Here is my setup:
RedHat Enterprise AS 3
./configure --enable-ssl --with-ssl=/usr --enable-so
configure seems to work, but a make fails with the error:
/usr/bin/ld:strings/apr_cpystrn.lo: file format no recognized: treating as linker script
Has anyone successfully compiled Apache with ICC, and if so, what settings were used?
Is there a way to view xmm regs, or for that matter __m128 variables, as four floating point values? Using a "p $xmm0" just gives me a bunch of hex - not really meaningful, you know?
The Intel 7.1 and 8.0 C/C++ compiler for Linux generate wrong addresses for multi-dimensional variable-length arrays when the -openmp flag is used. The example below illustrates how the first index is ignored in the address computations for a two-dimensional matrix (matrix[i] is always the same, regardless of i, and &matrix[i][i] is computed as &matrix[i]). The address calculations are also wrong if the matrix is declared as formal argument of a function. When -openmp is not specified, the address calculations are fine.
I hope some Intel guru solves this bug.
I'm using intel code coverage tool on linux for collecting coverage for c files.
When i try to merge the *.dyn files using 'profmerge', it fails with error
profmerge: merging dynamic file: 407bbe09_16585.dyn
./407bbe09_16585.dyn: No such file or directory
FATAL ERROR: fread on file ./407bbe09_16585.dyn failed
But the file ./407bbe09_16585.dyn exists.
Is this a known issue. anybody faced same issue.
I'm using ic++ v7.0. Consider the folowing loop (all arrays aligned to 16):
for(int x=0; x<16; x++)
AtanData[x] = atanf(TanData[x]);
The compiler complains "Loop not vectorized due to unvectorizable statement..." for the line containing the atan() call.
I thought the compiler was supposed to automatically use the short vector math library in these cases. ?
(AP_922).A fast precise implementation of 8x8 DCT.pdf
in this application the decompose of the matrix M8 is not right. anybody has found that? and who can correct it? please tell me the result.
how to optimize the program for dct(descrete cosine transform)? i am not very clear about how to achieve that. i have read some application notes found in the web of intel. in the algorithms presented inthose application notes, thay decomposed the tranformation matrix, and why they should do so? and howcan the program for dct benifit from such decompose on the transformation matrix?
thank you. i hope somebody can tell. best wishes.
I'm just starting to program using MMX, but i couldnt find old intel MMX application notes on the website. It seems the link has changed, i tried searching for it but couldnt find it. Can anyone help me what is the link to intel mmx application notes?
I'm trying to find
Color Conversion from YUV12 to RGB Using Intel MMX Technology
Old link, but cannot find it.
Ihave a problem with libimf.Using cpow() to calculate Exp(-I Pi/2) gives +I instead of -I. This is contrary to the behaviour of libm which gives the expected result. I get the same thing whether I use gcc or icc, so I'm pretty sure it's the library at fault, not the compiler. I couldn't find any discussion of this on the intel "software defects" list http://support.intel.com/support/performancetools/c/sb/CS-007699.htm
I have a couple of functions written in inline assembly code. (They're am_exp_ps and am_exp_eps from the approximate math library.) When I call them in a test program, they give correct results. However, when I use them in real code, it seems as though they are destroyingwhatever was in the xmm registers before they were called.
That is, the code flow is like this:
F32vec4 x, y, z;
x = F32vec4 (known values); // which has z in an xmm reg
...do some computations to get x;
y = am_exp_ps (x);
...and find that z has been changed!