Re: Introspection of exceptions that a function can throw
You'd hit a very big wall with separate compilation unless you can
inspect all the code, and know where to find it.
But you'd have a problem, for instance, if you are writing a plugin (.so
/ DLL) for a product for which you only have .di files.
Or even worse the other way round: if you want to allow people to write
plugins for your product, you can't know what they'll throw, even if
they have your code, unless you enforce a `nothrow` interface.
But I guess that if you're not doing any of this, it should be
possible... although I'd still do it as a separate pre-compilation step,
so it could be cached.
On 26/2/21 3:21, James Blachly wrote:
On 2/24/21 2:38 PM, Mark wrote:
Is there a way to obtain a list, at compile-time, of all the exception
types that a function might throw (directly or through a call to
another function)?
Thanks.
Crazy idea:
Could a program import its own source file as a string (`string source =
import('thisfile.d')`) and `-J` , then use a lexer/parser to
generate AST of the source code and extract exceptions potentially
thrown by given functions -- all at compile time?
Re: Optimizing for SIMD: best practices?(i.e. what features are allowed?)
On Thursday, 25 February 2021 at 11:28:14 UTC, z wrote:
How does one optimize code to make full use of the CPU's SIMD
capabilities?
Is there any way to guarantee that "packed" versions of SIMD
instructions will be used?(e.g. vmulps, vsqrtps, etc...)
To give some context, this is a sample of one of the functions
that could benefit from better SIMD usage :
float euclideanDistanceFixedSizeArray(float[3] a, float[3] b) {
float distance;
a[] -= b[];
a[] *= a[];
static foreach(size_t i; 0 .. 3/+typeof(a).length+/){
distance += a[i].abs;//abs required by the caller
}
return sqrt(distance);
}
vmovsd xmm0,qword ptr ds:[rdx]
vmovss xmm1,dword ptr ds:[rdx+8]
vmovsd xmm2,qword ptr ds:[rcx+4]
vsubps xmm0,xmm0,xmm2
vsubss xmm1,xmm1,dword ptr ds:[rcx+C]
vmulps xmm0,xmm0,xmm0
vmulss xmm1,xmm1,xmm1
vbroadcastss xmm2,dword ptr ds:[<__real@7fff>]
vandps xmm0,xmm0,xmm2
vpermilps xmm3,xmm0,F5
vaddss xmm0,xmm0,xmm3
vandps xmm1,xmm1,xmm2
vaddss xmm0,xmm0,xmm1
vsqrtss xmm0,xmm0,xmm0
vmovaps xmm6,xmmword ptr ss:[rsp+20]
add rsp,38
ret
I've tried to experiment with dynamic arrays of float[3] but
the output assembly seemed to be worse.[1](in short, it's
calling internal D functions which use "vxxxss" instructions
while performing many moves.)
Big thanks
[1] https://run.dlang.io/is/F3Xye3
If you are developing for deployment to a platform that has a
GPU, you might consider going SIMT (dcompute) rather than SIMD.
SIMT is a lot easier on the eyes. More importantly, if you're
targetting an SoC or console, or have relatively chunky
computations that allow you to work around the PCIe transit
costs, the path is open to very large performance improvements.
I've only been using dcompute for a week or so but so far it's
been great.
If your algorithims are very branchy, or you decide to stick with
multi-core/SIMD for any of a number of other good reasons, here
are a few things I learned before decamping to dcompute land that
may help:
1) LDC is pretty good at auto vectorization as you have
probably observed. Definitely worth a few iterations to try and
get the vectorizer engaged.
2) LDC auto vectorization was good but explicit __vector
programming is more predictable and was, at least for my tasks,
much faster. I couldn't persuade the auto vectorizer to "do the
right thing" throughout the hot path but perhaps you'll have
better luck.
3) LDC does a good job of going between T[N] <==>
__vector(T[N]) so using the static array types as your
input/output types and the __vector types as your compute types
works out well whenever you have to interface with an unaligned
world. LDC issues unaligned vector loads/stores for casts or full
array assigns: v = cast(VT)sa[]; or v[] = sa[]; These are quite
good on modern CPUs. To calibrate note that Ethan recently
talked about a 10% gain he experienced using full alignment,
IIRC, so there's that.
4) LDC also does a good job of discovering SIMD equivalents
given static foreach unrolled loops with explicit complie-time
indexing of vector element operands. You can use those along
with pragma(inline, true) to develop your own "intrinsics" that
supplement other libs.
5) If you adopt the __vector approach you'll have to handle the
partials manually. (array length % vec length != 0 indicates a
partial or tail fragment). If the classic (copying/padding)
approaches to such fragmentation don't work for you I'd suggest
using nested static functions that take ref T[N] inputs and
outputs. The main loops become very simple and the tail handling
reduces to loading stack allocated T[N] variables explicitly,
calling the static function, and unloading.
Good luck.
Re: How do I check if a type is assignable to null at compile time?
On Friday, 26 February 2021 at 05:34:26 UTC, Paul Backus wrote: On Friday, 26 February 2021 at 05:25:14 UTC, Jack wrote: I started with: enum isAssignableNull(T) = is(T : Object) || isPointer(T); but how do I cover all cases? Something like this should work: enum isAssignableNull(T) = __traits(compiles, (T t) => t = null); `isAssignableNull!(immutable void*)` is true with his definition but false with yours. Of course you are correct that you cannot assign to an immutable pointer.
Re: How do I check if a type is assignable to null at compile time?
On Friday, 26 February 2021 at 05:25:14 UTC, Jack wrote: I started with: enum isAssignableNull(T) = is(T : Object) || isPointer(T); but how do I cover all cases? If I understand what you mean by "is assignable to null", this should do it: --- enum bool isAssignableNull(T) = is(typeof(null) : T); // Tests: immutable string arrayslice = null; static assert(isAssignableToNull!(immutable string)); void function(int) funptr = null; static assert(isAssignableToNull!(typeof(funptr))); int[int] aa = null; static assert(isAssignableToNull!(int[int])); ---
Re: How do I check if a type is assignable to null at compile time?
On Friday, 26 February 2021 at 05:25:14 UTC, Jack wrote: I started with: enum isAssignableNull(T) = is(T : Object) || isPointer(T); but how do I cover all cases? Something like this should work: enum isAssignableNull(T) = __traits(compiles, (T t) => t = null);
How do I check if a type is assignable to null at compile time?
I started with: enum isAssignableNull(T) = is(T : Object) || isPointer(T); but how do I cover all cases?
Re: Optimizing for SIMD: best practices?(i.e. what features are allowed?)
On Thursday, 25 February 2021 at 11:28:14 UTC, z wrote:
float euclideanDistanceFixedSizeArray(float[3] a, float[3] b) {
Use __vector(float[4]), not float[3].
float distance;
The default value for float is float.nan. You need to explicitly
initialize it to 0.0f or something if you want this function to
actually do anything useful.
a[] -= b[];
a[] *= a[];
With __vector types, this can be simplified (not optimized) to
just:
a -= b;
a *= a;
static foreach(size_t i; 0 .. 3/+typeof(a).length+/){
distance += a[i].abs;//abs required by the caller
(a * a) above is always positive for real numbers. You don't need
the call to abs unless you're trying to guarantee that even nan
values will have a clear sign bit.
Also, there is no point to adding the first component to zero,
and copying element [0] from a SIMD register into a scalar is
free, so this can become:
float distance = a[0];
static foreach(size_t i; 1 .. 3)
distance += a[i];
}
return sqrt(distance);
}
Final assembly output (ldc 1.24.0 with -release -O3
-preview=intpromote -preview=dip1000 -m64 -mcpu=haswell
-fp-contract=fast -enable-cross-module-inlining):
vsubps xmm0, xmm1, xmm0
vmulps xmm0, xmm0, xmm0
vmovshdup xmm1, xmm0
vaddss xmm1, xmm0, xmm1
vpermilpd xmm0, xmm0, 1
vaddss xmm0, xmm0, xmm1
vsqrtss xmm0, xmm0, xmm0
ret
Re: Optimizing for SIMD: best practices?(i.e. what features are allowed?)
On Thursday, 25 February 2021 at 11:28:14 UTC, z wrote:
Is there any way to guarantee that "packed" versions of SIMD
instructions will be used?(e.g. vmulps, vsqrtps, etc...)
To give some context, this is a sample of one of the functions
that could benefit from better SIMD usage :
float euclideanDistanceFixedSizeArray(float[3] a, float[3] b) {
You need to use __vector(float[4]) instead of float[3] to tell
the compiler to pack multiple elements per SIMD register. Right
now your data lacks proper alignment for SIMD load/stores.
Beyond that, SIMD code is rather difficult to optimize. Code
written in ignorance or in a rush is unlikely to be meaningfully
faster than ordinary scalar code, unless the data flow is very
simple. You will probably get a bigger speedup for less effort
and pain by first minimizing heap allocations, maximizing
locality of reference, minimizing indirections, and minimizing
memory use. (And, of course, it should go without saying that
choosing an asymptotically efficient high-level algorithm is more
important than any micro-optimization for large data sets.)
Nevertheless, if you are up to the challenge, SIMD can sometimes
provide a final 2-3x speed boost.
Your algorithms will need to be designed to minimize mixing of
data between SIMD channels, as this forces the generation of lots
of extra instructions to swizzle the data, or worse to unpack and
repack it. Something like a Cartesian dot product or cross
product will benefit much less from SIMD than vector addition,
for example. Sometimes the amount of swizzling can be greatly
reduced with a little algebra, other times you might need to
refactor an array of structures into a structure of arrays.
Per-element conditional branches are very bad, and often
completely defeat the benefits of SIMD. For very short segments
of code (like conditional assignment), replace them with a SIMD
conditional move (vcmp and vblend). Bit-twiddling is your friend.
Finally, do not trust the compiler or the optimizer. People love
to make the claim that "The Compiler" is always better than
humans at micro-optimizations, but this is not at all the case
for SIMD code with current systems. I have found even LLVM to
produce quite bad SIMD code for complex algorithms, unless I
carefully structure my code to make it as easy as possible for
the optimizer to get to the final assembly I want. A sprinkling
of manual assembly code (directly, or via a library) is also
necessary to fill in certain instructions that the compiler
doesn't know when to use at all.
Resources I have found very helpful:
Matt Godbolt's Compiler Explorer online visual disassembler
(supports D):
https://godbolt.org/
Felix Cloutier's x86 and amd64 instruction reference:
https://www.felixcloutier.com/x86/
Agner Fog's optimization guide (especially the instruction
tables):
https://agner.org/optimize/
Re: Introspection of exceptions that a function can throw
On 2/24/21 2:38 PM, Mark wrote:
Is there a way to obtain a list, at compile-time, of all the exception
types that a function might throw (directly or through a call to another
function)?
Thanks.
Crazy idea:
Could a program import its own source file as a string (`string source =
import('thisfile.d')`) and `-J` , then use a lexer/parser to
generate AST of the source code and extract exceptions potentially
thrown by given functions -- all at compile time?
Re: Compile-Time Function Parameters That Aren't Types?
On Wednesday, 24 February 2021 at 20:15:08 UTC, Ali Çehreli wrote: On 2/23/21 7:52 PM, Kyle Ingraham wrote: Where would one find information on this There are Point and Polygon struct templates on the following page where one can pick e.g. the dimension (e.g. three dimensional space) by a size_t template parameter. http://ddili.org/ders/d.en/templates_more.html#ix_templates_more.value%20template%20parameter Ali Thank you very much Ali. This is another great example. Your book has been most helpful!
tiny alternative to std library
Hello, I noticed that importing some std libraries causes the build time to jump to around 1 - 3 secs. I started creating my own helper functions to avoid importing std for scripting and prototyping in order to keep the compile time at around 0.5 secs. I was wondering if anyone knows of any libraries that are geared towards something like this? Thanks
Re: DUB is not working correctly
On Thursday, 25 February 2021 at 17:34:29 UTC, Maxim wrote: In my opinion, that happens because of DUB can't accept my changes to the file. But how to make the manager admit it? I apologize to everyone who understood me wrong. If you use an ide to edit the DUB receipt maybe it's possible that you forgot to save the project before retriggering a rebuild. If so then the file used by DUB is still in its previous state.
Re: DUB is not working correctly
On Thursday, 25 February 2021 at 17:47:57 UTC, user1234 wrote: On Thursday, 25 February 2021 at 17:34:29 UTC, Maxim wrote: In my opinion, that happens because of DUB can't accept my changes to the file. But how to make the manager admit it? I apologize to everyone who understood me wrong. If you use an ide to edit the DUB receipt maybe it's possible that you forgot to save the project before retriggering a rebuild. If so then the file used by DUB is still in its previous state. I'm using a text editor.
Re: DUB is not working correctly
On Thursday, 25 February 2021 at 17:34:29 UTC, Maxim wrote: On Wednesday, 24 February 2021 at 16:13:48 UTC, Maxim wrote: [...] I think, I need to rephrase the question for the present situtation: how can I force DUB to change targetName? It doesn't read my changes in dub.json, and I don't know why. In my opinion, that happens because of DUB can't accept my changes to the file. But how to make the manager admit it? I apologize to everyone who understood me wrong. And my problem has been substituted by another! Now DUB says to me to set targetName with this problem: 'No target name set.'
Re: DUB is not working correctly
On Wednesday, 24 February 2021 at 16:13:48 UTC, Maxim wrote:
Hello, I have problems with working in dub environment. If I
try to init my project with 'dub init', all needed files will
be created successfully. However, when I run 'dub run', the
manager gives me an error:
'Configuration 'application' of package application contains no
source files. Please add {"targetType": "none"} to its package
description to avoid building it.
Package with target type "none" must have dependencies to
build.'
If I set targetType in dub.json to "none", the only message:
'Package with target type "none" must have dependencies to
build.'
will remain. When I set targetName to "app" or any other name,
the problem will appear again with the same message above. I
flicked through many forums and videos how to correctly install
dub, and nothing helped.
I am using dub 1.23.0 built on Sep 25 2020 on Windows 10 x64.
Will be very thankful for your help!
I think, I need to rephrase the question for the present
situtation: how can I force DUB to change targetName? It doesn't
read my changes in dub.json, and I don't know why. In my opinion,
that happens because of DUB can't accept my changes to the file.
But how to make the manager admit it? I apologize to everyone who
understood me wrong.
Re: byte array to string
On 2/24/21 10:58 PM, FeepingCreature wrote: On Thursday, 25 February 2021 at 06:57:57 UTC, FeepingCreature wrote: On Thursday, 25 February 2021 at 06:47:11 UTC, Mike wrote: hi all, If i have an array: byte[3] = [1,2,3]; How to get string "123" from it? Thanks in advance. string str = format!"%(%s)"(array); Er sorry, typo, that should be "%(%s%)". "Print the elements of the array, separated by nothing." Compare "%(%s, %)" for a comma separated list. I have an explanation of that syntax here: https://youtu.be/dRORNQIB2wA?t=981 Ali
Re: How to get output of piped process?
On Monday, 22 February 2021 at 13:23:40 UTC, Danny Arends wrote: On Friday, 19 February 2021 at 15:39:25 UTC, kdevel wrote: [...] Fortunately the D runtime /does/ take care and it throws---if the signal is ignored beforehand. I filed issue 21649. [...] Perhaps a bit late, It's never too late.™ :-) but this is how I deal with pipes and spawnShell. Read one byte at a time from stdout and stderr: https://github.com/DannyArends/DaNode/blob/master/danode/process.d Is this immune to SIGPIPE and is this design able to serve infinite streams? BTW: Why does run use spawnShell and not spawnProcess (would save one File object). If the design is not intended to serve infinite streams I would suggest to open two temporary files "out" and "err", delete them, and let the child process write stdout/stderr into those files. IFAICS this avoid threads, sleep, pipe and reading with fgetc.
Re: Optimizing for SIMD: best practices?(i.e. what features are allowed?)
On Thursday, 25 February 2021 at 11:28:14 UTC, z wrote: How does one optimize code to make full use of the CPU's SIMD capabilities? Is there any way to guarantee that "packed" versions of SIMD instructions will be used?(e.g. vmulps, vsqrtps, etc...) https://code.dlang.org/packages/intel-intrinsics
Re: DUB is not working correctly
On Thursday, 25 February 2021 at 13:07:08 UTC, Siemargl wrote: On Thursday, 25 February 2021 at 12:49:05 UTC, Maxim wrote: Wait, I have just noticed that you said about dub which came with dmd. So, was dub on your computer right after compiler installation or you did you do this manually? Many years DMD come together with dub. Look at dmd2\windows\bin\ folder I thought, if I run native dub (without installation from github), all will be fine, but it isn't.
Re: DUB is not working correctly
On Thursday, 25 February 2021 at 12:49:05 UTC, Maxim wrote: Wait, I have just noticed that you said about dub which came with dmd. So, was dub on your computer right after compiler installation or you did you do this manually? Many years DMD come together with dub. Look at dmd2\windows\bin\ folder
Re: DUB is not working correctly
On Wednesday, 24 February 2021 at 19:48:10 UTC, Siemargl wrote: On Wednesday, 24 February 2021 at 19:42:05 UTC, Maxim wrote: On Wednesday, 24 February 2021 at 19:35:46 UTC, Siemargl wrote: On Wednesday, 24 February 2021 at 19:18:32 UTC, Siemargl wrote: [...] Yes, i see th error. If in your dub.json change > "targetType": "executable", [...] Tried and everything seems to be okay with underscores but the error message about the targetType appears even when I start my new pure project without any dependencies. Just downloaded latest dmd 2.095.1, it comes with dub 1.24.1 Works fine on void dub init project, but underscore problem whith dsfml still here Wait, I have just noticed that you said about dub which came with dmd. So, was dub on your computer right after compiler installation or you did you do this manually?
Optimizing for SIMD: best practices?(i.e. what features are allowed?)
How does one optimize code to make full use of the CPU's SIMD
capabilities?
Is there any way to guarantee that "packed" versions of SIMD
instructions will be used?(e.g. vmulps, vsqrtps, etc...)
To give some context, this is a sample of one of the functions
that could benefit from better SIMD usage :
float euclideanDistanceFixedSizeArray(float[3] a, float[3] b) {
float distance;
a[] -= b[];
a[] *= a[];
static foreach(size_t i; 0 .. 3/+typeof(a).length+/){
distance += a[i].abs;//abs required by the caller
}
return sqrt(distance);
}
vmovsd xmm0,qword ptr ds:[rdx]
vmovss xmm1,dword ptr ds:[rdx+8]
vmovsd xmm2,qword ptr ds:[rcx+4]
vsubps xmm0,xmm0,xmm2
vsubss xmm1,xmm1,dword ptr ds:[rcx+C]
vmulps xmm0,xmm0,xmm0
vmulss xmm1,xmm1,xmm1
vbroadcastss xmm2,dword ptr ds:[<__real@7fff>]
vandps xmm0,xmm0,xmm2
vpermilps xmm3,xmm0,F5
vaddss xmm0,xmm0,xmm3
vandps xmm1,xmm1,xmm2
vaddss xmm0,xmm0,xmm1
vsqrtss xmm0,xmm0,xmm0
vmovaps xmm6,xmmword ptr ss:[rsp+20]
add rsp,38
ret
I've tried to experiment with dynamic arrays of float[3] but the
output assembly seemed to be worse.[1](in short, it's calling
internal D functions which use "vxxxss" instructions while
performing many moves.)
Big thanks
[1] https://run.dlang.io/is/F3Xye3
