Mesa (master): aco: implement 16-bit nir_op_fsign

[GitLab Mirror]

Module: Mesa
Branch: master
Commit: 0646562a17a9649461b60fd8723e91dbf527e4a6
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=0646562a17a9649461b60fd8723e91dbf527e4a6

Author: Samuel Pitoiset <samuel.pitoiset at gmail.com>
Date:   Fri Apr  3 14:55:01 2020 +0200

aco: implement 16-bit nir_op_fsign

Signed-off-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>
Reviewed-by: Daniel Schürmann <daniel at schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4452>

---

 src/amd/compiler/aco_instruction_selection.cpp | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/src/amd/compiler/aco_instruction_selection.cpp b/src/amd/compiler/aco_instruction_selection.cpp
index 4f7edaea96e..79c49afb7c2 100644
--- a/src/amd/compiler/aco_instruction_selection.cpp
+++ b/src/amd/compiler/aco_instruction_selection.cpp
@@ -2091,12 +2091,20 @@ void visit_alu_instr(isel_context *ctx, nir_alu_instr *instr)
    }
    case nir_op_fsign: {
       Temp src = as_vgpr(ctx, get_alu_src(ctx, instr->src[0]));
-      if (dst.size() == 1) {
+      if (dst.regClass() == v2b) {
+         Temp one = bld.copy(bld.def(v1), Operand(0x3c00u));
+         Temp minus_one = bld.copy(bld.def(v1), Operand(0xbc00u));
+         Temp cond = bld.vopc(aco_opcode::v_cmp_nlt_f16, bld.hint_vcc(bld.def(bld.lm)), Operand(0u), src);
+         src = bld.vop2(aco_opcode::v_cndmask_b32, bld.def(v1), one, src, cond);
+         cond = bld.vopc(aco_opcode::v_cmp_le_f16, bld.hint_vcc(bld.def(bld.lm)), Operand(0u), src);
+         Temp tmp = bld.vop2(aco_opcode::v_cndmask_b32, bld.def(v1), minus_one, src, cond);
+         bld.pseudo(aco_opcode::p_split_vector, Definition(dst), bld.def(v2b), tmp);
+      } else if (dst.regClass() == v1) {
          Temp cond = bld.vopc(aco_opcode::v_cmp_nlt_f32, bld.hint_vcc(bld.def(bld.lm)), Operand(0u), src);
          src = bld.vop2(aco_opcode::v_cndmask_b32, bld.def(v1), Operand(0x3f800000u), src, cond);
          cond = bld.vopc(aco_opcode::v_cmp_le_f32, bld.hint_vcc(bld.def(bld.lm)), Operand(0u), src);
          bld.vop2(aco_opcode::v_cndmask_b32, Definition(dst), Operand(0xbf800000u), src, cond);
-      } else if (dst.size() == 2) {
+      } else if (dst.regClass() == v2) {
          Temp cond = bld.vopc(aco_opcode::v_cmp_nlt_f64, bld.hint_vcc(bld.def(bld.lm)), Operand(0u), src);
          Temp tmp = bld.vop1(aco_opcode::v_mov_b32, bld.def(v1), Operand(0x3FF00000u));
          Temp upper = bld.vop2_e64(aco_opcode::v_cndmask_b32, bld.def(v1), tmp, emit_extract_vector(ctx, src, 1, v1), cond);