[AMDGPU] Switch scalarize global loads ON by default
Differential revision: https://reviews.llvm.org/D34407
llvm-svn: 307026
diff --git a/llvm/test/CodeGen/AMDGPU/add.ll b/llvm/test/CodeGen/AMDGPU/add.ll
index 7e4546d..6dcd7c2 100644
--- a/llvm/test/CodeGen/AMDGPU/add.ll
+++ b/llvm/test/CodeGen/AMDGPU/add.ll
@@ -5,9 +5,9 @@
;FUNC-LABEL: {{^}}test1:
;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
-;SI: v_add_i32_e32 [[REG:v[0-9]+]], vcc, {{v[0-9]+, v[0-9]+}}
-;SI-NOT: [[REG]]
-;SI: buffer_store_dword [[REG]],
+;SI: s_add_i32 s[[REG:[0-9]+]], {{s[0-9]+, s[0-9]+}}
+;SI: v_mov_b32_e32 v[[REG]], s[[REG]]
+;SI: buffer_store_dword v[[REG]],
define amdgpu_kernel void @test1(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
%b_ptr = getelementptr i32, i32 addrspace(1)* %in, i32 1
%a = load i32, i32 addrspace(1)* %in
@@ -21,8 +21,8 @@
;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
-;SI: v_add_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
-;SI: v_add_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
+;SI: s_add_i32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
+;SI: s_add_i32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
define amdgpu_kernel void @test2(<2 x i32> addrspace(1)* %out, <2 x i32> addrspace(1)* %in) {
%b_ptr = getelementptr <2 x i32>, <2 x i32> addrspace(1)* %in, i32 1
@@ -39,10 +39,10 @@
;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
;EG: ADD_INT {{[* ]*}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
-;SI: v_add_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
-;SI: v_add_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
-;SI: v_add_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
-;SI: v_add_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
+;SI: s_add_i32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
+;SI: s_add_i32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
+;SI: s_add_i32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
+;SI: s_add_i32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
define amdgpu_kernel void @test4(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)* %in) {
%b_ptr = getelementptr <4 x i32>, <4 x i32> addrspace(1)* %in, i32 1
diff --git a/llvm/test/CodeGen/AMDGPU/add_i128.ll b/llvm/test/CodeGen/AMDGPU/add_i128.ll
index 00a125c..d33965d 100644
--- a/llvm/test/CodeGen/AMDGPU/add_i128.ll
+++ b/llvm/test/CodeGen/AMDGPU/add_i128.ll
@@ -19,10 +19,10 @@
; Check that the SGPR add operand is correctly moved to a VGPR.
; GCN-LABEL: {{^}}sgpr_operand:
-; GCN: v_add_i32
-; GCN: v_addc_u32
-; GCN: v_addc_u32
-; GCN: v_addc_u32
+; GCN: s_add_u32
+; GCN: s_addc_u32
+; GCN: s_addc_u32
+; GCN: s_addc_u32
define amdgpu_kernel void @sgpr_operand(i128 addrspace(1)* noalias %out, i128 addrspace(1)* noalias %in, i128 %a) {
%foo = load i128, i128 addrspace(1)* %in, align 8
%result = add i128 %foo, %a
@@ -31,10 +31,10 @@
}
; GCN-LABEL: {{^}}sgpr_operand_reversed:
-; GCN: v_add_i32
-; GCN: v_addc_u32
-; GCN: v_addc_u32
-; GCN: v_addc_u32
+; GCN: s_add_u32
+; GCN: s_addc_u32
+; GCN: s_addc_u32
+; GCN: s_addc_u32
define amdgpu_kernel void @sgpr_operand_reversed(i128 addrspace(1)* noalias %out, i128 addrspace(1)* noalias %in, i128 %a) {
%foo = load i128, i128 addrspace(1)* %in, align 8
%result = add i128 %a, %foo
diff --git a/llvm/test/CodeGen/AMDGPU/add_i64.ll b/llvm/test/CodeGen/AMDGPU/add_i64.ll
index 62733d5..f673d91 100644
--- a/llvm/test/CodeGen/AMDGPU/add_i64.ll
+++ b/llvm/test/CodeGen/AMDGPU/add_i64.ll
@@ -19,8 +19,8 @@
; Check that the SGPR add operand is correctly moved to a VGPR.
; SI-LABEL: {{^}}sgpr_operand:
-; SI: v_add_i32
-; SI: v_addc_u32
+; SI: s_add_u32
+; SI: s_addc_u32
define amdgpu_kernel void @sgpr_operand(i64 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %in, i64 addrspace(1)* noalias %in_bar, i64 %a) {
%foo = load i64, i64 addrspace(1)* %in, align 8
%result = add i64 %foo, %a
@@ -32,8 +32,8 @@
; SGPR as other operand.
;
; SI-LABEL: {{^}}sgpr_operand_reversed:
-; SI: v_add_i32
-; SI: v_addc_u32
+; SI: s_add_u32
+; SI: s_addc_u32
define amdgpu_kernel void @sgpr_operand_reversed(i64 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %in, i64 %a) {
%foo = load i64, i64 addrspace(1)* %in, align 8
%result = add i64 %a, %foo
diff --git a/llvm/test/CodeGen/AMDGPU/and-gcn.ll b/llvm/test/CodeGen/AMDGPU/and-gcn.ll
index 2aec03a..ef11ae8 100644
--- a/llvm/test/CodeGen/AMDGPU/and-gcn.ll
+++ b/llvm/test/CodeGen/AMDGPU/and-gcn.ll
@@ -2,8 +2,7 @@
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}v_and_i64_br:
-; SI: v_and_b32
-; SI: v_and_b32
+; SI: s_and_b64
define amdgpu_kernel void @v_and_i64_br(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr, i64 addrspace(1)* %bptr) {
entry:
%tid = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0) #0
diff --git a/llvm/test/CodeGen/AMDGPU/and.ll b/llvm/test/CodeGen/AMDGPU/and.ll
index c356f8b..ee01901 100644
--- a/llvm/test/CodeGen/AMDGPU/and.ll
+++ b/llvm/test/CodeGen/AMDGPU/and.ll
@@ -8,8 +8,8 @@
; EG: AND_INT {{\*? *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; EG: AND_INT {{\*? *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
-; SI: v_and_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
-; SI: v_and_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
+; SI: s_and_b32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
+; SI: s_and_b32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
define amdgpu_kernel void @test2(<2 x i32> addrspace(1)* %out, <2 x i32> addrspace(1)* %in) {
%b_ptr = getelementptr <2 x i32>, <2 x i32> addrspace(1)* %in, i32 1
@@ -26,10 +26,11 @@
; EG: AND_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
; EG: AND_INT {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
-; SI: v_and_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
-; SI: v_and_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
-; SI: v_and_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
-; SI: v_and_b32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
+
+; SI: s_and_b32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
+; SI: s_and_b32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
+; SI: s_and_b32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
+; SI: s_and_b32 s{{[0-9]+, s[0-9]+, s[0-9]+}}
define amdgpu_kernel void @test4(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)* %in) {
%b_ptr = getelementptr <4 x i32>, <4 x i32> addrspace(1)* %in, i32 1
@@ -136,7 +137,9 @@
; FUNC-LABEL: {{^}}v_and_constant_i32
; SI: v_and_b32_e32 v{{[0-9]+}}, 0x12d687, v{{[0-9]+}}
define amdgpu_kernel void @v_and_constant_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %aptr) {
- %a = load i32, i32 addrspace(1)* %aptr, align 4
+ %tid = call i32 @llvm.r600.read.tidig.x() #0
+ %gep = getelementptr i32, i32 addrspace(1)* %aptr, i32 %tid
+ %a = load i32, i32 addrspace(1)* %gep, align 4
%and = and i32 %a, 1234567
store i32 %and, i32 addrspace(1)* %out, align 4
ret void
@@ -145,7 +148,9 @@
; FUNC-LABEL: {{^}}v_and_inline_imm_64_i32
; SI: v_and_b32_e32 v{{[0-9]+}}, 64, v{{[0-9]+}}
define amdgpu_kernel void @v_and_inline_imm_64_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %aptr) {
- %a = load i32, i32 addrspace(1)* %aptr, align 4
+ %tid = call i32 @llvm.r600.read.tidig.x() #0
+ %gep = getelementptr i32, i32 addrspace(1)* %aptr, i32 %tid
+ %a = load i32, i32 addrspace(1)* %gep, align 4
%and = and i32 %a, 64
store i32 %and, i32 addrspace(1)* %out, align 4
ret void
@@ -154,7 +159,9 @@
; FUNC-LABEL: {{^}}v_and_inline_imm_neg_16_i32
; SI: v_and_b32_e32 v{{[0-9]+}}, -16, v{{[0-9]+}}
define amdgpu_kernel void @v_and_inline_imm_neg_16_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %aptr) {
- %a = load i32, i32 addrspace(1)* %aptr, align 4
+ %tid = call i32 @llvm.r600.read.tidig.x() #0
+ %gep = getelementptr i32, i32 addrspace(1)* %aptr, i32 %tid
+ %a = load i32, i32 addrspace(1)* %gep, align 4
%and = and i32 %a, -16
store i32 %and, i32 addrspace(1)* %out, align 4
ret void
@@ -239,8 +246,11 @@
; SI: v_and_b32
; SI: v_and_b32
define amdgpu_kernel void @v_and_i64(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr, i64 addrspace(1)* %bptr) {
- %a = load i64, i64 addrspace(1)* %aptr, align 8
- %b = load i64, i64 addrspace(1)* %bptr, align 8
+ %tid = call i32 @llvm.r600.read.tidig.x() #0
+ %gep.a = getelementptr i64, i64 addrspace(1)* %aptr, i32 %tid
+ %a = load i64, i64 addrspace(1)* %gep.a, align 8
+ %gep.b = getelementptr i64, i64 addrspace(1)* %bptr, i32 %tid
+ %b = load i64, i64 addrspace(1)* %gep.b, align 8
%and = and i64 %a, %b
store i64 %and, i64 addrspace(1)* %out, align 8
ret void
@@ -251,7 +261,9 @@
; SI-DAG: v_and_b32_e32 {{v[0-9]+}}, 0x11e, {{v[0-9]+}}
; SI: buffer_store_dwordx2
define amdgpu_kernel void @v_and_constant_i64(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr) {
- %a = load i64, i64 addrspace(1)* %aptr, align 8
+ %tid = call i32 @llvm.r600.read.tidig.x() #0
+ %gep.a = getelementptr i64, i64 addrspace(1)* %aptr, i32 %tid
+ %a = load i64, i64 addrspace(1)* %gep.a, align 8
%and = and i64 %a, 1231231234567
store i64 %and, i64 addrspace(1)* %out, align 8
ret void
@@ -299,26 +311,30 @@
}
; FUNC-LABEL: {{^}}v_and_i64_32_bit_constant:
-; SI: buffer_load_dword [[VAL:v[0-9]+]]
+; SI: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]]
; SI-NOT: and
; SI: v_and_b32_e32 {{v[0-9]+}}, 0x12d687, [[VAL]]
; SI-NOT: and
; SI: buffer_store_dwordx2
define amdgpu_kernel void @v_and_i64_32_bit_constant(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr) {
- %a = load i64, i64 addrspace(1)* %aptr, align 8
+ %tid = call i32 @llvm.r600.read.tidig.x() #0
+ %gep.a = getelementptr i64, i64 addrspace(1)* %aptr, i32 %tid
+ %a = load i64, i64 addrspace(1)* %gep.a, align 8
%and = and i64 %a, 1234567
store i64 %and, i64 addrspace(1)* %out, align 8
ret void
}
; FUNC-LABEL: {{^}}v_and_inline_imm_i64:
-; SI: buffer_load_dword v{{[0-9]+}}
+; SI: {{buffer|flat}}_load_dword v{{[0-9]+}}
; SI-NOT: and
; SI: v_and_b32_e32 {{v[0-9]+}}, 64, {{v[0-9]+}}
; SI-NOT: and
; SI: buffer_store_dwordx2
define amdgpu_kernel void @v_and_inline_imm_i64(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr) {
- %a = load i64, i64 addrspace(1)* %aptr, align 8
+ %tid = call i32 @llvm.r600.read.tidig.x() #0
+ %gep.a = getelementptr i64, i64 addrspace(1)* %aptr, i32 %tid
+ %a = load i64, i64 addrspace(1)* %gep.a, align 8
%and = and i64 %a, 64
store i64 %and, i64 addrspace(1)* %out, align 8
ret void
@@ -326,13 +342,15 @@
; FIXME: Should be able to reduce load width
; FUNC-LABEL: {{^}}v_and_inline_neg_imm_i64:
-; SI: buffer_load_dwordx2 v{{\[}}[[VAL_LO:[0-9]+]]:[[VAL_HI:[0-9]+]]{{\]}}
+; SI: {{buffer|flat}}_load_dwordx2 v{{\[}}[[VAL_LO:[0-9]+]]:[[VAL_HI:[0-9]+]]{{\]}}
; SI-NOT: and
; SI: v_and_b32_e32 v[[VAL_LO]], -8, v[[VAL_LO]]
; SI-NOT: and
; SI: buffer_store_dwordx2 v{{\[}}[[VAL_LO]]:[[VAL_HI]]{{\]}}
define amdgpu_kernel void @v_and_inline_neg_imm_i64(i64 addrspace(1)* %out, i64 addrspace(1)* %aptr) {
- %a = load i64, i64 addrspace(1)* %aptr, align 8
+ %tid = call i32 @llvm.r600.read.tidig.x() #0
+ %gep.a = getelementptr i64, i64 addrspace(1)* %aptr, i32 %tid
+ %a = load i64, i64 addrspace(1)* %gep.a, align 8
%and = and i64 %a, -8
store i64 %and, i64 addrspace(1)* %out, align 8
ret void
@@ -549,5 +567,4 @@
store i64 %and, i64 addrspace(1)* %out, align 8
ret void
}
-
attributes #0 = { nounwind readnone }
diff --git a/llvm/test/CodeGen/AMDGPU/any_extend_vector_inreg.ll b/llvm/test/CodeGen/AMDGPU/any_extend_vector_inreg.ll
index c61c232..cdc60ab 100644
--- a/llvm/test/CodeGen/AMDGPU/any_extend_vector_inreg.ll
+++ b/llvm/test/CodeGen/AMDGPU/any_extend_vector_inreg.ll
@@ -2,9 +2,9 @@
; RUN: llc -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; GCN-LABEL: {{^}}any_extend_vector_inreg_v16i8_to_v4i32:
-; GCN: {{buffer|flat}}_load_dwordx4
-; GCN-DAG: {{buffer|flat}}_load_dwordx4
-; GCN-DAG: {{buffer|flat}}_load_dword
+; GCN: s_load_dwordx4
+; GCN-DAG: s_load_dwordx4
+; GCN-DAG: s_load_dword
; GCN: {{buffer|flat}}_store_byte
; GCN: {{buffer|flat}}_store_byte
diff --git a/llvm/test/CodeGen/AMDGPU/bitreverse.ll b/llvm/test/CodeGen/AMDGPU/bitreverse.ll
index 539373f..f29bfb4 100644
--- a/llvm/test/CodeGen/AMDGPU/bitreverse.ll
+++ b/llvm/test/CodeGen/AMDGPU/bitreverse.ll
@@ -2,6 +2,8 @@
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=FUNC %s
+declare i32 @llvm.amdgcn.workitem.id.x() #1
+
declare i16 @llvm.bitreverse.i16(i16) #1
declare i32 @llvm.bitreverse.i32(i32) #1
declare i64 @llvm.bitreverse.i64(i64) #1
@@ -42,12 +44,14 @@
}
; FUNC-LABEL: {{^}}v_brev_i32:
-; SI: buffer_load_dword [[VAL:v[0-9]+]],
+; SI: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; SI: v_bfrev_b32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; SI: buffer_store_dword [[RESULT]],
; SI: s_endpgm
define amdgpu_kernel void @v_brev_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) #0 {
- %val = load i32, i32 addrspace(1)* %valptr
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
+ %val = load i32, i32 addrspace(1)* %gep
%brev = call i32 @llvm.bitreverse.i32(i32 %val) #1
store i32 %brev, i32 addrspace(1)* %out
ret void
@@ -66,7 +70,9 @@
; SI: v_bfrev_b32_e32
; SI: v_bfrev_b32_e32
define amdgpu_kernel void @v_brev_v2i32(<2 x i32> addrspace(1)* noalias %out, <2 x i32> addrspace(1)* noalias %valptr) #0 {
- %val = load <2 x i32>, <2 x i32> addrspace(1)* %valptr
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <2 x i32>, <2 x i32> addrspace(1)* %valptr, i32 %tid
+ %val = load <2 x i32>, <2 x i32> addrspace(1)* %gep
%brev = call <2 x i32> @llvm.bitreverse.v2i32(<2 x i32> %val) #1
store <2 x i32> %brev, <2 x i32> addrspace(1)* %out
ret void
@@ -82,7 +88,9 @@
; FUNC-LABEL: {{^}}v_brev_i64:
; SI-NOT: v_or_b32_e64 v{{[0-9]+}}, 0, 0
define amdgpu_kernel void @v_brev_i64(i64 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %valptr) #0 {
- %val = load i64, i64 addrspace(1)* %valptr
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr i64, i64 addrspace(1)* %valptr, i32 %tid
+ %val = load i64, i64 addrspace(1)* %gep
%brev = call i64 @llvm.bitreverse.i64(i64 %val) #1
store i64 %brev, i64 addrspace(1)* %out
ret void
@@ -97,7 +105,9 @@
; FUNC-LABEL: {{^}}v_brev_v2i64:
define amdgpu_kernel void @v_brev_v2i64(<2 x i64> addrspace(1)* noalias %out, <2 x i64> addrspace(1)* noalias %valptr) #0 {
- %val = load <2 x i64>, <2 x i64> addrspace(1)* %valptr
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <2 x i64> , <2 x i64> addrspace(1)* %valptr, i32 %tid
+ %val = load <2 x i64>, <2 x i64> addrspace(1)* %gep
%brev = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> %val) #1
store <2 x i64> %brev, <2 x i64> addrspace(1)* %out
ret void
diff --git a/llvm/test/CodeGen/AMDGPU/bswap.ll b/llvm/test/CodeGen/AMDGPU/bswap.ll
index d2dacd7..eb3fc2f 100644
--- a/llvm/test/CodeGen/AMDGPU/bswap.ll
+++ b/llvm/test/CodeGen/AMDGPU/bswap.ll
@@ -10,7 +10,7 @@
declare <4 x i64> @llvm.bswap.v4i64(<4 x i64>) nounwind readnone
; FUNC-LABEL: @test_bswap_i32
-; SI: buffer_load_dword [[VAL:v[0-9]+]]
+; SI: s_load_dword [[VAL:s[0-9]+]]
; SI-DAG: v_alignbit_b32 [[TMP0:v[0-9]+]], [[VAL]], [[VAL]], 8
; SI-DAG: v_alignbit_b32 [[TMP1:v[0-9]+]], [[VAL]], [[VAL]], 24
; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0xff00ff
diff --git a/llvm/test/CodeGen/AMDGPU/cgp-addressing-modes-flat.ll b/llvm/test/CodeGen/AMDGPU/cgp-addressing-modes-flat.ll
index 5dec3e3..c114332 100644
--- a/llvm/test/CodeGen/AMDGPU/cgp-addressing-modes-flat.ll
+++ b/llvm/test/CodeGen/AMDGPU/cgp-addressing-modes-flat.ll
@@ -1,9 +1,9 @@
; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=bonaire < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-CI -check-prefix=OPT-CIVI %s
; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-VI -check-prefix=OPT-CIVI %s
; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=gfx900 -mattr=-flat-for-global < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-GFX9 %s
-; RUN: llc -march=amdgcn -mcpu=bonaire -mattr=-promote-alloca < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=CIVI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -mattr=-promote-alloca < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=CIVI %s
-; RUN: llc -march=amdgcn -mcpu=gfx900 -mattr=-flat-for-global -mattr=-promote-alloca < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s
+; RUN: llc -march=amdgcn -amdgpu-scalarize-global-loads=false -mcpu=bonaire -mattr=-promote-alloca < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=CIVI %s
+; RUN: llc -march=amdgcn -amdgpu-scalarize-global-loads=false -mcpu=tonga -mattr=-flat-for-global -mattr=-promote-alloca < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=CIVI %s
+; RUN: llc -march=amdgcn -amdgpu-scalarize-global-loads=false -mcpu=gfx900 -mattr=-flat-for-global -mattr=-promote-alloca < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s
; OPT-LABEL: @test_no_sink_flat_small_offset_i32(
; OPT-CIVI: getelementptr i32, i32 addrspace(4)* %in
@@ -40,7 +40,7 @@
; OPT-LABEL: @test_sink_noop_addrspacecast_flat_to_global_i32(
; OPT: getelementptr i32, i32 addrspace(4)* %out,
-; OPT-CI-NOT: getelementptr
+; rOPT-CI-NOT: getelementptr
; OPT: br i1
; OPT-CI: addrspacecast
diff --git a/llvm/test/CodeGen/AMDGPU/cgp-addressing-modes.ll b/llvm/test/CodeGen/AMDGPU/cgp-addressing-modes.ll
index c1cf56e..c01d834 100644
--- a/llvm/test/CodeGen/AMDGPU/cgp-addressing-modes.ll
+++ b/llvm/test/CodeGen/AMDGPU/cgp-addressing-modes.ll
@@ -1,9 +1,9 @@
; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=tahiti < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-SI %s
; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=bonaire < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-CI %s
; RUN: opt -S -codegenprepare -mtriple=amdgcn-unknown-unknown -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=OPT -check-prefix=OPT-VI %s
-; RUN: llc -march=amdgcn -mcpu=tahiti -mattr=-promote-alloca -amdgpu-sroa=0 < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=bonaire -mattr=-promote-alloca -amdgpu-sroa=0 < %s | FileCheck -check-prefix=GCN -check-prefix=CI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -mattr=-promote-alloca -amdgpu-sroa=0 < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -march=amdgcn -mcpu=tahiti -amdgpu-scalarize-global-loads=false -mattr=-promote-alloca -amdgpu-sroa=0 < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -march=amdgcn -mcpu=bonaire -amdgpu-scalarize-global-loads=false -mattr=-promote-alloca -amdgpu-sroa=0 < %s | FileCheck -check-prefix=GCN -check-prefix=CI %s
+; RUN: llc -march=amdgcn -mcpu=tonga -amdgpu-scalarize-global-loads=false -mattr=-flat-for-global -mattr=-promote-alloca -amdgpu-sroa=0 < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
target datalayout = "e-p:32:32-p1:64:64-p2:64:64-p3:32:32-p4:64:64-p5:32:32-p24:64:64-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64"
diff --git a/llvm/test/CodeGen/AMDGPU/coalescer_remat.ll b/llvm/test/CodeGen/AMDGPU/coalescer_remat.ll
index 3e1b76a..14b798b 100644
--- a/llvm/test/CodeGen/AMDGPU/coalescer_remat.ll
+++ b/llvm/test/CodeGen/AMDGPU/coalescer_remat.ll
@@ -12,7 +12,7 @@
; CHECK: v_mov_b32_e32 v{{[0-9]+}}, 0
; CHECK: v_mov_b32_e32 v{{[0-9]+}}, 0
; It's probably OK if this is slightly higher:
-; CHECK: ; NumVgprs: 8
+; CHECK: ; NumVgprs: 4
define amdgpu_kernel void @foobar(<4 x float> addrspace(1)* %out, <4 x float> addrspace(1)* %in, i32 %flag) {
entry:
%cmpflag = icmp eq i32 %flag, 1
diff --git a/llvm/test/CodeGen/AMDGPU/copy-illegal-type.ll b/llvm/test/CodeGen/AMDGPU/copy-illegal-type.ll
index d772d1b..e39bd60 100644
--- a/llvm/test/CodeGen/AMDGPU/copy-illegal-type.ll
+++ b/llvm/test/CodeGen/AMDGPU/copy-illegal-type.ll
@@ -5,35 +5,41 @@
declare i32 @llvm.amdgcn.workitem.id.y() nounwind readnone
; FUNC-LABEL: {{^}}test_copy_v4i8:
-; GCN: buffer_load_dword [[REG:v[0-9]+]]
+; GCN: {{buffer|flat}}_load_dword [[REG:v[0-9]+]]
; GCN: buffer_store_dword [[REG]]
; GCN: s_endpgm
define amdgpu_kernel void @test_copy_v4i8(<4 x i8> addrspace(1)* %out, <4 x i8> addrspace(1)* %in) nounwind {
- %val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4
+ %tid.x = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %in, i32 %tid.x
+ %val = load <4 x i8>, <4 x i8> addrspace(1)* %gep, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: {{^}}test_copy_v4i8_x2:
-; GCN: buffer_load_dword [[REG:v[0-9]+]]
+; GCN: {{buffer|flat}}_load_dword [[REG:v[0-9]+]]
; GCN: buffer_store_dword [[REG]]
; GCN: buffer_store_dword [[REG]]
; GCN: s_endpgm
define amdgpu_kernel void @test_copy_v4i8_x2(<4 x i8> addrspace(1)* %out0, <4 x i8> addrspace(1)* %out1, <4 x i8> addrspace(1)* %in) nounwind {
- %val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4
+ %tid.x = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %in, i32 %tid.x
+ %val = load <4 x i8>, <4 x i8> addrspace(1)* %gep, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out1, align 4
ret void
}
; FUNC-LABEL: {{^}}test_copy_v4i8_x3:
-; GCN: buffer_load_dword [[REG:v[0-9]+]]
+; GCN: {{buffer|flat}}_load_dword [[REG:v[0-9]+]]
; GCN: buffer_store_dword [[REG]]
; GCN: buffer_store_dword [[REG]]
; GCN: buffer_store_dword [[REG]]
; GCN: s_endpgm
define amdgpu_kernel void @test_copy_v4i8_x3(<4 x i8> addrspace(1)* %out0, <4 x i8> addrspace(1)* %out1, <4 x i8> addrspace(1)* %out2, <4 x i8> addrspace(1)* %in) nounwind {
- %val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4
+ %tid.x = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %in, i32 %tid.x
+ %val = load <4 x i8>, <4 x i8> addrspace(1)* %gep, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out1, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out2, align 4
@@ -41,14 +47,16 @@
}
; FUNC-LABEL: {{^}}test_copy_v4i8_x4:
-; GCN: buffer_load_dword [[REG:v[0-9]+]]
+; GCN: {{buffer|flat}}_load_dword [[REG:v[0-9]+]]
; GCN: buffer_store_dword [[REG]]
; GCN: buffer_store_dword [[REG]]
; GCN: buffer_store_dword [[REG]]
; GCN: buffer_store_dword [[REG]]
; GCN: s_endpgm
define amdgpu_kernel void @test_copy_v4i8_x4(<4 x i8> addrspace(1)* %out0, <4 x i8> addrspace(1)* %out1, <4 x i8> addrspace(1)* %out2, <4 x i8> addrspace(1)* %out3, <4 x i8> addrspace(1)* %in) nounwind {
- %val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4
+ %tid.x = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %in, i32 %tid.x
+ %val = load <4 x i8>, <4 x i8> addrspace(1)* %gep, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out1, align 4
store <4 x i8> %val, <4 x i8> addrspace(1)* %out2, align 4
@@ -57,7 +65,7 @@
}
; FUNC-LABEL: {{^}}test_copy_v4i8_extra_use:
-; GCN: buffer_load_dword
+; GCN: {{buffer|flat}}_load_dword
; GCN-DAG: v_lshrrev_b32
; GCN: v_and_b32
; GCN: v_or_b32
@@ -66,7 +74,9 @@
; GCN: s_endpgm
define amdgpu_kernel void @test_copy_v4i8_extra_use(<4 x i8> addrspace(1)* %out0, <4 x i8> addrspace(1)* %out1, <4 x i8> addrspace(1)* %in) nounwind {
- %val = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4
+ %tid.x = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %in, i32 %tid.x
+ %val = load <4 x i8>, <4 x i8> addrspace(1)* %gep, align 4
%add = add <4 x i8> %val, <i8 9, i8 9, i8 9, i8 9>
store <4 x i8> %val, <4 x i8> addrspace(1)* %out0, align 4
store <4 x i8> %add, <4 x i8> addrspace(1)* %out1, align 4
@@ -97,19 +107,21 @@
}
; FUNC-LABEL: {{^}}test_copy_v3i8_align4:
-; GCN: buffer_load_dword
+; GCN: {{buffer|flat}}_load_dword
; GCN-DAG: buffer_store_short v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0{{$}}
; GCN-DAG: buffer_store_byte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0 offset:2{{$}}
; GCN: s_endpgm
define amdgpu_kernel void @test_copy_v3i8_align4(<3 x i8> addrspace(1)* %out, <3 x i8> addrspace(1)* %in) nounwind {
- %val = load <3 x i8>, <3 x i8> addrspace(1)* %in, align 4
+ %tid.x = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <3 x i8>, <3 x i8> addrspace(1)* %in, i32 %tid.x
+ %val = load <3 x i8>, <3 x i8> addrspace(1)* %gep, align 4
store <3 x i8> %val, <3 x i8> addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: {{^}}test_copy_v3i8_align2:
-; GCN-DAG: buffer_load_ushort v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0{{$}}
-; GCN-DAG: buffer_load_ubyte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0 offset:2{{$}}
+; GCN-DAG: {{buffer|flat}}_load_ushort v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0{{$}}
+; GCN-DAG: {{buffer|flat}}_load_ubyte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0 offset:2{{$}}
; GCN-DAG: buffer_store_short v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0{{$}}
; GCN-DAG: buffer_store_byte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0 offset:2{{$}}
; GCN: s_endpgm
@@ -120,9 +132,9 @@
}
; FUNC-LABEL: {{^}}test_copy_v3i8_align1:
-; GCN: buffer_load_ubyte
-; GCN: buffer_load_ubyte
-; GCN: buffer_load_ubyte
+; GCN: {{buffer|flat}}_load_ubyte
+; GCN: {{buffer|flat}}_load_ubyte
+; GCN: {{buffer|flat}}_load_ubyte
; GCN: buffer_store_byte
; GCN: buffer_store_byte
@@ -135,10 +147,10 @@
}
; FUNC-LABEL: {{^}}test_copy_v4i8_volatile_load:
-; GCN: buffer_load_ubyte
-; GCN: buffer_load_ubyte
-; GCN: buffer_load_ubyte
-; GCN: buffer_load_ubyte
+; GCN: {{buffer|flat}}_load_ubyte
+; GCN: {{buffer|flat}}_load_ubyte
+; GCN: {{buffer|flat}}_load_ubyte
+; GCN: {{buffer|flat}}_load_ubyte
; GCN: buffer_store_dword
; GCN: s_endpgm
define amdgpu_kernel void @test_copy_v4i8_volatile_load(<4 x i8> addrspace(1)* %out, <4 x i8> addrspace(1)* %in) nounwind {
@@ -148,10 +160,10 @@
}
; FUNC-LABEL: {{^}}test_copy_v4i8_volatile_store:
-; GCN: buffer_load_ubyte
-; GCN: buffer_load_ubyte
-; GCN: buffer_load_ubyte
-; GCN: buffer_load_ubyte
+; GCN: {{buffer|flat}}_load_ubyte
+; GCN: {{buffer|flat}}_load_ubyte
+; GCN: {{buffer|flat}}_load_ubyte
+; GCN: {{buffer|flat}}_load_ubyte
; GCN: buffer_store_byte
; GCN: buffer_store_byte
; GCN: buffer_store_byte
diff --git a/llvm/test/CodeGen/AMDGPU/ctlz.ll b/llvm/test/CodeGen/AMDGPU/ctlz.ll
index 149c506..3031ee8 100644
--- a/llvm/test/CodeGen/AMDGPU/ctlz.ll
+++ b/llvm/test/CodeGen/AMDGPU/ctlz.ll
@@ -34,7 +34,7 @@
}
; FUNC-LABEL: {{^}}v_ctlz_i32:
-; GCN: buffer_load_dword [[VAL:v[0-9]+]],
+; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN-DAG: v_ffbh_u32_e32 [[CTLZ:v[0-9]+]], [[VAL]]
; GCN-DAG: v_cmp_ne_u32_e32 vcc, 0, [[CTLZ]]
; GCN: v_cndmask_b32_e32 [[RESULT:v[0-9]+]], 32, [[CTLZ]], vcc
@@ -44,14 +44,16 @@
; EG: FFBH_UINT
; EG: CNDE_INT
define amdgpu_kernel void @v_ctlz_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
- %val = load i32, i32 addrspace(1)* %valptr, align 4
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 false) nounwind readnone
store i32 %ctlz, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: {{^}}v_ctlz_v2i32:
-; GCN: buffer_load_dwordx2
+; GCN: {{buffer|flat}}_load_dwordx2
; GCN: v_ffbh_u32_e32
; GCN: v_ffbh_u32_e32
; GCN: buffer_store_dwordx2
@@ -62,14 +64,16 @@
; EG: FFBH_UINT
; EG: CNDE_INT
define amdgpu_kernel void @v_ctlz_v2i32(<2 x i32> addrspace(1)* noalias %out, <2 x i32> addrspace(1)* noalias %valptr) nounwind {
- %val = load <2 x i32>, <2 x i32> addrspace(1)* %valptr, align 8
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr <2 x i32>, <2 x i32> addrspace(1)* %valptr, i32 %tid
+ %val = load <2 x i32>, <2 x i32> addrspace(1)* %in.gep, align 8
%ctlz = call <2 x i32> @llvm.ctlz.v2i32(<2 x i32> %val, i1 false) nounwind readnone
store <2 x i32> %ctlz, <2 x i32> addrspace(1)* %out, align 8
ret void
}
; FUNC-LABEL: {{^}}v_ctlz_v4i32:
-; GCN: buffer_load_dwordx4
+; GCN: {{buffer|flat}}_load_dwordx4
; GCN: v_ffbh_u32_e32
; GCN: v_ffbh_u32_e32
; GCN: v_ffbh_u32_e32
@@ -90,14 +94,16 @@
; EG-DAG: FFBH_UINT
; EG-DAG: CNDE_INT
define amdgpu_kernel void @v_ctlz_v4i32(<4 x i32> addrspace(1)* noalias %out, <4 x i32> addrspace(1)* noalias %valptr) nounwind {
- %val = load <4 x i32>, <4 x i32> addrspace(1)* %valptr, align 16
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr <4 x i32>, <4 x i32> addrspace(1)* %valptr, i32 %tid
+ %val = load <4 x i32>, <4 x i32> addrspace(1)* %in.gep, align 16
%ctlz = call <4 x i32> @llvm.ctlz.v4i32(<4 x i32> %val, i1 false) nounwind readnone
store <4 x i32> %ctlz, <4 x i32> addrspace(1)* %out, align 16
ret void
}
; FUNC-LABEL: {{^}}v_ctlz_i8:
-; GCN: buffer_load_ubyte [[VAL:v[0-9]+]],
+; GCN: {{buffer|flat}}_load_ubyte [[VAL:v[0-9]+]],
; SI-DAG: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; VI-DAG: v_ffbh_u32_sdwa [[RESULT:v[0-9]+]], [[VAL]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0
; GCN: buffer_store_byte [[RESULT]],
@@ -168,12 +174,14 @@
}
; FUNC-LABEL: {{^}}v_ctlz_i32_sel_eq_neg1:
-; GCN: buffer_load_dword [[VAL:v[0-9]+]],
+; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm
- define amdgpu_kernel void @v_ctlz_i32_sel_eq_neg1(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
- %val = load i32, i32 addrspace(1)* %valptr
+define amdgpu_kernel void @v_ctlz_i32_sel_eq_neg1(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 false) nounwind readnone
%cmp = icmp eq i32 %val, 0
%sel = select i1 %cmp, i32 -1, i32 %ctlz
@@ -182,12 +190,14 @@
}
; FUNC-LABEL: {{^}}v_ctlz_i32_sel_ne_neg1:
-; GCN: buffer_load_dword [[VAL:v[0-9]+]],
+; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm
define amdgpu_kernel void @v_ctlz_i32_sel_ne_neg1(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
- %val = load i32, i32 addrspace(1)* %valptr
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 false) nounwind readnone
%cmp = icmp ne i32 %val, 0
%sel = select i1 %cmp, i32 %ctlz, i32 -1
@@ -197,13 +207,15 @@
; TODO: Should be able to eliminate select here as well.
; FUNC-LABEL: {{^}}v_ctlz_i32_sel_eq_bitwidth:
-; GCN: buffer_load_dword
+; GCN: {{buffer|flat}}_load_dword
; GCN: v_ffbh_u32_e32
; GCN: v_cmp
; GCN: v_cndmask
; GCN: s_endpgm
define amdgpu_kernel void @v_ctlz_i32_sel_eq_bitwidth(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
- %val = load i32, i32 addrspace(1)* %valptr
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 false) nounwind readnone
%cmp = icmp eq i32 %ctlz, 32
%sel = select i1 %cmp, i32 -1, i32 %ctlz
@@ -212,13 +224,15 @@
}
; FUNC-LABEL: {{^}}v_ctlz_i32_sel_ne_bitwidth:
-; GCN: buffer_load_dword
+; GCN: {{buffer|flat}}_load_dword
; GCN: v_ffbh_u32_e32
; GCN: v_cmp
; GCN: v_cndmask
; GCN: s_endpgm
define amdgpu_kernel void @v_ctlz_i32_sel_ne_bitwidth(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
- %val = load i32, i32 addrspace(1)* %valptr
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 false) nounwind readnone
%cmp = icmp ne i32 %ctlz, 32
%sel = select i1 %cmp, i32 %ctlz, i32 -1
@@ -242,7 +256,7 @@
}
; FUNC-LABEL: {{^}}v_ctlz_i16_sel_eq_neg1:
-; SI: buffer_load_ushort [[VAL:v[0-9]+]],
+; SI: {{buffer|flat}}_load_ushort [[VAL:v[0-9]+]],
; SI: v_ffbh_u32_e32 [[FFBH:v[0-9]+]], [[VAL]]
; SI: buffer_store_short [[FFBH]],
define amdgpu_kernel void @v_ctlz_i16_sel_eq_neg1(i16 addrspace(1)* noalias %out, i16 addrspace(1)* noalias %valptr) nounwind {
diff --git a/llvm/test/CodeGen/AMDGPU/ctlz_zero_undef.ll b/llvm/test/CodeGen/AMDGPU/ctlz_zero_undef.ll
index 48f3e44..c99ce86 100644
--- a/llvm/test/CodeGen/AMDGPU/ctlz_zero_undef.ll
+++ b/llvm/test/CodeGen/AMDGPU/ctlz_zero_undef.ll
@@ -29,21 +29,23 @@
}
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32:
-; GCN: buffer_load_dword [[VAL:v[0-9]+]],
+; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm
; EG: MEM_RAT_CACHELESS STORE_RAW [[RESULT:T[0-9]+\.[XYZW]]]
; EG: FFBH_UINT {{\*? *}}[[RESULT]]
define amdgpu_kernel void @v_ctlz_zero_undef_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
- %val = load i32, i32 addrspace(1)* %valptr, align 4
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone
store i32 %ctlz, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_v2i32:
-; GCN: buffer_load_dwordx2
+; GCN: {{buffer|flat}}_load_dwordx2
; GCN: v_ffbh_u32_e32
; GCN: v_ffbh_u32_e32
; GCN: buffer_store_dwordx2
@@ -52,14 +54,16 @@
; EG: FFBH_UINT {{\*? *}}[[RESULT]]
; EG: FFBH_UINT {{\*? *}}[[RESULT]]
define amdgpu_kernel void @v_ctlz_zero_undef_v2i32(<2 x i32> addrspace(1)* noalias %out, <2 x i32> addrspace(1)* noalias %valptr) nounwind {
- %val = load <2 x i32>, <2 x i32> addrspace(1)* %valptr, align 8
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr <2 x i32>, <2 x i32> addrspace(1)* %valptr, i32 %tid
+ %val = load <2 x i32>, <2 x i32> addrspace(1)* %in.gep, align 8
%ctlz = call <2 x i32> @llvm.ctlz.v2i32(<2 x i32> %val, i1 true) nounwind readnone
store <2 x i32> %ctlz, <2 x i32> addrspace(1)* %out, align 8
ret void
}
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_v4i32:
-; GCN: buffer_load_dwordx4
+; GCN: {{buffer|flat}}_load_dwordx4
; GCN: v_ffbh_u32_e32
; GCN: v_ffbh_u32_e32
; GCN: v_ffbh_u32_e32
@@ -72,18 +76,22 @@
; EG: FFBH_UINT {{\*? *}}[[RESULT]]
; EG: FFBH_UINT {{\*? *}}[[RESULT]]
define amdgpu_kernel void @v_ctlz_zero_undef_v4i32(<4 x i32> addrspace(1)* noalias %out, <4 x i32> addrspace(1)* noalias %valptr) nounwind {
- %val = load <4 x i32>, <4 x i32> addrspace(1)* %valptr, align 16
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr <4 x i32>, <4 x i32> addrspace(1)* %valptr, i32 %tid
+ %val = load <4 x i32>, <4 x i32> addrspace(1)* %in.gep, align 16
%ctlz = call <4 x i32> @llvm.ctlz.v4i32(<4 x i32> %val, i1 true) nounwind readnone
store <4 x i32> %ctlz, <4 x i32> addrspace(1)* %out, align 16
ret void
}
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i8:
-; GCN: buffer_load_ubyte [[VAL:v[0-9]+]],
+; GCN: {{buffer|flat}}_load_ubyte [[VAL:v[0-9]+]],
; GCN: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; GCN: buffer_store_byte [[RESULT]],
define amdgpu_kernel void @v_ctlz_zero_undef_i8(i8 addrspace(1)* noalias %out, i8 addrspace(1)* noalias %valptr) nounwind {
- %val = load i8, i8 addrspace(1)* %valptr
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i8, i8 addrspace(1)* %valptr, i32 %tid
+ %val = load i8, i8 addrspace(1)* %in.gep
%ctlz = call i8 @llvm.ctlz.i8(i8 %val, i1 true) nounwind readnone
store i8 %ctlz, i8 addrspace(1)* %out
ret void
@@ -145,11 +153,13 @@
}
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_eq_neg1:
-; GCN: buffer_load_dword [[VAL:v[0-9]+]],
+; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[RESULT]],
- define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_eq_neg1(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
- %val = load i32, i32 addrspace(1)* %valptr
+define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_eq_neg1(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone
%cmp = icmp eq i32 %val, 0
%sel = select i1 %cmp, i32 -1, i32 %ctlz
@@ -158,11 +168,13 @@
}
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_ne_neg1:
-; GCN: buffer_load_dword [[VAL:v[0-9]+]],
+; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN: v_ffbh_u32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[RESULT]],
define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_ne_neg1(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
- %val = load i32, i32 addrspace(1)* %valptr
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone
%cmp = icmp ne i32 %val, 0
%sel = select i1 %cmp, i32 %ctlz, i32 -1
@@ -186,15 +198,17 @@
}
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_eq_neg1_two_use:
-; GCN: buffer_load_dword [[VAL:v[0-9]+]],
+; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN-DAG: v_ffbh_u32_e32 [[RESULT0:v[0-9]+]], [[VAL]]
; GCN-DAG: v_cmp_eq_u32_e32 vcc, 0, [[VAL]]
; GCN-DAG: v_cndmask_b32_e64 [[RESULT1:v[0-9]+]], 0, 1, vcc
; GCN-DAG: buffer_store_dword [[RESULT0]]
; GCN-DAG: buffer_store_byte [[RESULT1]]
; GCN: s_endpgm
- define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_eq_neg1_two_use(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
- %val = load i32, i32 addrspace(1)* %valptr
+define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_eq_neg1_two_use(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone
%cmp = icmp eq i32 %val, 0
%sel = select i1 %cmp, i32 -1, i32 %ctlz
@@ -205,13 +219,15 @@
; Selected on wrong constant
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_eq_0:
-; GCN: buffer_load_dword
+; GCN: {{buffer|flat}}_load_dword
; GCN: v_ffbh_u32_e32
; GCN: v_cmp
; GCN: v_cndmask
; GCN: buffer_store_dword
- define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_eq_0(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
- %val = load i32, i32 addrspace(1)* %valptr
+define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_eq_0(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone
%cmp = icmp eq i32 %val, 0
%sel = select i1 %cmp, i32 0, i32 %ctlz
@@ -221,13 +237,15 @@
; Selected on wrong constant
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_ne_0:
-; GCN: buffer_load_dword
+; GCN: {{buffer|flat}}_load_dword
; GCN: v_ffbh_u32_e32
; GCN: v_cmp
; GCN: v_cndmask
; GCN: buffer_store_dword
define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_ne_0(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
- %val = load i32, i32 addrspace(1)* %valptr
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone
%cmp = icmp ne i32 %val, 0
%sel = select i1 %cmp, i32 %ctlz, i32 0
@@ -237,13 +255,15 @@
; Compare on wrong constant
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_eq_cmp_non0:
-; GCN: buffer_load_dword
+; GCN: {{buffer|flat}}_load_dword
; GCN: v_ffbh_u32_e32
; GCN: v_cmp
; GCN: v_cndmask
; GCN: buffer_store_dword
- define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_eq_cmp_non0(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
- %val = load i32, i32 addrspace(1)* %valptr
+define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_eq_cmp_non0(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone
%cmp = icmp eq i32 %val, 1
%sel = select i1 %cmp, i32 0, i32 %ctlz
@@ -253,13 +273,15 @@
; Selected on wrong constant
; FUNC-LABEL: {{^}}v_ctlz_zero_undef_i32_sel_ne_cmp_non0:
-; GCN: buffer_load_dword
+; GCN: {{buffer|flat}}_load_dword
; GCN: v_ffbh_u32_e32
; GCN: v_cmp
; GCN: v_cndmask
; GCN: buffer_store_dword
define amdgpu_kernel void @v_ctlz_zero_undef_i32_sel_ne_cmp_non0(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
- %val = load i32, i32 addrspace(1)* %valptr
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep
%ctlz = call i32 @llvm.ctlz.i32(i32 %val, i1 true) nounwind readnone
%cmp = icmp ne i32 %val, 1
%sel = select i1 %cmp, i32 %ctlz, i32 0
diff --git a/llvm/test/CodeGen/AMDGPU/ctpop.ll b/llvm/test/CodeGen/AMDGPU/ctpop.ll
index aa913ad..68b39ba 100644
--- a/llvm/test/CodeGen/AMDGPU/ctpop.ll
+++ b/llvm/test/CodeGen/AMDGPU/ctpop.ll
@@ -8,6 +8,8 @@
declare <8 x i32> @llvm.ctpop.v8i32(<8 x i32>) nounwind readnone
declare <16 x i32> @llvm.ctpop.v16i32(<16 x i32>) nounwind readnone
+declare i32 @llvm.r600.read.tidig.x() nounwind readnone
+
; FUNC-LABEL: {{^}}s_ctpop_i32:
; GCN: s_load_dword [[SVAL:s[0-9]+]],
; GCN: s_bcnt1_i32_b32 [[SRESULT:s[0-9]+]], [[SVAL]]
@@ -24,22 +26,24 @@
; XXX - Why 0 in register?
; FUNC-LABEL: {{^}}v_ctpop_i32:
-; GCN: buffer_load_dword [[VAL:v[0-9]+]],
+; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN: v_bcnt_u32_b32{{(_e64)*}} [[RESULT:v[0-9]+]], [[VAL]], 0
; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm
; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
- %val = load i32, i32 addrspace(1)* %in, align 4
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone
store i32 %ctpop, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: {{^}}v_ctpop_add_chain_i32:
-; GCN: buffer_load_dword [[VAL1:v[0-9]+]],
-; GCN: buffer_load_dword [[VAL0:v[0-9]+]],
+; GCN: {{buffer|flat}}_load_dword [[VAL0:v[0-9]+]],
+; GCN: {{buffer|flat}}_load_dword [[VAL1:v[0-9]+]],
; GCN: v_bcnt_u32_b32{{(_e64)*}} [[MIDRESULT:v[0-9]+]], [[VAL1]], 0
; SI: v_bcnt_u32_b32_e32 [[RESULT:v[0-9]+]], [[VAL0]], [[MIDRESULT]]
; VI: v_bcnt_u32_b32 [[RESULT:v[0-9]+]], [[VAL0]], [[MIDRESULT]]
@@ -49,8 +53,11 @@
; EG: BCNT_INT
; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_add_chain_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in0, i32 addrspace(1)* noalias %in1) nounwind {
- %val0 = load i32, i32 addrspace(1)* %in0, align 4
- %val1 = load i32, i32 addrspace(1)* %in1, align 4
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in0.gep = getelementptr i32, i32 addrspace(1)* %in0, i32 %tid
+ %in1.gep = getelementptr i32, i32 addrspace(1)* %in1, i32 %tid
+ %val0 = load i32, i32 addrspace(1)* %in0.gep, align 4
+ %val1 = load i32, i32 addrspace(1)* %in1.gep, align 4
%ctpop0 = call i32 @llvm.ctpop.i32(i32 %val0) nounwind readnone
%ctpop1 = call i32 @llvm.ctpop.i32(i32 %val1) nounwind readnone
%add = add i32 %ctpop0, %ctpop1
@@ -59,15 +66,17 @@
}
; FUNC-LABEL: {{^}}v_ctpop_add_sgpr_i32:
-; GCN: buffer_load_dword [[VAL0:v[0-9]+]],
+; GCN: {{buffer|flat}}_load_dword [[VAL0:v[0-9]+]],
; GCN: s_waitcnt
; GCN-NEXT: v_bcnt_u32_b32{{(_e64)*}} [[RESULT:v[0-9]+]], [[VAL0]], s{{[0-9]+}}
; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm
-define amdgpu_kernel void @v_ctpop_add_sgpr_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in0, i32 addrspace(1)* noalias %in1, i32 %sval) nounwind {
- %val0 = load i32, i32 addrspace(1)* %in0, align 4
- %ctpop0 = call i32 @llvm.ctpop.i32(i32 %val0) nounwind readnone
- %add = add i32 %ctpop0, %sval
+define amdgpu_kernel void @v_ctpop_add_sgpr_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in, i32 %sval) nounwind {
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep, align 4
+ %ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone
+ %add = add i32 %ctpop, %sval
store i32 %add, i32 addrspace(1)* %out, align 4
ret void
}
@@ -80,7 +89,9 @@
; EG: BCNT_INT
; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_v2i32(<2 x i32> addrspace(1)* noalias %out, <2 x i32> addrspace(1)* noalias %in) nounwind {
- %val = load <2 x i32>, <2 x i32> addrspace(1)* %in, align 8
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr <2 x i32>, <2 x i32> addrspace(1)* %in, i32 %tid
+ %val = load <2 x i32>, <2 x i32> addrspace(1)* %in.gep, align 8
%ctpop = call <2 x i32> @llvm.ctpop.v2i32(<2 x i32> %val) nounwind readnone
store <2 x i32> %ctpop, <2 x i32> addrspace(1)* %out, align 8
ret void
@@ -98,7 +109,9 @@
; EG: BCNT_INT
; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_v4i32(<4 x i32> addrspace(1)* noalias %out, <4 x i32> addrspace(1)* noalias %in) nounwind {
- %val = load <4 x i32>, <4 x i32> addrspace(1)* %in, align 16
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr <4 x i32>, <4 x i32> addrspace(1)* %in, i32 %tid
+ %val = load <4 x i32>, <4 x i32> addrspace(1)* %in.gep, align 16
%ctpop = call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %val) nounwind readnone
store <4 x i32> %ctpop, <4 x i32> addrspace(1)* %out, align 16
ret void
@@ -124,7 +137,9 @@
; EG: BCNT_INT
; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_v8i32(<8 x i32> addrspace(1)* noalias %out, <8 x i32> addrspace(1)* noalias %in) nounwind {
- %val = load <8 x i32>, <8 x i32> addrspace(1)* %in, align 32
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr <8 x i32>, <8 x i32> addrspace(1)* %in, i32 %tid
+ %val = load <8 x i32>, <8 x i32> addrspace(1)* %in.gep, align 32
%ctpop = call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %val) nounwind readnone
store <8 x i32> %ctpop, <8 x i32> addrspace(1)* %out, align 32
ret void
@@ -166,21 +181,25 @@
; EG: BCNT_INT
; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_v16i32(<16 x i32> addrspace(1)* noalias %out, <16 x i32> addrspace(1)* noalias %in) nounwind {
- %val = load <16 x i32>, <16 x i32> addrspace(1)* %in, align 32
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr <16 x i32>, <16 x i32> addrspace(1)* %in, i32 %tid
+ %val = load <16 x i32>, <16 x i32> addrspace(1)* %in.gep, align 32
%ctpop = call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %val) nounwind readnone
store <16 x i32> %ctpop, <16 x i32> addrspace(1)* %out, align 32
ret void
}
; FUNC-LABEL: {{^}}v_ctpop_i32_add_inline_constant:
-; GCN: buffer_load_dword [[VAL:v[0-9]+]],
+; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN: v_bcnt_u32_b32{{(_e64)*}} [[RESULT:v[0-9]+]], [[VAL]], 4
; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm
; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_i32_add_inline_constant(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
- %val = load i32, i32 addrspace(1)* %in, align 4
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone
%add = add i32 %ctpop, 4
store i32 %add, i32 addrspace(1)* %out, align 4
@@ -188,14 +207,16 @@
}
; FUNC-LABEL: {{^}}v_ctpop_i32_add_inline_constant_inv:
-; GCN: buffer_load_dword [[VAL:v[0-9]+]],
+; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN: v_bcnt_u32_b32{{(_e64)*}} [[RESULT:v[0-9]+]], [[VAL]], 4
; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm
; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_i32_add_inline_constant_inv(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
- %val = load i32, i32 addrspace(1)* %in, align 4
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone
%add = add i32 4, %ctpop
store i32 %add, i32 addrspace(1)* %out, align 4
@@ -203,14 +224,16 @@
}
; FUNC-LABEL: {{^}}v_ctpop_i32_add_literal:
-; GCN-DAG: buffer_load_dword [[VAL:v[0-9]+]],
+; GCN-DAG: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN-DAG: v_mov_b32_e32 [[LIT:v[0-9]+]], 0x1869f
; SI: v_bcnt_u32_b32_e32 [[RESULT:v[0-9]+]], [[VAL]], [[LIT]]
; VI: v_bcnt_u32_b32 [[RESULT:v[0-9]+]], [[VAL]], [[LIT]]
; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm
define amdgpu_kernel void @v_ctpop_i32_add_literal(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
- %val = load i32, i32 addrspace(1)* %in, align 4
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone
%add = add i32 %ctpop, 99999
store i32 %add, i32 addrspace(1)* %out, align 4
@@ -218,7 +241,7 @@
}
; FUNC-LABEL: {{^}}v_ctpop_i32_add_var:
-; GCN-DAG: buffer_load_dword [[VAL:v[0-9]+]],
+; GCN-DAG: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN-DAG: s_load_dword [[VAR:s[0-9]+]],
; GCN: v_bcnt_u32_b32{{(_e64)*}} [[RESULT:v[0-9]+]], [[VAL]], [[VAR]]
; GCN: buffer_store_dword [[RESULT]],
@@ -226,7 +249,9 @@
; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_i32_add_var(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in, i32 %const) nounwind {
- %val = load i32, i32 addrspace(1)* %in, align 4
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone
%add = add i32 %ctpop, %const
store i32 %add, i32 addrspace(1)* %out, align 4
@@ -234,7 +259,7 @@
}
; FUNC-LABEL: {{^}}v_ctpop_i32_add_var_inv:
-; GCN-DAG: buffer_load_dword [[VAL:v[0-9]+]],
+; GCN-DAG: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; GCN-DAG: s_load_dword [[VAR:s[0-9]+]],
; GCN: v_bcnt_u32_b32{{(_e64)*}} [[RESULT:v[0-9]+]], [[VAL]], [[VAR]]
; GCN: buffer_store_dword [[RESULT]],
@@ -242,7 +267,9 @@
; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_i32_add_var_inv(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in, i32 %const) nounwind {
- %val = load i32, i32 addrspace(1)* %in, align 4
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone
%add = add i32 %const, %ctpop
store i32 %add, i32 addrspace(1)* %out, align 4
@@ -250,18 +277,22 @@
}
; FUNC-LABEL: {{^}}v_ctpop_i32_add_vvar_inv:
-; GCN-DAG: buffer_load_dword [[VAL:v[0-9]+]], off, s[{{[0-9]+:[0-9]+}}], {{0$}}
-; GCN-DAG: buffer_load_dword [[VAR:v[0-9]+]], off, s[{{[0-9]+:[0-9]+}}], 0 offset:16
-; SI: v_bcnt_u32_b32_e32 [[RESULT:v[0-9]+]], [[VAL]], [[VAR]]
+; SI: buffer_load_dword [[VAR:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0 addr64
+; SI: buffer_load_dword [[VAL:v[0-9]+]], v[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0 addr64
+; SI: v_bcnt_u32_b32_e32 [[RESULT:v[0-9]+]], [[VAR]], [[VAL]]
+; VI: flat_load_dword [[VAL:v[0-9]+]], v[{{[0-9]+:[0-9]+}}]
+; VI: flat_load_dword [[VAR:v[0-9]+]], v[{{[0-9]+:[0-9]+}}]
; VI: v_bcnt_u32_b32 [[RESULT:v[0-9]+]], [[VAL]], [[VAR]]
; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm
; EG: BCNT_INT
define amdgpu_kernel void @v_ctpop_i32_add_vvar_inv(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in, i32 addrspace(1)* noalias %constptr) nounwind {
- %val = load i32, i32 addrspace(1)* %in, align 4
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep, align 4
%ctpop = call i32 @llvm.ctpop.i32(i32 %val) nounwind readnone
- %gep = getelementptr i32, i32 addrspace(1)* %constptr, i32 4
+ %gep = getelementptr i32, i32 addrspace(1)* %constptr, i32 %tid
%const = load i32, i32 addrspace(1)* %gep, align 4
%add = add i32 %const, %ctpop
store i32 %add, i32 addrspace(1)* %out, align 4
diff --git a/llvm/test/CodeGen/AMDGPU/ctpop64.ll b/llvm/test/CodeGen/AMDGPU/ctpop64.ll
index f18bd9f..4850370 100644
--- a/llvm/test/CodeGen/AMDGPU/ctpop64.ll
+++ b/llvm/test/CodeGen/AMDGPU/ctpop64.ll
@@ -1,6 +1,8 @@
; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=GCN -check-prefix=FUNC %s
+declare i32 @llvm.r600.read.tidig.x() nounwind readnone
+
declare i64 @llvm.ctpop.i64(i64) nounwind readnone
declare <2 x i64> @llvm.ctpop.v2i64(<2 x i64>) nounwind readnone
declare <4 x i64> @llvm.ctpop.v4i64(<4 x i64>) nounwind readnone
@@ -25,14 +27,16 @@
}
; FUNC-LABEL: {{^}}v_ctpop_i64:
-; GCN: buffer_load_dwordx2 v{{\[}}[[LOVAL:[0-9]+]]:[[HIVAL:[0-9]+]]{{\]}},
+; GCN: {{buffer|flat}}_load_dwordx2 v{{\[}}[[LOVAL:[0-9]+]]:[[HIVAL:[0-9]+]]{{\]}},
; GCN: v_bcnt_u32_b32{{(_e64)*}} [[MIDRESULT:v[0-9]+]], v[[LOVAL]], 0
; SI-NEXT: v_bcnt_u32_b32_e32 [[RESULT:v[0-9]+]], v[[HIVAL]], [[MIDRESULT]]
; VI-NEXT: v_bcnt_u32_b32 [[RESULT:v[0-9]+]], v[[HIVAL]], [[MIDRESULT]]
; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm
define amdgpu_kernel void @v_ctpop_i64(i32 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %in) nounwind {
- %val = load i64, i64 addrspace(1)* %in, align 8
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i64, i64 addrspace(1)* %in, i32 %tid
+ %val = load i64, i64 addrspace(1)* %in.gep, align 8
%ctpop = call i64 @llvm.ctpop.i64(i64 %val) nounwind readnone
%truncctpop = trunc i64 %ctpop to i32
store i32 %truncctpop, i32 addrspace(1)* %out, align 4
@@ -40,7 +44,7 @@
}
; FUNC-LABEL: {{^}}v_ctpop_i64_user:
-; GCN: buffer_load_dwordx2 v{{\[}}[[LOVAL:[0-9]+]]:[[HIVAL:[0-9]+]]{{\]}},
+; GCN: {{buffer|flat}}_load_dwordx2 v{{\[}}[[LOVAL:[0-9]+]]:[[HIVAL:[0-9]+]]{{\]}},
; GCN: v_bcnt_u32_b32{{(_e64)*}} [[MIDRESULT:v[0-9]+]], v[[LOVAL]], 0
; SI-NEXT: v_bcnt_u32_b32_e32 [[RESULT:v[0-9]+]], v[[HIVAL]], [[MIDRESULT]]
; VI-NEXT: v_bcnt_u32_b32 [[RESULT:v[0-9]+]], v[[HIVAL]], [[MIDRESULT]]
@@ -49,7 +53,9 @@
; GCN: buffer_store_dwordx2 v{{\[}}[[RESULT_LO]]:[[RESULT_HI]]{{\]}}
; GCN: s_endpgm
define amdgpu_kernel void @v_ctpop_i64_user(i64 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %in, i64 %s.val) nounwind {
- %val = load i64, i64 addrspace(1)* %in, align 8
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i64, i64 addrspace(1)* %in, i32 %tid
+ %val = load i64, i64 addrspace(1)* %in.gep, align 8
%ctpop = call i64 @llvm.ctpop.i64(i64 %val) nounwind readnone
%or = or i64 %ctpop, %s.val
store i64 %or, i64 addrspace(1)* %out
@@ -87,7 +93,9 @@
; GCN: v_bcnt_u32_b32
; GCN: s_endpgm
define amdgpu_kernel void @v_ctpop_v2i64(<2 x i32> addrspace(1)* noalias %out, <2 x i64> addrspace(1)* noalias %in) nounwind {
- %val = load <2 x i64>, <2 x i64> addrspace(1)* %in, align 16
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr <2 x i64>, <2 x i64> addrspace(1)* %in, i32 %tid
+ %val = load <2 x i64>, <2 x i64> addrspace(1)* %in.gep, align 16
%ctpop = call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %val) nounwind readnone
%truncctpop = trunc <2 x i64> %ctpop to <2 x i32>
store <2 x i32> %truncctpop, <2 x i32> addrspace(1)* %out, align 8
@@ -105,7 +113,9 @@
; GCN: v_bcnt_u32_b32
; GCN: s_endpgm
define amdgpu_kernel void @v_ctpop_v4i64(<4 x i32> addrspace(1)* noalias %out, <4 x i64> addrspace(1)* noalias %in) nounwind {
- %val = load <4 x i64>, <4 x i64> addrspace(1)* %in, align 32
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr <4 x i64>, <4 x i64> addrspace(1)* %in, i32 %tid
+ %val = load <4 x i64>, <4 x i64> addrspace(1)* %in.gep, align 32
%ctpop = call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %val) nounwind readnone
%truncctpop = trunc <4 x i64> %ctpop to <4 x i32>
store <4 x i32> %truncctpop, <4 x i32> addrspace(1)* %out, align 16
@@ -169,7 +179,8 @@
; FIXME: Should not have extra add
; FUNC-LABEL: {{^}}v_ctpop_i128:
-; GCN: buffer_load_dwordx4 v{{\[}}[[VAL0:[0-9]+]]:[[VAL3:[0-9]+]]{{\]}}, off, s{{\[[0-9]+:[0-9]+\]}}, 0{{$}}
+; SI: buffer_load_dwordx4 v{{\[}}[[VAL0:[0-9]+]]:[[VAL3:[0-9]+]]{{\]}}, v{{\[[0-9]+:[0-9]+\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0 addr64
+; VI: flat_load_dwordx4 v{{\[}}[[VAL0:[0-9]+]]:[[VAL3:[0-9]+]]{{\]}}, v{{\[[0-9]+:[0-9]+\]}}
; GCN-DAG: v_bcnt_u32_b32{{(_e64)*}} [[MIDRESULT0:v[0-9]+]], v{{[0-9]+}}, 0
; GCN-DAG: v_bcnt_u32_b32{{(_e32)*(_e64)*}} [[MIDRESULT1:v[0-9]+]], v[[VAL3]], [[MIDRESULT0]]
@@ -182,7 +193,9 @@
; GCN: buffer_store_dword [[RESULT]],
; GCN: s_endpgm
define amdgpu_kernel void @v_ctpop_i128(i32 addrspace(1)* noalias %out, i128 addrspace(1)* noalias %in) nounwind {
- %val = load i128, i128 addrspace(1)* %in, align 8
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i128, i128 addrspace(1)* %in, i32 %tid
+ %val = load i128, i128 addrspace(1)* %in.gep, align 8
%ctpop = call i128 @llvm.ctpop.i128(i128 %val) nounwind readnone
%truncctpop = trunc i128 %ctpop to i32
store i32 %truncctpop, i32 addrspace(1)* %out, align 4
diff --git a/llvm/test/CodeGen/AMDGPU/cttz_zero_undef.ll b/llvm/test/CodeGen/AMDGPU/cttz_zero_undef.ll
index 1fa6407..1bfd38d 100644
--- a/llvm/test/CodeGen/AMDGPU/cttz_zero_undef.ll
+++ b/llvm/test/CodeGen/AMDGPU/cttz_zero_undef.ll
@@ -5,6 +5,7 @@
declare i32 @llvm.cttz.i32(i32, i1) nounwind readnone
declare <2 x i32> @llvm.cttz.v2i32(<2 x i32>, i1) nounwind readnone
declare <4 x i32> @llvm.cttz.v4i32(<4 x i32>, i1) nounwind readnone
+declare i32 @llvm.r600.read.tidig.x() nounwind readnone
; FUNC-LABEL: {{^}}s_cttz_zero_undef_i32:
; SI: s_load_dword [[VAL:s[0-9]+]],
@@ -21,21 +22,23 @@
}
; FUNC-LABEL: {{^}}v_cttz_zero_undef_i32:
-; SI: buffer_load_dword [[VAL:v[0-9]+]],
+; SI: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]],
; SI: v_ffbl_b32_e32 [[RESULT:v[0-9]+]], [[VAL]]
; SI: buffer_store_dword [[RESULT]],
; SI: s_endpgm
; EG: MEM_RAT_CACHELESS STORE_RAW [[RESULT:T[0-9]+\.[XYZW]]]
; EG: FFBL_INT {{\*? *}}[[RESULT]]
define amdgpu_kernel void @v_cttz_zero_undef_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %valptr) nounwind {
- %val = load i32, i32 addrspace(1)* %valptr, align 4
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr i32, i32 addrspace(1)* %valptr, i32 %tid
+ %val = load i32, i32 addrspace(1)* %in.gep, align 4
%cttz = call i32 @llvm.cttz.i32(i32 %val, i1 true) nounwind readnone
store i32 %cttz, i32 addrspace(1)* %out, align 4
ret void
}
; FUNC-LABEL: {{^}}v_cttz_zero_undef_v2i32:
-; SI: buffer_load_dwordx2
+; SI: {{buffer|flat}}_load_dwordx2
; SI: v_ffbl_b32_e32
; SI: v_ffbl_b32_e32
; SI: buffer_store_dwordx2
@@ -44,14 +47,16 @@
; EG: FFBL_INT {{\*? *}}[[RESULT]]
; EG: FFBL_INT {{\*? *}}[[RESULT]]
define amdgpu_kernel void @v_cttz_zero_undef_v2i32(<2 x i32> addrspace(1)* noalias %out, <2 x i32> addrspace(1)* noalias %valptr) nounwind {
- %val = load <2 x i32>, <2 x i32> addrspace(1)* %valptr, align 8
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr <2 x i32>, <2 x i32> addrspace(1)* %valptr, i32 %tid
+ %val = load <2 x i32>, <2 x i32> addrspace(1)* %in.gep, align 8
%cttz = call <2 x i32> @llvm.cttz.v2i32(<2 x i32> %val, i1 true) nounwind readnone
store <2 x i32> %cttz, <2 x i32> addrspace(1)* %out, align 8
ret void
}
; FUNC-LABEL: {{^}}v_cttz_zero_undef_v4i32:
-; SI: buffer_load_dwordx4
+; SI: {{buffer|flat}}_load_dwordx4
; SI: v_ffbl_b32_e32
; SI: v_ffbl_b32_e32
; SI: v_ffbl_b32_e32
@@ -64,7 +69,9 @@
; EG: FFBL_INT {{\*? *}}[[RESULT]]
; EG: FFBL_INT {{\*? *}}[[RESULT]]
define amdgpu_kernel void @v_cttz_zero_undef_v4i32(<4 x i32> addrspace(1)* noalias %out, <4 x i32> addrspace(1)* noalias %valptr) nounwind {
- %val = load <4 x i32>, <4 x i32> addrspace(1)* %valptr, align 16
+ %tid = call i32 @llvm.r600.read.tidig.x()
+ %in.gep = getelementptr <4 x i32>, <4 x i32> addrspace(1)* %valptr, i32 %tid
+ %val = load <4 x i32>, <4 x i32> addrspace(1)* %in.gep, align 16
%cttz = call <4 x i32> @llvm.cttz.v4i32(<4 x i32> %val, i1 true) nounwind readnone
store <4 x i32> %cttz, <4 x i32> addrspace(1)* %out, align 16
ret void
diff --git a/llvm/test/CodeGen/AMDGPU/cvt_f32_ubyte.ll b/llvm/test/CodeGen/AMDGPU/cvt_f32_ubyte.ll
index 0328ce3..f839129 100644
--- a/llvm/test/CodeGen/AMDGPU/cvt_f32_ubyte.ll
+++ b/llvm/test/CodeGen/AMDGPU/cvt_f32_ubyte.ll
@@ -5,46 +5,52 @@
declare i32 @llvm.amdgcn.workitem.id.y() nounwind readnone
; GCN-LABEL: {{^}}load_i8_to_f32:
-; GCN: buffer_load_ubyte [[LOADREG:v[0-9]+]],
+; GCN: {{buffer|flat}}_load_ubyte [[LOADREG:v[0-9]+]],
; GCN-NOT: bfe
; GCN-NOT: lshr
; GCN: v_cvt_f32_ubyte0_e32 [[CONV:v[0-9]+]], [[LOADREG]]
; GCN: buffer_store_dword [[CONV]],
define amdgpu_kernel void @load_i8_to_f32(float addrspace(1)* noalias %out, i8 addrspace(1)* noalias %in) nounwind {
- %load = load i8, i8 addrspace(1)* %in, align 1
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr i8, i8 addrspace(1)* %in, i32 %tid
+ %load = load i8, i8 addrspace(1)* %gep, align 1
%cvt = uitofp i8 %load to float
store float %cvt, float addrspace(1)* %out, align 4
ret void
}
; GCN-LABEL: {{^}}load_v2i8_to_v2f32:
-; GCN: buffer_load_ushort [[LD:v[0-9]+]]
+; GCN: {{buffer|flat}}_load_ushort [[LD:v[0-9]+]]
; GCN-DAG: v_cvt_f32_ubyte1_e32 v[[HIRESULT:[0-9]+]], [[LD]]
; GCN-DAG: v_cvt_f32_ubyte0_e32 v[[LORESULT:[0-9]+]], [[LD]]
; GCN: buffer_store_dwordx2 v{{\[}}[[LORESULT]]:[[HIRESULT]]{{\]}},
define amdgpu_kernel void @load_v2i8_to_v2f32(<2 x float> addrspace(1)* noalias %out, <2 x i8> addrspace(1)* noalias %in) nounwind {
- %load = load <2 x i8>, <2 x i8> addrspace(1)* %in, align 2
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <2 x i8>, <2 x i8> addrspace(1)* %in, i32 %tid
+ %load = load <2 x i8>, <2 x i8> addrspace(1)* %gep, align 2
%cvt = uitofp <2 x i8> %load to <2 x float>
store <2 x float> %cvt, <2 x float> addrspace(1)* %out, align 16
ret void
}
; GCN-LABEL: {{^}}load_v3i8_to_v3f32:
-; GCN: buffer_load_dword [[VAL:v[0-9]+]]
+; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]]
; GCN-NOT: v_cvt_f32_ubyte3_e32
; GCN-DAG: v_cvt_f32_ubyte2_e32 v{{[0-9]+}}, [[VAL]]
; GCN-DAG: v_cvt_f32_ubyte1_e32 v[[HIRESULT:[0-9]+]], [[VAL]]
; GCN-DAG: v_cvt_f32_ubyte0_e32 v[[LORESULT:[0-9]+]], [[VAL]]
; GCN: buffer_store_dwordx2 v{{\[}}[[LORESULT]]:[[HIRESULT]]{{\]}},
define amdgpu_kernel void @load_v3i8_to_v3f32(<3 x float> addrspace(1)* noalias %out, <3 x i8> addrspace(1)* noalias %in) nounwind {
- %load = load <3 x i8>, <3 x i8> addrspace(1)* %in, align 4
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <3 x i8>, <3 x i8> addrspace(1)* %in, i32 %tid
+ %load = load <3 x i8>, <3 x i8> addrspace(1)* %gep, align 4
%cvt = uitofp <3 x i8> %load to <3 x float>
store <3 x float> %cvt, <3 x float> addrspace(1)* %out, align 16
ret void
}
; GCN-LABEL: {{^}}load_v4i8_to_v4f32:
-; GCN: buffer_load_dword [[LOADREG:v[0-9]+]]
+; GCN: {{buffer|flat}}_load_dword [[LOADREG:v[0-9]+]]
; GCN-NOT: bfe
; GCN-NOT: lshr
; GCN-DAG: v_cvt_f32_ubyte3_e32 v[[HIRESULT:[0-9]+]], [[LOADREG]]
@@ -53,7 +59,9 @@
; GCN-DAG: v_cvt_f32_ubyte0_e32 v[[LORESULT:[0-9]+]], [[LOADREG]]
; GCN: buffer_store_dwordx4 v{{\[}}[[LORESULT]]:[[HIRESULT]]{{\]}},
define amdgpu_kernel void @load_v4i8_to_v4f32(<4 x float> addrspace(1)* noalias %out, <4 x i8> addrspace(1)* noalias %in) nounwind {
- %load = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 4
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %in, i32 %tid
+ %load = load <4 x i8>, <4 x i8> addrspace(1)* %gep, align 4
%cvt = uitofp <4 x i8> %load to <4 x float>
store <4 x float> %cvt, <4 x float> addrspace(1)* %out, align 16
ret void
@@ -64,10 +72,10 @@
; FIXME: Packing bytes
; GCN-LABEL: {{^}}load_v4i8_to_v4f32_unaligned:
-; GCN: buffer_load_ubyte [[LOADREG3:v[0-9]+]]
-; GCN: buffer_load_ubyte [[LOADREG2:v[0-9]+]]
-; GCN: buffer_load_ubyte [[LOADREG1:v[0-9]+]]
-; GCN: buffer_load_ubyte [[LOADREG0:v[0-9]+]]
+; GCN: {{buffer|flat}}_load_ubyte [[LOADREG3:v[0-9]+]]
+; GCN: {{buffer|flat}}_load_ubyte [[LOADREG2:v[0-9]+]]
+; GCN: {{buffer|flat}}_load_ubyte [[LOADREG1:v[0-9]+]]
+; GCN: {{buffer|flat}}_load_ubyte [[LOADREG0:v[0-9]+]]
; GCN-DAG: v_lshlrev_b32
; GCN-DAG: v_or_b32
; GCN-DAG: v_cvt_f32_ubyte0_e32 v[[LORESULT:[0-9]+]],
@@ -77,7 +85,9 @@
; GCN: buffer_store_dwordx4
define amdgpu_kernel void @load_v4i8_to_v4f32_unaligned(<4 x float> addrspace(1)* noalias %out, <4 x i8> addrspace(1)* noalias %in) nounwind {
- %load = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 1
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %in, i32 %tid
+ %load = load <4 x i8>, <4 x i8> addrspace(1)* %gep, align 1
%cvt = uitofp <4 x i8> %load to <4 x float>
store <4 x float> %cvt, <4 x float> addrspace(1)* %out, align 16
ret void
@@ -124,14 +134,16 @@
; GCN-LABEL: {{^}}load_v7i8_to_v7f32:
; GCN: s_endpgm
define amdgpu_kernel void @load_v7i8_to_v7f32(<7 x float> addrspace(1)* noalias %out, <7 x i8> addrspace(1)* noalias %in) nounwind {
- %load = load <7 x i8>, <7 x i8> addrspace(1)* %in, align 1
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <7 x i8>, <7 x i8> addrspace(1)* %in, i32 %tid
+ %load = load <7 x i8>, <7 x i8> addrspace(1)* %gep, align 1
%cvt = uitofp <7 x i8> %load to <7 x float>
store <7 x float> %cvt, <7 x float> addrspace(1)* %out, align 16
ret void
}
; GCN-LABEL: {{^}}load_v8i8_to_v8f32:
-; GCN: buffer_load_dwordx2 v{{\[}}[[LOLOAD:[0-9]+]]:[[HILOAD:[0-9]+]]{{\]}},
+; GCN: {{buffer|flat}}_load_dwordx2 v{{\[}}[[LOLOAD:[0-9]+]]:[[HILOAD:[0-9]+]]{{\]}},
; GCN-NOT: bfe
; GCN-NOT: lshr
; GCN-DAG: v_cvt_f32_ubyte3_e32 v{{[0-9]+}}, v[[LOLOAD]]
@@ -147,19 +159,23 @@
; GCN: buffer_store_dwordx4
; GCN: buffer_store_dwordx4
define amdgpu_kernel void @load_v8i8_to_v8f32(<8 x float> addrspace(1)* noalias %out, <8 x i8> addrspace(1)* noalias %in) nounwind {
- %load = load <8 x i8>, <8 x i8> addrspace(1)* %in, align 8
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <8 x i8>, <8 x i8> addrspace(1)* %in, i32 %tid
+ %load = load <8 x i8>, <8 x i8> addrspace(1)* %gep, align 8
%cvt = uitofp <8 x i8> %load to <8 x float>
store <8 x float> %cvt, <8 x float> addrspace(1)* %out, align 16
ret void
}
; GCN-LABEL: {{^}}i8_zext_inreg_i32_to_f32:
-; GCN: buffer_load_dword [[LOADREG:v[0-9]+]],
+; GCN: {{buffer|flat}}_load_dword [[LOADREG:v[0-9]+]],
; GCN: v_add_i32_e32 [[ADD:v[0-9]+]], vcc, 2, [[LOADREG]]
; GCN-NEXT: v_cvt_f32_ubyte0_e32 [[CONV:v[0-9]+]], [[ADD]]
; GCN: buffer_store_dword [[CONV]],
define amdgpu_kernel void @i8_zext_inreg_i32_to_f32(float addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
- %load = load i32, i32 addrspace(1)* %in, align 4
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
+ %load = load i32, i32 addrspace(1)* %gep, align 4
%add = add i32 %load, 2
%inreg = and i32 %add, 255
%cvt = uitofp i32 %inreg to float
@@ -169,7 +185,9 @@
; GCN-LABEL: {{^}}i8_zext_inreg_hi1_to_f32:
define amdgpu_kernel void @i8_zext_inreg_hi1_to_f32(float addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
- %load = load i32, i32 addrspace(1)* %in, align 4
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
+ %load = load i32, i32 addrspace(1)* %gep, align 4
%inreg = and i32 %load, 65280
%shr = lshr i32 %inreg, 8
%cvt = uitofp i32 %shr to float
@@ -181,7 +199,9 @@
; them so it shouldn't really matter.
; GCN-LABEL: {{^}}i8_zext_i32_to_f32:
define amdgpu_kernel void @i8_zext_i32_to_f32(float addrspace(1)* noalias %out, i8 addrspace(1)* noalias %in) nounwind {
- %load = load i8, i8 addrspace(1)* %in, align 1
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr i8, i8 addrspace(1)* %in, i32 %tid
+ %load = load i8, i8 addrspace(1)* %gep, align 1
%ext = zext i8 %load to i32
%cvt = uitofp i32 %ext to float
store float %cvt, float addrspace(1)* %out, align 4
@@ -190,7 +210,9 @@
; GCN-LABEL: {{^}}v4i8_zext_v4i32_to_v4f32:
define amdgpu_kernel void @v4i8_zext_v4i32_to_v4f32(<4 x float> addrspace(1)* noalias %out, <4 x i8> addrspace(1)* noalias %in) nounwind {
- %load = load <4 x i8>, <4 x i8> addrspace(1)* %in, align 1
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <4 x i8>, <4 x i8> addrspace(1)* %in, i32 %tid
+ %load = load <4 x i8>, <4 x i8> addrspace(1)* %gep, align 1
%ext = zext <4 x i8> %load to <4 x i32>
%cvt = uitofp <4 x i32> %ext to <4 x float>
store <4 x float> %cvt, <4 x float> addrspace(1)* %out, align 16
@@ -198,12 +220,14 @@
}
; GCN-LABEL: {{^}}extract_byte0_to_f32:
-; GCN: buffer_load_dword [[VAL:v[0-9]+]]
+; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]]
; GCN-NOT: [[VAL]]
; GCN: v_cvt_f32_ubyte0_e32 [[CONV:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[CONV]]
define amdgpu_kernel void @extract_byte0_to_f32(float addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
- %val = load i32, i32 addrspace(1)* %in
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
+ %val = load i32, i32 addrspace(1)* %gep
%and = and i32 %val, 255
%cvt = uitofp i32 %and to float
store float %cvt, float addrspace(1)* %out
@@ -211,12 +235,14 @@
}
; GCN-LABEL: {{^}}extract_byte1_to_f32:
-; GCN: buffer_load_dword [[VAL:v[0-9]+]]
+; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]]
; GCN-NOT: [[VAL]]
; GCN: v_cvt_f32_ubyte1_e32 [[CONV:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[CONV]]
define amdgpu_kernel void @extract_byte1_to_f32(float addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
- %val = load i32, i32 addrspace(1)* %in
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
+ %val = load i32, i32 addrspace(1)* %gep
%srl = lshr i32 %val, 8
%and = and i32 %srl, 255
%cvt = uitofp i32 %and to float
@@ -225,12 +251,14 @@
}
; GCN-LABEL: {{^}}extract_byte2_to_f32:
-; GCN: buffer_load_dword [[VAL:v[0-9]+]]
+; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]]
; GCN-NOT: [[VAL]]
; GCN: v_cvt_f32_ubyte2_e32 [[CONV:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[CONV]]
define amdgpu_kernel void @extract_byte2_to_f32(float addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
- %val = load i32, i32 addrspace(1)* %in
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
+ %val = load i32, i32 addrspace(1)* %gep
%srl = lshr i32 %val, 16
%and = and i32 %srl, 255
%cvt = uitofp i32 %and to float
@@ -239,12 +267,14 @@
}
; GCN-LABEL: {{^}}extract_byte3_to_f32:
-; GCN: buffer_load_dword [[VAL:v[0-9]+]]
+; GCN: {{buffer|flat}}_load_dword [[VAL:v[0-9]+]]
; GCN-NOT: [[VAL]]
; GCN: v_cvt_f32_ubyte3_e32 [[CONV:v[0-9]+]], [[VAL]]
; GCN: buffer_store_dword [[CONV]]
define amdgpu_kernel void @extract_byte3_to_f32(float addrspace(1)* noalias %out, i32 addrspace(1)* noalias %in) nounwind {
- %val = load i32, i32 addrspace(1)* %in
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr i32, i32 addrspace(1)* %in, i32 %tid
+ %val = load i32, i32 addrspace(1)* %gep
%srl = lshr i32 %val, 24
%and = and i32 %srl, 255
%cvt = uitofp i32 %and to float
diff --git a/llvm/test/CodeGen/AMDGPU/early-if-convert-cost.ll b/llvm/test/CodeGen/AMDGPU/early-if-convert-cost.ll
index ace0159..7440498 100644
--- a/llvm/test/CodeGen/AMDGPU/early-if-convert-cost.ll
+++ b/llvm/test/CodeGen/AMDGPU/early-if-convert-cost.ll
@@ -1,4 +1,4 @@
-; RUN: llc -stress-early-ifcvt -amdgpu-early-ifcvt=1 -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -stress-early-ifcvt -amdgpu-early-ifcvt=1 -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; FIXME: Most of these cases that don't trigger because of broken cost
; heuristics. Should not need -stress-early-ifcvt
diff --git a/llvm/test/CodeGen/AMDGPU/early-if-convert.ll b/llvm/test/CodeGen/AMDGPU/early-if-convert.ll
index 9439130..792f0b1 100644
--- a/llvm/test/CodeGen/AMDGPU/early-if-convert.ll
+++ b/llvm/test/CodeGen/AMDGPU/early-if-convert.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=amdgcn -mcpu=verde -amdgpu-early-ifcvt=1 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -amdgpu-early-ifcvt=1 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; XUN: llc -march=amdgcn -mcpu=tonga -amdgpu-early-ifcvt=1 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; FIXME: This leaves behind a now unnecessary and with exec
diff --git a/llvm/test/CodeGen/AMDGPU/enable-no-signed-zeros-fp-math.ll b/llvm/test/CodeGen/AMDGPU/enable-no-signed-zeros-fp-math.ll
index 6eb1fc1..4e52a5f 100644
--- a/llvm/test/CodeGen/AMDGPU/enable-no-signed-zeros-fp-math.ll
+++ b/llvm/test/CodeGen/AMDGPU/enable-no-signed-zeros-fp-math.ll
@@ -2,6 +2,8 @@
; RUN: llc -march=amdgcn -enable-no-signed-zeros-fp-math=1 < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-UNSAFE %s
; RUN: llc -march=amdgcn -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-UNSAFE %s
+declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
+
; Test that the -enable-no-signed-zeros-fp-math flag works
; GCN-LABEL: {{^}}fneg_fsub_f32:
@@ -10,8 +12,11 @@
; GCN-UNSAFE-NOT: xor
define amdgpu_kernel void @fneg_fsub_f32(float addrspace(1)* %out, float addrspace(1)* %in) #0 {
- %b_ptr = getelementptr float, float addrspace(1)* %in, i32 1
- %a = load float, float addrspace(1)* %in, align 4
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %add = add i32 %tid, 1
+ %gep = getelementptr float, float addrspace(1)* %in, i32 %tid
+ %b_ptr = getelementptr float, float addrspace(1)* %in, i32 %add
+ %a = load float, float addrspace(1)* %gep, align 4
%b = load float, float addrspace(1)* %b_ptr, align 4
%result = fsub float %a, %b
%neg.result = fsub float -0.0, %result
diff --git a/llvm/test/CodeGen/AMDGPU/extractelt-to-trunc.ll b/llvm/test/CodeGen/AMDGPU/extractelt-to-trunc.ll
index 34999fa..3fb452d 100644
--- a/llvm/test/CodeGen/AMDGPU/extractelt-to-trunc.ll
+++ b/llvm/test/CodeGen/AMDGPU/extractelt-to-trunc.ll
@@ -1,5 +1,7 @@
; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
+
; Make sure the add and load are reduced to 32-bits even with the
; bitcast to vector.
; GCN-LABEL: {{^}}bitcast_int_to_vector_extract_0:
@@ -8,7 +10,9 @@
; GCN: v_add_i32_e32 [[ADD:v[0-9]+]], vcc, [[B]], [[A]]
; GCN: buffer_store_dword [[ADD]]
define amdgpu_kernel void @bitcast_int_to_vector_extract_0(i32 addrspace(1)* %out, i64 addrspace(1)* %in, i64 %b) {
- %a = load i64, i64 addrspace(1)* %in
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr i64, i64 addrspace(1)* %in, i32 %tid
+ %a = load i64, i64 addrspace(1)* %gep
%add = add i64 %a, %b
%val.bc = bitcast i64 %add to <2 x i32>
%extract = extractelement <2 x i32> %val.bc, i32 0
@@ -21,7 +25,9 @@
; GCN: v_add_f64
; GCN: buffer_store_dword v
define amdgpu_kernel void @bitcast_fp_to_vector_extract_0(i32 addrspace(1)* %out, double addrspace(1)* %in, double %b) {
- %a = load double, double addrspace(1)* %in
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr double, double addrspace(1)* %in, i32 %tid
+ %a = load double, double addrspace(1)* %gep
%add = fadd double %a, %b
%val.bc = bitcast double %add to <2 x i32>
%extract = extractelement <2 x i32> %val.bc, i32 0
@@ -34,7 +40,9 @@
; GCN: v_add_i32
; GCN: buffer_store_dword
define amdgpu_kernel void @bitcast_int_to_fpvector_extract_0(float addrspace(1)* %out, i64 addrspace(1)* %in, i64 %b) {
- %a = load i64, i64 addrspace(1)* %in
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr i64, i64 addrspace(1)* %in, i32 %tid
+ %a = load i64, i64 addrspace(1)* %gep
%add = add i64 %a, %b
%val.bc = bitcast i64 %add to <2 x float>
%extract = extractelement <2 x float> %val.bc, i32 0
diff --git a/llvm/test/CodeGen/AMDGPU/fabs.f16.ll b/llvm/test/CodeGen/AMDGPU/fabs.f16.ll
index 4e2ec4b..158e644 100644
--- a/llvm/test/CodeGen/AMDGPU/fabs.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fabs.f16.ll
@@ -134,7 +134,9 @@
; GFX9: v_and_b32_e32 [[FABS:v[0-9]+]], 0x7fff7fff, [[VAL]]
; GFX9: v_pk_mul_f16 v{{[0-9]+}}, [[FABS]], v{{[0-9]+$}}
define amdgpu_kernel void @v_fabs_fold_v2f16(<2 x half> addrspace(1)* %out, <2 x half> addrspace(1)* %in) #0 {
- %val = load <2 x half>, <2 x half> addrspace(1)* %in
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <2 x half>, <2 x half> addrspace(1)* %in, i32 %tid
+ %val = load <2 x half>, <2 x half> addrspace(1)* %gep
%fabs = call <2 x half> @llvm.fabs.v2f16(<2 x half> %val)
%fmul = fmul <2 x half> %fabs, %val
store <2 x half> %fmul, <2 x half> addrspace(1)* %out
diff --git a/llvm/test/CodeGen/AMDGPU/fadd.f16.ll b/llvm/test/CodeGen/AMDGPU/fadd.f16.ll
index 08199be..1d7652f 100644
--- a/llvm/test/CodeGen/AMDGPU/fadd.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fadd.f16.ll
@@ -2,8 +2,8 @@
; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}fadd_f16
-; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
-; GCN: buffer_load_ushort v[[B_F16:[0-9]+]]
+; GCN: {{buffer|flat}}_load_ushort v[[A_F16:[0-9]+]]
+; GCN: {{buffer|flat}}_load_ushort v[[B_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; SI: v_cvt_f32_f16_e32 v[[B_F32:[0-9]+]], v[[B_F16]]
; SI: v_add_f32_e32 v[[R_F32:[0-9]+]], v[[B_F32]], v[[A_F32]]
@@ -24,7 +24,7 @@
}
; GCN-LABEL: {{^}}fadd_f16_imm_a
-; GCN: buffer_load_ushort v[[B_F16:[0-9]+]]
+; GCN: {{buffer|flat}}_load_ushort v[[B_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[B_F32:[0-9]+]], v[[B_F16]]
; SI: v_add_f32_e32 v[[R_F32:[0-9]+]], 1.0, v[[B_F32]]
; SI: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[R_F32]]
@@ -42,7 +42,7 @@
}
; GCN-LABEL: {{^}}fadd_f16_imm_b
-; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
+; GCN: {{buffer|flat}}_load_ushort v[[A_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32:[0-9]+]], v[[A_F16]]
; SI: v_add_f32_e32 v[[R_F32:[0-9]+]], 2.0, v[[A_F32]]
; SI: v_cvt_f16_f32_e32 v[[R_F16:[0-9]+]], v[[R_F32]]
@@ -60,8 +60,8 @@
}
; GCN-LABEL: {{^}}fadd_v2f16:
-; GCN: buffer_load_dword v[[A_V2_F16:[0-9]+]]
-; GCN: buffer_load_dword v[[B_V2_F16:[0-9]+]]
+; GCN: {{buffer|flat}}_load_dword v[[A_V2_F16:[0-9]+]]
+; GCN: {{buffer|flat}}_load_dword v[[B_V2_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32_0:[0-9]+]], v[[A_V2_F16]]
; SI: v_lshrrev_b32_e32 v[[A_F16_1:[0-9]+]], 16, v[[A_V2_F16]]
@@ -70,8 +70,8 @@
; SI-DAG: v_cvt_f32_f16_e32 v[[A_F32_1:[0-9]+]], v[[A_F16_1]]
; SI-DAG: v_cvt_f32_f16_e32 v[[B_F32_1:[0-9]+]], v[[B_F16_1]]
-; SI: v_add_f32_e32 v[[R_F32_0:[0-9]+]], v[[B_F32_0]], v[[A_F32_0]]
-; SI: v_add_f32_e32 v[[R_F32_1:[0-9]+]], v[[B_F32_1]], v[[A_F32_1]]
+; SI-DAG: v_add_f32_e32 v[[R_F32_0:[0-9]+]], v[[B_F32_0]], v[[A_F32_0]]
+; SI-DAG: v_add_f32_e32 v[[R_F32_1:[0-9]+]], v[[B_F32_1]], v[[A_F32_1]]
; SI-DAG: v_cvt_f16_f32_e32 v[[R_F16_0:[0-9]+]], v[[R_F32_0]]
; SI-DAG: v_cvt_f16_f32_e32 v[[R_F16_1:[0-9]+]], v[[R_F32_1]]
; SI: v_lshlrev_b32_e32 v[[R_F16_HI:[0-9]+]], 16, v[[R_F16_1]]
@@ -88,15 +88,18 @@
<2 x half> addrspace(1)* %a,
<2 x half> addrspace(1)* %b) {
entry:
- %a.val = load <2 x half>, <2 x half> addrspace(1)* %a
- %b.val = load <2 x half>, <2 x half> addrspace(1)* %b
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep.a = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %a, i32 %tid
+ %gep.b = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %b, i32 %tid
+ %a.val = load <2 x half>, <2 x half> addrspace(1)* %gep.a
+ %b.val = load <2 x half>, <2 x half> addrspace(1)* %gep.b
%r.val = fadd <2 x half> %a.val, %b.val
store <2 x half> %r.val, <2 x half> addrspace(1)* %r
ret void
}
; GCN-LABEL: {{^}}fadd_v2f16_imm_a:
-; GCN-DAG: buffer_load_dword v[[B_V2_F16:[0-9]+]]
+; GCN-DAG: {{buffer|flat}}_load_dword v[[B_V2_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[B_F32_0:[0-9]+]], v[[B_V2_F16]]
; SI: v_lshrrev_b32_e32 v[[B_F16_1:[0-9]+]], 16, v[[B_V2_F16]]
; SI: v_cvt_f32_f16_e32 v[[B_F32_1:[0-9]+]], v[[B_F16_1]]
@@ -118,14 +121,16 @@
<2 x half> addrspace(1)* %r,
<2 x half> addrspace(1)* %b) {
entry:
- %b.val = load <2 x half>, <2 x half> addrspace(1)* %b
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep.b = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %b, i32 %tid
+ %b.val = load <2 x half>, <2 x half> addrspace(1)* %gep.b
%r.val = fadd <2 x half> <half 1.0, half 2.0>, %b.val
store <2 x half> %r.val, <2 x half> addrspace(1)* %r
ret void
}
; GCN-LABEL: {{^}}fadd_v2f16_imm_b:
-; GCN-DAG: buffer_load_dword v[[A_V2_F16:[0-9]+]]
+; GCN-DAG: {{buffer|flat}}_load_dword v[[A_V2_F16:[0-9]+]]
; SI: v_cvt_f32_f16_e32 v[[A_F32_0:[0-9]+]], v[[A_V2_F16]]
; SI-DAG: v_lshrrev_b32_e32 v[[A_F16_1:[0-9]+]], 16, v[[A_V2_F16]]
; SI: v_cvt_f32_f16_e32 v[[A_F32_1:[0-9]+]], v[[A_F16_1]]
@@ -147,8 +152,15 @@
<2 x half> addrspace(1)* %r,
<2 x half> addrspace(1)* %a) {
entry:
- %a.val = load <2 x half>, <2 x half> addrspace(1)* %a
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep.a = getelementptr inbounds <2 x half>, <2 x half> addrspace(1)* %a, i32 %tid
+ %a.val = load <2 x half>, <2 x half> addrspace(1)* %gep.a
%r.val = fadd <2 x half> %a.val, <half 2.0, half 1.0>
store <2 x half> %r.val, <2 x half> addrspace(1)* %r
ret void
}
+
+declare i32 @llvm.amdgcn.workitem.id.x() #1
+
+attributes #0 = { nounwind }
+attributes #1 = { nounwind readnone }
diff --git a/llvm/test/CodeGen/AMDGPU/fadd64.ll b/llvm/test/CodeGen/AMDGPU/fadd64.ll
index c936d98..8fd1f52 100644
--- a/llvm/test/CodeGen/AMDGPU/fadd64.ll
+++ b/llvm/test/CodeGen/AMDGPU/fadd64.ll
@@ -5,8 +5,11 @@
; CHECK: v_add_f64 {{v[[0-9]+:[0-9]+]}}, {{v[[0-9]+:[0-9]+]}}, {{v[[0-9]+:[0-9]+]}}
define amdgpu_kernel void @v_fadd_f64(double addrspace(1)* %out, double addrspace(1)* %in1,
double addrspace(1)* %in2) {
- %r0 = load double, double addrspace(1)* %in1
- %r1 = load double, double addrspace(1)* %in2
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep1 = getelementptr inbounds double, double addrspace(1)* %in1, i32 %tid
+ %gep2 = getelementptr inbounds double, double addrspace(1)* %in2, i32 %tid
+ %r0 = load double, double addrspace(1)* %gep1
+ %r1 = load double, double addrspace(1)* %gep2
%r2 = fadd double %r0, %r1
store double %r2, double addrspace(1)* %out
ret void
@@ -42,3 +45,8 @@
store <2 x double> %r2, <2 x double> addrspace(1)* %out
ret void
}
+
+declare i32 @llvm.amdgcn.workitem.id.x() #1
+
+attributes #0 = { nounwind }
+attributes #1 = { nounwind readnone }
diff --git a/llvm/test/CodeGen/AMDGPU/fcanonicalize.f16.ll b/llvm/test/CodeGen/AMDGPU/fcanonicalize.f16.ll
index 404358f..dd8e277 100644
--- a/llvm/test/CodeGen/AMDGPU/fcanonicalize.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fcanonicalize.f16.ll
@@ -5,6 +5,8 @@
declare half @llvm.canonicalize.f16(half) #0
declare <2 x half> @llvm.fabs.v2f16(<2 x half>) #0
declare <2 x half> @llvm.canonicalize.v2f16(<2 x half>) #0
+declare i32 @llvm.amdgcn.workitem.id.x() #0
+
; GCN-LABEL: {{^}}v_test_canonicalize_var_f16:
; GCN: v_mul_f16_e32 [[REG:v[0-9]+]], 1.0, {{v[0-9]+}}
@@ -213,7 +215,9 @@
; GFX9: v_pk_mul_f16 [[REG:v[0-9]+]], 1.0, {{v[0-9]+$}}
; GFX9: buffer_store_dword [[REG]]
define amdgpu_kernel void @v_test_canonicalize_var_v2f16(<2 x half> addrspace(1)* %out) #1 {
- %val = load <2 x half>, <2 x half> addrspace(1)* %out
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <2 x half>, <2 x half> addrspace(1)* %out, i32 %tid
+ %val = load <2 x half>, <2 x half> addrspace(1)* %gep
%canonicalized = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> %val)
store <2 x half> %canonicalized, <2 x half> addrspace(1)* %out
ret void
@@ -233,7 +237,9 @@
; GFX9: v_pk_mul_f16 [[REG:v[0-9]+]], 1.0, [[ABS]]{{$}}
; GCN: buffer_store_dword
define amdgpu_kernel void @v_test_canonicalize_fabs_var_v2f16(<2 x half> addrspace(1)* %out) #1 {
- %val = load <2 x half>, <2 x half> addrspace(1)* %out
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <2 x half>, <2 x half> addrspace(1)* %out, i32 %tid
+ %val = load <2 x half>, <2 x half> addrspace(1)* %gep
%val.fabs = call <2 x half> @llvm.fabs.v2f16(<2 x half> %val)
%canonicalized = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> %val.fabs)
store <2 x half> %canonicalized, <2 x half> addrspace(1)* %out
@@ -251,7 +257,9 @@
; GFX9: v_pk_mul_f16 [[REG:v[0-9]+]], 1.0, [[ABS]] neg_lo:[0,1] neg_hi:[0,1]{{$}}
; GCN: buffer_store_dword
define amdgpu_kernel void @v_test_canonicalize_fneg_fabs_var_v2f16(<2 x half> addrspace(1)* %out) #1 {
- %val = load <2 x half>, <2 x half> addrspace(1)* %out
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <2 x half>, <2 x half> addrspace(1)* %out, i32 %tid
+ %val = load <2 x half>, <2 x half> addrspace(1)* %gep
%val.fabs = call <2 x half> @llvm.fabs.v2f16(<2 x half> %val)
%val.fabs.fneg = fsub <2 x half> <half -0.0, half -0.0>, %val.fabs
%canonicalized = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> %val.fabs.fneg)
@@ -270,7 +278,9 @@
; GFX9: v_pk_mul_f16 [[REG:v[0-9]+]], 1.0, {{v[0-9]+}} neg_lo:[0,1] neg_hi:[0,1]{{$}}
; GFX9: buffer_store_dword [[REG]]
define amdgpu_kernel void @v_test_canonicalize_fneg_var_v2f16(<2 x half> addrspace(1)* %out) #1 {
- %val = load <2 x half>, <2 x half> addrspace(1)* %out
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr <2 x half>, <2 x half> addrspace(1)* %out, i32 %tid
+ %val = load <2 x half>, <2 x half> addrspace(1)* %gep
%fneg.val = fsub <2 x half> <half -0.0, half -0.0>, %val
%canonicalized = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> %fneg.val)
store <2 x half> %canonicalized, <2 x half> addrspace(1)* %out
diff --git a/llvm/test/CodeGen/AMDGPU/fcanonicalize.ll b/llvm/test/CodeGen/AMDGPU/fcanonicalize.ll
index 8c385f4..feb4c7b 100644
--- a/llvm/test/CodeGen/AMDGPU/fcanonicalize.ll
+++ b/llvm/test/CodeGen/AMDGPU/fcanonicalize.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
declare float @llvm.fabs.f32(float) #0
declare float @llvm.canonicalize.f32(float) #0
diff --git a/llvm/test/CodeGen/AMDGPU/fcmp.f16.ll b/llvm/test/CodeGen/AMDGPU/fcmp.f16.ll
index 7916226..13d5f7e 100644
--- a/llvm/test/CodeGen/AMDGPU/fcmp.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fcmp.f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}fcmp_f16_lt
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
diff --git a/llvm/test/CodeGen/AMDGPU/fcmp64.ll b/llvm/test/CodeGen/AMDGPU/fcmp64.ll
index b9e1921..95f7e0b 100644
--- a/llvm/test/CodeGen/AMDGPU/fcmp64.ll
+++ b/llvm/test/CodeGen/AMDGPU/fcmp64.ll
@@ -2,7 +2,7 @@
; RUN: llc < %s -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs | FileCheck %s
; CHECK-LABEL: {{^}}flt_f64:
-; CHECK: v_cmp_nge_f64_e32 vcc, {{v[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}}
+; CHECK: v_cmp_nge_f64_e32 vcc, {{s[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}}
define amdgpu_kernel void @flt_f64(i32 addrspace(1)* %out, double addrspace(1)* %in1,
double addrspace(1)* %in2) {
%r0 = load double, double addrspace(1)* %in1
@@ -14,7 +14,7 @@
}
; CHECK-LABEL: {{^}}fle_f64:
-; CHECK: v_cmp_ngt_f64_e32 vcc, {{v[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}}
+; CHECK: v_cmp_ngt_f64_e32 vcc, {{s[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}}
define amdgpu_kernel void @fle_f64(i32 addrspace(1)* %out, double addrspace(1)* %in1,
double addrspace(1)* %in2) {
%r0 = load double, double addrspace(1)* %in1
@@ -26,7 +26,7 @@
}
; CHECK-LABEL: {{^}}fgt_f64:
-; CHECK: v_cmp_nle_f64_e32 vcc, {{v[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}}
+; CHECK: v_cmp_nle_f64_e32 vcc, {{s[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}}
define amdgpu_kernel void @fgt_f64(i32 addrspace(1)* %out, double addrspace(1)* %in1,
double addrspace(1)* %in2) {
%r0 = load double, double addrspace(1)* %in1
@@ -38,7 +38,7 @@
}
; CHECK-LABEL: {{^}}fge_f64:
-; CHECK: v_cmp_nlt_f64_e32 vcc, {{v[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}}
+; CHECK: v_cmp_nlt_f64_e32 vcc, {{s[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}}
define amdgpu_kernel void @fge_f64(i32 addrspace(1)* %out, double addrspace(1)* %in1,
double addrspace(1)* %in2) {
%r0 = load double, double addrspace(1)* %in1
@@ -50,7 +50,7 @@
}
; CHECK-LABEL: {{^}}fne_f64:
-; CHECK: v_cmp_neq_f64_e32 vcc, {{v[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}}
+; CHECK: v_cmp_neq_f64_e32 vcc, {{s[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}}
define amdgpu_kernel void @fne_f64(double addrspace(1)* %out, double addrspace(1)* %in1,
double addrspace(1)* %in2) {
%r0 = load double, double addrspace(1)* %in1
@@ -62,7 +62,7 @@
}
; CHECK-LABEL: {{^}}feq_f64:
-; CHECK: v_cmp_nlg_f64_e32 vcc, {{v[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}}
+; CHECK: v_cmp_nlg_f64_e32 vcc, {{s[[0-9]+:[0-9]+], v[[0-9]+:[0-9]+]}}
define amdgpu_kernel void @feq_f64(double addrspace(1)* %out, double addrspace(1)* %in1,
double addrspace(1)* %in2) {
%r0 = load double, double addrspace(1)* %in1
diff --git a/llvm/test/CodeGen/AMDGPU/fconst64.ll b/llvm/test/CodeGen/AMDGPU/fconst64.ll
index 1255977..ca313d8 100644
--- a/llvm/test/CodeGen/AMDGPU/fconst64.ll
+++ b/llvm/test/CodeGen/AMDGPU/fconst64.ll
@@ -6,8 +6,15 @@
; CHECK-DAG: s_mov_b32 {{s[0-9]+}}, 0
define amdgpu_kernel void @fconst_f64(double addrspace(1)* %out, double addrspace(1)* %in) {
- %r1 = load double, double addrspace(1)* %in
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %gep = getelementptr inbounds double, double addrspace(1)* %in, i32 %tid
+ %r1 = load double, double addrspace(1)* %gep
%r2 = fadd double %r1, 5.000000e+00
store double %r2, double addrspace(1)* %out
ret void
}
+
+declare i32 @llvm.amdgcn.workitem.id.x() #1
+
+attributes #0 = { nounwind }
+attributes #1 = { nounwind readnone }
diff --git a/llvm/test/CodeGen/AMDGPU/fcopysign.f16.ll b/llvm/test/CodeGen/AMDGPU/fcopysign.f16.ll
index 4e2bf76..8e98424 100644
--- a/llvm/test/CodeGen/AMDGPU/fcopysign.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fcopysign.f16.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=GFX89 -check-prefix=GFX8 %s
-; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=GFX89 -check-prefix=GFX9 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=GFX89 -check-prefix=GFX8 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=GFX89 -check-prefix=GFX9 %s
declare half @llvm.copysign.f16(half, half)
declare float @llvm.copysign.f32(float, float)
@@ -9,16 +9,18 @@
declare <3 x half> @llvm.copysign.v3f16(<3 x half>, <3 x half>)
declare <4 x half> @llvm.copysign.v4f16(<4 x half>, <4 x half>)
+declare i32 @llvm.amdgcn.workitem.id.x()
+
; GCN-LABEL: {{^}}test_copysign_f16:
-; SI: buffer_load_ushort v[[SIGN:[0-9]+]]
-; SI: buffer_load_ushort v[[MAG:[0-9]+]]
+; SI: {{buffer|flat}}_load_ushort v[[SIGN:[0-9]+]]
+; SI: {{buffer|flat}}_load_ushort v[[MAG:[0-9]+]]
; SI: s_brev_b32 s[[CONST:[0-9]+]], -2
; SI-DAG: v_cvt_f32_f16_e32 v[[MAG_F32:[0-9]+]], v[[MAG]]
; SI-DAG: v_cvt_f32_f16_e32 v[[SIGN_F32:[0-9]+]], v[[SIGN]]
; SI: v_bfi_b32 v[[OUT_F32:[0-9]+]], s[[CONST]], v[[MAG_F32]], v[[SIGN_F32]]
; SI: v_cvt_f16_f32_e32 v[[OUT:[0-9]+]], v[[OUT_F32]]
-; GFX89: buffer_load_ushort v[[SIGN:[0-9]+]]
-; GFX89: buffer_load_ushort v[[MAG:[0-9]+]]
+; GFX89: {{buffer|flat}}_load_ushort v[[SIGN:[0-9]+]]
+; GFX89: {{buffer|flat}}_load_ushort v[[MAG:[0-9]+]]
; GFX89: s_movk_i32 s[[CONST:[0-9]+]], 0x7fff
; GFX89: v_bfi_b32 v[[OUT:[0-9]+]], s[[CONST]], v[[MAG]], v[[SIGN]]
; GCN: buffer_store_short v[[OUT]]
@@ -36,8 +38,8 @@
}
; GCN-LABEL: {{^}}test_copysign_out_f32_mag_f16_sign_f32:
-; GCN-DAG: buffer_load_ushort v[[MAG:[0-9]+]]
-; GCN-DAG: buffer_load_dword v[[SIGN:[0-9]+]]
+; GCN-DAG: {{buffer|flat}}_load_ushort v[[MAG:[0-9]+]]
+; GCN-DAG: {{buffer|flat}}_load_dword v[[SIGN:[0-9]+]]
; GCN-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2
; GCN-DAG: v_cvt_f32_f16_e32 v[[MAG_EXT:[0-9]+]], v[[MAG]]
; GCN: v_bfi_b32 v[[OUT:[0-9]+]], s[[CONST]], v[[MAG_EXT]], v[[SIGN]]
@@ -48,17 +50,20 @@
half addrspace(1)* %arg_mag,
float addrspace(1)* %arg_sign) {
entry:
- %mag = load half, half addrspace(1)* %arg_mag
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %arg_mag_gep = getelementptr half, half addrspace(1)* %arg_mag, i32 %tid
+ %mag = load half, half addrspace(1)* %arg_mag_gep
%mag.ext = fpext half %mag to float
- %sign = load float, float addrspace(1)* %arg_sign
+ %arg_sign_gep = getelementptr float, float addrspace(1)* %arg_sign, i32 %tid
+ %sign = load float, float addrspace(1)* %arg_sign_gep
%out = call float @llvm.copysign.f32(float %mag.ext, float %sign)
store float %out, float addrspace(1)* %arg_out
ret void
}
; GCN-LABEL: {{^}}test_copysign_out_f64_mag_f16_sign_f64:
-; GCN-DAG: buffer_load_ushort v[[MAG:[0-9]+]]
-; GCN-DAG: buffer_load_dwordx2 v{{\[}}[[SIGN_LO:[0-9]+]]:[[SIGN_HI:[0-9]+]]{{\]}}
+; GCN-DAG: {{buffer|flat}}_load_ushort v[[MAG:[0-9]+]]
+; GCN-DAG: {{buffer|flat}}_load_dwordx2 v{{\[}}[[SIGN_LO:[0-9]+]]:[[SIGN_HI:[0-9]+]]{{\]}}
; GCN-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2
; GCN-DAG: v_cvt_f32_f16_e32 v[[MAG_EXT:[0-9]+]], v[[MAG]]
; GCN-DAG: v_cvt_f64_f32_e32 v{{\[}}[[MAG_EXT_LO:[0-9]+]]:[[MAG_EXT_HI:[0-9]+]]{{\]}}, v[[MAG_EXT]]
@@ -70,17 +75,20 @@
half addrspace(1)* %arg_mag,
double addrspace(1)* %arg_sign) {
entry:
- %mag = load half, half addrspace(1)* %arg_mag
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %arg_mag_gep = getelementptr half, half addrspace(1)* %arg_mag, i32 %tid
+ %mag = load half, half addrspace(1)* %arg_mag_gep
%mag.ext = fpext half %mag to double
- %sign = load double, double addrspace(1)* %arg_sign
+ %arg_sign_gep = getelementptr double, double addrspace(1)* %arg_sign, i32 %tid
+ %sign = load double, double addrspace(1)* %arg_sign_gep
%out = call double @llvm.copysign.f64(double %mag.ext, double %sign)
store double %out, double addrspace(1)* %arg_out
ret void
}
; GCN-LABEL: {{^}}test_copysign_out_f32_mag_f32_sign_f16:
-; GCN-DAG: buffer_load_dword v[[MAG:[0-9]+]]
-; GCN-DAG: buffer_load_ushort v[[SIGN:[0-9]+]]
+; GCN-DAG: {{buffer|flat}}_load_dword v[[MAG:[0-9]+]]
+; GCN-DAG: {{buffer|flat}}_load_ushort v[[SIGN:[0-9]+]]
; GCN-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2
; SI-DAG: v_cvt_f32_f16_e32 v[[SIGN_F32:[0-9]+]], v[[SIGN]]
; SI: v_bfi_b32 v[[OUT:[0-9]+]], s[[CONST]], v[[MAG]], v[[SIGN_F32]]
@@ -93,8 +101,11 @@
float addrspace(1)* %arg_mag,
half addrspace(1)* %arg_sign) {
entry:
- %mag = load float, float addrspace(1)* %arg_mag
- %sign = load half, half addrspace(1)* %arg_sign
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %arg_mag_gep = getelementptr float, float addrspace(1)* %arg_mag, i32 %tid
+ %mag = load float, float addrspace(1)* %arg_mag_gep
+ %arg_sign_gep = getelementptr half, half addrspace(1)* %arg_sign, i32 %tid
+ %sign = load half, half addrspace(1)* %arg_sign_gep
%sign.ext = fpext half %sign to float
%out = call float @llvm.copysign.f32(float %mag, float %sign.ext)
store float %out, float addrspace(1)* %arg_out
@@ -102,8 +113,8 @@
}
; GCN-LABEL: {{^}}test_copysign_out_f64_mag_f64_sign_f16:
-; GCN-DAG: buffer_load_dwordx2 v{{\[}}[[MAG_LO:[0-9]+]]:[[MAG_HI:[0-9]+]]{{\]}}
-; GCN-DAG: buffer_load_ushort v[[SIGN:[0-9]+]]
+; GCN-DAG: {{buffer|flat}}_load_dwordx2 v{{\[}}[[MAG_LO:[0-9]+]]:[[MAG_HI:[0-9]+]]{{\]}}
+; GCN-DAG: {{buffer|flat}}_load_ushort v[[SIGN:[0-9]+]]
; GCN-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2
; SI-DAG: v_cvt_f32_f16_e32 v[[SIGN_F32:[0-9]+]], v[[SIGN]]
; SI: v_bfi_b32 v[[OUT_HI:[0-9]+]], s[[CONST]], v[[MAG_HI]], v[[SIGN_F32]]
@@ -116,8 +127,11 @@
double addrspace(1)* %arg_mag,
half addrspace(1)* %arg_sign) {
entry:
- %mag = load double, double addrspace(1)* %arg_mag
- %sign = load half, half addrspace(1)* %arg_sign
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %arg_mag_gep = getelementptr double, double addrspace(1)* %arg_mag, i32 %tid
+ %mag = load double, double addrspace(1)* %arg_mag_gep
+ %arg_sign_gep = getelementptr half, half addrspace(1)* %arg_sign, i32 %tid
+ %sign = load half, half addrspace(1)* %arg_sign_gep
%sign.ext = fpext half %sign to double
%out = call double @llvm.copysign.f64(double %mag, double %sign.ext)
store double %out, double addrspace(1)* %arg_out
@@ -125,8 +139,8 @@
}
; GCN-LABEL: {{^}}test_copysign_out_f16_mag_f16_sign_f32:
-; GCN-DAG: buffer_load_ushort v[[MAG:[0-9]+]]
-; GCN-DAG: buffer_load_dword v[[SIGN:[0-9]+]]
+; GCN-DAG: {{buffer|flat}}_load_ushort v[[MAG:[0-9]+]]
+; GCN-DAG: {{buffer|flat}}_load_dword v[[SIGN:[0-9]+]]
; SI-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2
; SI-DAG: v_cvt_f32_f16_e32 v[[MAG_F32:[0-9]+]], v[[MAG]]
; SI: v_bfi_b32 v[[OUT_F32:[0-9]+]], s[[CONST]], v[[MAG_F32]], v[[SIGN]]
@@ -141,8 +155,11 @@
half addrspace(1)* %arg_mag,
float addrspace(1)* %arg_sign) {
entry:
- %mag = load half, half addrspace(1)* %arg_mag
- %sign = load float, float addrspace(1)* %arg_sign
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %arg_mag_gep = getelementptr half, half addrspace(1)* %arg_mag, i32 %tid
+ %mag = load half, half addrspace(1)* %arg_mag_gep
+ %arg_sign_gep = getelementptr float, float addrspace(1)* %arg_sign, i32 %tid
+ %sign = load float, float addrspace(1)* %arg_sign_gep
%sign.trunc = fptrunc float %sign to half
%out = call half @llvm.copysign.f16(half %mag, half %sign.trunc)
store half %out, half addrspace(1)* %arg_out
@@ -150,8 +167,8 @@
}
; GCN-LABEL: {{^}}test_copysign_out_f16_mag_f16_sign_f64:
-; GCN-DAG: buffer_load_ushort v[[MAG:[0-9]+]]
-; GCN-DAG: buffer_load_dwordx2 v{{\[}}[[SIGN_LO:[0-9]+]]:[[SIGN_HI:[0-9]+]]{{\]}}
+; GCN-DAG: {{buffer|flat}}_load_ushort v[[MAG:[0-9]+]]
+; GCN-DAG: {{buffer|flat}}_load_dwordx2 v{{\[}}[[SIGN_LO:[0-9]+]]:[[SIGN_HI:[0-9]+]]{{\]}}
; SI-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2
; SI-DAG: v_cvt_f32_f16_e32 v[[MAG_F32:[0-9]+]], v[[MAG]]
; SI: v_bfi_b32 v[[OUT_F32:[0-9]+]], s[[CONST]], v[[MAG_F32]], v[[SIGN_HI]]
@@ -166,8 +183,11 @@
half addrspace(1)* %arg_mag,
double addrspace(1)* %arg_sign) {
entry:
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %arg_mag_gep = getelementptr half, half addrspace(1)* %arg_mag, i32 %tid
%mag = load half, half addrspace(1)* %arg_mag
- %sign = load double, double addrspace(1)* %arg_sign
+ %arg_sign_gep = getelementptr double, double addrspace(1)* %arg_sign, i32 %tid
+ %sign = load double, double addrspace(1)* %arg_sign_gep
%sign.trunc = fptrunc double %sign to half
%out = call half @llvm.copysign.f16(half %mag, half %sign.trunc)
store half %out, half addrspace(1)* %arg_out
@@ -175,8 +195,8 @@
}
; GCN-LABEL: {{^}}test_copysign_out_f16_mag_f32_sign_f16:
-; GCN-DAG: buffer_load_dword v[[MAG:[0-9]+]]
-; GCN-DAG: buffer_load_ushort v[[SIGN:[0-9]+]]
+; GCN-DAG: {{buffer|flat}}_load_dword v[[MAG:[0-9]+]]
+; GCN-DAG: {{buffer|flat}}_load_ushort v[[SIGN:[0-9]+]]
; SI-DAG: s_brev_b32 s[[CONST:[0-9]+]], -2
; SI-DAG: v_cvt_f16_f32_e32 v[[MAG_TRUNC:[0-9]+]], v[[MAG]]
; SI-DAG: v_cvt_f32_f16_e32 v[[SIGN_F32:[0-9]+]], v[[SIGN]]
@@ -193,9 +213,12 @@
float addrspace(1)* %arg_mag,
half addrspace(1)* %arg_sign) {
entry:
- %mag = load float, float addrspace(1)* %arg_mag
+ %tid = call i32 @llvm.amdgcn.workitem.id.x()
+ %arg_mag_gep = getelementptr float, float addrspace(1)* %arg_mag, i32 %tid
+ %mag = load float, float addrspace(1)* %arg_mag_gep
%mag.trunc = fptrunc float %mag to half
- %sign = load half, half addrspace(1)* %arg_sign
+ %arg_sign_gep = getelementptr half, half addrspace(1)* %arg_sign, i32 %tid
+ %sign = load half, half addrspace(1)* %arg_sign_gep
%out = call half @llvm.copysign.f16(half %mag.trunc, half %sign)
store half %out, half addrspace(1)* %arg_out
ret void
diff --git a/llvm/test/CodeGen/AMDGPU/fma-combine.ll b/llvm/test/CodeGen/AMDGPU/fma-combine.ll
index 4113ba8..ba6b53b 100644
--- a/llvm/test/CodeGen/AMDGPU/fma-combine.ll
+++ b/llvm/test/CodeGen/AMDGPU/fma-combine.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -fp-contract=fast < %s | FileCheck -check-prefix=SI-NOFMA -check-prefix=SI-SAFE -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs -fp-contract=fast < %s | FileCheck -check-prefix=SI-NOFMA -check-prefix=SI-SAFE -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs -fp-contract=fast -enable-no-infs-fp-math -enable-unsafe-fp-math -mattr=+fp32-denormals < %s | FileCheck -check-prefix=SI-FMA -check-prefix=SI-UNSAFE -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs -fp-contract=fast < %s | FileCheck -check-prefix=SI-NOFMA -check-prefix=SI-SAFE -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs -fp-contract=fast < %s | FileCheck -check-prefix=SI-NOFMA -check-prefix=SI-SAFE -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs -fp-contract=fast -enable-no-infs-fp-math -enable-unsafe-fp-math -mattr=+fp32-denormals < %s | FileCheck -check-prefix=SI-FMA -check-prefix=SI-UNSAFE -check-prefix=SI -check-prefix=FUNC %s
; Note: The SI-FMA conversions of type x * (y + 1) --> x * y + x would be
; beneficial even without fp32 denormals, but they do require no-infs-fp-math
diff --git a/llvm/test/CodeGen/AMDGPU/fma.f64.ll b/llvm/test/CodeGen/AMDGPU/fma.f64.ll
index 4d3f371..907121f 100644
--- a/llvm/test/CodeGen/AMDGPU/fma.f64.ll
+++ b/llvm/test/CodeGen/AMDGPU/fma.f64.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
declare double @llvm.fma.f64(double, double, double) nounwind readnone
declare <2 x double> @llvm.fma.v2f64(<2 x double>, <2 x double>, <2 x double>) nounwind readnone
diff --git a/llvm/test/CodeGen/AMDGPU/fma.ll b/llvm/test/CodeGen/AMDGPU/fma.ll
index 659cecb..6be4c45 100644
--- a/llvm/test/CodeGen/AMDGPU/fma.ll
+++ b/llvm/test/CodeGen/AMDGPU/fma.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
declare float @llvm.fma.f32(float, float, float) nounwind readnone
declare <2 x float> @llvm.fma.v2f32(<2 x float>, <2 x float>, <2 x float>) nounwind readnone
diff --git a/llvm/test/CodeGen/AMDGPU/fmul.f16.ll b/llvm/test/CodeGen/AMDGPU/fmul.f16.ll
index cd86409..eaf95a6 100644
--- a/llvm/test/CodeGen/AMDGPU/fmul.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmul.f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}fmul_f16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
diff --git a/llvm/test/CodeGen/AMDGPU/fmul64.ll b/llvm/test/CodeGen/AMDGPU/fmul64.ll
index f14233f2..d37d432 100644
--- a/llvm/test/CodeGen/AMDGPU/fmul64.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmul64.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=FUNC -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=FUNC -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=FUNC -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=FUNC -check-prefix=SI %s
; FUNC-LABEL: {{^}}fmul_f64:
; SI: v_mul_f64 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\]}}
diff --git a/llvm/test/CodeGen/AMDGPU/fmuladd.f32.ll b/llvm/test/CodeGen/AMDGPU/fmuladd.f32.ll
index e422550..58c3d7f 100644
--- a/llvm/test/CodeGen/AMDGPU/fmuladd.f32.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmuladd.f32.ll
@@ -1,12 +1,12 @@
-; RUN: llc -verify-machineinstrs -mcpu=tahiti -mattr=-fp32-denormals,+fast-fmaf -fp-contract=on < %s | FileCheck -check-prefixes=GCN,GCN-FLUSH-STRICT,GCN-FLUSH,SI-FLUSH,GCN-FLUSH-FASTFMA,GCN-FLUSH-FASTFMA-STRICT,SI %s
-; RUN: llc -verify-machineinstrs -mcpu=tahiti -mattr=+fp32-denormals,+fast-fmaf -fp-contract=on < %s | FileCheck -check-prefixes=GCN,GCN-DENORM-STRICT,GCN-DENORM,SI-DENORM,GCN-DENORM-FASTFMA,GCN-DENORM-FASTFMA-STRICT,SI %s
-; RUN: llc -verify-machineinstrs -mcpu=verde -mattr=-fp32-denormals,-fast-fmaf -fp-contract=on < %s | FileCheck -check-prefixes=GCN,GCN-FLUSH-STRICT,GCN-FLUSH,SI-FLUSH,GCN-FLUSH-SLOWFMA,GCN-FLUSH-SLOWFMA-STRICT,SI %s
-; RUN: llc -verify-machineinstrs -mcpu=verde -mattr=+fp32-denormals,-fast-fmaf -fp-contract=on < %s | FileCheck -check-prefixes=GCN,GCN-DENORM-STRICT,GCN-DENORM,SI-DENORM,GCN-DENORM-SLOWFMA,GCN-DENORM-SLOWFMA-STRICT,SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -mcpu=tahiti -mattr=-fp32-denormals,+fast-fmaf -fp-contract=on < %s | FileCheck -check-prefixes=GCN,GCN-FLUSH-STRICT,GCN-FLUSH,SI-FLUSH,GCN-FLUSH-FASTFMA,GCN-FLUSH-FASTFMA-STRICT,SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -mcpu=tahiti -mattr=+fp32-denormals,+fast-fmaf -fp-contract=on < %s | FileCheck -check-prefixes=GCN,GCN-DENORM-STRICT,GCN-DENORM,SI-DENORM,GCN-DENORM-FASTFMA,GCN-DENORM-FASTFMA-STRICT,SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -mcpu=verde -mattr=-fp32-denormals,-fast-fmaf -fp-contract=on < %s | FileCheck -check-prefixes=GCN,GCN-FLUSH-STRICT,GCN-FLUSH,SI-FLUSH,GCN-FLUSH-SLOWFMA,GCN-FLUSH-SLOWFMA-STRICT,SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -mcpu=verde -mattr=+fp32-denormals,-fast-fmaf -fp-contract=on < %s | FileCheck -check-prefixes=GCN,GCN-DENORM-STRICT,GCN-DENORM,SI-DENORM,GCN-DENORM-SLOWFMA,GCN-DENORM-SLOWFMA-STRICT,SI %s
-; RUN: llc -verify-machineinstrs -mcpu=tahiti -mattr=-fp32-denormals,+fast-fmaf -fp-contract=fast < %s | FileCheck -check-prefixes=GCN,GCN-FLUSH-CONTRACT,GCN-FLUSH,SI-FLUSH,GCN-FLUSH-FASTFMA,GCN-FLUSH-FASTFMA-CONTRACT,SI %s
-; RUN: llc -verify-machineinstrs -mcpu=tahiti -mattr=+fp32-denormals,+fast-fmaf -fp-contract=fast < %s | FileCheck -check-prefixes=GCN,GCN-DENORM-CONTRACT,GCN-DENORM,SI-DENORM,GCN-DENORM-FASTFMA,GCN-DENORM-FASTFMA-CONTRACT,SI %s
-; RUN: llc -verify-machineinstrs -mcpu=verde -mattr=-fp32-denormals,-fast-fmaf -fp-contract=fast < %s | FileCheck -check-prefixes=GCN,GCN-FLUSH-CONTRACT,GCN-FLUSH,SI-FLUSH,GCN-FLUSH-SLOWFMA,GCN-FLUSH-SLOWFMA-CONTRACT,SI %s
-; RUN: llc -verify-machineinstrs -mcpu=verde -mattr=+fp32-denormals,-fast-fmaf -fp-contract=fast < %s | FileCheck -check-prefixes=GCN,GCN-DENORM-CONTRACT,GCN-DENORM,SI-DENORM,GCN-DENORM-SLOWFMA,GCN-DENORM-SLOWFMA-CONTRACT,SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -mcpu=tahiti -mattr=-fp32-denormals,+fast-fmaf -fp-contract=fast < %s | FileCheck -check-prefixes=GCN,GCN-FLUSH-CONTRACT,GCN-FLUSH,SI-FLUSH,GCN-FLUSH-FASTFMA,GCN-FLUSH-FASTFMA-CONTRACT,SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -mcpu=tahiti -mattr=+fp32-denormals,+fast-fmaf -fp-contract=fast < %s | FileCheck -check-prefixes=GCN,GCN-DENORM-CONTRACT,GCN-DENORM,SI-DENORM,GCN-DENORM-FASTFMA,GCN-DENORM-FASTFMA-CONTRACT,SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -mcpu=verde -mattr=-fp32-denormals,-fast-fmaf -fp-contract=fast < %s | FileCheck -check-prefixes=GCN,GCN-FLUSH-CONTRACT,GCN-FLUSH,SI-FLUSH,GCN-FLUSH-SLOWFMA,GCN-FLUSH-SLOWFMA-CONTRACT,SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -mcpu=verde -mattr=+fp32-denormals,-fast-fmaf -fp-contract=fast < %s | FileCheck -check-prefixes=GCN,GCN-DENORM-CONTRACT,GCN-DENORM,SI-DENORM,GCN-DENORM-SLOWFMA,GCN-DENORM-SLOWFMA-CONTRACT,SI %s
; Test all permutations of: fp32 denormals, fast fp contract, fp contract enabled for fmuladd, fmaf fast/slow.
diff --git a/llvm/test/CodeGen/AMDGPU/fmuladd.f64.ll b/llvm/test/CodeGen/AMDGPU/fmuladd.f64.ll
index 86e91e0..8d91a56 100644
--- a/llvm/test/CodeGen/AMDGPU/fmuladd.f64.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmuladd.f64.ll
@@ -1,9 +1,9 @@
-; RUN: llc -march=amdgcn -mcpu=tahiti -fp-contract=on -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-STRICTSI %s
-; RUN: llc -march=amdgcn -mcpu=verde -fp-contract=on -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,SI %s
-; RUN: llc -march=amdgcn -mcpu=tahiti -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,SI %s
-; RUN: llc -march=amdgcn -mcpu=verde -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,SI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -fp-contract=on -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,VI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -fp-contract=on -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-STRICTSI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -fp-contract=on -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -fp-contract=on -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -fp-contract=fast -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,VI %s
; GCN-LABEL: {{^}}fmuladd_f64:
; GCN: v_fma_f64 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\]}}
diff --git a/llvm/test/CodeGen/AMDGPU/fmuladd.v2f16.ll b/llvm/test/CodeGen/AMDGPU/fmuladd.v2f16.ll
index 6246100..b50a26c 100644
--- a/llvm/test/CodeGen/AMDGPU/fmuladd.v2f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmuladd.v2f16.ll
@@ -1,12 +1,12 @@
-; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-fp64-fp16-denormals -fp-contract=on -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,GFX9-FLUSH,GFX9 %s
-; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-fp64-fp16-denormals -fp-contract=on -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,GFX9-FLUSH,GFX9 %s
-; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-fp64-fp16-denormals -fp-contract=fast -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,GFX9-FLUSH,GFX9 %s
-; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-fp64-fp16-denormals -fp-contract=fast -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,GFX9-FLUSH,GFX9 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-fp64-fp16-denormals -fp-contract=on -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,GFX9-FLUSH,GFX9 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-fp64-fp16-denormals -fp-contract=on -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,GFX9-FLUSH,GFX9 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-fp64-fp16-denormals -fp-contract=fast -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,GFX9-FLUSH,GFX9 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-fp64-fp16-denormals -fp-contract=fast -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,GFX9-FLUSH,GFX9 %s
-; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=+fp64-fp16-denormals -fp-contract=on -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,GFX9-DENORM-STRICT,GFX9-DENORM,GFX9 %s
-; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=+fp64-fp16-denormals -fp-contract=on -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,GFX9-DENORM-STRICT,GFX9-DENORM,GFX9 %s
-; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=+fp64-fp16-denormals -fp-contract=fast -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,GFX9-DENORM-CONTRACT,GFX9-DENORM,GFX9 %s
-; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=+fp64-fp16-denormals -fp-contract=fast -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,GFX9-DENORM-CONTRACT,GFX9-DENORM,GFX9 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=+fp64-fp16-denormals -fp-contract=on -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,GFX9-DENORM-STRICT,GFX9-DENORM,GFX9 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=+fp64-fp16-denormals -fp-contract=on -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-STRICT,GFX9-DENORM-STRICT,GFX9-DENORM,GFX9 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=+fp64-fp16-denormals -fp-contract=fast -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,GFX9-DENORM-CONTRACT,GFX9-DENORM,GFX9 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=+fp64-fp16-denormals -fp-contract=fast -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefixes=GCN,GCN-CONTRACT,GFX9-DENORM-CONTRACT,GFX9-DENORM,GFX9 %s
declare i32 @llvm.amdgcn.workitem.id.x() #1
declare <2 x half> @llvm.fmuladd.v2f16(<2 x half>, <2 x half>, <2 x half>) #1
diff --git a/llvm/test/CodeGen/AMDGPU/fneg-fabs.ll b/llvm/test/CodeGen/AMDGPU/fneg-fabs.ll
index 0a7346f..3f20ca7 100644
--- a/llvm/test/CodeGen/AMDGPU/fneg-fabs.ll
+++ b/llvm/test/CodeGen/AMDGPU/fneg-fabs.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}fneg_fabs_fadd_f32:
; SI-NOT: and
diff --git a/llvm/test/CodeGen/AMDGPU/fneg.f16.ll b/llvm/test/CodeGen/AMDGPU/fneg.f16.ll
index 2d94726..25ed208 100644
--- a/llvm/test/CodeGen/AMDGPU/fneg.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fneg.f16.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=kaveri -mtriple=amdgcn--amdhsa -verify-machineinstrs < %s | FileCheck -check-prefix=CI -check-prefix=CIVI -check-prefix=GCN %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mtriple=amdgcn--amdhsa -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=CIVI -check-prefix=GCN -check-prefix=GFX89 %s
-; RUN: llc -march=amdgcn -mcpu=gfx901 -mtriple=amdgcn--amdhsa -verify-machineinstrs < %s | FileCheck -check-prefix=GFX9 -check-prefix=GCN -check-prefix=GFX89 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=kaveri -mtriple=amdgcn--amdhsa -verify-machineinstrs < %s | FileCheck -check-prefix=CI -check-prefix=CIVI -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mtriple=amdgcn--amdhsa -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=CIVI -check-prefix=GCN -check-prefix=GFX89 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mtriple=amdgcn--amdhsa -verify-machineinstrs < %s | FileCheck -check-prefix=GFX9 -check-prefix=GCN -check-prefix=GFX89 %s
; FIXME: Should be able to do scalar op
; GCN-LABEL: {{^}}s_fneg_f16:
diff --git a/llvm/test/CodeGen/AMDGPU/fp32_to_fp16.ll b/llvm/test/CodeGen/AMDGPU/fp32_to_fp16.ll
index 2c6b1cb..579a145 100644
--- a/llvm/test/CodeGen/AMDGPU/fp32_to_fp16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fp32_to_fp16.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
declare i16 @llvm.convert.to.fp16.f32(float) nounwind readnone
diff --git a/llvm/test/CodeGen/AMDGPU/fpext.f16.ll b/llvm/test/CodeGen/AMDGPU/fpext.f16.ll
index 15cc73b..c483fea 100644
--- a/llvm/test/CodeGen/AMDGPU/fpext.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fpext.f16.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SIVI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=GFX89 %s
-; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 -check-prefix=GFX89 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SIVI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=GFX89 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 -check-prefix=GFX89 %s
; GCN-LABEL: {{^}}fpext_f16_to_f32
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
diff --git a/llvm/test/CodeGen/AMDGPU/fptosi.f16.ll b/llvm/test/CodeGen/AMDGPU/fptosi.f16.ll
index f310618..6f23df1 100644
--- a/llvm/test/CodeGen/AMDGPU/fptosi.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fptosi.f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}fptosi_f16_to_i16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
diff --git a/llvm/test/CodeGen/AMDGPU/fptoui.f16.ll b/llvm/test/CodeGen/AMDGPU/fptoui.f16.ll
index 7641c08..335e41f 100644
--- a/llvm/test/CodeGen/AMDGPU/fptoui.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fptoui.f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}fptoui_f16_to_i16
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
diff --git a/llvm/test/CodeGen/AMDGPU/fptrunc.f16.ll b/llvm/test/CodeGen/AMDGPU/fptrunc.f16.ll
index bc72f44..32ef5ed 100644
--- a/llvm/test/CodeGen/AMDGPU/fptrunc.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fptrunc.f16.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SIVI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=SIVI %s
-; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global,-fp64-fp16-denormals -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SIVI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=SIVI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global,-fp64-fp16-denormals -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s
; GCN-LABEL: {{^}}fptrunc_f32_to_f16:
; GCN: buffer_load_dword v[[A_F32:[0-9]+]]
diff --git a/llvm/test/CodeGen/AMDGPU/fract.f64.ll b/llvm/test/CodeGen/AMDGPU/fract.f64.ll
index 9a56cbe..1314dfe 100644
--- a/llvm/test/CodeGen/AMDGPU/fract.f64.ll
+++ b/llvm/test/CodeGen/AMDGPU/fract.f64.ll
@@ -1,9 +1,9 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -enable-unsafe-fp-math -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-UNSAFE -check-prefix=SI-UNSAFE -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -enable-unsafe-fp-math -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-UNSAFE -check-prefix=VI-UNSAFE -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -enable-unsafe-fp-math -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-UNSAFE -check-prefix=SI-UNSAFE -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -enable-unsafe-fp-math -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-UNSAFE -check-prefix=VI-UNSAFE -check-prefix=FUNC %s
declare double @llvm.fabs.f64(double) #0
declare double @llvm.floor.f64(double) #0
diff --git a/llvm/test/CodeGen/AMDGPU/fract.ll b/llvm/test/CodeGen/AMDGPU/fract.ll
index 207fe28..0233165 100644
--- a/llvm/test/CodeGen/AMDGPU/fract.ll
+++ b/llvm/test/CodeGen/AMDGPU/fract.ll
@@ -1,8 +1,8 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-SAFE -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-SAFE -check-prefix=GCN -check-prefix=CI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-SAFE -check-prefix=GCN -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN-UNSAFE -check-prefix=GCN %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN-UNSAFE -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-SAFE -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-SAFE -check-prefix=GCN -check-prefix=CI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN-SAFE -check-prefix=GCN -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN-UNSAFE -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN-UNSAFE -check-prefix=GCN %s
declare float @llvm.fabs.f32(float) #0
declare float @llvm.floor.f32(float) #0
diff --git a/llvm/test/CodeGen/AMDGPU/frem.ll b/llvm/test/CodeGen/AMDGPU/frem.ll
index 9778069..bf63129 100644
--- a/llvm/test/CodeGen/AMDGPU/frem.ll
+++ b/llvm/test/CodeGen/AMDGPU/frem.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=CI -check-prefix=GCN -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=CI -check-prefix=GCN -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=CI -check-prefix=GCN -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=CI -check-prefix=GCN -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}frem_f32:
; GCN-DAG: buffer_load_dword [[X:v[0-9]+]], {{.*$}}
diff --git a/llvm/test/CodeGen/AMDGPU/fsqrt.f64.ll b/llvm/test/CodeGen/AMDGPU/fsqrt.f64.ll
index 453d8fb..186757e 100644
--- a/llvm/test/CodeGen/AMDGPU/fsqrt.f64.ll
+++ b/llvm/test/CodeGen/AMDGPU/fsqrt.f64.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}v_safe_fsqrt_f64:
; GCN: v_sqrt_f64_e32 {{v\[[0-9]+:[0-9]+\], v\[[0-9]+:[0-9]+\]}}
diff --git a/llvm/test/CodeGen/AMDGPU/fsqrt.ll b/llvm/test/CodeGen/AMDGPU/fsqrt.ll
index a0fd341..6bd9a0d 100644
--- a/llvm/test/CodeGen/AMDGPU/fsqrt.ll
+++ b/llvm/test/CodeGen/AMDGPU/fsqrt.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
; Run with unsafe-fp-math to make sure nothing tries to turn this into 1 / rsqrt(x)
diff --git a/llvm/test/CodeGen/AMDGPU/fsub.f16.ll b/llvm/test/CodeGen/AMDGPU/fsub.f16.ll
index fa00c06..4731b06 100644
--- a/llvm/test/CodeGen/AMDGPU/fsub.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/fsub.f16.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SIVI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX89 -check-prefix=VI -check-prefix=SIVI %s
-; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX89 -check-prefix=GFX9 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SIVI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX89 -check-prefix=VI -check-prefix=SIVI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX89 -check-prefix=GFX9 %s
; GCN-LABEL: {{^}}fsub_f16:
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
diff --git a/llvm/test/CodeGen/AMDGPU/fsub.ll b/llvm/test/CodeGen/AMDGPU/fsub.ll
index e7a92d9..153b9ad 100644
--- a/llvm/test/CodeGen/AMDGPU/fsub.ll
+++ b/llvm/test/CodeGen/AMDGPU/fsub.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}v_fsub_f32:
; SI: v_subrev_f32_e32 {{v[0-9]+}}, {{v[0-9]+}}, {{v[0-9]+}}
diff --git a/llvm/test/CodeGen/AMDGPU/fsub64.ll b/llvm/test/CodeGen/AMDGPU/fsub64.ll
index dc33241..73f1a69 100644
--- a/llvm/test/CodeGen/AMDGPU/fsub64.ll
+++ b/llvm/test/CodeGen/AMDGPU/fsub64.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
declare double @llvm.fabs.f64(double) #0
diff --git a/llvm/test/CodeGen/AMDGPU/ftrunc.f64.ll b/llvm/test/CodeGen/AMDGPU/ftrunc.f64.ll
index 1f72ec6..bb2a6ba 100644
--- a/llvm/test/CodeGen/AMDGPU/ftrunc.f64.ll
+++ b/llvm/test/CodeGen/AMDGPU/ftrunc.f64.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=SI < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=bonaire < %s | FileCheck -check-prefix=CI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=CI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=SI < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire < %s | FileCheck -check-prefix=CI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=CI -check-prefix=FUNC %s
declare double @llvm.trunc.f64(double) nounwind readnone
declare <2 x double> @llvm.trunc.v2f64(<2 x double>) nounwind readnone
diff --git a/llvm/test/CodeGen/AMDGPU/global-extload-i16.ll b/llvm/test/CodeGen/AMDGPU/global-extload-i16.ll
index 19e592f..4e50f99 100644
--- a/llvm/test/CodeGen/AMDGPU/global-extload-i16.ll
+++ b/llvm/test/CodeGen/AMDGPU/global-extload-i16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; XUN: llc -march=r600 -mcpu=cypress < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FIXME: cypress is broken because the bigger testcases spill and it's not implemented
diff --git a/llvm/test/CodeGen/AMDGPU/half.ll b/llvm/test/CodeGen/AMDGPU/half.ll
index 41ae5a4..64a495e 100644
--- a/llvm/test/CodeGen/AMDGPU/half.ll
+++ b/llvm/test/CodeGen/AMDGPU/half.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; half args should be promoted to float for SI and lower.
diff --git a/llvm/test/CodeGen/AMDGPU/imm.ll b/llvm/test/CodeGen/AMDGPU/imm.ll
index c2668a0..8cda01a1 100644
--- a/llvm/test/CodeGen/AMDGPU/imm.ll
+++ b/llvm/test/CodeGen/AMDGPU/imm.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; Use a 64-bit value with lo bits that can be represented as an inline constant
; GCN-LABEL: {{^}}i64_imm_inline_lo:
diff --git a/llvm/test/CodeGen/AMDGPU/immv216.ll b/llvm/test/CodeGen/AMDGPU/immv216.ll
index cd3502b..b29e93d 100644
--- a/llvm/test/CodeGen/AMDGPU/immv216.ll
+++ b/llvm/test/CodeGen/AMDGPU/immv216.ll
@@ -1,6 +1,6 @@
-; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s
-; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
-; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=kaveri -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=kaveri -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI %s
; FIXME: Merge into imm.ll
; GCN-LABEL: {{^}}store_inline_imm_neg_0.0_v2i16:
diff --git a/llvm/test/CodeGen/AMDGPU/indirect-addressing-si.ll b/llvm/test/CodeGen/AMDGPU/indirect-addressing-si.ll
index 0d20c32..62200b9 100644
--- a/llvm/test/CodeGen/AMDGPU/indirect-addressing-si.ll
+++ b/llvm/test/CodeGen/AMDGPU/indirect-addressing-si.ll
@@ -1,7 +1,7 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=MOVREL %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=MOVREL %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -amdgpu-vgpr-index-mode -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=IDXMODE %s
-; RUN: llc -march=amdgcn -mcpu=gfx900 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=IDXMODE %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=MOVREL %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=MOVREL %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -amdgpu-vgpr-index-mode -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=IDXMODE %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx900 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=IDXMODE %s
; Tests for indirect addressing on SI, which is implemented using dynamic
; indexing of vectors.
diff --git a/llvm/test/CodeGen/AMDGPU/invariant-load-no-alias-store.ll b/llvm/test/CodeGen/AMDGPU/invariant-load-no-alias-store.ll
index 5cd965d..eea2619 100644
--- a/llvm/test/CodeGen/AMDGPU/invariant-load-no-alias-store.ll
+++ b/llvm/test/CodeGen/AMDGPU/invariant-load-no-alias-store.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=amdgcn -amdgpu-load-store-vectorizer=0 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -amdgpu-load-store-vectorizer=0 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; GatherAllAliases gives up on trying to analyze cases where the
; pointer may have been loaded from an aliased store, so make sure
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.class.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.class.f16.ll
index f08d4b6..06dc2cc 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.class.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.class.f16.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.fabs.f16(half %a)
declare i1 @llvm.amdgcn.class.f16(half %a, i32 %b)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.class.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.class.ll
index 1fcdac5..f71b975 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.class.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.class.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
declare i1 @llvm.amdgcn.class.f32(float, i32) #1
declare i1 @llvm.amdgcn.class.f64(double, i32) #1
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.div.fmas.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.div.fmas.ll
index 2cc63ae..1b3e09a 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.div.fmas.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.div.fmas.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -strict-whitespace -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -strict-whitespace -check-prefix=GCN -check-prefix=SI %s
; XUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -strict-whitespace -check-prefix=GCN -check-prefix=VI %s
; FIXME: Enable for VI.
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ldexp.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ldexp.f16.ll
index fe211d3..7068f45 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ldexp.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ldexp.f16.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.amdgcn.ldexp.f16(half %a, i32 %b)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sbfe.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sbfe.ll
index 593c958..871b8c4 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sbfe.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sbfe.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}bfe_i32_arg_arg_arg:
; GCN: v_bfe_i32
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sffbh.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sffbh.ll
index 495e36b..39370e4 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sffbh.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sffbh.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
declare i32 @llvm.amdgcn.sffbh.i32(i32) #1
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.trig.preop.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.trig.preop.ll
index e0cec21..8468aa3 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.trig.preop.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.trig.preop.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
declare double @llvm.amdgcn.trig.preop.f64(double, i32) nounwind readnone
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ubfe.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ubfe.ll
index 92e3a10..68fd08f 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ubfe.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ubfe.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}bfe_u32_arg_arg_arg:
; GCN: v_bfe_u32
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.ceil.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.ceil.f16.ll
index 0604a49..e6dacfa 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.ceil.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.ceil.f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.ceil.f16(half %a)
declare <2 x half> @llvm.ceil.v2f16(<2 x half> %a)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.cos.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.cos.f16.ll
index d836ea3..0f4f6d9 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.cos.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.cos.f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.cos.f16(half %a)
declare <2 x half> @llvm.cos.v2f16(<2 x half> %a)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.exp2.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.exp2.f16.ll
index 5757142..8a20b9a 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.exp2.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.exp2.f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.exp2.f16(half %a)
declare <2 x half> @llvm.exp2.v2f16(<2 x half> %a)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.floor.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.floor.f16.ll
index 6a18141..a551f67 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.floor.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.floor.f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.floor.f16(half %a)
declare <2 x half> @llvm.floor.v2f16(<2 x half> %a)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.fma.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.fma.f16.ll
index 3f4fba7..da0841d 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.fma.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.fma.f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.fma.f16(half %a, half %b, half %c)
declare <2 x half> @llvm.fma.v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %c)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.fmuladd.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.fmuladd.f16.ll
index 806723e..72cd834 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.fmuladd.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.fmuladd.f16.ll
@@ -1,7 +1,7 @@
-; RUN: llc -march=amdgcn -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SI-FLUSH %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=VI-FLUSH %s
-; RUN: llc -march=amdgcn -mattr=+fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SI-DENORM %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=+fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=VI-DENORM %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SI-FLUSH %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=VI-FLUSH %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mattr=+fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SI-DENORM %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=+fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=VI-DENORM %s
declare half @llvm.fmuladd.f16(half %a, half %b, half %c)
declare <2 x half> @llvm.fmuladd.v2f16(<2 x half> %a, <2 x half> %b, <2 x half> %c)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.log2.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.log2.f16.ll
index 773eb55..7e5553e 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.log2.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.log2.f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.log2.f16(half %a)
declare <2 x half> @llvm.log2.v2f16(<2 x half> %a)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.maxnum.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.maxnum.f16.ll
index 8f4b314..393ee21 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.maxnum.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.maxnum.f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.maxnum.f16(half %a, half %b)
declare <2 x half> @llvm.maxnum.v2f16(<2 x half> %a, <2 x half> %b)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.minnum.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.minnum.f16.ll
index 1a86286..afd05ec 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.minnum.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.minnum.f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.minnum.f16(half %a, half %b)
declare <2 x half> @llvm.minnum.v2f16(<2 x half> %a, <2 x half> %b)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.rint.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.rint.f16.ll
index 30cb969..0747063 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.rint.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.rint.f16.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SIVI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SIVI -check-prefix=VI -check-prefix=GFX89 %s
-; RUN: llc -march=amdgcn -mcpu=gfx900 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX89 -check-prefix=GFX9 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=SIVI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SIVI -check-prefix=VI -check-prefix=GFX89 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx900 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GFX89 -check-prefix=GFX9 %s
declare half @llvm.rint.f16(half %a)
declare <2 x half> @llvm.rint.v2f16(<2 x half> %a)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.sin.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.sin.f16.ll
index eb1f32c..d768f7f 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.sin.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.sin.f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.sin.f16(half %a)
declare <2 x half> @llvm.sin.v2f16(<2 x half> %a)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.sqrt.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.sqrt.f16.ll
index 46ee6526..f7da626 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.sqrt.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.sqrt.f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.sqrt.f16(half %a)
declare <2 x half> @llvm.sqrt.v2f16(<2 x half> %a)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.trunc.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.trunc.f16.ll
index dc7182a..b22defb 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.trunc.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.trunc.f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
declare half @llvm.trunc.f16(half %a)
declare <2 x half> @llvm.trunc.v2f16(<2 x half> %a)
diff --git a/llvm/test/CodeGen/AMDGPU/load-global-f32.ll b/llvm/test/CodeGen/AMDGPU/load-global-f32.ll
index bd6fea5..77557a5 100644
--- a/llvm/test/CodeGen/AMDGPU/load-global-f32.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-global-f32.ll
@@ -1,9 +1,9 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
-; RUN: llc -mtriple=amdgcn-amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn-amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=cayman < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cayman < %s | FileCheck -check-prefix=R600 -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}global_load_f32:
; GCN-NOHSA: buffer_load_dword v{{[0-9]+}}
diff --git a/llvm/test/CodeGen/AMDGPU/load-global-f64.ll b/llvm/test/CodeGen/AMDGPU/load-global-f64.ll
index 5b772e1..84214b7 100644
--- a/llvm/test/CodeGen/AMDGPU/load-global-f64.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-global-f64.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
-; RUN: llc -mtriple=amdgcn-amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn-amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}global_load_f64:
; GCN-NOHSA: buffer_load_dwordx2 [[VAL:v\[[0-9]+:[0-9]+\]]]
diff --git a/llvm/test/CodeGen/AMDGPU/load-global-i16.ll b/llvm/test/CodeGen/AMDGPU/load-global-i16.ll
index e3415b9..cb2495d 100644
--- a/llvm/test/CodeGen/AMDGPU/load-global-i16.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-global-i16.ll
@@ -1,8 +1,8 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-NOHSA,GCN-NOHSA-SI,FUNC %s
-; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-HSA,FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-NOHSA,GCN-NOHSA-VI,FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=EGCM -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=cayman -verify-machineinstrs < %s | FileCheck -check-prefix=CM -check-prefix=EGCM -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-NOHSA,GCN-NOHSA-SI,FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-HSA,FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-NOHSA,GCN-NOHSA-VI,FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=EGCM -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cayman -verify-machineinstrs < %s | FileCheck -check-prefix=CM -check-prefix=EGCM -check-prefix=FUNC %s
; FIXME: r600 is broken because the bigger testcases spill and it's not implemented
diff --git a/llvm/test/CodeGen/AMDGPU/load-global-i32.ll b/llvm/test/CodeGen/AMDGPU/load-global-i32.ll
index 5df32c1..6360d39 100644
--- a/llvm/test/CodeGen/AMDGPU/load-global-i32.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-global-i32.ll
@@ -1,7 +1,7 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
-; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}global_load_i32:
diff --git a/llvm/test/CodeGen/AMDGPU/load-global-i64.ll b/llvm/test/CodeGen/AMDGPU/load-global-i64.ll
index de16b6c..c71db0b 100644
--- a/llvm/test/CodeGen/AMDGPU/load-global-i64.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-global-i64.ll
@@ -1,9 +1,9 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
-; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-HSA -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=cayman < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cayman < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}global_load_i64:
; GCN-NOHSA: buffer_load_dwordx2 [[VAL:v\[[0-9]+:[0-9]+\]]]
diff --git a/llvm/test/CodeGen/AMDGPU/load-global-i8.ll b/llvm/test/CodeGen/AMDGPU/load-global-i8.ll
index fc0cbf9..3fe6bd26 100644
--- a/llvm/test/CodeGen/AMDGPU/load-global-i8.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-global-i8.ll
@@ -1,8 +1,8 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-NOHSA,SI,FUNC %s
-; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-HSA,SI,FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-NOHSA,VI,FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=cayman -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-NOHSA,SI,FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-HSA,SI,FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-NOHSA,VI,FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cayman -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}global_load_i8:
diff --git a/llvm/test/CodeGen/AMDGPU/load-weird-sizes.ll b/llvm/test/CodeGen/AMDGPU/load-weird-sizes.ll
index d6162c3..f9ba624 100644
--- a/llvm/test/CodeGen/AMDGPU/load-weird-sizes.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-weird-sizes.ll
@@ -1,8 +1,8 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI-NOHSA -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -mtriple=amdgcn-amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=FUNC -check-prefix=CI-HSA -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI-NOHSA -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=EG -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=cayman < %s | FileCheck -check-prefix=R600 -check-prefix=CM -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI-NOHSA -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn-amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=FUNC -check-prefix=CI-HSA -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI-NOHSA -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cayman < %s | FileCheck -check-prefix=R600 -check-prefix=CM -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}load_i24:
; SI: {{flat|buffer}}_load_ubyte
diff --git a/llvm/test/CodeGen/AMDGPU/max.ll b/llvm/test/CodeGen/AMDGPU/max.ll
index ffcdac0..6387c9f 100644
--- a/llvm/test/CodeGen/AMDGPU/max.ll
+++ b/llvm/test/CodeGen/AMDGPU/max.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=cypress < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}v_test_imax_sge_i32:
diff --git a/llvm/test/CodeGen/AMDGPU/merge-stores.ll b/llvm/test/CodeGen/AMDGPU/merge-stores.ll
index dfd5b97..6b0ec48 100644
--- a/llvm/test/CodeGen/AMDGPU/merge-stores.ll
+++ b/llvm/test/CodeGen/AMDGPU/merge-stores.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs -amdgpu-load-store-vectorizer=0 < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=GCN-AA %s
-; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs -amdgpu-load-store-vectorizer=0 < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=GCN-AA %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs -amdgpu-load-store-vectorizer=0 < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=GCN-AA %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire -verify-machineinstrs -amdgpu-load-store-vectorizer=0 < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=GCN-AA %s
; This test is mostly to test DAG store merging, so disable the vectorizer.
; Run with devices with different unaligned load restrictions.
diff --git a/llvm/test/CodeGen/AMDGPU/mubuf.ll b/llvm/test/CodeGen/AMDGPU/mubuf.ll
index b23b211..9766649 100644
--- a/llvm/test/CodeGen/AMDGPU/mubuf.ll
+++ b/llvm/test/CodeGen/AMDGPU/mubuf.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=amdgcn -show-mc-encoding -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -show-mc-encoding -verify-machineinstrs < %s | FileCheck %s
declare i32 @llvm.amdgcn.workitem.id.x() readnone
diff --git a/llvm/test/CodeGen/AMDGPU/mul.ll b/llvm/test/CodeGen/AMDGPU/mul.ll
index 57c50c9..a029078 100644
--- a/llvm/test/CodeGen/AMDGPU/mul.ll
+++ b/llvm/test/CodeGen/AMDGPU/mul.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG %s -check-prefix=FUNC
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG %s -check-prefix=FUNC
; mul24 and mad24 are affected
diff --git a/llvm/test/CodeGen/AMDGPU/no-shrink-extloads.ll b/llvm/test/CodeGen/AMDGPU/no-shrink-extloads.ll
index 8a7bf6d..500e4cb 100644
--- a/llvm/test/CodeGen/AMDGPU/no-shrink-extloads.ll
+++ b/llvm/test/CodeGen/AMDGPU/no-shrink-extloads.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
diff --git a/llvm/test/CodeGen/AMDGPU/or.ll b/llvm/test/CodeGen/AMDGPU/or.ll
index eb08284..8e6885c 100644
--- a/llvm/test/CodeGen/AMDGPU/or.ll
+++ b/llvm/test/CodeGen/AMDGPU/or.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}or_v2i32:
diff --git a/llvm/test/CodeGen/AMDGPU/promote-alloca-invariant-markers.ll b/llvm/test/CodeGen/AMDGPU/promote-alloca-invariant-markers.ll
index f83eb56..776b151 100644
--- a/llvm/test/CodeGen/AMDGPU/promote-alloca-invariant-markers.ll
+++ b/llvm/test/CodeGen/AMDGPU/promote-alloca-invariant-markers.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=amdgcn -mattr=+promote-alloca -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mattr=+promote-alloca -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
declare {}* @llvm.invariant.start.p0i8(i64, i8* nocapture) #0
declare void @llvm.invariant.end.p0i8({}*, i64, i8* nocapture) #0
diff --git a/llvm/test/CodeGen/AMDGPU/reduce-load-width-alignment.ll b/llvm/test/CodeGen/AMDGPU/reduce-load-width-alignment.ll
index ecb513c..d8c7438 100644
--- a/llvm/test/CodeGen/AMDGPU/reduce-load-width-alignment.ll
+++ b/llvm/test/CodeGen/AMDGPU/reduce-load-width-alignment.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}reduce_i64_load_align_4_width_to_i32:
; GCN: buffer_load_dword [[VAL:v[0-9]+]]
diff --git a/llvm/test/CodeGen/AMDGPU/reorder-stores.ll b/llvm/test/CodeGen/AMDGPU/reorder-stores.ll
index ff40692..260b32e 100644
--- a/llvm/test/CodeGen/AMDGPU/reorder-stores.ll
+++ b/llvm/test/CodeGen/AMDGPU/reorder-stores.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn < %s | FileCheck -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn < %s | FileCheck -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=SI %s
; SI-LABEL: {{^}}no_reorder_v2f64_global_load_store:
; SI: buffer_load_dwordx4
diff --git a/llvm/test/CodeGen/AMDGPU/rotl.i64.ll b/llvm/test/CodeGen/AMDGPU/rotl.i64.ll
index 2664907..fa29d78 100644
--- a/llvm/test/CodeGen/AMDGPU/rotl.i64.ll
+++ b/llvm/test/CodeGen/AMDGPU/rotl.i64.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=BOTH %s
-; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=BOTH %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=BOTH %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=BOTH %s
; BOTH-LABEL: {{^}}s_rotl_i64:
; BOTH-DAG: s_lshl_b64
diff --git a/llvm/test/CodeGen/AMDGPU/rotr.i64.ll b/llvm/test/CodeGen/AMDGPU/rotr.i64.ll
index 9eda479..af58b40 100644
--- a/llvm/test/CodeGen/AMDGPU/rotr.i64.ll
+++ b/llvm/test/CodeGen/AMDGPU/rotr.i64.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=BOTH %s
-; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=BOTH %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=BOTH %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=BOTH %s
; BOTH-LABEL: {{^}}s_rotr_i64:
; BOTH-DAG: s_sub_i32
diff --git a/llvm/test/CodeGen/AMDGPU/rsq.ll b/llvm/test/CodeGen/AMDGPU/rsq.ll
index 9462683..6cbc8c3 100644
--- a/llvm/test/CodeGen/AMDGPU/rsq.ll
+++ b/llvm/test/CodeGen/AMDGPU/rsq.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -mattr=-fp32-denormals -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=SI-UNSAFE -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mattr=-fp32-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=SI-SAFE -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mattr=-fp32-denormals -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=SI-UNSAFE -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mattr=-fp32-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=SI-SAFE -check-prefix=SI %s
declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone
declare float @llvm.sqrt.f32(float) nounwind readnone
diff --git a/llvm/test/CodeGen/AMDGPU/s_movk_i32.ll b/llvm/test/CodeGen/AMDGPU/s_movk_i32.ll
index a131aaa..797fbc2 100644
--- a/llvm/test/CodeGen/AMDGPU/s_movk_i32.ll
+++ b/llvm/test/CodeGen/AMDGPU/s_movk_i32.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
; SI-LABEL: {{^}}s_movk_i32_k0:
; SI-DAG: s_mov_b32 [[LO_S_IMM:s[0-9]+]], 0xffff{{$}}
diff --git a/llvm/test/CodeGen/AMDGPU/saddo.ll b/llvm/test/CodeGen/AMDGPU/saddo.ll
index 586a455..09e87d5 100644
--- a/llvm/test/CodeGen/AMDGPU/saddo.ll
+++ b/llvm/test/CodeGen/AMDGPU/saddo.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs< %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=SI -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress -verify-machineinstrs< %s
declare { i32, i1 } @llvm.sadd.with.overflow.i32(i32, i32) nounwind readnone
declare { i64, i1 } @llvm.sadd.with.overflow.i64(i64, i64) nounwind readnone
diff --git a/llvm/test/CodeGen/AMDGPU/salu-to-valu.ll b/llvm/test/CodeGen/AMDGPU/salu-to-valu.ll
index 6e1dd16..d5b2fa0b67 100644
--- a/llvm/test/CodeGen/AMDGPU/salu-to-valu.ll
+++ b/llvm/test/CodeGen/AMDGPU/salu-to-valu.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=CI %s
-; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI --check-prefix=GCN-HSA %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-NOHSA -check-prefix=CI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=CI --check-prefix=GCN-HSA %s
declare i32 @llvm.amdgcn.workitem.id.x() #0
declare i32 @llvm.amdgcn.workitem.id.y() #0
diff --git a/llvm/test/CodeGen/AMDGPU/scalar_to_vector.ll b/llvm/test/CodeGen/AMDGPU/scalar_to_vector.ll
index 62d0d93..8c74ec6 100644
--- a/llvm/test/CodeGen/AMDGPU/scalar_to_vector.ll
+++ b/llvm/test/CodeGen/AMDGPU/scalar_to_vector.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; XXX - Why the packing?
; GCN-LABEL: {{^}}scalar_to_vector_v2i32:
diff --git a/llvm/test/CodeGen/AMDGPU/schedule-global-loads.ll b/llvm/test/CodeGen/AMDGPU/schedule-global-loads.ll
index 44d4608..2dddba8 100644
--- a/llvm/test/CodeGen/AMDGPU/schedule-global-loads.ll
+++ b/llvm/test/CodeGen/AMDGPU/schedule-global-loads.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=FUNC -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=FUNC -check-prefix=SI %s
; FIXME: This currently doesn't do a great job of clustering the
; loads, which end up with extra moves between them. Right now, it
diff --git a/llvm/test/CodeGen/AMDGPU/scratch-buffer.ll b/llvm/test/CodeGen/AMDGPU/scratch-buffer.ll
index 6b1e859..4ae9871 100644
--- a/llvm/test/CodeGen/AMDGPU/scratch-buffer.ll
+++ b/llvm/test/CodeGen/AMDGPU/scratch-buffer.ll
@@ -1,5 +1,5 @@
-; RUN: llc -verify-machineinstrs -march=amdgcn < %s | FileCheck -check-prefix=GCN %s
-; RUN: llc -verify-machineinstrs -march=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -march=amdgcn < %s | FileCheck -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -march=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=GCN %s
; When a frame index offset is more than 12-bits, make sure we don't store
; it in mubuf's offset field.
diff --git a/llvm/test/CodeGen/AMDGPU/sdiv.ll b/llvm/test/CodeGen/AMDGPU/sdiv.ll
index 7ec6ca8..305107f 100644
--- a/llvm/test/CodeGen/AMDGPU/sdiv.ll
+++ b/llvm/test/CodeGen/AMDGPU/sdiv.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; The code generated by sdiv is long and complex and may frequently change.
; The goal of this test is to make sure the ISel doesn't fail.
diff --git a/llvm/test/CodeGen/AMDGPU/sdwa-peephole.ll b/llvm/test/CodeGen/AMDGPU/sdwa-peephole.ll
index 0dc7cc3..9892250 100644
--- a/llvm/test/CodeGen/AMDGPU/sdwa-peephole.ll
+++ b/llvm/test/CodeGen/AMDGPU/sdwa-peephole.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=fiji -amdgpu-sdwa-peephole=0 -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=NOSDWA -check-prefix=GCN %s
-; RUN: llc -march=amdgcn -mcpu=fiji -amdgpu-sdwa-peephole -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=SDWA -check-prefix=GCN %s
-; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-sdwa-peephole -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=GFX9 -check-prefix=SDWA -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -amdgpu-sdwa-peephole=0 -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=NOSDWA -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -amdgpu-sdwa-peephole -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=SDWA -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx900 -amdgpu-sdwa-peephole -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=GFX9 -check-prefix=SDWA -check-prefix=GCN %s
; GCN-LABEL: {{^}}add_shr_i32:
; NOSDWA: v_lshrrev_b32_e32 v[[DST:[0-9]+]], 16, v{{[0-9]+}}
diff --git a/llvm/test/CodeGen/AMDGPU/select-vectors.ll b/llvm/test/CodeGen/AMDGPU/select-vectors.ll
index ebbc675..b77ebcf 100644
--- a/llvm/test/CodeGen/AMDGPU/select-vectors.ll
+++ b/llvm/test/CodeGen/AMDGPU/select-vectors.ll
@@ -1,6 +1,6 @@
-; RUN: llc -verify-machineinstrs -march=amdgcn < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -verify-machineinstrs -march=amdgcn -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
-; RUN: llc -verify-machineinstrs -march=amdgcn -mcpu=gfx900 -mattr=-flat-for-global < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -march=amdgcn < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -march=amdgcn -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -verify-machineinstrs -march=amdgcn -mcpu=gfx900 -mattr=-flat-for-global < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 %s
; Test expansion of scalar selects on vectors.
; Evergreen not enabled since it seems to be having problems with doubles.
diff --git a/llvm/test/CodeGen/AMDGPU/select.f16.ll b/llvm/test/CodeGen/AMDGPU/select.f16.ll
index 92ee2eb..62627e5 100644
--- a/llvm/test/CodeGen/AMDGPU/select.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/select.f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}select_f16:
; GCN: buffer_load_ushort v[[A_F16:[0-9]+]]
diff --git a/llvm/test/CodeGen/AMDGPU/sext-in-reg.ll b/llvm/test/CodeGen/AMDGPU/sext-in-reg.ll
index 160fb6a..5b4d9ed 100644
--- a/llvm/test/CodeGen/AMDGPU/sext-in-reg.ll
+++ b/llvm/test/CodeGen/AMDGPU/sext-in-reg.ll
@@ -1,7 +1,7 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=GFX89 -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 -check-prefix=GFX89 -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=cypress < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=GFX89 -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefix=GCN -check-prefix=GFX9 -check-prefix=GFX89 -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FIXME: i16 promotion pass ruins the scalar cases when legal.
; FIXME: r600 fails verifier
diff --git a/llvm/test/CodeGen/AMDGPU/sgpr-copy-duplicate-operand.ll b/llvm/test/CodeGen/AMDGPU/sgpr-copy-duplicate-operand.ll
index fb0bbaa..8250bad 100644
--- a/llvm/test/CodeGen/AMDGPU/sgpr-copy-duplicate-operand.ll
+++ b/llvm/test/CodeGen/AMDGPU/sgpr-copy-duplicate-operand.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs< %s | FileCheck -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs< %s | FileCheck -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs< %s | FileCheck -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs< %s | FileCheck -check-prefix=SI %s
; Copy VGPR -> SGPR used twice as an instruction operand, which is then
; used in an REG_SEQUENCE that also needs to be handled.
diff --git a/llvm/test/CodeGen/AMDGPU/sgpr-copy.ll b/llvm/test/CodeGen/AMDGPU/sgpr-copy.ll
index 9310511..3b24cf8 100644
--- a/llvm/test/CodeGen/AMDGPU/sgpr-copy.ll
+++ b/llvm/test/CodeGen/AMDGPU/sgpr-copy.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck %s
; CHECK-LABEL: {{^}}phi1:
; CHECK: s_buffer_load_dword [[DST:s[0-9]]], {{s\[[0-9]+:[0-9]+\]}}, 0x0
diff --git a/llvm/test/CodeGen/AMDGPU/shift-and-i128-ubfe.ll b/llvm/test/CodeGen/AMDGPU/shift-and-i128-ubfe.ll
index 4f7b61a..ac9cc98 100644
--- a/llvm/test/CodeGen/AMDGPU/shift-and-i128-ubfe.ll
+++ b/llvm/test/CodeGen/AMDGPU/shift-and-i128-ubfe.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; Extract the high bit of the 1st quarter
; GCN-LABEL: {{^}}v_uextract_bit_31_i128:
diff --git a/llvm/test/CodeGen/AMDGPU/shift-and-i64-ubfe.ll b/llvm/test/CodeGen/AMDGPU/shift-and-i64-ubfe.ll
index c70eb9b..670287b 100644
--- a/llvm/test/CodeGen/AMDGPU/shift-and-i64-ubfe.ll
+++ b/llvm/test/CodeGen/AMDGPU/shift-and-i64-ubfe.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; Make sure 64-bit BFE pattern does a 32-bit BFE on the relevant half.
diff --git a/llvm/test/CodeGen/AMDGPU/shift-i64-opts.ll b/llvm/test/CodeGen/AMDGPU/shift-i64-opts.ll
index 5306e19..f3faa39 100644
--- a/llvm/test/CodeGen/AMDGPU/shift-i64-opts.ll
+++ b/llvm/test/CodeGen/AMDGPU/shift-i64-opts.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -mcpu=tahiti < %s | FileCheck -check-prefix=FAST64 -check-prefix=GCN %s
-; RUN: llc -march=amdgcn -mcpu=bonaire < %s | FileCheck -check-prefix=SLOW64 -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti < %s | FileCheck -check-prefix=FAST64 -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire < %s | FileCheck -check-prefix=SLOW64 -check-prefix=GCN %s
; lshr (i64 x), c: c > 32 => reg_sequence lshr (i32 hi_32(x)), (c - 32), 0
diff --git a/llvm/test/CodeGen/AMDGPU/shl.ll b/llvm/test/CodeGen/AMDGPU/shl.ll
index edc313e..13ac914 100644
--- a/llvm/test/CodeGen/AMDGPU/shl.ll
+++ b/llvm/test/CodeGen/AMDGPU/shl.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s
; XUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
declare i32 @llvm.r600.read.tidig.x() #0
diff --git a/llvm/test/CodeGen/AMDGPU/si-triv-disjoint-mem-access.ll b/llvm/test/CodeGen/AMDGPU/si-triv-disjoint-mem-access.ll
index 348c720..1710918 100644
--- a/llvm/test/CodeGen/AMDGPU/si-triv-disjoint-mem-access.ll
+++ b/llvm/test/CodeGen/AMDGPU/si-triv-disjoint-mem-access.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=amdgcn -mcpu=bonaire -enable-amdgpu-aa=0 -verify-machineinstrs -enable-misched -enable-aa-sched-mi < %s | FileCheck -check-prefix=FUNC -check-prefix=CI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire -enable-amdgpu-aa=0 -verify-machineinstrs -enable-misched -enable-aa-sched-mi < %s | FileCheck -check-prefix=FUNC -check-prefix=CI %s
declare void @llvm.amdgcn.tbuffer.store.i32(i32, <4 x i32>, i32, i32, i32, i32, i32, i32, i1, i1)
declare void @llvm.amdgcn.tbuffer.store.v4i32(<4 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i1, i1)
diff --git a/llvm/test/CodeGen/AMDGPU/sign_extend.ll b/llvm/test/CodeGen/AMDGPU/sign_extend.ll
index 3e452c2..c80945f 100644
--- a/llvm/test/CodeGen/AMDGPU/sign_extend.ll
+++ b/llvm/test/CodeGen/AMDGPU/sign_extend.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,SI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,VI %s
; GCN-LABEL: {{^}}s_sext_i1_to_i32:
; GCN: v_cndmask_b32_e64
diff --git a/llvm/test/CodeGen/AMDGPU/sitofp.f16.ll b/llvm/test/CodeGen/AMDGPU/sitofp.f16.ll
index 574d1c0..0bcef99 100644
--- a/llvm/test/CodeGen/AMDGPU/sitofp.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/sitofp.f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}sitofp_i16_to_f16
; GCN: buffer_load_{{sshort|ushort}} v[[A_I16:[0-9]+]]
diff --git a/llvm/test/CodeGen/AMDGPU/sminmax.ll b/llvm/test/CodeGen/AMDGPU/sminmax.ll
index 827d672..eced4a0 100644
--- a/llvm/test/CodeGen/AMDGPU/sminmax.ll
+++ b/llvm/test/CodeGen/AMDGPU/sminmax.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}s_abs_i32:
; GCN: s_abs_i32
diff --git a/llvm/test/CodeGen/AMDGPU/sminmax.v2i16.ll b/llvm/test/CodeGen/AMDGPU/sminmax.v2i16.ll
index a9aac2d..2726342 100644
--- a/llvm/test/CodeGen/AMDGPU/sminmax.v2i16.ll
+++ b/llvm/test/CodeGen/AMDGPU/sminmax.v2i16.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefix=GFX9 -check-prefix=GCN %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=CIVI -check-prefix=GCN %s
-; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=CI -check-prefix=CIVI -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx901 -mattr=-flat-for-global -verify-machineinstrs -enable-packed-inlinable-literals < %s | FileCheck -check-prefix=GFX9 -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=CIVI -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=bonaire -verify-machineinstrs < %s | FileCheck -check-prefix=CI -check-prefix=CIVI -check-prefix=GCN %s
; GCN-LABEL: {{^}}s_abs_v2i16:
; GFX9: s_load_dword [[VAL:s[0-9]+]]
diff --git a/llvm/test/CodeGen/AMDGPU/spill-cfg-position.ll b/llvm/test/CodeGen/AMDGPU/spill-cfg-position.ll
index 1ca0919..cbf9f37 100644
--- a/llvm/test/CodeGen/AMDGPU/spill-cfg-position.ll
+++ b/llvm/test/CodeGen/AMDGPU/spill-cfg-position.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs -stress-regalloc=6 < %s | FileCheck %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs -stress-regalloc=6 < %s | FileCheck %s
; Inline spiller can decide to move a spill as early as possible in the basic block.
; It will skip phis and label, but we also need to make sure it skips instructions
diff --git a/llvm/test/CodeGen/AMDGPU/sra.ll b/llvm/test/CodeGen/AMDGPU/sra.ll
index 44cfdf6..74618b2 100644
--- a/llvm/test/CodeGen/AMDGPU/sra.ll
+++ b/llvm/test/CodeGen/AMDGPU/sra.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
declare i32 @llvm.r600.read.tidig.x() #0
diff --git a/llvm/test/CodeGen/AMDGPU/srem.ll b/llvm/test/CodeGen/AMDGPU/srem.ll
index e067258..51eaf9a 100644
--- a/llvm/test/CodeGen/AMDGPU/srem.ll
+++ b/llvm/test/CodeGen/AMDGPU/srem.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=SI < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood < %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=SI < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s
define amdgpu_kernel void @srem_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
%den_ptr = getelementptr i32, i32 addrspace(1)* %in, i32 1
diff --git a/llvm/test/CodeGen/AMDGPU/srl.ll b/llvm/test/CodeGen/AMDGPU/srl.ll
index cb40ecf..8878b45 100644
--- a/llvm/test/CodeGen/AMDGPU/srl.ll
+++ b/llvm/test/CodeGen/AMDGPU/srl.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN -check-prefix=FUNC %s
; XUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=GCN -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
declare i32 @llvm.r600.read.tidig.x() #0
diff --git a/llvm/test/CodeGen/AMDGPU/ssubo.ll b/llvm/test/CodeGen/AMDGPU/ssubo.ll
index 1356323..d65c2ad 100644
--- a/llvm/test/CodeGen/AMDGPU/ssubo.ll
+++ b/llvm/test/CodeGen/AMDGPU/ssubo.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs< %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=SI -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs< %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress -verify-machineinstrs< %s
declare { i32, i1 } @llvm.ssub.with.overflow.i32(i32, i32) nounwind readnone
declare { i64, i1 } @llvm.ssub.with.overflow.i64(i64, i64) nounwind readnone
diff --git a/llvm/test/CodeGen/AMDGPU/sub.ll b/llvm/test/CodeGen/AMDGPU/sub.ll
index e7655df..46f1b12 100644
--- a/llvm/test/CodeGen/AMDGPU/sub.ll
+++ b/llvm/test/CodeGen/AMDGPU/sub.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
declare i32 @llvm.r600.read.tidig.x() readnone
diff --git a/llvm/test/CodeGen/AMDGPU/trunc-bitcast-vector.ll b/llvm/test/CodeGen/AMDGPU/trunc-bitcast-vector.ll
index f900403..77a6820 100644
--- a/llvm/test/CodeGen/AMDGPU/trunc-bitcast-vector.ll
+++ b/llvm/test/CodeGen/AMDGPU/trunc-bitcast-vector.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck --check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck --check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck --check-prefix=VI %s
; CHECK-LABEL: {{^}}trunc_i64_bitcast_v2i32:
; CHECK: buffer_load_dword v
diff --git a/llvm/test/CodeGen/AMDGPU/trunc.ll b/llvm/test/CodeGen/AMDGPU/trunc.ll
index 0c91d52..da038f4 100644
--- a/llvm/test/CodeGen/AMDGPU/trunc.ll
+++ b/llvm/test/CodeGen/AMDGPU/trunc.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs< %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -verify-machineinstrs< %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
-; RUN: llc -march=r600 -mcpu=cypress < %s | FileCheck -check-prefix=EG %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs< %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -verify-machineinstrs< %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress < %s | FileCheck -check-prefix=EG %s
declare i32 @llvm.r600.read.tidig.x() nounwind readnone
diff --git a/llvm/test/CodeGen/AMDGPU/uaddo.ll b/llvm/test/CodeGen/AMDGPU/uaddo.ll
index 632ccaa..f6a7263 100644
--- a/llvm/test/CodeGen/AMDGPU/uaddo.ll
+++ b/llvm/test/CodeGen/AMDGPU/uaddo.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,SI,FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,VI,FUNC %s
-; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefixes=EG,FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,SI,FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,VI,FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefixes=EG,FUNC %s
; FUNC-LABEL: {{^}}s_uaddo_i64_zext:
; GCN: s_add_u32
diff --git a/llvm/test/CodeGen/AMDGPU/udiv.ll b/llvm/test/CodeGen/AMDGPU/udiv.ll
index d9dab0d..1d68377 100644
--- a/llvm/test/CodeGen/AMDGPU/udiv.ll
+++ b/llvm/test/CodeGen/AMDGPU/udiv.ll
@@ -1,9 +1,9 @@
-; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs -mattr=-fp32-denormals < %s | FileCheck -check-prefix=SI -check-prefix=FUNC -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs -mattr=-fp32-denormals < %s | FileCheck -check-prefix=SI -check-prefix=FUNC -check-prefix=VI %s
-; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=fiji -mattr=+fp32-denormals < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn--amdhsa -mcpu=fiji -mattr=+fp32-denormals < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
-; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}udiv_i32:
; EG-NOT: SETGE_INT
diff --git a/llvm/test/CodeGen/AMDGPU/uitofp.f16.ll b/llvm/test/CodeGen/AMDGPU/uitofp.f16.ll
index 0c3b0fc..eaa1d07 100644
--- a/llvm/test/CodeGen/AMDGPU/uitofp.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/uitofp.f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs -enable-unsafe-fp-math < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}uitofp_i16_to_f16
; GCN: buffer_load_ushort v[[A_I16:[0-9]+]]
diff --git a/llvm/test/CodeGen/AMDGPU/urem.ll b/llvm/test/CodeGen/AMDGPU/urem.ll
index fb4eab4..823c918 100644
--- a/llvm/test/CodeGen/AMDGPU/urem.ll
+++ b/llvm/test/CodeGen/AMDGPU/urem.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; The code generated by urem is long and complex and may frequently
; change. The goal of this test is to make sure the ISel doesn't fail
diff --git a/llvm/test/CodeGen/AMDGPU/usubo.ll b/llvm/test/CodeGen/AMDGPU/usubo.ll
index d1f454f..99c759a 100644
--- a/llvm/test/CodeGen/AMDGPU/usubo.ll
+++ b/llvm/test/CodeGen/AMDGPU/usubo.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,SI,FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,VI,FUNC %s
-; RUN: llc -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefixes=EG,FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,SI,FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,VI,FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=cypress -verify-machineinstrs < %s | FileCheck -check-prefixes=EG,FUNC %s
; FUNC-LABEL: {{^}}s_usubo_i64_zext:
; GCN: s_sub_u32
diff --git a/llvm/test/CodeGen/AMDGPU/v_mac.ll b/llvm/test/CodeGen/AMDGPU/v_mac.ll
index 2b96f7d..5f3c6a2 100644
--- a/llvm/test/CodeGen/AMDGPU/v_mac.ll
+++ b/llvm/test/CodeGen/AMDGPU/v_mac.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=VI-FLUSH -check-prefix=GCN %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=+fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=VI-DENORM -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=VI-FLUSH -check-prefix=GCN %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=+fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=VI -check-prefix=VI-DENORM -check-prefix=GCN %s
; GCN-LABEL: {{^}}mac_vvv:
; GCN: buffer_load_dword [[A:v[0-9]+]], off, s[{{[0-9]+:[0-9]+}}], 0{{$}}
diff --git a/llvm/test/CodeGen/AMDGPU/v_mac_f16.ll b/llvm/test/CodeGen/AMDGPU/v_mac_f16.ll
index ce4a69d..3f18b95 100644
--- a/llvm/test/CodeGen/AMDGPU/v_mac_f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/v_mac_f16.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=amdgcn -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mattr=-fp64-fp16-denormals -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -mattr=-fp64-fp16-denormals,-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VI %s
; GCN-LABEL: {{^}}mac_f16:
; GCN: {{buffer|flat}}_load_ushort v[[A_F16:[0-9]+]]
diff --git a/llvm/test/CodeGen/AMDGPU/vectorize-global-local.ll b/llvm/test/CodeGen/AMDGPU/vectorize-global-local.ll
index 90cf34e..381ff5b 100644
--- a/llvm/test/CodeGen/AMDGPU/vectorize-global-local.ll
+++ b/llvm/test/CodeGen/AMDGPU/vectorize-global-local.ll
@@ -1,4 +1,4 @@
-; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck %s
; CHECK-DAG: flat_load_dwordx4
; CHECK-DAG: flat_load_dwordx4
; CHECK-DAG: flat_load_dwordx4
diff --git a/llvm/test/CodeGen/AMDGPU/xor.ll b/llvm/test/CodeGen/AMDGPU/xor.ll
index 57a082a..db3e42a 100644
--- a/llvm/test/CodeGen/AMDGPU/xor.ll
+++ b/llvm/test/CodeGen/AMDGPU/xor.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
; FUNC-LABEL: {{^}}xor_v2i32: