mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-26 04:32:44 +01:00
fb51a50eef
More or less all the instructions defined in the v8.2a full-fp16 extension are defined as UNPREDICTABLE if you put them in an IT block (Thumb) or use with any condition other than AL (ARM). LLVM didn't know that, and was happy to conditionalise them. In order to force these instructions to count as not predicable, I had to make a small Tablegen change. The code generation back end mostly decides if an instruction was predicable by looking for something it can identify as a predicate operand; there's an isPredicable bit flag that overrides that check in the positive direction, but nothing that overrides it in the negative direction. (I considered the alternative approach of actually removing the predicate operand from those instructions, but thought that it would be more painful overall for instructions differing only in data type to have different shapes of operand list. This way, the only code that has to notice the difference is the if-converter.) So I've added an isUnpredicable bit alongside isPredicable, and set that bit on the right subset of FP16 instructions, and also on the VSEL, VMAXNM/VMINNM and VRINT[ANPM] families which should be unpredicable for all data types. I've included a couple of representative regression tests, both of which previously caused an fp16 instruction to be conditionalised in ARM state and (with -arm-no-restrict-it) to be put in an IT block in Thumb. Reviewers: SjoerdMeijer, t.p.northover, efriedma Reviewed By: efriedma Subscribers: jdoerfert, javed.absar, kristof.beyls, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D57823 llvm-svn: 354768
101 lines
2.6 KiB
LLVM
101 lines
2.6 KiB
LLVM
; RUN: llc -O3 -mtriple=armv8a-none-eabi -mattr=+fullfp16 -o - %s | FileCheck %s
|
|
; RUN: llc -O3 -mtriple=thumbv8a-none-eabi -mattr=+fullfp16 -arm-no-restrict-it -o - %s | FileCheck %s
|
|
|
|
; Require the vmul.f16 not to be predicated, because it's illegal to
|
|
; do so with fp16 instructions
|
|
define half @conditional_fmul_f16(half* %p) {
|
|
; CHECK-LABEL: conditional_fmul_f16:
|
|
; CHECK: vmul.f16
|
|
entry:
|
|
%p1 = getelementptr half, half* %p, i32 1
|
|
%a = load half, half* %p, align 2
|
|
%threshold = load half, half* %p1, align 2
|
|
%flag = fcmp ogt half %a, %threshold
|
|
br i1 %flag, label %mul, label %out
|
|
|
|
mul:
|
|
%p2 = getelementptr half, half* %p, i32 2
|
|
%mult = load half, half* %p2, align 2
|
|
%b = fmul half %a, %mult
|
|
br label %out
|
|
|
|
out:
|
|
%sel = phi half [ %a, %entry ], [ %b, %mul ]
|
|
ret half %sel
|
|
}
|
|
|
|
; Expect that the corresponding vmul.f32 _will_ be predicated (to make
|
|
; sure the previous test is really testing something)
|
|
define float @conditional_fmul_f32(float* %p) {
|
|
; CHECK-LABEL: conditional_fmul_f32:
|
|
; CHECK: vmulgt.f32
|
|
entry:
|
|
%p1 = getelementptr float, float* %p, i32 1
|
|
%a = load float, float* %p, align 2
|
|
%threshold = load float, float* %p1, align 2
|
|
%flag = fcmp ogt float %a, %threshold
|
|
br i1 %flag, label %mul, label %out
|
|
|
|
mul:
|
|
%p2 = getelementptr float, float* %p, i32 2
|
|
%mult = load float, float* %p2, align 2
|
|
%b = fmul float %a, %mult
|
|
br label %out
|
|
|
|
out:
|
|
%sel = phi float [ %a, %entry ], [ %b, %mul ]
|
|
ret float %sel
|
|
}
|
|
|
|
; Require the two comparisons to be done with unpredicated vcmp.f16
|
|
; instructions (again, it is illegal to predicate them)
|
|
define void @chained_comparisons_f16(half* %p) {
|
|
; CHECK-LABEL: chained_comparisons_f16:
|
|
; CHECK: vcmp.f16
|
|
; CHECK: vcmp.f16
|
|
entry:
|
|
%p1 = getelementptr half, half* %p, i32 1
|
|
|
|
%a = load half, half* %p, align 2
|
|
%b = load half, half* %p1, align 2
|
|
|
|
%aflag = fcmp oeq half %a, 0xH0000
|
|
%bflag = fcmp oeq half %b, 0xH0000
|
|
%flag = or i1 %aflag, %bflag
|
|
br i1 %flag, label %call, label %out
|
|
|
|
call:
|
|
call void @external_function()
|
|
br label %out
|
|
|
|
out:
|
|
ret void
|
|
}
|
|
|
|
; Again, do the corresponding test with 32-bit floats and check that
|
|
; the second comparison _is_ predicated on the result of the first.
|
|
define void @chained_comparisons_f32(float* %p) {
|
|
; CHECK-LABEL: chained_comparisons_f32:
|
|
; CHECK: vcmp.f32
|
|
; CHECK: vcmpne.f32
|
|
entry:
|
|
%p1 = getelementptr float, float* %p, i32 1
|
|
|
|
%a = load float, float* %p, align 2
|
|
%b = load float, float* %p1, align 2
|
|
|
|
%aflag = fcmp oeq float %a, 0x00000000
|
|
%bflag = fcmp oeq float %b, 0x00000000
|
|
%flag = or i1 %aflag, %bflag
|
|
br i1 %flag, label %call, label %out
|
|
|
|
call:
|
|
call void @external_function()
|
|
br label %out
|
|
|
|
out:
|
|
ret void
|
|
}
|
|
|
|
declare void @external_function()
|