radeonsi: use v_mad_f32 for fma

v_fma_f32 runs at FP64 rate (= slow). Alien Isolation and F1 2015 seem
to use fma for all d3d multiply-add instructions, which is silly.

This tries to restore performance for those games.

The main difference between v_mad_f32 and v_fma_f32 is that v_mad doesn't
support denormals, which we don't enable anyway, because they are slow too.

Also, there is code size reduction:
  Totals from affected shaders:
  VGPRS: 109796 -> 109808 (0.01 %)
  Spilled SGPRs: 29995 -> 30022 (0.09 %)
  Spilled VGPRs: 12 -> 13 (8.33 %) <-- it's just one shader going from 12 to 13
  Code Size: 6667596 -> 6476356 (-2.87 %) bytes
  Max Waves: 26931 -> 26899 (-0.12 %)

I've not actually tested real performance.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

diff --git a/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c b/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c
index d382496387fbd09569c3f17ee101d89e9f2e0e9d..4c603fb1aadef27e6ea644afb7510e5f80402ca6 100644 (file)
--- a/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c
+++ b/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c
@@ -1769,8 +1769,8 @@ void radeon_llvm_context_init(struct radeon_llvm_context *ctx, const char *tripl
                HAVE_LLVM >= 0x0308 ? "llvm.exp2.f32" : "llvm.AMDIL.exp.";
        bld_base->op_actions[TGSI_OPCODE_FLR].emit = build_tgsi_intrinsic_nomem;
        bld_base->op_actions[TGSI_OPCODE_FLR].intr_name = "llvm.floor.f32";
-       bld_base->op_actions[TGSI_OPCODE_FMA].emit = build_tgsi_intrinsic_nomem;
-       bld_base->op_actions[TGSI_OPCODE_FMA].intr_name = "llvm.fma.f32";
+       bld_base->op_actions[TGSI_OPCODE_FMA].emit =
+               bld_base->op_actions[TGSI_OPCODE_MAD].emit;
        bld_base->op_actions[TGSI_OPCODE_FRC].emit = emit_frac;
        bld_base->op_actions[TGSI_OPCODE_F2I].emit = emit_f2i;
        bld_base->op_actions[TGSI_OPCODE_F2U].emit = emit_f2u;