radeonsi: support 8 non-ClipVertex clip planes instead of 6

If there are more than 6 planes without gl_ClipVertex and gl_ClipDistance,
add "gl_ClipVertex = gl_Position;" to support up to 8 planes.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35529>

src/amd/ci/radeonsi-mendocino-fails.txt

@@ -411,7 +411,6 @@ spec@khr_texture_compression_astc@miptree-gles srgb-fp,Fail
 spec@khr_texture_compression_astc@miptree-gles srgb-fp@sRGB decode full precision,Fail
 spec@khr_texture_compression_astc@sliced-3d-miptree-gles srgb-fp,Fail
 spec@khr_texture_compression_astc@sliced-3d-miptree-gles srgb-fp@sRGB decode full precision,Fail
-spec@!opengl 1.0@gl-1.0-user-clip-all-planes,Fail
 spec@!opengl 1.0@rasterpos,Fail
 spec@!opengl 1.0@rasterpos@glsl_vs_gs_linked,Fail
 spec@!opengl 1.0@rasterpos@glsl_vs_tes_linked,Fail

src/amd/ci/radeonsi-raven-fails.txt

@@ -30,8 +30,6 @@ spec@khr_texture_compression_astc@miptree-gles srgb-fp@sRGB decode full precisio
 spec@khr_texture_compression_astc@sliced-3d-miptree-gles srgb-fp,Fail
 spec@khr_texture_compression_astc@sliced-3d-miptree-gles srgb-fp@sRGB decode full precision,Fail
 
-spec@!opengl 1.0@gl-1.0-user-clip-all-planes,Fail
-
 spec@!opengl 1.1@line-smooth-stipple,Fail
 
 # since Debian 12 (bookworm) uprev

src/amd/ci/radeonsi-stoney-fails.txt

@@ -1,6 +1,5 @@
 glx@glx-multi-window-single-context,Fail
 glx@glx-swap-pixmap-bad,Fail
-spec@!opengl 1.0@gl-1.0-user-clip-all-planes,Fail
 spec@!opengl 1.0@rasterpos,Fail
 spec@!opengl 1.0@rasterpos@glsl_vs_gs_linked,Fail
 spec@!opengl 1.0@rasterpos@glsl_vs_tes_linked,Fail

src/amd/ci/radeonsi-vangogh-fails.txt

@@ -1,4 +1,3 @@
-spec@!opengl 1.0@gl-1.0-user-clip-all-planes,Fail
 spec@!opengl 1.0@rasterpos,Fail
 spec@!opengl 1.0@rasterpos@glsl_vs_gs_linked,Fail
 spec@!opengl 1.0@rasterpos@glsl_vs_tes_linked,Fail

src/gallium/drivers/radeonsi/ci/gfx10-navi10-fail.csv

@@ -7,7 +7,6 @@ glx@glx-visuals-depth -pixmap,Fail
 glx@glx-visuals-stencil -pixmap,Fail
 glx@glx_arb_create_context_no_error@no error,Fail
 glx@glx_ext_no_config_context@no fbconfig,Fail
-spec@!opengl 1.0@gl-1.0-user-clip-all-planes,Fail
 spec@!opengl 1.1@line-smooth-stipple,Fail
 spec@arb_bindless_texture@compiler@samplers@arith-bound-sampler-texture2d.frag,Crash
 spec@arb_program_interface_query@arb_program_interface_query-getprogramresourceindex,Fail

src/gallium/drivers/radeonsi/ci/gfx10_3-navi21-fail.csv

@@ -7,7 +7,6 @@ glx@glx-multi-window-single-context,Fail
 glx@glx-swap-event_interval,Fail
 glx@glx_arb_create_context_no_error@no error,Fail
 glx@glx_ext_no_config_context@no fbconfig,Fail
-spec@!opengl 1.0@gl-1.0-user-clip-all-planes,Fail
 spec@!opengl 1.1@line-smooth-stipple,Fail
 spec@arb_pipeline_statistics_query@arb_pipeline_statistics_query-frag,Fail
 spec@arb_program_interface_query@arb_program_interface_query-getprogramresourceindex,Fail

src/gallium/drivers/radeonsi/ci/gfx11-navi31-fail.csv

@@ -9,7 +9,6 @@ glx@glx-swap-event_interval,Fail
 glx@glx_ext_no_config_context@no fbconfig,Fail
 glx@glx-visuals-depth -pixmap,Warn
 glx@glx-visuals-stencil -pixmap,Fail
-spec@!opengl 1.0@gl-1.0-user-clip-all-planes,Fail
 spec@!opengl 1.1@line-smooth-stipple,Fail
 spec@arb_pipeline_statistics_query@arb_pipeline_statistics_query-frag,Fail
 spec@arb_shader_texture_lod@execution@arb_shader_texture_lod-texgradcube,Fail

src/gallium/drivers/radeonsi/ci/gfx12-gfx1200-fail.csv

@@ -5,7 +5,6 @@ glx@glx-swap-event_interval,Fail
 glx@glx_ext_no_config_context@no fbconfig,Fail
 glx@glx-visuals-depth -pixmap,Fail
 glx@glx-visuals-stencil -pixmap,Fail
-spec@!opengl 1.0@gl-1.0-user-clip-all-planes,Fail
 spec@!opengl 1.1@line-smooth-stipple,Fail
 spec@arb_pipeline_statistics_query@arb_pipeline_statistics_query-clip,Fail
 spec@arb_pipeline_statistics_query@arb_pipeline_statistics_query-frag,Fail

src/gallium/drivers/radeonsi/ci/gfx6-tahiti-fail.csv

@@ -28,7 +28,6 @@ spec@khr_texture_compression_astc@sliced-3d-miptree-gles srgb-fp@sRGB decode ful
 spec@khr_texture_compression_astc@sliced-3d-miptree-gl srgb-fp,Fail
 spec@khr_texture_compression_astc@sliced-3d-miptree-gl srgb-fp@sRGB decode full precision,Fail
 spec@oes_shader_io_blocks@compiler@layout-location-aliasing.vert,Fail
-spec@!opengl 1.0@gl-1.0-user-clip-all-planes,Fail
 spec@!opengl 1.1@line-smooth-stipple,Fail
 
 # These are hw bugs related to GS and primitive restart

src/gallium/drivers/radeonsi/ci/gfx7-hawaii-fail.csv

@@ -15,7 +15,6 @@ glx@glx_ext_import_context@make current- multi process,Fail
 glx@glx_ext_import_context@make current- single process,Fail
 glx@glx_ext_import_context@query context info,Fail
 glx@glx_ext_no_config_context@no fbconfig,Fail
-spec@!opengl 1.0@gl-1.0-user-clip-all-planes,Fail
 spec@!opengl 1.1@line-smooth-stipple,Fail
 spec@!opengl 1.1@texwrap formats bordercolor-swizzled,Fail
 spec@!opengl 1.1@texwrap formats bordercolor-swizzled@GL_R3_G3_B2- swizzled- border color only,Fail

src/gallium/drivers/radeonsi/ci/gfx8-polaris11-fail.csv

@@ -5,7 +5,6 @@ glx@glx-multi-window-single-context,Fail
 glx@glx-swap-event_interval,Fail
 glx@glx-visuals-depth -pixmap,Fail
 glx@glx_ext_no_config_context@no fbconfig,Fail
-spec@!opengl 1.0@gl-1.0-user-clip-all-planes,Fail
 spec@!opengl 1.1@line-smooth-stipple,Fail
 spec@!opengl 1.1@texwrap formats bordercolor-swizzled,Fail
 spec@!opengl 1.1@texwrap formats bordercolor-swizzled@GL_R3_G3_B2- swizzled- border color only,Fail

src/gallium/drivers/radeonsi/ci/gfx9-raven-fail.csv

@@ -19,7 +19,6 @@ spec@arb_texture_query_lod@compiler@built-in-functions@texturequerylod-samplercu
 spec@arb_texture_query_lod@compiler@built-in-functions@texturequerylod-samplercubeshadow.frag,Crash
 spec@arb_uniform_buffer_object@referenced-by-shader,Fail
 glx@glx_ext_no_config_context@no fbconfig,Fail
-spec@!opengl 1.0@gl-1.0-user-clip-all-planes,Fail
 spec@arb_bindless_texture@compiler@samplers@arith-bound-sampler-texture2d.frag,Crash
 spec@arb_program_interface_query@arb_program_interface_query-getprogramresourceindex,Fail
 spec@arb_program_interface_query@arb_program_interface_query-getprogramresourceindex@'vs_input2[1][0]' on GL_PROGRAM_INPUT,Fail

src/gallium/drivers/radeonsi/ci/gfx9-vega20-fail.csv

@@ -18,7 +18,6 @@ glx@glx_ext_import_context@make current- multi process,Fail
 glx@glx_ext_import_context@make current- single process,Fail
 glx@glx_ext_import_context@query context info,Fail
 glx@glx_ext_no_config_context@no fbconfig,Fail
-spec@!opengl 1.0@gl-1.0-user-clip-all-planes,Fail
 spec@!opengl 1.1@line-smooth-stipple,Fail
 spec@!opengl 3.2@gl-3.2-adj-prims cull-back pv-last,Fail
 spec@!opengl 3.2@gl-3.2-adj-prims cull-front pv-last,Fail

src/gallium/drivers/radeonsi/si_shader.c

@@ -994,6 +994,7 @@ static void si_dump_shader_key(const struct si_shader *shader, FILE *f)
         stage == MESA_SHADER_VERTEX) &&
        !key->ge.as_es && !key->ge.as_ls) {
       fprintf(f, "  mono.remove_streamout = 0x%x\n", key->ge.mono.remove_streamout);
+      fprintf(f, "  mono.write_pos_to_clipvertex = %u\n", key->ge.mono.write_pos_to_clipvertex);
       fprintf(f, "  opt.kill_outputs = 0x%" PRIx64 "\n", key->ge.opt.kill_outputs);
       fprintf(f, "  opt.kill_clip_distances = 0x%x\n", key->ge.opt.kill_clip_distances);
       fprintf(f, "  opt.kill_pointsize = %u\n", key->ge.opt.kill_pointsize);
@@ -1121,6 +1122,7 @@ static void si_lower_ngg(struct si_shader *shader, nir_shader *nir,
       .vs_output_param_offset = temp_info->vs_output_param_offset,
       .has_param_exports = shader->info.nr_param_exports,
       .export_clipdist_mask = shader->info.clipdist_mask | shader->info.culldist_mask,
+      .write_pos_to_clipvertex = shader->key.ge.mono.write_pos_to_clipvertex,
       .force_vrs = sel->screen->options.vrs2x2,
       .use_gfx12_xfb_intrinsic = !nir->info.use_aco_amd,
       .skip_viewport_state_culling = sel->info.writes_viewport_index,
@@ -1534,7 +1536,7 @@ static void run_late_optimization_and_lowering_passes(struct si_nir_shader_ctx *
          NIR_PASS_V(nir, ac_nir_lower_legacy_vs,
                     sel->screen->info.gfx_level,
                     shader->info.clipdist_mask | shader->info.culldist_mask,
-                    false, false,
+                    shader->key.ge.mono.write_pos_to_clipvertex, false,
                     ctx->temp_info.vs_output_param_offset,
                     shader->info.nr_param_exports,
                     shader->key.ge.mono.u.vs_export_prim_id,
@@ -1566,6 +1568,7 @@ static void run_late_optimization_and_lowering_passes(struct si_nir_shader_ctx *
          .has_pipeline_stats_query = sel->screen->use_ngg,
          .gfx_level = sel->screen->info.gfx_level,
          .export_clipdist_mask = shader->info.clipdist_mask | shader->info.culldist_mask,
+         .write_pos_to_clipvertex = shader->key.ge.mono.write_pos_to_clipvertex,
          .param_offsets = ctx->temp_info.vs_output_param_offset,
          .has_param_exports = shader->info.nr_param_exports,
          .disable_streamout = !shader->info.num_streamout_vec4s,

src/gallium/drivers/radeonsi/si_shader.h

@@ -635,6 +635,11 @@ struct si_shader_key_ge {
 
       /* Gfx12: When no streamout buffers are bound, streamout must be disabled. */
       unsigned remove_streamout : 1;
+
+      /* Emulate 8 clip planes using the CLIP_VERTEX output. Only 6 clip planes are supported
+       * without this.
+       */
+      unsigned write_pos_to_clipvertex : 1;
    } mono;
 
    /* Optimization flags for asynchronous compilation only. */

src/gallium/drivers/radeonsi/si_shader_variant_info.c

@@ -259,8 +259,10 @@ void si_get_shader_variant_info(struct si_shader *shader,
             shader->info.num_streamout_vec4s = DIV_ROUND_UP(num_streamout_dwords, 4);
          }
 
-         shader->info.clipdist_mask = shader->selector->info.clipdist_mask &
-                                      ~shader->key.ge.opt.kill_clip_distances;
+         shader->info.clipdist_mask =
+            (shader->key.ge.mono.write_pos_to_clipvertex ?
+               SI_USER_CLIP_PLANE_MASK : shader->selector->info.clipdist_mask) &
+            ~shader->key.ge.opt.kill_clip_distances;
          shader->info.culldist_mask = BITFIELD_RANGE(nir->info.clip_distance_array_size,
                                                      nir->info.cull_distance_array_size);
       }

src/gallium/drivers/radeonsi/si_state_shaders.cpp

@@ -2474,6 +2474,16 @@ static void si_get_vs_key_outputs(struct si_context *sctx, struct si_shader_sele
 
    if (sctx->gfx_level >= GFX12)
       key->ge.mono.remove_streamout = key->ge.opt.remove_streamout;
+
+   /* The fixed-func hw only supports 6 clip planes, while gl_ClipVertex supports 8. */
+   if (!vs->info.clipdist_mask &&
+       sctx->queued.named.rasterizer->clip_plane_enable & BITFIELD_RANGE(6, 2)) {
+      key->ge.mono.write_pos_to_clipvertex = 1;
+      key->ge.opt.kill_clip_distances = SI_USER_CLIP_PLANE_MASK &
+                                        ~sctx->queued.named.rasterizer->clip_plane_enable;
+   } else {
+      key->ge.mono.write_pos_to_clipvertex = 0;
+   }
 }
 
 static void si_clear_vs_key_outputs(struct si_context *sctx, struct si_shader_selector *vs,
@@ -2486,6 +2496,7 @@ static void si_clear_vs_key_outputs(struct si_context *sctx, struct si_shader_se
    key->ge.opt.ngg_vs_streamout_num_verts_per_prim = 0;
    key->ge.mono.u.vs_export_prim_id = 0;
    key->ge.mono.remove_streamout = 0;
+   key->ge.mono.write_pos_to_clipvertex = 0;
 }
 
 void si_ps_key_update_framebuffer(struct si_context *sctx)