[llvm] eee82dc - AMDGPU/GlobalISel: Add some additional IR tests for zextload

[Matt Arsenault via llvm-commits]

Author: Matt Arsenault
Date: 2022-04-11T19:43:37-04:00
New Revision: eee82dc66d61c842f9edff1ee80315b5a8d80870

URL: https://github.com/llvm/llvm-project/commit/eee82dc66d61c842f9edff1ee80315b5a8d80870
DIFF: https://github.com/llvm/llvm-project/commit/eee82dc66d61c842f9edff1ee80315b5a8d80870.diff

LOG: AMDGPU/GlobalISel: Add some additional IR tests for zextload

Added: 
    

Modified: 
    llvm/test/CodeGen/AMDGPU/GlobalISel/zextload.ll

Removed: 
    


################################################################################
diff  --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/zextload.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/zextload.ll
index 7ce0f51136679..90c1137134991 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/zextload.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/zextload.ll
@@ -4,6 +4,121 @@
 ; RUN: llc -global-isel -mtriple=amdgcn-amd-amdpal -mcpu=tahiti < %s | FileCheck --check-prefix=GFX6 %s
 ; RUN: llc -global-isel -mtriple=amdgcn-amd-amdpal -mcpu=gfx1010 < %s | FileCheck --check-prefix=GFX10 %s
 
+define i32 @zextload_global_i1_to_i32(i1 addrspace(1)* %ptr) {
+; GFX9-LABEL: zextload_global_i1_to_i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_and_b32_e32 v0, 1, v0
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX8-LABEL: zextload_global_i1_to_i32:
+; GFX8:       ; %bb.0:
+; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX8-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX8-NEXT:    s_waitcnt vmcnt(0)
+; GFX8-NEXT:    v_and_b32_e32 v0, 1, v0
+; GFX8-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX6-LABEL: zextload_global_i1_to_i32:
+; GFX6:       ; %bb.0:
+; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_mov_b32 s6, 0
+; GFX6-NEXT:    s_mov_b32 s7, 0xf000
+; GFX6-NEXT:    s_mov_b64 s[4:5], 0
+; GFX6-NEXT:    buffer_load_ubyte v0, v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
+; GFX6-NEXT:    v_and_b32_e32 v0, 1, v0
+; GFX6-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: zextload_global_i1_to_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_and_b32_e32 v0, 1, v0
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %load = load i1, i1 addrspace(1)* %ptr
+  %ext = zext i1 %load to i32
+  ret i32 %ext
+}
+
+define i32 @zextload_global_i8_to_i32(i8 addrspace(1)* %ptr) {
+; GFX9-LABEL: zextload_global_i8_to_i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX8-LABEL: zextload_global_i8_to_i32:
+; GFX8:       ; %bb.0:
+; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX8-NEXT:    flat_load_ubyte v0, v[0:1]
+; GFX8-NEXT:    s_waitcnt vmcnt(0)
+; GFX8-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX6-LABEL: zextload_global_i8_to_i32:
+; GFX6:       ; %bb.0:
+; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_mov_b32 s6, 0
+; GFX6-NEXT:    s_mov_b32 s7, 0xf000
+; GFX6-NEXT:    s_mov_b64 s[4:5], 0
+; GFX6-NEXT:    buffer_load_ubyte v0, v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
+; GFX6-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: zextload_global_i8_to_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    global_load_ubyte v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %load = load i8, i8 addrspace(1)* %ptr
+  %ext = zext i8 %load to i32
+  ret i32 %ext
+}
+
+define i32 @zextload_global_i16_to_i32(i16 addrspace(1)* %ptr) {
+; GFX9-LABEL: zextload_global_i16_to_i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    global_load_ushort v0, v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX8-LABEL: zextload_global_i16_to_i32:
+; GFX8:       ; %bb.0:
+; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX8-NEXT:    flat_load_ushort v0, v[0:1]
+; GFX8-NEXT:    s_waitcnt vmcnt(0)
+; GFX8-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX6-LABEL: zextload_global_i16_to_i32:
+; GFX6:       ; %bb.0:
+; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    s_mov_b32 s6, 0
+; GFX6-NEXT:    s_mov_b32 s7, 0xf000
+; GFX6-NEXT:    s_mov_b64 s[4:5], 0
+; GFX6-NEXT:    buffer_load_ushort v0, v[0:1], s[4:7], 0 addr64
+; GFX6-NEXT:    s_waitcnt vmcnt(0)
+; GFX6-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: zextload_global_i16_to_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    global_load_ushort v0, v[0:1], off
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+  %load = load i16, i16 addrspace(1)* %ptr
+  %ext = zext i16 %load to i32
+  ret i32 %ext
+}
+
 define i64 @zextload_global_i1_to_i64(i1 addrspace(1)* %ptr) {
 ; GFX9-LABEL: zextload_global_i1_to_i64:
 ; GFX9:       ; %bb.0: