[llvm] [AMDGPU] Replace dynamic VGPR feature with attribute (PR #133444)

[LLVM Continuous Integration via llvm-commits]

llvm-ci wrote:

LLVM Buildbot has detected a new failure on builder `llvm-clang-x86_64-gcc-ubuntu` running on `sie-linux-worker3` while building `llvm` at step 6 "test-build-unified-tree-check-all".

Full details are available at: https://lab.llvm.org/buildbot/#/builders/174/builds/19892

<details>
<summary>Here is the relevant piece of the build log for the reference</summary>

```
Step 6 (test-build-unified-tree-check-all) failure: test (failure)
******************** TEST 'LLVM :: CodeGen/AMDGPU/pal-metadata-3.0-callable-dvgpr.ll' FAILED ********************
Exit Code: 1

Command Output (stderr):
--
/home/buildbot/buildbot-root/llvm-clang-x86_64-gcc-ubuntu/build/bin/llc -mtriple=amdgcn--amdpal -mcpu=gfx1200 -verify-machineinstrs < /home/buildbot/buildbot-root/llvm-clang-x86_64-gcc-ubuntu/llvm-project/llvm/test/CodeGen/AMDGPU/pal-metadata-3.0-callable-dvgpr.ll | /home/buildbot/buildbot-root/llvm-clang-x86_64-gcc-ubuntu/build/bin/FileCheck /home/buildbot/buildbot-root/llvm-clang-x86_64-gcc-ubuntu/llvm-project/llvm/test/CodeGen/AMDGPU/pal-metadata-3.0-callable-dvgpr.ll # RUN: at line 1
+ /home/buildbot/buildbot-root/llvm-clang-x86_64-gcc-ubuntu/build/bin/llc -mtriple=amdgcn--amdpal -mcpu=gfx1200 -verify-machineinstrs
+ /home/buildbot/buildbot-root/llvm-clang-x86_64-gcc-ubuntu/build/bin/FileCheck /home/buildbot/buildbot-root/llvm-clang-x86_64-gcc-ubuntu/llvm-project/llvm/test/CodeGen/AMDGPU/pal-metadata-3.0-callable-dvgpr.ll
/home/buildbot/buildbot-root/llvm-clang-x86_64-gcc-ubuntu/llvm-project/llvm/test/CodeGen/AMDGPU/pal-metadata-3.0-callable-dvgpr.ll:83:15: error: CHECK-NEXT: expected string not found in input
; CHECK-NEXT: .sgpr_count: 0x1
              ^
<stdin>:933:14: note: scanning from here
 .lds_size: 0
             ^
<stdin>:934:2: note: possible intended match here
 .sgpr_count: 0x21
 ^

Input file: <stdin>
Check file: /home/buildbot/buildbot-root/llvm-clang-x86_64-gcc-ubuntu/llvm-project/llvm/test/CodeGen/AMDGPU/pal-metadata-3.0-callable-dvgpr.ll

-dump-input=help explains the following input dump.

Input was:
<<<<<<
           1:  .amdgcn_target "amdgcn-unknown-amdpal--gfx1200" 
           2:  .text 
           3:  .globl no_stack ; -- Begin function no_stack 
           4:  .p2align 2 
           5:  .type no_stack, at function 
           6: no_stack: ; @no_stack 
           7: ; %bb.0: 
           8:  s_wait_loadcnt_dscnt 0x0 
           9:  s_wait_expcnt 0x0 
          10:  s_wait_samplecnt 0x0 
          11:  s_wait_bvhcnt 0x0 
          12:  s_wait_kmcnt 0x0 
          13:  v_add_f32_e32 v0, 1.0, v0 
          14:  s_setpc_b64 s[30:31] 
          15: .Lfunc_end0: 
          16:  .size no_stack, .Lfunc_end0-no_stack 
          17:  ; -- End function 
          18:  .set no_stack.num_vgpr, 1 
          19:  .set no_stack.num_agpr, 0 
          20:  .set no_stack.numbered_sgpr, 32 
          21:  .set no_stack.private_seg_size, 0 
          22:  .set no_stack.uses_vcc, 0 
          23:  .set no_stack.uses_flat_scratch, 0 
          24:  .set no_stack.has_dyn_sized_stack, 0 
          25:  .set no_stack.has_recursion, 0 
...

```

</details>

https://github.com/llvm/llvm-project/pull/133444