X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fgallium%2Fdrivers%2Fsvga%2Fsvga_tgsi_decl_sm30.c;h=42a3593ed2aa3a98597a4cabcdfcd8440633f721;hb=ca531aeeb120cdda966120a17cdf50aa14c3129b;hp=6f4087048636b25cde8a2c0a19d3787521ead46c;hpb=49a6f5e68eeb6b23bb040cfc1f93befc2f1eb35a;p=mesa.git

diff --git a/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c b/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c
index 6f408704863..42a3593ed2a 100644
--- a/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c
+++ b/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c
@@ -30,9 +30,16 @@
 
 #include "svga_tgsi_emit.h"
 
-static boolean translate_vs_ps_semantic( struct tgsi_declaration_semantic semantic,
-                                         unsigned *usage,
-                                         unsigned *idx )
+
+/**
+ * Translate TGSI semantic info into SVGA3d semantic info.
+ * This is called for VS outputs and PS inputs only.
+ */
+static boolean
+translate_vs_ps_semantic(struct svga_shader_emitter *emit,
+                         struct tgsi_declaration_semantic semantic,
+                         unsigned *usage,
+                         unsigned *idx)
 {
    switch (semantic.Name) {
    case TGSI_SEMANTIC_POSITION:  
@@ -40,7 +47,6 @@ static boolean translate_vs_ps_semantic( struct tgsi_declaration_semantic semant
       *usage = SVGA3D_DECLUSAGE_POSITION;
       break;
    case TGSI_SEMANTIC_COLOR:     
-
       *idx = semantic.Index;
       *usage = SVGA3D_DECLUSAGE_COLOR;
       break;
@@ -58,13 +64,21 @@ static boolean translate_vs_ps_semantic( struct tgsi_declaration_semantic semant
       *usage = SVGA3D_DECLUSAGE_PSIZE;
       break;
    case TGSI_SEMANTIC_GENERIC:   
-      *idx = semantic.Index + 1; /* texcoord[0] is reserved for fog & position */
+      *idx = svga_remap_generic_index(emit->key.generic_remap_table,
+                                      semantic.Index);
       *usage = SVGA3D_DECLUSAGE_TEXCOORD;
       break;
    case TGSI_SEMANTIC_NORMAL:    
       *idx = semantic.Index;
       *usage = SVGA3D_DECLUSAGE_NORMAL;
       break;
+   case TGSI_SEMANTIC_CLIPDIST:
+   case TGSI_SEMANTIC_CLIPVERTEX:
+      /* XXX at this time we don't support clip distance or clip vertices */
+      debug_warn_once("unsupported clip distance/vertex attribute\n");
+      *usage = SVGA3D_DECLUSAGE_TEXCOORD;
+      *idx = 0;
+      return TRUE;
    default:
       assert(0);
       *usage = SVGA3D_DECLUSAGE_TEXCOORD;
@@ -76,14 +90,24 @@ static boolean translate_vs_ps_semantic( struct tgsi_declaration_semantic semant
 }
 
 
-static boolean emit_decl( struct svga_shader_emitter *emit,
-                          SVGA3dShaderDestToken reg,
-                          unsigned usage, 
-                          unsigned index )
+/**
+ * Emit a PS input (or VS depth/fog output) register declaration.
+ * For example, if usage = SVGA3D_DECLUSAGE_TEXCOORD, reg.num = 1, and
+ * index = 3, we'll emit "dcl_texcoord3 v1".
+ */
+static boolean
+emit_decl(struct svga_shader_emitter *emit,
+          SVGA3dShaderDestToken reg,
+          unsigned usage, 
+          unsigned index)
 {
    SVGA3DOpDclArgs dcl;
    SVGA3dShaderInstToken opcode;
 
+   /* check values against bitfield sizes */
+   assert(index < 16);
+   assert(usage <= SVGA3D_DECLUSAGE_MAX);
+
    opcode = inst_token( SVGA3DOP_DCL );
    dcl.values[0] = 0;
    dcl.values[1] = 0;
@@ -93,16 +117,20 @@ static boolean emit_decl( struct svga_shader_emitter *emit,
    dcl.index = index;
    dcl.values[0] |= 1<<31;
 
-   return  (emit_instruction(emit, opcode) &&
-            svga_shader_emit_dwords( emit, dcl.values, Elements(dcl.values)));
+   return (emit_instruction(emit, opcode) &&
+           svga_shader_emit_dwords( emit, dcl.values, ARRAY_SIZE(dcl.values)));
 }
 
-static boolean emit_vface_decl( struct svga_shader_emitter *emit )
+
+/**
+ * Emit declaration for PS front/back-face input register.
+ */
+static boolean
+emit_vface_decl(struct svga_shader_emitter *emit)
 {
    if (!emit->emitted_vface) {
       SVGA3dShaderDestToken reg =
-         dst_register( SVGA3DREG_MISCTYPE,
-                       SVGA3DMISCREG_FACE );
+         dst_register(SVGA3DREG_MISCTYPE, SVGA3DMISCREG_FACE);
 
       if (!emit_decl( emit, reg, 0, 0 ))
          return FALSE;
@@ -112,13 +140,17 @@ static boolean emit_vface_decl( struct svga_shader_emitter *emit )
    return TRUE;
 }
 
+
+/**
+ * Emit PS input register to pass depth/fog coordinates.
+ * Note that this always goes into texcoord[0].
+ */
 static boolean
 ps30_input_emit_depth_fog( struct svga_shader_emitter *emit,
                            struct src_register *out )
 {
    struct src_register reg;
 
-
    if (emit->emitted_depth_fog) {
       *out = emit->ps_depth_fog;
       return TRUE;
@@ -137,9 +169,15 @@ ps30_input_emit_depth_fog( struct svga_shader_emitter *emit,
    return emit_decl( emit, dst( reg ), SVGA3D_DECLUSAGE_TEXCOORD, 0 );
 }
 
-static boolean ps30_input( struct svga_shader_emitter *emit,
-                           struct tgsi_declaration_semantic semantic,
-                           unsigned idx )
+
+/**
+ * Process a PS input declaration.
+ * We'll emit a declaration like "dcl_texcoord1 v2"
+ */
+static boolean
+ps30_input(struct svga_shader_emitter *emit,
+           struct tgsi_declaration_semantic semantic,
+           unsigned idx)
 {
    unsigned usage, index;
    SVGA3dShaderDestToken reg;
@@ -178,10 +216,10 @@ static boolean ps30_input( struct svga_shader_emitter *emit,
 
       return emit_decl( emit, reg, 0, 0 );
    }
-   else if (emit->key.fkey.light_twoside &&
+   else if (emit->key.fs.light_twoside &&
             (semantic.Name == TGSI_SEMANTIC_COLOR)) {
 
-      if (!translate_vs_ps_semantic( semantic, &usage, &index ))
+      if (!translate_vs_ps_semantic( emit, semantic, &usage, &index ))
          return FALSE;
 
       emit->internal_color_idx[emit->internal_color_count] = idx;
@@ -195,7 +233,7 @@ static boolean ps30_input( struct svga_shader_emitter *emit,
          return FALSE;
 
       semantic.Name = TGSI_SEMANTIC_BCOLOR;
-      if (!translate_vs_ps_semantic( semantic, &usage, &index ))
+      if (!translate_vs_ps_semantic( emit, semantic, &usage, &index ))
          return FALSE;
 
       if (emit->ps30_input_count >= SVGA3D_INPUTREG_MAX)
@@ -234,7 +272,7 @@ static boolean ps30_input( struct svga_shader_emitter *emit,
    }
    else {
 
-      if (!translate_vs_ps_semantic( semantic, &usage, &index ))
+      if (!translate_vs_ps_semantic( emit, semantic, &usage, &index ))
          return FALSE;
 
       if (emit->ps30_input_count >= SVGA3D_INPUTREG_MAX)
@@ -247,9 +285,9 @@ static boolean ps30_input( struct svga_shader_emitter *emit,
          return FALSE;
 
       if (semantic.Name == TGSI_SEMANTIC_GENERIC &&
-          emit->key.fkey.sprite_origin_lower_left &&
+          emit->key.sprite_origin_lower_left &&
           index >= 1 &&
-          emit->key.fkey.tex[index - 1].sprite_texgen) {
+          emit->key.tex[index - 1].sprite_texgen) {
          /* This is a sprite texture coord with lower-left origin.
           * We need to invert the texture T coordinate since the SVGA3D
           * device only supports an upper-left origin.
@@ -278,24 +316,47 @@ static boolean ps30_input( struct svga_shader_emitter *emit,
 }
 
 
-/* PS output registers are the same as 2.0
+/**
+ * Process a PS output declaration.
+ * Note that we don't actually emit a SVGA3DOpDcl for PS outputs.
+ * \idx  register index, such as OUT[2] (not semantic index)
  */
-static boolean ps30_output( struct svga_shader_emitter *emit,
-                            struct tgsi_declaration_semantic semantic,
-                            unsigned idx )
+static boolean
+ps30_output(struct svga_shader_emitter *emit,
+            struct tgsi_declaration_semantic semantic,
+            unsigned idx)
 {
-   SVGA3dShaderDestToken reg;
-
    switch (semantic.Name) {
    case TGSI_SEMANTIC_COLOR:
-      if (emit->unit == PIPE_SHADER_FRAGMENT &&
-          emit->key.fkey.white_fragments) {
-
-         emit->output_map[idx] = dst_register( SVGA3DREG_TEMP,
-                                               emit->nr_hw_temp++ );
-         emit->temp_col[idx] = emit->output_map[idx];
-         emit->true_col[idx] = dst_register( SVGA3DREG_COLOROUT, 
-                                              semantic.Index );
+      if (emit->unit == PIPE_SHADER_FRAGMENT) {
+         if (emit->key.fs.white_fragments) {
+            /* Used for XOR logicop mode */
+            emit->output_map[idx] = dst_register( SVGA3DREG_TEMP,
+                                                  emit->nr_hw_temp++ );
+            emit->temp_color_output[idx] = emit->output_map[idx];
+            emit->true_color_output[idx] = dst_register(SVGA3DREG_COLOROUT, 
+                                                        semantic.Index);
+         }
+         else if (emit->key.fs.write_color0_to_n_cbufs) {
+            /* We'll write color output [0] to all render targets.
+             * Prepare all the output registers here, but only when the
+             * semantic.Index == 0 so we don't do this more than once.
+             */
+            if (semantic.Index == 0) {
+               unsigned i;
+               for (i = 0; i < emit->key.fs.write_color0_to_n_cbufs; i++) {
+                  emit->output_map[idx+i] = dst_register(SVGA3DREG_TEMP,
+                                                     emit->nr_hw_temp++);
+                  emit->temp_color_output[i] = emit->output_map[idx+i];
+                  emit->true_color_output[i] = dst_register(SVGA3DREG_COLOROUT,
+                                                            i);
+               }
+            }
+         }
+         else {
+            emit->output_map[idx] =
+               dst_register(SVGA3DREG_COLOROUT, semantic.Index);
+         }
       }
       else {
          emit->output_map[idx] = dst_register( SVGA3DREG_COLOROUT, 
@@ -311,7 +372,8 @@ static boolean ps30_output( struct svga_shader_emitter *emit,
       break;
    default:
       assert(0);
-      reg = dst_register( SVGA3DREG_COLOROUT, 0 );
+      /* A wild stab in the dark. */
+      emit->output_map[idx] = dst_register( SVGA3DREG_COLOROUT, 0 );
       break;
    }
 
@@ -319,11 +381,14 @@ static boolean ps30_output( struct svga_shader_emitter *emit,
 }
 
 
-/* We still make up the input semantics the same as in 2.0
+/**
+ * Declare a VS input register.
+ * We still make up the input semantics the same as in 2.0
  */
-static boolean vs30_input( struct svga_shader_emitter *emit,
-                           struct tgsi_declaration_semantic semantic,
-                           unsigned idx )
+static boolean
+vs30_input(struct svga_shader_emitter *emit,
+           struct tgsi_declaration_semantic semantic,
+           unsigned idx)
 {
    SVGA3DOpDclArgs dcl;
    SVGA3dShaderInstToken opcode;
@@ -333,39 +398,28 @@ static boolean vs30_input( struct svga_shader_emitter *emit,
    dcl.values[0] = 0;
    dcl.values[1] = 0;
 
-   if (emit->key.vkey.zero_stride_vertex_elements & (1 << idx)) {
-      unsigned i;
-      unsigned offset = 0;
-      unsigned start_idx = emit->info.file_max[TGSI_FILE_CONSTANT] + 1;
-      /* adjust for prescale constants */
-      start_idx += emit->key.vkey.need_prescale ? 2 : 0;
-      /* compute the offset from the start of zero stride constants */
-      for (i = 0; i < PIPE_MAX_ATTRIBS && i < idx; ++i) {
-         if (emit->key.vkey.zero_stride_vertex_elements & (1<<i))
-            ++offset;
-      }
-      emit->input_map[idx] = src_register( SVGA3DREG_CONST,
-                                           start_idx + offset );
-   } else {
-      emit->input_map[idx] = src_register( SVGA3DREG_INPUT, idx );
-      dcl.dst = dst_register( SVGA3DREG_INPUT, idx );
+   emit->input_map[idx] = src_register( SVGA3DREG_INPUT, idx );
+   dcl.dst = dst_register( SVGA3DREG_INPUT, idx );
 
-      assert(dcl.dst.reserved0);
+   assert(dcl.dst.reserved0);
 
-      svga_generate_vdecl_semantics( idx, &usage, &index );
+   svga_generate_vdecl_semantics( idx, &usage, &index );
 
-      dcl.usage = usage;
-      dcl.index = index;
-      dcl.values[0] |= 1<<31;
+   dcl.usage = usage;
+   dcl.index = index;
+   dcl.values[0] |= 1<<31;
 
-      return  (emit_instruction(emit, opcode) &&
-               svga_shader_emit_dwords( emit, dcl.values, Elements(dcl.values)));
-   }
-   return TRUE;
+   return (emit_instruction(emit, opcode) &&
+           svga_shader_emit_dwords( emit, dcl.values, ARRAY_SIZE(dcl.values)));
 }
 
-static boolean vs30_output_emit_depth_fog( struct svga_shader_emitter *emit,
-                                           SVGA3dShaderDestToken *out )
+
+/**
+ * Declare VS output for holding depth/fog.
+ */
+static boolean
+vs30_output_emit_depth_fog(struct svga_shader_emitter *emit,
+                           SVGA3dShaderDestToken *out)
 {
    SVGA3dShaderDestToken reg;
 
@@ -383,12 +437,16 @@ static boolean vs30_output_emit_depth_fog( struct svga_shader_emitter *emit,
    return emit_decl( emit, reg, SVGA3D_DECLUSAGE_TEXCOORD, 0 );
 }
 
-/* VS3.0 outputs have proper declarations and semantic info for
+
+/**
+ * Declare a VS output.
+ * VS3.0 outputs have proper declarations and semantic info for
  * matching against PS inputs.
  */
-static boolean vs30_output( struct svga_shader_emitter *emit,
-                         struct tgsi_declaration_semantic semantic,
-                         unsigned idx )
+static boolean
+vs30_output(struct svga_shader_emitter *emit,
+            struct tgsi_declaration_semantic semantic,
+            unsigned idx)
 {
    SVGA3DOpDclArgs dcl;
    SVGA3dShaderInstToken opcode;
@@ -398,7 +456,7 @@ static boolean vs30_output( struct svga_shader_emitter *emit,
    dcl.values[0] = 0;
    dcl.values[1] = 0;
 
-   if (!translate_vs_ps_semantic( semantic, &usage, &index ))
+   if (!translate_vs_ps_semantic( emit, semantic, &usage, &index ))
       return FALSE;
 
    if (emit->vs30_output_count >= SVGA3D_OUTPUTREG_MAX)
@@ -429,7 +487,7 @@ static boolean vs30_output( struct svga_shader_emitter *emit,
       /* This has the effect of not declaring psiz (below) and not 
        * emitting the final MOV to true_psiz in the postamble.
        */
-      if (!emit->key.vkey.allow_psiz)
+      if (!emit->key.vs.allow_psiz)
          return TRUE;
 
       emit->true_psiz = dcl.dst;
@@ -450,14 +508,34 @@ static boolean vs30_output( struct svga_shader_emitter *emit,
       emit->output_map[idx] = dcl.dst;
    }
 
+   return (emit_instruction(emit, opcode) &&
+           svga_shader_emit_dwords( emit, dcl.values, ARRAY_SIZE(dcl.values)));
+}
+
 
-   return  (emit_instruction(emit, opcode) &&
-            svga_shader_emit_dwords( emit, dcl.values, Elements(dcl.values)));
+/** Translate PIPE_TEXTURE_x to SVGA3DSAMP_x */
+static ubyte
+svga_tgsi_sampler_type(const struct svga_shader_emitter *emit, int idx)
+{
+   switch (emit->sampler_target[idx]) {
+   case TGSI_TEXTURE_1D:
+      return SVGA3DSAMP_2D;
+   case TGSI_TEXTURE_2D:
+   case TGSI_TEXTURE_RECT:
+      return SVGA3DSAMP_2D;
+   case TGSI_TEXTURE_3D:
+      return SVGA3DSAMP_VOLUME;
+   case TGSI_TEXTURE_CUBE:
+      return SVGA3DSAMP_CUBE;
+   }
+
+   return SVGA3DSAMP_UNKNOWN;
 }
 
-static boolean ps30_sampler( struct svga_shader_emitter *emit,
-                          struct tgsi_declaration_semantic semantic,
-                          unsigned idx )
+
+static boolean
+ps30_sampler( struct svga_shader_emitter *emit,
+              unsigned idx )
 {
    SVGA3DOpDclArgs dcl;
    SVGA3dShaderInstToken opcode;
@@ -470,32 +548,40 @@ static boolean ps30_sampler( struct svga_shader_emitter *emit,
    dcl.type = svga_tgsi_sampler_type( emit, idx );
    dcl.values[0] |= 1<<31;
 
-   return  (emit_instruction(emit, opcode) &&
-            svga_shader_emit_dwords( emit, dcl.values, Elements(dcl.values)));
+   return (emit_instruction(emit, opcode) &&
+           svga_shader_emit_dwords( emit, dcl.values, ARRAY_SIZE(dcl.values)));
 }
 
+boolean
+svga_shader_emit_samplers_decl( struct svga_shader_emitter *emit )
+{
+   unsigned i;
 
-boolean svga_translate_decl_sm30( struct svga_shader_emitter *emit,
-                             const struct tgsi_full_declaration *decl )
+   for (i = 0; i < emit->num_samplers; i++) {
+      if (!ps30_sampler(emit, i))
+         return FALSE;
+   }
+   return TRUE;
+}
+
+boolean
+svga_translate_decl_sm30( struct svga_shader_emitter *emit,
+                          const struct tgsi_full_declaration *decl )
 {
    unsigned first = decl->Range.First;
    unsigned last = decl->Range.Last;
-   unsigned semantic = 0;
-   unsigned semantic_idx = 0;
    unsigned idx;
 
-   if (decl->Declaration.Semantic) {
-      semantic = decl->Semantic.Name;
-      semantic_idx = decl->Semantic.Index;
-   }
-
    for( idx = first; idx <= last; idx++ ) {
-      boolean ok;
+      boolean ok = TRUE;
 
       switch (decl->Declaration.File) {
       case TGSI_FILE_SAMPLER:
          assert (emit->unit == PIPE_SHADER_FRAGMENT);
-         ok = ps30_sampler( emit, decl->Semantic, idx );
+         /* just keep track of the number of samplers here.
+          * Will emit the declaration in the helpers function.
+          */
+         emit->num_samplers = MAX2(emit->num_samplers, decl->Range.Last + 1);
          break;
 
       case TGSI_FILE_INPUT:
@@ -512,6 +598,14 @@ boolean svga_translate_decl_sm30( struct svga_shader_emitter *emit,
             ok = ps30_output( emit, decl->Semantic, idx );
          break;
 
+      case TGSI_FILE_SAMPLER_VIEW:
+         {
+            unsigned unit = decl->Range.First;
+            assert(decl->Range.First == decl->Range.Last);
+            emit->sampler_target[unit] = decl->SamplerView.Resource;
+         }
+         break;
+
       default:
          /* don't need to declare other vars */
          ok = TRUE;
@@ -523,6 +617,3 @@ boolean svga_translate_decl_sm30( struct svga_shader_emitter *emit,
 
    return TRUE;
 }
-
-
-