Replace swizzles on the LHS with additional swizzles on the RHS and a
write mask in the assignment instruction. As part of this add
ir_assignment::set_lhs. Ideally we'd make ir_assignment::lhs private
to prevent erroneous writes, but that would require a lot of code
butchery at this point.
Add ir_assignment constructor that takes an explicit write mask. This
is required for ir_assignment::clone, but it can also be used in other
places. Without this, ir_assignment clones lose their write masks,
and incorrect IR is generated in optimization passes.
Add ir_assignment::whole_variable_written method. This method gets
the variable on the LHS if the whole variable is written or NULL
otherwise. This is different from
ir->lhs->whole_variable_referenced() because the latter has no
knowledge of the write mask stored in the ir_assignment.
Gut all code from ir_to_mesa that handled swizzles on the LHS of
assignments. There is probably some other refactoring that could be
done here, but that can be left for another day.
Calling exit() on a memory failure probably made sense for the
standalone preprocessor, but doesn't seem too appealing as part of
the GL library. Also, we don't use it in the main compiler.
Mesa will do the mapping at _mesa_add_sampler() time. Fixes assertion
failures in debug builds, which might have caught real problems with
multiple samplers linked in a row.
Instead of using a linker-assigned location (since samplers don't
actually take up uniform space, being a link-time choice), use the
sampler's varaible pointer as a hash key.
The new compiler doesn't generate Mesa IR at compile time, and that
compile time code previously wouldn't have reflected the link time
code that actually got used. But do dump the info log of the compile
regardless.
In most cases, we needed to be reparenting the cloned IR to a
different context (for example, to the linked shader instead of the
unlinked shader), or optimization before the reparent would cause
memory usage of the original object to grow and grow.
This is a zero-ing function, (like calloc), to avoid bugs due to
accessing uninitialized values. Thanks to valgrind for noticing the
use of uninitialized values.
The symbol_header structure that tracks symbols with a particular name
may have a different (longer) life time than the symbols it tracks.
Not keeping a local copy of the name can lead to use-after-free
errors. For example, the following sequence would trigger such an
error:
char *copy = strdup(name);
_mesa_symbol_table_push_scope(st);
_mesa_symbol_table_add_symbol(st, 0, name, NULL);
_mesa_symbol_table_pop_scope(st);
free(name);
_mesa_symbol_table_find_symbol(st, 0, copy);
With this change, the symbol table keeps a local copy of the name that
has the same life time as the symbol_header for that name. This
resolves some use-after-free errors with built-in functions in the
GLSL compiler.
The non-named parameter grammar understandably doesn't set the
identifier field. Fixes intermittent failures about void main(void)
{} having a named void parameter.
So many problems here. One is that we can't do the quadrant handling
for all the channels at the same time, so we call the float(y, x)
version multiple times. I'd also left out the x == 0 handling. Also,
the quadrant handling was broken for y == 0, so there was a funny
discontinuity on the +x side if you plugged in obvious values to test.
I generated the atan(float y, float x) code from a short segment of
GLSL and pasted it in by hand. It would be nice to automate that
somehow.
Fixes:
glsl-fs-atan-1
glsl-fs-atan-2
This prevents using uninitialized data in _msea_glsl_error in some
cases, (including at least 6 piglit tests). Thanks to valgrind for
pointing out the problem!
Otherwise, ir_function_inlining will see the body of the function from
the unlinked version of the shader, which won't have had the lowering
passes done on it or linking's variable remapping.
