[llvm] 741e1ca - [AMDGPU] upgrade test to use auto-generated llc checks (#145832)
via llvm-commits
llvm-commits at lists.llvm.org
Thu Jun 26 02:00:56 PDT 2025
Author: Sameer Sahasrabuddhe
Date: 2025-06-26T14:30:53+05:30
New Revision: 741e1ca3b33146538a3a33c26d3e992ac65dac08
URL: https://github.com/llvm/llvm-project/commit/741e1ca3b33146538a3a33c26d3e992ac65dac08
DIFF: https://github.com/llvm/llvm-project/commit/741e1ca3b33146538a3a33c26d3e992ac65dac08.diff
LOG: [AMDGPU] upgrade test to use auto-generated llc checks (#145832)
This makes it easier to fix the failure in #145720
Added:
Modified:
llvm/test/CodeGen/AMDGPU/noclobber-barrier.ll
Removed:
################################################################################
diff --git a/llvm/test/CodeGen/AMDGPU/noclobber-barrier.ll b/llvm/test/CodeGen/AMDGPU/noclobber-barrier.ll
index a8fab161e1ffb..34260c49ff92c 100644
--- a/llvm/test/CodeGen/AMDGPU/noclobber-barrier.ll
+++ b/llvm/test/CodeGen/AMDGPU/noclobber-barrier.ll
@@ -1,4 +1,5 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: opt -mtriple=amdgcn -mcpu=gfx900 -amdgpu-aa -amdgpu-aa-wrapper -amdgpu-annotate-uniform -S < %s | FileCheck %s
; RUN: llc -mtriple=amdgcn -mcpu=gfx900 -verify-machineinstrs -amdgpu-atomic-optimizer-strategy=None < %s | FileCheck -check-prefix=GCN %s
@@ -7,16 +8,6 @@
@LDS = linkonce_odr hidden local_unnamed_addr addrspace(3) global i32 poison
-; GCN-LABEL: {{^}}simple_barrier:
-; GCN: s_load_dword s
-; GCN: s_waitcnt lgkmcnt(0)
-; GCN: s_barrier
-; GCN: s_waitcnt lgkmcnt(0)
-; GCN: ; wave barrier
-; GCN-NOT: global_load_dword
-; GCN: s_load_dword s
-; GCN-NOT: global_load_dword
-; GCN: global_store_dword
define amdgpu_kernel void @simple_barrier(ptr addrspace(1) %arg) {
; CHECK-LABEL: @simple_barrier(
; CHECK-NEXT: bb:
@@ -32,6 +23,22 @@ define amdgpu_kernel void @simple_barrier(ptr addrspace(1) %arg) {
; CHECK-NEXT: store i32 [[I3]], ptr addrspace(1) [[I4]], align 4
; CHECK-NEXT: ret void
;
+; GCN-LABEL: simple_barrier:
+; GCN: ; %bb.0: ; %bb
+; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GCN-NEXT: v_mov_b32_e32 v0, 0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_load_dword s2, s[0:1], 0x0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_barrier
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: ; wave barrier
+; GCN-NEXT: s_load_dword s3, s[0:1], 0x4
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_add_i32 s2, s3, s2
+; GCN-NEXT: v_mov_b32_e32 v1, s2
+; GCN-NEXT: global_store_dword v0, v1, s[0:1] offset:8
+; GCN-NEXT: s_endpgm
bb:
%i = load i32, ptr addrspace(1) %arg, align 4
fence syncscope("workgroup") release
@@ -46,15 +53,6 @@ bb:
ret void
}
-; GCN-LABEL: {{^}}memory_phi_no_clobber:
-; GCN: s_load_dword s
-; GCN: s_waitcnt lgkmcnt(0)
-; GCN: s_waitcnt lgkmcnt(0)
-; GCN: s_barrier
-; GCN-NOT: global_load_dword
-; GCN: s_load_dword s
-; GCN-NOT: global_load_dword
-; GCN: global_store_dword
define amdgpu_kernel void @memory_phi_no_clobber(ptr addrspace(1) %arg, i1 %cond) {
; CHECK-LABEL: @memory_phi_no_clobber(
; CHECK-NEXT: bb:
@@ -74,6 +72,32 @@ define amdgpu_kernel void @memory_phi_no_clobber(ptr addrspace(1) %arg, i1 %cond
; CHECK-NEXT: store i32 [[I3]], ptr addrspace(1) [[I4]], align 4
; CHECK-NEXT: ret void
;
+; GCN-LABEL: memory_phi_no_clobber:
+; GCN: ; %bb.0: ; %bb
+; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GCN-NEXT: s_load_dword s2, s[4:5], 0x2c
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_bitcmp0_b32 s2, 0
+; GCN-NEXT: s_load_dword s4, s[0:1], 0x0
+; GCN-NEXT: s_mov_b64 s[2:3], -1
+; GCN-NEXT: s_cbranch_scc0 .LBB1_2
+; GCN-NEXT: ; %bb.1: ; %if.else
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_mov_b64 s[2:3], 0
+; GCN-NEXT: .LBB1_2: ; %Flow
+; GCN-NEXT: s_andn2_b64 vcc, exec, s[2:3]
+; GCN-NEXT: s_cbranch_vccnz .LBB1_4
+; GCN-NEXT: ; %bb.3: ; %if.then
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_barrier
+; GCN-NEXT: .LBB1_4: ; %if.end
+; GCN-NEXT: s_load_dword s2, s[0:1], 0x4
+; GCN-NEXT: v_mov_b32_e32 v0, 0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_add_i32 s2, s2, s4
+; GCN-NEXT: v_mov_b32_e32 v1, s2
+; GCN-NEXT: global_store_dword v0, v1, s[0:1] offset:8
+; GCN-NEXT: s_endpgm
bb:
%i = load i32, ptr addrspace(1) %arg, align 4
br i1 %cond, label %if.then, label %if.else
@@ -95,12 +119,6 @@ if.end:
ret void
}
-; GCN-LABEL: {{^}}memory_phi_clobber1:
-; GCN: s_load_dword s
-; GCN: s_barrier
-; GCN: global_store_dword
-; GCN: global_load_dword
-; GCN: global_store_dword
define amdgpu_kernel void @memory_phi_clobber1(ptr addrspace(1) %arg, i1 %cond) {
; CHECK-LABEL: @memory_phi_clobber1(
; CHECK-NEXT: bb:
@@ -121,6 +139,33 @@ define amdgpu_kernel void @memory_phi_clobber1(ptr addrspace(1) %arg, i1 %cond)
; CHECK-NEXT: store i32 [[I3]], ptr addrspace(1) [[I4]], align 4
; CHECK-NEXT: ret void
;
+; GCN-LABEL: memory_phi_clobber1:
+; GCN: ; %bb.0: ; %bb
+; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GCN-NEXT: s_load_dword s2, s[4:5], 0x2c
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_bitcmp0_b32 s2, 0
+; GCN-NEXT: s_load_dword s4, s[0:1], 0x0
+; GCN-NEXT: s_mov_b64 s[2:3], -1
+; GCN-NEXT: s_cbranch_scc0 .LBB2_2
+; GCN-NEXT: ; %bb.1: ; %if.else
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_barrier
+; GCN-NEXT: s_mov_b64 s[2:3], 0
+; GCN-NEXT: .LBB2_2: ; %Flow
+; GCN-NEXT: s_andn2_b64 vcc, exec, s[2:3]
+; GCN-NEXT: s_cbranch_vccnz .LBB2_4
+; GCN-NEXT: ; %bb.3: ; %if.then
+; GCN-NEXT: v_mov_b32_e32 v0, 0
+; GCN-NEXT: v_mov_b32_e32 v1, 1
+; GCN-NEXT: global_store_dword v0, v1, s[0:1] offset:12
+; GCN-NEXT: .LBB2_4: ; %if.end
+; GCN-NEXT: v_mov_b32_e32 v0, 0
+; GCN-NEXT: global_load_dword v1, v0, s[0:1] offset:4
+; GCN-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
+; GCN-NEXT: v_add_u32_e32 v1, s4, v1
+; GCN-NEXT: global_store_dword v0, v1, s[0:1] offset:8
+; GCN-NEXT: s_endpgm
bb:
%i = load i32, ptr addrspace(1) %arg, align 4
br i1 %cond, label %if.then, label %if.else
@@ -143,12 +188,6 @@ if.end:
ret void
}
-; GCN-LABEL: {{^}}memory_phi_clobber2:
-; GCN-DAG: s_load_dword s
-; GCN-DAG: global_store_dword
-; GCN: s_barrier
-; GCN: global_load_dword
-; GCN: global_store_dword
define amdgpu_kernel void @memory_phi_clobber2(ptr addrspace(1) %arg, i1 %cond) {
; CHECK-LABEL: @memory_phi_clobber2(
; CHECK-NEXT: bb:
@@ -169,6 +208,33 @@ define amdgpu_kernel void @memory_phi_clobber2(ptr addrspace(1) %arg, i1 %cond)
; CHECK-NEXT: store i32 [[I3]], ptr addrspace(1) [[I4]], align 4
; CHECK-NEXT: ret void
;
+; GCN-LABEL: memory_phi_clobber2:
+; GCN: ; %bb.0: ; %bb
+; GCN-NEXT: s_load_dword s2, s[4:5], 0x2c
+; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_bitcmp0_b32 s2, 0
+; GCN-NEXT: s_mov_b64 s[2:3], -1
+; GCN-NEXT: s_cbranch_scc0 .LBB3_2
+; GCN-NEXT: ; %bb.1: ; %if.else
+; GCN-NEXT: v_mov_b32_e32 v0, 0
+; GCN-NEXT: v_mov_b32_e32 v1, 1
+; GCN-NEXT: global_store_dword v0, v1, s[0:1] offset:12
+; GCN-NEXT: s_mov_b64 s[2:3], 0
+; GCN-NEXT: .LBB3_2: ; %Flow
+; GCN-NEXT: s_load_dword s4, s[0:1], 0x0
+; GCN-NEXT: s_andn2_b64 vcc, exec, s[2:3]
+; GCN-NEXT: s_cbranch_vccnz .LBB3_4
+; GCN-NEXT: ; %bb.3: ; %if.then
+; GCN-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
+; GCN-NEXT: s_barrier
+; GCN-NEXT: .LBB3_4: ; %if.end
+; GCN-NEXT: v_mov_b32_e32 v0, 0
+; GCN-NEXT: global_load_dword v1, v0, s[0:1] offset:4
+; GCN-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
+; GCN-NEXT: v_add_u32_e32 v1, s4, v1
+; GCN-NEXT: global_store_dword v0, v1, s[0:1] offset:8
+; GCN-NEXT: s_endpgm
bb:
%i = load i32, ptr addrspace(1) %arg, align 4
br i1 %cond, label %if.then, label %if.else
@@ -191,11 +257,6 @@ if.end:
ret void
}
-; GCN-LABEL: {{^}}no_clobbering_loop1:
-; GCN: s_load_dword s
-; GCN: s_load_dword s
-; GCN-NOT: global_load_dword
-; GCN: global_store_dword
define amdgpu_kernel void @no_clobbering_loop1(ptr addrspace(1) %arg, i1 %cc) {
; CHECK-LABEL: @no_clobbering_loop1(
; CHECK-NEXT: bb:
@@ -212,6 +273,30 @@ define amdgpu_kernel void @no_clobbering_loop1(ptr addrspace(1) %arg, i1 %cc) {
; CHECK: end:
; CHECK-NEXT: ret void
;
+; GCN-LABEL: no_clobbering_loop1:
+; GCN: ; %bb.0: ; %bb
+; GCN-NEXT: s_load_dword s0, s[4:5], 0x2c
+; GCN-NEXT: s_load_dwordx2 s[2:3], s[4:5], 0x24
+; GCN-NEXT: v_mov_b32_e32 v0, 0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_bitcmp1_b32 s0, 0
+; GCN-NEXT: s_load_dword s4, s[2:3], 0x0
+; GCN-NEXT: s_cselect_b64 s[0:1], -1, 0
+; GCN-NEXT: s_xor_b64 s[0:1], s[0:1], -1
+; GCN-NEXT: v_cndmask_b32_e64 v1, 0, 1, s[0:1]
+; GCN-NEXT: v_cmp_ne_u32_e64 s[0:1], 1, v1
+; GCN-NEXT: .LBB4_1: ; %while.cond
+; GCN-NEXT: ; =>This Inner Loop Header: Depth=1
+; GCN-NEXT: s_load_dword s5, s[2:3], 0x4
+; GCN-NEXT: s_and_b64 vcc, exec, s[0:1]
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_add_i32 s5, s5, s4
+; GCN-NEXT: v_mov_b32_e32 v1, s5
+; GCN-NEXT: global_store_dword v0, v1, s[2:3] offset:8
+; GCN-NEXT: ; wave barrier
+; GCN-NEXT: s_cbranch_vccnz .LBB4_1
+; GCN-NEXT: ; %bb.2: ; %end
+; GCN-NEXT: s_endpgm
bb:
%i = load i32, ptr addrspace(1) %arg, align 4
br label %while.cond
@@ -229,11 +314,6 @@ end:
ret void
}
-; GCN-LABEL: {{^}}no_clobbering_loop2:
-; GCN: s_load_dword s
-; GCN: s_load_dword s
-; GCN-NOT: global_load_dword
-; GCN: global_store_dword
define amdgpu_kernel void @no_clobbering_loop2(ptr addrspace(1) noalias %arg, ptr addrspace(1) noalias %out, i32 %n) {
; CHECK-LABEL: @no_clobbering_loop2(
; CHECK-NEXT: bb:
@@ -253,6 +333,28 @@ define amdgpu_kernel void @no_clobbering_loop2(ptr addrspace(1) noalias %arg, pt
; CHECK-NEXT: store i32 [[I3]], ptr addrspace(1) [[OUT:%.*]], align 4
; CHECK-NEXT: ret void
;
+; GCN-LABEL: no_clobbering_loop2:
+; GCN: ; %bb.0: ; %bb
+; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GCN-NEXT: s_load_dword s6, s[4:5], 0x34
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_load_dword s4, s[0:1], 0x0
+; GCN-NEXT: .LBB5_1: ; %while.cond
+; GCN-NEXT: ; =>This Inner Loop Header: Depth=1
+; GCN-NEXT: s_load_dword s5, s[0:1], 0x0
+; GCN-NEXT: ; wave barrier
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_add_i32 s4, s5, s4
+; GCN-NEXT: s_add_u32 s0, s0, 4
+; GCN-NEXT: s_addc_u32 s1, s1, 0
+; GCN-NEXT: s_add_i32 s6, s6, -1
+; GCN-NEXT: s_cmp_eq_u32 s6, 0
+; GCN-NEXT: s_cbranch_scc1 .LBB5_1
+; GCN-NEXT: ; %bb.2: ; %end
+; GCN-NEXT: v_mov_b32_e32 v0, 0
+; GCN-NEXT: v_mov_b32_e32 v1, s4
+; GCN-NEXT: global_store_dword v0, v1, s[2:3]
+; GCN-NEXT: s_endpgm
bb:
%i = load i32, ptr addrspace(1) %arg, align 4
br label %while.cond
@@ -273,10 +375,6 @@ end:
ret void
}
-; GCN-LABEL: {{^}}clobbering_loop:
-; GCN: s_load_dword s
-; GCN: global_load_dword
-; GCN: global_store_dword
define amdgpu_kernel void @clobbering_loop(ptr addrspace(1) %arg, ptr addrspace(1) %out, i1 %cc) {
; CHECK-LABEL: @clobbering_loop(
; CHECK-NEXT: bb:
@@ -293,6 +391,29 @@ define amdgpu_kernel void @clobbering_loop(ptr addrspace(1) %arg, ptr addrspace(
; CHECK: end:
; CHECK-NEXT: ret void
;
+; GCN-LABEL: clobbering_loop:
+; GCN: ; %bb.0: ; %bb
+; GCN-NEXT: s_load_dword s0, s[4:5], 0x34
+; GCN-NEXT: s_load_dwordx4 s[8:11], s[4:5], 0x24
+; GCN-NEXT: v_mov_b32_e32 v0, 0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_bitcmp1_b32 s0, 0
+; GCN-NEXT: s_load_dword s2, s[8:9], 0x0
+; GCN-NEXT: s_cselect_b64 s[0:1], -1, 0
+; GCN-NEXT: s_xor_b64 s[0:1], s[0:1], -1
+; GCN-NEXT: v_cndmask_b32_e64 v1, 0, 1, s[0:1]
+; GCN-NEXT: v_cmp_ne_u32_e64 s[0:1], 1, v1
+; GCN-NEXT: .LBB6_1: ; %while.cond
+; GCN-NEXT: ; =>This Inner Loop Header: Depth=1
+; GCN-NEXT: global_load_dword v1, v0, s[8:9] offset:4
+; GCN-NEXT: s_and_b64 vcc, exec, s[0:1]
+; GCN-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
+; GCN-NEXT: v_add_u32_e32 v1, s2, v1
+; GCN-NEXT: global_store_dword v0, v1, s[10:11] offset:4
+; GCN-NEXT: ; wave barrier
+; GCN-NEXT: s_cbranch_vccnz .LBB6_1
+; GCN-NEXT: ; %bb.2: ; %end
+; GCN-NEXT: s_endpgm
bb:
%i = load i32, ptr addrspace(1) %arg, align 4
br label %while.cond
@@ -310,11 +431,6 @@ end:
ret void
}
-; GCN-LABEL: {{^}}clobber_by_atomic_load:
-; GCN: s_load_dword s
-; GCN: global_load_dword {{.*}} glc
-; GCN: global_load_dword
-; GCN: global_store_dword
define amdgpu_kernel void @clobber_by_atomic_load(ptr addrspace(1) %arg) {
; CHECK-LABEL: @clobber_by_atomic_load(
; CHECK-NEXT: bb:
@@ -328,6 +444,21 @@ define amdgpu_kernel void @clobber_by_atomic_load(ptr addrspace(1) %arg) {
; CHECK-NEXT: store i32 [[I3]], ptr addrspace(1) [[I4]], align 4
; CHECK-NEXT: ret void
;
+; GCN-LABEL: clobber_by_atomic_load:
+; GCN: ; %bb.0: ; %bb
+; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GCN-NEXT: v_mov_b32_e32 v0, 0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_load_dword s2, s[0:1], 0x0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: global_load_dword v1, v0, s[0:1] offset:8 glc
+; GCN-NEXT: s_waitcnt vmcnt(0)
+; GCN-NEXT: buffer_wbinvl1_vol
+; GCN-NEXT: global_load_dword v1, v0, s[0:1] offset:12
+; GCN-NEXT: s_waitcnt vmcnt(0)
+; GCN-NEXT: v_add_u32_e32 v1, s2, v1
+; GCN-NEXT: global_store_dword v0, v1, s[0:1] offset:16
+; GCN-NEXT: s_endpgm
bb:
%i = load i32, ptr addrspace(1) %arg, align 4
%gep = getelementptr inbounds i32, ptr addrspace(1) %arg, i64 2
@@ -340,12 +471,6 @@ bb:
ret void
}
-; GCN-LABEL: {{^}}no_alias_store:
-; GCN: ds_write_b32
-; GCN: s_barrier
-; GCN: s_load_dword s
-; GCN-NOT: global_load_dword
-; GCN: global_store_dword
define protected amdgpu_kernel void @no_alias_store(ptr addrspace(1) %in, ptr addrspace(1) %out) {
; CHECK-LABEL: @no_alias_store(
; CHECK-NEXT: entry:
@@ -357,6 +482,18 @@ define protected amdgpu_kernel void @no_alias_store(ptr addrspace(1) %in, ptr ad
; CHECK-NEXT: store i32 [[LD]], ptr addrspace(1) [[OUT:%.*]], align 4
; CHECK-NEXT: ret void
;
+; GCN-LABEL: no_alias_store:
+; GCN: ; %bb.0: ; %entry
+; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GCN-NEXT: v_mov_b32_e32 v0, 0
+; GCN-NEXT: ds_write_b32 v0, v0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_barrier
+; GCN-NEXT: s_load_dword s0, s[0:1], 0x0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: v_mov_b32_e32 v1, s0
+; GCN-NEXT: global_store_dword v0, v1, s[2:3]
+; GCN-NEXT: s_endpgm
entry:
store i32 0, ptr addrspace(3) @LDS, align 4
fence syncscope("workgroup") release
@@ -367,11 +504,6 @@ entry:
ret void
}
-; GCN-LABEL: {{^}}may_alias_store:
-; GCN: global_store_dword
-; GCN: s_barrier
-; GCN: global_load_dword
-; GCN: global_store_dword
define protected amdgpu_kernel void @may_alias_store(ptr addrspace(1) %in, ptr addrspace(1) %out) {
; CHECK-LABEL: @may_alias_store(
; CHECK-NEXT: entry:
@@ -383,6 +515,18 @@ define protected amdgpu_kernel void @may_alias_store(ptr addrspace(1) %in, ptr a
; CHECK-NEXT: store i32 [[LD]], ptr addrspace(1) [[OUT]], align 4
; CHECK-NEXT: ret void
;
+; GCN-LABEL: may_alias_store:
+; GCN: ; %bb.0: ; %entry
+; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GCN-NEXT: v_mov_b32_e32 v0, 0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: global_store_dword v0, v0, s[2:3]
+; GCN-NEXT: s_waitcnt vmcnt(0)
+; GCN-NEXT: s_barrier
+; GCN-NEXT: global_load_dword v1, v0, s[0:1]
+; GCN-NEXT: s_waitcnt vmcnt(0)
+; GCN-NEXT: global_store_dword v0, v1, s[2:3]
+; GCN-NEXT: s_endpgm
entry:
store i32 0, ptr addrspace(1) %out, align 4
fence syncscope("workgroup") release
@@ -393,12 +537,6 @@ entry:
ret void
}
-; GCN-LABEL: {{^}}no_alias_volatile_store:
-; GCN: ds_write_b32
-; GCN: s_barrier
-; GCN: s_load_dword s
-; GCN-NOT: global_load_dword
-; GCN: global_store_dword
define protected amdgpu_kernel void @no_alias_volatile_store(ptr addrspace(1) %in, ptr addrspace(1) %out) {
; CHECK-LABEL: @no_alias_volatile_store(
; CHECK-NEXT: entry:
@@ -410,6 +548,18 @@ define protected amdgpu_kernel void @no_alias_volatile_store(ptr addrspace(1) %i
; CHECK-NEXT: store i32 [[LD]], ptr addrspace(1) [[OUT:%.*]], align 4
; CHECK-NEXT: ret void
;
+; GCN-LABEL: no_alias_volatile_store:
+; GCN: ; %bb.0: ; %entry
+; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GCN-NEXT: v_mov_b32_e32 v0, 0
+; GCN-NEXT: ds_write_b32 v0, v0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_barrier
+; GCN-NEXT: s_load_dword s0, s[0:1], 0x0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: v_mov_b32_e32 v1, s0
+; GCN-NEXT: global_store_dword v0, v1, s[2:3]
+; GCN-NEXT: s_endpgm
entry:
store volatile i32 0, ptr addrspace(3) @LDS, align 4
fence syncscope("workgroup") release
@@ -420,11 +570,6 @@ entry:
ret void
}
-; GCN-LABEL: {{^}}no_alias_atomic_rmw_relaxed:
-; GCN: ds_add_u32
-; GCN: s_load_dword s
-; GCN-NOT: global_load_dword
-; GCN: global_store_dword
define protected amdgpu_kernel void @no_alias_atomic_rmw_relaxed(ptr addrspace(1) %in, ptr addrspace(1) %out) {
; CHECK-LABEL: @no_alias_atomic_rmw_relaxed(
; CHECK-NEXT: entry:
@@ -433,6 +578,18 @@ define protected amdgpu_kernel void @no_alias_atomic_rmw_relaxed(ptr addrspace(1
; CHECK-NEXT: store i32 [[LD]], ptr addrspace(1) [[OUT:%.*]], align 4
; CHECK-NEXT: ret void
;
+; GCN-LABEL: no_alias_atomic_rmw_relaxed:
+; GCN: ; %bb.0: ; %entry
+; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GCN-NEXT: v_mov_b32_e32 v0, 5
+; GCN-NEXT: v_mov_b32_e32 v1, 0
+; GCN-NEXT: ds_add_u32 v1, v0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_load_dword s0, s[0:1], 0x0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: v_mov_b32_e32 v0, s0
+; GCN-NEXT: global_store_dword v1, v0, s[2:3]
+; GCN-NEXT: s_endpgm
entry:
%unused = atomicrmw add ptr addrspace(3) @LDS, i32 5 monotonic
%ld = load i32, ptr addrspace(1) %in, align 4
@@ -440,11 +597,6 @@ entry:
ret void
}
-; GCN-LABEL: {{^}}no_alias_atomic_cmpxchg:
-; GCN: ds_cmpst_b32
-; GCN: s_load_dword s
-; GCN-NOT: global_load_dword
-; GCN: global_store_dword
define protected amdgpu_kernel void @no_alias_atomic_cmpxchg(ptr addrspace(1) %in, ptr addrspace(1) %out, i32 %swap) {
; CHECK-LABEL: @no_alias_atomic_cmpxchg(
; CHECK-NEXT: entry:
@@ -456,6 +608,22 @@ define protected amdgpu_kernel void @no_alias_atomic_cmpxchg(ptr addrspace(1) %i
; CHECK-NEXT: store i32 [[LD]], ptr addrspace(1) [[OUT:%.*]], align 4
; CHECK-NEXT: ret void
;
+; GCN-LABEL: no_alias_atomic_cmpxchg:
+; GCN: ; %bb.0: ; %entry
+; GCN-NEXT: s_load_dword s6, s[4:5], 0x34
+; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GCN-NEXT: v_mov_b32_e32 v0, 7
+; GCN-NEXT: v_mov_b32_e32 v1, 0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: v_mov_b32_e32 v2, s6
+; GCN-NEXT: ds_cmpst_b32 v1, v0, v2
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_barrier
+; GCN-NEXT: s_load_dword s0, s[0:1], 0x0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: v_mov_b32_e32 v0, s0
+; GCN-NEXT: global_store_dword v1, v0, s[2:3]
+; GCN-NEXT: s_endpgm
entry:
%unused = cmpxchg ptr addrspace(3) @LDS, i32 7, i32 %swap seq_cst monotonic
fence syncscope("workgroup") release
@@ -466,11 +634,6 @@ entry:
ret void
}
-; GCN-LABEL: {{^}}no_alias_atomic_rmw:
-; GCN: ds_add_u32
-; GCN: s_load_dword s
-; GCN-NOT: global_load_dword
-; GCN: global_store_dword
define protected amdgpu_kernel void @no_alias_atomic_rmw(ptr addrspace(1) %in, ptr addrspace(1) %out) {
; CHECK-LABEL: @no_alias_atomic_rmw(
; CHECK-NEXT: entry:
@@ -482,6 +645,20 @@ define protected amdgpu_kernel void @no_alias_atomic_rmw(ptr addrspace(1) %in, p
; CHECK-NEXT: store i32 [[LD]], ptr addrspace(1) [[OUT:%.*]], align 4
; CHECK-NEXT: ret void
;
+; GCN-LABEL: no_alias_atomic_rmw:
+; GCN: ; %bb.0: ; %entry
+; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GCN-NEXT: v_mov_b32_e32 v0, 5
+; GCN-NEXT: v_mov_b32_e32 v1, 0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: ds_add_u32 v1, v0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: s_barrier
+; GCN-NEXT: s_load_dword s0, s[0:1], 0x0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: v_mov_b32_e32 v0, s0
+; GCN-NEXT: global_store_dword v1, v0, s[2:3]
+; GCN-NEXT: s_endpgm
entry:
%unused = atomicrmw add ptr addrspace(3) @LDS, i32 5 seq_cst
fence syncscope("workgroup") release
@@ -492,10 +669,6 @@ entry:
ret void
}
-; GCN-LABEL: {{^}}may_alias_atomic_cmpxchg:
-; GCN: global_atomic_cmpswap
-; GCN: global_load_dword
-; GCN: global_store_dword
define protected amdgpu_kernel void @may_alias_atomic_cmpxchg(ptr addrspace(1) %in, ptr addrspace(1) %out, i32 %swap) {
; CHECK-LABEL: @may_alias_atomic_cmpxchg(
; CHECK-NEXT: entry:
@@ -507,6 +680,22 @@ define protected amdgpu_kernel void @may_alias_atomic_cmpxchg(ptr addrspace(1) %
; CHECK-NEXT: store i32 [[LD]], ptr addrspace(1) [[OUT]], align 4
; CHECK-NEXT: ret void
;
+; GCN-LABEL: may_alias_atomic_cmpxchg:
+; GCN: ; %bb.0: ; %entry
+; GCN-NEXT: s_load_dword s6, s[4:5], 0x34
+; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GCN-NEXT: v_mov_b32_e32 v2, 0
+; GCN-NEXT: v_mov_b32_e32 v1, 7
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: v_mov_b32_e32 v0, s6
+; GCN-NEXT: global_atomic_cmpswap v2, v[0:1], s[2:3]
+; GCN-NEXT: s_waitcnt vmcnt(0)
+; GCN-NEXT: buffer_wbinvl1_vol
+; GCN-NEXT: s_barrier
+; GCN-NEXT: global_load_dword v0, v2, s[0:1]
+; GCN-NEXT: s_waitcnt vmcnt(0)
+; GCN-NEXT: global_store_dword v2, v0, s[2:3]
+; GCN-NEXT: s_endpgm
entry:
%unused = cmpxchg ptr addrspace(1) %out, i32 7, i32 %swap seq_cst monotonic
fence syncscope("workgroup") release
@@ -517,10 +706,6 @@ entry:
ret void
}
-; GCN-LABEL: {{^}}may_alias_atomic_rmw:
-; GCN: global_atomic_add
-; GCN: global_load_dword
-; GCN: global_store_dword
define protected amdgpu_kernel void @may_alias_atomic_rmw(ptr addrspace(1) %in, ptr addrspace(1) %out) {
; CHECK-LABEL: @may_alias_atomic_rmw(
; CHECK-NEXT: entry:
@@ -532,6 +717,20 @@ define protected amdgpu_kernel void @may_alias_atomic_rmw(ptr addrspace(1) %in,
; CHECK-NEXT: store i32 [[LD]], ptr addrspace(1) [[OUT]], align 4
; CHECK-NEXT: ret void
;
+; GCN-LABEL: may_alias_atomic_rmw:
+; GCN: ; %bb.0: ; %entry
+; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GCN-NEXT: v_mov_b32_e32 v0, 0
+; GCN-NEXT: v_mov_b32_e32 v1, 5
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: global_atomic_add v0, v1, s[2:3]
+; GCN-NEXT: s_waitcnt vmcnt(0)
+; GCN-NEXT: buffer_wbinvl1_vol
+; GCN-NEXT: s_barrier
+; GCN-NEXT: global_load_dword v1, v0, s[0:1]
+; GCN-NEXT: s_waitcnt vmcnt(0)
+; GCN-NEXT: global_store_dword v0, v1, s[2:3]
+; GCN-NEXT: s_endpgm
entry:
%unused = atomicrmw add ptr addrspace(1) %out, i32 5 syncscope("agent") seq_cst
fence syncscope("workgroup") release
@@ -542,12 +741,6 @@ entry:
ret void
}
-; GCN-LABEL: {{^}}no_alias_atomic_rmw_then_clobber:
-; GCN: global_store_dword
-; GCN: global_store_dword
-; GCN: ds_add_u32
-; GCN: global_load_dword
-; GCN: global_store_dword
define protected amdgpu_kernel void @no_alias_atomic_rmw_then_clobber(ptr addrspace(1) %in, ptr addrspace(1) %out, ptr addrspace(1) noalias %noalias) {
; CHECK-LABEL: @no_alias_atomic_rmw_then_clobber(
; CHECK-NEXT: entry:
@@ -561,6 +754,24 @@ define protected amdgpu_kernel void @no_alias_atomic_rmw_then_clobber(ptr addrsp
; CHECK-NEXT: store i32 [[LD]], ptr addrspace(1) [[OUT]], align 4
; CHECK-NEXT: ret void
;
+; GCN-LABEL: no_alias_atomic_rmw_then_clobber:
+; GCN: ; %bb.0: ; %entry
+; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GCN-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GCN-NEXT: v_mov_b32_e32 v0, 1
+; GCN-NEXT: v_mov_b32_e32 v1, 0
+; GCN-NEXT: v_mov_b32_e32 v2, 2
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: global_store_dword v1, v0, s[2:3]
+; GCN-NEXT: global_store_dword v1, v2, s[6:7]
+; GCN-NEXT: v_mov_b32_e32 v0, 5
+; GCN-NEXT: ds_add_u32 v1, v0
+; GCN-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
+; GCN-NEXT: s_barrier
+; GCN-NEXT: global_load_dword v0, v1, s[0:1]
+; GCN-NEXT: s_waitcnt vmcnt(0)
+; GCN-NEXT: global_store_dword v1, v0, s[2:3]
+; GCN-NEXT: s_endpgm
entry:
store i32 1, ptr addrspace(1) %out, align 4
store i32 2, ptr addrspace(1) %noalias, align 4
@@ -573,12 +784,6 @@ entry:
ret void
}
-; GCN-LABEL: {{^}}no_alias_atomic_rmw_then_no_alias_store:
-; GCN: global_store_dword
-; GCN: ds_add_u32
-; GCN: s_load_dword s
-; GCN-NOT: global_load_dword
-; GCN: global_store_dword
define protected amdgpu_kernel void @no_alias_atomic_rmw_then_no_alias_store(ptr addrspace(1) %in, ptr addrspace(1) %out, ptr addrspace(1) noalias %noalias) {
; CHECK-LABEL: @no_alias_atomic_rmw_then_no_alias_store(
; CHECK-NEXT: entry:
@@ -591,6 +796,23 @@ define protected amdgpu_kernel void @no_alias_atomic_rmw_then_no_alias_store(ptr
; CHECK-NEXT: store i32 [[LD]], ptr addrspace(1) [[OUT:%.*]], align 4
; CHECK-NEXT: ret void
;
+; GCN-LABEL: no_alias_atomic_rmw_then_no_alias_store:
+; GCN: ; %bb.0: ; %entry
+; GCN-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GCN-NEXT: v_mov_b32_e32 v0, 2
+; GCN-NEXT: v_mov_b32_e32 v1, 0
+; GCN-NEXT: v_mov_b32_e32 v2, 5
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: global_store_dword v1, v0, s[6:7]
+; GCN-NEXT: ds_add_u32 v1, v2
+; GCN-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0)
+; GCN-NEXT: s_barrier
+; GCN-NEXT: s_load_dword s0, s[0:1], 0x0
+; GCN-NEXT: s_waitcnt lgkmcnt(0)
+; GCN-NEXT: v_mov_b32_e32 v0, s0
+; GCN-NEXT: global_store_dword v1, v0, s[2:3]
+; GCN-NEXT: s_endpgm
entry:
store i32 2, ptr addrspace(1) %noalias, align 4
%unused = atomicrmw add ptr addrspace(3) @LDS, i32 5 seq_cst
More information about the llvm-commits
mailing list