i965/fs: Migrate register spills and fills to the IR builder.

Yes, it's incorrect to use the 0-th channel enable group
unconditionally without considering the execution and regioning
controls of the instruction that uses the spilled value, but it
matches the previous behaviour exactly, the builder just makes the
preexisting problem more obvious because emitting an instruction of
non-native SIMD width without having called .group() or .exec_all()
explicitly would have led to an assertion failure.

I'll fix the problem in a follow-up series, as the solution is going
to be non-trivial.

Reviewed-by: Matt Turner <mattst88@gmail.com>

diff --git a/src/mesa/drivers/dri/i965/brw_fs_reg_allocate.cpp b/src/mesa/drivers/dri/i965/brw_fs_reg_allocate.cpp
index 582d0993f1c9096a7aad6b9fc6becdb10ef35916..3faf49ab8aab433d5cd0f19e629e26c4d221fb2b 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_fs_reg_allocate.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_reg_allocate.cpp
@@ -30,6 +30,8 @@
 #include "glsl/glsl_types.h"
 #include "glsl/ir_optimization.h"
 
+using namespace brw;
+
 static void
 assign_reg(unsigned *reg_hw_locations, fs_reg *reg)
 {
@@ -696,25 +698,24 @@ fs_visitor::emit_unspill(bblock_t *block, fs_inst *inst, fs_reg dst,
       dst.width = 16;
    }
 
+   const fs_builder ibld = bld.annotate(inst->annotation, inst->ir)
+                              .group(reg_size * 8, 0)
+                              .at(block, inst);
+
    for (int i = 0; i < count / reg_size; i++) {
       /* The gen7 descriptor-based offset is 12 bits of HWORD units. */
       bool gen7_read = devinfo->gen >= 7 && spill_offset < (1 << 12) * REG_SIZE;
-
-      fs_inst *unspill_inst =
-         new(mem_ctx) fs_inst(gen7_read ?
-                              SHADER_OPCODE_GEN7_SCRATCH_READ :
-                              SHADER_OPCODE_GEN4_SCRATCH_READ,
-                              dst);
+      fs_inst *unspill_inst = ibld.emit(gen7_read ?
+                                        SHADER_OPCODE_GEN7_SCRATCH_READ :
+                                        SHADER_OPCODE_GEN4_SCRATCH_READ,
+                                        dst);
       unspill_inst->offset = spill_offset;
-      unspill_inst->ir = inst->ir;
-      unspill_inst->annotation = inst->annotation;
       unspill_inst->regs_written = reg_size;
 
       if (!gen7_read) {
          unspill_inst->base_mrf = 14;
          unspill_inst->mlen = 1; /* header contains offset */
       }
-      inst->insert_before(block, unspill_inst);
 
       dst.reg_offset += reg_size;
       spill_offset += reg_size * REG_SIZE;
@@ -732,17 +733,17 @@ fs_visitor::emit_spill(bblock_t *block, fs_inst *inst, fs_reg src,
       reg_size = 2;
    }
 
+   const fs_builder ibld = bld.annotate(inst->annotation, inst->ir)
+                              .group(reg_size * 8, 0)
+                              .at(block, inst->next);
+
    for (int i = 0; i < count / reg_size; i++) {
       fs_inst *spill_inst =
-         new(mem_ctx) fs_inst(SHADER_OPCODE_GEN4_SCRATCH_WRITE,
-                              reg_size * 8, reg_null_f, src);
+         ibld.emit(SHADER_OPCODE_GEN4_SCRATCH_WRITE, bld.null_reg_f(), src);
       src.reg_offset += reg_size;
       spill_inst->offset = spill_offset + i * reg_size * REG_SIZE;
-      spill_inst->ir = inst->ir;
-      spill_inst->annotation = inst->annotation;
       spill_inst->mlen = 1 + reg_size; /* header, value */
       spill_inst->base_mrf = spill_base_mrf;
-      inst->insert_after(block, spill_inst);
    }
 }