mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2024-12-15 04:30:12 +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
111 lines
2.8 KiB
LLVM
111 lines
2.8 KiB
LLVM
; Test 32-bit floating-point loads for z13.
|
|
;
|
|
; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z13 | FileCheck %s
|
|
|
|
; Test that we use LDE instead of LE - low end of the LE range.
|
|
define float @f1(float *%src) {
|
|
; CHECK-LABEL: f1:
|
|
; CHECK: lde %f0, 0(%r2)
|
|
; CHECK: br %r14
|
|
%val = load float, float *%src
|
|
ret float %val
|
|
}
|
|
|
|
; Test that we use LDE instead of LE - high end of the LE range.
|
|
define float @f2(float *%src) {
|
|
; CHECK-LABEL: f2:
|
|
; CHECK: lde %f0, 4092(%r2)
|
|
; CHECK: br %r14
|
|
%ptr = getelementptr float, float *%src, i64 1023
|
|
%val = load float, float *%ptr
|
|
ret float %val
|
|
}
|
|
|
|
; Check the next word up, which should use LEY instead of LDE.
|
|
define float @f3(float *%src) {
|
|
; CHECK-LABEL: f3:
|
|
; CHECK: ley %f0, 4096(%r2)
|
|
; CHECK: br %r14
|
|
%ptr = getelementptr float, float *%src, i64 1024
|
|
%val = load float, float *%ptr
|
|
ret float %val
|
|
}
|
|
|
|
; Check the high end of the aligned LEY range.
|
|
define float @f4(float *%src) {
|
|
; CHECK-LABEL: f4:
|
|
; CHECK: ley %f0, 524284(%r2)
|
|
; CHECK: br %r14
|
|
%ptr = getelementptr float, float *%src, i64 131071
|
|
%val = load float, float *%ptr
|
|
ret float %val
|
|
}
|
|
|
|
; Check the next word up, which needs separate address logic.
|
|
; Other sequences besides this one would be OK.
|
|
define float @f5(float *%src) {
|
|
; CHECK-LABEL: f5:
|
|
; CHECK: agfi %r2, 524288
|
|
; CHECK: lde %f0, 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr = getelementptr float, float *%src, i64 131072
|
|
%val = load float, float *%ptr
|
|
ret float %val
|
|
}
|
|
|
|
; Check the high end of the negative aligned LEY range.
|
|
define float @f6(float *%src) {
|
|
; CHECK-LABEL: f6:
|
|
; CHECK: ley %f0, -4(%r2)
|
|
; CHECK: br %r14
|
|
%ptr = getelementptr float, float *%src, i64 -1
|
|
%val = load float, float *%ptr
|
|
ret float %val
|
|
}
|
|
|
|
; Check the low end of the LEY range.
|
|
define float @f7(float *%src) {
|
|
; CHECK-LABEL: f7:
|
|
; CHECK: ley %f0, -524288(%r2)
|
|
; CHECK: br %r14
|
|
%ptr = getelementptr float, float *%src, i64 -131072
|
|
%val = load float, float *%ptr
|
|
ret float %val
|
|
}
|
|
|
|
; Check the next word down, which needs separate address logic.
|
|
; Other sequences besides this one would be OK.
|
|
define float @f8(float *%src) {
|
|
; CHECK-LABEL: f8:
|
|
; CHECK: agfi %r2, -524292
|
|
; CHECK: lde %f0, 0(%r2)
|
|
; CHECK: br %r14
|
|
%ptr = getelementptr float, float *%src, i64 -131073
|
|
%val = load float, float *%ptr
|
|
ret float %val
|
|
}
|
|
|
|
; Check that LDE allows an index.
|
|
define float @f9(i64 %src, i64 %index) {
|
|
; CHECK-LABEL: f9:
|
|
; CHECK: lde %f0, 4092({{%r3,%r2|%r2,%r3}})
|
|
; CHECK: br %r14
|
|
%add1 = add i64 %src, %index
|
|
%add2 = add i64 %add1, 4092
|
|
%ptr = inttoptr i64 %add2 to float *
|
|
%val = load float, float *%ptr
|
|
ret float %val
|
|
}
|
|
|
|
; Check that LEY allows an index.
|
|
define float @f10(i64 %src, i64 %index) {
|
|
; CHECK-LABEL: f10:
|
|
; CHECK: ley %f0, 4096({{%r3,%r2|%r2,%r3}})
|
|
; CHECK: br %r14
|
|
%add1 = add i64 %src, %index
|
|
%add2 = add i64 %add1, 4096
|
|
%ptr = inttoptr i64 %add2 to float *
|
|
%val = load float, float *%ptr
|
|
ret float %val
|
|
}
|