aco: Fix operand of s_bcnt1_i32_b64 in emit_boolean_reduce.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>

diff --git a/src/amd/compiler/aco_instruction_selection.cpp b/src/amd/compiler/aco_instruction_selection.cpp
index 807cc02264cda9c679326c6418f8bf89534b85df..9eb8fff92df430be6440c4a2922f4895405bf48e 100644 (file)
--- a/src/amd/compiler/aco_instruction_selection.cpp
+++ b/src/amd/compiler/aco_instruction_selection.cpp
@@ -5223,7 +5223,7 @@ Temp emit_boolean_reduce(isel_context *ctx, nir_op op, unsigned cluster_size, Te
    } else if (op == nir_op_ixor && cluster_size == 64) {
       //subgroupXor(val) -> s_bcnt1_i32_b64(val & exec) & 1
       Temp tmp = bld.sop2(aco_opcode::s_and_b64, bld.def(s2), bld.def(s1, scc), src, Operand(exec, s2));
-      tmp = bld.sop1(aco_opcode::s_bcnt1_i32_b64, bld.def(s2), bld.def(s1, scc), tmp);
+      tmp = bld.sop1(aco_opcode::s_bcnt1_i32_b64, bld.def(s1), bld.def(s1, scc), tmp);
       tmp = bld.sop2(aco_opcode::s_and_b32, bld.def(s1), bld.def(s1, scc), tmp, Operand(1u)).def(1).getTemp();
       return bool_to_vector_condition(ctx, tmp);
    } else {