From 7a3e5dd2ec1ac81317e1af075fdd0fe16aa67b1d Mon Sep 17 00:00:00 2001 From: Georg Lehmann Date: Tue, 8 Aug 2023 18:48:59 +0200 Subject: [PATCH] aco: use s_bitreplicate_b64_b32 to set exec to 0xffff0000ffff0000 Foz-DB Navi21: Totals from 29 (0.02% of 132657) affected shaders: Instrs: 19342 -> 19301 (-0.21%) Reviewed-by: Rhys Perry Part-of: --- src/amd/compiler/aco_lower_to_hw_instr.cpp | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/src/amd/compiler/aco_lower_to_hw_instr.cpp b/src/amd/compiler/aco_lower_to_hw_instr.cpp index ea43c3f0cef..8aa12328623 100644 --- a/src/amd/compiler/aco_lower_to_hw_instr.cpp +++ b/src/amd/compiler/aco_lower_to_hw_instr.cpp @@ -869,10 +869,11 @@ emit_reduction(lower_context* ctx, aco_opcode op, ReduceOp reduce_op, unsigned c emit_dpp_op(ctx, tmp, tmp, tmp, vtmp, reduce_op, src.size(), dpp_row_sr(8), 0xf, 0xf, false, identity); if (ctx->program->gfx_level >= GFX10) { - bld.sop2(aco_opcode::s_bfm_b32, Definition(exec_lo, s1), Operand::c32(16u), - Operand::c32(16u)); if (ctx->program->wave_size == 64) { - bld.sop2(aco_opcode::s_bfm_b32, Definition(exec_hi, s1), Operand::c32(16u), + bld.sop1(aco_opcode::s_bitreplicate_b64_b32, Definition(exec, s2), + Operand::c32(0xff00ff00u)); + } else { + bld.sop2(aco_opcode::s_bfm_b32, Definition(exec_lo, s1), Operand::c32(16u), Operand::c32(16u)); } for (unsigned i = 0; i < src.size(); i++) {