similar to: [LLVMdev] LIT Any/All options

Hi Linfeng, On 06/02/17 02:51 PM, Linfeng Zhang wrote: > However, the critical thing is that all the states in each stage when > processing input[i] are reused by the next input[i+1]. That is > input[i+1] must wait input[i] for 1 stage, and input[i+2] must wait > input[i+1] for 1 stage, etc. That is indeed the tricky part... and the one I think you could do slightly differently. If

Hi Linfeng, On 06/02/17 07:18 PM, Linfeng Zhang wrote: > This is a great idea. But the order (psEncC->shapingLPCOrder) can be > configured to 12, 14, 16, 20 and 24 according to complexity parameter. > > It's hard to get a universal function to handle all these orders > efficiently. Any suggestions? I can think of two ways of handling larger orders. The obvious one is

Hi Jean-Marc, Attached is the silk_warped_autocorrelation_FIX_neon() which implements your idea. Speed improvement vs the previous optimization: Complexity 0-4: Doesn't call this function. Complexity 5: 2.1% (order = 16) Complexity 6: 1.0% (order = 20) Complexity 8: 0.1% (order = 24) Complexity 10: 0.1% (order = 24) Code size of silk_warped_autocorrelation_FIX_neon() changes from 2,644

I was running a sampling syntax based on a data frame (ago) of 160 rows and 25 columns. Below are the column names: > names(ago) [1] "SubID" "AGR1" "AGR2" "AGR3" "AGR4" "AGR5" "AGR6" "AGR7" "AGR8" [10] "AGR9" "AGR10" "WAGR1" "WAGR2"

I have a code generation question for ARM with VFP and NEON. I am generating code for the following function as a test: void FloatingPointTest(float f1, float f2, float f3) { float f4 = f1 * f2; if (f4 > f3) printf("%f\n",f2); else printf("%f\n",f3); } I have tried compiling with: 1. -mfloat-abi=softfp and -mfpu=neon 2.

Hi Linfeng, Thanks for the updated patch. I'll have a look and get back to you. When you report speedup percentages, is that relative to the entire encoder or relative to just that function in C? Also, what's the speedup compared to master? Cheers, Jean-Marc On 05/04/17 12:14 PM, Linfeng Zhang wrote: > I attached a new patch with small cleanup (disassembly is identical as > the

Hi Linfeng, I had a closer look at your patch and the code looks good -- and slightly simpler than I had anticipated, so that's good. I did some profiling on a Cortex A57 and I've been seeing slightly less improvement than you're reporting, more like 3.5% at complexity 8. It appears that the warped autocorrelation function itself is only faster by a factor of about 1.35. That's a

Sorry to bother you kind folks again with my questions. I am trying to learn as much as I can about all this, and I will admit that I don't have the proper background, but I hope that someone can at least point me in the correct direction. I have created a test matrix for what I want to do: s1 s2 s3 s4 s5 s1 10 5 0 8 7 s2 5 10 0 0 5 s3 0 0 10 0 0 s4 8 0 0 10 0 s5 7

This is a great idea. But the order (psEncC->shapingLPCOrder) can be configured to 12, 14, 16, 20 and 24 according to complexity parameter. It's hard to get a universal function to handle all these orders efficiently. Any suggestions? Thanks, Linfeng On Mon, Feb 6, 2017 at 12:40 PM, Jean-Marc Valin <jmvalin at jmvalin.ca> wrote: > Hi Linfeng, > > On 06/02/17 02:51 PM,

Hi everybody, consider please an optimization problem: minimize sum S1+S2 Subject to : y - x =< A + S1 x - y =< A + S2 and we want to add two more constraints: y - x =< B - S3 x - y =< B - S4 where A is a small constant value and B is a large constant value, S1 and S2 are surplus and S3

Install the gmp package, run your code, and then try this: bu <- gmp::as.bigq(u) bs4 <- bu[1] + bu[2] + bu[3] + bu[4] + bu[5] s4 <- as.double(bs4) s1 - s4 ## [1] 0 s2[[2]] - s4 ## [1] 7.105427e-15 s3 - s4 ## [1] 7.105427e-15 identical(s1, s4) ## [1] TRUE `bs4` is the exact sum of the binary rationals in your `u` vector; `s4` is the closest double precision to this exact sum.

Hi everyone, I have dataset which I take random samples of it 5 times. each time I get the mean for rows for each sample. at the end I need to calculate the Average of all means for each sample and each row. to clear it up I give an example: say this is my dataset. X8 X9X10X12 X13 X14 X15 X16X17X18X19 X20 X21 X22 s1 0 1 0 0 0 1 0 0 0 0 0 0 0 0 s2 0 0 0

I attached a new patch with small cleanup (disassembly is identical as the last patch). We have done the same internal testing as usual. Also, attached 2 failed temporary versions which try to reduce code size (just for code review reference purpose). The new patch of silk_warped_autocorrelation_FIX_neon() has a code size of 3,228 bytes (with gcc). smaller_slower.c has a code size of 2,304

I've looked at it a bit more and an updated ninja is part of it. You also need build.ninja to contain 'pool = console' on the lit rules. I added it manually to try it out but presumably a cmake update is needed to add this automatically. ________________________________ From: Renato Golin [renato.golin at linaro.org] Sent: 23 December 2014 00:55 To: David Blaikie Cc: LLVM Dev; Daniel

Hi Jean-Marc, Thanks a lot for reviewing this huge assembly function! silk_warped_autocorrelation_FIX_c()'s kernel part is for( n = 0; n < length; n++ ) { tmp1_QS = silk_LSHIFT32( (opus_int32)input[ n ], QS ); /* Loop over allpass sections */ for( i = 0; i < order; i++ ) { /* Output of allpass section */ tmp2_QS = silk_SMLAWB(

Hi Jean-Marc, Thanks for your suggestions. Will get back to you once we have some updates. Linfeng On Mon, Feb 6, 2017 at 5:47 PM, Jean-Marc Valin <jmvalin at jmvalin.ca> wrote: > Hi Linfeng, > > On 06/02/17 07:18 PM, Linfeng Zhang wrote: > > This is a great idea. But the order (psEncC->shapingLPCOrder) can be > > configured to 12, 14, 16, 20 and 24 according to

My simple functions were to compare the result with the gfortran compiler sum() function. I thought that the Fortran sum could not be less precise than R. I was wrong. I am impressed. The R sum does in fact match the result if we use the Kahan algorithm. P. I am glad to see that R sum() is more accurate than the gfortran compiler sum. On 16/03/18 11:37 AM, luke-tierney at uiowa.edu wrote:

Dear R users: I have some difficulties analizing data with mixed effects NLME and the last version of R. More concretely, I have a repeated measures design with a single group and 2 experimental factors (say A and B) and my interest is to compare additive and nonadditive models. suj rv A B 1 s1 4 a1 b1 2 s1 5 a1 b2 3 s1 7 a1 b3 4 s1 1 a2

Hello, The example shows a sequence of songtypes a bird has sang. The entire list contains several thousand songs which the bird has produced consecutively. Is there any convenient way in R to produce a contingency table which shows how often a special song type was sang after a special other songtype? Or within a determined frame of e.g. 10 consecutively produced songs? Example: Nr Songtype 1

I have coded a time series from simulated data: simtimeseries <- arima.sim(n=1024,list(order=c(4,0,0),ar=c(2.7607, -3.8106, 2.6535, -0.9258),sd=sqrt(1))) #show roots are outside unit circle plot.ts(simtimeseries, xlab="", ylab="", main="Time Series of Simulated Data") # Yule ---------------------------------------------------------------------------- q1 <-