AMDGPU: Cleanup min/max legacy tests

Also add some more tests in preparation for
a future patch.

llvm-svn: 339526

4 files changed, 664 insertions, 124 deletions

diff --git a/llvm/test/CodeGen/AMDGPU/fmax_legacy.f64.ll b/llvm/test/CodeGen/AMDGPU/fmax_legacy.f64.ll
index 083346e9d1c..11730796850 100644
--- a/llvm/test/CodeGen/AMDGPU/fmax_legacy.f64.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmax_legacy.f64.ll
@@ -1,10 +1,48 @@
-; RUN: llc -march=amdgcn < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; Make sure we don't try to form FMAX_LEGACY nodes with f64
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=amdgcn-- -mcpu=tahiti < %s | FileCheck -check-prefix=SI %s
+; RUN: llc -mtriple=amdgcn-- -mcpu=fiji < %s | FileCheck -check-prefix=VI %s
 
-declare i32 @llvm.amdgcn.workitem.id.x() #1
+; Make sure we don't try to form FMAX_LEGACY nodes with f64
 
-; FUNC-LABEL: @test_fmax_legacy_uge_f64
 define amdgpu_kernel void @test_fmax_legacy_uge_f64(double addrspace(1)* %out, double addrspace(1)* %in) #0 {
+; SI-LABEL: test_fmax_legacy_uge_f64:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s10, 0
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b64 s[8:9], s[2:3]
+; SI-NEXT:    v_mov_b32_e32 v1, 0
+; SI-NEXT:    buffer_load_dwordx4 v[0:3], v[0:1], s[8:11], 0 addr64
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cmp_nlt_f64_e32 vcc, v[0:1], v[2:3]
+; SI-NEXT:    v_cndmask_b32_e32 v1, v3, v1, vcc
+; SI-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: test_fmax_legacy_uge_f64:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_cmp_nlt_f64_e32 vcc, v[0:1], v[2:3]
+; VI-NEXT:    v_cndmask_b32_e32 v1, v3, v1, vcc
+; VI-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
+; VI-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; VI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %gep.0 = getelementptr double, double addrspace(1)* %in, i32 %tid
   %gep.1 = getelementptr double, double addrspace(1)* %gep.0, i32 1
@@ -18,8 +56,45 @@ define amdgpu_kernel void @test_fmax_legacy_uge_f64(double addrspace(1)* %out, d
   ret void
 }
 
-; FUNC-LABEL: @test_fmax_legacy_oge_f64
 define amdgpu_kernel void @test_fmax_legacy_oge_f64(double addrspace(1)* %out, double addrspace(1)* %in) #0 {
+; SI-LABEL: test_fmax_legacy_oge_f64:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s10, 0
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b64 s[8:9], s[2:3]
+; SI-NEXT:    v_mov_b32_e32 v1, 0
+; SI-NEXT:    buffer_load_dwordx4 v[0:3], v[0:1], s[8:11], 0 addr64
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cmp_ge_f64_e32 vcc, v[0:1], v[2:3]
+; SI-NEXT:    v_cndmask_b32_e32 v1, v3, v1, vcc
+; SI-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: test_fmax_legacy_oge_f64:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_cmp_ge_f64_e32 vcc, v[0:1], v[2:3]
+; VI-NEXT:    v_cndmask_b32_e32 v1, v3, v1, vcc
+; VI-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
+; VI-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; VI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %gep.0 = getelementptr double, double addrspace(1)* %in, i32 %tid
   %gep.1 = getelementptr double, double addrspace(1)* %gep.0, i32 1
@@ -33,8 +108,45 @@ define amdgpu_kernel void @test_fmax_legacy_oge_f64(double addrspace(1)* %out, d
   ret void
 }
 
-; FUNC-LABEL: @test_fmax_legacy_ugt_f64
 define amdgpu_kernel void @test_fmax_legacy_ugt_f64(double addrspace(1)* %out, double addrspace(1)* %in) #0 {
+; SI-LABEL: test_fmax_legacy_ugt_f64:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s10, 0
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b64 s[8:9], s[2:3]
+; SI-NEXT:    v_mov_b32_e32 v1, 0
+; SI-NEXT:    buffer_load_dwordx4 v[0:3], v[0:1], s[8:11], 0 addr64
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cmp_nle_f64_e32 vcc, v[0:1], v[2:3]
+; SI-NEXT:    v_cndmask_b32_e32 v1, v3, v1, vcc
+; SI-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: test_fmax_legacy_ugt_f64:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_cmp_nle_f64_e32 vcc, v[0:1], v[2:3]
+; VI-NEXT:    v_cndmask_b32_e32 v1, v3, v1, vcc
+; VI-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
+; VI-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; VI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %gep.0 = getelementptr double, double addrspace(1)* %in, i32 %tid
   %gep.1 = getelementptr double, double addrspace(1)* %gep.0, i32 1
@@ -48,8 +160,45 @@ define amdgpu_kernel void @test_fmax_legacy_ugt_f64(double addrspace(1)* %out, d
   ret void
 }
 
-; FUNC-LABEL: @test_fmax_legacy_ogt_f64
 define amdgpu_kernel void @test_fmax_legacy_ogt_f64(double addrspace(1)* %out, double addrspace(1)* %in) #0 {
+; SI-LABEL: test_fmax_legacy_ogt_f64:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s10, 0
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b64 s[8:9], s[2:3]
+; SI-NEXT:    v_mov_b32_e32 v1, 0
+; SI-NEXT:    buffer_load_dwordx4 v[0:3], v[0:1], s[8:11], 0 addr64
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cmp_gt_f64_e32 vcc, v[0:1], v[2:3]
+; SI-NEXT:    v_cndmask_b32_e32 v1, v3, v1, vcc
+; SI-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: test_fmax_legacy_ogt_f64:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_cmp_gt_f64_e32 vcc, v[0:1], v[2:3]
+; VI-NEXT:    v_cndmask_b32_e32 v1, v3, v1, vcc
+; VI-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
+; VI-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; VI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %gep.0 = getelementptr double, double addrspace(1)* %in, i32 %tid
   %gep.1 = getelementptr double, double addrspace(1)* %gep.0, i32 1
@@ -63,5 +212,7 @@ define amdgpu_kernel void @test_fmax_legacy_ogt_f64(double addrspace(1)* %out, d
   ret void
 }
 
+declare i32 @llvm.amdgcn.workitem.id.x() #1
+
 attributes #0 = { nounwind }
 attributes #1 = { nounwind readnone }
diff --git a/llvm/test/CodeGen/AMDGPU/fmax_legacy.ll b/llvm/test/CodeGen/AMDGPU/fmax_legacy.ll
index ebc88fa4103..6a1f7966c30 100644
--- a/llvm/test/CodeGen/AMDGPU/fmax_legacy.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmax_legacy.ll
@@ -1,14 +1,14 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=SI-SAFE -check-prefix=FUNC %s
-; RUN: llc -enable-no-nans-fp-math -enable-no-signed-zeros-fp-math -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI-NONAN -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=GCN-SAFE -check-prefix=FUNC %s
+; RUN: llc -enable-no-nans-fp-math -enable-no-signed-zeros-fp-math -march=amdgcn -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN-NONAN -check-prefix=GCN -check-prefix=FUNC %s
+; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -enable-var-scope -check-prefix=EG -check-prefix=FUNC %s
 
 declare i32 @llvm.r600.read.tidig.x() #1
 
-; FUNC-LABEL: @test_fmax_legacy_uge_f32
-; SI: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
-; SI: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
-; SI-SAFE: v_max_legacy_f32_e32 {{v[0-9]+}}, [[B]], [[A]]
-; SI-NONAN: v_max_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
+; FUNC-LABEL: {{^}}test_fmax_legacy_uge_f32:
+; GCN: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
+; GCN: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
+; GCN-SAFE: v_max_legacy_f32_e32 {{v[0-9]+}}, [[B]], [[A]]
+; GCN-NONAN: v_max_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
 
 ; EG: MAX
 define amdgpu_kernel void @test_fmax_legacy_uge_f32(float addrspace(1)* %out, float addrspace(1)* %in) #0 {
@@ -25,11 +25,37 @@ define amdgpu_kernel void @test_fmax_legacy_uge_f32(float addrspace(1)* %out, fl
   ret void
 }
 
-; FUNC-LABEL: @test_fmax_legacy_oge_f32
-; SI: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
-; SI: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
-; SI-SAFE: v_max_legacy_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
-; SI-NONAN: v_max_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
+; FUNC-LABEL: {{^}}test_fmax_legacy_uge_f32_nnan_src:
+; GCN: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
+; GCN: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
+; GCN-DAG: v_add_f32_e32 [[ADD_A:v[0-9]+]], 1.0, [[A]]
+; GCN-DAG: v_add_f32_e32 [[ADD_B:v[0-9]+]], 2.0, [[B]]
+
+; GCN-SAFE: v_max_legacy_f32_e32 {{v[0-9]+}}, [[ADD_B]], [[ADD_A]]
+; GCN-NONAN: v_max_f32_e32 {{v[0-9]+}}, [[ADD_A]], [[ADD_B]]
+
+; EG: MAX
+define amdgpu_kernel void @test_fmax_legacy_uge_f32_nnan_src(float addrspace(1)* %out, float addrspace(1)* %in) #0 {
+  %tid = call i32 @llvm.r600.read.tidig.x() #1
+  %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
+  %gep.1 = getelementptr float, float addrspace(1)* %gep.0, i32 1
+
+  %a = load volatile float, float addrspace(1)* %gep.0, align 4
+  %b = load volatile float, float addrspace(1)* %gep.1, align 4
+  %a.nnan = fadd nnan float %a, 1.0
+  %b.nnan = fadd nnan float %b, 2.0
+
+  %cmp = fcmp uge float %a.nnan, %b.nnan
+  %val = select i1 %cmp, float %a.nnan, float %b.nnan
+  store float %val, float addrspace(1)* %out, align 4
+  ret void
+}
+
+; FUNC-LABEL: {{^}}test_fmax_legacy_oge_f32:
+; GCN: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
+; GCN: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
+; GCN-SAFE: v_max_legacy_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
+; GCN-NONAN: v_max_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
 ; EG: MAX
 define amdgpu_kernel void @test_fmax_legacy_oge_f32(float addrspace(1)* %out, float addrspace(1)* %in) #0 {
   %tid = call i32 @llvm.r600.read.tidig.x() #1
@@ -45,11 +71,11 @@ define amdgpu_kernel void @test_fmax_legacy_oge_f32(float addrspace(1)* %out, fl
   ret void
 }
 
-; FUNC-LABEL: @test_fmax_legacy_ugt_f32
-; SI: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
-; SI: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
-; SI-SAFE: v_max_legacy_f32_e32 {{v[0-9]+}}, [[B]], [[A]]
-; SI-NONAN: v_max_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
+; FUNC-LABEL: {{^}}test_fmax_legacy_ugt_f32:
+; GCN: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
+; GCN: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
+; GCN-SAFE: v_max_legacy_f32_e32 {{v[0-9]+}}, [[B]], [[A]]
+; GCN-NONAN: v_max_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
 ; EG: MAX
 define amdgpu_kernel void @test_fmax_legacy_ugt_f32(float addrspace(1)* %out, float addrspace(1)* %in) #0 {
   %tid = call i32 @llvm.r600.read.tidig.x() #1
@@ -65,11 +91,11 @@ define amdgpu_kernel void @test_fmax_legacy_ugt_f32(float addrspace(1)* %out, fl
   ret void
 }
 
-; FUNC-LABEL: @test_fmax_legacy_ogt_f32
-; SI: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
-; SI: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
-; SI-SAFE: v_max_legacy_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
-; SI-NONAN: v_max_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
+; FUNC-LABEL: {{^}}test_fmax_legacy_ogt_f32:
+; GCN: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
+; GCN: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
+; GCN-SAFE: v_max_legacy_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
+; GCN-NONAN: v_max_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
 ; EG: MAX
 define amdgpu_kernel void @test_fmax_legacy_ogt_f32(float addrspace(1)* %out, float addrspace(1)* %in) #0 {
   %tid = call i32 @llvm.r600.read.tidig.x() #1
@@ -86,10 +112,10 @@ define amdgpu_kernel void @test_fmax_legacy_ogt_f32(float addrspace(1)* %out, fl
 }
 
 ; FUNC-LABEL: {{^}}test_fmax_legacy_ogt_v1f32:
-; SI: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
-; SI: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
-; SI-SAFE: v_max_legacy_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
-; SI-NONAN: v_max_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
+; GCN: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
+; GCN: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
+; GCN-SAFE: v_max_legacy_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
+; GCN-NONAN: v_max_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
 ; EG: MAX
 define amdgpu_kernel void @test_fmax_legacy_ogt_v1f32(<1 x float> addrspace(1)* %out, <1 x float> addrspace(1)* %in) #0 {
   %tid = call i32 @llvm.r600.read.tidig.x() #1
@@ -106,12 +132,12 @@ define amdgpu_kernel void @test_fmax_legacy_ogt_v1f32(<1 x float> addrspace(1)*
 }
 
 ; FUNC-LABEL: {{^}}test_fmax_legacy_ogt_v3f32:
-; SI-SAFE: v_max_legacy_f32_e32
-; SI-SAFE: v_max_legacy_f32_e32
-; SI-SAFE: v_max_legacy_f32_e32
-; SI-NONAN: v_max_f32_e32
-; SI-NONAN: v_max_f32_e32
-; SI-NONAN: v_max_f32_e32
+; GCN-SAFE: v_max_legacy_f32_e32
+; GCN-SAFE: v_max_legacy_f32_e32
+; GCN-SAFE: v_max_legacy_f32_e32
+; GCN-NONAN: v_max_f32_e32
+; GCN-NONAN: v_max_f32_e32
+; GCN-NONAN: v_max_f32_e32
 define amdgpu_kernel void @test_fmax_legacy_ogt_v3f32(<3 x float> addrspace(1)* %out, <3 x float> addrspace(1)* %in) #0 {
   %tid = call i32 @llvm.r600.read.tidig.x() #1
   %gep.0 = getelementptr <3 x float>, <3 x float> addrspace(1)* %in, i32 %tid
@@ -126,13 +152,13 @@ define amdgpu_kernel void @test_fmax_legacy_ogt_v3f32(<3 x float> addrspace(1)*
   ret void
 }
 
-; FUNC-LABEL: @test_fmax_legacy_ogt_f32_multi_use
-; SI: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
-; SI: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
-; SI-NOT: v_max_
-; SI: v_cmp_gt_f32
-; SI-NEXT: v_cndmask_b32
-; SI-NOT: v_max_
+; FUNC-LABEL: {{^}}test_fmax_legacy_ogt_f32_multi_use:
+; GCN: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
+; GCN: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
+; GCN-NOT: v_max_
+; GCN: v_cmp_gt_f32
+; GCN-NEXT: v_cndmask_b32
+; GCN-NOT: v_max_
 
 ; EG: MAX
 define amdgpu_kernel void @test_fmax_legacy_ogt_f32_multi_use(float addrspace(1)* %out0, i1 addrspace(1)* %out1, float addrspace(1)* %in) #0 {
diff --git a/llvm/test/CodeGen/AMDGPU/fmin_legacy.f64.ll b/llvm/test/CodeGen/AMDGPU/fmin_legacy.f64.ll
index 99bc114831c..306b4c25fb9 100644
--- a/llvm/test/CodeGen/AMDGPU/fmin_legacy.f64.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmin_legacy.f64.ll
@@ -1,20 +1,150 @@
-; RUN: llc -march=amdgcn < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=amdgcn-- -mcpu=tahiti < %s | FileCheck -check-prefix=SI %s
+; RUN: llc -mtriple=amdgcn-- -mcpu=fiji < %s | FileCheck -check-prefix=VI %s
 
-declare i32 @llvm.amdgcn.workitem.id.x() #1
+define amdgpu_kernel void @test_fmin_legacy_uge_f64(double addrspace(1)* %out, double addrspace(1)* %in) #0 {
+; SI-LABEL: test_fmin_legacy_uge_f64:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s10, 0
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b64 s[8:9], s[2:3]
+; SI-NEXT:    v_mov_b32_e32 v1, 0
+; SI-NEXT:    buffer_load_dwordx4 v[0:3], v[0:1], s[8:11], 0 addr64
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cmp_nlt_f64_e32 vcc, v[0:1], v[2:3]
+; SI-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc
+; SI-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: test_fmin_legacy_uge_f64:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_cmp_nlt_f64_e32 vcc, v[0:1], v[2:3]
+; VI-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc
+; VI-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
+; VI-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; VI-NEXT:    s_endpgm
+  %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
+  %gep.0 = getelementptr double, double addrspace(1)* %in, i32 %tid
+  %gep.1 = getelementptr double, double addrspace(1)* %gep.0, i32 1
+
+  %a = load double, double addrspace(1)* %gep.0, align 8
+  %b = load double, double addrspace(1)* %gep.1, align 8
+
+  %cmp = fcmp uge double %a, %b
+  %val = select i1 %cmp, double %b, double %a
+  store double %val, double addrspace(1)* %out, align 8
+  ret void
+}
+
+define amdgpu_kernel void @test_fmin_legacy_ugt_f64(double addrspace(1)* %out, double addrspace(1)* %in) #0 {
+; SI-LABEL: test_fmin_legacy_ugt_f64:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s10, 0
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b64 s[8:9], s[2:3]
+; SI-NEXT:    v_mov_b32_e32 v1, 0
+; SI-NEXT:    buffer_load_dwordx4 v[0:3], v[0:1], s[8:11], 0 addr64
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cmp_nle_f64_e32 vcc, v[0:1], v[2:3]
+; SI-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc
+; SI-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: test_fmin_legacy_ugt_f64:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_cmp_nle_f64_e32 vcc, v[0:1], v[2:3]
+; VI-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc
+; VI-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
+; VI-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; VI-NEXT:    s_endpgm
+  %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
+  %gep.0 = getelementptr double, double addrspace(1)* %in, i32 %tid
+  %gep.1 = getelementptr double, double addrspace(1)* %gep.0, i32 1
+
+  %a = load double, double addrspace(1)* %gep.0, align 8
+  %b = load double, double addrspace(1)* %gep.1, align 8
 
-; FUNC-LABEL: @test_fmin_legacy_f64
-define amdgpu_kernel void @test_fmin_legacy_f64(<4 x double> addrspace(1)* %out, <4 x double> inreg %reg0) #0 {
-   %r0 = extractelement <4 x double> %reg0, i32 0
-   %r1 = extractelement <4 x double> %reg0, i32 1
-   %r2 = fcmp uge double %r0, %r1
-   %r3 = select i1 %r2, double %r1, double %r0
-   %vec = insertelement <4 x double> undef, double %r3, i32 0
-   store <4 x double> %vec, <4 x double> addrspace(1)* %out, align 16
-   ret void
+  %cmp = fcmp ugt double %a, %b
+  %val = select i1 %cmp, double %b, double %a
+  store double %val, double addrspace(1)* %out, align 8
+  ret void
 }
 
-; FUNC-LABEL: @test_fmin_legacy_ule_f64
 define amdgpu_kernel void @test_fmin_legacy_ule_f64(double addrspace(1)* %out, double addrspace(1)* %in) #0 {
+; SI-LABEL: test_fmin_legacy_ule_f64:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s10, 0
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b64 s[8:9], s[2:3]
+; SI-NEXT:    v_mov_b32_e32 v1, 0
+; SI-NEXT:    buffer_load_dwordx4 v[0:3], v[0:1], s[8:11], 0 addr64
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cmp_ngt_f64_e32 vcc, v[0:1], v[2:3]
+; SI-NEXT:    v_cndmask_b32_e32 v1, v3, v1, vcc
+; SI-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: test_fmin_legacy_ule_f64:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_cmp_ngt_f64_e32 vcc, v[0:1], v[2:3]
+; VI-NEXT:    v_cndmask_b32_e32 v1, v3, v1, vcc
+; VI-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
+; VI-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; VI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %gep.0 = getelementptr double, double addrspace(1)* %in, i32 %tid
   %gep.1 = getelementptr double, double addrspace(1)* %gep.0, i32 1
@@ -28,8 +158,45 @@ define amdgpu_kernel void @test_fmin_legacy_ule_f64(double addrspace(1)* %out, d
   ret void
 }
 
-; FUNC-LABEL: @test_fmin_legacy_ole_f64
-define amdgpu_kernel void @test_fmin_legacy_ole_f64(double addrspace(1)* %out, double addrspace(1)* %in) #0 {
+define amdgpu_kernel void @test_fmin_legacy_ult_f64(double addrspace(1)* %out, double addrspace(1)* %in) #0 {
+; SI-LABEL: test_fmin_legacy_ult_f64:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s10, 0
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b64 s[8:9], s[2:3]
+; SI-NEXT:    v_mov_b32_e32 v1, 0
+; SI-NEXT:    buffer_load_dwordx4 v[0:3], v[0:1], s[8:11], 0 addr64
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cmp_nge_f64_e32 vcc, v[0:1], v[2:3]
+; SI-NEXT:    v_cndmask_b32_e32 v1, v3, v1, vcc
+; SI-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: test_fmin_legacy_ult_f64:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_cmp_nge_f64_e32 vcc, v[0:1], v[2:3]
+; VI-NEXT:    v_cndmask_b32_e32 v1, v3, v1, vcc
+; VI-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
+; VI-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; VI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %gep.0 = getelementptr double, double addrspace(1)* %in, i32 %tid
   %gep.1 = getelementptr double, double addrspace(1)* %gep.0, i32 1
@@ -37,14 +204,51 @@ define amdgpu_kernel void @test_fmin_legacy_ole_f64(double addrspace(1)* %out, d
   %a = load double, double addrspace(1)* %gep.0, align 8
   %b = load double, double addrspace(1)* %gep.1, align 8
 
-  %cmp = fcmp ole double %a, %b
+  %cmp = fcmp ult double %a, %b
   %val = select i1 %cmp, double %a, double %b
   store double %val, double addrspace(1)* %out, align 8
   ret void
 }
 
-; FUNC-LABEL: @test_fmin_legacy_olt_f64
-define amdgpu_kernel void @test_fmin_legacy_olt_f64(double addrspace(1)* %out, double addrspace(1)* %in) #0 {
+define amdgpu_kernel void @test_fmin_legacy_oge_f64(double addrspace(1)* %out, double addrspace(1)* %in) #0 {
+; SI-LABEL: test_fmin_legacy_oge_f64:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s10, 0
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b64 s[8:9], s[2:3]
+; SI-NEXT:    v_mov_b32_e32 v1, 0
+; SI-NEXT:    buffer_load_dwordx4 v[0:3], v[0:1], s[8:11], 0 addr64
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cmp_ge_f64_e32 vcc, v[0:1], v[2:3]
+; SI-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc
+; SI-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: test_fmin_legacy_oge_f64:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_cmp_ge_f64_e32 vcc, v[0:1], v[2:3]
+; VI-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc
+; VI-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
+; VI-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; VI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %gep.0 = getelementptr double, double addrspace(1)* %in, i32 %tid
   %gep.1 = getelementptr double, double addrspace(1)* %gep.0, i32 1
@@ -52,14 +256,155 @@ define amdgpu_kernel void @test_fmin_legacy_olt_f64(double addrspace(1)* %out, d
   %a = load double, double addrspace(1)* %gep.0, align 8
   %b = load double, double addrspace(1)* %gep.1, align 8
 
-  %cmp = fcmp olt double %a, %b
+  %cmp = fcmp oge double %a, %b
+  %val = select i1 %cmp, double %b, double %a
+  store double %val, double addrspace(1)* %out, align 8
+  ret void
+}
+
+define amdgpu_kernel void @test_fmin_legacy_ogt_f64(double addrspace(1)* %out, double addrspace(1)* %in) #0 {
+; SI-LABEL: test_fmin_legacy_ogt_f64:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s10, 0
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b64 s[8:9], s[2:3]
+; SI-NEXT:    v_mov_b32_e32 v1, 0
+; SI-NEXT:    buffer_load_dwordx4 v[0:3], v[0:1], s[8:11], 0 addr64
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cmp_gt_f64_e32 vcc, v[0:1], v[2:3]
+; SI-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc
+; SI-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: test_fmin_legacy_ogt_f64:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_cmp_gt_f64_e32 vcc, v[0:1], v[2:3]
+; VI-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc
+; VI-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
+; VI-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; VI-NEXT:    s_endpgm
+  %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
+  %gep.0 = getelementptr double, double addrspace(1)* %in, i32 %tid
+  %gep.1 = getelementptr double, double addrspace(1)* %gep.0, i32 1
+
+  %a = load double, double addrspace(1)* %gep.0, align 8
+  %b = load double, double addrspace(1)* %gep.1, align 8
+
+  %cmp = fcmp ogt double %a, %b
+  %val = select i1 %cmp, double %b, double %a
+  store double %val, double addrspace(1)* %out, align 8
+  ret void
+}
+
+define amdgpu_kernel void @test_fmin_legacy_ole_f64(double addrspace(1)* %out, double addrspace(1)* %in) #0 {
+; SI-LABEL: test_fmin_legacy_ole_f64:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s10, 0
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b64 s[8:9], s[2:3]
+; SI-NEXT:    v_mov_b32_e32 v1, 0
+; SI-NEXT:    buffer_load_dwordx4 v[0:3], v[0:1], s[8:11], 0 addr64
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cmp_le_f64_e32 vcc, v[0:1], v[2:3]
+; SI-NEXT:    v_cndmask_b32_e32 v1, v3, v1, vcc
+; SI-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: test_fmin_legacy_ole_f64:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_cmp_le_f64_e32 vcc, v[0:1], v[2:3]
+; VI-NEXT:    v_cndmask_b32_e32 v1, v3, v1, vcc
+; VI-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
+; VI-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; VI-NEXT:    s_endpgm
+  %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
+  %gep.0 = getelementptr double, double addrspace(1)* %in, i32 %tid
+  %gep.1 = getelementptr double, double addrspace(1)* %gep.0, i32 1
+
+  %a = load double, double addrspace(1)* %gep.0, align 8
+  %b = load double, double addrspace(1)* %gep.1, align 8
+
+  %cmp = fcmp ole double %a, %b
   %val = select i1 %cmp, double %a, double %b
   store double %val, double addrspace(1)* %out, align 8
   ret void
 }
 
-; FUNC-LABEL: @test_fmin_legacy_ult_f64
-define amdgpu_kernel void @test_fmin_legacy_ult_f64(double addrspace(1)* %out, double addrspace(1)* %in) #0 {
+define amdgpu_kernel void @test_fmin_legacy_olt_f64(double addrspace(1)* %out, double addrspace(1)* %in) #0 {
+; SI-LABEL: test_fmin_legacy_olt_f64:
+; SI:       ; %bb.0:
+; SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; SI-NEXT:    s_mov_b32 s7, 0xf000
+; SI-NEXT:    s_mov_b32 s10, 0
+; SI-NEXT:    s_mov_b32 s11, s7
+; SI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; SI-NEXT:    s_waitcnt lgkmcnt(0)
+; SI-NEXT:    s_mov_b64 s[8:9], s[2:3]
+; SI-NEXT:    v_mov_b32_e32 v1, 0
+; SI-NEXT:    buffer_load_dwordx4 v[0:3], v[0:1], s[8:11], 0 addr64
+; SI-NEXT:    s_mov_b32 s6, -1
+; SI-NEXT:    s_mov_b32 s4, s0
+; SI-NEXT:    s_mov_b32 s5, s1
+; SI-NEXT:    s_waitcnt vmcnt(0)
+; SI-NEXT:    v_cmp_lt_f64_e32 vcc, v[0:1], v[2:3]
+; SI-NEXT:    v_cndmask_b32_e32 v1, v3, v1, vcc
+; SI-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
+; SI-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; SI-NEXT:    s_endpgm
+;
+; VI-LABEL: test_fmin_legacy_olt_f64:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
+; VI-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VI-NEXT:    v_cmp_lt_f64_e32 vcc, v[0:1], v[2:3]
+; VI-NEXT:    v_cndmask_b32_e32 v1, v3, v1, vcc
+; VI-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
+; VI-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; VI-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %gep.0 = getelementptr double, double addrspace(1)* %in, i32 %tid
   %gep.1 = getelementptr double, double addrspace(1)* %gep.0, i32 1
@@ -67,11 +412,13 @@ define amdgpu_kernel void @test_fmin_legacy_ult_f64(double addrspace(1)* %out, d
   %a = load double, double addrspace(1)* %gep.0, align 8
   %b = load double, double addrspace(1)* %gep.1, align 8
 
-  %cmp = fcmp ult double %a, %b
+  %cmp = fcmp olt double %a, %b
   %val = select i1 %cmp, double %a, double %b
   store double %val, double addrspace(1)* %out, align 8
   ret void
 }
 
+declare i32 @llvm.amdgcn.workitem.id.x() #1
+
 attributes #0 = { nounwind }
 attributes #1 = { nounwind readnone }
diff --git a/llvm/test/CodeGen/AMDGPU/fmin_legacy.ll b/llvm/test/CodeGen/AMDGPU/fmin_legacy.ll
index a354bf375e9..e0acbaf59db 100644
--- a/llvm/test/CodeGen/AMDGPU/fmin_legacy.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmin_legacy.ll
@@ -1,6 +1,6 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI-SAFE -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -enable-no-nans-fp-math -enable-no-signed-zeros-fp-math  -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI-NONAN -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN-SAFE -check-prefix=GCN -check-prefix=FUNC %s
+; RUN: llc -enable-no-nans-fp-math -enable-no-signed-zeros-fp-math -march=amdgcn -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN-NONAN -check-prefix=GCN -check-prefix=FUNC %s
+; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -enable-var-scope -check-prefix=EG -check-prefix=FUNC %s
 
 declare i32 @llvm.r600.read.tidig.x() #1
 
@@ -10,25 +10,24 @@ declare i32 @llvm.r600.read.tidig.x() #1
 
 ; FUNC-LABEL: {{^}}s_test_fmin_legacy_subreg_inputs_f32:
 ; EG: MIN *
-; SI-SAFE: v_min_legacy_f32_e32 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}
-; SI-NONAN: v_min_f32_e32 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}
-define amdgpu_kernel void @s_test_fmin_legacy_subreg_inputs_f32(<4 x float> addrspace(1)* %out, <4 x float> inreg %reg0) #0 {
+; GCN-SAFE: v_min_legacy_f32_e32 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}
+; GCN-NONAN: v_min_f32_e32 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}
+define amdgpu_kernel void @s_test_fmin_legacy_subreg_inputs_f32(float addrspace(1)* %out, <4 x float> %reg0) #0 {
    %r0 = extractelement <4 x float> %reg0, i32 0
    %r1 = extractelement <4 x float> %reg0, i32 1
    %r2 = fcmp uge float %r0, %r1
    %r3 = select i1 %r2, float %r1, float %r0
-   %vec = insertelement <4 x float> undef, float %r3, i32 0
-   store <4 x float> %vec, <4 x float> addrspace(1)* %out, align 16
+   store float %r3, float addrspace(1)* %out
    ret void
 }
 
 ; FUNC-LABEL: {{^}}s_test_fmin_legacy_ule_f32:
-; SI-DAG: s_load_dwordx2 s{{\[}}[[A:[0-9]+]]:[[B:[0-9]+]]{{\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0xb
+; GCN-DAG: s_load_dwordx2 s{{\[}}[[A:[0-9]+]]:[[B:[0-9]+]]{{\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0xb
 
-; SI-DAG: v_mov_b32_e32 [[VB:v[0-9]+]], s[[B]]
+; GCN-DAG: v_mov_b32_e32 [[VB:v[0-9]+]], s[[B]]
 
-; SI-SAFE: v_min_legacy_f32_e64 {{v[0-9]+}}, [[VB]], s[[A]]
-; SI-NONAN: v_min_f32_e32 {{v[0-9]+}}, s[[A]], [[VB]]
+; GCN-SAFE: v_min_legacy_f32_e64 {{v[0-9]+}}, [[VB]], s[[A]]
+; GCN-NONAN: v_min_f32_e32 {{v[0-9]+}}, s[[A]], [[VB]]
 
 define amdgpu_kernel void @s_test_fmin_legacy_ule_f32(float addrspace(1)* %out, float %a, float %b) #0 {
   %cmp = fcmp ule float %a, %b
@@ -37,11 +36,28 @@ define amdgpu_kernel void @s_test_fmin_legacy_ule_f32(float addrspace(1)* %out,
   ret void
 }
 
-; FUNC-LABEL: @test_fmin_legacy_ule_f32
-; SI: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
-; SI: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
-; SI-SAFE: v_min_legacy_f32_e32 {{v[0-9]+}}, [[B]], [[A]]
-; SI-NONAN: v_min_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
+; GCN-LABEL: {{^}}s_test_fmin_legacy_ule_f32_nnan_src:
+; GCN: s_load_dwordx2 s{{\[}}[[A:[0-9]+]]:[[B:[0-9]+]]{{\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0xb
+
+; GCN-DAG: v_add_f32_e64 [[ADD_A:v[0-9]+]], s[[A]], 1.0
+; GCN-DAG: v_add_f32_e64 [[ADD_B:v[0-9]+]], s[[B]], 2.0
+
+; GCN-SAFE: v_min_legacy_f32_e32 {{v[0-9]+}}, [[ADD_B]], [[ADD_A]]
+; GCN-NONAN: v_min_f32_e32 {{v[0-9]+}}, [[ADD_A]], [[ADD_B]]
+define amdgpu_kernel void @s_test_fmin_legacy_ule_f32_nnan_src(float addrspace(1)* %out, float %a, float %b) #0 {
+  %a.nnan = fadd nnan float %a, 1.0
+  %b.nnan = fadd nnan float %b, 2.0
+  %cmp = fcmp ule float %a.nnan, %b.nnan
+  %val = select i1 %cmp, float %a.nnan, float %b.nnan
+  store float %val, float addrspace(1)* %out, align 4
+  ret void
+}
+
+; FUNC-LABEL: {{^}}test_fmin_legacy_ule_f32:
+; GCN: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
+; GCN: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
+; GCN-SAFE: v_min_legacy_f32_e32 {{v[0-9]+}}, [[B]], [[A]]
+; GCN-NONAN: v_min_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
 define amdgpu_kernel void @test_fmin_legacy_ule_f32(float addrspace(1)* %out, float addrspace(1)* %in) #0 {
   %tid = call i32 @llvm.r600.read.tidig.x() #1
   %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
@@ -56,11 +72,11 @@ define amdgpu_kernel void @test_fmin_legacy_ule_f32(float addrspace(1)* %out, fl
   ret void
 }
 
-; FUNC-LABEL: @test_fmin_legacy_ole_f32
-; SI: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
-; SI: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
-; SI-SAFE: v_min_legacy_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
-; SI-NONAN: v_min_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
+; FUNC-LABEL: {{^}}test_fmin_legacy_ole_f32:
+; GCN: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
+; GCN: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
+; GCN-SAFE: v_min_legacy_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
+; GCN-NONAN: v_min_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
 define amdgpu_kernel void @test_fmin_legacy_ole_f32(float addrspace(1)* %out, float addrspace(1)* %in) #0 {
   %tid = call i32 @llvm.r600.read.tidig.x() #1
   %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
@@ -75,11 +91,11 @@ define amdgpu_kernel void @test_fmin_legacy_ole_f32(float addrspace(1)* %out, fl
   ret void
 }
 
-; FUNC-LABEL: @test_fmin_legacy_olt_f32
-; SI: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
-; SI: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
-; SI-SAFE: v_min_legacy_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
-; SI-NONAN: v_min_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
+; FUNC-LABEL: {{^}}test_fmin_legacy_olt_f32:
+; GCN: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
+; GCN: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
+; GCN-SAFE: v_min_legacy_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
+; GCN-NONAN: v_min_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
 define amdgpu_kernel void @test_fmin_legacy_olt_f32(float addrspace(1)* %out, float addrspace(1)* %in) #0 {
   %tid = call i32 @llvm.r600.read.tidig.x() #1
   %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
@@ -94,11 +110,11 @@ define amdgpu_kernel void @test_fmin_legacy_olt_f32(float addrspace(1)* %out, fl
   ret void
 }
 
-; FUNC-LABEL: @test_fmin_legacy_ult_f32
-; SI: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
-; SI: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
-; SI-SAFE: v_min_legacy_f32_e32 {{v[0-9]+}}, [[B]], [[A]]
-; SI-NONAN: v_min_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
+; FUNC-LABEL: {{^}}test_fmin_legacy_ult_f32:
+; GCN: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
+; GCN: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
+; GCN-SAFE: v_min_legacy_f32_e32 {{v[0-9]+}}, [[B]], [[A]]
+; GCN-NONAN: v_min_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
 define amdgpu_kernel void @test_fmin_legacy_ult_f32(float addrspace(1)* %out, float addrspace(1)* %in) #0 {
   %tid = call i32 @llvm.r600.read.tidig.x() #1
   %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid
@@ -114,10 +130,10 @@ define amdgpu_kernel void @test_fmin_legacy_ult_f32(float addrspace(1)* %out, fl
 }
 
 ; FUNC-LABEL: {{^}}test_fmin_legacy_ult_v1f32:
-; SI: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
-; SI: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
-; SI-SAFE: v_min_legacy_f32_e32 {{v[0-9]+}}, [[B]], [[A]]
-; SI-NONAN: v_min_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
+; GCN: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
+; GCN: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
+; GCN-SAFE: v_min_legacy_f32_e32 {{v[0-9]+}}, [[B]], [[A]]
+; GCN-NONAN: v_min_f32_e32 {{v[0-9]+}}, [[A]], [[B]]
 define amdgpu_kernel void @test_fmin_legacy_ult_v1f32(<1 x float> addrspace(1)* %out, <1 x float> addrspace(1)* %in) #0 {
   %tid = call i32 @llvm.r600.read.tidig.x() #1
   %gep.0 = getelementptr <1 x float>, <1 x float> addrspace(1)* %in, i32 %tid
@@ -133,13 +149,13 @@ define amdgpu_kernel void @test_fmin_legacy_ult_v1f32(<1 x float> addrspace(1)*
 }
 
 ; FUNC-LABEL: {{^}}test_fmin_legacy_ult_v2f32:
-; SI: buffer_load_dwordx2
-; SI: buffer_load_dwordx2
-; SI-SAFE: v_min_legacy_f32_e32
-; SI-SAFE: v_min_legacy_f32_e32
+; GCN: buffer_load_dwordx2
+; GCN: buffer_load_dwordx2
+; GCN-SAFE: v_min_legacy_f32_e32
+; GCN-SAFE: v_min_legacy_f32_e32
 
-; SI-NONAN: v_min_f32_e32
-; SI-NONAN: v_min_f32_e32
+; GCN-NONAN: v_min_f32_e32
+; GCN-NONAN: v_min_f32_e32
 define amdgpu_kernel void @test_fmin_legacy_ult_v2f32(<2 x float> addrspace(1)* %out, <2 x float> addrspace(1)* %in) #0 {
   %tid = call i32 @llvm.r600.read.tidig.x() #1
   %gep.0 = getelementptr <2 x float>, <2 x float> addrspace(1)* %in, i32 %tid
@@ -155,13 +171,13 @@ define amdgpu_kernel void @test_fmin_legacy_ult_v2f32(<2 x float> addrspace(1)*
 }
 
 ; FUNC-LABEL: {{^}}test_fmin_legacy_ult_v3f32:
-; SI-SAFE: v_min_legacy_f32_e32
-; SI-SAFE: v_min_legacy_f32_e32
-; SI-SAFE: v_min_legacy_f32_e32
+; GCN-SAFE: v_min_legacy_f32_e32
+; GCN-SAFE: v_min_legacy_f32_e32
+; GCN-SAFE: v_min_legacy_f32_e32
 
-; SI-NONAN: v_min_f32_e32
-; SI-NONAN: v_min_f32_e32
-; SI-NONAN: v_min_f32_e32
+; GCN-NONAN: v_min_f32_e32
+; GCN-NONAN: v_min_f32_e32
+; GCN-NONAN: v_min_f32_e32
 define amdgpu_kernel void @test_fmin_legacy_ult_v3f32(<3 x float> addrspace(1)* %out, <3 x float> addrspace(1)* %in) #0 {
   %tid = call i32 @llvm.r600.read.tidig.x() #1
   %gep.0 = getelementptr <3 x float>, <3 x float> addrspace(1)* %in, i32 %tid
@@ -176,14 +192,14 @@ define amdgpu_kernel void @test_fmin_legacy_ult_v3f32(<3 x float> addrspace(1)*
   ret void
 }
 
-; FUNC-LABEL: @test_fmin_legacy_ole_f32_multi_use
-; SI: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
-; SI: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
-; SI-NOT: v_min
-; SI: v_cmp_le_f32
-; SI-NEXT: v_cndmask_b32
-; SI-NOT: v_min
-; SI: s_endpgm
+; FUNC-LABEL: {{^}}test_fmin_legacy_ole_f32_multi_use:
+; GCN: buffer_load_dword [[A:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64{{$}}
+; GCN: buffer_load_dword [[B:v[0-9]+]], {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 addr64 offset:4
+; GCN-NOT: v_min
+; GCN: v_cmp_le_f32
+; GCN-NEXT: v_cndmask_b32
+; GCN-NOT: v_min
+; GCN: s_endpgm
 define amdgpu_kernel void @test_fmin_legacy_ole_f32_multi_use(float addrspace(1)* %out0, i1 addrspace(1)* %out1, float addrspace(1)* %in) #0 {
   %tid = call i32 @llvm.r600.read.tidig.x() #1
   %gep.0 = getelementptr float, float addrspace(1)* %in, i32 %tid