radeon/r200: drop legacy texture heap code

[mesa.git] / src / mesa / drivers / dri / r200 / r200_vertprog.c

diff --git a/src/mesa/drivers/dri/r200/r200_vertprog.c b/src/mesa/drivers/dri/r200/r200_vertprog.c
index 4960d481d5da337006c344457fb6135f981513ca..888f91db73d21149a9f5c936e44b9ef6cc2b640a 100644 (file)
--- a/src/mesa/drivers/dri/r200/r200_vertprog.c
+++ b/src/mesa/drivers/dri/r200/r200_vertprog.c
@@ -30,18 +30,20 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
  *   Aapo Tahkola <aet@rasterburn.org>
  *   Roland Scheidegger <rscheidegger_lists@hispeed.ch>
  */
-#include "glheader.h"
-#include "macros.h"
-#include "enums.h"
-#include "program.h"
+#include "main/glheader.h"
+#include "main/macros.h"
+#include "main/enums.h"
+#include "shader/program.h"
+#include "shader/prog_instruction.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_statevars.h"
+#include "shader/programopt.h"
+#include "tnl/tnl.h"
 
 #include "r200_context.h"
 #include "r200_vertprog.h"
 #include "r200_ioctl.h"
 #include "r200_tcl.h"
-#include "program_instruction.h"
-#include "programopt.h"
-#include "tnl/tnl.h"
 
 #if SWIZZLE_X != VSF_IN_COMPONENT_X || \
     SWIZZLE_Y != VSF_IN_COMPONENT_Y || \
@@ -152,7 +154,7 @@ static GLboolean r200VertexProgUpdateParams(GLcontext *ctx, struct r200_vertex_p
    return GL_TRUE;
 }
 
-static __inline unsigned long t_dst_mask(GLuint mask)
+static INLINE unsigned long t_dst_mask(GLuint mask)
 {
    /* WRITEMASK_* is equivalent to VSF_FLAG_* */
    return mask & VSF_FLAG_ALL;
@@ -213,6 +215,7 @@ static unsigned long t_src_class(enum register_file file)
    case PROGRAM_LOCAL_PARAM:
    case PROGRAM_ENV_PARAM:
    case PROGRAM_NAMED_PARAM:
+   case PROGRAM_CONSTANT:
    case PROGRAM_STATE_VAR:
       return VSF_IN_CLASS_PARAM;
    /*
@@ -226,7 +229,7 @@ static unsigned long t_src_class(enum register_file file)
    }
 }
 
-static __inline unsigned long t_swizzle(GLubyte swizzle)
+static INLINE unsigned long t_swizzle(GLubyte swizzle)
 {
 /* this is in fact a NOP as the Mesa SWIZZLE_* are all identical to VSF_IN_COMPONENT_* */
    return swizzle;
@@ -406,6 +409,7 @@ static GLboolean r200_translate_vertex_program(GLcontext *ctx, struct r200_verte
    int fog_temp_i = 0;
    int free_inputs;
    int array_count = 0;
+   int u_temp_used;
 
    vp->native = GL_FALSE;
    vp->translated = GL_TRUE;
@@ -462,7 +466,7 @@ static GLboolean r200_translate_vertex_program(GLcontext *ctx, struct r200_verte
       base e isn't directly available neither. */
    if ((mesa_vp->Base.OutputsWritten & (1 << VERT_RESULT_FOGC)) && !vp->fogpidx) {
       struct gl_program_parameter_list *paramList;
-      GLint tokens[6] = { STATE_FOG_PARAMS, 0, 0, 0, 0, 0 };
+      gl_state_index tokens[STATE_LENGTH] = { STATE_FOG_PARAMS, 0, 0, 0, 0 };
       paramList = mesa_vp->Base.Parameters;
       vp->fogpidx = _mesa_add_state_reference(paramList, tokens);
    }
@@ -742,9 +746,16 @@ static GLboolean r200_translate_vertex_program(GLcontext *ctx, struct r200_verte
         goto next;
 
       case OPCODE_MAD:
+        /* only 2 read ports into temp memory thus may need the macro op MAD_2
+           instead (requiring 2 clocks) if all inputs are in temp memory
+           (and, only if they actually reference 3 distinct temps) */
         hw_op=(src[0].File == PROGRAM_TEMPORARY &&
            src[1].File == PROGRAM_TEMPORARY &&
-           src[2].File == PROGRAM_TEMPORARY) ? R200_VPI_OUT_OP_MAD_2 : R200_VPI_OUT_OP_MAD;
+           src[2].File == PROGRAM_TEMPORARY &&
+           (((src[0].RelAddr << 8) | src[0].Index) != ((src[1].RelAddr << 8) | src[1].Index)) &&
+           (((src[0].RelAddr << 8) | src[0].Index) != ((src[2].RelAddr << 8) | src[2].Index)) &&
+           (((src[1].RelAddr << 8) | src[1].Index) != ((src[2].RelAddr << 8) | src[2].Index))) ?
+           R200_VPI_OUT_OP_MAD_2 : R200_VPI_OUT_OP_MAD;
 
         o_inst->op = MAKE_VSF_OP(hw_op, t_dst(&dst),
            t_dst_mask(dst.WriteMask));
@@ -872,8 +883,11 @@ else {
       case OPCODE_XPD:
         /* mul r0, r1.yzxw, r2.zxyw
            mad r0, -r2.yzxw, r1.zxyw, r0
-           NOTE: might need MAD_2
          */
+        hw_op=(src[0].File == PROGRAM_TEMPORARY &&
+           src[1].File == PROGRAM_TEMPORARY &&
+           (((src[0].RelAddr << 8) | src[0].Index) != ((src[1].RelAddr << 8) | src[1].Index))) ?
+           R200_VPI_OUT_OP_MAD_2 : R200_VPI_OUT_OP_MAD;
 
         o_inst->op = MAKE_VSF_OP(R200_VPI_OUT_OP_MUL,
            (u_temp_i << R200_VPI_OUT_REG_INDEX_SHIFT) | R200_VSF_OUT_CLASS_TMP,
@@ -899,7 +913,7 @@ else {
         o_inst++;
         u_temp_i--;
 
-        o_inst->op = MAKE_VSF_OP(R200_VPI_OUT_OP_MAD, t_dst(&dst),
+        o_inst->op = MAKE_VSF_OP(hw_op, t_dst(&dst),
                t_dst_mask(dst.WriteMask));
 
         o_inst->src0 = MAKE_VSF_SOURCE(t_src_index(vp, &src[1]),
@@ -1049,14 +1063,15 @@ else {
          dofogfix = 0;
       }
 
+      u_temp_used = (R200_VSF_MAX_TEMPS - 1) - u_temp_i;
       if (mesa_vp->Base.NumNativeTemporaries <
-        (mesa_vp->Base.NumTemporaries + (R200_VSF_MAX_TEMPS - 1 - u_temp_i))) {
+        (mesa_vp->Base.NumTemporaries + u_temp_used)) {
         mesa_vp->Base.NumNativeTemporaries =
-           mesa_vp->Base.NumTemporaries + (R200_VSF_MAX_TEMPS - 1 - u_temp_i);
+           mesa_vp->Base.NumTemporaries + u_temp_used;
       }
-      if (u_temp_i < mesa_vp->Base.NumTemporaries) {
+      if ((mesa_vp->Base.NumTemporaries + u_temp_used) > R200_VSF_MAX_TEMPS) {
         if (R200_DEBUG & DEBUG_FALLBACKS) {
-           fprintf(stderr, "Ran out of temps, num temps %d, us %d\n", mesa_vp->Base.NumTemporaries, u_temp_i);
+           fprintf(stderr, "Ran out of temps, num temps %d, us %d\n", mesa_vp->Base.NumTemporaries, u_temp_used);
         }
         return GL_FALSE;
       }
@@ -1095,9 +1110,9 @@ void r200SetupVertexProg( GLcontext *ctx ) {
    }
    /* could optimize setting up vertex progs away for non-tcl hw */
    fallback = !(vp->native && r200VertexProgUpdateParams(ctx, vp) &&
-      rmesa->r200Screen->drmSupportsVertexProgram);
+      rmesa->radeon.radeonScreen->drmSupportsVertexProgram);
    TCL_FALLBACK(ctx, R200_TCL_FALLBACK_VERTEX_PROGRAM, fallback);
-   if (rmesa->TclFallback) return;
+   if (rmesa->radeon.TclFallback) return;
 
    R200_STATECHANGE( rmesa, vap );
    /* FIXME: fglrx sets R200_VAP_SINGLE_BUF_STATE_ENABLE too. Do we need it?