3 files changed, 89 insertions, 14 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/merge-stores.ll b/llvm/test/CodeGen/AMDGPU/merge-stores.ll
index 87148ae9f69..62d372a398a 100644
--- a/llvm/test/CodeGen/AMDGPU/merge-stores.ll
+++ b/llvm/test/CodeGen/AMDGPU/merge-stores.ll
@@ -121,10 +121,7 @@ define void @merge_global_store_4_constants_i32(i32 addrspace(1)* %out) #0 {
 }
 
 ; GCN-LABEL: {{^}}merge_global_store_4_constants_f32_order:
-; XGCN: buffer_store_dwordx4
-; GCN: buffer_store_dword v
-; GCN: buffer_store_dword v
-; GCN: buffer_store_dwordx2 v
+; GCN: buffer_store_dwordx4
 define void @merge_global_store_4_constants_f32_order(float addrspace(1)* %out) #0 {
   %out.gep.1 = getelementptr float, float addrspace(1)* %out, i32 1
   %out.gep.2 = getelementptr float, float addrspace(1)* %out, i32 2
@@ -137,17 +134,9 @@ define void @merge_global_store_4_constants_f32_order(float addrspace(1)* %out)
   ret void
 }
 
-; First store is out of order. Because of order of combines, the
-; consecutive store fails because only some of the stores have been
-; replaced with integer constant stores, and then won't merge because
-; the types are different.
-
+; First store is out of order.
 ; GCN-LABEL: {{^}}merge_global_store_4_constants_f32:
-; XGCN: buffer_store_dwordx4
-; GCN: buffer_store_dword v
-; GCN: buffer_store_dword v
-; GCN: buffer_store_dword v
-; GCN: buffer_store_dword v
+; GCN: buffer_store_dwordx4
 define void @merge_global_store_4_constants_f32(float addrspace(1)* %out) #0 {
   %out.gep.1 = getelementptr float, float addrspace(1)* %out, i32 1
   %out.gep.2 = getelementptr float, float addrspace(1)* %out, i32 2
@@ -160,6 +149,29 @@ define void @merge_global_store_4_constants_f32(float addrspace(1)* %out) #0 {
   ret void
 }
 
+; FIXME: Should be able to merge this
+; GCN-LABEL: {{^}}merge_global_store_4_constants_mixed_i32_f32:
+; XGCN: buffer_store_dwordx4
+; GCN: buffer_store_dword
+; GCN: buffer_store_dword
+; GCN: buffer_store_dword
+; GCN: buffer_store_dword
+; GCN: s_endpgm
+define void @merge_global_store_4_constants_mixed_i32_f32(float addrspace(1)* %out) #0 {
+  %out.gep.1 = getelementptr float, float addrspace(1)* %out, i32 1
+  %out.gep.2 = getelementptr float, float addrspace(1)* %out, i32 2
+  %out.gep.3 = getelementptr float, float addrspace(1)* %out, i32 3
+
+  %out.gep.1.bc = bitcast float addrspace(1)* %out.gep.1 to i32 addrspace(1)*
+  %out.gep.3.bc = bitcast float addrspace(1)* %out.gep.3 to i32 addrspace(1)*
+
+  store i32 11, i32 addrspace(1)* %out.gep.1.bc
+  store float 2.0, float addrspace(1)* %out.gep.2
+  store i32 17, i32 addrspace(1)* %out.gep.3.bc
+  store float 8.0, float addrspace(1)* %out
+  ret void
+}
+
 ; GCN-LABEL: {{^}}merge_global_store_3_constants_i32:
 ; SI-DAG: buffer_store_dwordx2
 ; SI-DAG: buffer_store_dword
diff --git a/llvm/test/CodeGen/PowerPC/vector-merge-store-fp-constants.ll b/llvm/test/CodeGen/PowerPC/vector-merge-store-fp-constants.ll
new file mode 100644
index 00000000000..db92f20c352
--- /dev/null
+++ b/llvm/test/CodeGen/PowerPC/vector-merge-store-fp-constants.ll
@@ -0,0 +1,28 @@
+; RUN: llc -march=ppc64 -mtriple=ppc64-apple-darwin < %s | FileCheck %s
+
+; CHECK-LABEL: {{^}}_merge_8_float_zero_stores:
+; CHECK: li [[ZEROREG:r[0-9]+]], 0
+; CHECK-DAG: std [[ZEROREG]], 0([[PTR:r[0-9]+]])
+; CHECK-DAG: std [[ZEROREG]], 8([[PTR]])
+; CHECK-DAG: std [[ZEROREG]], 16([[PTR]])
+; CHECK-DAG: std [[ZEROREG]], 24([[PTR]])
+; CHECK: blr
+define void @merge_8_float_zero_stores(float* %ptr) {
+  %idx0 = getelementptr float, float* %ptr, i64 0
+  %idx1 = getelementptr float, float* %ptr, i64 1
+  %idx2 = getelementptr float, float* %ptr, i64 2
+  %idx3 = getelementptr float, float* %ptr, i64 3
+  %idx4 = getelementptr float, float* %ptr, i64 4
+  %idx5 = getelementptr float, float* %ptr, i64 5
+  %idx6 = getelementptr float, float* %ptr, i64 6
+  %idx7 = getelementptr float, float* %ptr, i64 7
+  store float 0.0, float* %idx0, align 4
+  store float 0.0, float* %idx1, align 4
+  store float 0.0, float* %idx2, align 4
+  store float 0.0, float* %idx3, align 4
+  store float 0.0, float* %idx4, align 4
+  store float 0.0, float* %idx5, align 4
+  store float 0.0, float* %idx6, align 4
+  store float 0.0, float* %idx7, align 4
+  ret void
+}
diff --git a/llvm/test/CodeGen/X86/vector-merge-store-fp-constants.ll b/llvm/test/CodeGen/X86/vector-merge-store-fp-constants.ll
new file mode 100644
index 00000000000..a6fb32d48a7
--- /dev/null
+++ b/llvm/test/CodeGen/X86/vector-merge-store-fp-constants.ll
@@ -0,0 +1,35 @@
+; RUN: llc -march=x86-64 -mtriple=x86_64-unknown-unknown < %s | FileCheck -check-prefix=DEFAULTCPU -check-prefix=ALL %s
+; RUN: llc -march=x86-64 -mcpu=x86-64 -mtriple=x86_64-unknown-unknown < %s | FileCheck -check-prefix=X8664CPU -check-prefix=ALL %s
+
+
+; ALL-LABEL: {{^}}merge_8_float_zero_stores:
+
+; DEFAULTCPU-DAG: movq $0, ([[PTR:%[a-z]+]])
+; DEFAULTCPU-DAG: movq $0, 8([[PTR]])
+; DEFAULTCPU-DAG: movq $0, 16([[PTR]])
+; DEFAULTCPU-DAG: movq $0, 24([[PTR]])
+
+; X8664CPU: xorps [[ZEROREG:%xmm[0-9]+]], [[ZEROREG]]
+; X8664CPU-DAG: movups [[ZEROREG]], ([[PTR:%[a-z]+]])
+; X8664CPU-DAG: movups [[ZEROREG]], 16([[PTR:%[a-z]+]])
+
+; ALL: retq
+define void @merge_8_float_zero_stores(float* %ptr) {
+  %idx0 = getelementptr float, float* %ptr, i64 0
+  %idx1 = getelementptr float, float* %ptr, i64 1
+  %idx2 = getelementptr float, float* %ptr, i64 2
+  %idx3 = getelementptr float, float* %ptr, i64 3
+  %idx4 = getelementptr float, float* %ptr, i64 4
+  %idx5 = getelementptr float, float* %ptr, i64 5
+  %idx6 = getelementptr float, float* %ptr, i64 6
+  %idx7 = getelementptr float, float* %ptr, i64 7
+  store float 0.0, float* %idx0, align 4
+  store float 0.0, float* %idx1, align 4
+  store float 0.0, float* %idx2, align 4
+  store float 0.0, float* %idx3, align 4
+  store float 0.0, float* %idx4, align 4
+  store float 0.0, float* %idx5, align 4
+  store float 0.0, float* %idx6, align 4
+  store float 0.0, float* %idx7, align 4
+  ret void
+}