[x265] [PATCH] asm: interp_8tap_horiz pp and ps sse2

[dave]

On 04/30/2015 12:43 PM, Steve Borho wrote:
> On 04/30, dave wrote:
>> On 04/30/2015 09:28 AM, Steve Borho wrote:
>>> On 04/30, dave wrote:
>>>> I submitted a patch fixing this but I was only able to test it for 64-bit.
>>>> When trying to build for 32-bit I ran into the following build error.
>>>>
>>>> Linking CXX executable x265
>>>> libx265.so.56: undefined reference to `dlsym'
>>>> libx265.so.56: undefined reference to `dlopen'
>>>> collect2: error: ld returned 1 exit status
>>>> make[2]: *** [x265] Error 1
>>>> make[1]: *** [CMakeFiles/cli.dir/all] Error 2
>>>> make: *** [all] Error 2
>>> do you have a 32bit version of libdl installed?
>>>
>> Yes.
>>
>> ls -l lib32/libdl*
>> -rw-r--r-- 1 root root 9768 Feb 22 03:45 lib32/libdl-2.13.so
>> lrwxrwxrwx 1 root root   13 Feb 22 03:45 lib32/libdl.so.2 -> libdl-2.13.so
> Not sure what's causing this. cmake should be checking for libdl and
> adding that to the library list if it is present. Can you tell if -ldl
> is being passed to the linker?
>
When I toggle cmake to show all options,

LIBDL    LIBDL-NOTFOUND

I see the same for LIBRT and all the NUMA stuff but I am guessing those 
aren't required.

I turned on CMAKE_VERBOSE_MAKEFILE and logged the output, no -ldl or any 
mention of libdl.

I changed LIBDL-NOTFOUND to /lib32/libdl.so.2 and was able to build and 
test.

32-bit..
./test/TestBench --testbench interp | grep luma_h
luma_hpp[  4x4]        2.01x      1867.65       3762.72
luma_hps[  4x4]        1.90x      4550.19       8629.97
luma_hpp[  8x8]        2.38x      6087.71       14482.83
luma_hps[  8x8]        1.95x      11010.22      21510.71
luma_hpp[16x16]        2.71x      23922.61      64745.14
luma_hps[16x16]        2.33x      32951.08      76901.44
luma_hpp[32x32]        2.81x      94440.19      265020.06
luma_hps[32x32]        2.24x      111068.30      248300.09
luma_hpp[64x64]        2.61x      376556.16      983163.88
luma_hps[64x64]        2.19x      402714.62      882300.50
luma_hpp[  8x4]        2.34x      3110.44       7265.88
luma_hps[  8x4]        1.92x      8213.40       15790.23
luma_hpp[  4x8]        2.11x      3627.76       7652.79
luma_hps[  4x8]        1.91x      6084.59       11629.97
luma_hpp[ 16x8]        2.72x      11995.22      32633.96
luma_hps[ 16x8]        2.32x      21585.11      50085.57
luma_hpp[ 8x16]        2.37x      12222.63      29025.33
luma_hps[ 8x16]        1.99x      16649.99      33118.82
luma_hpp[32x16]        2.63x      47409.25      124461.49
luma_hps[32x16]        2.23x      65630.83      146272.73
luma_hpp[16x32]        2.71x      47632.11      128912.72
luma_hps[16x32]        2.34x      55819.18      130416.08
luma_hpp[64x32]        2.60x      189231.47      492174.06
luma_hps[64x32]        2.19x      221716.19      485533.44
luma_hpp[32x64]        2.65x      188249.28      498872.19
luma_hps[32x64]        2.24x      201548.78      451323.59
luma_hpp[16x12]        2.70x      18042.62      48710.91
luma_hps[16x12]        2.35x      27225.15      63847.45
luma_hpp[12x16]        2.31x      18982.63      43757.87
luma_hps[12x16]        1.92x      25584.95      49156.58
luma_hpp[ 16x4]        2.87x      6058.06       17385.17
luma_hps[ 16x4]        2.31x      15944.99      36844.97
luma_hpp[ 4x16]        2.14x      7312.49       15642.68
luma_hps[ 4x16]        1.96x      8990.00       17631.11
luma_hpp[32x24]        2.63x      70876.62      186707.23
luma_hps[32x24]        2.23x      88340.04      196993.19
luma_hpp[24x32]        2.66x      71127.98      189195.44
luma_hps[24x32]        2.26x      83403.52      188429.67
luma_hpp[ 32x8]        2.63x      23754.62      62378.40
luma_hps[ 32x8]        2.23x      42879.46      95483.74
luma_hpp[ 8x32]        2.44x      23913.38      58450.92
luma_hps[ 8x32]        1.98x      28051.65      55436.18
luma_hpp[64x48]        2.61x      282386.12      736332.88
luma_hps[64x48]        2.19x      312061.47      683854.44
luma_hpp[48x64]        2.62x      282449.75      741238.62
luma_hps[48x64]        2.21x      302046.09      667564.06
luma_hpp[64x16]        2.60x      94443.56      245561.27
luma_hps[64x16]        2.19x      130718.59      285776.12
luma_hpp[16x64]        2.73x      94895.83      258934.14
luma_hps[16x64]        2.34x      101004.30      236416.59


thanks on the libdl tip.