KonstantinSeurer
/
mesa
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

nir/lower_locals_to_regs: Do an ad-hoc copy propagate on our generated MOV. 

I noticed the inefficiency in NIR-to-TGSI output while trying to debug a
failure handling some arrays in r600.  While this makes reading CTS
shaders easier, the effect in the real world is pretty limited.  From
softpipe shader-db:

total instructions in shared programs: 2929840 -> 2929836 (<.01%)
instructions in affected programs: 118 -> 114 (-3.39%)

Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14321>
This commit is contained in:
Emma Anholt 2021-12-28 10:03:35 -08:00 committed by Marge Bot
parent ef325d4650
commit 61400f8a2d
1 changed files with 20 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
src/compiler/nir/nir_lower_locals_to_regs.c
View File

@ -227,7 +227,27 @@ lower_locals_to_regs_block(nir_block *block,
nir_src reg_src = get_deref_reg_src(deref, state);
nir_alu_instr *mov = nir_alu_instr_create(b->shader, nir_op_mov);
nir_src_copy(&mov->src[0].src, &intrin->src[1]);
/* The normal NIR SSA copy propagate pass can't happen after this pass,
* so do an ad-hoc copy propagate since this ALU op can do swizzles
* while the deref couldn't.
*/
if (mov->src[0].src.is_ssa) {
nir_instr *parent = mov->src[0].src.ssa->parent_instr;
if (parent->type == nir_instr_type_alu) {
nir_alu_instr *parent_alu = nir_instr_as_alu(parent);
if (parent_alu->op == nir_op_mov && parent_alu->src[0].src.is_ssa) {
for (unsigned i = 0; i < intrin->num_components; i++)
mov->src[0].swizzle[i] = parent_alu->src[0].swizzle[mov->src[0].swizzle[i]];
mov->src[0].abs = parent_alu->src[0].abs;
mov->src[0].negate = parent_alu->src[0].negate;
mov->src[0].src = parent_alu->src[0].src;
}
}
}
mov->dest.write_mask = nir_intrinsic_write_mask(intrin);
mov->dest.dest.is_ssa = false;
mov->dest.dest.reg.reg = reg_src.reg.reg;