How to add a macro with clBuildProgram for OpenCL - opencl

In my kernel i have this defined.
#define ACTIVATION_FUNCTION(X) (1.7159f*tanh(2.0f/3.0f*X))
I would like to define it in the clBuildProgram call, such i can alter the kernel at runtime. How can i do this?

You can use the -D argument to the OpenCL compiler, by passing it in the options parameter of the clBuildProgram function. Passing -D x=y, is equivalent to adding #define x y at the top of your kernel file. Similarly, passing -D x is equivalent to adding #define x (for any x and y, of course).
In your case, you probably want to pass something like this:
-D ACTIVATION_FUNCTION(X)=(1.7159f*tanh(2.0f/3.0f*X))
Which you can then change as you see fit, directly from your program, at runtime.
Note you could also open the kernel file and write the define directly into it, as an alternative solution, but this is probably the cleanest way. Just be careful with newlines, I'm not sure how well they are handled.
Ref. this documentation page on clBuildProgram, "Preprocessor Options" section.

Related

My Emacs substitute tab into whitespace automatically

This has been bothering me, which hook should I check to prevent this from happening (which makes the Makefile fail)
Put this in your init file:
(setq-default indent-tabs-mode nil)
By default, that option is non-nil, which means that Emacs uses TAB chars when it can, instead of just SPC chars.
Update after your comment:
Maybe you have that code in your init file, but I suspect that other code that you are also using (e.g. loading or autoloading) sets the value to non-nil in your makefile buffer.
That's easy to test: in that buffer, do C-h v indent-tabs-mode.
If the value is non-nil, then check the code that you invoke to set up the makefile buffer - makefile-mode or whatever code sets up the buffer. I'll bet that this is your problem.
Remember: setq-default sets the default value, which is the value the variable will have in any buffer, unless you set the (local) value to something else in that buffer. Something is setting the value to non-nil in that buffer -that's my guess. It's up to use to find that something.
If you cannot find it by checking the code that sets up the makefile buffer, then bisect your init file (and if necessary files, it loads) recursively to find the culprit.

What are kernel blocks in OpenCL?

In the article "How to set up Xcode to run OpenCL code, and how to verify the kernels before building" NeXTCoder referred to some code as the "Short Answer", i.e. https://developer.apple.com/library/mac/#documentation/Performance/Conceptual/OpenCL_MacProgGuide/XCodeHelloWorld/XCodeHelloWorld.html.
In that code the author says "Wrap your kernel code into a kernel block:" without explaining what is a "kernel block". (The OpenCL Programmer Guide for Mac OS X by Apple makes no mention of kernel block.)
The host program calls "square_kernel" but the sample kernel is called "square" and the sample kernel block is labelled "kernelName" (in italics). Can you please tell me how to put the 3 pieces together:kernel, kernel block & host program to run in Xcode 5.1? I only have one kernel. Thanks.
It's not really jargon. It's closure-like entity.
OpenCL C 2.0 adds support for the clang block syntax. You use the ^ operator to declare a Block variable and to indicate the beginning of a Block literal. The body of the Block itself is contained within {}, as shown in the example (as usual with C, ; indicates the end of the statement).The Block is able to make use of variables from the same scope in which it was defined.
Example:
int multiplier = 7;
int (^myBlock)(int) = ^(int num) {
return num * multiplier;
};
printf(“%d\n”, myBlock(3));
// prints 21
Source:
https://www.khronos.org/registry/cl/sdk/2.1/docs/man/xhtml/blocks.html
The term "kernel block" only seems to be a jargon to refer to the "part of the code that is the kernel". Particularly, the kernel block in this case is simply the function that is declared to be a kernel, by adding kernel before its declaration. Or, even simpler, and from the way how the term is used on this website, I would say that "kernel block" is the same as "kernel".
The kernelName (in italics) is a placeholder. The code there shows the general pattern of how to define any kernel:
It is prefixed with kernel
It returns void
It has a name ... the kernelName, which may for example be square
It has several input- and output parameters
The reason why the kernel is called square, but invoked with square_kernel seems to be some magic that is done by XCode: It seems to read the .cl file, and creates a .h file that contains additional declarations that are derived from the .cl file (as can be seen in this question, where a kernel called rebound is defined, and GCL generated a rebound_kernel declaration).

How to denote that a command line argument is optional when printing usage

Assume that I have a script that can be run in either of the following ways.
./foo arg1 arg2
./foo
Is there a generally accepted way to denote that arg1 and arg2 aren't mandatory arguments when printing the correct usage of the command?
I've sometimes noticed usage printed with the arguments wrapped in brackets like in the following usage printout.
Usage: ./foo [arg1] [arg2]
Do these brackets mean that the argument is optional or is there another generally accepted way to denote that an argument is optional?
I suppose this is as much a standard as anything.
The Open Group Base Specifications Issue 7
IEEE Std 1003.1, 2013 Edition
Copyright © 2001-2013 The IEEE and The Open Group
Ch. 12 - Utility Conventions
Although it doesn't seem to mention many things I have commonly seen over the years used to denote various meanings:
square brackets [optional option]
angle brackets <required argument>
curly braces {default values}
parenthesis (miscellaneous info)
Edit: I should add, that these are just conventions. The important thing is to pick a convention which is sensible, clearly state your convention, and stick to it consistently. Be flexible and create conventions which seem to be most frequently encountered on your target platform(s). They will be the easiest for users to adapt to.
I personally have not seen a 'standard' that denotes that a switch is optional (like how there's a standard that defines how certain languages are written for example), as it really is personal choice, but according to IBM's docs and the Wiki, along with numerous shell scripts I've personally seen (and command line options from various programs), and the IEEE, the 'defacto' is to treat square bracketed ([]) parameters as optional parameters. Example from Linux:
ping (output trimmed...)
usage: ping [-c count] [-t ttl] host
where [-c count] and [-t ttl] are optional parameters but host is not (as defined in the help).
I personally follow the defacto as well by using [] to mean they are optional parameters and make sure to note that in the usage of that script/program.
I should note that a computer standard should define how something happens and its failure paths (either true fail or undefined behavior). Something along the lines of the command line interpreter _shall_ treat arguments as optional when enclosed in square brackets, and _shall_ treat X as Y when Z, etc.. Much like the ISO C standard says how a function shall be formed for it to be valid (otherwise it fails). Given that there are no command line interpreters, from ASH to ZSH and everything in between, that fail a script for treating [] as anything but optional, one could say there is not a true standard.
Yes, the square brackets indicate optional arguments in Unix man pages.
From "man man":
[-abc] any or all arguments within [ ] are optional.
I've never wondered if they're formally specified somewhere, I've always just assumed they come from conventions used in abstract algebra, in particular, in BNF grammars.

Passing arguments to execl

I want to create my own pipeline like in Unix terminal (just to practice). It should take applications to execute in quotes like that:
pipeline "ls -l" "grep" ....
I know that I should use fork(), execl() (exec*) and API to redirect stdin and stdout. But are there any alternatives for execl to execute app with arguments using just one argument which includes application path and arguments? Is there a way not to parse manually ls -l but pass it as one argument to execl?
If you have only a single command line instead of an argument vector, let the shell do the parsing for you:
execl("/bin/sh", "sh", "-c", the_command_line, NULL);
Of course, don't let untrusted remote user input into this command line. But if you are dealing with untrusted remote user input to begin with, you should try to arrange to pass actual a list of isolated arguments to the target application as per normal usage of exec[vl], not a command line.
Realistically, you can only really use execl() when the number of arguments to the command are known at compile time. In a shell, you'll normally use execv() or execvp() instead; these can handle an arbitrary number of arguments to the command to be executed. In theory, you use execv() when the path name of the command is given and execvp() (which does a PATH-based search for the command) when it isn't. However, execvp() handles the 'path given' case, so simply use execvp().
So, for your pipeline command, you'll end up with one child using something equivalent to:
char *args_1[] = { "ls", "-l", 0 };
execvp(args_1[0], args_1);
The other child will end up using something equivalent to:
char *args_2[] = { "grep", "pattern", 0 };
execvp(args_2[0], args_2);
Except, of course, that you'll have created those strings from the command line arguments instead of by initialization as shown. Note that grep requires a pattern to search for.
You've still got plumbing issues to resolve. Make sure you close enough pipe file descriptors. When you dup() or dup2() a pipe to standard input or standard output, you close both the file descriptors from the pipe() function.

gdb: logging something instead of breaking?

Is it possible to have gdb log something to the terminal instead of breaking on it? For example I would like to set a 'breakpoint' on some method and have gdb print self as well as the parameters each time the method is invoked. Basically I want to insert print statements into arbitrary places without actually recompiling.
thanks for any suggestions
This is what i have so far after these helpful comments:
define logFoo
b fooMethod
commands
po self
end
end
GDB doesn't seem to like the nested end statements though. any thoughts?
You can use Breakpoint Command Lists. There is an example how to do it.
For example, here is how you could use
breakpoint commands to print the value
of x at entry to foo whenever x is
positive.
break foo if x>0
commands
silent
printf "x is %d\n",x
cont
end
Use a breakpoint as usual, and set a macro to log and continue:
define c
print "foo"
cont
c
end
No, this is not possible. You can only hook into the symbols of the code and machine code. If you want to log output you will need a logging functionality.
If you are tracing specific errors try conditional breakpoints and watch variables.
EDIT:
Even while not directly loggin it could be an alternative to use GDB command files

Resources