[llvm] [SPARC] Add a test for `half` support (NFC) (PR #152723)
Trevor Gross via llvm-commits
llvm-commits at lists.llvm.org
Fri Aug 8 12:20:48 PDT 2025
https://github.com/tgross35 updated https://github.com/llvm/llvm-project/pull/152723
>From e2be3c1e46924ced0fda72324c4bf51d0b540c6e Mon Sep 17 00:00:00 2001
From: Trevor Gross <tmgross at umich.edu>
Date: Fri, 8 Aug 2025 08:03:18 -0500
Subject: [PATCH] [SPARC] Add a test for half support (NFC)
To prepare for updates to SPARC `half` support, add a version of the
`half.ll` test that is present on other targets.
---
llvm/test/CodeGen/SPARC/half.ll | 888 ++++++++++++++++++++++++++++++++
1 file changed, 888 insertions(+)
create mode 100644 llvm/test/CodeGen/SPARC/half.ll
diff --git a/llvm/test/CodeGen/SPARC/half.ll b/llvm/test/CodeGen/SPARC/half.ll
new file mode 100644
index 0000000000000..e7acdead5283e
--- /dev/null
+++ b/llvm/test/CodeGen/SPARC/half.ll
@@ -0,0 +1,888 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+
+; RUN: llc %s -o - -mtriple=sparc-linux-gnu | FileCheck %s -check-prefixes=CHECK,SMARC32,V8
+; RUN: llc %s -o - -mtriple=sparcel-linux-gnu | FileCheck %s -check-prefixes=CHECK,SMARC32,V8
+; RUN: llc %s -o - -mtriple=sparc-linux-gnu -mattr=v9 | FileCheck %s -check-prefixes=CHECK,SMARC32,V9
+; RUN: llc %s -o - -mtriple=sparc64-unknown-linux | FileCheck %s -check-prefixes=CHECK,SPARC64
+
+; Tests for various operations on half precison float. Much of the test is
+; copied from test/CodeGen/X86/half.ll.
+
+define void @store(half %x, ptr %p) nounwind {
+; SMARC32-LABEL: store:
+; SMARC32: ! %bb.0:
+; SMARC32-NEXT: save %sp, -96, %sp
+; SMARC32-NEXT: call __truncsfhf2
+; SMARC32-NEXT: mov %i0, %o0
+; SMARC32-NEXT: sth %o0, [%i1]
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: store:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -176, %sp
+; SPARC64-NEXT: call __truncsfhf2
+; SPARC64-NEXT: nop
+; SPARC64-NEXT: sth %o0, [%i1]
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ store half %x, ptr %p
+ ret void
+}
+
+define half @return(ptr %p) nounwind {
+; SMARC32-LABEL: return:
+; SMARC32: ! %bb.0:
+; SMARC32-NEXT: save %sp, -96, %sp
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: lduh [%i0], %o0
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: return:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -176, %sp
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: lduh [%i0], %o0
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %r = load half, ptr %p
+ ret half %r
+}
+
+define dso_local double @loadd(ptr nocapture readonly %a) local_unnamed_addr nounwind {
+; SMARC32-LABEL: loadd:
+; SMARC32: ! %bb.0: ! %entry
+; SMARC32-NEXT: save %sp, -96, %sp
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: lduh [%i0+2], %o0
+; SMARC32-NEXT: fstod %f0, %f0
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: loadd:
+; SPARC64: ! %bb.0: ! %entry
+; SPARC64-NEXT: save %sp, -176, %sp
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: lduh [%i0+2], %o0
+; SPARC64-NEXT: fstod %f0, %f0
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+entry:
+ %arrayidx = getelementptr inbounds i16, ptr %a, i64 1
+ %0 = load i16, ptr %arrayidx, align 2
+ %1 = tail call double @llvm.convert.from.fp16.f64(i16 %0)
+ ret double %1
+}
+
+define dso_local float @loadf(ptr nocapture readonly %a) local_unnamed_addr nounwind {
+; SMARC32-LABEL: loadf:
+; SMARC32: ! %bb.0: ! %entry
+; SMARC32-NEXT: save %sp, -96, %sp
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: lduh [%i0+2], %o0
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: loadf:
+; SPARC64: ! %bb.0: ! %entry
+; SPARC64-NEXT: save %sp, -176, %sp
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: lduh [%i0+2], %o0
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+entry:
+ %arrayidx = getelementptr inbounds i16, ptr %a, i64 1
+ %0 = load i16, ptr %arrayidx, align 2
+ %1 = tail call float @llvm.convert.from.fp16.f32(i16 %0)
+ ret float %1
+}
+
+define dso_local void @stored(ptr nocapture %a, double %b) local_unnamed_addr nounwind {
+; SMARC32-LABEL: stored:
+; SMARC32: ! %bb.0: ! %entry
+; SMARC32-NEXT: save %sp, -112, %sp
+; SMARC32-NEXT: mov %i2, %i3
+; SMARC32-NEXT: mov %i1, %i2
+; SMARC32-NEXT: std %i2, [%fp+-8]
+; SMARC32-NEXT: ldd [%fp+-8], %f0
+; SMARC32-NEXT: std %f0, [%fp+-16]
+; SMARC32-NEXT: call __truncdfhf2
+; SMARC32-NEXT: ldd [%fp+-16], %o0
+; SMARC32-NEXT: sth %o0, [%i0]
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: stored:
+; SPARC64: ! %bb.0: ! %entry
+; SPARC64-NEXT: save %sp, -176, %sp
+; SPARC64-NEXT: fmovd %f2, %f0
+; SPARC64-NEXT: call __truncdfhf2
+; SPARC64-NEXT: nop
+; SPARC64-NEXT: sth %o0, [%i0]
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+entry:
+ %0 = tail call i16 @llvm.convert.to.fp16.f64(double %b)
+ store i16 %0, ptr %a, align 2
+ ret void
+}
+
+define dso_local void @storef(ptr nocapture %a, float %b) local_unnamed_addr nounwind {
+; SMARC32-LABEL: storef:
+; SMARC32: ! %bb.0: ! %entry
+; SMARC32-NEXT: save %sp, -96, %sp
+; SMARC32-NEXT: call __truncsfhf2
+; SMARC32-NEXT: mov %i1, %o0
+; SMARC32-NEXT: sth %o0, [%i0]
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: storef:
+; SPARC64: ! %bb.0: ! %entry
+; SPARC64-NEXT: save %sp, -176, %sp
+; SPARC64-NEXT: fmovs %f3, %f1
+; SPARC64-NEXT: call __truncsfhf2
+; SPARC64-NEXT: nop
+; SPARC64-NEXT: sth %o0, [%i0]
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+entry:
+ %0 = tail call i16 @llvm.convert.to.fp16.f32(float %b)
+ store i16 %0, ptr %a, align 2
+ ret void
+}
+
+define void @test_load_store(ptr %in, ptr %out) nounwind {
+; CHECK-LABEL: test_load_store:
+; CHECK: ! %bb.0:
+; CHECK-NEXT: lduh [%o0], %o0
+; CHECK-NEXT: retl
+; CHECK-NEXT: sth %o0, [%o1]
+ %val = load half, ptr %in
+ store half %val, ptr %out
+ ret void
+}
+
+define i16 @test_bitcast_from_half(ptr %addr) nounwind {
+; CHECK-LABEL: test_bitcast_from_half:
+; CHECK: ! %bb.0:
+; CHECK-NEXT: retl
+; CHECK-NEXT: lduh [%o0], %o0
+ %val = load half, ptr %addr
+ %val_int = bitcast half %val to i16
+ ret i16 %val_int
+}
+
+define void @test_bitcast_to_half(ptr %addr, i16 %in) nounwind {
+; CHECK-LABEL: test_bitcast_to_half:
+; CHECK: ! %bb.0:
+; CHECK-NEXT: retl
+; CHECK-NEXT: sth %o1, [%o0]
+ %val_fp = bitcast i16 %in to half
+ store half %val_fp, ptr %addr
+ ret void
+}
+
+define half @from_bits(i16 %x) nounwind {
+; SMARC32-LABEL: from_bits:
+; SMARC32: ! %bb.0:
+; SMARC32-NEXT: save %sp, -96, %sp
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: mov %i0, %o0
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: from_bits:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -176, %sp
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: srl %i0, 0, %o0
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %res = bitcast i16 %x to half
+ ret half %res
+}
+
+define i16 @to_bits(half %x) nounwind {
+; SMARC32-LABEL: to_bits:
+; SMARC32: ! %bb.0:
+; SMARC32-NEXT: save %sp, -96, %sp
+; SMARC32-NEXT: call __truncsfhf2
+; SMARC32-NEXT: mov %i0, %o0
+; SMARC32-NEXT: sethi 4194240, %i0
+; SMARC32-NEXT: andn %o0, %i0, %i0
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: to_bits:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -176, %sp
+; SPARC64-NEXT: call __truncsfhf2
+; SPARC64-NEXT: nop
+; SPARC64-NEXT: sethi 63, %i0
+; SPARC64-NEXT: or %i0, 1023, %i0
+; SPARC64-NEXT: and %o0, %i0, %i0
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %res = bitcast half %x to i16
+ ret i16 %res
+}
+
+define float @test_extend32(ptr %addr) nounwind {
+; SMARC32-LABEL: test_extend32:
+; SMARC32: ! %bb.0:
+; SMARC32-NEXT: save %sp, -96, %sp
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: lduh [%i0], %o0
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: test_extend32:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -176, %sp
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: lduh [%i0], %o0
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %val16 = load half, ptr %addr
+ %val32 = fpext half %val16 to float
+ ret float %val32
+}
+
+define double @test_extend64(ptr %addr) nounwind {
+; SMARC32-LABEL: test_extend64:
+; SMARC32: ! %bb.0:
+; SMARC32-NEXT: save %sp, -96, %sp
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: lduh [%i0], %o0
+; SMARC32-NEXT: fstod %f0, %f0
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: test_extend64:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -176, %sp
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: lduh [%i0], %o0
+; SPARC64-NEXT: fstod %f0, %f0
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %val16 = load half, ptr %addr
+ %val32 = fpext half %val16 to double
+ ret double %val32
+}
+
+define void @test_trunc32(float %in, ptr %addr) nounwind {
+; SMARC32-LABEL: test_trunc32:
+; SMARC32: ! %bb.0:
+; SMARC32-NEXT: save %sp, -96, %sp
+; SMARC32-NEXT: call __truncsfhf2
+; SMARC32-NEXT: mov %i0, %o0
+; SMARC32-NEXT: sth %o0, [%i1]
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: test_trunc32:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -176, %sp
+; SPARC64-NEXT: call __truncsfhf2
+; SPARC64-NEXT: nop
+; SPARC64-NEXT: sth %o0, [%i1]
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %val16 = fptrunc float %in to half
+ store half %val16, ptr %addr
+ ret void
+}
+
+define void @test_trunc64(double %in, ptr %addr) nounwind {
+; SMARC32-LABEL: test_trunc64:
+; SMARC32: ! %bb.0:
+; SMARC32-NEXT: save %sp, -112, %sp
+; SMARC32-NEXT: ! kill: def $i1 killed $i1 killed $i0_i1 def $i0_i1
+; SMARC32-NEXT: ! kill: def $i0 killed $i0 killed $i0_i1 def $i0_i1
+; SMARC32-NEXT: std %i0, [%fp+-8]
+; SMARC32-NEXT: ldd [%fp+-8], %f0
+; SMARC32-NEXT: std %f0, [%fp+-16]
+; SMARC32-NEXT: call __truncdfhf2
+; SMARC32-NEXT: ldd [%fp+-16], %o0
+; SMARC32-NEXT: sth %o0, [%i2]
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: test_trunc64:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -176, %sp
+; SPARC64-NEXT: call __truncdfhf2
+; SPARC64-NEXT: nop
+; SPARC64-NEXT: sth %o0, [%i1]
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %val16 = fptrunc double %in to half
+ store half %val16, ptr %addr
+ ret void
+}
+
+define i64 @test_fptosi_i64(ptr %p) nounwind {
+; SMARC32-LABEL: test_fptosi_i64:
+; SMARC32: ! %bb.0:
+; SMARC32-NEXT: save %sp, -96, %sp
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: lduh [%i0], %o0
+; SMARC32-NEXT: st %f0, [%fp+-4]
+; SMARC32-NEXT: call __fixsfdi
+; SMARC32-NEXT: ld [%fp+-4], %o0
+; SMARC32-NEXT: mov %o0, %i0
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore %g0, %o1, %o1
+;
+; SPARC64-LABEL: test_fptosi_i64:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -192, %sp
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: lduh [%i0], %o0
+; SPARC64-NEXT: fstox %f0, %f0
+; SPARC64-NEXT: std %f0, [%fp+2039]
+; SPARC64-NEXT: ldx [%fp+2039], %i0
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %a = load half, ptr %p, align 2
+ %r = fptosi half %a to i64
+ ret i64 %r
+}
+
+define void @test_sitofp_i64(i64 %a, ptr %p) nounwind {
+; SMARC32-LABEL: test_sitofp_i64:
+; SMARC32: ! %bb.0:
+; SMARC32-NEXT: save %sp, -96, %sp
+; SMARC32-NEXT: mov %i1, %o1
+; SMARC32-NEXT: call __floatdisf
+; SMARC32-NEXT: mov %i0, %o0
+; SMARC32-NEXT: st %f0, [%fp+-4]
+; SMARC32-NEXT: call __truncsfhf2
+; SMARC32-NEXT: ld [%fp+-4], %o0
+; SMARC32-NEXT: sth %o0, [%i2]
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: test_sitofp_i64:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -192, %sp
+; SPARC64-NEXT: stx %i0, [%fp+2039]
+; SPARC64-NEXT: ldd [%fp+2039], %f0
+; SPARC64-NEXT: call __truncsfhf2
+; SPARC64-NEXT: fxtos %f0, %f1
+; SPARC64-NEXT: sth %o0, [%i1]
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %r = sitofp i64 %a to half
+ store half %r, ptr %p
+ ret void
+}
+
+define i64 @test_fptoui_i64(ptr %p) nounwind {
+; SMARC32-LABEL: test_fptoui_i64:
+; SMARC32: ! %bb.0:
+; SMARC32-NEXT: save %sp, -96, %sp
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: lduh [%i0], %o0
+; SMARC32-NEXT: st %f0, [%fp+-4]
+; SMARC32-NEXT: call __fixunssfdi
+; SMARC32-NEXT: ld [%fp+-4], %o0
+; SMARC32-NEXT: mov %o0, %i0
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore %g0, %o1, %o1
+;
+; SPARC64-LABEL: test_fptoui_i64:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -192, %sp
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: lduh [%i0], %o0
+; SPARC64-NEXT: sethi %h44(.LCPI17_0), %i0
+; SPARC64-NEXT: add %i0, %m44(.LCPI17_0), %i0
+; SPARC64-NEXT: sllx %i0, 12, %i0
+; SPARC64-NEXT: ld [%i0+%l44(.LCPI17_0)], %f1
+; SPARC64-NEXT: fsubs %f0, %f1, %f2
+; SPARC64-NEXT: fstox %f2, %f2
+; SPARC64-NEXT: std %f2, [%fp+2031]
+; SPARC64-NEXT: fstox %f0, %f2
+; SPARC64-NEXT: std %f2, [%fp+2039]
+; SPARC64-NEXT: ldx [%fp+2031], %i0
+; SPARC64-NEXT: sethi 0, %i1
+; SPARC64-NEXT: or %i1, 0, %i1
+; SPARC64-NEXT: sethi 2097152, %i2
+; SPARC64-NEXT: or %i2, 0, %i2
+; SPARC64-NEXT: sllx %i2, 32, %i2
+; SPARC64-NEXT: ldx [%fp+2039], %i3
+; SPARC64-NEXT: or %i2, %i1, %i1
+; SPARC64-NEXT: xor %i0, %i1, %i0
+; SPARC64-NEXT: fcmps %fcc0, %f0, %f1
+; SPARC64-NEXT: movl %fcc0, %i3, %i0
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %a = load half, ptr %p, align 2
+ %r = fptoui half %a to i64
+ ret i64 %r
+}
+
+define void @test_uitofp_i64(i64 %a, ptr %p) nounwind {
+; SMARC32-LABEL: test_uitofp_i64:
+; SMARC32: ! %bb.0:
+; SMARC32-NEXT: save %sp, -96, %sp
+; SMARC32-NEXT: mov %i1, %o1
+; SMARC32-NEXT: call __floatundisf
+; SMARC32-NEXT: mov %i0, %o0
+; SMARC32-NEXT: st %f0, [%fp+-4]
+; SMARC32-NEXT: call __truncsfhf2
+; SMARC32-NEXT: ld [%fp+-4], %o0
+; SMARC32-NEXT: sth %o0, [%i2]
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: test_uitofp_i64:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -192, %sp
+; SPARC64-NEXT: stx %i0, [%fp+2031]
+; SPARC64-NEXT: srlx %i0, 1, %i2
+; SPARC64-NEXT: and %i0, 1, %i3
+; SPARC64-NEXT: or %i3, %i2, %i2
+; SPARC64-NEXT: stx %i2, [%fp+2039]
+; SPARC64-NEXT: ldd [%fp+2031], %f0
+; SPARC64-NEXT: ldd [%fp+2039], %f2
+; SPARC64-NEXT: fxtos %f0, %f1
+; SPARC64-NEXT: fxtos %f2, %f0
+; SPARC64-NEXT: fadds %f0, %f0, %f0
+; SPARC64-NEXT: call __truncsfhf2
+; SPARC64-NEXT: fmovrslz %i0, %f0, %f1
+; SPARC64-NEXT: sth %o0, [%i1]
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %r = uitofp i64 %a to half
+ store half %r, ptr %p
+ ret void
+}
+
+define <4 x float> @test_extend32_vec4(ptr %p) nounwind {
+; SMARC32-LABEL: test_extend32_vec4:
+; SMARC32: ! %bb.0:
+; SMARC32-NEXT: save %sp, -104, %sp
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: lduh [%i0], %o0
+; SMARC32-NEXT: st %f0, [%fp+-4] ! 4-byte Folded Spill
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: lduh [%i0+2], %o0
+; SMARC32-NEXT: st %f0, [%fp+-8] ! 4-byte Folded Spill
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: lduh [%i0+4], %o0
+; SMARC32-NEXT: st %f0, [%fp+-12] ! 4-byte Folded Spill
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: lduh [%i0+6], %o0
+; SMARC32-NEXT: fmovs %f0, %f3
+; SMARC32-NEXT: ld [%fp+-4], %f0 ! 4-byte Folded Reload
+; SMARC32-NEXT: ld [%fp+-8], %f1 ! 4-byte Folded Reload
+; SMARC32-NEXT: ld [%fp+-12], %f2 ! 4-byte Folded Reload
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: test_extend32_vec4:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -192, %sp
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: lduh [%i0], %o0
+; SPARC64-NEXT: st %f0, [%fp+2043] ! 4-byte Folded Spill
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: lduh [%i0+2], %o0
+; SPARC64-NEXT: st %f0, [%fp+2039] ! 4-byte Folded Spill
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: lduh [%i0+4], %o0
+; SPARC64-NEXT: st %f0, [%fp+2035] ! 4-byte Folded Spill
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: lduh [%i0+6], %o0
+; SPARC64-NEXT: fmovs %f0, %f3
+; SPARC64-NEXT: ld [%fp+2043], %f0 ! 4-byte Folded Reload
+; SPARC64-NEXT: ld [%fp+2039], %f1 ! 4-byte Folded Reload
+; SPARC64-NEXT: ld [%fp+2035], %f2 ! 4-byte Folded Reload
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %a = load <4 x half>, ptr %p, align 8
+ %b = fpext <4 x half> %a to <4 x float>
+ ret <4 x float> %b
+}
+
+define <4 x double> @test_extend64_vec4(ptr %p) nounwind {
+; SMARC32-LABEL: test_extend64_vec4:
+; SMARC32: ! %bb.0:
+; SMARC32-NEXT: save %sp, -120, %sp
+; SMARC32-NEXT: ld [%fp+64], %i1
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: lduh [%i0+4], %o0
+; SMARC32-NEXT: st %f0, [%fp+-8] ! 4-byte Folded Spill
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: lduh [%i0+2], %o0
+; SMARC32-NEXT: st %f0, [%fp+-16] ! 4-byte Folded Spill
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: lduh [%i0], %o0
+; SMARC32-NEXT: lduh [%i0+6], %o0
+; SMARC32-NEXT: fstod %f0, %f0
+; SMARC32-NEXT: std %f0, [%fp+-24] ! 8-byte Folded Spill
+; SMARC32-NEXT: ld [%fp+-16], %f0 ! 4-byte Folded Reload
+; SMARC32-NEXT: fstod %f0, %f0
+; SMARC32-NEXT: std %f0, [%fp+-16] ! 8-byte Folded Spill
+; SMARC32-NEXT: ld [%fp+-8], %f0 ! 4-byte Folded Reload
+; SMARC32-NEXT: fstod %f0, %f0
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: std %f0, [%fp+-8]
+; SMARC32-NEXT: fstod %f0, %f0
+; SMARC32-NEXT: std %f0, [%i1+24]
+; SMARC32-NEXT: ldd [%fp+-8], %f0 ! 8-byte Folded Reload
+; SMARC32-NEXT: std %f0, [%i1+16]
+; SMARC32-NEXT: ldd [%fp+-16], %f0 ! 8-byte Folded Reload
+; SMARC32-NEXT: std %f0, [%i1+8]
+; SMARC32-NEXT: ldd [%fp+-24], %f0 ! 8-byte Folded Reload
+; SMARC32-NEXT: std %f0, [%i1]
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: test_extend64_vec4:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -208, %sp
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: lduh [%i0], %o0
+; SPARC64-NEXT: fstod %f0, %f0
+; SPARC64-NEXT: std %f0, [%fp+2039] ! 8-byte Folded Spill
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: lduh [%i0+2], %o0
+; SPARC64-NEXT: fstod %f0, %f0
+; SPARC64-NEXT: std %f0, [%fp+2031] ! 8-byte Folded Spill
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: lduh [%i0+4], %o0
+; SPARC64-NEXT: fstod %f0, %f0
+; SPARC64-NEXT: std %f0, [%fp+2023] ! 8-byte Folded Spill
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: lduh [%i0+6], %o0
+; SPARC64-NEXT: fstod %f0, %f6
+; SPARC64-NEXT: ldd [%fp+2039], %f0 ! 8-byte Folded Reload
+; SPARC64-NEXT: ldd [%fp+2031], %f2 ! 8-byte Folded Reload
+; SPARC64-NEXT: ldd [%fp+2023], %f4 ! 8-byte Folded Reload
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %a = load <4 x half>, ptr %p, align 8
+ %b = fpext <4 x half> %a to <4 x double>
+ ret <4 x double> %b
+}
+
+define void @test_trunc32_vec4(<4 x float> %a, ptr %p) nounwind {
+; SMARC32-LABEL: test_trunc32_vec4:
+; SMARC32: ! %bb.0:
+; SMARC32-NEXT: save %sp, -96, %sp
+; SMARC32-NEXT: call __truncsfhf2
+; SMARC32-NEXT: mov %i0, %o0
+; SMARC32-NEXT: mov %o0, %i0
+; SMARC32-NEXT: call __truncsfhf2
+; SMARC32-NEXT: mov %i1, %o0
+; SMARC32-NEXT: mov %o0, %i1
+; SMARC32-NEXT: call __truncsfhf2
+; SMARC32-NEXT: mov %i2, %o0
+; SMARC32-NEXT: mov %o0, %i2
+; SMARC32-NEXT: call __truncsfhf2
+; SMARC32-NEXT: mov %i3, %o0
+; SMARC32-NEXT: sth %o0, [%i4+6]
+; SMARC32-NEXT: sth %i2, [%i4+4]
+; SMARC32-NEXT: sth %i1, [%i4+2]
+; SMARC32-NEXT: sth %i0, [%i4]
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: test_trunc32_vec4:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -192, %sp
+; SPARC64-NEXT: st %f7, [%fp+2043] ! 4-byte Folded Spill
+; SPARC64-NEXT: st %f5, [%fp+2039] ! 4-byte Folded Spill
+; SPARC64-NEXT: call __truncsfhf2
+; SPARC64-NEXT: st %f3, [%fp+2035]
+; SPARC64-NEXT: mov %o0, %i0
+; SPARC64-NEXT: call __truncsfhf2
+; SPARC64-NEXT: ld [%fp+2035], %f1
+; SPARC64-NEXT: mov %o0, %i1
+; SPARC64-NEXT: call __truncsfhf2
+; SPARC64-NEXT: ld [%fp+2039], %f1
+; SPARC64-NEXT: mov %o0, %i2
+; SPARC64-NEXT: call __truncsfhf2
+; SPARC64-NEXT: ld [%fp+2043], %f1
+; SPARC64-NEXT: sth %o0, [%i4+6]
+; SPARC64-NEXT: sth %i2, [%i4+4]
+; SPARC64-NEXT: sth %i1, [%i4+2]
+; SPARC64-NEXT: sth %i0, [%i4]
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %v = fptrunc <4 x float> %a to <4 x half>
+ store <4 x half> %v, ptr %p
+ ret void
+}
+
+define void @test_trunc64_vec4(<4 x double> %a, ptr %p) nounwind {
+; SMARC32-LABEL: test_trunc64_vec4:
+; SMARC32: ! %bb.0:
+; SMARC32-NEXT: save %sp, -160, %sp
+; SMARC32-NEXT: ld [%fp+92], %g2
+; SMARC32-NEXT: ld [%fp+96], %g3
+; SMARC32-NEXT: ! kill: def $i5 killed $i5 killed $i4_i5 def $i4_i5
+; SMARC32-NEXT: ! kill: def $i3 killed $i3 killed $i2_i3 def $i2_i3
+; SMARC32-NEXT: ! kill: def $i1 killed $i1 killed $i0_i1 def $i0_i1
+; SMARC32-NEXT: std %g2, [%fp+-32]
+; SMARC32-NEXT: ! kill: def $i4 killed $i4 killed $i4_i5 def $i4_i5
+; SMARC32-NEXT: std %i4, [%fp+-24]
+; SMARC32-NEXT: ! kill: def $i2 killed $i2 killed $i2_i3 def $i2_i3
+; SMARC32-NEXT: std %i2, [%fp+-16]
+; SMARC32-NEXT: ! kill: def $i0 killed $i0 killed $i0_i1 def $i0_i1
+; SMARC32-NEXT: std %i0, [%fp+-8]
+; SMARC32-NEXT: ld [%fp+100], %i3
+; SMARC32-NEXT: ldd [%fp+-8], %f0
+; SMARC32-NEXT: ldd [%fp+-16], %f2
+; SMARC32-NEXT: ldd [%fp+-24], %f4
+; SMARC32-NEXT: ldd [%fp+-32], %f6
+; SMARC32-NEXT: std %f0, [%fp+-40]
+; SMARC32-NEXT: std %f2, [%fp+-48]
+; SMARC32-NEXT: std %f4, [%fp+-56]
+; SMARC32-NEXT: std %f6, [%fp+-64]
+; SMARC32-NEXT: call __truncdfhf2
+; SMARC32-NEXT: ldd [%fp+-40], %o0
+; SMARC32-NEXT: mov %o0, %i0
+; SMARC32-NEXT: call __truncdfhf2
+; SMARC32-NEXT: ldd [%fp+-48], %o0
+; SMARC32-NEXT: mov %o0, %i1
+; SMARC32-NEXT: call __truncdfhf2
+; SMARC32-NEXT: ldd [%fp+-56], %o0
+; SMARC32-NEXT: mov %o0, %i2
+; SMARC32-NEXT: call __truncdfhf2
+; SMARC32-NEXT: ldd [%fp+-64], %o0
+; SMARC32-NEXT: sth %o0, [%i3+6]
+; SMARC32-NEXT: sth %i2, [%i3+4]
+; SMARC32-NEXT: sth %i1, [%i3+2]
+; SMARC32-NEXT: sth %i0, [%i3]
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: test_trunc64_vec4:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -208, %sp
+; SPARC64-NEXT: std %f6, [%fp+2039] ! 8-byte Folded Spill
+; SPARC64-NEXT: std %f4, [%fp+2031] ! 8-byte Folded Spill
+; SPARC64-NEXT: call __truncdfhf2
+; SPARC64-NEXT: std %f2, [%fp+2023]
+; SPARC64-NEXT: mov %o0, %i0
+; SPARC64-NEXT: call __truncdfhf2
+; SPARC64-NEXT: ldd [%fp+2023], %f0
+; SPARC64-NEXT: mov %o0, %i1
+; SPARC64-NEXT: call __truncdfhf2
+; SPARC64-NEXT: ldd [%fp+2031], %f0
+; SPARC64-NEXT: mov %o0, %i2
+; SPARC64-NEXT: call __truncdfhf2
+; SPARC64-NEXT: ldd [%fp+2039], %f0
+; SPARC64-NEXT: sth %o0, [%i4+6]
+; SPARC64-NEXT: sth %i2, [%i4+4]
+; SPARC64-NEXT: sth %i1, [%i4+2]
+; SPARC64-NEXT: sth %i0, [%i4]
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %v = fptrunc <4 x double> %a to <4 x half>
+ store <4 x half> %v, ptr %p
+ ret void
+}
+
+define float @test_sitofp_fadd_i32(i32 %a, ptr %b) nounwind {
+; SMARC32-LABEL: test_sitofp_fadd_i32:
+; SMARC32: ! %bb.0:
+; SMARC32-NEXT: save %sp, -104, %sp
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: lduh [%i1], %o0
+; SMARC32-NEXT: st %i0, [%fp+-4]
+; SMARC32-NEXT: ld [%fp+-4], %f1
+; SMARC32-NEXT: st %f0, [%fp+-12] ! 4-byte Folded Spill
+; SMARC32-NEXT: fitos %f1, %f0
+; SMARC32-NEXT: st %f0, [%fp+-8]
+; SMARC32-NEXT: call __truncsfhf2
+; SMARC32-NEXT: ld [%fp+-8], %o0
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: nop
+; SMARC32-NEXT: ld [%fp+-12], %f1 ! 4-byte Folded Reload
+; SMARC32-NEXT: fadds %f1, %f0, %f0
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: test_sitofp_fadd_i32:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -192, %sp
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: lduh [%i1], %o0
+; SPARC64-NEXT: st %f0, [%fp+2039] ! 4-byte Folded Spill
+; SPARC64-NEXT: st %i0, [%fp+2043]
+; SPARC64-NEXT: ld [%fp+2043], %f0
+; SPARC64-NEXT: call __truncsfhf2
+; SPARC64-NEXT: fitos %f0, %f1
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: nop
+; SPARC64-NEXT: ld [%fp+2039], %f1 ! 4-byte Folded Reload
+; SPARC64-NEXT: fadds %f1, %f0, %f0
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %tmp0 = load half, ptr %b
+ %tmp1 = sitofp i32 %a to half
+ %tmp2 = fadd half %tmp0, %tmp1
+ %tmp3 = fpext half %tmp2 to float
+ ret float %tmp3
+}
+
+define half @PR40273(half) nounwind {
+; V8-LABEL: PR40273:
+; V8: ! %bb.0:
+; V8-NEXT: save %sp, -96, %sp
+; V8-NEXT: call __truncsfhf2
+; V8-NEXT: mov %i0, %o0
+; V8-NEXT: call __extendhfsf2
+; V8-NEXT: nop
+; V8-NEXT: sethi %hi(.LCPI24_0), %i0
+; V8-NEXT: ld [%i0+%lo(.LCPI24_0)], %f1
+; V8-NEXT: fcmps %f0, %f1
+; V8-NEXT: nop
+; V8-NEXT: fbne .LBB24_2
+; V8-NEXT: nop
+; V8-NEXT: ! %bb.1:
+; V8-NEXT: ba .LBB24_3
+; V8-NEXT: mov %g0, %i0
+; V8-NEXT: .LBB24_2:
+; V8-NEXT: mov 4, %i0
+; V8-NEXT: .LBB24_3:
+; V8-NEXT: sethi %hi(.LCPI24_1), %i1
+; V8-NEXT: add %i1, %lo(.LCPI24_1), %i1
+; V8-NEXT: ld [%i1+%i0], %f0
+; V8-NEXT: ret
+; V8-NEXT: restore
+;
+; V9-LABEL: PR40273:
+; V9: ! %bb.0:
+; V9-NEXT: save %sp, -96, %sp
+; V9-NEXT: call __truncsfhf2
+; V9-NEXT: mov %i0, %o0
+; V9-NEXT: call __extendhfsf2
+; V9-NEXT: nop
+; V9-NEXT: sethi %hi(.LCPI24_0), %i0
+; V9-NEXT: ld [%i0+%lo(.LCPI24_0)], %f1
+; V9-NEXT: mov %g0, %i0
+; V9-NEXT: fcmps %fcc0, %f0, %f1
+; V9-NEXT: movne %fcc0, 4, %i0
+; V9-NEXT: sethi %hi(.LCPI24_1), %i1
+; V9-NEXT: add %i1, %lo(.LCPI24_1), %i1
+; V9-NEXT: ld [%i1+%i0], %f0
+; V9-NEXT: ret
+; V9-NEXT: restore
+;
+; SPARC64-LABEL: PR40273:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -176, %sp
+; SPARC64-NEXT: call __truncsfhf2
+; SPARC64-NEXT: nop
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: nop
+; SPARC64-NEXT: sethi %h44(.LCPI24_0), %i0
+; SPARC64-NEXT: add %i0, %m44(.LCPI24_0), %i0
+; SPARC64-NEXT: sllx %i0, 12, %i0
+; SPARC64-NEXT: ld [%i0+%l44(.LCPI24_0)], %f1
+; SPARC64-NEXT: mov %g0, %i0
+; SPARC64-NEXT: fcmps %fcc0, %f0, %f1
+; SPARC64-NEXT: movne %fcc0, 4, %i0
+; SPARC64-NEXT: sethi %h44(.LCPI24_1), %i1
+; SPARC64-NEXT: add %i1, %m44(.LCPI24_1), %i1
+; SPARC64-NEXT: sllx %i1, 12, %i1
+; SPARC64-NEXT: add %i1, %l44(.LCPI24_1), %i1
+; SPARC64-NEXT: ld [%i1+%i0], %f0
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %2 = fcmp une half %0, 0xH0000
+ %3 = uitofp i1 %2 to half
+ ret half %3
+}
+
+define half @fabs(half %x) nounwind {
+; SMARC32-LABEL: fabs:
+; SMARC32: ! %bb.0:
+; SMARC32-NEXT: save %sp, -96, %sp
+; SMARC32-NEXT: call __truncsfhf2
+; SMARC32-NEXT: mov %i0, %o0
+; SMARC32-NEXT: call __extendhfsf2
+; SMARC32-NEXT: nop
+; SMARC32-NEXT: fabss %f0, %f0
+; SMARC32-NEXT: ret
+; SMARC32-NEXT: restore
+;
+; SPARC64-LABEL: fabs:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -176, %sp
+; SPARC64-NEXT: call __truncsfhf2
+; SPARC64-NEXT: nop
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: nop
+; SPARC64-NEXT: fabss %f0, %f0
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %a = call half @llvm.fabs.f16(half %x)
+ ret half %a
+}
+
+define half @fcopysign(half %x, half %y) nounwind {
+; V8-LABEL: fcopysign:
+; V8: ! %bb.0:
+; V8-NEXT: save %sp, -96, %sp
+; V8-NEXT: call __truncsfhf2
+; V8-NEXT: mov %i0, %o0
+; V8-NEXT: call __extendhfsf2
+; V8-NEXT: nop
+; V8-NEXT: sethi 2097152, %i0
+; V8-NEXT: and %i1, %i0, %i0
+; V8-NEXT: cmp %i0, 0
+; V8-NEXT: be .LBB26_2
+; V8-NEXT: fabss %f0, %f0
+; V8-NEXT: ! %bb.1:
+; V8-NEXT: fnegs %f0, %f0
+; V8-NEXT: .LBB26_2:
+; V8-NEXT: ret
+; V8-NEXT: restore
+;
+; V9-LABEL: fcopysign:
+; V9: ! %bb.0:
+; V9-NEXT: save %sp, -96, %sp
+; V9-NEXT: call __truncsfhf2
+; V9-NEXT: mov %i0, %o0
+; V9-NEXT: call __extendhfsf2
+; V9-NEXT: nop
+; V9-NEXT: sethi 2097152, %i0
+; V9-NEXT: and %i1, %i0, %i0
+; V9-NEXT: fabss %f0, %f0
+; V9-NEXT: fnegs %f0, %f1
+; V9-NEXT: cmp %i0, 0
+; V9-NEXT: fmovsne %icc, %f1, %f0
+; V9-NEXT: ret
+; V9-NEXT: restore
+;
+; SPARC64-LABEL: fcopysign:
+; SPARC64: ! %bb.0:
+; SPARC64-NEXT: save %sp, -192, %sp
+; SPARC64-NEXT: call __truncsfhf2
+; SPARC64-NEXT: st %f3, [%fp+2039]
+; SPARC64-NEXT: call __extendhfsf2
+; SPARC64-NEXT: nop
+; SPARC64-NEXT: ld [%fp+2039], %f1 ! 4-byte Folded Reload
+; SPARC64-NEXT: st %f1, [%fp+2043]
+; SPARC64-NEXT: ld [%fp+2043], %i0
+; SPARC64-NEXT: sethi 2097152, %i1
+; SPARC64-NEXT: and %i0, %i1, %i0
+; SPARC64-NEXT: fabss %f0, %f0
+; SPARC64-NEXT: fnegs %f0, %f1
+; SPARC64-NEXT: cmp %i0, 0
+; SPARC64-NEXT: fmovsne %icc, %f1, %f0
+; SPARC64-NEXT: ret
+; SPARC64-NEXT: restore
+ %a = call half @llvm.copysign.f16(half %x, half %y)
+ ret half %a
+}
More information about the llvm-commits
mailing list