[Mesa-dev] [RFC PATCH] nir: extend nir_shader_compiler_options with is_scalar

[Christian Gmeiner]

As not every (upcoming) backend compiler is happy with
nir_lower_xxx_to_scalar lowerings do them only if the backend
is scalar (and not vec4) based.

Signed-off-by: Christian Gmeiner <christian.gmeiner at gmail.com>
---
 src/amd/vulkan/radv_shader.c                | 1 +
 src/broadcom/compiler/nir_to_vir.c          | 1 +
 src/compiler/nir/nir.h                      | 3 +++
 src/gallium/drivers/freedreno/ir3/ir3_nir.c | 1 +
 src/gallium/drivers/radeonsi/si_get.c       | 1 +
 src/gallium/drivers/vc4/vc4_program.c       | 1 +
 src/intel/compiler/brw_compiler.c           | 1 +
 src/mesa/state_tracker/st_glsl_to_nir.cpp   | 8 ++++++--
 8 files changed, 15 insertions(+), 2 deletions(-)

diff --git a/src/amd/vulkan/radv_shader.c b/src/amd/vulkan/radv_shader.c
index 171802eede..22f61cb23d 100644
--- a/src/amd/vulkan/radv_shader.c
+++ b/src/amd/vulkan/radv_shader.c
@@ -49,6 +49,7 @@
 #include "util/string_buffer.h"
 
 static const struct nir_shader_compiler_options nir_options = {
+	.is_scalar = true,
 	.vertex_id_zero_based = true,
 	.lower_scmp = true,
 	.lower_flrp32 = true,
diff --git a/src/broadcom/compiler/nir_to_vir.c b/src/broadcom/compiler/nir_to_vir.c
index 595689d244..b4f29554ac 100644
--- a/src/broadcom/compiler/nir_to_vir.c
+++ b/src/broadcom/compiler/nir_to_vir.c
@@ -1876,6 +1876,7 @@ nir_to_vir(struct v3d_compile *c)
 }
 
 const nir_shader_compiler_options v3d_nir_options = {
+        .is_scalar = true,
         .lower_all_io_to_temps = true,
         .lower_extract_byte = true,
         .lower_extract_word = true,
diff --git a/src/compiler/nir/nir.h b/src/compiler/nir/nir.h
index c676331000..4796e3e88c 100644
--- a/src/compiler/nir/nir.h
+++ b/src/compiler/nir/nir.h
@@ -1883,6 +1883,9 @@ typedef struct nir_shader_compiler_options {
     */
    bool native_integers;
 
+   /* Indicates that the hw makes use of a scalar ISA */
+   bool is_scalar;
+
    /* Indicates that the driver only has zero-based vertex id */
    bool vertex_id_zero_based;
 
diff --git a/src/gallium/drivers/freedreno/ir3/ir3_nir.c b/src/gallium/drivers/freedreno/ir3/ir3_nir.c
index 0b637bb99b..a921c14565 100644
--- a/src/gallium/drivers/freedreno/ir3/ir3_nir.c
+++ b/src/gallium/drivers/freedreno/ir3/ir3_nir.c
@@ -47,6 +47,7 @@ static const nir_shader_compiler_options options = {
 		.lower_ldexp = true,
 		.fuse_ffma = true,
 		.native_integers = true,
+		.is_scalar = true,
 		.vertex_id_zero_based = true,
 		.lower_extract_byte = true,
 		.lower_extract_word = true,
diff --git a/src/gallium/drivers/radeonsi/si_get.c b/src/gallium/drivers/radeonsi/si_get.c
index e9657602f3..621ff86814 100644
--- a/src/gallium/drivers/radeonsi/si_get.c
+++ b/src/gallium/drivers/radeonsi/si_get.c
@@ -510,6 +510,7 @@ static const struct nir_shader_compiler_options nir_options = {
 	.lower_unpack_unorm_4x8 = true,
 	.lower_extract_byte = true,
 	.lower_extract_word = true,
+	.is_scalar = true,
 	.max_unroll_iterations = 32,
 	.native_integers = true,
 };
diff --git a/src/gallium/drivers/vc4/vc4_program.c b/src/gallium/drivers/vc4/vc4_program.c
index be80a851d2..1d7d0bdb0a 100644
--- a/src/gallium/drivers/vc4/vc4_program.c
+++ b/src/gallium/drivers/vc4/vc4_program.c
@@ -2191,6 +2191,7 @@ static const nir_shader_compiler_options nir_options = {
         .lower_ldexp = true,
         .lower_negate = true,
         .native_integers = true,
+        .is_scalar = true,
         .max_unroll_iterations = 32,
 };
 
diff --git a/src/intel/compiler/brw_compiler.c b/src/intel/compiler/brw_compiler.c
index 9340317492..9794362f5d 100644
--- a/src/intel/compiler/brw_compiler.c
+++ b/src/intel/compiler/brw_compiler.c
@@ -57,6 +57,7 @@
    .lower_unpack_snorm_4x8 = true,                                            \
    .lower_unpack_unorm_2x16 = true,                                           \
    .lower_unpack_unorm_4x8 = true,                                            \
+   .is_scalar = true,                                                         \
    .vs_inputs_dual_locations = true,                                          \
    .max_unroll_iterations = 32
 
diff --git a/src/mesa/state_tracker/st_glsl_to_nir.cpp b/src/mesa/state_tracker/st_glsl_to_nir.cpp
index 0bd9c4e4e3..9fd64166ae 100644
--- a/src/mesa/state_tracker/st_glsl_to_nir.cpp
+++ b/src/mesa/state_tracker/st_glsl_to_nir.cpp
@@ -262,13 +262,17 @@ st_nir_assign_uniform_locations(struct gl_program *prog,
 static void
 st_nir_opts(nir_shader *nir)
 {
+   const struct nir_shader_compiler_options *options = nir->options;
    bool progress;
    do {
       progress = false;
 
       NIR_PASS_V(nir, nir_lower_vars_to_ssa);
-      NIR_PASS_V(nir, nir_lower_alu_to_scalar);
-      NIR_PASS_V(nir, nir_lower_phis_to_scalar);
+
+      if (options->is_scalar) {
+         NIR_PASS_V(nir, nir_lower_alu_to_scalar);
+         NIR_PASS_V(nir, nir_lower_phis_to_scalar);
+      }
 
       NIR_PASS_V(nir, nir_lower_64bit_pack);
       NIR_PASS(progress, nir, nir_copy_prop);
-- 
2.14.3