<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body text="#000000" bgcolor="#FFFFFF">
<p><br>
</p>
<pre class="moz-signature" cols="72">-------------
Best regards,
Alexey Bataev</pre>
<div class="moz-cite-prefix">17.01.2019 13:41, Finkel, Hal J. пишет:<br>
</div>
<blockquote type="cite"
cite="mid:9d7bf137-9881-31b4-7a72-c767fd372786@anl.gov">
<pre class="moz-quote-pre" wrap="">On 1/17/19 12:18 PM, Alexey Bataev wrote:
</pre>
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">
-------------
Best regards,
Alexey Bataev
17.01.2019 13:11, Finkel, Hal J. пишет:
</pre>
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">On 1/17/19 11:44 AM, Alexey Bataev wrote:
</pre>
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">-------------
Best regards,
Alexey Bataev
17.01.2019 12:40, Finkel, Hal J. пишет:
</pre>
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">On 1/17/19 11:11 AM, Alexey Bataev wrote:
</pre>
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">The compiler does not know anything about the layout on the host when
it compiles for the device.
</pre>
</blockquote>
<pre class="moz-quote-pre" wrap="">No, the compiler does know about the host layout (e.g., can't we
construct this by calling getContext().getAuxTargetInfo(), or similar?).
</pre>
</blockquote>
<pre class="moz-quote-pre" wrap="">The compiler knows, but the backend is unable to correctly represent
unsupported types.
</pre>
</blockquote>
<pre class="moz-quote-pre" wrap="">To any extent to which this is true, this is something that we can fix.
</pre>
<blockquote type="cite">
<blockquote type="cite">
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">We cannot do anything with the types that are not supported by the
target device and we cannot use the layout from the host. And it is
user responsibility to write and use the code that is compatible with
the the target devices.
He/she does not need to use macros/void* types, there are templates.
</pre>
</blockquote>
<pre class="moz-quote-pre" wrap="">No. This doesn't solve the problem (because you still need to share the
instantiations between the devices). Also, even if it did, does not
address the legacy-code problem that the feature is intended to address.
The user already has classes and data on the host and wishes to access
*parts* of that data on the device. We should make as much of that work
as possible.
</pre>
</blockquote>
<pre class="moz-quote-pre" wrap="">Design your code to be compatible with the both, the host and the device.
</pre>
</blockquote>
<pre class="moz-quote-pre" wrap="">Your vantage point here is certainly not the same as mine. From my
perspective, we already have many hundreds of millions of lines of code
and we're incrementally poring that code to use OpenMP and accelerators.
This is not a new top-down design process. We already have data
structures and we want to make accessing those data structures from the
accelerator as easy as possible. We should require as little code change
as possible - we'll have more than enough code change necessary for hot
code regions, and we'd like to minimize necessary changes elsewhere.
Anything that we can find a way to support in this regard, we should.
</pre>
</blockquote>
<pre class="moz-quote-pre" wrap="">But we should not break the compiler because of this. Plus, don't you
think that we may break code reliability?
</pre>
</blockquote>
<pre class="moz-quote-pre" wrap="">
I'm proposing no such thing. Or, you say "break" and I say "properly
implement" :-)
</pre>
</blockquote>
<p>I don't think that what you're saying is "properly implement".
For me it sounds like "let's break the compiler in favor of user
code". Or "let's add a hack in favor of user code".<br>
</p>
<blockquote type="cite"
cite="mid:9d7bf137-9881-31b4-7a72-c767fd372786@anl.gov">
<pre class="moz-quote-pre" wrap="">
We would still need to produce an error in cases where we'd need to
produce an operation that we couldn't lower. We could also produce
runtime calls and let linking fail, but an error is better.
</pre>
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">Introduce possible security issues in the code emitted by the compiler
that does not take device restrictions into the account?
</pre>
</blockquote>
<pre class="moz-quote-pre" wrap="">
This is not a concern - we'd still fail to produce code that we couldn't
lower, but only if we need operations that are problematic, not based on
transitive type referencing.
Thanks again,
Hal
</pre>
<blockquote type="cite">
<blockquote type="cite">
<blockquote type="cite">
<blockquote type="cite">
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">You cannot use classes, which use types incompatible with the device.
There is a problem with the data layout on the device and we just
don't know how to represent such classes on the device.
</pre>
</blockquote>
<pre class="moz-quote-pre" wrap="">There's no reason for this to be true. To be clear, the model of a
shared address space only makes sense, from a user perspective, if the
data layout is the same between the host and the target. Not mostly
similar, but the same. Otherwise, users will constantly be tracking down
subtle data-layout incompatibilities.
</pre>
</blockquote>
<pre class="moz-quote-pre" wrap="">That's why he/she must use only types, which are supported and have
the same data layout on the host and on the device. Otherwise, there
is data layout incompatibility.
</pre>
</blockquote>
<pre class="moz-quote-pre" wrap="">Again, the model only makes sense if the host and device share
data-layout rules. This is true even for types on which we cannot
(efficiently) operate on the accelerator.
Thanks again,
Hal
</pre>
<blockquote type="cite">
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">Thanks again,
Hal
</pre>
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">-------------
Best regards,
Alexey Bataev
17.01.2019 11:47, Finkel, Hal J. пишет:
</pre>
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">On 1/17/19 9:52 AM, Alexey Bataev wrote:
</pre>
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">Because the type is not compatible with the target device.
</pre>
</blockquote>
<pre class="moz-quote-pre" wrap="">But it's not that simple. The situation is that the programming
environment supports the type, but *operations* on that type are not
supported in certain contexts (e.g., when compiled for a certain
device). As you point out, we already need to move in this explicit
direction by, for example, allowing typedefs for types that are not
supported in all contexts, function declarations, and so on. In the end,
we should allow our users to design their classes and abstractions using
good software-engineering practice without worrying about access-context
partitioning.
Also, the other problem here is that the function I used as an example
is a very common C++ idiom. There are a lot of classes with function
that return a reference to themselves. Classes can have lots of data
members, and those members might not be accessed on the device (even if
the class itself might be accessed on the device). We're moving to a
world in which unified memory is common - the promise of this technology
is that configuration data and complex data structures, which might be
occasionally accessed (but for which explicitly managing data movement
is not performance relevant) are handled transparently. If use of these
data structures is transitively poisoned by use of any type not
supported on the device (including by pointers to types that use those
types), then we'll force unhelpful and technically-unnecessary
refactoring, thus reducing the value of the feature.
In the current implementation we pre-process the source twice, and so we
can:
1. Use ifdefs to change the data memebers when compiling for different
targets. This is hard to get right because, in order to keep the data
layout otherwise the same, the user needs to understand the layout rules
in order to put something in the structure that is supported on the
target and keeps the layout the same (this is very error prone). Also,
if we move to a single-preprocessing-stage model, this no longer works.
2. Replace all pointers to relevant types with void*, or similar, and
use a lot of casts. This is also bad.
We shouldn't be forcing users to play these games. The compiler knows
the layout on the host and it can use it on the target. The fact that
some operations on some types might not be supported on the target is
not relevant to handling pointers/references to containing types.
Thanks again,
Hal
</pre>
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">-------------
Best regards,
Alexey Bataev
17.01.2019 10:50, Finkel, Hal J. пишет:
</pre>
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">On 1/17/19 9:27 AM, Alexey Bataev wrote:
</pre>
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">It should be compilable for the device only iff function foo is not used
on the device.
</pre>
</blockquote>
<pre class="moz-quote-pre" wrap="">Says whom? I disagree. This function should work on the device. Why
should it not?
-Hal
</pre>
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">-------------
Best regards,
Alexey Bataev
17.01.2019 10:24, Finkel, Hal J. пишет:
</pre>
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">On 1/17/19 4:05 AM, Alexey Bataev wrote:
</pre>
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">Best regards,
Alexey Bataev
</pre>
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">17 янв. 2019 г., в 0:46, Finkel, Hal J. <a class="moz-txt-link-rfc2396E" href="mailto:hfinkel@anl.gov"><hfinkel@anl.gov></a> написал(а):
</pre>
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">On 1/16/19 8:45 AM, Alexey Bataev wrote:
Yes, I thought about this. But we need to delay the diagnostic until
the Codegen phase. What I need is the way to associate the diagnostic
with the function so that this diagnostic is available in CodeGen.
Also, we need to postpone the diagnotics not only for functions,
but,for example, for some types. For example, __float128 type is not
supported by CUDA. We can get error messages when we ran into
something like `typedef __float128 SomeOtherType` (say, in some system
header files) and get the error diagnostic when we compile for the
device. Though, actually, this type is not used in the device code,
the diagnostic is still emitted and we need to delay too and emit it
only iff the type is used in the device code.
</pre>
</blockquote>
<pre class="moz-quote-pre" wrap="">This should be fixed for CUDA too, right?
Also, we still get to have pointers to aggregates containing those types
on the device, right?
</pre>
</blockquote>
<pre class="moz-quote-pre" wrap="">No, why? This is not allowed and should be diagnosed too. If somebody tries somehow to use not allowed type for the device variables/functions - it should be diagnosed.
</pre>
</blockquote>
<pre class="moz-quote-pre" wrap="">Because this should be allowed. If I have:
struct X {
int a;
__float128 b;
};
and we have some function which does this:
X *foo(X *x) {
return x;
}
We'll certainly want this function to compile for all targets, even if
there's no __float128 support on some accelerator. The whole model only
really makes sense if the accelerator shares the aggregate-layout rules
of the host, and this is a needless hassle for users if this causes an
error (especially in a unified-memory environment where configuration
data structures, etc. are shared between devices).
Thanks again,
Hal
</pre>
<blockquote type="cite">
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">Thanks again,
Hal
</pre>
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">-------------
Best regards,
Alexey Bataev
15.01.2019 17:33, John McCall пишет:
</pre>
<blockquote type="cite">
<blockquote type="cite">
<pre class="moz-quote-pre" wrap="">On 15 Jan 2019, at 17:20, Alexey Bataev wrote:
This is not only for asm, we need to delay all target-specific
diagnostics.
I'm not saying that we need to move the host diagnostic, only the
diagnostic for the device compilation.
As for Cuda, it is a little but different. In Cuda the programmer
must explicitly mark the device functions, while in OpenMP it must
be done implicitly. Thus, we cannot reuse the solution used for Cuda.
</pre>
</blockquote>
<pre class="moz-quote-pre" wrap="">All it means is that you can't just use the solution used for CUDA
"off the shelf". The basic idea of associating diagnostics with the
current function and then emitting those diagnostics later when you
realize that you have to emit that function is still completely
applicable.
John.
</pre>
</blockquote>
</blockquote>
<pre class="moz-quote-pre" wrap="">--
Hal Finkel
Lead, Compiler Technology and Programming Languages
Leadership Computing Facility
Argonne National Laboratory
</pre>
</blockquote>
</blockquote>
</blockquote>
</blockquote>
</blockquote>
</blockquote>
</blockquote>
</blockquote>
</blockquote>
</blockquote>
</blockquote>
</blockquote>
</blockquote>
</body>
</html>