Skip to content

[RISCV] Lower vector_splice on zvfhmin/zvfbfmin #112579

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 3 commits into from
Oct 16, 2024
Merged

[RISCV] Lower vector_splice on zvfhmin/zvfbfmin #112579

merged 3 commits into from
Oct 16, 2024

Conversation

lukel97
Copy link
Contributor

@lukel97 lukel97 commented Oct 16, 2024

Similar to other permutation ops, we can just reuse the existing lowering.

@lukel97
Precommit tests
fe7e653
@lukel97
[RISCV] Lower vector_splice on zvfhmin/zvfbfmin
118f8ba 
Similar to other permutation ops, we can just reuse the existing lowering.
@llvmbot
Copy link
Member

llvmbot commented Oct 16, 2024 •
edited
Loading

@llvm/pr-subscribers-llvm-analysis

@llvm/pr-subscribers-backend-risc-v

Author: Luke Lau (lukel97)

Changes

Similar to other permutation ops, we can just reuse the existing lowering.

Patch is 25.54 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/112579.diff

3 Files Affected:

  • (modified) llvm/lib/Target/RISCV/RISCVISelLowering.cpp (+1-1)
  • (modified) llvm/test/Analysis/CostModel/RISCV/splice.ll (+24-2)
  • (modified) llvm/test/CodeGen/RISCV/rvv/vector-splice.ll (+331-2)
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index bf333b7b790167..076ed173f64e2e 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -1076,7 +1076,7 @@ RISCVTargetLowering::RISCVTargetLowering(const TargetMachine &TM,
       setOperationAction({ISD::INSERT_VECTOR_ELT, ISD::CONCAT_VECTORS,
                           ISD::INSERT_SUBVECTOR, ISD::EXTRACT_SUBVECTOR,
                           ISD::VECTOR_DEINTERLEAVE, ISD::VECTOR_INTERLEAVE,
-                          ISD::VECTOR_REVERSE},
+                          ISD::VECTOR_REVERSE, ISD::VECTOR_SPLICE},
                          VT, Custom);
       MVT EltVT = VT.getVectorElementType();
       if (isTypeLegal(EltVT))
diff --git a/llvm/test/Analysis/CostModel/RISCV/splice.ll b/llvm/test/Analysis/CostModel/RISCV/splice.ll
index 8d7d1576a532da..ddfaa8c13d425f 100644
--- a/llvm/test/Analysis/CostModel/RISCV/splice.ll
+++ b/llvm/test/Analysis/CostModel/RISCV/splice.ll
@@ -1,6 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
-; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -S -mtriple=riscv64 -mattr=+v,+f,+d,+zfh,+zvfh | FileCheck %s
-; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -S -mtriple=riscv64 -mattr=+v,+f,+d,+zfh,+zvfhmin | FileCheck %s
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -S -mtriple=riscv64 -mattr=+v,+f,+d,+zfh,+zvfh,+zvfbfmin | FileCheck %s
+; RUN: opt < %s -passes="print<cost-model>" 2>&1 -disable-output -S -mtriple=riscv64 -mattr=+v,+f,+d,+zfh,+zvfhmin,+zvfbfmin | FileCheck %s
 ; RUN: opt < %s -passes="print<cost-model>" -cost-kind=code-size 2>&1 -disable-output -S -mtriple=riscv64 -mattr=+v,+f,+d,+zfh,+zvfh | FileCheck %s --check-prefix=SIZE
 ; RUN: opt < %s -passes="print<cost-model>" -cost-kind=code-size 2>&1 -disable-output -S -mtriple=riscv64 -mattr=+v,+f,+d,+zfh,+zvfhmin | FileCheck %s --check-prefix=SIZE
 
@@ -34,6 +34,13 @@ define void @vector_splice() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %splice.nxv16i64 = call <vscale x 16 x i64> @llvm.vector.splice.nxv16i64(<vscale x 16 x i64> zeroinitializer, <vscale x 16 x i64> zeroinitializer, i32 -1)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %splice.nxv32i64 = call <vscale x 32 x i64> @llvm.vector.splice.nxv32i64(<vscale x 32 x i64> zeroinitializer, <vscale x 32 x i64> zeroinitializer, i32 -1)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 128 for instruction: %splice.nxv64i64 = call <vscale x 64 x i64> @llvm.vector.splice.nxv64i64(<vscale x 64 x i64> zeroinitializer, <vscale x 64 x i64> zeroinitializer, i32 -1)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splice.nxv1bf16 = call <vscale x 1 x bfloat> @llvm.vector.splice.nxv1bf16(<vscale x 1 x bfloat> zeroinitializer, <vscale x 1 x bfloat> zeroinitializer, i32 -1)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splice.nxv2bf16 = call <vscale x 2 x bfloat> @llvm.vector.splice.nxv2bf16(<vscale x 2 x bfloat> zeroinitializer, <vscale x 2 x bfloat> zeroinitializer, i32 -1)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splice.nxv4bf16 = call <vscale x 4 x bfloat> @llvm.vector.splice.nxv4bf16(<vscale x 4 x bfloat> zeroinitializer, <vscale x 4 x bfloat> zeroinitializer, i32 -1)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %splice.nxv8bf16 = call <vscale x 8 x bfloat> @llvm.vector.splice.nxv8bf16(<vscale x 8 x bfloat> zeroinitializer, <vscale x 8 x bfloat> zeroinitializer, i32 -1)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %splice.nxv16bf16 = call <vscale x 16 x bfloat> @llvm.vector.splice.nxv16bf16(<vscale x 16 x bfloat> zeroinitializer, <vscale x 16 x bfloat> zeroinitializer, i32 -1)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %splice.nxv32bf16 = call <vscale x 32 x bfloat> @llvm.vector.splice.nxv32bf16(<vscale x 32 x bfloat> zeroinitializer, <vscale x 32 x bfloat> zeroinitializer, i32 -1)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %splice.nxv64bf16 = call <vscale x 64 x bfloat> @llvm.vector.splice.nxv64bf16(<vscale x 64 x bfloat> zeroinitializer, <vscale x 64 x bfloat> zeroinitializer, i32 -1)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splice.nxv1f16 = call <vscale x 1 x half> @llvm.vector.splice.nxv1f16(<vscale x 1 x half> zeroinitializer, <vscale x 1 x half> zeroinitializer, i32 -1)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splice.nxv2f16 = call <vscale x 2 x half> @llvm.vector.splice.nxv2f16(<vscale x 2 x half> zeroinitializer, <vscale x 2 x half> zeroinitializer, i32 -1)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splice.nxv4f16 = call <vscale x 4 x half> @llvm.vector.splice.nxv4f16(<vscale x 4 x half> zeroinitializer, <vscale x 4 x half> zeroinitializer, i32 -1)
@@ -86,6 +93,13 @@ define void @vector_splice() {
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %splice.nxv16i64 = call <vscale x 16 x i64> @llvm.vector.splice.nxv16i64(<vscale x 16 x i64> zeroinitializer, <vscale x 16 x i64> zeroinitializer, i32 -1)
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %splice.nxv32i64 = call <vscale x 32 x i64> @llvm.vector.splice.nxv32i64(<vscale x 32 x i64> zeroinitializer, <vscale x 32 x i64> zeroinitializer, i32 -1)
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %splice.nxv64i64 = call <vscale x 64 x i64> @llvm.vector.splice.nxv64i64(<vscale x 64 x i64> zeroinitializer, <vscale x 64 x i64> zeroinitializer, i32 -1)
+; SIZE-NEXT:  Cost Model: Invalid cost for instruction: %splice.nxv1bf16 = call <vscale x 1 x bfloat> @llvm.vector.splice.nxv1bf16(<vscale x 1 x bfloat> zeroinitializer, <vscale x 1 x bfloat> zeroinitializer, i32 -1)
+; SIZE-NEXT:  Cost Model: Invalid cost for instruction: %splice.nxv2bf16 = call <vscale x 2 x bfloat> @llvm.vector.splice.nxv2bf16(<vscale x 2 x bfloat> zeroinitializer, <vscale x 2 x bfloat> zeroinitializer, i32 -1)
+; SIZE-NEXT:  Cost Model: Invalid cost for instruction: %splice.nxv4bf16 = call <vscale x 4 x bfloat> @llvm.vector.splice.nxv4bf16(<vscale x 4 x bfloat> zeroinitializer, <vscale x 4 x bfloat> zeroinitializer, i32 -1)
+; SIZE-NEXT:  Cost Model: Invalid cost for instruction: %splice.nxv8bf16 = call <vscale x 8 x bfloat> @llvm.vector.splice.nxv8bf16(<vscale x 8 x bfloat> zeroinitializer, <vscale x 8 x bfloat> zeroinitializer, i32 -1)
+; SIZE-NEXT:  Cost Model: Invalid cost for instruction: %splice.nxv16bf16 = call <vscale x 16 x bfloat> @llvm.vector.splice.nxv16bf16(<vscale x 16 x bfloat> zeroinitializer, <vscale x 16 x bfloat> zeroinitializer, i32 -1)
+; SIZE-NEXT:  Cost Model: Invalid cost for instruction: %splice.nxv32bf16 = call <vscale x 32 x bfloat> @llvm.vector.splice.nxv32bf16(<vscale x 32 x bfloat> zeroinitializer, <vscale x 32 x bfloat> zeroinitializer, i32 -1)
+; SIZE-NEXT:  Cost Model: Invalid cost for instruction: %splice.nxv64bf16 = call <vscale x 64 x bfloat> @llvm.vector.splice.nxv64bf16(<vscale x 64 x bfloat> zeroinitializer, <vscale x 64 x bfloat> zeroinitializer, i32 -1)
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splice.nxv1f16 = call <vscale x 1 x half> @llvm.vector.splice.nxv1f16(<vscale x 1 x half> zeroinitializer, <vscale x 1 x half> zeroinitializer, i32 -1)
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splice.nxv2f16 = call <vscale x 2 x half> @llvm.vector.splice.nxv2f16(<vscale x 2 x half> zeroinitializer, <vscale x 2 x half> zeroinitializer, i32 -1)
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %splice.nxv4f16 = call <vscale x 4 x half> @llvm.vector.splice.nxv4f16(<vscale x 4 x half> zeroinitializer, <vscale x 4 x half> zeroinitializer, i32 -1)
@@ -141,6 +155,14 @@ define void @vector_splice() {
   %splice.nxv32i64 = call <vscale x 32 x i64> @llvm.vector.splice.nxv32i64(<vscale x 32 x i64> zeroinitializer, <vscale x 32 x i64> zeroinitializer, i32 -1)
   %splice.nxv64i64 = call <vscale x 64 x i64> @llvm.vector.splice.nxv64i64(<vscale x 64 x i64> zeroinitializer, <vscale x 64 x i64> zeroinitializer, i32 -1)
 
+  %splice.nxv1bf16 = call <vscale x 1 x bfloat> @llvm.vector.splice.nxv1bf16(<vscale x 1 x bfloat> zeroinitializer, <vscale x 1 x bfloat> zeroinitializer, i32 -1)
+  %splice.nxv2bf16 = call <vscale x 2 x bfloat> @llvm.vector.splice.nxv2bf16(<vscale x 2 x bfloat> zeroinitializer, <vscale x 2 x bfloat> zeroinitializer, i32 -1)
+  %splice.nxv4bf16 = call <vscale x 4 x bfloat> @llvm.vector.splice.nxv4bf16(<vscale x 4 x bfloat> zeroinitializer, <vscale x 4 x bfloat> zeroinitializer, i32 -1)
+  %splice.nxv8bf16 = call <vscale x 8 x bfloat> @llvm.vector.splice.nxv8bf16(<vscale x 8 x bfloat> zeroinitializer, <vscale x 8 x bfloat> zeroinitializer, i32 -1)
+  %splice.nxv16bf16 = call <vscale x 16 x bfloat> @llvm.vector.splice.nxv16bf16(<vscale x 16 x bfloat> zeroinitializer, <vscale x 16 x bfloat> zeroinitializer, i32 -1)
+  %splice.nxv32bf16 = call <vscale x 32 x bfloat> @llvm.vector.splice.nxv32bf16(<vscale x 32 x bfloat> zeroinitializer, <vscale x 32 x bfloat> zeroinitializer, i32 -1)
+  %splice.nxv64bf16 = call <vscale x 64 x bfloat> @llvm.vector.splice.nxv64bf16(<vscale x 64 x bfloat> zeroinitializer, <vscale x 64 x bfloat> zeroinitializer, i32 -1)
+
   %splice.nxv1f16 = call <vscale x 1 x half> @llvm.vector.splice.nxv1f16(<vscale x 1 x half> zeroinitializer, <vscale x 1 x half> zeroinitializer, i32 -1)
   %splice.nxv2f16 = call <vscale x 2 x half> @llvm.vector.splice.nxv2f16(<vscale x 2 x half> zeroinitializer, <vscale x 2 x half> zeroinitializer, i32 -1)
   %splice.nxv4f16 = call <vscale x 4 x half> @llvm.vector.splice.nxv4f16(<vscale x 4 x half> zeroinitializer, <vscale x 4 x half> zeroinitializer, i32 -1)
diff --git a/llvm/test/CodeGen/RISCV/rvv/vector-splice.ll b/llvm/test/CodeGen/RISCV/rvv/vector-splice.ll
index 8cb6fed2f588a7..c9cb6dc6397c3c 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vector-splice.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vector-splice.ll
@@ -1,6 +1,8 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -mtriple riscv32 -mattr=+m,+f,+d,+v,+zfh,+zvfh < %s | FileCheck %s
-; RUN: llc -mtriple riscv64 -mattr=+m,+f,+d,+v,+zfh,+zvfh < %s | FileCheck %s
+; RUN: llc -mtriple riscv32 -mattr=+m,+f,+d,+v,+zvfhmin,+zvfbfmin < %s | FileCheck %s --check-prefixes=CHECK
+; RUN: llc -mtriple riscv64 -mattr=+m,+f,+d,+v,+zvfhmin,+zvfbfmin < %s | FileCheck %s --check-prefixes=CHECK
+; RUN: llc -mtriple riscv32 -mattr=+m,+f,+d,+v,+zvfh,+zvfbfmin < %s | FileCheck %s --check-prefixes=CHECK
+; RUN: llc -mtriple riscv64 -mattr=+m,+f,+d,+v,+zvfh,+zvfbfmin < %s | FileCheck %s --check-prefixes=CHECK
 
 ; Tests assume VLEN=128 or vscale_range_min=2.
 
@@ -1533,6 +1535,333 @@ define <vscale x 8 x i64> @splice_nxv8i64_offset_max(<vscale x 8 x i64> %a, <vsc
   ret <vscale x 8 x i64> %res
 }
 
+declare <vscale x 1 x bfloat> @llvm.vector.splice.nxv1bf16(<vscale x 1 x bfloat>, <vscale x 1 x bfloat>, i32)
+
+define <vscale x 1 x bfloat> @splice_nxv1bf16_offset_zero(<vscale x 1 x bfloat> %a, <vscale x 1 x bfloat> %b) #0 {
+; CHECK-LABEL: splice_nxv1bf16_offset_zero:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    ret
+  %res = call <vscale x 1 x bfloat> @llvm.vector.splice.nxv1bf16(<vscale x 1 x bfloat> %a, <vscale x 1 x bfloat> %b, i32 0)
+  ret <vscale x 1 x bfloat> %res
+}
+
+define <vscale x 1 x bfloat> @splice_nxv1bf16_offset_negone(<vscale x 1 x bfloat> %a, <vscale x 1 x bfloat> %b) #0 {
+; CHECK-LABEL: splice_nxv1bf16_offset_negone:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    csrr a0, vlenb
+; CHECK-NEXT:    srli a0, a0, 3
+; CHECK-NEXT:    addi a0, a0, -1
+; CHECK-NEXT:    vsetvli a1, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vslidedown.vx v8, v8, a0
+; CHECK-NEXT:    vslideup.vi v8, v9, 1
+; CHECK-NEXT:    ret
+  %res = call <vscale x 1 x bfloat> @llvm.vector.splice.nxv1bf16(<vscale x 1 x bfloat> %a, <vscale x 1 x bfloat> %b, i32 -1)
+  ret <vscale x 1 x bfloat> %res
+}
+
+define <vscale x 1 x bfloat> @splice_nxv1bf16_offset_min(<vscale x 1 x bfloat> %a, <vscale x 1 x bfloat> %b) #0 {
+; CHECK-LABEL: splice_nxv1bf16_offset_min:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    csrr a0, vlenb
+; CHECK-NEXT:    srli a0, a0, 3
+; CHECK-NEXT:    addi a0, a0, -2
+; CHECK-NEXT:    vsetivli zero, 2, e16, mf4, ta, ma
+; CHECK-NEXT:    vslidedown.vx v8, v8, a0
+; CHECK-NEXT:    vsetvli a0, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vslideup.vi v8, v9, 2
+; CHECK-NEXT:    ret
+  %res = call <vscale x 1 x bfloat> @llvm.vector.splice.nxv1bf16(<vscale x 1 x bfloat> %a, <vscale x 1 x bfloat> %b, i32 -2)
+  ret <vscale x 1 x bfloat> %res
+}
+
+define <vscale x 1 x bfloat> @splice_nxv1bf16_offset_max(<vscale x 1 x bfloat> %a, <vscale x 1 x bfloat> %b) #0 {
+; CHECK-LABEL: splice_nxv1bf16_offset_max:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    csrr a0, vlenb
+; CHECK-NEXT:    srli a0, a0, 3
+; CHECK-NEXT:    addi a0, a0, -1
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf4, ta, ma
+; CHECK-NEXT:    vslidedown.vi v8, v8, 1
+; CHECK-NEXT:    vsetvli a1, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vslideup.vx v8, v9, a0
+; CHECK-NEXT:    ret
+  %res = call <vscale x 1 x bfloat> @llvm.vector.splice.nxv1bf16(<vscale x 1 x bfloat> %a, <vscale x 1 x bfloat> %b, i32 1)
+  ret <vscale x 1 x bfloat> %res
+}
+
+declare <vscale x 2 x bfloat> @llvm.vector.splice.nxv2bf16(<vscale x 2 x bfloat>, <vscale x 2 x bfloat>, i32)
+
+define <vscale x 2 x bfloat> @splice_nxv2bf16_offset_zero(<vscale x 2 x bfloat> %a, <vscale x 2 x bfloat> %b) #0 {
+; CHECK-LABEL: splice_nxv2bf16_offset_zero:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    ret
+  %res = call <vscale x 2 x bfloat> @llvm.vector.splice.nxv2bf16(<vscale x 2 x bfloat> %a, <vscale x 2 x bfloat> %b, i32 0)
+  ret <vscale x 2 x bfloat> %res
+}
+
+define <vscale x 2 x bfloat> @splice_nxv2bf16_offset_negone(<vscale x 2 x bfloat> %a, <vscale x 2 x bfloat> %b) #0 {
+; CHECK-LABEL: splice_nxv2bf16_offset_negone:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    csrr a0, vlenb
+; CHECK-NEXT:    srli a0, a0, 2
+; CHECK-NEXT:    addi a0, a0, -1
+; CHECK-NEXT:    vsetvli a1, zero, e16, mf2, ta, ma
+; CHECK-NEXT:    vslidedown.vx v8, v8, a0
+; CHECK-NEXT:    vslideup.vi v8, v9, 1
+; CHECK-NEXT:    ret
+  %res = call <vscale x 2 x bfloat> @llvm.vector.splice.nxv2bf16(<vscale x 2 x bfloat> %a, <vscale x 2 x bfloat> %b, i32 -1)
+  ret <vscale x 2 x bfloat> %res
+}
+
+define <vscale x 2 x bfloat> @splice_nxv2bf16_offset_min(<vscale x 2 x bfloat> %a, <vscale x 2 x bfloat> %b) #0 {
+; CHECK-LABEL: splice_nxv2bf16_offset_min:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    csrr a0, vlenb
+; CHECK-NEXT:    srli a0, a0, 2
+; CHECK-NEXT:    addi a0, a0, -4
+; CHECK-NEXT:    vsetivli zero, 4, e16, mf2, ta, ma
+; CHECK-NEXT:    vslidedown.vx v8, v8, a0
+; CHECK-NEXT:    vsetvli a0, zero, e16, mf2, ta, ma
+; CHECK-NEXT:    vslideup.vi v8, v9, 4
+; CHECK-NEXT:    ret
+  %res = call <vscale x 2 x bfloat> @llvm.vector.splice.nxv2bf16(<vscale x 2 x bfloat> %a, <vscale x 2 x bfloat> %b, i32 -4)
+  ret <vscale x 2 x bfloat> %res
+}
+
+define <vscale x 2 x bfloat> @splice_nxv2bf16_offset_max(<vscale x 2 x bfloat> %a, <vscale x 2 x bfloat> %b) #0 {
+; CHECK-LABEL: splice_nxv2bf16_offset_max:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    csrr a0, vlenb
+; CHECK-NEXT:    srli a0, a0, 2
+; CHECK-NEXT:    addi a0, a0, -3
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, ma
+; CHECK-NEXT:    vslidedown.vi v8, v8, 3
+; CHECK-NEXT:    vsetvli a1, zero, e16, mf2, ta, ma
+; CHECK-NEXT:    vslideup.vx v8, v9, a0
+; CHECK-NEXT:    ret
+  %res = call <vscale x 2 x bfloat> @llvm.vector.splice.nxv2bf16(<vscale x 2 x bfloat> %a, <vscale x 2 x bfloat> %b, i32 3)
+  ret <vscale x 2 x bfloat> %res
+}
+
+declare <vscale x 4 x bfloat> @llvm.vector.splice.nxv4bf16(<vscale x 4 x bfloat>, <vscale x 4 x bfloat>, i32)
+
+define <vscale x 4 x bfloat> @splice_nxv4bf16_offset_zero(<vscale x 4 x bfloat> %a, <vscale x 4 x bfloat> %b) #0 {
+; CHECK-LABEL: splice_nxv4bf16_offset_zero:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    ret
+  %res = call <vscale x 4 x bfloat> @llvm.vector.splice.nxv4bf16(<vscale x 4 x bfloat> %a, <vscale x 4 x bfloat> %b, i32 0)
+  ret <vscale x 4 x bfloat> %res
+}
+
+define <vscale x 4 x bfloat> @splice_nxv4bf16_offset_negone(<vscale x 4 x bfloat> %a, <vscale x 4 x bfloat> %b) #0 {
+; CHECK-LABEL: splice_nxv4bf16_offset_negone:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    csrr a0, vlenb
+; CHECK-NEXT:    srli a0, a0, 1
+; CHECK-NEXT:    addi a0, a0, -1
+; CHECK-NEXT:    vsetvli a1, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vslidedown.vx v8, v8, a0
+; CHECK-NEXT:    vslideup.vi v8, v9, 1
+; CHECK-NEXT:    ret
+  %res = call <vscale x 4 x bfloat> @llvm.vector.splice.nxv4bf16(<vscale x 4 x bfloat> %a, <vscale x 4 x bfloat> %b, i32 -1)
+  ret <vscale x 4 x bfloat> %res
+}
+
+define <vscale x 4 x bfloat> @splice_nxv4bf16_offset_min(<vscale x 4 x bfloat> %a, <vscale x 4 x bfloat> %b) #0 {
+; CHECK-LABEL: splice_nxv4bf16_offset_min:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    csrr a0, vlenb
+; CHECK-NEXT:    srli a0, a0, 1
+; CHECK-NEXT:    addi a0, a0, -8
+; CHECK-NEXT:    vsetivli zero, 8, e16, m1, ta, ma
+; CHECK-NEXT:    vslidedown.vx v8, v8, a0
+; CHECK-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vslideup.vi v8, v9, 8
+; CHECK-NEXT:    ret
+  %res = call <vscale x 4 x bfloat> @llvm.vector.splice.nxv4bf16(<vscale x 4 x bfloat> %a, <vscale x 4 x bfloat> %b, i32 -8)
+  ret <vscale x 4 x bfloat> %res
+}
+
+define <vscale x 4 x bfloat> @splice_nxv4bf16_offset_max(<vscale x 4 x bfloat> %a, <vscale x 4 x bfloat> %b) #0 {
+; CHECK-LABEL: splice_nxv4bf16_offset_max:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    csrr a0, vlenb
+; CHECK-NEXT:    srli a0, a0, 1
+; CHECK-NEXT:    addi a0, a0, -7
+; CHECK-NEXT:    vsetvli zero, a0, e16, m1, ta, ma
+; CHECK-NEXT:    vslidedown.vi v8, v8, 7
+; CHECK-NEXT:    vsetvli a1, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vslideup.vx v8, v9, a0
+; CHECK-NEXT:    ret
+  %res = call <vscale x 4 x bfloat> @llvm.vector.splice.nxv4bf16(<vscale x 4 x bfloat> %a, <vscale x 4 x bfloat> %b, i32 7)
+  ret <vscale x 4 x bfloat> %res
+}
+
+declare <vscale x 8 x bfloat> @llvm.vector.splice.nxv8bf16(<vscale x 8 x bfloat>, <vscale x 8 x bfloat>, i32)
+
+define <vscale x 8 x bfloat> @splice_nxv8bf16_offset_zero(<vscale x 8 x bfloat> %a, <vscale x 8 x bfloat> %b) #0 {
+; CHECK-LABEL: splice_nxv8bf16_offset_zero:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    ret
+  %res = call <vscale x 8 x bfloat> @llvm.vector.splice.nxv8bf16(<vscale x 8 x bfloat> %a, <vscale x 8 x bfloat> %b, i32 0)
+  ret <vscale x 8 x bfloat> %res
+}
+
+define <vscale x 8 x bfloat> @splice_nxv8bf16_offset_negone(<vscale x 8 x bfloat> %a, <vscale x 8 x bfloat> %b) #0 {
+; CHECK-LABEL: splice_nxv8bf16_offset_negone:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    csrr a0, vlenb
+; CHECK-NEXT:    addi a0, a0, -1
+; CHECK-NEXT:    vsetivli zero, 1, e16, m2, ta, ma
+; CHECK-NEXT:    vslidedown.vx v8, v8, a0
+; CHECK-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vslideup.vi v8, v10, 1
+; CHECK-NEXT:    ret
+  %res = call <vscale x 8 x bfloat> @llvm.vector.splice.nxv8bf16(<vscale x 8 x bfloat> %a, <vscale x 8 x bfloat> %b, i32 -1)
+  ret <vscale x 8 x bfloat> %res
+}
+
+define <vscale x 8 x bfloat> @splice_nxv8bf16_offset_min(<vscale x 8 x bfloat> %a, <vscale x 8 x bfloat> %b) #0 {
+; CHECK-LABEL: splice_nxv8bf16_offset_min:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    csrr a0, vlenb
+; CHECK-NEXT:    addi a0, a0, -16
+; CHECK-NEXT:    vsetivli zero, 16, e16, m2, ta, ma
+; CHECK-NEXT:    vslidedown.vx v8, v8, a0
+; CHECK-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vslideup.vi v8, v10, 16
+; CHECK-NEXT:    ret
+  %res = call <vscale x 8 x bfloat> @llvm.vector.splice.nxv8bf16(<vscale x 8 x bfloat> %a, <vscale x 8 x bfloat> %b, i32 -16)
+  ret <vscale x 8 x bfloat> %res
+}
+
+define <vscale x 8 x bfloat> @splice_nxv8bf16_offset_max(<vscale x 8 x bfloat> %a, <vscale x 8 x bfloat> %b) #0 {
+; CHECK-LABEL: splice_nxv8bf16_offset_max:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    csrr a0, vlenb
+; CHECK-NEXT:    addi a0, a0, -15
+; CHECK-NEXT:    vsetvli zero, a0, e16,...
[truncated]

topperc
topperc approved these changes Oct 16, 2024
View reviewed changes
Copy link
Collaborator

@topperc topperc left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

lukel97
lukel97 commented Oct 16, 2024
View reviewed changes
llvm/test/Analysis/CostModel/RISCV/splice.ll
; SIZE-NEXT: Cost Model: Invalid cost for instruction: %splice.nxv8bf16 = call <vscale x 8 x bfloat> @llvm.vector.splice.nxv8bf16(<vscale x 8 x bfloat> zeroinitializer, <vscale x 8 x bfloat> zeroinitializer, i32 -1)
; SIZE-NEXT: Cost Model: Invalid cost for instruction: %splice.nxv16bf16 = call <vscale x 16 x bfloat> @llvm.vector.splice.nxv16bf16(<vscale x 16 x bfloat> zeroinitializer, <vscale x 16 x bfloat> zeroinitializer, i32 -1)
; SIZE-NEXT: Cost Model: Invalid cost for instruction: %splice.nxv32bf16 = call <vscale x 32 x bfloat> @llvm.vector.splice.nxv32bf16(<vscale x 32 x bfloat> zeroinitializer, <vscale x 32 x bfloat> zeroinitializer, i32 -1)
; SIZE-NEXT: Cost Model: Invalid cost for instruction: %splice.nxv64bf16 = call <vscale x 64 x bfloat> @llvm.vector.splice.nxv64bf16(<vscale x 64 x bfloat> zeroinitializer, <vscale x 64 x bfloat> zeroinitializer, i32 -1)
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not sure why the size is returning an invalid cost but the throughput is OK. Will look at it in another PR

@lukel97 lukel97 merged commit e88bcc1 into llvm:main Oct 16, 2024
8 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@topperc topperc topperc approved these changes

@preames preames Awaiting requested review from preames

@wangpc-pp wangpc-pp Awaiting requested review from wangpc-pp

Assignees
No one assigned
Labels
backend:RISC-V llvm:analysis
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@lukel97 @llvmbot @topperc