Tan1 again

jorge · 1 Jun 2005

unkargherth wrote: So tan1 gets calculated only for first of the 4 SSE rows, true?

Yeah, you're right. It's a bug and very easy to fix. Thanks for reporting it. If it's urgent I'll try to have it fixed for the next release.

unkargherth · 1 Jun 2005

jorge wrote: unkargherth wrote: So tan1 gets calculated only for first of the 4 SSE rows, true?

Yeah, you're right. It's a bug and very easy to fix. Thanks for reporting it. If it's urgent I'll try to have it fixed for the next release.

Well, it's relatively urgent. I'm locked on some filters by this, but ther are enough filters now, so it's urgency depends n how much cost to correct. Decide yourself jorge. I Have no idea on your priporities and

Anyway , IMHO, this should be solved sometime in near future

unkargherth · 2 Jun 2005

Well it seems dificult to tame this baby..

Now ( 0.9.3u) it works for the 4 SSE rows ( Thanks jorge !!!)

But i found another minibug on it 🙁

this

polyin i; polyout o; float t; t=0.5*i; o=tan1(t);

works perfectly, but

polyin i; polyout o; o=tan1(i*0.5);

fails miserably

I know that's reallya "not so urgent" bug, because there is a workaround I'm only notifying it

AS usual, a new guiness record. check the date of first post on this an the date of this to know the time for bug correct6ion.... Awesome Jorge

jorge · 2 Jun 2005

unkargherth wrote: I know that's reallya "not so urgent" bug, because there is a workaround I'm only notifying it

Thanks unkargherth, those bugs are really easy to fix, it's more a question of finding and reporting them.

I've uploaded a new version that should fix the tan1(i*0.5) bug, let me know if it works. The link is the same as before: [old link]

unkargherth · 2 Jun 2005

Re-downloaded and nothing changed..... 🙁.At least for me

Anyway, bad idea to track bugs based on only one person view

Anybody could confirm or rebate?

jorge · 2 Jun 2005

unkargherth, out=tan1(in*0.5) should generate the following assembler code ( unless there's something I'm missing ):

polyin in;polyout out;float temp=0.0; float F0P500000=0.500000; float F65536P000000=65536.000000; float F16384P000000=16384.000000; movaps xmm0,in; mulps xmm0,F0P500000; movaps xmm1,xmm0; mulps xmm1,F65536P000000; addps xmm1,F16384P000000; movaps temp,xmm1; fld temp[0]; fistp temp[0]; mov eax,temp[0]; and eax,65535; shl eax,2; add eax,sine; fld [eax]; fstp temp[0]; fld temp[1]; fistp temp[1]; mov eax,temp[1]; and eax,65535; shl eax,2; add eax,sine; fld [eax]; fstp temp[1]; fld temp[2]; fistp temp[2]; mov eax,temp[2]; and eax,65535; shl eax,2; add eax,sine; fld [eax]; fstp temp[2]; fld temp[3]; fistp temp[3]; mov eax,temp[3]; and eax,65535; shl eax,2; add eax,sine; fld [eax]; fstp temp[3]; movaps xmm1,temp; movaps xmm2,xmm0; mulps xmm2,F65536P000000; movaps temp,xmm2; fld temp[0]; fistp temp[0]; mov eax,temp[0]; and eax,65535; shl eax,2; add eax,sine; fld [eax]; fstp temp[0]; fld temp[1]; fistp temp[1]; mov eax,temp[1]; and eax,65535; shl eax,2; add eax,sine; fld [eax]; fstp temp[1]; fld temp[2]; fistp temp[2]; mov eax,temp[2]; and eax,65535; shl eax,2; add eax,sine; fld [eax]; fstp temp[2]; fld temp[3]; fistp temp[3]; mov eax,temp[3]; and eax,65535; shl eax,2; add eax,sine; fld [eax]; fstp temp[3]; divps temp,xmm1; movaps xmm0,temp; //Assignment> sLeft=xmm0 movaps out,xmm0;

which is what's happening here.

unkargherth · 2 Jun 2005

yes it does, but....

take the code module. Feed it with a sine osc out with freq 0.01 and try to draw the ouput from code module. Nothing

But if you do the same using the intermediate var variant, it draws nicely

So i suspect that something stange is happening