KonstantinSeurer
/
mesa
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

aco: implement 16-bit nir_op_fsign 

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4452>
This commit is contained in:
Samuel Pitoiset 2020-04-03 14:55:01 +02:00
parent 6793ae1c5e
commit 0646562a17
1 changed files with 10 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
src/amd/compiler/aco_instruction_selection.cpp
View File

@ -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);