It's rather easy to produce two constant multiplies separated by other
multiplies while writing a BRDF shader, and non-obvious enough in the
resulting codegen that I didn't catch it in my demo code until just
recently. Cuts 3 965 instructions from my demo (<1%), and 20 from
glsl-fs-raytrace (1.3%).
All the current HW backends transform subtract to adding the negation,
so I haven't bothered peepholing it back out in Mesa IR. This allows
some subtract of subtract to get removed in ir_algebraic.
Whereas constant folding evaluates constant expressions at rvalue
nodes, constant propagation tracks constant components of vectors
across execution to replace (possibly swizzled) variable dereferences
with constant values, triggering possible constant folding or reduced
variable liveness.
This lets us track copies across basic block boundaries. The loop
doesn't get a filled out list of available copies into it yet, though.
glsl-fs-raytrace drops from 585 to 535 Mesa IR instructions out of the
compiler, and it appears that Yo Frankie's largest shaders decrease in
Mesa IR count by about 10% as well.
This can be supported on r600 without using the endian swapper, and is a
better fit for (typical) uploads using GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV
anyway.
Currently only ir_binop_equal and ir_binop_nequal are supported, but
soon all of the relational operators will be added. Making this
change now will simplify those commits.
This fixes the assert added in LLVM 2.8:
assert(getType()->isIntOrIntVectorTy() &&
"Tried to create an integer operation on a non-integer type!")
But it also fixes some subtle bugs, since we should've been doing this
since LLVM 2.6 anyway.
Includes a modified patch from steckdenis@yahoo.fr for the
FNeg instructions in emit_fetch(); thanks for pointing those out.
http://bugs.freedesktop.org/29404http://bugs.freedesktop.org/29407
Signed-off-by: José Fonseca <jfonseca@vmware.com>
In this case, we were incorrectly prioritizing PIPE_TRANSFER_DONTBLOCK over
PIPE_TRANSFER_UNSYNCHRONIZED.
This can lead to failure in the Mesa VBO draw paths that end up specifying
both, but don't expect map to fail (in particular, the problem manifested as
a leak of buffer objects in teapot with other changes).
This makes it compatible with the modified DRM interface in drm-radeon-testing.
Also, now you need to set RADEON_HYPERZ=1 to be able to use hyperz.
It's not bug-free yet.
Two integers were being operated on as
a vector of floats in draw_llvm_generate().
This bug got uncovered by fixing this bug:
http://bugs.freedesktop.org/29407
This is a follow-up patch to commit
f4511c4835.
Files that include tnl_dd/t_dd_dmatmp.h now need to also include
m_xform.h as t_context.h no longer includes it.
s_fragprog.h
Include mtype.h for GLcontext symbol.
Include s_span.h for SWspan symbol.
s_fragprog.c
Include s_context.h now that it is removed from s_fragprog.h.
