glsl: Use array deref for access to vector components

We've assumed that we could lower per-component vector access from

  vec[i] = scalar

to

  vec = ir_triop_vector_insert(vec, scalar, i)

but with SSBOs (and compute shader SLM and tesselation outputs) this is
no longer valid. If a vector is "externally visible", multiple threads
can write independent components simultaneously. With lowering to
ir_triop_vector_insert, each thread read the entire vector, changes one
component, then writes out the entire vector. This is racy.

Instead of generating a ir_binop_vector_extract when we see v[i], we
generate ir_dereference_array. We then add a lowering pass to lower the
ir_dereference_array to ir_binop_vector_extract for rvalues and for to
vector_insert for lvalues in a separate lowering pass.

The resulting IR is the same as before, but we now have a window between
ast->ir conversion and the lowering pass where v[i] appears in the IR as
an array deref. This lets us run lowering passes that lower the vector
access to I/O (eg for SSBO load/store) before we lower the per-component
access to full vector writes.

Reviewed-by: Jordan Justen <[email protected]>
Signed-off-by: Kristian Høgsberg Kristensen <[email protected]>

1 files changed, 0 insertions, 43 deletions

diff --git a/src/glsl/ast_to_hir.cpp b/src/glsl/ast_to_hir.cpp
index 6f5f3c1b245..9d341e8cf92 100644
--- a/src/glsl/ast_to_hir.cpp
+++ b/src/glsl/ast_to_hir.cpp
@@ -850,43 +850,6 @@ do_assignment(exec_list *instructions, struct _mesa_glsl_parse_state *state,
 {
    void *ctx = state;
    bool error_emitted = (lhs->type->is_error() || rhs->type->is_error());
-   ir_rvalue *extract_channel = NULL;
-
-   /* If the assignment LHS comes back as an ir_binop_vector_extract
-    * expression, move it to the RHS as an ir_triop_vector_insert.
-    */
-   if (lhs->ir_type == ir_type_expression) {
-      ir_expression *const lhs_expr = lhs->as_expression();
-
-      if (unlikely(lhs_expr->operation == ir_binop_vector_extract)) {
-         ir_rvalue *new_rhs =
-            validate_assignment(state, lhs_loc, lhs,
-                                rhs, is_initializer);
-
-         if (new_rhs == NULL) {
-            return lhs;
-         } else {
-            /* This converts:
-             * - LHS: (expression float vector_extract <vec> <channel>)
-             * - RHS: <scalar>
-             * into:
-             * - LHS: <vec>
-             * - RHS: (expression vec2 vector_insert <vec> <channel> <scalar>)
-             *
-             * The LHS type is now a vector instead of a scalar.  Since GLSL
-             * allows assignments to be used as rvalues, we need to re-extract
-             * the channel from assignment_temp when returning the rvalue.
-             */
-            extract_channel = lhs_expr->operands[1];
-            rhs = new(ctx) ir_expression(ir_triop_vector_insert,
-                                         lhs_expr->operands[0]->type,
-                                         lhs_expr->operands[0],
-                                         new_rhs,
-                                         extract_channel);
-            lhs = lhs_expr->operands[0]->clone(ctx, NULL);
-         }
-      }
-   }
 
    ir_variable *lhs_var = lhs->variable_referenced();
    if (lhs_var)
@@ -984,12 +947,6 @@ do_assignment(exec_list *instructions, struct _mesa_glsl_parse_state *state,
       }
       ir_rvalue *rvalue = new(ctx) ir_dereference_variable(var);
 
-      if (extract_channel) {
-         rvalue = new(ctx) ir_expression(ir_binop_vector_extract,
-                                         rvalue,
-                                         extract_channel->clone(ctx, NULL));
-      }
-
       *out_rvalue = rvalue;
    } else {
       if (!error_emitted)