Tutorial part 2: Creating a trivial machine code function

Consider this C function:

int square (int i)
{
  return i * i;
}

How can we construct this at run-time using libgccjit?

First we need to include the relevant header:

#include <libgccjit.h>

All state associated with compilation is associated with a gcc_jit_context *.

Create one using gcc_jit_context_acquire():

gcc_jit_context *ctxt;
ctxt = gcc_jit_context_acquire ();

The JIT library has a system of types. It is statically-typed: every expression is of a specific type, fixed at compile-time. In our example, all of the expressions are of the C int type, so let’s obtain this from the context, as a gcc_jit_type *, using gcc_jit_context_get_type():

gcc_jit_type *int_type =
  gcc_jit_context_get_type (ctxt, GCC_JIT_TYPE_INT);

gcc_jit_type * is an example of a “contextual” object: every entity in the API is associated with a gcc_jit_context *.

Memory management is easy: all such “contextual” objects are automatically cleaned up for you when the context is released, using gcc_jit_context_release():

gcc_jit_context_release (ctxt);

so you don’t need to manually track and cleanup all objects, just the contexts.

Although the API is C-based, there is a form of class hierarchy, which looks like this:

+- gcc_jit_object
    +- gcc_jit_location
    +- gcc_jit_type
       +- gcc_jit_struct
    +- gcc_jit_field
    +- gcc_jit_function
    +- gcc_jit_block
    +- gcc_jit_rvalue
        +- gcc_jit_lvalue
           +- gcc_jit_param

There are casting methods for upcasting from subclasses to parent classes. For example, gcc_jit_type_as_object():

gcc_jit_object *obj = gcc_jit_type_as_object (int_type);

One thing you can do with a gcc_jit_object * is to ask it for a human-readable description, using gcc_jit_object_get_debug_string():

printf ("obj: %s\n", gcc_jit_object_get_debug_string (obj));

giving this text on stdout:

obj: int

This is invaluable when debugging.

Let’s create the function. To do so, we first need to construct its single parameter, specifying its type and giving it a name, using gcc_jit_context_new_param():

gcc_jit_param *param_i =
  gcc_jit_context_new_param (ctxt, NULL, int_type, "i");

Now we can create the function, using gcc_jit_context_new_function():

gcc_jit_function *func =
  gcc_jit_context_new_function (ctxt, NULL,
                                GCC_JIT_FUNCTION_EXPORTED,
                                int_type,
                                "square",
                                1, &param_i,
                                0);

To define the code within the function, we must create basic blocks containing statements.

Every basic block contains a list of statements, eventually terminated by a statement that either returns, or jumps to another basic block.

Our function has no control-flow, so we just need one basic block:

gcc_jit_block *block = gcc_jit_function_new_block (func, NULL);

Our basic block is relatively simple: it immediately terminates by returning the value of an expression.

We can build the expression using gcc_jit_context_new_binary_op():

gcc_jit_rvalue *expr =
  gcc_jit_context_new_binary_op (
    ctxt, NULL,
    GCC_JIT_BINARY_OP_MULT, int_type,
    gcc_jit_param_as_rvalue (param_i),
    gcc_jit_param_as_rvalue (param_i));

A gcc_jit_rvalue * is another example of a gcc_jit_object * subclass. We can upcast it using gcc_jit_rvalue_as_object() and as before print it with gcc_jit_object_get_debug_string().

printf ("expr: %s\n",
        gcc_jit_object_get_debug_string (
          gcc_jit_rvalue_as_object (expr)));

giving this output:

expr: i * i

Creating the expression in itself doesn’t do anything; we have to add this expression to a statement within the block. In this case, we use it to build a return statement, which terminates the basic block:

gcc_jit_block_end_with_return (block, NULL, expr);

OK, we’ve populated the context. We can now compile it using gcc_jit_context_compile():

gcc_jit_result *result;
result = gcc_jit_context_compile (ctxt);

and get a gcc_jit_result *.

We can now use gcc_jit_result_get_code() to look up a specific machine code routine within the result, in this case, the function we created above.

void *fn_ptr = gcc_jit_result_get_code (result, "square");
if (!fn_ptr)
  {
    fprintf (stderr, "NULL fn_ptr");
    goto error;
  }

We can now cast the pointer to an appropriate function pointer type, and then call it:

typedef int (*fn_type) (int);
fn_type square = (fn_type)fn_ptr;
printf ("result: %d", square (5));

result: 25

Options

To get more information on what’s going on, you can set debugging flags on the context using gcc_jit_context_set_bool_option().

Setting GCC_JIT_BOOL_OPTION_DUMP_INITIAL_GIMPLE will dump a C-like representation to stderr when you compile (GCC’s “GIMPLE” representation):

gcc_jit_context_set_bool_option (
  ctxt,
  GCC_JIT_BOOL_OPTION_DUMP_INITIAL_GIMPLE,
  1);
result = gcc_jit_context_compile (ctxt);

square (signed int i)
{
  signed int D.260;

  entry:
  D.260 = i * i;
  return D.260;
}

We can see the generated machine code in assembler form (on stderr) by setting GCC_JIT_BOOL_OPTION_DUMP_GENERATED_CODE on the context before compiling:

gcc_jit_context_set_bool_option (
  ctxt,
  GCC_JIT_BOOL_OPTION_DUMP_GENERATED_CODE,
  1);
result = gcc_jit_context_compile (ctxt);

      .file   "fake.c"
      .text
      .globl  square
      .type   square, @function
square:
.LFB6:
      .cfi_startproc
      pushq   %rbp
      .cfi_def_cfa_offset 16
      .cfi_offset 6, -16
      movq    %rsp, %rbp
      .cfi_def_cfa_register 6
      movl    %edi, -4(%rbp)
.L14:
      movl    -4(%rbp), %eax
      imull   -4(%rbp), %eax
      popq    %rbp
      .cfi_def_cfa 7, 8
      ret
      .cfi_endproc
.LFE6:
      .size   square, .-square
      .ident  "GCC: (GNU) 4.9.0 20131023 (Red Hat 0.2-0.5.1920c315ff984892399893b380305ab36e07b455.fc20)"
      .section       .note.GNU-stack,"",@progbits

By default, no optimizations are performed, the equivalent of GCC’s -O0 option. We can turn things up to e.g. -O3 by calling gcc_jit_context_set_int_option() with GCC_JIT_INT_OPTION_OPTIMIZATION_LEVEL:

gcc_jit_context_set_int_option (
  ctxt,
  GCC_JIT_INT_OPTION_OPTIMIZATION_LEVEL,
  3);

      .file   "fake.c"
      .text
      .p2align 4,,15
      .globl  square
      .type   square, @function
square:
.LFB7:
      .cfi_startproc
.L16:
      movl    %edi, %eax
      imull   %edi, %eax
      ret
      .cfi_endproc
.LFE7:
      .size   square, .-square
      .ident  "GCC: (GNU) 4.9.0 20131023 (Red Hat 0.2-0.5.1920c315ff984892399893b380305ab36e07b455.fc20)"
      .section        .note.GNU-stack,"",@progbits

Naturally this has only a small effect on such a trivial function.

Full example

Here’s what the above looks like as a complete program:

/* Usage example for libgccjit.so
   Copyright (C) 2014 Free Software Foundation, Inc.

This file is part of GCC.

GCC is free software; you can redistribute it and/or modify it
under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 3, or (at your option)
any later version.

GCC is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
General Public License for more details.

You should have received a copy of the GNU General Public License
along with GCC; see the file COPYING3.  If not see
<http://www.gnu.org/licenses/>.  */

#include <libgccjit.h>

#include <stdlib.h>
#include <stdio.h>

void
create_code (gcc_jit_context *ctxt)
{
  /* Let's try to inject the equivalent of:

      int square (int i)
      {
        return i * i;
      }
  */
  gcc_jit_type *int_type =
    gcc_jit_context_get_type (ctxt, GCC_JIT_TYPE_INT);
  gcc_jit_param *param_i =
    gcc_jit_context_new_param (ctxt, NULL, int_type, "i");
  gcc_jit_function *func =
    gcc_jit_context_new_function (ctxt, NULL,
                                  GCC_JIT_FUNCTION_EXPORTED,
                                  int_type,
                                  "square",
                                  1, &param_i,
                                  0);

  gcc_jit_block *block = gcc_jit_function_new_block (func, NULL);

  gcc_jit_rvalue *expr =
    gcc_jit_context_new_binary_op (
      ctxt, NULL,
      GCC_JIT_BINARY_OP_MULT, int_type,
      gcc_jit_param_as_rvalue (param_i),
      gcc_jit_param_as_rvalue (param_i));

   gcc_jit_block_end_with_return (block, NULL, expr);
}

int
main (int argc, char **argv)
{
  gcc_jit_context *ctxt = NULL;
  gcc_jit_result *result = NULL;

  /* Get a "context" object for working with the library.  */
  ctxt = gcc_jit_context_acquire ();
  if (!ctxt)
    {
      fprintf (stderr, "NULL ctxt");
      goto error;
    }

  /* Set some options on the context.
     Let's see the code being generated, in assembler form.  */
  gcc_jit_context_set_bool_option (
    ctxt,
    GCC_JIT_BOOL_OPTION_DUMP_GENERATED_CODE,
    0);

  /* Populate the context.  */
  create_code (ctxt);

  /* Compile the code.  */
  result = gcc_jit_context_compile (ctxt);
  if (!result)
    {
      fprintf (stderr, "NULL result");
      goto error;
    }

  /* Extract the generated code from "result".  */
  void *fn_ptr = gcc_jit_result_get_code (result, "square");
  if (!fn_ptr)
     {
       fprintf (stderr, "NULL fn_ptr");
       goto error;
     }

  typedef int (*fn_type) (int);
  fn_type square = (fn_type)fn_ptr;
  printf ("result: %d", square (5));

 error:
  gcc_jit_context_release (ctxt);
  gcc_jit_result_release (result);
  return 0;
}

Building and running it:

$ gcc \
    tut02-square.c \
    -o tut02-square \
    -lgccjit

# Run the built program:
$ ./tut02-square
result: 25