<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
</head>
<body text="#000000" bgcolor="#FFFFFF">
Hi Jyotirmoy Bhattacharya,<br>
<br>
I've vectorized the outer loop of your example with RV, the Region
Vectorizer (<a class="moz-txt-link-freetext" href="https://github.com/cdl-saarland/rv">https://github.com/cdl-saarland/rv</a>). I've attached the
full IR. This is the code i got for the inner most loop:<br>
<br>
<style type="text/css">pre.cjk { font-family: "NSimSun",monospace; }p { margin-bottom: 0.1in; line-height: 120%; }</style>
<pre class="western">for.body4.rv: ; preds = %for.body4.rv, %for.body4.lr.ph.rv
%indvars.iv14 = phi i64 [ %0, %for.body4.lr.ph.rv ], [ %indvars.iv.next21, %for.body4.rv ]
%u1.03315 = phi <4 x double> [ zeroinitializer, %for.body4.lr.ph.rv ], [ %u0.03216, %for.body4.rv ]
%u0.03216 = phi <4 x double> [ zeroinitializer, %for.body4.lr.ph.rv ], [ %add_SIMD, %for.body4.rv ]
%mul5_SIMD = fmul <4 x double> %mul_SIMD, %u0.03216
%sub_SIMD = fsub <4 x double> %mul5_SIMD, %u1.03315
%arrayidx717 = getelementptr inbounds double, double* %coeffs, i64 %indvars.iv14
%scal_load18 = load double, double* %arrayidx717, align 8
%.splatinsert19 = insertelement <4 x double> undef, double %scal_load18, i32 0
%.splat20 = shufflevector <4 x double> %.splatinsert19, <4 x double> undef, <4 x i32> zeroinitializer
%add_SIMD = fadd <4 x double> %sub_SIMD, %.splat20
%indvars.iv.next21 = add nsw i64 %indvars.iv14, -1
%cmp222 = icmp sgt i64 %indvars.iv14, 0
br i1 %cmp222, label %for.body4.rv, label %for.cond.cleanup3.loopexit.rv
}</pre>
<br>
To reproduce this get the release_38 branch of RV from github and do
as follows:<br>
<br>
File cheby.c:<br>
<tt>void cheby_eval(double *coeffs,int n,double *xs,double *ys,int
m)</tt><br>
<tt>{
</tt>
<pre wrap=""> #pragma omp simd
for (int i=0;i<m;i++){
double x = xs[i];
double u0=0,u1=0,u2=0;
for (int k=n;k>=0;k--){
u2 = u1;
u1 = u0;
u0 = 2*x*u1-u2+coeffs[k];
}
ys[i] = 0.5*(coeffs[0]+u0-u2);
}
}
<EOF>
</pre>
1. Compile to IR w/o any of LLVM's vectorizers:<br>
<tt>clang -O3 -fno-vectorize -fno-slp-vectorize cheby.c -c
-emit-llvm -S -o cheby.ll</tt><br>
<br>
2. Run the IR through RV's cmd line vectorizer<br>
<tt>./bin/rvTool -loopvec -w 4 -i cheby.ll -k cheby_eval -o
cheby.rv.ll</tt><br>
<br>
I'd like to add your code to our test suite on github if that is ok
with you.<br>
<br>
Please get in touch with me if you have any other outer loops that
could/should be vectorized.<br>
<br>
Regards,<br>
Simon<br>
<br>
<div class="moz-cite-prefix">On 05/10/2017 04:09 PM, via llvm-dev
wrote:
</div>
<blockquote type="cite"
cite="mid:mailman.25285.1494425357.1282.llvm-dev@lists.llvm.org">
<pre wrap="">
I have the following C++ code that evaluates a Chebyshev polynomial using
Clenshaw's algorithm
void cheby_eval(double *coeffs,int n,double *xs,double *ys,int m)
{
#pragma omp simd
for (int i=0;i<m;i++){
double x = xs[i];
double u0=0,u1=0,u2=0;
for (int k=n;k>=0;k--){
u2 = u1;
u1 = u0;
u0 = 2*x*u1-u2+coeffs[k];
}
ys[i] = 0.5*(coeffs[0]+u0-u2);
}
}
I'm hoping for an autovectorization of the outer loop so that the inner
loop operates on vectors.
When compiled with
clang++ -O3 -march=haswell -Rpass-analysis=loop-vectorize -S chebyshev.cc
using clang++ 3.8.1-23, no vectorization happens and I get the message
chebyshev.cc:19:18: remark: loop not vectorized: cannot identify array
bounds
[-Rpass-analysis=loop-vectorize]
ys[i] = 0.5*(coeffs[0]+u0-u2);
^
chebyshev.cc:21:1: remark: loop not vectorized: value that could not be
identified as reduction is used outside the loop
[-Rpass-analysis=loop-vectorize]
On the same code icc vectorizes the outer loop as expected.
I was wondering if there are small ways in which I can change my code to
help LLVM's autovectorizer to succeed. I would also appreciate any pointers
to documentation or LLVM source that can help me better understand how
autovectorization of outer loops works.
Regards,
Jyotirmoy Bhattacharya
PS. The interesting part of icc's assembler output is
..B1.4: # Preds ..B1.8 ..B1.3
xorl %r15d, %r15d #14.5
xorl %ebx, %ebx #14.21
testq %rsi, %rsi #14.21
vmovupd (%rdx,%r9,8), %ymm3 #12.16
vxorpd %ymm5, %ymm5, %ymm5 #13.14
vmovdqa %ymm1, %ymm4 #13.19
vmovdqa %ymm1, %ymm2 #13.24
jl ..B1.8 # Prob 2% #14.21
..B1.5: # Preds ..B1.4
vaddpd %ymm3, %ymm3, %ymm3 #17.14
..B1.6: # Preds ..B1.6 ..B1.5
vmovapd %ymm4, %ymm2 #20.3
incq %r15 #14.5
vmovapd %ymm5, %ymm4 #20.3
vfmsub213pd %ymm2, %ymm3, %ymm5 #17.19
vbroadcastsd (%r11,%rbx,8), %ymm6 #17.22
decq %rbx
vaddpd %ymm5, %ymm6, %ymm5 #17.22
cmpq %r10, %r15 #14.5
jb ..B1.6 # Prob 82% #14.5
..B1.8: # Preds ..B1.6 ..B1.4
vbroadcastsd (%rdi), %ymm3 #19.18
vaddpd %ymm3, %ymm5, %ymm4 #19.28
vsubpd %ymm2, %ymm4, %ymm2 #19.31
vmulpd %ymm2, %ymm0, %ymm5 #19.31
vmovupd %ymm5, (%rcx,%r9,8) #19.5
addq $4, %r9 #11.3
cmpq %r8, %r9 #11.3
jb ..B1.4 # Prob 82% #11
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <a class="moz-txt-link-rfc2396E" href="http://lists.llvm.org/pipermail/llvm-dev/attachments/20170510/9a48b564/attachment-0001.html"><http://lists.llvm.org/pipermail/llvm-dev/attachments/20170510/9a48b564/attachment-0001.html></a>
------------------------------
</pre>
</blockquote>
<br>
<pre class="moz-signature" cols="72">--
Simon Moll
Researcher / PhD Student
Compiler Design Lab (Prof. Hack)
Saarland University, Computer Science
Building E1.3, Room 4.31
Tel. +49 (0)681 302-57521 : <a class="moz-txt-link-abbreviated" href="mailto:moll@cs.uni-saarland.de">moll@cs.uni-saarland.de</a>
Fax. +49 (0)681 302-3065 : <a class="moz-txt-link-freetext" href="http://compilers.cs.uni-saarland.de/people/moll">http://compilers.cs.uni-saarland.de/people/moll</a></pre>
</body>
</html>