i965: Do NIR shader cloning in the caller.

This moves nir_shader_clone() to the driver-specific compile function,
rather than the shared src/intel/compiler code.  This allows i965 to do
key-specific passes before calling brw_compile_*.  Vulkan should not
need this cloning as it doesn't compile multiple variants.

We do need to continue cloning in the compute shader code because we
lower various things in NIR based on the SIMD width.

Reviewed-by: Alejandro Piñeiro <[email protected]>

6 files changed, 10 insertions, 15 deletions

diff --git a/src/intel/compiler/brw_compiler.h b/src/intel/compiler/brw_compiler.h
index 8adc1ff8115..e4f4d83c8e8 100644
--- a/src/intel/compiler/brw_compiler.h
+++ b/src/intel/compiler/brw_compiler.h
@@ -1239,7 +1239,7 @@ brw_compile_vs(const struct brw_compiler *compiler, void *log_data,
                void *mem_ctx,
                const struct brw_vs_prog_key *key,
                struct brw_vs_prog_data *prog_data,
-               const struct nir_shader *shader,
+               struct nir_shader *shader,
                int shader_time_index,
                char **error_str);
 
@@ -1254,7 +1254,7 @@ brw_compile_tcs(const struct brw_compiler *compiler,
                 void *mem_ctx,
                 const struct brw_tcs_prog_key *key,
                 struct brw_tcs_prog_data *prog_data,
-                const struct nir_shader *nir,
+                struct nir_shader *nir,
                 int shader_time_index,
                 char **error_str);
 
@@ -1269,7 +1269,7 @@ brw_compile_tes(const struct brw_compiler *compiler, void *log_data,
                 const struct brw_tes_prog_key *key,
                 const struct brw_vue_map *input_vue_map,
                 struct brw_tes_prog_data *prog_data,
-                const struct nir_shader *shader,
+                struct nir_shader *shader,
                 struct gl_program *prog,
                 int shader_time_index,
                 char **error_str);
@@ -1284,7 +1284,7 @@ brw_compile_gs(const struct brw_compiler *compiler, void *log_data,
                void *mem_ctx,
                const struct brw_gs_prog_key *key,
                struct brw_gs_prog_data *prog_data,
-               const struct nir_shader *shader,
+               struct nir_shader *shader,
                struct gl_program *prog,
                int shader_time_index,
                char **error_str);
@@ -1331,7 +1331,7 @@ brw_compile_fs(const struct brw_compiler *compiler, void *log_data,
                void *mem_ctx,
                const struct brw_wm_prog_key *key,
                struct brw_wm_prog_data *prog_data,
-               const struct nir_shader *shader,
+               struct nir_shader *shader,
                struct gl_program *prog,
                int shader_time_index8,
                int shader_time_index16,
diff --git a/src/intel/compiler/brw_fs.cpp b/src/intel/compiler/brw_fs.cpp
index 3e083723471..aa29c8a4deb 100644
--- a/src/intel/compiler/brw_fs.cpp
+++ b/src/intel/compiler/brw_fs.cpp
@@ -7119,7 +7119,7 @@ brw_compile_fs(const struct brw_compiler *compiler, void *log_data,
                void *mem_ctx,
                const struct brw_wm_prog_key *key,
                struct brw_wm_prog_data *prog_data,
-               const nir_shader *src_shader,
+               nir_shader *shader,
                struct gl_program *prog,
                int shader_time_index8, int shader_time_index16,
                int shader_time_index32, bool allow_spilling,
@@ -7128,7 +7128,6 @@ brw_compile_fs(const struct brw_compiler *compiler, void *log_data,
 {
    const struct gen_device_info *devinfo = compiler->devinfo;
 
-   nir_shader *shader = nir_shader_clone(mem_ctx, src_shader);
    shader = brw_nir_apply_sampler_key(shader, compiler, &key->tex, true);
    brw_nir_lower_fs_inputs(shader, devinfo, key);
    brw_nir_lower_fs_outputs(shader);
diff --git a/src/intel/compiler/brw_shader.cpp b/src/intel/compiler/brw_shader.cpp
index 798c799a828..b77bd798d17 100644
--- a/src/intel/compiler/brw_shader.cpp
+++ b/src/intel/compiler/brw_shader.cpp
@@ -1189,7 +1189,7 @@ brw_compile_tes(const struct brw_compiler *compiler,
                 const struct brw_tes_prog_key *key,
                 const struct brw_vue_map *input_vue_map,
                 struct brw_tes_prog_data *prog_data,
-                const nir_shader *src_shader,
+                nir_shader *nir,
                 struct gl_program *prog,
                 int shader_time_index,
                 char **error_str)
@@ -1198,7 +1198,6 @@ brw_compile_tes(const struct brw_compiler *compiler,
    const bool is_scalar = compiler->scalar_stage[MESA_SHADER_TESS_EVAL];
    const unsigned *assembly;
 
-   nir_shader *nir = nir_shader_clone(mem_ctx, src_shader);
    nir->info.inputs_read = key->inputs_read;
    nir->info.patch_inputs_read = key->patch_inputs_read;
 
diff --git a/src/intel/compiler/brw_vec4.cpp b/src/intel/compiler/brw_vec4.cpp
index 74a4d09fc79..6ed07b72f31 100644
--- a/src/intel/compiler/brw_vec4.cpp
+++ b/src/intel/compiler/brw_vec4.cpp
@@ -2828,12 +2828,11 @@ brw_compile_vs(const struct brw_compiler *compiler, void *log_data,
                void *mem_ctx,
                const struct brw_vs_prog_key *key,
                struct brw_vs_prog_data *prog_data,
-               const nir_shader *src_shader,
+               nir_shader *shader,
                int shader_time_index,
                char **error_str)
 {
    const bool is_scalar = compiler->scalar_stage[MESA_SHADER_VERTEX];
-   nir_shader *shader = nir_shader_clone(mem_ctx, src_shader);
    shader = brw_nir_apply_sampler_key(shader, compiler, &key->tex, is_scalar);
 
    const unsigned *assembly = NULL;
diff --git a/src/intel/compiler/brw_vec4_gs_visitor.cpp b/src/intel/compiler/brw_vec4_gs_visitor.cpp
index c4c876a0d03..63ff27e5e08 100644
--- a/src/intel/compiler/brw_vec4_gs_visitor.cpp
+++ b/src/intel/compiler/brw_vec4_gs_visitor.cpp
@@ -615,7 +615,7 @@ brw_compile_gs(const struct brw_compiler *compiler, void *log_data,
                void *mem_ctx,
                const struct brw_gs_prog_key *key,
                struct brw_gs_prog_data *prog_data,
-               const nir_shader *src_shader,
+               nir_shader *shader,
                struct gl_program *prog,
                int shader_time_index,
                char **error_str)
@@ -625,7 +625,6 @@ brw_compile_gs(const struct brw_compiler *compiler, void *log_data,
    c.key = *key;
 
    const bool is_scalar = compiler->scalar_stage[MESA_SHADER_GEOMETRY];
-   nir_shader *shader = nir_shader_clone(mem_ctx, src_shader);
 
    /* The GLSL linker will have already matched up GS inputs and the outputs
     * of prior stages.  The driver does extend VS outputs in some cases, but
diff --git a/src/intel/compiler/brw_vec4_tcs.cpp b/src/intel/compiler/brw_vec4_tcs.cpp
index 18c6dae8c10..be0969dda12 100644
--- a/src/intel/compiler/brw_vec4_tcs.cpp
+++ b/src/intel/compiler/brw_vec4_tcs.cpp
@@ -378,7 +378,7 @@ brw_compile_tcs(const struct brw_compiler *compiler,
                 void *mem_ctx,
                 const struct brw_tcs_prog_key *key,
                 struct brw_tcs_prog_data *prog_data,
-                const nir_shader *src_shader,
+                nir_shader *nir,
                 int shader_time_index,
                 char **error_str)
 {
@@ -387,7 +387,6 @@ brw_compile_tcs(const struct brw_compiler *compiler,
    const bool is_scalar = compiler->scalar_stage[MESA_SHADER_TESS_CTRL];
    const unsigned *assembly;
 
-   nir_shader *nir = nir_shader_clone(mem_ctx, src_shader);
    nir->info.outputs_written = key->outputs_written;
    nir->info.patch_outputs_written = key->patch_outputs_written;