glsl: Rework assignments with write_masks to have LHS chan count match RHS.

It turns out that most people new to this IR are surprised when an
assignment to (say) 3 components on the LHS takes 4 components on the
RHS.  It also makes for quite strange IR output:

(assign (constant bool (1)) (x) (var_ref color) (swiz x (var_ref v) ))
(assign (constant bool (1)) (y) (var_ref color) (swiz yy (var_ref v) ))
(assign (constant bool (1)) (z) (var_ref color) (swiz zzz (var_ref v) ))

But even worse, even we get it wrong, as shown by this line of our
current step(float, vec4):

(assign (constant bool (1)) (w)
	(var_ref t)
	(expression float b2f (expression bool >=
		    (swiz w (var_ref x))(var_ref edge))))

where we try to assign a float to the writemasked-out x channel and
don't supply anything for the actual w channel we're writing.  Drivers
right now just get lucky since ir_to_mesa spams the float value across
all the source channels of a vec4.

Instead, the RHS will now have a number of components equal to the
number of components actually being written.  Hopefully this confuses
everyone less, and it also makes codegen for a scalar target simpler.

Reviewed-by: Kenneth Graunke <[email protected]>
Reviewed-by: Ian Romanick <[email protected]>

1 files changed, 27 insertions, 1 deletions

diff --git a/src/mesa/program/ir_to_mesa.cpp b/src/mesa/program/ir_to_mesa.cpp
index f24dce79f65..de0cb1ac5bd 100644
--- a/src/mesa/program/ir_to_mesa.cpp
+++ b/src/mesa/program/ir_to_mesa.cpp
@@ -1577,13 +1577,39 @@ ir_to_mesa_visitor::visit(ir_assignment *ir)
       assert(!ir->lhs->type->is_scalar() && !ir->lhs->type->is_vector());
       l.writemask = WRITEMASK_XYZW;
    } else if (ir->lhs->type->is_scalar()) {
-      /* FINISHME: This hack makes writing to gl_FragData, which lives in the
+      /* FINISHME: This hack makes writing to gl_FragDepth, which lives in the
        * FINISHME: W component of fragment shader output zero, work correctly.
        */
       l.writemask = WRITEMASK_XYZW;
    } else {
+      int swizzles[4];
+      int first_enabled_chan = 0;
+      int rhs_chan = 0;
+
       assert(ir->lhs->type->is_vector());
       l.writemask = ir->write_mask;
+
+      for (int i = 0; i < 4; i++) {
+	 if (l.writemask & (1 << i)) {
+	    first_enabled_chan = GET_SWZ(r.swizzle, i);
+	    break;
+	 }
+      }
+
+      /* Swizzle a small RHS vector into the channels being written.
+       *
+       * glsl ir treats write_mask as dictating how many channels are
+       * present on the RHS while Mesa IR treats write_mask as just
+       * showing which channels of the vec4 RHS get written.
+       */
+      for (int i = 0; i < 4; i++) {
+	 if (l.writemask & (1 << i))
+	    swizzles[i] = GET_SWZ(r.swizzle, rhs_chan++);
+	 else
+	    swizzles[i] = first_enabled_chan;
+      }
+      r.swizzle = MAKE_SWIZZLE4(swizzles[0], swizzles[1],
+				swizzles[2], swizzles[3]);
    }
 
    assert(l.file != PROGRAM_UNDEFINED);