[llvm] [AMDGPU] Enable volatile and non-temporal for loads to LDS (PR #153244)

[Robert Imschweiler via llvm-commits]

================
@@ -218,3 +218,172 @@ main_body:
   ret void
 }
 
+define amdgpu_ps void @global_load_lds_dword_volatile(ptr addrspace(1) nocapture %gptr, ptr addrspace(3) inreg %lptr) {
+; GFX90A-LABEL: global_load_lds_dword_volatile:
+; GFX90A:       ; %bb.0: ; %main_body
+; GFX90A-NEXT:    s_mov_b32 m0, s0
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    global_load_dword v[0:1], off lds
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    global_load_dword v[0:1], off offset:256 lds
+; GFX90A-NEXT:    global_load_dword v[0:1], off offset:512 lds
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: global_load_lds_dword_volatile:
+; GFX942:       ; %bb.0: ; %main_body
+; GFX942-NEXT:    s_mov_b32 m0, s0
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    global_load_lds_dword v[0:1], off sc0 sc1
+; GFX942-NEXT:    s_waitcnt vmcnt(0)
+; GFX942-NEXT:    global_load_lds_dword v[0:1], off offset:256
+; GFX942-NEXT:    global_load_lds_dword v[0:1], off offset:512
+; GFX942-NEXT:    s_endpgm
+;
+; GFX10-LABEL: global_load_lds_dword_volatile:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    s_mov_b32 m0, s0
+; GFX10-NEXT:    global_load_dword v[0:1], off lds
+; GFX10-NEXT:    global_load_dword v[0:1], off offset:256 lds
+; GFX10-NEXT:    global_load_dword v[0:1], off offset:512 lds
+; GFX10-NEXT:    s_endpgm
+;
+; GFX942-GISEL-LABEL: global_load_lds_dword_volatile:
+; GFX942-GISEL:       ; %bb.0: ; %main_body
+; GFX942-GISEL-NEXT:    s_mov_b32 m0, s0
+; GFX942-GISEL-NEXT:    s_nop 0
+; GFX942-GISEL-NEXT:    global_load_lds_dword v[0:1], off sc0 sc1
+; GFX942-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX942-GISEL-NEXT:    global_load_lds_dword v[0:1], off offset:256
+; GFX942-GISEL-NEXT:    global_load_lds_dword v[0:1], off offset:512
+; GFX942-GISEL-NEXT:    s_endpgm
+main_body:
+  call void @llvm.amdgcn.load.to.lds.p1(ptr addrspace(1) %gptr, ptr addrspace(3) %lptr, i32 4, i32 0, i32 2147483648)
+  call void @llvm.amdgcn.load.to.lds.p1(ptr addrspace(1) %gptr, ptr addrspace(3) %lptr, i32 4, i32 256, i32 0)
+  call void @llvm.amdgcn.load.to.lds.p1(ptr addrspace(1) %gptr, ptr addrspace(3) %lptr, i32 4, i32 512, i32 0)
+  ret void
+}
+
+define amdgpu_ps void @buffer_load_lds_dword_volatile(ptr addrspace(7) nocapture inreg %gptr, i32 %off, ptr addrspace(3) inreg %lptr) {
+; GFX90A-LABEL: buffer_load_lds_dword_volatile:
+; GFX90A:       ; %bb.0: ; %main_body
+; GFX90A-NEXT:    v_add_u32_e32 v0, s4, v0
+; GFX90A-NEXT:    s_mov_b32 m0, s5
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    buffer_load_dword v0, s[0:3], 0 offen lds
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    buffer_load_dword v0, s[0:3], 0 offen offset:256 lds
+; GFX90A-NEXT:    buffer_load_dword v0, s[0:3], 0 offen offset:512 lds
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: buffer_load_lds_dword_volatile:
+; GFX942:       ; %bb.0: ; %main_body
+; GFX942-NEXT:    v_add_u32_e32 v0, s4, v0
+; GFX942-NEXT:    s_mov_b32 m0, s5
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    buffer_load_dword v0, s[0:3], 0 offen sc0 sc1 lds
+; GFX942-NEXT:    s_waitcnt vmcnt(0)
+; GFX942-NEXT:    buffer_load_dword v0, s[0:3], 0 offen offset:256 lds
+; GFX942-NEXT:    buffer_load_dword v0, s[0:3], 0 offen offset:512 lds
+; GFX942-NEXT:    s_endpgm
+;
+; GFX10-LABEL: buffer_load_lds_dword_volatile:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    v_add_nc_u32_e32 v0, s4, v0
+; GFX10-NEXT:    s_mov_b32 m0, s5
+; GFX10-NEXT:    buffer_load_dword v0, s[0:3], 0 offen lds
----------------
ro-i wrote:

same here

https://github.com/llvm/llvm-project/pull/153244