[llvm] [X86] Add test coverage for the concatenation of ISD::FFLOOR intrinsics (PR #170168)
via llvm-commits
llvm-commits at lists.llvm.org
Mon Dec 1 08:45:48 PST 2025
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-backend-x86
Author: Simon Pilgrim (RKSimon)
<details>
<summary>Changes</summary>
These were missed in #<!-- -->170160
---
Full diff: https://github.com/llvm/llvm-project/pull/170168.diff
1 Files Affected:
- (added) llvm/test/CodeGen/X86/combine-ffloor.ll (+178)
``````````diff
diff --git a/llvm/test/CodeGen/X86/combine-ffloor.ll b/llvm/test/CodeGen/X86/combine-ffloor.ll
new file mode 100644
index 0000000000000..35c28d2c85399
--- /dev/null
+++ b/llvm/test/CodeGen/X86/combine-ffloor.ll
@@ -0,0 +1,178 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64-v2 | FileCheck %s --check-prefixes=SSE
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=sandybridge | FileCheck %s --check-prefixes=AVX,AVX1OR2,AVX1
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64-v3 | FileCheck %s --check-prefixes=AVX,AVX1OR2,AVX2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64-v4 | FileCheck %s --check-prefixes=AVX,AVX512
+
+define <4 x double> @concat_floor_v4f64_v2f64(<2 x double> %a0, <2 x double> %a1) {
+; SSE-LABEL: concat_floor_v4f64_v2f64:
+; SSE: # %bb.0:
+; SSE-NEXT: roundpd $9, %xmm0, %xmm0
+; SSE-NEXT: roundpd $9, %xmm1, %xmm1
+; SSE-NEXT: retq
+;
+; AVX-LABEL: concat_floor_v4f64_v2f64:
+; AVX: # %bb.0:
+; AVX-NEXT: vroundpd $9, %xmm0, %xmm0
+; AVX-NEXT: vroundpd $9, %xmm1, %xmm1
+; AVX-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-NEXT: retq
+ %v0 = call <2 x double> @llvm.floor.v2f64(<2 x double> %a0)
+ %v1 = call <2 x double> @llvm.floor.v2f64(<2 x double> %a1)
+ %res = shufflevector <2 x double> %v0, <2 x double> %v1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ ret <4 x double> %res
+}
+
+define <8 x float> @concat_floor_v8f32_v4f32(<4 x float> %a0, <4 x float> %a1) {
+; SSE-LABEL: concat_floor_v8f32_v4f32:
+; SSE: # %bb.0:
+; SSE-NEXT: roundps $9, %xmm0, %xmm0
+; SSE-NEXT: roundps $9, %xmm1, %xmm1
+; SSE-NEXT: retq
+;
+; AVX-LABEL: concat_floor_v8f32_v4f32:
+; AVX: # %bb.0:
+; AVX-NEXT: vroundps $9, %xmm0, %xmm0
+; AVX-NEXT: vroundps $9, %xmm1, %xmm1
+; AVX-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-NEXT: retq
+ %v0 = call <4 x float> @llvm.floor.v4f32(<4 x float> %a0)
+ %v1 = call <4 x float> @llvm.floor.v4f32(<4 x float> %a1)
+ %res = shufflevector <4 x float> %v0, <4 x float> %v1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ ret <8 x float> %res
+}
+
+define <8 x double> @concat_floor_v8f64_v2f64(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, <2 x double> %a3) {
+; SSE-LABEL: concat_floor_v8f64_v2f64:
+; SSE: # %bb.0:
+; SSE-NEXT: roundpd $9, %xmm0, %xmm0
+; SSE-NEXT: roundpd $9, %xmm1, %xmm1
+; SSE-NEXT: roundpd $9, %xmm2, %xmm2
+; SSE-NEXT: roundpd $9, %xmm3, %xmm3
+; SSE-NEXT: retq
+;
+; AVX1OR2-LABEL: concat_floor_v8f64_v2f64:
+; AVX1OR2: # %bb.0:
+; AVX1OR2-NEXT: vroundpd $9, %xmm0, %xmm0
+; AVX1OR2-NEXT: vroundpd $9, %xmm1, %xmm1
+; AVX1OR2-NEXT: vroundpd $9, %xmm2, %xmm2
+; AVX1OR2-NEXT: vroundpd $9, %xmm3, %xmm3
+; AVX1OR2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1OR2-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm1
+; AVX1OR2-NEXT: retq
+;
+; AVX512-LABEL: concat_floor_v8f64_v2f64:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vroundpd $9, %xmm0, %xmm0
+; AVX512-NEXT: vroundpd $9, %xmm1, %xmm1
+; AVX512-NEXT: vroundpd $9, %xmm2, %xmm2
+; AVX512-NEXT: vroundpd $9, %xmm3, %xmm3
+; AVX512-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2
+; AVX512-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX512-NEXT: vinsertf64x4 $1, %ymm2, %zmm0, %zmm0
+; AVX512-NEXT: retq
+ %v0 = call <2 x double> @llvm.floor.v2f64(<2 x double> %a0)
+ %v1 = call <2 x double> @llvm.floor.v2f64(<2 x double> %a1)
+ %v2 = call <2 x double> @llvm.floor.v2f64(<2 x double> %a2)
+ %v3 = call <2 x double> @llvm.floor.v2f64(<2 x double> %a3)
+ %r01 = shufflevector <2 x double> %v0, <2 x double> %v1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ %r23 = shufflevector <2 x double> %v2, <2 x double> %v3, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ %res = shufflevector <4 x double> %r01, <4 x double> %r23, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ ret <8 x double> %res
+}
+
+define <16 x float> @concat_floor_v16f32_v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, <4 x float> %a3) {
+; SSE-LABEL: concat_floor_v16f32_v4f32:
+; SSE: # %bb.0:
+; SSE-NEXT: roundps $9, %xmm0, %xmm0
+; SSE-NEXT: roundps $9, %xmm1, %xmm1
+; SSE-NEXT: roundps $9, %xmm2, %xmm2
+; SSE-NEXT: roundps $9, %xmm3, %xmm3
+; SSE-NEXT: retq
+;
+; AVX1OR2-LABEL: concat_floor_v16f32_v4f32:
+; AVX1OR2: # %bb.0:
+; AVX1OR2-NEXT: vroundps $9, %xmm0, %xmm0
+; AVX1OR2-NEXT: vroundps $9, %xmm1, %xmm1
+; AVX1OR2-NEXT: vroundps $9, %xmm2, %xmm2
+; AVX1OR2-NEXT: vroundps $9, %xmm3, %xmm3
+; AVX1OR2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1OR2-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm1
+; AVX1OR2-NEXT: retq
+;
+; AVX512-LABEL: concat_floor_v16f32_v4f32:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vroundps $9, %xmm0, %xmm0
+; AVX512-NEXT: vroundps $9, %xmm1, %xmm1
+; AVX512-NEXT: vroundps $9, %xmm2, %xmm2
+; AVX512-NEXT: vroundps $9, %xmm3, %xmm3
+; AVX512-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2
+; AVX512-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX512-NEXT: vinsertf64x4 $1, %ymm2, %zmm0, %zmm0
+; AVX512-NEXT: retq
+ %v0 = call <4 x float> @llvm.floor.v4f32(<4 x float> %a0)
+ %v1 = call <4 x float> @llvm.floor.v4f32(<4 x float> %a1)
+ %v2 = call <4 x float> @llvm.floor.v4f32(<4 x float> %a2)
+ %v3 = call <4 x float> @llvm.floor.v4f32(<4 x float> %a3)
+ %r01 = shufflevector <4 x float> %v0, <4 x float> %v1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ %r23 = shufflevector <4 x float> %v2, <4 x float> %v3, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ %res = shufflevector <8 x float> %r01, <8 x float> %r23, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+ ret <16 x float> %res
+}
+
+define <8 x double> @concat_floor_v8f64_v4f64(<4 x double> %a0, <4 x double> %a1) {
+; SSE-LABEL: concat_floor_v8f64_v4f64:
+; SSE: # %bb.0:
+; SSE-NEXT: roundpd $9, %xmm0, %xmm0
+; SSE-NEXT: roundpd $9, %xmm1, %xmm1
+; SSE-NEXT: roundpd $9, %xmm2, %xmm2
+; SSE-NEXT: roundpd $9, %xmm3, %xmm3
+; SSE-NEXT: retq
+;
+; AVX1OR2-LABEL: concat_floor_v8f64_v4f64:
+; AVX1OR2: # %bb.0:
+; AVX1OR2-NEXT: vroundpd $9, %ymm0, %ymm0
+; AVX1OR2-NEXT: vroundpd $9, %ymm1, %ymm1
+; AVX1OR2-NEXT: retq
+;
+; AVX512-LABEL: concat_floor_v8f64_v4f64:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vroundpd $9, %ymm0, %ymm0
+; AVX512-NEXT: vroundpd $9, %ymm1, %ymm1
+; AVX512-NEXT: vinsertf64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512-NEXT: retq
+ %v0 = call <4 x double> @llvm.floor.v4f64(<4 x double> %a0)
+ %v1 = call <4 x double> @llvm.floor.v4f64(<4 x double> %a1)
+ %res = shufflevector <4 x double> %v0, <4 x double> %v1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ ret <8 x double> %res
+}
+
+define <16 x float> @concat_floor_v16f32_v8f32(<8 x float> %a0, <8 x float> %a1) {
+; SSE-LABEL: concat_floor_v16f32_v8f32:
+; SSE: # %bb.0:
+; SSE-NEXT: roundps $9, %xmm0, %xmm0
+; SSE-NEXT: roundps $9, %xmm1, %xmm1
+; SSE-NEXT: roundps $9, %xmm2, %xmm2
+; SSE-NEXT: roundps $9, %xmm3, %xmm3
+; SSE-NEXT: retq
+;
+; AVX1OR2-LABEL: concat_floor_v16f32_v8f32:
+; AVX1OR2: # %bb.0:
+; AVX1OR2-NEXT: vroundps $9, %ymm0, %ymm0
+; AVX1OR2-NEXT: vroundps $9, %ymm1, %ymm1
+; AVX1OR2-NEXT: retq
+;
+; AVX512-LABEL: concat_floor_v16f32_v8f32:
+; AVX512: # %bb.0:
+; AVX512-NEXT: vroundps $9, %ymm0, %ymm0
+; AVX512-NEXT: vroundps $9, %ymm1, %ymm1
+; AVX512-NEXT: vinsertf64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512-NEXT: retq
+ %v0 = call <8 x float> @llvm.floor.v8f32(<8 x float> %a0)
+ %v1 = call <8 x float> @llvm.floor.v8f32(<8 x float> %a1)
+ %res = shufflevector <8 x float> %v0, <8 x float> %v1, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+ ret <16 x float> %res
+}
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; AVX1: {{.*}}
+; AVX2: {{.*}}
``````````
</details>
https://github.com/llvm/llvm-project/pull/170168
More information about the llvm-commits
mailing list