From 3a175b54a44389cf18f43960027c2df297d1de0a Mon Sep 17 00:00:00 2001
From: Georg Lehmann <dadschoorse@gmail.com>
Date: Sat, 6 Sep 2025 21:02:18 +0200
Subject: [PATCH] aco,nir: support subdword v_permlane_b16

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38389>
---
 .../instruction_selection/aco_select_nir_intrinsics.cpp        | 3 ++-
 src/compiler/nir/nir_intrinsics.py                             | 2 +-
 2 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/src/amd/compiler/instruction_selection/aco_select_nir_intrinsics.cpp b/src/amd/compiler/instruction_selection/aco_select_nir_intrinsics.cpp
index 92fee82b7bf..53ca8824385 100644
--- a/src/amd/compiler/instruction_selection/aco_select_nir_intrinsics.cpp
+++ b/src/amd/compiler/instruction_selection/aco_select_nir_intrinsics.cpp
@@ -4707,7 +4707,8 @@ visit_intrinsic(isel_context* ctx, nir_intrinsic_instr* instr)
 
       if (src.regClass() == s1) {
          bld.copy(Definition(dst), src);
-      } else if (dst.regClass() == v1 && src.regClass() == v1) {
+      } else if (dst.type() == RegType::vgpr && src.type() == RegType::vgpr && dst.size() == 1 &&
+                 src.size() == 1) {
          bld.vop3(aco_opcode::v_permlane16_b32, Definition(dst), src,
                   bld.as_uniform(get_ssa_temp(ctx, instr->src[1].ssa)),
                   bld.as_uniform(get_ssa_temp(ctx, instr->src[2].ssa)));
diff --git a/src/compiler/nir/nir_intrinsics.py b/src/compiler/nir/nir_intrinsics.py
index fc98e474202..c97dd34831a 100644
--- a/src/compiler/nir/nir_intrinsics.py
+++ b/src/compiler/nir/nir_intrinsics.py
@@ -623,7 +623,7 @@ intrinsic("write_invocation_amd", src_comp=[0, 0, 1], dest_comp=0, bit_sizes=src
 # src = [ mask, addition ]
 intrinsic("mbcnt_amd", src_comp=[1, 1], dest_comp=1, bit_sizes=[32], flags=[CAN_REORDER, CAN_ELIMINATE])
 # Compiled to v_permlane16_b32. src = [ value, lanesel_lo, lanesel_hi ]
-intrinsic("lane_permute_16_amd", src_comp=[1, 1, 1], dest_comp=1, bit_sizes=[32], flags=[CAN_ELIMINATE])
+intrinsic("lane_permute_16_amd", src_comp=[1, 1, 1], dest_comp=1, bit_sizes=src0, flags=[CAN_ELIMINATE])
 # subgroup shuffle up/down with cluster size 16.
 # base in [-15, -1]: DPP_ROW_SR
 # base in [  1, 15]: DPP_ROW_SL, otherwise invalid.