; -enable-misched=false makes the register usage more predictable
-; -regalloc=basic just makes the test run faster
-; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=basic -vgpr-regalloc=basic -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX9
-; RUN: llc -march=amdgcn -mcpu=gfx90a -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=basic -vgpr-regalloc=basic -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX90A
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=basic -vgpr-regalloc=basic -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10WGP-WAVE32
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=+wavefrontsize64 -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=basic -vgpr-regalloc=basic -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10WGP-WAVE64
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=+cumode -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=basic -vgpr-regalloc=basic -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10CU-WAVE32
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=+cumode,+wavefrontsize64 -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=basic -vgpr-regalloc=basic -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10CU-WAVE64
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=basic -vgpr-regalloc=basic -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX11WGP-WAVE32
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize64 -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=basic -vgpr-regalloc=basic -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX11WGP-WAVE64
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=+cumode -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=basic -vgpr-regalloc=basic -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX11CU-WAVE32
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=+cumode,+wavefrontsize64 -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=basic -vgpr-regalloc=basic -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX11CU-WAVE64
+; -regalloc=fast just makes the test run faster
+; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=fast -vgpr-regalloc=fast -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX9
+; RUN: llc -march=amdgcn -mcpu=gfx90a -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=fast -vgpr-regalloc=fast -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX90A
+; RUN: llc -march=amdgcn -mcpu=gfx1010 -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=fast -vgpr-regalloc=fast -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10WGP-WAVE32
+; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=+wavefrontsize64 -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=fast -vgpr-regalloc=fast -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10WGP-WAVE64
+; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=+cumode -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=fast -vgpr-regalloc=fast -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10CU-WAVE32
+; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=+cumode,+wavefrontsize64 -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=fast -vgpr-regalloc=fast -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX10CU-WAVE64
+; RUN: llc -march=amdgcn -mcpu=gfx1100 -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=fast -vgpr-regalloc=fast -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX11WGP-WAVE32
+; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize64 -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=fast -vgpr-regalloc=fast -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX11WGP-WAVE64
+; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=+cumode -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=fast -vgpr-regalloc=fast -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX11CU-WAVE32
+; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=+cumode,+wavefrontsize64 -amdgpu-function-calls=false -enable-misched=false -sgpr-regalloc=fast -vgpr-regalloc=fast -verify-machineinstrs < %s | FileCheck %s --check-prefixes=GCN,GFX11CU-WAVE64
define internal void @use256vgprs() {
%v0 = call i32 asm sideeffect "; def $0", "=v"()