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

aco: add support for texturing with clamped LOD 

This is a requirement for the shaderResourceMinLod feature which
allows to clamp LOD. This uses all image_sample_*_cl variants.

All dEQP-VK.glsl.texture_functions.texture*clamp.* pass.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4989>
This commit is contained in:
Samuel Pitoiset 2020-05-11 16:33:14 +02:00 committed by Marge Bot
parent 47a769143b
commit aaf5706aa3
1 changed files with 39 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

40
src/amd/compiler/aco_instruction_selection.cpp
View File

@ -8170,9 +8170,11 @@ void visit_tex(isel_context *ctx, nir_tex_instr *instr)
{ {
Builder bld(ctx->program, ctx->block); Builder bld(ctx->program, ctx->block);
bool has_bias = false, has_lod = false, level_zero = false, has_compare = false, bool has_bias = false, has_lod = false, level_zero = false, has_compare = false,
has_offset = false, has_ddx = false, has_ddy = false, has_derivs = false, has_sample_index = false; has_offset = false, has_ddx = false, has_ddy = false, has_derivs = false, has_sample_index = false,
has_clamped_lod = false;
Temp resource, sampler, fmask_ptr, bias = Temp(), compare = Temp(), sample_index = Temp(), Temp resource, sampler, fmask_ptr, bias = Temp(), compare = Temp(), sample_index = Temp(),
lod = Temp(), offset = Temp(), ddx = Temp(), ddy = Temp(); lod = Temp(), offset = Temp(), ddx = Temp(), ddy = Temp(),
clamped_lod = Temp();
std::vector<Temp> coords; std::vector<Temp> coords;
std::vector<Temp> derivs; std::vector<Temp> derivs;
nir_const_value *sample_index_cv = NULL; nir_const_value *sample_index_cv = NULL;
@ -8208,6 +8210,10 @@ void visit_tex(isel_context *ctx, nir_tex_instr *instr)
} }
break; break;
} }
case nir_tex_src_min_lod:
clamped_lod = get_ssa_temp(ctx, instr->src[i].src.ssa);
has_clamped_lod = true;
break;
case nir_tex_src_comparator: case nir_tex_src_comparator:
if (instr->is_shadow) { if (instr->is_shadow) {
compare = get_ssa_temp(ctx, instr->src[i].src.ssa); compare = get_ssa_temp(ctx, instr->src[i].src.ssa);
@ -8608,6 +8614,8 @@ void visit_tex(isel_context *ctx, nir_tex_instr *instr)
args.emplace_back(sample_index); args.emplace_back(sample_index);
if (has_lod) if (has_lod)
args.emplace_back(lod); args.emplace_back(lod);
if (has_clamped_lod)
args.emplace_back(clamped_lod);
Temp arg = bld.tmp(RegClass(RegType::vgpr, args.size())); Temp arg = bld.tmp(RegClass(RegType::vgpr, args.size()));
aco_ptr<Instruction> vec{create_instruction<Pseudo_instruction>(aco_opcode::p_create_vector, Format::PSEUDO, args.size(), 1)}; aco_ptr<Instruction> vec{create_instruction<Pseudo_instruction>(aco_opcode::p_create_vector, Format::PSEUDO, args.size(), 1)};
@ -8652,7 +8660,21 @@ void visit_tex(isel_context *ctx, nir_tex_instr *instr)
// TODO: would be better to do this by adding offsets, but needs the opcodes ordered. // TODO: would be better to do this by adding offsets, but needs the opcodes ordered.
aco_opcode opcode = aco_opcode::image_sample; aco_opcode opcode = aco_opcode::image_sample;
if (has_offset) { /* image_sample_*_o */ if (has_offset) { /* image_sample_*_o */
if (has_clamped_lod) {
if (has_compare) { if (has_compare) {
opcode = aco_opcode::image_sample_c_cl_o;
if (has_derivs)
opcode = aco_opcode::image_sample_c_d_cl_o;
if (has_bias)
opcode = aco_opcode::image_sample_c_b_cl_o;
} else {
opcode = aco_opcode::image_sample_cl_o;
if (has_derivs)
opcode = aco_opcode::image_sample_d_cl_o;
if (has_bias)
opcode = aco_opcode::image_sample_b_cl_o;
}
} else if (has_compare) {
opcode = aco_opcode::image_sample_c_o; opcode = aco_opcode::image_sample_c_o;
if (has_derivs) if (has_derivs)
opcode = aco_opcode::image_sample_c_d_o; opcode = aco_opcode::image_sample_c_d_o;
@ -8673,6 +8695,20 @@ void visit_tex(isel_context *ctx, nir_tex_instr *instr)
if (has_lod) if (has_lod)
opcode = aco_opcode::image_sample_l_o; opcode = aco_opcode::image_sample_l_o;
} }
} else if (has_clamped_lod) { /* image_sample_*_cl */
if (has_compare) {
opcode = aco_opcode::image_sample_c_cl;
if (has_derivs)
opcode = aco_opcode::image_sample_c_d_cl;
if (has_bias)
opcode = aco_opcode::image_sample_c_b_cl;
} else {
opcode = aco_opcode::image_sample_cl;
if (has_derivs)
opcode = aco_opcode::image_sample_d_cl;
if (has_bias)
opcode = aco_opcode::image_sample_b_cl;
}
} else { /* no offset */ } else { /* no offset */
if (has_compare) { if (has_compare) {
opcode = aco_opcode::image_sample_c; opcode = aco_opcode::image_sample_c;