i915: fix broken indirect constant buffer reads

The READ message's msg_control value can be 0 or 1 to indicate that the
Oword should be read into the lower or upper half of the target register.
It seems that the other half of the register gets clobbered though.  So
we read into two dest registers then use a MOV to combine the upper/lower
halves.

diff --git a/src/mesa/drivers/dri/i965/brw_eu.h b/src/mesa/drivers/dri/i965/brw_eu.h
index 896e67dbfe914523c0f33e47a3d9fd5b832c5377..62c98bd8bb364f703283848ad5827dbd32ebae4c 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_eu.h
+++ b/src/mesa/drivers/dri/i965/brw_eu.h
@@ -867,6 +867,7 @@ void brw_dp_READ_4( struct brw_compile *p,
 
 void brw_dp_READ_4_vs( struct brw_compile *p,
                        struct brw_reg dest,
+                       GLuint oword,
                        GLboolean relAddr,
                        struct brw_reg addrReg,
                        GLuint location,

diff --git a/src/mesa/drivers/dri/i965/brw_eu_emit.c b/src/mesa/drivers/dri/i965/brw_eu_emit.c
index df2141660c088f63d54d1c5af1830a7ac18f83c3..60ea44f7a9688cf75cbd272446404084289e0781 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_eu_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_eu_emit.c
@@ -1009,6 +1009,7 @@ void brw_dp_READ_4( struct brw_compile *p,
  */
 void brw_dp_READ_4_vs(struct brw_compile *p,
                       struct brw_reg dest,
+                      GLuint oword,
                       GLboolean relAddr,
                       struct brw_reg addrReg,
                       GLuint location,
@@ -1016,6 +1017,7 @@ void brw_dp_READ_4_vs(struct brw_compile *p,
 {
    GLuint msg_reg_nr = 1;
 
+   assert(oword < 2);
    /*
    printf("vs const read msg, location %u, msg_reg_nr %d\n",
           location, msg_reg_nr);
@@ -1061,56 +1063,13 @@ void brw_dp_READ_4_vs(struct brw_compile *p,
 
       brw_set_dp_read_message(insn,
                              bind_table_index,
-                             0,  /* msg_control (0 means 1 Oword, lower half) */
+                             oword,  /* 0 = lower Oword, 1 = upper Oword */
                              BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */
                              0, /* source cache = data cache */
                              1, /* msg_length */
                              1, /* response_length (1 Oword) */
                              0); /* eot */
    }
-
-   if (relAddr) {
-      /* second read to get second constant */
-      msg_reg_nr++;
-      {
-         /* Setup MRF[1] with location/offset into const buffer */
-         struct brw_reg b;
-
-         brw_push_insn_state(p);
-         brw_set_compression_control(p, BRW_COMPRESSION_NONE);
-         brw_set_mask_control(p, BRW_MASK_DISABLE);
-         brw_set_predicate_control(p, BRW_PREDICATE_NONE);
-
-         b = brw_message_reg(msg_reg_nr);
-         b = retype(b, BRW_REGISTER_TYPE_UD);
-         addrReg = suboffset(addrReg, 1); /* upper half of addrReg */
-         brw_ADD(p, b, addrReg, brw_imm_ud(location));
-
-         brw_pop_insn_state(p);
-      }
-
-      {
-         struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND);
-
-         insn->header.predicate_control = BRW_PREDICATE_NONE;
-         insn->header.compression_control = BRW_COMPRESSION_NONE; 
-         insn->header.destreg__conditonalmod = msg_reg_nr;
-         insn->header.mask_control = BRW_MASK_DISABLE;
-         /*insn->header.access_mode = BRW_ALIGN_16;*/
-
-         brw_set_dest(insn, dest);
-         brw_set_src0(insn, brw_null_reg());
-
-         brw_set_dp_read_message(insn,
-                                 bind_table_index,
-                                 1,  /* msg_control (1 means 1 Oword, upper half) */
-                                 BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */
-                                 0, /* source cache = data cache */
-                                 1, /* msg_length */
-                                 1, /* response_length (1 Oword) */
-                                 0); /* eot */
-      }
-   }
 }
 
 

diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c
index 0c86c23ad919b0f9c282050f39c0c9b3b3b06f3f..524f1211cee9d6dd9dfb81be5bd6d9a65f791676 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_vs_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c
@@ -708,10 +708,12 @@ get_constant(struct brw_vs_compile *c,
    const struct prog_src_register *src = &inst->SrcReg[argIndex];
    struct brw_compile *p = &c->func;
    struct brw_reg const_reg;
+   struct brw_reg const2_reg;
 
    assert(argIndex < 3);
 
    if (c->current_const[argIndex].index != src->Index || src->RelAddr) {
+      struct brw_reg addrReg = c->regs[PROGRAM_ADDRESS][0];
 
       c->current_const[argIndex].index = src->Index;
 
@@ -719,19 +721,46 @@ get_constant(struct brw_vs_compile *c,
       printf("  fetch const[%d] for arg %d into reg %d\n",
              src->Index, argIndex, c->current_const[argIndex].reg.nr);
 #endif
-
       /* need to fetch the constant now */
       brw_dp_READ_4_vs(p,
-                       c->current_const[argIndex].reg, /* writeback dest */
-                       src->RelAddr,                   /* relative indexing? */
-                       c->regs[PROGRAM_ADDRESS][0],    /* address register */
-                       16 * src->Index,                /* byte offset */
-                       SURF_INDEX_VERT_CONST_BUFFER    /* binding table index */
+                       c->current_const[argIndex].reg,/* writeback dest */
+                       0,                             /* oword */
+                       src->RelAddr,                  /* relative indexing? */
+                       addrReg,                       /* address register */
+                       16 * src->Index,               /* byte offset */
+                       SURF_INDEX_VERT_CONST_BUFFER   /* binding table index */
                        );
+
+      if (src->RelAddr) {
+         /* second read */
+         const2_reg = get_tmp(c);
+
+         /* use upper half of address reg for second read */
+         addrReg = stride(addrReg, 0, 4, 0);
+         addrReg.subnr = 16;
+
+         brw_dp_READ_4_vs(p,
+                          const2_reg,              /* writeback dest */
+                          1,                       /* oword */
+                          src->RelAddr,            /* relative indexing? */
+                          addrReg,                 /* address register */
+                          16 * src->Index,         /* byte offset */
+                          SURF_INDEX_VERT_CONST_BUFFER
+                          );
+      }
    }
 
    const_reg = c->current_const[argIndex].reg;
-   if (!src->RelAddr) {
+
+   if (src->RelAddr) {
+      /* merge the two Owords into the constant register */
+      /* const_reg[7..4] = const2_reg[7..4] */
+      brw_MOV(p,
+              suboffset(stride(const_reg, 0, 4, 1), 4),
+              suboffset(stride(const2_reg, 0, 4, 1), 4));
+      release_tmp(c, const2_reg);
+   }
+   else {
       /* replicate lower four floats into upper half (to get XYZWXYZW) */
       const_reg = stride(const_reg, 0, 4, 0);
       const_reg.subnr = 0;