fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-09 04:38:03 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

nv50/ir: optimize shl + and

Browse source 
Address loading can often end up as shl + shr + shl combinations. The
latter two are equal shifts, which get converted into an and mask.
However if the previous shl is more than the mask is trying to remove
(in terms of low bits), we can just remove the and entirely. This
reduces some large shaders by as many as 3% of instructions (out of 2K).

total instructions in shared programs : 6495509 -> 6491076 (-0.07%)
total gprs used in shared programs    : 954621 -> 954623 (0.00%)

                local        gpr       inst      bytes
    helped           0           0        1014        1014
      hurt           0           2           0           0

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
This commit is contained in:
Ilia Mirkin 2017-01-13 00:41:11 -05:00
parent 5ba380c226
commit 19963231a3
1 changed files with 11 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

11
src/gallium/drivers/nouveau/codegen/nv50_ir_peephole.cpp
View file

@ -1260,6 +1260,17 @@ ConstantFolding::opnd(Instruction *i, ImmediateValue &imm0, int s)
i->op = OP_EXTBF;
i->setSrc(0, src->getSrc(0));
i->setSrc(1, new_ImmediateValue(prog, ext));
} else if (src->op == OP_SHL &&
src->src(1).getImmediate(imm1) &&
i->src(t).mod == Modifier(0) &&
util_is_power_of_two(~imm0.reg.data.u32 + 1) &&
util_last_bit(~imm0.reg.data.u32) <= imm1.reg.data.u32) {
i->op = OP_MOV;
i->setSrc(s, NULL);
if (t) {
i->setSrc(0, i->getSrc(t));
i->setSrc(t, NULL);
}
}
}
break;