aco: move call to store_output_to_temps in store_ls_or_es_output earlier

Skips get_intrinsic_io_basic_offset()

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4667>

diff --git a/src/amd/compiler/aco_instruction_selection.cpp b/src/amd/compiler/aco_instruction_selection.cpp
index 31a5e410e9f8042222107f567ea2f4d8a729ceda..a76f1bb25af72e4785d6accf687dc6fe27da10dc 100644 (file)
--- a/src/amd/compiler/aco_instruction_selection.cpp
+++ b/src/amd/compiler/aco_instruction_selection.cpp
@@ -3777,11 +3777,6 @@ void visit_store_ls_or_es_output(isel_context *ctx, nir_intrinsic_instr *instr)
 {
    Builder bld(ctx->program, ctx->block);
 
-   std::pair<Temp, unsigned> offs = get_intrinsic_io_basic_offset(ctx, instr, 4u);
-   Temp src = get_ssa_temp(ctx, instr->src[0].ssa);
-   unsigned write_mask = nir_intrinsic_write_mask(instr);
-   unsigned elem_size_bytes = instr->src[0].ssa->bit_size / 8u;
-
    if (ctx->tcs_in_out_eq && store_output_to_temps(ctx, instr)) {
       /* When the TCS only reads this output directly and for the same vertices as its invocation id, it is unnecessary to store the VS output to LDS. */
       bool indirect_write;
@@ -3790,6 +3785,11 @@ void visit_store_ls_or_es_output(isel_context *ctx, nir_intrinsic_instr *instr)
          return;
    }
 
+   std::pair<Temp, unsigned> offs = get_intrinsic_io_basic_offset(ctx, instr, 4u);
+   Temp src = get_ssa_temp(ctx, instr->src[0].ssa);
+   unsigned write_mask = nir_intrinsic_write_mask(instr);
+   unsigned elem_size_bytes = instr->src[0].ssa->bit_size / 8u;
+
    if (ctx->stage == vertex_es || ctx->stage == tess_eval_es) {
       /* GFX6-8: ES stage is not merged into GS, data is passed from ES to GS in VMEM. */
       Temp esgs_ring = bld.smem(aco_opcode::s_load_dwordx4, bld.def(s4), ctx->program->private_segment_buffer, Operand(RING_ESGS_VS * 16u));