mirror of
https://github.com/RPCSX/llvm.git
synced 2024-12-11 05:35:11 +00:00
cf0fa9b9dd
The z13 vector facility includes some instructions that operate only on the high f64 in a v2f64, effectively extending the FP register set from 16 to 32 registers. It's still better to use the old instructions if the operands happen to fit though, since the older instructions have a shorter encoding. Based on a patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@236524 91177308-0d34-0410-b5e6-96231b3b80d8
122 lines
3.6 KiB
LLVM
122 lines
3.6 KiB
LLVM
; Test 64-bit floating-point subtraction.
|
|
;
|
|
; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 \
|
|
; RUN: | FileCheck -check-prefix=CHECK -check-prefix=CHECK-SCALAR %s
|
|
; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z13 | FileCheck %s
|
|
|
|
declare double @foo()
|
|
|
|
; Check register subtraction.
|
|
define double @f1(double %f1, double %f2) {
|
|
; CHECK-LABEL: f1:
|
|
; CHECK: sdbr %f0, %f2
|
|
; CHECK: br %r14
|
|
%res = fsub double %f1, %f2
|
|
ret double %res
|
|
}
|
|
|
|
; Check the low end of the SDB range.
|
|
define double @f2(double %f1, double *%ptr) {
|
|
; CHECK-LABEL: f2:
|
|
; CHECK: sdb %f0, 0(%r2)
|
|
; CHECK: br %r14
|
|
%f2 = load double , double *%ptr
|
|
%res = fsub double %f1, %f2
|
|
ret double %res
|
|
}
|
|
|
|
; Check the high end of the aligned SDB range.
|
|
define double @f3(double %f1, double *%base) {
|
|
; CHECK-LABEL: f3:
|
|
; CHECK: sdb %f0, 4088(%r2)
|
|
; CHECK: br %r14
|
|
%ptr = getelementptr double, double *%base, i64 511
|
|
%f2 = load double , double *%ptr
|
|
%res = fsub double %f1, %f2
|
|
ret double %res
|
|
}
|
|
|
|
; Check the next doubleword up, which needs separate address logic.
|
|
; Other sequences besides this one would be OK.
|
|
define double @f4(double %f1, double *%base) {
|
|
; CHECK-LABEL: f4:
|
|
; CHECK: aghi %r2, 4096
|
|
; CHECK: sdb %f0, 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr = getelementptr double, double *%base, i64 512
|
|
%f2 = load double , double *%ptr
|
|
%res = fsub double %f1, %f2
|
|
ret double %res
|
|
}
|
|
|
|
; Check negative displacements, which also need separate address logic.
|
|
define double @f5(double %f1, double *%base) {
|
|
; CHECK-LABEL: f5:
|
|
; CHECK: aghi %r2, -8
|
|
; CHECK: sdb %f0, 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr = getelementptr double, double *%base, i64 -1
|
|
%f2 = load double , double *%ptr
|
|
%res = fsub double %f1, %f2
|
|
ret double %res
|
|
}
|
|
|
|
; Check that SDB allows indices.
|
|
define double @f6(double %f1, double *%base, i64 %index) {
|
|
; CHECK-LABEL: f6:
|
|
; CHECK: sllg %r1, %r3, 3
|
|
; CHECK: sdb %f0, 800(%r1,%r2)
|
|
; CHECK: br %r14
|
|
%ptr1 = getelementptr double, double *%base, i64 %index
|
|
%ptr2 = getelementptr double, double *%ptr1, i64 100
|
|
%f2 = load double , double *%ptr2
|
|
%res = fsub double %f1, %f2
|
|
ret double %res
|
|
}
|
|
|
|
; Check that subtractions of spilled values can use SDB rather than SDBR.
|
|
define double @f7(double *%ptr0) {
|
|
; CHECK-LABEL: f7:
|
|
; CHECK: brasl %r14, foo@PLT
|
|
; CHECK-SCALAR: sdb %f0, 16{{[04]}}(%r15)
|
|
; CHECK: br %r14
|
|
%ptr1 = getelementptr double, double *%ptr0, i64 2
|
|
%ptr2 = getelementptr double, double *%ptr0, i64 4
|
|
%ptr3 = getelementptr double, double *%ptr0, i64 6
|
|
%ptr4 = getelementptr double, double *%ptr0, i64 8
|
|
%ptr5 = getelementptr double, double *%ptr0, i64 10
|
|
%ptr6 = getelementptr double, double *%ptr0, i64 12
|
|
%ptr7 = getelementptr double, double *%ptr0, i64 14
|
|
%ptr8 = getelementptr double, double *%ptr0, i64 16
|
|
%ptr9 = getelementptr double, double *%ptr0, i64 18
|
|
%ptr10 = getelementptr double, double *%ptr0, i64 20
|
|
|
|
%val0 = load double , double *%ptr0
|
|
%val1 = load double , double *%ptr1
|
|
%val2 = load double , double *%ptr2
|
|
%val3 = load double , double *%ptr3
|
|
%val4 = load double , double *%ptr4
|
|
%val5 = load double , double *%ptr5
|
|
%val6 = load double , double *%ptr6
|
|
%val7 = load double , double *%ptr7
|
|
%val8 = load double , double *%ptr8
|
|
%val9 = load double , double *%ptr9
|
|
%val10 = load double , double *%ptr10
|
|
|
|
%ret = call double @foo()
|
|
|
|
%sub0 = fsub double %ret, %val0
|
|
%sub1 = fsub double %sub0, %val1
|
|
%sub2 = fsub double %sub1, %val2
|
|
%sub3 = fsub double %sub2, %val3
|
|
%sub4 = fsub double %sub3, %val4
|
|
%sub5 = fsub double %sub4, %val5
|
|
%sub6 = fsub double %sub5, %val6
|
|
%sub7 = fsub double %sub6, %val7
|
|
%sub8 = fsub double %sub7, %val8
|
|
%sub9 = fsub double %sub8, %val9
|
|
%sub10 = fsub double %sub9, %val10
|
|
|
|
ret double %sub10
|
|
}
|