virgl: Fix missing fog output in certain cases

For fragment ARB programs, the fog co-ordinate input
can be requested independently of any other part of the
graphics pipeline. Through the use of an OPTION declaration.

This can lead to a case where a fragment shader expects
a fog co-ordinate input, but it is not being supplied
by the preceding shader stage.

This patch will fix it by forcing it to 0 when it detects
the case, which should neutralize the fog color.

Reviewed-by: Gert Wollny <gert.wollny@collabora.com>

src/vrend_renderer.c

@@ -3420,11 +3420,21 @@ static inline void vrend_sync_shader_io(struct vrend_sub_context *sub_ctx,
      break;
    }
 
-   if (next_type != -1 && sub_ctx->shaders[next_type])
+   if (next_type != -1 && sub_ctx->shaders[next_type]) {
       key->output = sub_ctx->shaders[next_type]->sinfo.in;
 
-}
+      if (type == PIPE_SHADER_VERTEX && next_type == PIPE_SHADER_FRAGMENT) {
+         if (sub_ctx->shaders[type]) {
+            uint32_t fog_input = sub_ctx->shaders[next_type]->sinfo.fog_input_mask;
+            uint32_t fog_output = sub_ctx->shaders[type]->sinfo.fog_output_mask;
 
+            //We only want to issue the fixup for inputs not fed by the outputs of the
+            //previous stage
+            key->vs.fog_fixup_mask = (fog_input ^ fog_output) & fog_input;
+         }
+      }
+   }
+}
 
 static inline void vrend_fill_shader_key(struct vrend_sub_context *sub_ctx,
                                          struct vrend_shader_selector *sel,

src/vrend_shader.c

@@ -219,6 +219,9 @@ struct dump_ctx {
    struct vrend_array *sampler_arrays;
    uint32_t num_sampler_arrays;
 
+   uint32_t fog_input_mask;
+   uint32_t fog_output_mask;
+
    int num_consts;
    int num_imm;
    struct immed imm[MAX_IMMEDIATE];
@@ -898,12 +901,18 @@ static int lookup_image_array(const struct dump_ctx *ctx, int index)
 }
 
 static boolean
-iter_inputs(struct tgsi_iterate_context *iter,
+iter_decls(struct tgsi_iterate_context *iter,
            struct tgsi_full_declaration *decl)
 {
    struct dump_ctx *ctx = (struct dump_ctx *)iter;
    switch (decl->Declaration.File) {
    case TGSI_FILE_INPUT:
+      /* Tag used semantic fog inputs */
+      if (decl->Semantic.Name == TGSI_SEMANTIC_FOG) {
+         ctx->fog_input_mask |= (1 << decl->Semantic.Index);
+      }
+
+      if (ctx->prog_type == TGSI_PROCESSOR_FRAGMENT) {
          for (uint32_t j = 0; j < ctx->num_inputs; j++) {
             if (ctx->inputs[j].name == decl->Semantic.Name &&
                 ctx->inputs[j].sid == decl->Semantic.Index &&
@@ -915,6 +924,17 @@ iter_inputs(struct tgsi_iterate_context *iter,
          ctx->inputs[ctx->num_inputs].last = decl->Range.Last;
          ctx->num_inputs++;
       }
+      break;
+
+   case TGSI_FILE_OUTPUT:
+      if (decl->Semantic.Name == TGSI_SEMANTIC_FOG) {
+         ctx->fog_output_mask |= (1 << decl->Semantic.Index);
+      }
+      break;
+
+   default:
+      break;
+   }
    return true;
 }
 
@@ -2039,6 +2059,40 @@ static void emit_clip_dist_movs(const struct dump_ctx *ctx,
    }
 }
 
+static void emit_fog_fixup_hdr(const struct dump_ctx *ctx,
+                               struct vrend_glsl_strbufs *glsl_strbufs)
+{
+   uint32_t fixup_mask = ctx->key->vs.fog_fixup_mask;
+   int semantic;
+   const char *prefix = get_stage_output_name_prefix(TGSI_PROCESSOR_VERTEX);
+
+   while (fixup_mask) {
+      semantic = ffs(fixup_mask) - 1;
+
+      emit_hdrf(glsl_strbufs, "out vec4 %s_f%d;\n", prefix, semantic);
+      fixup_mask &= (~(1 << semantic));
+   }
+}
+
+static void emit_fog_fixup_write(const struct dump_ctx *ctx,
+                                 struct vrend_glsl_strbufs *glsl_strbufs)
+{
+   uint32_t fixup_mask = ctx->key->vs.fog_fixup_mask;
+   int semantic;
+   const char *prefix = get_stage_output_name_prefix(TGSI_PROCESSOR_VERTEX);
+
+   while (fixup_mask) {
+      semantic = ffs(fixup_mask) - 1;
+
+      /*
+      *  Force unwritten fog outputs to 0,0,0,1
+      */
+      emit_buff(glsl_strbufs, "%s_f%d = vec4(0.0, 0.0, 0.0, 1.0);\n",
+               prefix, semantic);
+      fixup_mask &= (~(1 << semantic));
+   }
+}
+
 #define emit_arit_op2(op) emit_buff(&ctx->glsl_strbufs, "%s = %s(%s((%s %s %s))%s);\n", dsts[0], get_string(dinfo.dstconv), get_string(dinfo.dtypeprefix), srcs[0], op, srcs[1], writemask)
 #define emit_op1(op) emit_buff(&ctx->glsl_strbufs, "%s = %s(%s(%s(%s))%s);\n", dsts[0], get_string(dinfo.dstconv), get_string(dinfo.dtypeprefix), op, srcs[0], writemask)
 #define emit_compare(op) emit_buff(&ctx->glsl_strbufs, "%s = %s(%s((%s(%s(%s), %s(%s))))%s);\n", dsts[0], get_string(dinfo.dstconv), get_string(dinfo.dtypeprefix), op, get_string(sinfo.svec4), srcs[0], get_string(sinfo.svec4), srcs[1], writemask)
@@ -2056,6 +2110,9 @@ static void handle_vertex_proc_exit(const struct dump_ctx *ctx,
 
     if (!ctx->key->gs_present && !ctx->key->tes_present)
        emit_prescale(glsl_strbufs);
+
+    if (ctx->key->vs.fog_fixup_mask)
+       emit_fog_fixup_write(ctx, glsl_strbufs);
 }
 
 static void emit_fragment_logicop(const struct dump_ctx *ctx,
@@ -6323,6 +6380,9 @@ static void emit_ios_vs(const struct dump_ctx *ctx,
       }
    }
 
+   if (ctx->key->vs.fog_fixup_mask)
+      emit_fog_fixup_hdr(ctx, glsl_strbufs);
+
    emit_winsys_correction(glsl_strbufs);
 
    if (ctx->has_clipvertex) {
@@ -6880,6 +6940,8 @@ static void fill_sinfo(const struct dump_ctx *ctx, struct vrend_shader_info *sin
    sinfo->images_used_mask = ctx->images_used_mask;
    sinfo->num_consts = ctx->num_consts;
    sinfo->ubo_used_mask = ctx->ubo_used_mask;
+   sinfo->fog_input_mask = ctx->fog_input_mask;
+   sinfo->fog_output_mask = ctx->fog_output_mask;
 
    sinfo->ssbo_used_mask = ctx->ssbo_used_mask;
 
@@ -6998,8 +7060,9 @@ bool vrend_convert_shader(const struct vrend_context *rctx,
    memset(&ctx, 0, sizeof(struct dump_ctx));
 
    /* First pass to deal with edge cases. */
-   if (ctx.prog_type == TGSI_PROCESSOR_FRAGMENT)
+   if (ctx.prog_type == TGSI_PROCESSOR_FRAGMENT ||
-      ctx.iter.iterate_declaration = iter_inputs;
+       ctx.prog_type == TGSI_PROCESSOR_VERTEX)
+      ctx.iter.iterate_declaration = iter_decls;
    ctx.iter.iterate_instruction = analyze_instruction;
    bret = tgsi_iterate_shader(tokens, &ctx.iter);
    if (bret == false)

src/vrend_shader.h

@@ -116,6 +116,8 @@ struct vrend_shader_info {
    uint32_t shadow_samp_mask;
    uint32_t attrib_input_mask;
    uint32_t fs_blend_equation_advanced;
+   uint32_t fog_input_mask;
+   uint32_t fog_output_mask;
 
    int num_consts;
    int num_inputs;
@@ -155,6 +157,7 @@ struct vrend_shader_key {
       struct {
          uint32_t attrib_signed_int_bitmask;
          uint32_t attrib_unsigned_int_bitmask;
+         uint32_t fog_fixup_mask;
       } vs;
    };