5c2a133b13
This change is part of this proposal: https://discourse.llvm.org/t/rfc-all-the-math-intrinsics/78294 - `Builtins.td` - Add f16 support for libm atan2 builtin - `CGBuiltin.cpp` - Emit constraint atan2 intrinsic for clang builtin - `clang/test/CodeGenCXX/builtin-calling-conv.cpp` - Use erff instead of atan2 for clang builtin to lib call calling convention check, now that atan2 maps to an intrinsic. - add atan2 cases to llvm.experimental.constrained tests for more backends: ARM, PowerPC, RISCV, SystemZ. - LangRef.rst: add llvm.experimental.constrained.atan2, revise llvm.atan2 description. Last part of Implement the atan2 HLSL Function. Fixes #70096.
195 lines
6.4 KiB
LLVM
195 lines
6.4 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -mtriple powerpc64le < %s | FileCheck %s
|
|
|
|
; Check constrained ops converted to call
|
|
define void @test(ptr %cast) strictfp {
|
|
; CHECK-LABEL: test:
|
|
; CHECK: # %bb.0: # %root
|
|
; CHECK-NEXT: mflr 0
|
|
; CHECK-NEXT: .cfi_def_cfa_offset 64
|
|
; CHECK-NEXT: .cfi_offset lr, 16
|
|
; CHECK-NEXT: .cfi_offset r29, -24
|
|
; CHECK-NEXT: .cfi_offset r30, -16
|
|
; CHECK-NEXT: std 29, -24(1) # 8-byte Folded Spill
|
|
; CHECK-NEXT: std 30, -16(1) # 8-byte Folded Spill
|
|
; CHECK-NEXT: stdu 1, -64(1)
|
|
; CHECK-NEXT: addi 30, 3, -8
|
|
; CHECK-NEXT: li 29, 255
|
|
; CHECK-NEXT: std 0, 80(1)
|
|
; CHECK-NEXT: .p2align 5
|
|
; CHECK-NEXT: .LBB0_1: # %for.body
|
|
; CHECK-NEXT: #
|
|
; CHECK-NEXT: lfdu 1, 8(30)
|
|
; CHECK-NEXT: bl cos
|
|
; CHECK-NEXT: nop
|
|
; CHECK-NEXT: addi 29, 29, -1
|
|
; CHECK-NEXT: stfd 1, 0(30)
|
|
; CHECK-NEXT: cmpldi 29, 0
|
|
; CHECK-NEXT: bc 12, 1, .LBB0_1
|
|
; CHECK-NEXT: # %bb.2: # %exit
|
|
; CHECK-NEXT: addi 1, 1, 64
|
|
; CHECK-NEXT: ld 0, 16(1)
|
|
; CHECK-NEXT: ld 30, -16(1) # 8-byte Folded Reload
|
|
; CHECK-NEXT: ld 29, -24(1) # 8-byte Folded Reload
|
|
; CHECK-NEXT: mtlr 0
|
|
; CHECK-NEXT: blr
|
|
root:
|
|
br label %for.body
|
|
|
|
exit:
|
|
ret void
|
|
|
|
for.body:
|
|
%i = phi i64 [ 0, %root ], [ %next, %for.body ]
|
|
%idx = getelementptr inbounds double, ptr %cast, i64 %i
|
|
%val = load double, ptr %idx
|
|
%cos = tail call nnan ninf nsz arcp double @llvm.experimental.constrained.cos.f64(double %val, metadata !"round.dynamic", metadata !"fpexcept.strict")
|
|
store double %cos, ptr %idx, align 8
|
|
%next = add nuw nsw i64 %i, 1
|
|
%cond = icmp eq i64 %next, 255
|
|
br i1 %cond, label %exit, label %for.body
|
|
}
|
|
|
|
; Check constrained ops converted to native instruction
|
|
define void @test2(ptr %cast) strictfp {
|
|
; CHECK-LABEL: test2:
|
|
; CHECK: # %bb.0: # %entry
|
|
; CHECK-NEXT: li 4, 255
|
|
; CHECK-NEXT: addi 3, 3, -8
|
|
; CHECK-NEXT: mtctr 4
|
|
; CHECK-NEXT: .p2align 4
|
|
; CHECK-NEXT: .LBB1_1: # %for.body
|
|
; CHECK-NEXT: #
|
|
; CHECK-NEXT: lfdu 0, 8(3)
|
|
; CHECK-NEXT: xssqrtdp 0, 0
|
|
; CHECK-NEXT: stfd 0, 0(3)
|
|
; CHECK-NEXT: bdnz .LBB1_1
|
|
; CHECK-NEXT: # %bb.2: # %exit
|
|
; CHECK-NEXT: blr
|
|
entry:
|
|
br label %for.body
|
|
|
|
for.body:
|
|
%i = phi i64 [ 0, %entry ], [ %next, %for.body ]
|
|
%idx = getelementptr inbounds double, ptr %cast, i64 %i
|
|
%val = load double, ptr %idx
|
|
%cos = tail call nnan ninf nsz arcp double @llvm.experimental.constrained.sqrt.f64(double %val, metadata !"round.dynamic", metadata !"fpexcept.strict")
|
|
store double %cos, ptr %idx, align 8
|
|
%next = add nuw nsw i64 %i, 1
|
|
%cond = icmp eq i64 %next, 255
|
|
br i1 %cond, label %exit, label %for.body
|
|
|
|
exit:
|
|
ret void
|
|
}
|
|
|
|
; Check constrained ops converted to call
|
|
define void @testTan(ptr %cast) strictfp {
|
|
; CHECK-LABEL: testTan:
|
|
; CHECK: # %bb.0: # %root
|
|
; CHECK-NEXT: mflr 0
|
|
; CHECK-NEXT: .cfi_def_cfa_offset 64
|
|
; CHECK-NEXT: .cfi_offset lr, 16
|
|
; CHECK-NEXT: .cfi_offset r29, -24
|
|
; CHECK-NEXT: .cfi_offset r30, -16
|
|
; CHECK-NEXT: std 29, -24(1) # 8-byte Folded Spill
|
|
; CHECK-NEXT: std 30, -16(1) # 8-byte Folded Spill
|
|
; CHECK-NEXT: stdu 1, -64(1)
|
|
; CHECK-NEXT: addi 30, 3, -8
|
|
; CHECK-NEXT: li 29, 255
|
|
; CHECK-NEXT: std 0, 80(1)
|
|
; CHECK-NEXT: .p2align 5
|
|
; CHECK-NEXT: .LBB2_1: # %for.body
|
|
; CHECK-NEXT: #
|
|
; CHECK-NEXT: lfdu 1, 8(30)
|
|
; CHECK-NEXT: bl tan
|
|
; CHECK-NEXT: nop
|
|
; CHECK-NEXT: addi 29, 29, -1
|
|
; CHECK-NEXT: stfd 1, 0(30)
|
|
; CHECK-NEXT: cmpldi 29, 0
|
|
; CHECK-NEXT: bc 12, 1, .LBB2_1
|
|
; CHECK-NEXT: # %bb.2: # %exit
|
|
; CHECK-NEXT: addi 1, 1, 64
|
|
; CHECK-NEXT: ld 0, 16(1)
|
|
; CHECK-NEXT: ld 30, -16(1) # 8-byte Folded Reload
|
|
; CHECK-NEXT: ld 29, -24(1) # 8-byte Folded Reload
|
|
; CHECK-NEXT: mtlr 0
|
|
; CHECK-NEXT: blr
|
|
root:
|
|
br label %for.body
|
|
|
|
exit:
|
|
ret void
|
|
|
|
for.body:
|
|
%i = phi i64 [ 0, %root ], [ %next, %for.body ]
|
|
%idx = getelementptr inbounds double, ptr %cast, i64 %i
|
|
%val = load double, ptr %idx
|
|
%tan = tail call nnan ninf nsz arcp double @llvm.experimental.constrained.tan.f64(double %val, metadata !"round.dynamic", metadata !"fpexcept.strict")
|
|
store double %tan, ptr %idx, align 8
|
|
%next = add nuw nsw i64 %i, 1
|
|
%cond = icmp eq i64 %next, 255
|
|
br i1 %cond, label %exit, label %for.body
|
|
}
|
|
|
|
; Check constrained ops converted to call
|
|
define void @testAtan2(ptr %cast1, ptr %cast2) strictfp {
|
|
; CHECK-LABEL: testAtan2:
|
|
; CHECK: # %bb.0: # %root
|
|
; CHECK-NEXT: mflr 0
|
|
; CHECK-NEXT: .cfi_def_cfa_offset 64
|
|
; CHECK-NEXT: .cfi_offset lr, 16
|
|
; CHECK-NEXT: .cfi_offset r28, -32
|
|
; CHECK-NEXT: .cfi_offset r29, -24
|
|
; CHECK-NEXT: .cfi_offset r30, -16
|
|
; CHECK-NEXT: std 28, -32(1) # 8-byte Folded Spill
|
|
; CHECK-NEXT: std 29, -24(1) # 8-byte Folded Spill
|
|
; CHECK-NEXT: std 30, -16(1) # 8-byte Folded Spill
|
|
; CHECK-NEXT: stdu 1, -64(1)
|
|
; CHECK-NEXT: addi 30, 3, -8
|
|
; CHECK-NEXT: addi 29, 4, -8
|
|
; CHECK-NEXT: li 28, 255
|
|
; CHECK-NEXT: std 0, 80(1)
|
|
; CHECK-NEXT: .p2align 5
|
|
; CHECK-NEXT: .LBB3_1: # %for.body
|
|
; CHECK-NEXT: #
|
|
; CHECK-NEXT: lfdu 2, 8(29)
|
|
; CHECK-NEXT: lfdu 1, 8(30)
|
|
; CHECK-NEXT: bl atan2
|
|
; CHECK-NEXT: nop
|
|
; CHECK-NEXT: addi 28, 28, -1
|
|
; CHECK-NEXT: stfd 1, 0(30)
|
|
; CHECK-NEXT: cmpldi 28, 0
|
|
; CHECK-NEXT: bc 12, 1, .LBB3_1
|
|
; CHECK-NEXT: # %bb.2: # %exit
|
|
; CHECK-NEXT: addi 1, 1, 64
|
|
; CHECK-NEXT: ld 0, 16(1)
|
|
; CHECK-NEXT: ld 30, -16(1) # 8-byte Folded Reload
|
|
; CHECK-NEXT: ld 29, -24(1) # 8-byte Folded Reload
|
|
; CHECK-NEXT: ld 28, -32(1) # 8-byte Folded Reload
|
|
; CHECK-NEXT: mtlr 0
|
|
; CHECK-NEXT: blr
|
|
root:
|
|
br label %for.body
|
|
|
|
exit:
|
|
ret void
|
|
|
|
for.body:
|
|
%i = phi i64 [ 0, %root ], [ %next, %for.body ]
|
|
%idx1 = getelementptr inbounds double, ptr %cast1, i64 %i
|
|
%idx2 = getelementptr inbounds double, ptr %cast2, i64 %i
|
|
%val1 = load double, ptr %idx1
|
|
%val2 = load double, ptr %idx2
|
|
%tan = tail call nnan ninf nsz arcp double @llvm.experimental.constrained.atan2.f64(double %val1, double %val2, metadata !"round.dynamic", metadata !"fpexcept.strict")
|
|
store double %tan, ptr %idx1, align 8
|
|
%next = add nuw nsw i64 %i, 1
|
|
%cond = icmp eq i64 %next, 255
|
|
br i1 %cond, label %exit, label %for.body
|
|
}
|
|
|
|
declare double @llvm.experimental.constrained.cos.f64(double, metadata, metadata)
|
|
declare double @llvm.experimental.constrained.tan.f64(double, metadata, metadata)
|
|
declare double @llvm.experimental.constrained.atan2.f64(double, double, metadata, metadata)
|
|
declare double @llvm.experimental.constrained.sqrt.f64(double, metadata, metadata)
|