pan/bi: Allow specializing bifrost_nir_options by arch

We need different settings for Bifrost and Valhall. Keeping everything static
simplifies lifetimes.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20906>

diff --git a/src/panfrost/compiler/bifrost_compile.h b/src/panfrost/compiler/bifrost_compile.h
index 58d265b..86f966f 100644 (file)
--- a/src/panfrost/compiler/bifrost_compile.h
+++ b/src/panfrost/compiler/bifrost_compile.h
@@ -33,67 +33,69 @@ void bifrost_compile_shader_nir(nir_shader *nir,
                                 struct util_dynarray *binary,
                                 struct pan_shader_info *info);
 
-static const nir_shader_compiler_options bifrost_nir_options = {
-   .lower_scmp = true,
-   .lower_flrp16 = true,
-   .lower_flrp32 = true,
-   .lower_flrp64 = true,
-   .lower_ffract = true,
-   .lower_fmod = true,
-   .lower_fdiv = true,
-   .lower_isign = true,
-   .lower_find_lsb = true,
-   .lower_ifind_msb = true,
-   .lower_fdph = true,
-   .lower_fsqrt = true,
+#define DEFINE_OPTIONS(arch)                                                   \
+   static const nir_shader_compiler_options bifrost_nir_options_v##arch = {    \
+      .lower_scmp = true,                                                      \
+      .lower_flrp16 = true,                                                    \
+      .lower_flrp32 = true,                                                    \
+      .lower_flrp64 = true,                                                    \
+      .lower_ffract = true,                                                    \
+      .lower_fmod = true,                                                      \
+      .lower_fdiv = true,                                                      \
+      .lower_isign = true,                                                     \
+      .lower_find_lsb = true,                                                  \
+      .lower_ifind_msb = true,                                                 \
+      .lower_fdph = true,                                                      \
+      .lower_fsqrt = true,                                                     \
+                                                                               \
+      .lower_fsign = true,                                                     \
+                                                                               \
+      .lower_bitfield_insert_to_shifts = true,                                 \
+      .lower_bitfield_extract_to_shifts = true,                                \
+      .lower_insert_byte = true,                                               \
+      .lower_rotate = true,                                                    \
+                                                                               \
+      .lower_pack_half_2x16 = true,                                            \
+      .lower_pack_unorm_2x16 = true,                                           \
+      .lower_pack_snorm_2x16 = true,                                           \
+      .lower_pack_unorm_4x8 = true,                                            \
+      .lower_pack_snorm_4x8 = true,                                            \
+      .lower_unpack_half_2x16 = true,                                          \
+      .lower_unpack_unorm_2x16 = true,                                         \
+      .lower_unpack_snorm_2x16 = true,                                         \
+      .lower_unpack_unorm_4x8 = true,                                          \
+      .lower_unpack_snorm_4x8 = true,                                          \
+      .lower_pack_split = true,                                                \
+                                                                               \
+      .lower_doubles_options =                                                 \
+         nir_lower_dmod, /* TODO: Don't lower supported 64-bit operations */   \
+      .lower_int64_options = ~0, /* TODO: Use IMULD on v7 */                   \
+      .lower_mul_high = true,                                                  \
+      .lower_fisnormal = true,                                                 \
+      .lower_uadd_carry = true,                                                \
+      .lower_usub_borrow = true,                                               \
+                                                                               \
+      .has_fsub = true,                                                        \
+      .has_isub = true,                                                        \
+      .vectorize_io = true,                                                    \
+      .vectorize_vec2_16bit = true,                                            \
+      .fuse_ffma16 = true,                                                     \
+      .fuse_ffma32 = true,                                                     \
+      .fuse_ffma64 = true,                                                     \
+      .use_interpolated_input_intrinsics = true,                               \
+                                                                               \
+      .lower_uniforms_to_ubo = true,                                           \
+                                                                               \
+      .has_cs_global_id = true,                                                \
+      .lower_cs_local_index_to_id = true,                                      \
+      .max_unroll_iterations = 32,                                             \
+      .force_indirect_unrolling =                                              \
+         (nir_var_shader_in | nir_var_shader_out | nir_var_function_temp),     \
+      .force_indirect_unrolling_sampler = true,                                \
+                                                                               \
+      .use_scoped_barrier = true,                                              \
+   };
 
-   .lower_fsign = true,
-
-   .lower_bitfield_insert_to_shifts = true,
-   .lower_bitfield_extract_to_shifts = true,
-   .lower_insert_byte = true,
-   .lower_rotate = true,
-
-   .lower_pack_half_2x16 = true,
-   .lower_pack_unorm_2x16 = true,
-   .lower_pack_snorm_2x16 = true,
-   .lower_pack_unorm_4x8 = true,
-   .lower_pack_snorm_4x8 = true,
-   .lower_unpack_half_2x16 = true,
-   .lower_unpack_unorm_2x16 = true,
-   .lower_unpack_snorm_2x16 = true,
-   .lower_unpack_unorm_4x8 = true,
-   .lower_unpack_snorm_4x8 = true,
-   .lower_pack_split = true,
-
-   .lower_doubles_options = nir_lower_dmod,
-   /* TODO: Don't lower supported 64-bit operations */
-   .lower_int64_options = ~0,
-   /* TODO: Use IMULD on v7 */
-   .lower_mul_high = true,
-   .lower_fisnormal = true,
-   .lower_uadd_carry = true,
-   .lower_usub_borrow = true,
-
-   .has_fsub = true,
-   .has_isub = true,
-   .vectorize_io = true,
-   .vectorize_vec2_16bit = true,
-   .fuse_ffma16 = true,
-   .fuse_ffma32 = true,
-   .fuse_ffma64 = true,
-   .use_interpolated_input_intrinsics = true,
-
-   .lower_uniforms_to_ubo = true,
-
-   .has_cs_global_id = true,
-   .lower_cs_local_index_to_id = true,
-   .max_unroll_iterations = 32,
-   .force_indirect_unrolling =
-      (nir_var_shader_in | nir_var_shader_out | nir_var_function_temp),
-   .force_indirect_unrolling_sampler = true,
-
-   .use_scoped_barrier = true,
-};
+DEFINE_OPTIONS(6);
 
 #endif

diff --git a/src/panfrost/lib/pan_shader.c b/src/panfrost/lib/pan_shader.c
index 4386edd..1fe8491 100644 (file)
--- a/src/panfrost/lib/pan_shader.c
+++ b/src/panfrost/lib/pan_shader.c
@@ -36,7 +36,7 @@ const nir_shader_compiler_options *
 GENX(pan_shader_get_compiler_options)(void)
 {
 #if PAN_ARCH >= 6
-   return &bifrost_nir_options;
+   return &bifrost_nir_options_v6;
 #else
    return &midgard_nir_options;
 #endif