intel/fs,vec4: Stuff the constant data from NIR in the end of the program
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6244>
diff --git a/src/intel/compiler/brw_compiler.h b/src/intel/compiler/brw_compiler.h
index 8df2ee5..679a787 100644
--- a/src/intel/compiler/brw_compiler.h
+++ b/src/intel/compiler/brw_compiler.h
@@ -685,6 +685,9 @@
unsigned program_size;
+ unsigned const_data_size;
+ unsigned const_data_offset;
+
/** Does this program pull from any UBO or other constant buffers? */
bool has_ubo_pull;
diff --git a/src/intel/compiler/brw_fs.cpp b/src/intel/compiler/brw_fs.cpp
index 388fd8d..460a365 100644
--- a/src/intel/compiler/brw_fs.cpp
+++ b/src/intel/compiler/brw_fs.cpp
@@ -8784,6 +8784,8 @@
stats = stats ? stats + 1 : NULL;
}
+ g.add_const_data(shader->constant_data, shader->constant_data_size);
+
delete v8;
delete v16;
delete v32;
@@ -9167,6 +9169,8 @@
v->performance_analysis.require(), stats);
}
+ g.add_const_data(src_shader->constant_data, src_shader->constant_data_size);
+
ret = g.get_assembly();
delete v8;
diff --git a/src/intel/compiler/brw_fs.h b/src/intel/compiler/brw_fs.h
index c24438a..0843f6e 100644
--- a/src/intel/compiler/brw_fs.h
+++ b/src/intel/compiler/brw_fs.h
@@ -478,6 +478,7 @@
struct shader_stats shader_stats,
const brw::performance &perf,
struct brw_compile_stats *stats);
+ void add_const_data(void *data, unsigned size);
const unsigned *get_assembly();
private:
diff --git a/src/intel/compiler/brw_fs_generator.cpp b/src/intel/compiler/brw_fs_generator.cpp
index df35744..74c6cd3 100644
--- a/src/intel/compiler/brw_fs_generator.cpp
+++ b/src/intel/compiler/brw_fs_generator.cpp
@@ -2595,6 +2595,16 @@
return start_offset;
}
+void
+fs_generator::add_const_data(void *data, unsigned size)
+{
+ assert(prog_data->const_data_size == 0);
+ if (size > 0) {
+ prog_data->const_data_size = size;
+ prog_data->const_data_offset = brw_append_data(p, data, size, 32);
+ }
+}
+
const unsigned *
fs_generator::get_assembly()
{
diff --git a/src/intel/compiler/brw_shader.cpp b/src/intel/compiler/brw_shader.cpp
index e78f2cb..3219a01 100644
--- a/src/intel/compiler/brw_shader.cpp
+++ b/src/intel/compiler/brw_shader.cpp
@@ -1374,6 +1374,8 @@
g.generate_code(v.cfg, 8, v.shader_stats,
v.performance_analysis.require(), stats);
+ g.add_const_data(nir->constant_data, nir->constant_data_size);
+
assembly = g.get_assembly();
} else {
brw::vec4_tes_visitor v(compiler, log_data, key, prog_data,
diff --git a/src/intel/compiler/brw_vec4.cpp b/src/intel/compiler/brw_vec4.cpp
index 864ece7..fe7db43 100644
--- a/src/intel/compiler/brw_vec4.cpp
+++ b/src/intel/compiler/brw_vec4.cpp
@@ -3001,6 +3001,7 @@
}
g.generate_code(v.cfg, 8, v.shader_stats,
v.performance_analysis.require(), stats);
+ g.add_const_data(shader->constant_data, shader->constant_data_size);
assembly = g.get_assembly();
}
diff --git a/src/intel/compiler/brw_vec4_generator.cpp b/src/intel/compiler/brw_vec4_generator.cpp
index 43fcdf7..e9142c2 100644
--- a/src/intel/compiler/brw_vec4_generator.cpp
+++ b/src/intel/compiler/brw_vec4_generator.cpp
@@ -2281,5 +2281,12 @@
generate_code(p, compiler, log_data, nir, prog_data, cfg, perf, stats);
+ assert(prog_data->base.const_data_size == 0);
+ if (nir->constant_data_size > 0) {
+ prog_data->base.const_data_size = nir->constant_data_size;
+ prog_data->base.const_data_offset =
+ brw_append_data(p, nir->constant_data, nir->constant_data_size, 32);
+ }
+
return brw_get_program(p, &prog_data->base.program_size);
}
diff --git a/src/intel/compiler/brw_vec4_gs_visitor.cpp b/src/intel/compiler/brw_vec4_gs_visitor.cpp
index 58f6f12..b286bfd 100644
--- a/src/intel/compiler/brw_vec4_gs_visitor.cpp
+++ b/src/intel/compiler/brw_vec4_gs_visitor.cpp
@@ -867,6 +867,7 @@
}
g.generate_code(v.cfg, 8, v.shader_stats,
v.performance_analysis.require(), stats);
+ g.add_const_data(shader->constant_data, shader->constant_data_size);
return g.get_assembly();
}
}
diff --git a/src/intel/compiler/brw_vec4_tcs.cpp b/src/intel/compiler/brw_vec4_tcs.cpp
index e9a6880..29c0d9e 100644
--- a/src/intel/compiler/brw_vec4_tcs.cpp
+++ b/src/intel/compiler/brw_vec4_tcs.cpp
@@ -483,6 +483,8 @@
g.generate_code(v.cfg, 8, v.shader_stats,
v.performance_analysis.require(), stats);
+ g.add_const_data(nir->constant_data, nir->constant_data_size);
+
assembly = g.get_assembly();
} else {
vec4_tcs_visitor v(compiler, log_data, key, prog_data,
