浏览代码

nir/builder: Don't emit no-op swizzles

The nir_swizzle helper is used some on it's own but it's also called by
nir_channel and nir_channels which are used everywhere.  It's pretty
quick to check while we're walking the swizzle anyway whether or not
it's an identity swizzle.  If it is, we now don't bother emitting the
instruction.  Sure, copy-prop will clean it up for us but there's no
sense making more work for the optimizer than we have to.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
tags/19.1-branchpoint
Jason Ekstrand 6 年前
父节点
当前提交
743700be1f
共有 1 个文件被更改,包括 9 次插入1 次删除
  1. 9
    1
      src/compiler/nir/nir_builder.h

+ 9
- 1
src/compiler/nir/nir_builder.h 查看文件

@@ -497,8 +497,16 @@ nir_swizzle(nir_builder *build, nir_ssa_def *src, const unsigned *swiz,
assert(num_components <= NIR_MAX_VEC_COMPONENTS);
nir_alu_src alu_src = { NIR_SRC_INIT };
alu_src.src = nir_src_for_ssa(src);
for (unsigned i = 0; i < num_components && i < NIR_MAX_VEC_COMPONENTS; i++)

bool is_identity_swizzle = true;
for (unsigned i = 0; i < num_components && i < NIR_MAX_VEC_COMPONENTS; i++) {
if (swiz[i] != i)
is_identity_swizzle = false;
alu_src.swizzle[i] = swiz[i];
}

if (num_components == src->num_components && is_identity_swizzle)
return src;

return use_fmov ? nir_fmov_alu(build, alu_src, num_components) :
nir_imov_alu(build, alu_src, num_components);

正在加载...
取消
保存