freedreno/ir3: fix mad copy propagation special case

[mesa.git] / src / freedreno / ir3 / ir3_cp.c

diff --git a/src/freedreno/ir3/ir3_cp.c b/src/freedreno/ir3/ir3_cp.c
index 9bd97b690e34fa095441f17009370ec0b7024e12..3039e50c3e4c0bde217abb70651244880c6028bb 100644 (file)
--- a/src/freedreno/ir3/ir3_cp.c
+++ b/src/freedreno/ir3/ir3_cp.c
@@ -32,6 +32,9 @@
 #include "ir3_compiler.h"
 #include "ir3_shader.h"
 
+#define swap(a, b) \
+       do { __typeof(a) __tmp = (a); (a) = (b); (b) = __tmp; } while (0)
+
 /*
  * Copy Propagate:
  */
@@ -360,6 +363,36 @@ unuse(struct ir3_instruction *instr)
        }
 }
 
+/**
+ * Handles the special case of the 2nd src (n == 1) to "normal" mad
+ * instructions, which cannot reference a constant.  See if it is
+ * possible to swap the 1st and 2nd sources.
+ */
+static bool
+try_swap_mad_two_srcs(struct ir3_instruction *instr, unsigned new_flags)
+{
+       if (!is_mad(instr->opc))
+               return false;
+
+       /* NOTE: pre-swap first two src's before valid_flags(),
+        * which might try to dereference the n'th src:
+        */
+       swap(instr->regs[0 + 1], instr->regs[1 + 1]);
+
+       bool valid_swap =
+               /* can we propagate mov if we move 2nd src to first? */
+               valid_flags(instr, 0, new_flags) &&
+               /* and does first src fit in second slot? */
+               valid_flags(instr, 1, instr->regs[1 + 1]->flags);
+
+       if (!valid_swap) {
+               /* put things back the way they were: */
+               swap(instr->regs[0 + 1], instr->regs[1 + 1]);
+       }   /* otherwise leave things swapped */
+
+       return valid_swap;
+}
+
 /**
  * Handle cp for a given src register.  This additionally handles
  * the cases of collapsing immedate/const (which replace the src
@@ -423,15 +456,8 @@ reg_cp(struct ir3_cp_ctx *ctx, struct ir3_instruction *instr,
                         * src prior to multiply) can swap their first two srcs if
                         * src[0] is !CONST and src[1] is CONST:
                         */
-                       if ((n == 1) && is_mad(instr->opc) &&
-                                       !(instr->regs[0 + 1]->flags & (IR3_REG_CONST | IR3_REG_RELATIV)) &&
-                                       valid_flags(instr, 0, new_flags & ~IR3_REG_IMMED)) {
-                               /* swap src[0] and src[1]: */
-                               struct ir3_register *tmp;
-                               tmp = instr->regs[0 + 1];
-                               instr->regs[0 + 1] = instr->regs[1 + 1];
-                               instr->regs[1 + 1] = tmp;
-
+                       if ((n == 1) && try_swap_mad_two_srcs(instr, new_flags)) {
+                               /* we swapped, so now we are dealing with 1st src: */
                                n = 0;
                        } else {
                                return;