diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fcmp.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fcmp.ll
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fcmp.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fcmp.ll
@@ -1,5 +1,6 @@
-; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,SI %s
-; RUN: llc -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,VI %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX %s
+; RUN: llc -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,VI %s
 
 declare i64 @llvm.amdgcn.fcmp.f32(float, float, i32) #0
 declare i64 @llvm.amdgcn.fcmp.f64(double, double, i32) #0
@@ -8,18 +9,68 @@
 declare i64 @llvm.amdgcn.fcmp.f16(half, half, i32) #0
 declare half @llvm.fabs.f16(half) #0
 
-; GCN-LABEL: {{^}}v_fcmp_f32_oeq_with_fabs:
-; GCN: v_cmp_eq_f32_e64 {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}}, |{{v[0-9]+}}|
 define amdgpu_kernel void @v_fcmp_f32_oeq_with_fabs(i64 addrspace(1)* %out, float %src, float %a) {
+; GFX-LABEL: v_fcmp_f32_oeq_with_fabs:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s7, 0xf000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_mov_b32_e32 v0, s3
+; GFX-NEXT:    v_cmp_eq_f32_e64 s[2:3], s2, |v0|
+; GFX-NEXT:    s_mov_b32 s6, -1
+; GFX-NEXT:    s_mov_b32 s4, s0
+; GFX-NEXT:    s_mov_b32 s5, s1
+; GFX-NEXT:    v_mov_b32_e32 v0, s2
+; GFX-NEXT:    v_mov_b32_e32 v1, s3
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_oeq_with_fabs:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s3
+; VI-NEXT:    v_cmp_eq_f32_e64 s[2:3], s2, |v0|
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %temp = call float @llvm.fabs.f32(float %a)
   %result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float %temp, i32 1)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f32_oeq_both_operands_with_fabs:
-; GCN: v_cmp_eq_f32_e64 {{s\[[0-9]+:[0-9]+\]}}, |{{s[0-9]+}}|, |{{v[0-9]+}}|
 define amdgpu_kernel void @v_fcmp_f32_oeq_both_operands_with_fabs(i64 addrspace(1)* %out, float %src, float %a) {
+; GFX-LABEL: v_fcmp_f32_oeq_both_operands_with_fabs:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s7, 0xf000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_mov_b32_e32 v0, s3
+; GFX-NEXT:    v_cmp_eq_f32_e64 s[2:3], |s2|, |v0|
+; GFX-NEXT:    s_mov_b32 s6, -1
+; GFX-NEXT:    s_mov_b32 s4, s0
+; GFX-NEXT:    s_mov_b32 s5, s1
+; GFX-NEXT:    v_mov_b32_e32 v0, s2
+; GFX-NEXT:    v_mov_b32_e32 v1, s3
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_oeq_both_operands_with_fabs:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s3
+; VI-NEXT:    v_cmp_eq_f32_e64 s[2:3], |s2|, |v0|
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %temp = call float @llvm.fabs.f32(float %a)
   %src_input = call float @llvm.fabs.f32(float %src)
   %result = call i64 @llvm.amdgcn.fcmp.f32(float %src_input, float %temp, i32 1)
@@ -27,227 +78,902 @@
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f32:
-; GCN-NOT: v_cmp_eq_f32_e64
 define amdgpu_kernel void @v_fcmp_f32(i64 addrspace(1)* %out, float %src) {
+; GCN-LABEL: v_fcmp_f32:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 -1)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f32_oeq:
-; GCN: v_cmp_eq_f32_e64
 define amdgpu_kernel void @v_fcmp_f32_oeq(i64 addrspace(1)* %out, float %src) {
+; GFX-LABEL: v_fcmp_f32_oeq:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_eq_f32_e64 s[4:5], s2, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_oeq:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_eq_f32_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 1)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f32_one:
-; GCN: v_cmp_neq_f32_e64
 define amdgpu_kernel void @v_fcmp_f32_one(i64 addrspace(1)* %out, float %src) {
+; GFX-LABEL: v_fcmp_f32_one:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_neq_f32_e64 s[4:5], s2, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_one:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_neq_f32_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 6)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f32_ogt:
-; GCN: v_cmp_gt_f32_e64
 define amdgpu_kernel void @v_fcmp_f32_ogt(i64 addrspace(1)* %out, float %src) {
+; GFX-LABEL: v_fcmp_f32_ogt:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_gt_f32_e64 s[4:5], s2, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_ogt:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_gt_f32_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 2)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f32_oge:
-; GCN: v_cmp_ge_f32_e64
 define amdgpu_kernel void @v_fcmp_f32_oge(i64 addrspace(1)* %out, float %src) {
+; GFX-LABEL: v_fcmp_f32_oge:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_ge_f32_e64 s[4:5], s2, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_oge:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_ge_f32_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 3)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f32_olt:
-; GCN: v_cmp_lt_f32_e64
 define amdgpu_kernel void @v_fcmp_f32_olt(i64 addrspace(1)* %out, float %src) {
+; GFX-LABEL: v_fcmp_f32_olt:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_lt_f32_e64 s[4:5], s2, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_olt:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_lt_f32_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 4)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f32_ole:
-; GCN: v_cmp_le_f32_e64
 define amdgpu_kernel void @v_fcmp_f32_ole(i64 addrspace(1)* %out, float %src) {
+; GFX-LABEL: v_fcmp_f32_ole:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_le_f32_e64 s[4:5], s2, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_ole:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_le_f32_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 5)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
 
-; GCN-LABEL: {{^}}v_fcmp_f32_ueq:
-; GCN: v_cmp_nlg_f32_e64
 define amdgpu_kernel void @v_fcmp_f32_ueq(i64 addrspace(1)* %out, float %src) {
+; GFX-LABEL: v_fcmp_f32_ueq:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_nlg_f32_e64 s[4:5], s2, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_ueq:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_nlg_f32_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 9)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f32_une:
-; GCN: v_cmp_neq_f32_e64
 define amdgpu_kernel void @v_fcmp_f32_une(i64 addrspace(1)* %out, float %src) {
+; GFX-LABEL: v_fcmp_f32_une:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_neq_f32_e64 s[4:5], s2, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_une:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_neq_f32_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 14)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f32_ugt:
-; GCN: v_cmp_nle_f32_e64
 define amdgpu_kernel void @v_fcmp_f32_ugt(i64 addrspace(1)* %out, float %src) {
+; GFX-LABEL: v_fcmp_f32_ugt:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_nle_f32_e64 s[4:5], s2, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_ugt:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_nle_f32_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 10)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f32_uge:
-; GCN: v_cmp_nlt_f32_e64
 define amdgpu_kernel void @v_fcmp_f32_uge(i64 addrspace(1)* %out, float %src) {
+; GFX-LABEL: v_fcmp_f32_uge:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_nlt_f32_e64 s[4:5], s2, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_uge:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_nlt_f32_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 11)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f32_ult:
-; GCN: v_cmp_nge_f32_e64
 define amdgpu_kernel void @v_fcmp_f32_ult(i64 addrspace(1)* %out, float %src) {
+; GFX-LABEL: v_fcmp_f32_ult:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_nge_f32_e64 s[4:5], s2, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_ult:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_nge_f32_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 12)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f32_ule:
-; GCN: v_cmp_ngt_f32_e64
 define amdgpu_kernel void @v_fcmp_f32_ule(i64 addrspace(1)* %out, float %src) {
+; GFX-LABEL: v_fcmp_f32_ule:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_ngt_f32_e64 s[4:5], s2, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f32_ule:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x42c80000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_ngt_f32_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f32(float %src, float 100.00, i32 13)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f64_oeq:
-; GCN: v_cmp_eq_f64_e64
 define amdgpu_kernel void @v_fcmp_f64_oeq(i64 addrspace(1)* %out, double %src) {
+; GFX-LABEL: v_fcmp_f64_oeq:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s7, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0
+; GFX-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_eq_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX-NEXT:    s_mov_b32 s6, -1
+; GFX-NEXT:    s_mov_b32 s4, s0
+; GFX-NEXT:    s_mov_b32 s5, s1
+; GFX-NEXT:    v_mov_b32_e32 v0, s2
+; GFX-NEXT:    v_mov_b32_e32 v1, s3
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_oeq:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0
+; VI-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_eq_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 1)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f64_one:
-; GCN: v_cmp_neq_f64_e64
 define amdgpu_kernel void @v_fcmp_f64_one(i64 addrspace(1)* %out, double %src) {
+; GFX-LABEL: v_fcmp_f64_one:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s7, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0
+; GFX-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_neq_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX-NEXT:    s_mov_b32 s6, -1
+; GFX-NEXT:    s_mov_b32 s4, s0
+; GFX-NEXT:    s_mov_b32 s5, s1
+; GFX-NEXT:    v_mov_b32_e32 v0, s2
+; GFX-NEXT:    v_mov_b32_e32 v1, s3
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_one:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0
+; VI-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_neq_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 6)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f64_ogt:
-; GCN: v_cmp_gt_f64_e64
 define amdgpu_kernel void @v_fcmp_f64_ogt(i64 addrspace(1)* %out, double %src) {
+; GFX-LABEL: v_fcmp_f64_ogt:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s7, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0
+; GFX-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_gt_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX-NEXT:    s_mov_b32 s6, -1
+; GFX-NEXT:    s_mov_b32 s4, s0
+; GFX-NEXT:    s_mov_b32 s5, s1
+; GFX-NEXT:    v_mov_b32_e32 v0, s2
+; GFX-NEXT:    v_mov_b32_e32 v1, s3
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_ogt:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0
+; VI-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_gt_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 2)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f64_oge:
-; GCN: v_cmp_ge_f64_e64
 define amdgpu_kernel void @v_fcmp_f64_oge(i64 addrspace(1)* %out, double %src) {
+; GFX-LABEL: v_fcmp_f64_oge:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s7, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0
+; GFX-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_ge_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX-NEXT:    s_mov_b32 s6, -1
+; GFX-NEXT:    s_mov_b32 s4, s0
+; GFX-NEXT:    s_mov_b32 s5, s1
+; GFX-NEXT:    v_mov_b32_e32 v0, s2
+; GFX-NEXT:    v_mov_b32_e32 v1, s3
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_oge:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0
+; VI-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_ge_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 3)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f64_olt:
-; GCN: v_cmp_lt_f64_e64
 define amdgpu_kernel void @v_fcmp_f64_olt(i64 addrspace(1)* %out, double %src) {
+; GFX-LABEL: v_fcmp_f64_olt:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s7, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0
+; GFX-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_lt_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX-NEXT:    s_mov_b32 s6, -1
+; GFX-NEXT:    s_mov_b32 s4, s0
+; GFX-NEXT:    s_mov_b32 s5, s1
+; GFX-NEXT:    v_mov_b32_e32 v0, s2
+; GFX-NEXT:    v_mov_b32_e32 v1, s3
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_olt:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0
+; VI-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_lt_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 4)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f64_ole:
-; GCN: v_cmp_le_f64_e64
 define amdgpu_kernel void @v_fcmp_f64_ole(i64 addrspace(1)* %out, double %src) {
+; GFX-LABEL: v_fcmp_f64_ole:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s7, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0
+; GFX-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_le_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX-NEXT:    s_mov_b32 s6, -1
+; GFX-NEXT:    s_mov_b32 s4, s0
+; GFX-NEXT:    s_mov_b32 s5, s1
+; GFX-NEXT:    v_mov_b32_e32 v0, s2
+; GFX-NEXT:    v_mov_b32_e32 v1, s3
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_ole:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0
+; VI-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_le_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 5)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f64_ueq:
-; GCN: v_cmp_nlg_f64_e64
 define amdgpu_kernel void @v_fcmp_f64_ueq(i64 addrspace(1)* %out, double %src) {
+; GFX-LABEL: v_fcmp_f64_ueq:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s7, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0
+; GFX-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_nlg_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX-NEXT:    s_mov_b32 s6, -1
+; GFX-NEXT:    s_mov_b32 s4, s0
+; GFX-NEXT:    s_mov_b32 s5, s1
+; GFX-NEXT:    v_mov_b32_e32 v0, s2
+; GFX-NEXT:    v_mov_b32_e32 v1, s3
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_ueq:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0
+; VI-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_nlg_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 9)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f64_une:
-; GCN: v_cmp_neq_f64_e64
 define amdgpu_kernel void @v_fcmp_f64_une(i64 addrspace(1)* %out, double %src) {
+; GFX-LABEL: v_fcmp_f64_une:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s7, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0
+; GFX-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_neq_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX-NEXT:    s_mov_b32 s6, -1
+; GFX-NEXT:    s_mov_b32 s4, s0
+; GFX-NEXT:    s_mov_b32 s5, s1
+; GFX-NEXT:    v_mov_b32_e32 v0, s2
+; GFX-NEXT:    v_mov_b32_e32 v1, s3
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_une:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0
+; VI-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_neq_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 14)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f64_ugt:
-; GCN: v_cmp_nle_f64_e64
 define amdgpu_kernel void @v_fcmp_f64_ugt(i64 addrspace(1)* %out, double %src) {
+; GFX-LABEL: v_fcmp_f64_ugt:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s7, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0
+; GFX-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_nle_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX-NEXT:    s_mov_b32 s6, -1
+; GFX-NEXT:    s_mov_b32 s4, s0
+; GFX-NEXT:    s_mov_b32 s5, s1
+; GFX-NEXT:    v_mov_b32_e32 v0, s2
+; GFX-NEXT:    v_mov_b32_e32 v1, s3
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_ugt:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0
+; VI-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_nle_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 10)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f64_uge:
-; GCN: v_cmp_nlt_f64_e64
 define amdgpu_kernel void @v_fcmp_f64_uge(i64 addrspace(1)* %out, double %src) {
+; GFX-LABEL: v_fcmp_f64_uge:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s7, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0
+; GFX-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_nlt_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX-NEXT:    s_mov_b32 s6, -1
+; GFX-NEXT:    s_mov_b32 s4, s0
+; GFX-NEXT:    s_mov_b32 s5, s1
+; GFX-NEXT:    v_mov_b32_e32 v0, s2
+; GFX-NEXT:    v_mov_b32_e32 v1, s3
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_uge:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0
+; VI-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_nlt_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 11)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f64_ult:
-; GCN: v_cmp_nge_f64_e64
 define amdgpu_kernel void @v_fcmp_f64_ult(i64 addrspace(1)* %out, double %src) {
+; GFX-LABEL: v_fcmp_f64_ult:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s7, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0
+; GFX-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_nge_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX-NEXT:    s_mov_b32 s6, -1
+; GFX-NEXT:    s_mov_b32 s4, s0
+; GFX-NEXT:    s_mov_b32 s5, s1
+; GFX-NEXT:    v_mov_b32_e32 v0, s2
+; GFX-NEXT:    v_mov_b32_e32 v1, s3
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_ult:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0
+; VI-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_nge_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 12)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f64_ule:
-; GCN: v_cmp_ngt_f64_e64
 define amdgpu_kernel void @v_fcmp_f64_ule(i64 addrspace(1)* %out, double %src) {
+; GFX-LABEL: v_fcmp_f64_ule:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s7, 0xf000
+; GFX-NEXT:    v_mov_b32_e32 v0, 0
+; GFX-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cmp_ngt_f64_e64 s[2:3], s[2:3], v[0:1]
+; GFX-NEXT:    s_mov_b32 s6, -1
+; GFX-NEXT:    s_mov_b32 s4, s0
+; GFX-NEXT:    s_mov_b32 s5, s1
+; GFX-NEXT:    v_mov_b32_e32 v0, s2
+; GFX-NEXT:    v_mov_b32_e32 v1, s3
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f64_ule:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0
+; VI-NEXT:    v_mov_b32_e32 v1, 0x40590000
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_ngt_f64_e64 s[2:3], s[2:3], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f64(double %src, double 100.00, i32 13)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f16_oeq_with_fabs:
-; VI: v_cmp_eq_f16_e64 {{s\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}}, |{{v[0-9]+}}|
 
-; SI: v_cvt_f32_f16_e32 [[CVT0:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cvt_f32_f16_e64 [[CVT1:v[0-9]+]], |s{{[0-9]+}}|
-; SI: v_cmp_eq_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[CVT0]], [[CVT1]]
 define amdgpu_kernel void @v_fcmp_f16_oeq_with_fabs(i64 addrspace(1)* %out, half %src, half %a) {
+; GFX-LABEL: v_fcmp_f16_oeq_with_fabs:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cvt_f32_f16_e32 v0, s2
+; GFX-NEXT:    s_lshr_b32 s2, s2, 16
+; GFX-NEXT:    v_cvt_f32_f16_e64 v1, |s2|
+; GFX-NEXT:    v_cmp_eq_f32_e64 s[4:5], v0, v1
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_oeq_with_fabs:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_lshr_b32 s3, s2, 16
+; VI-NEXT:    v_mov_b32_e32 v0, s3
+; VI-NEXT:    v_cmp_eq_f16_e64 s[2:3], s2, |v0|
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %temp = call half @llvm.fabs.f16(half %a)
   %result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half %temp, i32 1)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f16_oeq_both_operands_with_fabs:
-; VI: v_cmp_eq_f16_e64 {{s\[[0-9]+:[0-9]+\]}}, |{{s[0-9]+}}|, |{{v[0-9]+}}|
 
-; SI: v_cvt_f32_f16_e64 [[CVT0:v[0-9]+]], |s{{[0-9]+}}|
-; SI: v_cvt_f32_f16_e64 [[CVT1:v[0-9]+]], |s{{[0-9]+}}|
-; SI: v_cmp_eq_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[CVT0]], [[CVT1]]
 define amdgpu_kernel void @v_fcmp_f16_oeq_both_operands_with_fabs(i64 addrspace(1)* %out, half %src, half %a) {
+; GFX-LABEL: v_fcmp_f16_oeq_both_operands_with_fabs:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cvt_f32_f16_e64 v0, |s2|
+; GFX-NEXT:    s_lshr_b32 s2, s2, 16
+; GFX-NEXT:    v_cvt_f32_f16_e64 v1, |s2|
+; GFX-NEXT:    v_cmp_eq_f32_e64 s[4:5], v0, v1
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_oeq_both_operands_with_fabs:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    s_lshr_b32 s3, s2, 16
+; VI-NEXT:    v_mov_b32_e32 v0, s3
+; VI-NEXT:    v_cmp_eq_f16_e64 s[2:3], |s2|, |v0|
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %temp = call half @llvm.fabs.f16(half %a)
   %src_input = call half @llvm.fabs.f16(half %src)
   %result = call i64 @llvm.amdgcn.fcmp.f16(half %src_input, half %temp, i32 1)
@@ -255,153 +981,430 @@
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f16:
-; GCN-NOT: v_cmp_eq_
 define amdgpu_kernel void @v_fcmp_f16(i64 addrspace(1)* %out, half %src) {
+; GCN-LABEL: v_fcmp_f16:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 -1)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f16_oeq:
-; VI: v_cmp_eq_f16_e64
 
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_eq_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
 define amdgpu_kernel void @v_fcmp_f16_oeq(i64 addrspace(1)* %out, half %src) {
+; GFX-LABEL: v_fcmp_f16_oeq:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    s_mov_b32 s4, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cvt_f32_f16_e32 v0, s2
+; GFX-NEXT:    v_cmp_eq_f32_e64 s[4:5], s4, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_oeq:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x5640
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_eq_f16_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 1)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f16_one:
-; VI: v_cmp_neq_f16_e64
 
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_neq_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
 define amdgpu_kernel void @v_fcmp_f16_one(i64 addrspace(1)* %out, half %src) {
+; GFX-LABEL: v_fcmp_f16_one:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    s_mov_b32 s4, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cvt_f32_f16_e32 v0, s2
+; GFX-NEXT:    v_cmp_neq_f32_e64 s[4:5], s4, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_one:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x5640
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_neq_f16_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 6)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f16_ogt:
-; VI: v_cmp_gt_f16_e64
 
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_lt_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
 define amdgpu_kernel void @v_fcmp_f16_ogt(i64 addrspace(1)* %out, half %src) {
+; GFX-LABEL: v_fcmp_f16_ogt:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    s_mov_b32 s4, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cvt_f32_f16_e32 v0, s2
+; GFX-NEXT:    v_cmp_lt_f32_e64 s[4:5], s4, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_ogt:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x5640
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_gt_f16_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 2)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f16_oge:
-; VI: v_cmp_ge_f16_e64
 
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_le_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
 define amdgpu_kernel void @v_fcmp_f16_oge(i64 addrspace(1)* %out, half %src) {
+; GFX-LABEL: v_fcmp_f16_oge:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    s_mov_b32 s4, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cvt_f32_f16_e32 v0, s2
+; GFX-NEXT:    v_cmp_le_f32_e64 s[4:5], s4, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_oge:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x5640
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_ge_f16_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 3)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f16_olt:
-; VI: v_cmp_lt_f16_e64
 
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_gt_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
 define amdgpu_kernel void @v_fcmp_f16_olt(i64 addrspace(1)* %out, half %src) {
+; GFX-LABEL: v_fcmp_f16_olt:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    s_mov_b32 s4, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cvt_f32_f16_e32 v0, s2
+; GFX-NEXT:    v_cmp_gt_f32_e64 s[4:5], s4, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_olt:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x5640
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_lt_f16_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 4)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f16_ole:
-; VI: v_cmp_le_f16_e64
 
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_ge_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
 define amdgpu_kernel void @v_fcmp_f16_ole(i64 addrspace(1)* %out, half %src) {
+; GFX-LABEL: v_fcmp_f16_ole:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    s_mov_b32 s4, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cvt_f32_f16_e32 v0, s2
+; GFX-NEXT:    v_cmp_ge_f32_e64 s[4:5], s4, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_ole:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x5640
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_le_f16_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 5)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f16_ueq:
-; VI: v_cmp_nlg_f16_e64
 
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_nlg_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
 define amdgpu_kernel void @v_fcmp_f16_ueq(i64 addrspace(1)* %out, half %src) {
+; GFX-LABEL: v_fcmp_f16_ueq:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    s_mov_b32 s4, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cvt_f32_f16_e32 v0, s2
+; GFX-NEXT:    v_cmp_nlg_f32_e64 s[4:5], s4, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_ueq:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x5640
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_nlg_f16_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 9)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f16_une:
-; VI: v_cmp_neq_f16_e64
 
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_neq_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
 define amdgpu_kernel void @v_fcmp_f16_une(i64 addrspace(1)* %out, half %src) {
+; GFX-LABEL: v_fcmp_f16_une:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    s_mov_b32 s4, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cvt_f32_f16_e32 v0, s2
+; GFX-NEXT:    v_cmp_neq_f32_e64 s[4:5], s4, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_une:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x5640
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_neq_f16_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 14)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f16_ugt:
-; VI: v_cmp_nle_f16_e64
 
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_nge_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
 define amdgpu_kernel void @v_fcmp_f16_ugt(i64 addrspace(1)* %out, half %src) {
+; GFX-LABEL: v_fcmp_f16_ugt:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    s_mov_b32 s4, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cvt_f32_f16_e32 v0, s2
+; GFX-NEXT:    v_cmp_nge_f32_e64 s[4:5], s4, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_ugt:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x5640
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_nle_f16_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 10)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f16_uge:
-; VI: v_cmp_nlt_f16_e64
 
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_ngt_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
 define amdgpu_kernel void @v_fcmp_f16_uge(i64 addrspace(1)* %out, half %src) {
+; GFX-LABEL: v_fcmp_f16_uge:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    s_mov_b32 s4, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cvt_f32_f16_e32 v0, s2
+; GFX-NEXT:    v_cmp_ngt_f32_e64 s[4:5], s4, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_uge:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x5640
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_nlt_f16_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 11)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f16_ult:
-; VI: v_cmp_nge_f16_e64
 
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_nle_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
 define amdgpu_kernel void @v_fcmp_f16_ult(i64 addrspace(1)* %out, half %src) {
+; GFX-LABEL: v_fcmp_f16_ult:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    s_mov_b32 s4, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cvt_f32_f16_e32 v0, s2
+; GFX-NEXT:    v_cmp_nle_f32_e64 s[4:5], s4, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_ult:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x5640
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_nge_f16_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 12)
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
-; GCN-LABEL: {{^}}v_fcmp_f16_ule:
-; VI: v_cmp_ngt_f16_e64
 
-; SI-DAG: s_mov_b32 [[K:s[0-9]+]], 0x42c80000
-; SI-DAG: v_cvt_f32_f16_e32 [[CVT:v[0-9]+]], s{{[0-9]+}}
-; SI: v_cmp_nlt_f32_e64 s{{\[[0-9]+:[0-9]+\]}}, [[K]], [[CVT]]
 define amdgpu_kernel void @v_fcmp_f16_ule(i64 addrspace(1)* %out, half %src) {
+; GFX-LABEL: v_fcmp_f16_ule:
+; GFX:       ; %bb.0:
+; GFX-NEXT:    s_load_dword s2, s[0:1], 0xb
+; GFX-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GFX-NEXT:    s_mov_b32 s3, 0xf000
+; GFX-NEXT:    s_mov_b32 s4, 0x42c80000
+; GFX-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX-NEXT:    v_cvt_f32_f16_e32 v0, s2
+; GFX-NEXT:    v_cmp_nlt_f32_e64 s[4:5], s4, v0
+; GFX-NEXT:    s_mov_b32 s2, -1
+; GFX-NEXT:    v_mov_b32_e32 v0, s4
+; GFX-NEXT:    v_mov_b32_e32 v1, s5
+; GFX-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX-NEXT:    s_endpgm
+;
+; VI-LABEL: v_fcmp_f16_ule:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v0, 0x5640
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_cmp_ngt_f16_e64 s[2:3], s2, v0
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; VI-NEXT:    s_endpgm
   %result = call i64 @llvm.amdgcn.fcmp.f16(half %src, half 100.00, i32 13)
   store i64 %result, i64 addrspace(1)* %out
   ret void