i965/vec4: Don't coalesce regs in Gen6 MATH ops if reswizzle/writemask needed

Gen6 MATH instructions can not execute in align16 mode, so swizzles or
writemasking are not allowed.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=92033
Reviewed-by: Matt Turner <mattst88@gmail.com>

diff --git a/src/mesa/drivers/dri/i965/brw_ir_vec4.h b/src/mesa/drivers/dri/i965/brw_ir_vec4.h
index 6e8b16139d3fd118eaecc2aeadbcbe2fb43b03a8..96dd633e117c405fd65a25a4a3bfdb74405c5b83 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_ir_vec4.h
+++ b/src/mesa/drivers/dri/i965/brw_ir_vec4.h
@@ -175,7 +175,8 @@ public:
 
    bool is_send_from_grf();
    unsigned regs_read(unsigned arg) const;
-   bool can_reswizzle(int dst_writemask, int swizzle, int swizzle_mask);
+   bool can_reswizzle(const struct brw_device_info *devinfo, int dst_writemask,
+                      int swizzle, int swizzle_mask);
    void reswizzle(int dst_writemask, int swizzle);
    bool can_do_source_mods(const struct brw_device_info *devinfo);
 

diff --git a/src/mesa/drivers/dri/i965/brw_vec4.cpp b/src/mesa/drivers/dri/i965/brw_vec4.cpp
index 6e52d79e05357e0831db077f75c4397525ed17e8..c6510453479a4be533d276cbaeccd145918890ac 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_vec4.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4.cpp
@@ -941,10 +941,18 @@ vec4_visitor::opt_set_dependency_control()
 }
 
 bool
-vec4_instruction::can_reswizzle(int dst_writemask,
+vec4_instruction::can_reswizzle(const struct brw_device_info *devinfo,
+                                int dst_writemask,
                                 int swizzle,
                                 int swizzle_mask)
 {
+   /* Gen6 MATH instructions can not execute in align16 mode, so swizzles
+    * or writemasking are not allowed.
+    */
+   if (devinfo->gen == 6 && is_math() &&
+       (swizzle != BRW_SWIZZLE_XYZW || dst_writemask != WRITEMASK_XYZW))
+      return false;
+
    /* If this instruction sets anything not referenced by swizzle, then we'd
     * totally break it when we reswizzle.
     */
@@ -1099,7 +1107,7 @@ vec4_visitor::opt_register_coalesce()
                break;
 
             /* If we can't handle the swizzle, bail. */
-            if (!scan_inst->can_reswizzle(inst->dst.writemask,
+            if (!scan_inst->can_reswizzle(devinfo, inst->dst.writemask,
                                           inst->src[0].swizzle,
                                           chans_needed)) {
                break;