Skip to content
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

Enable customer lowering for fabs_v16f16 with AVX2 #72914

Merged
merged 1 commit into from
Nov 21, 2023
Merged

Enable customer lowering for fabs_v16f16 with AVX2 #72914

merged 1 commit into from
Nov 21, 2023

Conversation

david-xl
Copy link
Contributor

This is part-2 change to improve codegen for vec_fabs. In this patch, v16f16 and v132f16 fabs are improved.

There will be at least two followups patches after this one.

  1. fixing the ISEL crash when fabs.v32f16 uses custom lowering with AVX512
  2. better expansion for v16f16, v32f16 types on AVX1 subtargets.

@llvmbot
Copy link
Collaborator

llvmbot commented Nov 20, 2023

@llvm/pr-subscribers-backend-x86

Author: David Li (david-xl)

Changes

This is part-2 change to improve codegen for vec_fabs. In this patch, v16f16 and v132f16 fabs are improved.

There will be at least two followups patches after this one.

  1. fixing the ISEL crash when fabs.v32f16 uses custom lowering with AVX512
  2. better expansion for v16f16, v32f16 types on AVX1 subtargets.

Patch is 108.50 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/72914.diff

2 Files Affected:

  • (modified) llvm/lib/Target/X86/X86ISelLowering.cpp (+3)
  • (modified) llvm/test/CodeGen/X86/vec_fabs.ll (+23-1725)
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index deeab311320f0a2..5ec9b18f458a5bf 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -1596,6 +1596,8 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
       setOperationAction(ISD::STORE,              VT, Custom);
     }
     setF16Action(MVT::v16f16, Expand);
+    if (Subtarget.hasAVX2())
+      setOperationAction(ISD::FABS, MVT::v16f16, Custom);
     setOperationAction(ISD::FADD, MVT::v16f16, Expand);
     setOperationAction(ISD::FSUB, MVT::v16f16, Expand);
     setOperationAction(ISD::FMUL, MVT::v16f16, Expand);
@@ -1695,6 +1697,7 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
 
     for (auto VT : { MVT::v1i1, MVT::v2i1, MVT::v4i1, MVT::v8i1 })
       setOperationAction(ISD::EXTRACT_SUBVECTOR, VT, Custom);
+
   }
 
   // This block controls legalization for 512-bit operations with 8/16/32/64 bit
diff --git a/llvm/test/CodeGen/X86/vec_fabs.ll b/llvm/test/CodeGen/X86/vec_fabs.ll
index f691cb76bc684e6..ececfce210f563d 100644
--- a/llvm/test/CodeGen/X86/vec_fabs.ll
+++ b/llvm/test/CodeGen/X86/vec_fabs.ll
@@ -515,564 +515,17 @@ define <16 x half> @fabs_v16f16(ptr %p) {
 ;
 ; X86-AVX2-LABEL: fabs_v16f16:
 ; X86-AVX2:       # %bb.0:
-; X86-AVX2-NEXT:    pushl %esi
-; X86-AVX2-NEXT:    .cfi_def_cfa_offset 8
-; X86-AVX2-NEXT:    subl $372, %esp # imm = 0x174
-; X86-AVX2-NEXT:    .cfi_def_cfa_offset 380
-; X86-AVX2-NEXT:    .cfi_offset %esi, -8
-; X86-AVX2-NEXT:    movl {{[0-9]+}}(%esp), %esi
-; X86-AVX2-NEXT:    vmovdqa (%esi), %xmm0
-; X86-AVX2-NEXT:    vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    vmovaps 16(%esi), %xmm1
-; X86-AVX2-NEXT:    vmovups %xmm1, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    vpextrw $0, %xmm0, (%esp)
-; X86-AVX2-NEXT:    calll __extendhfsf2
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vpsrld $16, %xmm0, %xmm0
-; X86-AVX2-NEXT:    vpextrw $0, %xmm0, (%esp)
-; X86-AVX2-NEXT:    fstps {{[0-9]+}}(%esp)
-; X86-AVX2-NEXT:    vbroadcastss {{.*#+}} xmm1 = [NaN,NaN,NaN,NaN]
-; X86-AVX2-NEXT:    vmovups %xmm1, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-AVX2-NEXT:    vandps %xmm1, %xmm0, %xmm0
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __extendhfsf2
-; X86-AVX2-NEXT:    vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vmovss %xmm0, (%esp)
-; X86-AVX2-NEXT:    fstps {{[0-9]+}}(%esp)
-; X86-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-AVX2-NEXT:    vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __truncsfhf2
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vmovss %xmm0, (%esp)
-; X86-AVX2-NEXT:    calll __truncsfhf2
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vpextrw $0, %xmm0, (%esp)
-; X86-AVX2-NEXT:    calll __extendhfsf2
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vpsrlq $48, %xmm0, %xmm0
-; X86-AVX2-NEXT:    vpextrw $0, %xmm0, (%esp)
-; X86-AVX2-NEXT:    fstps {{[0-9]+}}(%esp)
-; X86-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-AVX2-NEXT:    vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __extendhfsf2
-; X86-AVX2-NEXT:    vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vmovss %xmm0, (%esp)
-; X86-AVX2-NEXT:    fstps {{[0-9]+}}(%esp)
-; X86-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-AVX2-NEXT:    vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __truncsfhf2
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vmovd %xmm0, (%esp)
-; X86-AVX2-NEXT:    vpinsrw $0, 4(%esi), %xmm0, %xmm0
-; X86-AVX2-NEXT:    vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __truncsfhf2
-; X86-AVX2-NEXT:    vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vpextrw $0, %xmm0, (%esp)
-; X86-AVX2-NEXT:    calll __extendhfsf2
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vpsrlq $48, %xmm0, %xmm0
-; X86-AVX2-NEXT:    vpextrw $0, %xmm0, (%esp)
-; X86-AVX2-NEXT:    fstps {{[0-9]+}}(%esp)
-; X86-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-AVX2-NEXT:    vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __extendhfsf2
-; X86-AVX2-NEXT:    vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vmovss %xmm0, (%esp)
-; X86-AVX2-NEXT:    fstps {{[0-9]+}}(%esp)
-; X86-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-AVX2-NEXT:    vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __truncsfhf2
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vmovd %xmm0, (%esp)
-; X86-AVX2-NEXT:    vpinsrw $0, 20(%esi), %xmm0, %xmm0
-; X86-AVX2-NEXT:    vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __truncsfhf2
-; X86-AVX2-NEXT:    vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vpextrw $0, %xmm0, (%esp)
-; X86-AVX2-NEXT:    calll __extendhfsf2
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; X86-AVX2-NEXT:    vpextrw $0, %xmm0, (%esp)
-; X86-AVX2-NEXT:    fstps {{[0-9]+}}(%esp)
-; X86-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-AVX2-NEXT:    vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __extendhfsf2
-; X86-AVX2-NEXT:    vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vmovss %xmm0, (%esp)
-; X86-AVX2-NEXT:    fstps {{[0-9]+}}(%esp)
-; X86-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-AVX2-NEXT:    vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __truncsfhf2
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vmovd %xmm0, (%esp)
-; X86-AVX2-NEXT:    vpinsrw $0, 8(%esi), %xmm0, %xmm0
-; X86-AVX2-NEXT:    vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __truncsfhf2
-; X86-AVX2-NEXT:    vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vpextrw $0, %xmm0, (%esp)
-; X86-AVX2-NEXT:    calll __extendhfsf2
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; X86-AVX2-NEXT:    vpextrw $0, %xmm0, (%esp)
-; X86-AVX2-NEXT:    fstps {{[0-9]+}}(%esp)
-; X86-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-AVX2-NEXT:    vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __extendhfsf2
-; X86-AVX2-NEXT:    vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vmovss %xmm0, (%esp)
-; X86-AVX2-NEXT:    fstps {{[0-9]+}}(%esp)
-; X86-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-AVX2-NEXT:    vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __truncsfhf2
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vmovd %xmm0, (%esp)
-; X86-AVX2-NEXT:    vpinsrw $0, 24(%esi), %xmm0, %xmm0
-; X86-AVX2-NEXT:    vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __truncsfhf2
-; X86-AVX2-NEXT:    vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vpextrw $0, %xmm0, (%esp)
-; X86-AVX2-NEXT:    calll __extendhfsf2
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; X86-AVX2-NEXT:    vpextrw $0, %xmm0, (%esp)
-; X86-AVX2-NEXT:    fstps {{[0-9]+}}(%esp)
-; X86-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-AVX2-NEXT:    vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __extendhfsf2
-; X86-AVX2-NEXT:    vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vmovss %xmm0, (%esp)
-; X86-AVX2-NEXT:    fstps {{[0-9]+}}(%esp)
-; X86-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-AVX2-NEXT:    vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __truncsfhf2
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vmovd %xmm0, (%esp)
-; X86-AVX2-NEXT:    vpinsrw $0, 12(%esi), %xmm0, %xmm0
-; X86-AVX2-NEXT:    vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __truncsfhf2
-; X86-AVX2-NEXT:    vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vpextrw $0, %xmm0, (%esp)
-; X86-AVX2-NEXT:    calll __extendhfsf2
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; X86-AVX2-NEXT:    vpextrw $0, %xmm0, (%esp)
-; X86-AVX2-NEXT:    fstps {{[0-9]+}}(%esp)
-; X86-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-AVX2-NEXT:    vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __extendhfsf2
-; X86-AVX2-NEXT:    vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vmovss %xmm0, (%esp)
-; X86-AVX2-NEXT:    fstps {{[0-9]+}}(%esp)
-; X86-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-AVX2-NEXT:    vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __truncsfhf2
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vmovd %xmm0, (%esp)
-; X86-AVX2-NEXT:    vpinsrw $0, 28(%esi), %xmm0, %xmm0
-; X86-AVX2-NEXT:    vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __truncsfhf2
-; X86-AVX2-NEXT:    vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vpextrw $0, %xmm0, (%esp)
-; X86-AVX2-NEXT:    calll __extendhfsf2
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vpsrld $16, %xmm0, %xmm0
-; X86-AVX2-NEXT:    vpextrw $0, %xmm0, (%esp)
-; X86-AVX2-NEXT:    fstps {{[0-9]+}}(%esp)
-; X86-AVX2-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-AVX2-NEXT:    vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __extendhfsf2
-; X86-AVX2-NEXT:    vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vmovss %xmm0, (%esp)
-; X86-AVX2-NEXT:    fstps {{[0-9]+}}(%esp)
-; X86-AVX2-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
-; X86-AVX2-NEXT:    vpand {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    calll __truncsfhf2
-; X86-AVX2-NEXT:    vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
-; X86-AVX2-NEXT:    vmovss %xmm1, (%esp)
-; X86-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
-; X86-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
-; X86-AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
-; X86-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
-; X86-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1],xmm2[2],mem[2],xmm2[3],mem[3]
-; X86-AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm2, %ymm1
-; X86-AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
-; X86-AVX2-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
-; X86-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
-; X86-AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0
-; X86-AVX2-NEXT:    vmovdqu %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
-; X86-AVX2-NEXT:    vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
-; X86-AVX2-NEXT:    vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
-; X86-AVX2-NEXT:    vzeroupper
-; X86-AVX2-NEXT:    calll __truncsfhf2
-; X86-AVX2-NEXT:    vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
-; X86-AVX2-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; X86-AVX2-NEXT:    vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
-; X86-AVX2-NEXT:    vpunpckldq {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; X86-AVX2-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
-; X86-AVX2-NEXT:    vpunpcklqdq {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; X86-AVX2-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; X86-AVX2-NEXT:    addl $372, %esp # imm = 0x174
-; X86-AVX2-NEXT:    .cfi_def_cfa_offset 8
-; X86-AVX2-NEXT:    popl %esi
-; X86-AVX2-NEXT:    .cfi_def_cfa_offset 4
+; X86-AVX2-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-AVX2-NEXT:    vpbroadcastw {{.*#+}} ymm0 = [NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN]
+; X86-AVX2-NEXT:    vpand (%eax), %ymm0, %ymm0
 ; X86-AVX2-NEXT:    retl
 ;
-; X86-AVX512VL-LABEL: fabs_v16f16:
-; X86-AVX512VL:       # %bb.0:
-; X86-AVX512VL-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-AVX512VL-NEXT:    movzwl 28(%eax), %ecx
-; X86-AVX512VL-NEXT:    vmovd %ecx, %xmm0
-; X86-AVX512VL-NEXT:    vcvtph2ps %xmm0, %xmm1
-; X86-AVX512VL-NEXT:    vpbroadcastd {{.*#+}} xmm0 = [NaN,NaN,NaN,NaN]
-; X86-AVX512VL-NEXT:    vpand %xmm0, %xmm1, %xmm1
-; X86-AVX512VL-NEXT:    vcvtps2ph $4, %xmm1, %xmm1
-; X86-AVX512VL-NEXT:    vmovd %xmm1, %ecx
-; X86-AVX512VL-NEXT:    vpinsrw $0, %ecx, %xmm0, %xmm3
-; X86-AVX512VL-NEXT:    vmovdqa (%eax), %xmm1
-; X86-AVX512VL-NEXT:    vmovdqa 16(%eax), %xmm2
-; X86-AVX512VL-NEXT:    vpsrldq {{.*#+}} xmm4 = xmm2[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; X86-AVX512VL-NEXT:    vpextrw $0, %xmm4, %ecx
-; X86-AVX512VL-NEXT:    movzwl %cx, %ecx
-; X86-AVX512VL-NEXT:    vmovd %ecx, %xmm4
-; X86-AVX512VL-NEXT:    vcvtph2ps %xmm4, %xmm4
-; X86-AVX512VL-NEXT:    vpand %xmm0, %xmm4, %xmm4
-; X86-AVX512VL-NEXT:    vcvtps2ph $4, %xmm4, %xmm4
-; X86-AVX512VL-NEXT:    vmovd %xmm4, %ecx
-; X86-AVX512VL-NEXT:    vpinsrw $0, %ecx, %xmm0, %xmm4
-; X86-AVX512VL-NEXT:    vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
-; X86-AVX512VL-NEXT:    movzwl 12(%eax), %ecx
-; X86-AVX512VL-NEXT:    vmovd %ecx, %xmm4
-; X86-AVX512VL-NEXT:    vcvtph2ps %xmm4, %xmm4
-; X86-AVX512VL-NEXT:    vpand %xmm0, %xmm4, %xmm4
-; X86-AVX512VL-NEXT:    vcvtps2ph $4, %xmm4, %xmm4
-; X86-AVX512VL-NEXT:    vmovd %xmm4, %ecx
-; X86-AVX512VL-NEXT:    vpinsrw $0, %ecx, %xmm0, %xmm4
-; X86-AVX512VL-NEXT:    vpsrldq {{.*#+}} xmm5 = xmm1[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; X86-AVX512VL-NEXT:    vpextrw $0, %xmm5, %ecx
-; X86-AVX512VL-NEXT:    movzwl %cx, %ecx
-; X86-AVX512VL-NEXT:    vmovd %ecx, %xmm5
-; X86-AVX512VL-NEXT:    vcvtph2ps %xmm5, %xmm5
-; X86-AVX512VL-NEXT:    vpand %xmm0, %xmm5, %xmm5
-; X86-AVX512VL-NEXT:    vcvtps2ph $4, %xmm5, %xmm5
-; X86-AVX512VL-NEXT:    vmovd %xmm5, %ecx
-; X86-AVX512VL-NEXT:    vpinsrw $0, %ecx, %xmm0, %xmm5
-; X86-AVX512VL-NEXT:    vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
-; X86-AVX512VL-NEXT:    vinserti128 $1, %xmm3, %ymm4, %ymm3
-; X86-AVX512VL-NEXT:    movzwl 24(%eax), %ecx
-; X86-AVX512VL-NEXT:    vmovd %ecx, %xmm4
-; X86-AVX512VL-NEXT:    vcvtph2ps %xmm4, %xmm4
-; X86-AVX512VL-NEXT:    vpand %xmm0, %xmm4, %xmm4
-; X86-AVX512VL-NEXT:    vcvtps2ph $4, %xmm4, %xmm4
-; X86-AVX512VL-NEXT:    vmovd %xmm4, %ecx
-; X86-AVX512VL-NEXT:    vpinsrw $0, %ecx, %xmm0, %xmm4
-; X86-AVX512VL-NEXT:    vpsrldq {{.*#+}} xmm5 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; X86-AVX512VL-NEXT:    vpextrw $0, %xmm5, %ecx
-; X86-AVX512VL-NEXT:    movzwl %cx, %ecx
-; X86-AVX512VL-NEXT:    vmovd %ecx, %xmm5
-; X86-AVX512VL-NEXT:    vcvtph2ps %xmm5, %xmm5
-; X86-AVX512VL-NEXT:    vpand %xmm0, %xmm5, %xmm5
-; X86-AVX512VL-NEXT:    vcvtps2ph $4, %xmm5, %xmm5
-; X86-AVX512VL-N...
[truncated]

@david-xl david-xl requested review from RKSimon and atanasyan and removed request for atanasyan November 20, 2023 21:08
RKSimon
RKSimon approved these changes Nov 21, 2023
View reviewed changes
Copy link
Collaborator

@RKSimon RKSimon left a comment

Choose a reason for hiding this comment

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

LGTM

@david-xl david-xl merged commit 30e8dcd into llvm:main Nov 21, 2023
3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@RKSimon RKSimon RKSimon approved these changes

Assignees
No one assigned
Labels
backend:X86
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@david-xl @llvmbot @RKSimon