[llvm] [DXIL] Add lowering for `ceil` (PR #87043)

via llvm-commits llvm-commits at lists.llvm.org
Fri Mar 29 10:32:19 PDT 2024


llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-backend-directx

Author: Helena Kotas (hekota)

<details>
<summary>Changes</summary>

Add lowering of llvm.ceil intrinsics to DXIL ops.

Fixes #<!-- -->86984

---
Full diff: https://github.com/llvm/llvm-project/pull/87043.diff


3 Files Affected:

- (modified) llvm/lib/Target/DirectX/DXIL.td (+3) 
- (added) llvm/test/CodeGen/DirectX/ceil.ll (+20) 
- (added) llvm/test/CodeGen/DirectX/ceil_error.ll (+10) 


``````````diff
diff --git a/llvm/lib/Target/DirectX/DXIL.td b/llvm/lib/Target/DirectX/DXIL.td
index c5d7ee76275f86..a9ae47c5c0c5a0 100644
--- a/llvm/lib/Target/DirectX/DXIL.td
+++ b/llvm/lib/Target/DirectX/DXIL.td
@@ -292,6 +292,9 @@ def Round : DXILOpMapping<26, unary, int_round,
 def Floor : DXILOpMapping<27, unary, int_floor,
                          "Returns the largest integer that is less than or equal to the input.",
                          [llvm_halforfloat_ty, LLVMMatchType<0>]>;
+def Ceil : DXILOpMapping<28, unary, int_ceil,
+                         "Returns the smallest integer that is greater than or equal to the input.",
+                         [llvm_halforfloat_ty, LLVMMatchType<0>]>;
 def Trunc : DXILOpMapping<29, unary, int_trunc,
                          "Returns the specified value truncated to the integer component.",
                          [llvm_halforfloat_ty, LLVMMatchType<0>]>;
diff --git a/llvm/test/CodeGen/DirectX/ceil.ll b/llvm/test/CodeGen/DirectX/ceil.ll
new file mode 100644
index 00000000000000..15854714678014
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/ceil.ll
@@ -0,0 +1,20 @@
+; RUN: opt -S -dxil-op-lower < %s | FileCheck %s
+
+; Make sure dxil operation function calls for ceil are generated for float and half.
+
+define noundef float @ceil_float(float noundef %a) {
+entry:
+; CHECK:call float @dx.op.unary.f32(i32 28, float %{{.*}})
+  %elt.ceil = call float @llvm.ceil.f32(float %a)
+  ret float %elt.ceil
+}
+
+define noundef half @ceil_half(half noundef %a) {
+entry:
+; CHECK:call half @dx.op.unary.f16(i32 28, half %{{.*}})
+  %elt.ceil = call half @llvm.ceil.f16(half %a)
+  ret half %elt.ceil
+}
+
+declare half @llvm.ceil.f16(half)
+declare float @llvm.ceil.f32(float)
diff --git a/llvm/test/CodeGen/DirectX/ceil_error.ll b/llvm/test/CodeGen/DirectX/ceil_error.ll
new file mode 100644
index 00000000000000..1b554d8715566e
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/ceil_error.ll
@@ -0,0 +1,10 @@
+; RUN: not opt -S -dxil-op-lower %s 2>&1 | FileCheck %s
+
+; DXIL operation ceil does not support double overload type
+; CHECK: LLVM ERROR: Invalid Overload Type
+
+define noundef double @ceil_double(double noundef %a) {
+entry:
+  %elt.ceil = call double @llvm.ceil.f64(double %a)
+  ret double %elt.ceil
+}

``````````

</details>


https://github.com/llvm/llvm-project/pull/87043


More information about the llvm-commits mailing list