23

GoogleCodeExporter commented 9 years ago 
ping. this fell through the cracks...

Original comment by jz...@google.com on 16 Jul 2014 at 6:48

GoogleCodeExporter commented 9 years ago

Original comment by hku...@google.com on 17 Jul 2014 at 6:33

GoogleCodeExporter commented 9 years ago 
Still failing

Original comment by johannkoenig@google.com on 6 Nov 2014 at 1:21

GoogleCodeExporter commented 9 years ago 
https://gerrit.chromium.org/gerrit/68921

The test exercised tx_type 0 but the code is never called that way in the 
encoder:
// iht
void vp9_iht4x4_add(TX_TYPE tx_type, const tran_low_t *input, uint8_t *dest, 
int stride, int eob) {
  if (tx_type == DCT_DCT)
    vp9_idct4x4_add(input, dest, stride, eob);
  else
    vp9_iht4x4_16_add(input, dest, stride, tx_type);
}

The updated patch contains all the tx_types and calls the C code if the DCT_DCT 
is specified. It would probably be better to fix the test and the reference 
code to indicate this is not an expected code path.

Original comment by johannkoenig@google.com on 15 Dec 2014 at 10:53