[llvm] [Arm64EC] Add support for `fp128` (PR #206980)

Folkert de Vries via llvm-commits llvm-commits at lists.llvm.org
Wed Jul 1 06:33:02 PDT 2026


https://github.com/folkertdev created https://github.com/llvm/llvm-project/pull/206980

fixes https://github.com/llvm/llvm-project/issues/94434

Analogue to https://github.com/llvm/llvm-project/pull/152843 but for `fp128`.

Testing is a bit thin still, we need to fix some other problems like https://github.com/llvm/llvm-project/issues/144006 so that the arm64ec target can be added to more tests.

>From 44b39541a45152942ba67daaa327c4b0a34c5344 Mon Sep 17 00:00:00 2001
From: Folkert de Vries <folkert at folkertdev.nl>
Date: Wed, 1 Jul 2026 15:16:14 +0200
Subject: [PATCH] [Arm64EC] Add support for `fp128`

---
 .../AArch64/AArch64Arm64ECCallLowering.cpp    | 21 +++++++---
 .../CodeGen/AArch64/arm64ec-entry-thunks.ll   | 11 +++---
 .../CodeGen/AArch64/arm64ec-exit-thunks.ll    | 38 ++++++++++---------
 .../CodeGen/Generic/fp128-exp10-libcall.ll    |  5 ++-
 .../CodeGen/Generic/fp128-math-libcalls.ll    |  8 ++--
 5 files changed, 49 insertions(+), 34 deletions(-)

diff --git a/llvm/lib/Target/AArch64/AArch64Arm64ECCallLowering.cpp b/llvm/lib/Target/AArch64/AArch64Arm64ECCallLowering.cpp
index 44f76f9e8772b..147fa490c32a3 100644
--- a/llvm/lib/Target/AArch64/AArch64Arm64ECCallLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64Arm64ECCallLowering.cpp
@@ -333,9 +333,16 @@ ThunkArgInfo AArch64Arm64ECCallLowering::canonicalizeThunkType(
     return direct(T);
   }
 
+  if (T->isFP128Ty()) {
+    // Prefix with `llvm` since MSVC doesn't specify `_Float128`
+    Out << "__llvm_q__";
+    // f128 uses sret for compatibility with GCC.
+    return pointerIndirection(T);
+  }
+
   if (T->isFloatingPointTy()) {
-    report_fatal_error("Only 16, 32, and 64 bit floating points are supported "
-                       "for ARM64EC thunks");
+    report_fatal_error("Only 16, 32, 64, and 128 bit IEEE floating points "
+                       "are supported for ARM64EC thunks");
   }
 
   auto &DL = M->getDataLayout();
@@ -350,7 +357,7 @@ ThunkArgInfo AArch64Arm64ECCallLowering::canonicalizeThunkType(
     uint64_t ElementSizePerBytes = DL.getTypeSizeInBits(ElementTy) / 8;
     uint64_t TotalSizeBytes = ElementCnt * ElementSizePerBytes;
     if (ElementTy->isHalfTy() || ElementTy->isFloatTy() ||
-        ElementTy->isDoubleTy()) {
+        ElementTy->isDoubleTy() || ElementTy->isFP128Ty()) {
       if (ElementTy->isHalfTy())
         // Prefix with `llvm` since MSVC doesn't specify `_Float16`
         Out << "__llvm_H__";
@@ -358,6 +365,9 @@ ThunkArgInfo AArch64Arm64ECCallLowering::canonicalizeThunkType(
         Out << "F";
       else if (ElementTy->isDoubleTy())
         Out << "D";
+      else if (ElementTy->isFP128Ty())
+        // Prefix with `llvm` since MSVC doesn't specify `_Float128`
+        Out << "__llvm_Q__";
       Out << TotalSizeBytes;
       if (Alignment.value() >= 16 && !Ret)
         Out << "a" << Alignment.value();
@@ -370,9 +380,8 @@ ThunkArgInfo AArch64Arm64ECCallLowering::canonicalizeThunkType(
         return pointerIndirection(T);
       }
     } else if (T->isFloatingPointTy()) {
-      report_fatal_error(
-          "Only 16, 32, and 64 bit floating points are supported "
-          "for ARM64EC thunks");
+      report_fatal_error("Only 16, 32, 64, and 128 bit IEEE floating points "
+                         "are supported for ARM64EC thunks");
     }
   }
 
diff --git a/llvm/test/CodeGen/AArch64/arm64ec-entry-thunks.ll b/llvm/test/CodeGen/AArch64/arm64ec-entry-thunks.ll
index 09fe884940c6a..e1ffd7655c232 100644
--- a/llvm/test/CodeGen/AArch64/arm64ec-entry-thunks.ll
+++ b/llvm/test/CodeGen/AArch64/arm64ec-entry-thunks.ll
@@ -85,10 +85,10 @@ define i64 @simple_integers(i8, i16, i32, i64) nounwind {
   ret i64 0
 }
 
-; NOTE: Only half, float, and double are supported.
-define double @simple_floats(half, float, double) nounwind {
-; CHECK-LABEL:    .def    $ientry_thunk$cdecl$d$__llvm_h__fd;
-; CHECK:          .section        .wowthk$aa,"xr",discard,$ientry_thunk$cdecl$d$__llvm_h__fd
+; NOTE: Only half, float, double, and fp128 are supported.
+define double @simple_floats(half, float, double, fp128) nounwind {
+; CHECK-LABEL:    .def    $ientry_thunk$cdecl$d$__llvm_h__fd__llvm_q__;
+; CHECK:          .section        .wowthk$aa,"xr",discard,$ientry_thunk$cdecl$d$__llvm_h__fd__llvm_q__
 ; CHECK:          // %bb.0:
 ; CHECK-NEXT:     stp     q6, q7, [sp, #-176]!            // 32-byte Folded Spill
 ; CHECK-NEXT:     .seh_save_any_reg_px    q6, 176
@@ -105,6 +105,7 @@ define double @simple_floats(half, float, double) nounwind {
 ; CHECK-NEXT:     add     x29, sp, #160
 ; CHECK-NEXT:     .seh_add_fp     160
 ; CHECK-NEXT:     .seh_endprologue
+; CHECK-NEXT:     ldr     q3, [x3]
 ; CHECK-NEXT:     blr     x9
 ; CHECK-NEXT:     adrp    x8, __os_arm64x_dispatch_ret
 ; CHECK-NEXT:     ldr     x0, [x8, :lo12:__os_arm64x_dispatch_ret]
@@ -602,7 +603,7 @@ start:
 ; CHECK-NEXT:     .symidx $ientry_thunk$cdecl$i8$i8i8i8i8
 ; CHECK-NEXT:     .word   1
 ; CHECK-NEXT:     .symidx "#simple_floats"
-; CHECK-NEXT:     .symidx $ientry_thunk$cdecl$d$__llvm_h__fd
+; CHECK-NEXT:     .symidx $ientry_thunk$cdecl$d$__llvm_h__fd__llvm_q__
 ; CHECK-NEXT:     .word   1
 ; CHECK-NEXT:     .symidx "#has_varargs"
 ; CHECK-NEXT:     .symidx $ientry_thunk$cdecl$v$varargs
diff --git a/llvm/test/CodeGen/AArch64/arm64ec-exit-thunks.ll b/llvm/test/CodeGen/AArch64/arm64ec-exit-thunks.ll
index 6fba6a3974574..22ca2bb59fd87 100644
--- a/llvm/test/CodeGen/AArch64/arm64ec-exit-thunks.ll
+++ b/llvm/test/CodeGen/AArch64/arm64ec-exit-thunks.ll
@@ -93,26 +93,28 @@ declare i64 @simple_integers(i8, i16, i32, i64) nounwind;
 ; CHECK-NEXT:     .seh_endfunclet
 ; CHECK-NEXT:     .seh_endproc
 
-; NOTE: Only half, float, and double are supported.
-declare double @simple_floats(half, float, double) nounwind;
-; CHECK-LABEL:    .def    $iexit_thunk$cdecl$d$__llvm_h__fd;
-; CHECK:          .section        .wowthk$aa,"xr",discard,$iexit_thunk$cdecl$d$__llvm_h__fd
+; NOTE: Only half, float, double, and fp128 are supported.
+declare double @simple_floats(half, float, double, fp128) nounwind;
+; CHECK-LABEL:    .def    $iexit_thunk$cdecl$d$__llvm_h__fd__llvm_q__;
+; CHECK:          .section        .wowthk$aa,"xr",discard,$iexit_thunk$cdecl$d$__llvm_h__fd__llvm_q__
 ; CHECK:          // %bb.0:
-; CHECK-NEXT:     sub     sp, sp, #48
-; CHECK-NEXT:     .seh_stackalloc 48
-; CHECK-NEXT:     stp     x29, x30, [sp, #32]             // 16-byte Folded Spill
-; CHECK-NEXT:     .seh_save_fplr  32
-; CHECK-NEXT:     add     x29, sp, #32
-; CHECK-NEXT:     .seh_add_fp     32
+; CHECK-NEXT:     sub     sp, sp, #64
+; CHECK-NEXT:     .seh_stackalloc 64
+; CHECK-NEXT:     stp     x29, x30, [sp, #48]             // 16-byte Folded Spill
+; CHECK-NEXT:     .seh_save_fplr  48
+; CHECK-NEXT:     add     x29, sp, #48
+; CHECK-NEXT:     .seh_add_fp     48
 ; CHECK-NEXT:     .seh_endprologue
 ; CHECK-NEXT:     adrp    x8, __os_arm64x_dispatch_call_no_redirect
+; CHECK-NEXT:     sub     x3, x29, #16
+; CHECK-NEXT:     stur    q3, [x29, #-16]
 ; CHECK-NEXT:     ldr     x16, [x8, :lo12:__os_arm64x_dispatch_call_no_redirect]
 ; CHECK-NEXT:     blr     x16
 ; CHECK-NEXT:     .seh_startepilogue
-; CHECK-NEXT:     ldp     x29, x30, [sp, #32]             // 16-byte Folded Reload
-; CHECK-NEXT:     .seh_save_fplr  32
-; CHECK-NEXT:     add     sp, sp, #48
-; CHECK-NEXT:     .seh_stackalloc 48
+; CHECK-NEXT:     ldp     x29, x30, [sp, #48]             // 16-byte Folded Reload
+; CHECK-NEXT:     .seh_save_fplr  48
+; CHECK-NEXT:     add     sp, sp, #64
+; CHECK-NEXT:     .seh_stackalloc 64
 ; CHECK-NEXT:     .seh_endepilogue
 ; CHECK-NEXT:     ret
 ; CHECK-NEXT:     .seh_endfunclet
@@ -129,8 +131,8 @@ declare double @simple_floats(half, float, double) nounwind;
 ; CHECK-NEXT:     adrp    x11, simple_floats
 ; CHECK-NEXT:     add     x11, x11, :lo12:simple_floats
 ; CHECK-NEXT:     ldr     x8, [x8, :lo12:__os_arm64x_check_icall]
-; CHECK-NEXT:     adrp    x10, $iexit_thunk$cdecl$d$__llvm_h__fd
-; CHECK-NEXT:     add     x10, x10, :lo12:$iexit_thunk$cdecl$d$__llvm_h__fd
+; CHECK-NEXT:     adrp    x10, $iexit_thunk$cdecl$d$__llvm_h__fd__llvm_q__
+; CHECK-NEXT:     add     x10, x10, :lo12:$iexit_thunk$cdecl$d$__llvm_h__fd__llvm_q__
 ; CHECK-NEXT:     blr     x8
 ; CHECK-NEXT:     .seh_startepilogue
 ; CHECK-NEXT:     ldr     x30, [sp], #16                  // 8-byte Folded Reload
@@ -615,7 +617,7 @@ declare void @"??@md5mangleaaaaaaaaaaaaaaaaaaaaaaa@"()
 ; CHECK-NEXT:     .symidx simple_integers
 ; CHECK-NEXT:     .word   0
 ; CHECK-NEXT:     .symidx simple_floats
-; CHECK-NEXT:     .symidx $iexit_thunk$cdecl$d$__llvm_h__fd
+; CHECK-NEXT:     .symidx $iexit_thunk$cdecl$d$__llvm_h__fd__llvm_q__
 ; CHECK-NEXT:     .word   4
 ; CHECK-NEXT:     .symidx "#simple_floats$exit_thunk"
 ; CHECK-NEXT:     .symidx simple_floats
@@ -678,7 +680,7 @@ declare void @"??@md5mangleaaaaaaaaaaaaaaaaaaaaaaa@"()
 define void @func_caller() nounwind {
   call void @no_op()
   call i64 @simple_integers(i8 0, i16 0, i32 0, i64 0)
-  call double @simple_floats(half 0.0, float 0.0, double 0.0)
+  call double @simple_floats(half 0.0, float 0.0, double 0.0, fp128 0.0)
   call void (...) @has_varargs()
   %c = alloca i8
   call void @has_sret(ptr sret([100 x i8]) %c)
diff --git a/llvm/test/CodeGen/Generic/fp128-exp10-libcall.ll b/llvm/test/CodeGen/Generic/fp128-exp10-libcall.ll
index 5e97f03c2bc3b..50956b5e93f38 100644
--- a/llvm/test/CodeGen/Generic/fp128-exp10-libcall.ll
+++ b/llvm/test/CodeGen/Generic/fp128-exp10-libcall.ll
@@ -13,7 +13,10 @@
 ; RUN: %if x86-registered-target     %{ llc < %s -mtriple=i686-unknown-linux-musl      | FileCheck %s --check-prefixes=CHECK-ALL,CHECK-USELD %}
 ; RUN: %if x86-registered-target     %{ llc < %s -mtriple=x86_64-unknown-linux-gnu     | FileCheck %s --check-prefixes=CHECK-ALL,CHECK-F128  %}
 ; RUN: %if x86-registered-target     %{ llc < %s -mtriple=x86_64-unknown-linux-musl    | FileCheck %s --check-prefixes=CHECK-ALL,CHECK-USELD %}
-; RUN %if x86-registered-target     %{ llc < %s -mtriple=x86_64-pc-windows-msvc        | FileCheck %s --check-prefixes=CHECK-ALL,CHECK-F128  %}
+
+; FIXME(#144006): Windows-MSVC should also be run but has a ldexp selection failure.
+; %if x86-registered-target     %{ llc < %s -mtriple=x86_64-pc-windows-msvc       | FileCheck %s --check-prefixes=CHECK-ALL,CHECK-F128  %}
+; %if aarch64-registered-target %{ llc < %s -mtriple=arm64ec-pc-windows-msvc      | FileCheck %s --check-prefixes=CHECK-ALL,CHECK-F128  %}
 
 ; ERR: error: no libcall available for fexp10
 define fp128 @test_exp10(fp128 %a) {
diff --git a/llvm/test/CodeGen/Generic/fp128-math-libcalls.ll b/llvm/test/CodeGen/Generic/fp128-math-libcalls.ll
index f759c94621381..ed8b9b0789c2e 100644
--- a/llvm/test/CodeGen/Generic/fp128-math-libcalls.ll
+++ b/llvm/test/CodeGen/Generic/fp128-math-libcalls.ll
@@ -30,10 +30,10 @@
 ; RUN: %if x86-registered-target     %{ llc < %s -mtriple=i686-unknown-linux-musl      | FileCheck %s --check-prefixes=CHECK-ALL,CHECK-USELD %}
 ; RUN: %if x86-registered-target     %{ llc < %s -mtriple=x86_64-unknown-linux-gnu     | FileCheck %s --check-prefixes=CHECK-ALL,CHECK-F128  %}
 ; RUN: %if x86-registered-target     %{ llc < %s -mtriple=x86_64-unknown-linux-musl    | FileCheck %s --check-prefixes=CHECK-ALL,CHECK-USELD %}
-;
-; FIXME(#144006): Windows-MSVC should also be run but has a ldexp selection
-; failure.
-; %if x86-registered-target     %{ llc < %s -mtriple=x86_64-pc-windows-msvc       -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK-ALL,CHECK-F128  %}
+
+; FIXME(#144006): Windows-MSVC should also be run but has a ldexp selection failure.
+; %if x86-registered-target     %{ llc < %s -mtriple=x86_64-pc-windows-msvc       | FileCheck %s --check-prefixes=CHECK-ALL,CHECK-F128  %}
+; %if aarch64-registered-target %{ llc < %s -mtriple=arm64ec-pc-windows-msvc      | FileCheck %s --check-prefixes=CHECK-ALL,CHECK-F128  %}
 
 define fp128 @test_acos(fp128 %a) {
 ; CHECK-ALL-LABEL:  test_acos:



More information about the llvm-commits mailing list