[llvm] [AMDGPU] Add the 3-dword image_gather4 variant for packed D16 + TFE (PR #215972)

[Matt Arsenault via llvm-commits]

================
@@ -0,0 +1,19 @@
+; RUN: llc < %s -mtriple=amdgpu9.08 | FileCheck -check-prefix=GCN %s
+
+; GCN-LABEL: {{^}}image_gather4_b_2d_v4f16_tfe_agpr:
+; GCN: image_gather4_b v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], s[0:7], s[8:11] dmask:0x4 tfe d16{{$}}
----------------
arsenm wrote:

This isn't using AGPRs, it's using VGPRs and copying to AGPRs. This would need to be a 90a target, and tricked into using AGPRs with asm 

https://github.com/llvm/llvm-project/pull/215972