diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index a6ba6e518899f..3f6ac9d1e6404 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -18862,27 +18862,38 @@ SDValue DAGCombiner::visitFPOW(SDNode *N) {
 
 static SDValue foldFPToIntToFP(SDNode *N, const SDLoc &DL, SelectionDAG &DAG,
                                const TargetLowering &TLI) {
-  // We only do this if the target has legal ftrunc. Otherwise, we'd likely be
-  // replacing casts with a libcall. We also must be allowed to ignore -0.0
-  // because FTRUNC will return -0.0 for (-1.0, -0.0), but using integer
-  // conversions would return +0.0.
+  // We can fold the fpto[us]i -> [us]itofp pattern into a single ftrunc.
+  // If NoSignedZerosFPMath is enabled, this is a direct replacement.
+  // Otherwise, for strict math, we must handle edge cases:
+  // 1. For unsigned conversions, use FABS to handle negative cases. Take -0.0
+  // as example, it first becomes integer 0, and is converted back to +0.0.
+  // FTRUNC on its own could produce -0.0.
+
   // FIXME: We should be able to use node-level FMF here.
-  // TODO: If strict math, should we use FABS (+ range check for signed cast)?
   EVT VT = N->getValueType(0);
-  if (!TLI.isOperationLegal(ISD::FTRUNC, VT) ||
-      !DAG.getTarget().Options.NoSignedZerosFPMath)
+  if (!TLI.isOperationLegal(ISD::FTRUNC, VT))
     return SDValue();
 
   // fptosi/fptoui round towards zero, so converting from FP to integer and
   // back is the same as an 'ftrunc': [us]itofp (fpto[us]i X) --> ftrunc X
   SDValue N0 = N->getOperand(0);
   if (N->getOpcode() == ISD::SINT_TO_FP && N0.getOpcode() == ISD::FP_TO_SINT &&
-      N0.getOperand(0).getValueType() == VT)
-    return DAG.getNode(ISD::FTRUNC, DL, VT, N0.getOperand(0));
+      N0.getOperand(0).getValueType() == VT) {
+    if (DAG.getTarget().Options.NoSignedZerosFPMath)
+      return DAG.getNode(ISD::FTRUNC, DL, VT, N0.getOperand(0));
+  }
 
   if (N->getOpcode() == ISD::UINT_TO_FP && N0.getOpcode() == ISD::FP_TO_UINT &&
-      N0.getOperand(0).getValueType() == VT)
-    return DAG.getNode(ISD::FTRUNC, DL, VT, N0.getOperand(0));
+      N0.getOperand(0).getValueType() == VT) {
+    if (DAG.getTarget().Options.NoSignedZerosFPMath)
+      return DAG.getNode(ISD::FTRUNC, DL, VT, N0.getOperand(0));
+
+    // Strict math: use FABS to handle negative inputs correctly.
+    if (TLI.isFAbsFree(VT)) {
+      SDValue Abs = DAG.getNode(ISD::FABS, DL, VT, N0.getOperand(0));
+      return DAG.getNode(ISD::FTRUNC, DL, VT, Abs);
+    }
+  }
 
   return SDValue();
 }
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-cvt-fp-int-fp.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-cvt-fp-int-fp.ll
index f4ae66a3b2259..4ad5b38b256fe 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-cvt-fp-int-fp.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-cvt-fp-int-fp.ll
@@ -70,6 +70,12 @@ define half @t3(half %x)  {
 ; CHECK-NEXT:    // kill: def $h0 killed $h0 killed $z0
 ; CHECK-NEXT:    ret
 ;
+; USE-NEON-NO-GPRS-LABEL: t3:
+; USE-NEON-NO-GPRS:       // %bb.0: // %entry
+; USE-NEON-NO-GPRS-NEXT:    fcvtzs h0, h0
+; USE-NEON-NO-GPRS-NEXT:    scvtf h0, h0
+; USE-NEON-NO-GPRS-NEXT:    ret
+;
 ; NONEON-NOSVE-LABEL: t3:
 ; NONEON-NOSVE:       // %bb.0: // %entry
 ; NONEON-NOSVE-NEXT:    fcvt s0, h0
@@ -147,6 +153,12 @@ define half @t6(half %x)  {
 ; CHECK-NEXT:    // kill: def $h0 killed $h0 killed $z0
 ; CHECK-NEXT:    ret
 ;
+; USE-NEON-NO-GPRS-LABEL: t6:
+; USE-NEON-NO-GPRS:       // %bb.0: // %entry
+; USE-NEON-NO-GPRS-NEXT:    fcvtzu h0, h0
+; USE-NEON-NO-GPRS-NEXT:    ucvtf h0, h0
+; USE-NEON-NO-GPRS-NEXT:    ret
+;
 ; NONEON-NOSVE-LABEL: t6:
 ; NONEON-NOSVE:       // %bb.0: // %entry
 ; NONEON-NOSVE-NEXT:    fcvt s0, h0
diff --git a/llvm/test/CodeGen/AMDGPU/fptoui_uitofp.ll b/llvm/test/CodeGen/AMDGPU/fptoui_uitofp.ll
new file mode 100644
index 0000000000000..49204f84acb85
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/fptoui_uitofp.ll
@@ -0,0 +1,296 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+; RUN: llc -mtriple=amdgcn -mcpu=gfx600 < %s | FileCheck -check-prefix=GFX6 %s
+; RUN: llc -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefix=GFX9 %s
+
+define amdgpu_kernel void @fptoui_f32_to_i16_to_f32(ptr addrspace(1) %out, float %x) {
+; GFX6-LABEL: fptoui_f32_to_i16_to_f32:
+; GFX6:       ; %bb.0: ; %entry
+; GFX6-NEXT:    s_load_dword s6, s[4:5], 0xb
+; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
+; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX6-NEXT:    v_trunc_f32_e64 v0, |s6|
+; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GFX6-NEXT:    s_endpgm
+;
+; GFX9-LABEL: fptoui_f32_to_i16_to_f32:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x2c
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_trunc_f32_e64 v1, |s2|
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+entry:
+  %ui = fptoui float %x to i16
+  %fp = uitofp i16 %ui to float
+  store float %fp, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @fptoui_f32_to_i32_to_f32(ptr addrspace(1) %out, float %x) {
+; GFX6-LABEL: fptoui_f32_to_i32_to_f32:
+; GFX6:       ; %bb.0: ; %entry
+; GFX6-NEXT:    s_load_dword s6, s[4:5], 0xb
+; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
+; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX6-NEXT:    v_trunc_f32_e64 v0, |s6|
+; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GFX6-NEXT:    s_endpgm
+;
+; GFX9-LABEL: fptoui_f32_to_i32_to_f32:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x2c
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_trunc_f32_e64 v1, |s2|
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+entry:
+  %ui = fptoui float %x to i32
+  %fp = uitofp i32 %ui to float
+  store float %fp, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @fptoui_f32_to_i64_to_f32(ptr addrspace(1) %out, float %x) {
+; GFX6-LABEL: fptoui_f32_to_i64_to_f32:
+; GFX6:       ; %bb.0: ; %entry
+; GFX6-NEXT:    s_load_dword s6, s[4:5], 0xb
+; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
+; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX6-NEXT:    v_trunc_f32_e64 v0, |s6|
+; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GFX6-NEXT:    s_endpgm
+;
+; GFX9-LABEL: fptoui_f32_to_i64_to_f32:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x2c
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_trunc_f32_e64 v1, |s2|
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+entry:
+  %ui = fptoui float %x to i64
+  %fp = uitofp i64 %ui to float
+  store float %fp, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @fptoui_f16_to_i16_to_f16(ptr addrspace(1) %out, half %x) {
+; GFX6-LABEL: fptoui_f16_to_i16_to_f16:
+; GFX6:       ; %bb.0: ; %entry
+; GFX6-NEXT:    s_load_dword s0, s[4:5], 0xb
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
+; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX6-NEXT:    v_cvt_f32_f16_e32 v0, s0
+; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX6-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; GFX6-NEXT:    v_cvt_f32_u32_e32 v0, v0
+; GFX6-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX6-NEXT:    buffer_store_short v0, off, s[0:3], 0
+; GFX6-NEXT:    s_endpgm
+;
+; GFX9-LABEL: fptoui_f16_to_i16_to_f16:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x2c
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_trunc_f16_e64 v1, |s2|
+; GFX9-NEXT:    global_store_short v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+entry:
+  %ui = fptoui half %x to i16
+  %fp = uitofp i16 %ui to half
+  store half %fp, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @fptoui_f16_to_i32_to_f16(ptr addrspace(1) %out, half %x) {
+; GFX6-LABEL: fptoui_f16_to_i32_to_f16:
+; GFX6:       ; %bb.0: ; %entry
+; GFX6-NEXT:    s_load_dword s0, s[4:5], 0xb
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
+; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX6-NEXT:    v_cvt_f32_f16_e64 v0, |s0|
+; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX6-NEXT:    v_trunc_f32_e32 v0, v0
+; GFX6-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX6-NEXT:    buffer_store_short v0, off, s[0:3], 0
+; GFX6-NEXT:    s_endpgm
+;
+; GFX9-LABEL: fptoui_f16_to_i32_to_f16:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x2c
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_trunc_f16_e64 v1, |s2|
+; GFX9-NEXT:    global_store_short v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+entry:
+  %ui = fptoui half %x to i32
+  %fp = uitofp i32 %ui to half
+  store half %fp, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @fptoui_f16_to_i64_to_f16(ptr addrspace(1) %out, half %x) {
+; GFX6-LABEL: fptoui_f16_to_i64_to_f16:
+; GFX6:       ; %bb.0: ; %entry
+; GFX6-NEXT:    s_load_dword s0, s[4:5], 0xb
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
+; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX6-NEXT:    v_cvt_f32_f16_e64 v0, |s0|
+; GFX6-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x9
+; GFX6-NEXT:    v_trunc_f32_e32 v0, v0
+; GFX6-NEXT:    v_cvt_f16_f32_e32 v0, v0
+; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX6-NEXT:    buffer_store_short v0, off, s[0:3], 0
+; GFX6-NEXT:    s_endpgm
+;
+; GFX9-LABEL: fptoui_f16_to_i64_to_f16:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x2c
+; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_trunc_f16_e64 v1, |s2|
+; GFX9-NEXT:    global_store_short v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+entry:
+  %ui = fptoui half %x to i64
+  %fp = uitofp i64 %ui to half
+  store half %fp, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @fptoui_f64_to_i16_to_f64(ptr addrspace(1) %out, double %x) {
+; GFX6-LABEL: fptoui_f64_to_i16_to_f64:
+; GFX6:       ; %bb.0: ; %entry
+; GFX6-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
+; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX6-NEXT:    v_cvt_u32_f64_e32 v0, s[2:3]
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
+; GFX6-NEXT:    v_cvt_f64_u32_e32 v[0:1], v0
+; GFX6-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX6-NEXT:    s_endpgm
+;
+; GFX9-LABEL: fptoui_f64_to_i16_to_f64:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_trunc_f64_e64 v[0:1], |s[2:3]|
+; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX9-NEXT:    s_endpgm
+entry:
+  %ui = fptoui double %x to i16
+  %fp = uitofp i16 %ui to double
+  store double %fp, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @fptoui_f64_to_i32_to_f64(ptr addrspace(1) %out, double %x) {
+; GFX6-LABEL: fptoui_f64_to_i32_to_f64:
+; GFX6:       ; %bb.0: ; %entry
+; GFX6-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
+; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX6-NEXT:    v_cvt_u32_f64_e32 v0, s[2:3]
+; GFX6-NEXT:    s_mov_b32 s3, 0xf000
+; GFX6-NEXT:    s_mov_b32 s2, -1
+; GFX6-NEXT:    v_cvt_f64_u32_e32 v[0:1], v0
+; GFX6-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GFX6-NEXT:    s_endpgm
+;
+; GFX9-LABEL: fptoui_f64_to_i32_to_f64:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_trunc_f64_e64 v[0:1], |s[2:3]|
+; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX9-NEXT:    s_endpgm
+entry:
+  %ui = fptoui double %x to i32
+  %fp = uitofp i32 %ui to double
+  store double %fp, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @fptoui_f64_to_i64_to_f64(ptr addrspace(1) %out, double %x) {
+; GFX6-LABEL: fptoui_f64_to_i64_to_f64:
+; GFX6:       ; %bb.0: ; %entry
+; GFX6-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
+; GFX6-NEXT:    s_mov_b32 s6, -1
+; GFX6-NEXT:    s_mov_b32 s5, 0xfffff
+; GFX6-NEXT:    s_mov_b32 s4, s6
+; GFX6-NEXT:    v_not_b32_e32 v0, 31
+; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX6-NEXT:    s_bfe_u32 s7, s3, 0xb0014
+; GFX6-NEXT:    s_addk_i32 s7, 0xfc01
+; GFX6-NEXT:    s_lshr_b64 s[4:5], s[4:5], s7
+; GFX6-NEXT:    s_and_b32 s8, s3, 0x80000000
+; GFX6-NEXT:    s_andn2_b64 s[4:5], s[2:3], s[4:5]
+; GFX6-NEXT:    s_cmp_lt_i32 s7, 0
+; GFX6-NEXT:    s_cselect_b32 s4, 0, s4
+; GFX6-NEXT:    s_cselect_b32 s5, s8, s5
+; GFX6-NEXT:    s_cmp_gt_i32 s7, 51
+; GFX6-NEXT:    s_cselect_b32 s3, s3, s5
+; GFX6-NEXT:    s_cselect_b32 s2, s2, s4
+; GFX6-NEXT:    v_ldexp_f64 v[0:1], s[2:3], v0
+; GFX6-NEXT:    v_mov_b32_e32 v4, -1
+; GFX6-NEXT:    v_fract_f64_e32 v[2:3], v[0:1]
+; GFX6-NEXT:    v_mov_b32_e32 v5, 0x3fefffff
+; GFX6-NEXT:    v_min_f64 v[2:3], v[2:3], v[4:5]
+; GFX6-NEXT:    v_cmp_class_f64_e64 vcc, v[0:1], 3
+; GFX6-NEXT:    s_mov_b32 s4, 0
+; GFX6-NEXT:    v_cndmask_b32_e32 v2, v2, v0, vcc
+; GFX6-NEXT:    v_cndmask_b32_e32 v3, v3, v1, vcc
+; GFX6-NEXT:    v_add_f64 v[0:1], v[0:1], -v[2:3]
+; GFX6-NEXT:    v_mov_b32_e32 v2, s2
+; GFX6-NEXT:    s_mov_b32 s5, 0xc1f00000
+; GFX6-NEXT:    v_mov_b32_e32 v3, s3
+; GFX6-NEXT:    v_fma_f64 v[2:3], v[0:1], s[4:5], v[2:3]
+; GFX6-NEXT:    v_cvt_u32_f64_e32 v0, v[0:1]
+; GFX6-NEXT:    v_cvt_u32_f64_e32 v2, v[2:3]
+; GFX6-NEXT:    s_mov_b32 s7, 0xf000
+; GFX6-NEXT:    s_mov_b32 s4, s0
+; GFX6-NEXT:    v_cvt_f64_u32_e32 v[0:1], v0
+; GFX6-NEXT:    v_cvt_f64_u32_e32 v[2:3], v2
+; GFX6-NEXT:    s_mov_b32 s5, s1
+; GFX6-NEXT:    v_ldexp_f64 v[0:1], v[0:1], 32
+; GFX6-NEXT:    v_add_f64 v[0:1], v[0:1], v[2:3]
+; GFX6-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GFX6-NEXT:    s_endpgm
+;
+; GFX9-LABEL: fptoui_f64_to_i64_to_f64:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX9-NEXT:    v_mov_b32_e32 v2, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    v_trunc_f64_e64 v[0:1], |s[2:3]|
+; GFX9-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX9-NEXT:    s_endpgm
+entry:
+  %ui = fptoui double %x to i64
+  %fp = uitofp i64 %ui to double
+  store double %fp, ptr addrspace(1) %out
+  ret void
+}