[llvm] [AMDGPU][MC][True16] VOP3dot instruction update for true16/fake16 (PR #113474)

[Brox Chen via llvm-commits]

================
@@ -1983,105 +1983,228 @@
 # W64: v_div_scale_f64 v[254:255], vcc, 0xaf123456, -vcc, -1 clamp div:2 ; encoding: [0xfe,0xea,0xfd,0xd6,0xff,0xd4,0x04,0x5b,0x56,0x34,0x12,0xaf]
 0xfe,0xea,0xfd,0xd6,0xff,0xd4,0x04,0x5b,0x56,0x34,0x12,0xaf
 
-# GFX11: v_dot2_bf16_bf16 v5, v1, v2, s3         ; encoding: [0x05,0x00,0x67,0xd6,0x01,0x05,0x0e,0x00]
+# W32-REAL16: v_dot2_bf16_bf16 v5.l, v1, v2, s3       ; encoding: [0x05,0x00,0x67,0xd6,0x01,0x05,0x0e,0x00]
+# W32-FAKE16: v_dot2_bf16_bf16 v5, v1, v2, s3         ; encoding: [0x05,0x00,0x67,0xd6,0x01,0x05,0x0e,0x00]
+# W64-REAL16: v_dot2_bf16_bf16 v5.l, v1, v2, s3       ; encoding: [0x05,0x00,0x67,0xd6,0x01,0x05,0x0e,0x00]
+# W64-FAKE16: v_dot2_bf16_bf16 v5, v1, v2, s3         ; encoding: [0x05,0x00,0x67,0xd6,0x01,0x05,0x0e,0x00]
----------------
broxigarchen wrote:

I think we can keep it as it for now since fake16 testline will be removed later. I can merge those in the later patch

https://github.com/llvm/llvm-project/pull/113474