ac: generate FMA for inexact instructions for radeonsi

[mesa.git] / src / amd / llvm / ac_nir_to_llvm.c

diff --git a/src/amd/llvm/ac_nir_to_llvm.c b/src/amd/llvm/ac_nir_to_llvm.c
index 627f5d2d931212ccff4dfc57db393c0348acc49d..03717191e243fd8b897d42e08f7c7c5639255d0f 100644 (file)
--- a/src/amd/llvm/ac_nir_to_llvm.c
+++ b/src/amd/llvm/ac_nir_to_llvm.c
@@ -589,6 +589,10 @@ static void visit_alu(struct ac_nir_context *ctx, const nir_alu_instr *instr)
        unsigned num_components = instr->dest.dest.ssa.num_components;
        unsigned src_components;
        LLVMTypeRef def_type = get_def_type(ctx, &instr->dest.dest.ssa);
+       bool saved_inexact = false;
+
+       if (instr->exact)
+               saved_inexact = ac_disable_inexact_math(ctx->ac.builder);
 
        assert(nir_op_infos[instr->op].num_inputs <= ARRAY_SIZE(src));
        switch (instr->op) {
@@ -1182,6 +1186,9 @@ static void visit_alu(struct ac_nir_context *ctx, const nir_alu_instr *instr)
                result = ac_to_integer_or_pointer(&ctx->ac, result);
                ctx->ssa_defs[instr->dest.dest.ssa.index] = result;
        }
+
+       if (instr->exact)
+               ac_restore_inexact_math(ctx->ac.builder, saved_inexact);
 }
 
 static void visit_load_const(struct ac_nir_context *ctx,