[llvm] [AMDGPU][GFX12] VOP encoding and codegen - add support for v_cvt fp8/… (PR #78414)

[Matt Arsenault via llvm-commits]

================
@@ -0,0 +1,105 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -march=amdgcn -mcpu=gfx1200 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX12 %s
+
+define amdgpu_cs float @test_cvt_f32_bf8_byte0(i32 %a) {
+; GFX12-LABEL: test_cvt_f32_bf8_byte0:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    v_cvt_f32_bf8_dpp v0, v0 quad_perm:[0,1,2,3] row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX12-NEXT:    ; return to shader part epilog
+  %tmp0 = call i32 @llvm.amdgcn.mov.dpp.i32(i32 %a, i32 228, i32 15, i32 15, i1 1)
+  %ret = tail call float @llvm.amdgcn.cvt.f32.bf8(i32 %tmp0, i32 0)
+  ret float %ret
+}
+
+define amdgpu_cs float @test_cvt_f32_bf8_byte2(i32 %a) {
+; GFX12-LABEL: test_cvt_f32_bf8_byte2:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    v_cvt_f32_bf8_e64_dpp v0, v0 op_sel:[0,1] quad_perm:[0,1,2,3] row_mask:0xf bank_mask:0xf bound_ctrl:1
+; GFX12-NEXT:    ; return to shader part epilog
+  %tmp0 = call i32 @llvm.amdgcn.mov.dpp.i32(i32 %a, i32 228, i32 15, i32 15, i1 1)
+  %ret = tail call float @llvm.amdgcn.cvt.f32.bf8(i32 %tmp0, i32 2)
----------------
arsenm wrote:

No test with byte1?

https://github.com/llvm/llvm-project/pull/78414