freedreno/ir3: allow inputs with the same location

turnip can have multiple inputs with the same location, and different
location_frac.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3109>

diff --git a/src/freedreno/ir3/ir3_compiler_nir.c b/src/freedreno/ir3/ir3_compiler_nir.c
index 698d98c6c6a5e867d9ed4710a10637d9c3eaed10..656c85e67f18bcabd72afe302deeb9ed1f9b2611 100644 (file)
--- a/src/freedreno/ir3/ir3_compiler_nir.c
+++ b/src/freedreno/ir3/ir3_compiler_nir.c
@@ -2704,7 +2704,7 @@ setup_input(struct ir3_context *ctx, nir_variable *in)
                return;
 
        so->inputs[n].slot = slot;
-       so->inputs[n].compmask = (1 << (ncomp + frac)) - 1;
+       so->inputs[n].compmask |= (1 << (ncomp + frac)) - 1;
        so->inputs_count = MAX2(so->inputs_count, n + 1);
        so->inputs[n].interpolate = in->data.interpolation;
 
@@ -2767,17 +2767,25 @@ setup_input(struct ir3_context *ctx, nir_variable *in)
                        ctx->inputs[idx] = instr;
                }
        } else if (ctx->so->type == MESA_SHADER_VERTEX) {
-               /* We shouldn't have fractional input for VS input.. that only shows
-                * up with varying packing
-                */
-               assert(frac == 0);
+               struct ir3_instruction *input = NULL, *in;
+               struct ir3_instruction *components[4];
+               unsigned mask = (1 << (ncomp + frac)) - 1;
 
-               struct ir3_instruction *input = create_input(ctx, (1 << ncomp) - 1);
-               struct ir3_instruction *components[ncomp];
+               foreach_input(in, ctx->ir) {
+                       if (in->input.inidx == n) {
+                               input = in;
+                               break;
+                       }
+               }
 
-               input->input.inidx = n;
+               if (!input) {
+                       input = create_input(ctx, mask);
+                       input->input.inidx = n;
+               } else {
+                       input->regs[0]->wrmask |= mask;
+               }
 
-               ir3_split_dest(ctx->block, components, input, 0, ncomp);
+               ir3_split_dest(ctx->block, components, input, frac, ncomp);
 
                for (int i = 0; i < ncomp; i++) {
                        unsigned idx = (n * 4) + i + frac;