tgsi: Determine which shader input channels are effectively.

TGSI's UsageMask flag is never set. We can move this logic into
tgsi_ureg, but there there are still cases where's not used, so this
seems a better place for now.

diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.c b/src/gallium/auxiliary/tgsi/tgsi_scan.c
index e0c5d3d3d613d28d557eea3360be54ad78ca27d1..7ee272b94981b4605f29e95c984493bed5bf2c70 100644 (file)
--- a/src/gallium/auxiliary/tgsi/tgsi_scan.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_scan.c
@@ -36,6 +36,7 @@
 
 #include "util/u_math.h"
 #include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
 #include "tgsi/tgsi_scan.h"
 
 
@@ -84,19 +85,21 @@ tgsi_scan_shader(const struct tgsi_token *tokens,
          {
             const struct tgsi_full_instruction *fullinst
                = &parse.FullToken.FullInstruction;
+            uint i;
 
             assert(fullinst->Instruction.Opcode < TGSI_OPCODE_LAST);
             info->opcode_count[fullinst->Instruction.Opcode]++;
 
-            /* check if we read the frag shader FOG or FACE inputs */
-            if (procType == TGSI_PROCESSOR_FRAGMENT) {
-               uint i;
-               for (i = 0; i < fullinst->Instruction.NumSrcRegs; i++) {
-                  const struct tgsi_full_src_register *src =
-                     &fullinst->Src[i];
+            for (i = 0; i < fullinst->Instruction.NumSrcRegs; i++) {
+               const struct tgsi_full_src_register *src =
+                  &fullinst->Src[i];
+               int ind = src->Register.Index;
+
+               /* check if we read the frag shader FOG or FACE inputs */
+               if (procType == TGSI_PROCESSOR_FRAGMENT) {
                   if (src->Register.File == TGSI_FILE_INPUT ||
                       src->Register.File == TGSI_FILE_SYSTEM_VALUE) {
-                     const int ind = src->Register.Index;
+                     assert(ind >= 0);
                      if (info->input_semantic_name[ind] == TGSI_SEMANTIC_FOG) {
                         info->uses_fogcoord = TRUE;
                      }
@@ -105,6 +108,21 @@ tgsi_scan_shader(const struct tgsi_token *tokens,
                      }
                   }
                }
+
+               /* Mark which inputs are effectively used */
+               if (src->Register.File == TGSI_FILE_INPUT) {
+                  unsigned usage_mask;
+                  usage_mask = tgsi_util_get_inst_usage_mask(fullinst, i);
+                  if (src->Register.Indirect) {
+                     for (ind = 0; ind < info->num_inputs; ++ind) {
+                        info->input_usage_mask[ind] |= usage_mask;
+                     }
+                  } else {
+                     assert(ind >= 0);
+                     assert(ind < PIPE_MAX_SHADER_INPUTS);
+                     info->input_usage_mask[ind] |= usage_mask;
+                  }
+               }
             }
 
             info->num_instructions++;

diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.h b/src/gallium/auxiliary/tgsi/tgsi_scan.h
index 27de33f7990c72192ce1247c055a467bd97c66fe..f43541429d7578e214d3a040e8c2552ab7c84e42 100644 (file)
--- a/src/gallium/auxiliary/tgsi/tgsi_scan.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_scan.h
@@ -45,6 +45,7 @@ struct tgsi_shader_info
    ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; /**< TGSI_SEMANTIC_x */
    ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS];
    ubyte input_interpolate[PIPE_MAX_SHADER_INPUTS];
+   ubyte input_usage_mask[PIPE_MAX_SHADER_INPUTS];
    ubyte input_cylindrical_wrap[PIPE_MAX_SHADER_INPUTS];
    ubyte output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; /**< TGSI_SEMANTIC_x */
    ubyte output_semantic_index[PIPE_MAX_SHADER_OUTPUTS];

diff --git a/src/gallium/auxiliary/tgsi/tgsi_util.c b/src/gallium/auxiliary/tgsi/tgsi_util.c
index 0a7e4105a801acc1418bcd81b886ee67092c070b..3ec54964169bd528bed0407e0ef233fd74a300b3 100644 (file)
--- a/src/gallium/auxiliary/tgsi/tgsi_util.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_util.c
@@ -163,3 +163,149 @@ tgsi_util_set_full_src_register_sign_mode(
       assert( 0 );
    }
 }
+
+/**
+ * Determine which channels of the specificed src register are effectively
+ * used by this instruction.
+ */
+unsigned
+tgsi_util_get_inst_usage_mask(const struct tgsi_full_instruction *inst,
+                              unsigned src_idx)
+{
+   const struct tgsi_full_src_register *src = &inst->Src[src_idx];
+   unsigned write_mask = inst->Dst[0].Register.WriteMask;
+   unsigned read_mask;
+   unsigned usage_mask;
+   unsigned chan;
+
+   switch (inst->Instruction.Opcode) {
+   case TGSI_OPCODE_MOV:
+   case TGSI_OPCODE_ARL:
+   case TGSI_OPCODE_ARR:
+   case TGSI_OPCODE_RCP:
+   case TGSI_OPCODE_MUL:
+   case TGSI_OPCODE_DIV:
+   case TGSI_OPCODE_ADD:
+   case TGSI_OPCODE_MIN:
+   case TGSI_OPCODE_MAX:
+   case TGSI_OPCODE_SLT:
+   case TGSI_OPCODE_SGE:
+   case TGSI_OPCODE_MAD:
+   case TGSI_OPCODE_SUB:
+   case TGSI_OPCODE_LRP:
+   case TGSI_OPCODE_CND:
+   case TGSI_OPCODE_FRC:
+   case TGSI_OPCODE_CEIL:
+   case TGSI_OPCODE_CLAMP:
+   case TGSI_OPCODE_FLR:
+   case TGSI_OPCODE_ROUND:
+   case TGSI_OPCODE_POW:
+   case TGSI_OPCODE_ABS:
+   case TGSI_OPCODE_COS:
+   case TGSI_OPCODE_SIN:
+   case TGSI_OPCODE_DDX:
+   case TGSI_OPCODE_DDY:
+   case TGSI_OPCODE_SEQ:
+   case TGSI_OPCODE_SGT:
+   case TGSI_OPCODE_SLE:
+   case TGSI_OPCODE_SNE:
+   case TGSI_OPCODE_SSG:
+   case TGSI_OPCODE_CMP:
+   case TGSI_OPCODE_TRUNC:
+   case TGSI_OPCODE_NOT:
+   case TGSI_OPCODE_AND:
+   case TGSI_OPCODE_OR:
+   case TGSI_OPCODE_XOR:
+   case TGSI_OPCODE_SAD:
+      /* Channel-wise operations */
+      read_mask = write_mask;
+      break;
+
+   case TGSI_OPCODE_EX2:
+   case TGSI_OPCODE_LG2:
+   case TGSI_OPCODE_RCC:
+      read_mask = TGSI_WRITEMASK_X;
+      break;
+
+   case TGSI_OPCODE_SCS:
+      read_mask = write_mask & TGSI_WRITEMASK_XY ? TGSI_WRITEMASK_X : 0;
+      break;
+
+   case TGSI_OPCODE_EXP:
+   case TGSI_OPCODE_LOG:
+      read_mask = write_mask & TGSI_WRITEMASK_XYZ ? TGSI_WRITEMASK_X : 0;
+      break;
+
+   case TGSI_OPCODE_DP2A:
+      read_mask = src_idx == 2 ? TGSI_WRITEMASK_X : TGSI_WRITEMASK_XY;
+      break;
+
+   case TGSI_OPCODE_DP2:
+      read_mask = TGSI_WRITEMASK_XY;
+      break;
+
+   case TGSI_OPCODE_DP3:
+      read_mask = TGSI_WRITEMASK_XYZ;
+      break;
+
+   case TGSI_OPCODE_DP4:
+      read_mask = TGSI_WRITEMASK_XYZW;
+      break;
+
+   case TGSI_OPCODE_DPH:
+      read_mask = src_idx == 0 ? TGSI_WRITEMASK_XYZ : TGSI_WRITEMASK_XYZW;
+      break;
+
+   case TGSI_OPCODE_TEX:
+   case TGSI_OPCODE_TXD:
+   case TGSI_OPCODE_TXB:
+   case TGSI_OPCODE_TXL:
+   case TGSI_OPCODE_TXP:
+      if (src_idx == 0) {
+         switch (inst->Texture.Texture) {
+         case TGSI_TEXTURE_1D:
+         case TGSI_TEXTURE_SHADOW1D:
+            read_mask = TGSI_WRITEMASK_X;
+            break;
+
+         case TGSI_TEXTURE_2D:
+         case TGSI_TEXTURE_RECT:
+         case TGSI_TEXTURE_SHADOW2D:
+         case TGSI_TEXTURE_SHADOWRECT:
+            read_mask = TGSI_WRITEMASK_XY;
+            break;
+
+         case TGSI_TEXTURE_3D:
+         case TGSI_TEXTURE_CUBE:
+            read_mask = TGSI_WRITEMASK_XYZ;
+            break;
+
+         default:
+            assert(0);
+            read_mask = 0;
+         }
+
+         if (inst->Instruction.Opcode != TGSI_OPCODE_TEX) {
+            read_mask |= TGSI_WRITEMASK_W;
+         }
+      } else {
+         /* A safe approximation */
+         read_mask = TGSI_WRITEMASK_XYZW;
+      }
+      break;
+
+   default:
+      /* Assume all channels are read */
+      read_mask = TGSI_WRITEMASK_XYZW;
+      break;
+   }
+
+   usage_mask = 0;
+   for (chan = 0; chan < 4; ++chan) {
+      if (read_mask & (1 << chan)) {
+         usage_mask |= 1 << tgsi_util_get_full_src_register_swizzle(src, chan);
+      }
+   }
+
+   return usage_mask;
+}

diff --git a/src/gallium/auxiliary/tgsi/tgsi_util.h b/src/gallium/auxiliary/tgsi/tgsi_util.h
index 19ee2e7cf2a024ae854e13e3ed8084635b95f3b7..04702ba9826bd3a6bb36dd55048046a8e77eddb5 100644 (file)
--- a/src/gallium/auxiliary/tgsi/tgsi_util.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_util.h
@@ -34,6 +34,7 @@ extern "C" {
 
 struct tgsi_src_register;
 struct tgsi_full_src_register;
+struct tgsi_full_instruction;
 
 void *
 tgsi_align_128bit(
@@ -71,6 +72,10 @@ tgsi_util_set_full_src_register_sign_mode(
    struct tgsi_full_src_register *reg,
    unsigned sign_mode );
 
+unsigned
+tgsi_util_get_inst_usage_mask(const struct tgsi_full_instruction *inst,
+                              unsigned src_idx);
+
 #if defined __cplusplus
 }
 #endif