search for: rerol

On 3 June 2015 at 19:57, Benjamin Kramer <benny.kra at gmail.com> wrote: > There's a loop reroll pass in LLVM trunk that should do exactly this transformation. Though that's a loop pass (runOnLoop). What you could do is add a previous pass that would recognize the pattern and create a loop of 1 iteration around the code and then run the reroll pass. If your pattern recognition is too go...

Hi all I am working on k-means algorithm (in R: kmeans( ) ). The R-help advice us to try several random start in order to avoid local minimum. Does one know if there is a procedure that automaticly run this rerolling and select the best partition ? Or any studies that gives clues on the number of rerolling ? Thanks for helping. Christophe

...would work, Hal's patch would re-roll, and if I base my analysis on what was there before, I won't detect it. >> > You would detect it doing interleaved data in the vectorizer and generate exactly the code that GCC does. > > Which granted will be less efficient than if we reroll and vectorize in which case we should see a speed up of 8 (8 x i8). > > I have not looked at whether Hal's code could be improved to handle your uncanonical example. > >> Since Hal hinted on leaving that as an API for the vectorizer, I thought that maybe it'd be worth runn...

On Tue, Feb 17, 2004 at 02:20:46PM +0100, Michael Nottebrock wrote: [distfile rerolls] > I didn't know that I was supposed to perform a security audit and I did not do > so. So if anyone happens to have the old distfile still around, please send > it my way, cause I don't. I suggest next time instead of marking a port as > BROKEN= Checksum mismatch, mark it...

Hey guys, in an HPC project I am working on I am given an LLVM program consisting of a linear sequence of repetitive junks of code with an uniform memory access pattern. Each code junk does the following: 1) loads some memory, 2) performs some arithmetic operations, 3) stores the result back to memory. The memory stride between consecutive junks is constant over the whole program, thus the

...the quota error. But a *move* just >>> happily piles on to the overquota namespace. Is that normal? >> >> Probably needs a bit more thinking, but I guess the attached patch would >> help. >> >> > I appreciate the patch! Esp on a Weds night. I applied and rerolled > dovecot, but I can still move messages into the over-quota namespace. How about this updated patch? > Out of curiosity, in the Quota wiki page, it mentions that 'in theory there > could be e.g. "user quota" and "domain quota" roots'. That's also super...

On 23 Nov 2016, at 0.49, Mark Moseley <moseleymark at gmail.com> wrote: > > If I move messages between namespaces, it appears to ignore the quotas I've > set on them. A *copy* will trigger the quota error. But a *move* just > happily piles on to the overquota namespace. Is that normal? Probably needs a bit more thinking, but I guess the attached patch would help.

...ed loop code takes longer than the non-unrolled loop). There might > be a clean way of doing this; I'll think about it. I don't really understand the issue here, can you elaborate on when this might be a win? I really don't like "speculatively unroll, try to do something, then reroll". That is terrible for compile time and just strikes me as poor design :-) -Chris

...kes longer than the non-unrolled loop). There might > > be a clean way of doing this; I'll think about it. > > I don't really understand the issue here, can you elaborate on when this might be a win? I really don't like "speculatively unroll, try to do something, then reroll". That is terrible for compile time and just strikes me as poor design :-) >From Ayal's e-mail, it seems that the gcc vectorizer contains specialized unrolling code to handle these kinds of cases. With appropriate refactoring, perhaps that is the best solution. -Hal > > -Ch...

On Mon, 2012-02-06 at 14:26 -0800, Chris Lattner wrote: > On Feb 2, 2012, at 7:56 PM, Hal Finkel wrote: > > As some of you may know, I committed my basic-block autovectorization > > pass a few days ago. I encourage anyone interested to try it out (pass > > -vectorize to opt or -mllvm -vectorize to clang) and provide feedback. > > Especially in combination with

...ed for int-to-fp conversion at -O0 r203054 - The PPC global base register cannot be r0 r199832 - Fix pr18515. r200288 - Handle spilling the PPC GPRC_NOR0 register class r199763 - Fix pointer info on PPC byval stores r202192 - Account for 128-bit integer operations in PPCCTRLoops r198425 - Fix loop rerolling pass failure with non-consant loop lower bound I apologize the delay; I've not had a chance to refine my list until this morning. Thanks again, Hal ----- Original Message ----- > From: "Hal Finkel" <hfinkel at anl.gov> > To: "Tom Stellard" <tom at stell...

...stead of O(log N), where N is alphabet size), if the huffman tree would be left (or right; I prefer left) aligned. Ask me for pointers if you need convincing. The code I submit now could be more clean, I know. Seems to work though. Weird. unrolled && rerolled the decodevs loops. envelope.c if (fabs(...) < min) creates horrible assembler (gcc 2.95, x86), so changed to if (... < min && ... > -min). muchos faster. lsp.c put the fromdB() before linearmap cales.h todB_nn() for non-negative values. fabs(...

On Thu, Dec 17, 2015 at 9:21 AM, Andy Ayers <andya at microsoft.com> wrote: > While your bb count distribution is extremely likely to be some kind of power-law like distribution, it's not guaranteed. > > Also you might think about operations that can amplify (rerolling) or appear to amplify (TRE) or diminish BB counts, and how you'd go about reclassifying block hotness. yes -- this is more of a problem for Sample based FDO because the samples are collected after such transformations. We have a paper discussing that (to appear cgo2016). thanks, Dvaid &...

...on them. A *copy* will trigger the quota error. But a *move* just > > happily piles on to the overquota namespace. Is that normal? > > Probably needs a bit more thinking, but I guess the attached patch would > help. > > I appreciate the patch! Esp on a Weds night. I applied and rerolled dovecot, but I can still move messages into the over-quota namespace. I threw some i_debug's into quota_roots_equal() (and one right at the top), but I don't ever see them in the debug logs. But both "ctx->moving" and "src_box == NULL" are true, so it never calls...

...st > >>> happily piles on to the overquota namespace. Is that normal? > >> > >> Probably needs a bit more thinking, but I guess the attached patch would > >> help. > >> > >> > > I appreciate the patch! Esp on a Weds night. I applied and rerolled > > dovecot, but I can still move messages into the over-quota namespace. > > How about this updated patch? > > Nope, still lets me move messages into the over-quota namespace. Both these are true in quota_check: ctx->moving quota_move_requires_check > > Out of c...

I saw that there is a local root exploit in the wild. http://blog.kagesenshi.org/2008/02/local-root-exploit-on-wild.html And I see my centos box still has: 2.6.18-53.1.4.el5 yum says there are no updates... am I safe? Valent.

...s on to the overquota namespace. Is that normal? >> >> >> >> Probably needs a bit more thinking, but I guess the attached patch >> would >> >> help. >> >> >> >> >> > I appreciate the patch! Esp on a Weds night. I applied and rerolled >> > dovecot, but I can still move messages into the over-quota namespace. >> >> How about this updated patch? >> >> > Nope, still lets me move messages into the over-quota namespace. > > Both these are true in quota_check: > > ctx->moving >...

On Wed, Dec 16, 2015 at 6:00 PM, Andy Ayers <andya at microsoft.com> wrote: > I’ve done similar rankings with profile data and found it worked out pretty > well. > > > > In my case it was ranking function hotness but the idea was similar: sort by > weight, compute various percentile cutoff points, and use those to classify. > I put in some compensation for truly

...kes longer than the non-unrolled loop). There might > > be a clean way of doing this; I'll think about it. > > I don't really understand the issue here, can you elaborate on when this might be a win? I really don't like "speculatively unroll, try to do something, then reroll". That is terrible for compile time and just strikes me as poor design :-) > > -Chris > -- Hal Finkel Postdoctoral Appointee Leadership Computing Facility Argonne National Laboratory

So I'm running nfs to get content to my web servers. Now I've had this problem 2 times (about 2 weeks since the last occurrence). I use drbd on the nfs server for redundancy. Now to my problem: All my web sites stopped responding so I started by checking dmesg and there I found a bunch of this errors || Aug 11 16:00:39 web03 kernel: lockd: server 192.168.20.22 not responding, timed out