freedreno/ir3: Emit link map as byte or dwords offsets as needed

Stages that load inputs with ldlw (TCS, GS) need byte offsets, stages
that load with ldg (TES) need dwords offsets.

Signed-off-by: Kristian H. Kristensen <hoegsberg@google.com>
Acked-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Rob Clark <robdclark@gmail.com>

diff --git a/src/gallium/drivers/freedreno/ir3/ir3_gallium.c b/src/gallium/drivers/freedreno/ir3/ir3_gallium.c
index 9fad7446080300e7cb7e4e7aeab1f95a287ea264..8b329687978ce2529640e34d98338ccfc4bec01b 100644 (file)
--- a/src/gallium/drivers/freedreno/ir3/ir3_gallium.c
+++ b/src/gallium/drivers/freedreno/ir3/ir3_gallium.c
@@ -400,13 +400,27 @@ link_geometry_stages(const struct ir3_shader_variant *producer,
                const struct ir3_shader_variant *consumer,
                uint32_t *locs)
 {
-       uint32_t num_loc = 0;
+       uint32_t num_loc = 0, factor;
+
+       switch (consumer->type) {
+       case MESA_SHADER_TESS_CTRL:
+       case MESA_SHADER_GEOMETRY:
+               /* These stages load with ldlw, which expects byte offsets. */
+               factor = 4;
+               break;
+       case MESA_SHADER_TESS_EVAL:
+               /* The tess eval shader uses ldg, which takes dword offsets. */
+               factor = 1;
+               break;
+       default:
+               unreachable("bad shader stage");
+       }
 
        nir_foreach_variable(in_var, &consumer->shader->nir->inputs) {
                nir_foreach_variable(out_var, &producer->shader->nir->outputs) {
                        if (in_var->data.location == out_var->data.location) {
                                locs[in_var->data.driver_location] =
-                                       producer->shader->output_loc[out_var->data.driver_location] * 4;
+                                       producer->shader->output_loc[out_var->data.driver_location] * factor;
 
                                debug_assert(num_loc <= in_var->data.driver_location + 1);
                                num_loc = in_var->data.driver_location + 1;