mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2025-01-31 20:51:52 +01:00
0663a19f9d
The assert that caused this to be reverted should be fixed now. Original commit message: This patch changes our defualt legalization behavior for 16, 32, and 64 bit vectors with i8/i16/i32/i64 scalar types from promotion to widening. For example, v8i8 will now be widened to v16i8 instead of promoted to v8i16. This keeps the elements widths the same and pads with undef elements. We believe this is a better legalization strategy. But it carries some issues due to the fragmented vector ISA. For example, i8 shifts and multiplies get widened and then later have to be promoted/split into vXi16 vectors. This has the potential to cause regressions so we wanted to get it in early in the 10.0 cycle so we have plenty of time to address them. Next steps will be to merge tests that explicitly test the command line option. And then we can remove the option and its associated code. llvm-svn: 368183
91 lines
2.8 KiB
LLVM
91 lines
2.8 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -mtriple=i686-unknown-unknown -mcpu=generic -mattr=+sse4.2 < %s | FileCheck %s
|
|
; RUN: llc -mtriple=i686-unknown-unknown -mcpu=atom < %s | FileCheck -check-prefix=ATOM %s
|
|
|
|
; Scheduler causes produce a different instruction order
|
|
|
|
; bitcast a v4i16 to v2i32
|
|
|
|
define void @convert(<2 x i32>* %dst, <4 x i16>* %src) nounwind {
|
|
; CHECK-LABEL: convert:
|
|
; CHECK: # %bb.0: # %entry
|
|
; CHECK-NEXT: pushl %eax
|
|
; CHECK-NEXT: movl $0, (%esp)
|
|
; CHECK-NEXT: pcmpeqd %xmm0, %xmm0
|
|
; CHECK-NEXT: cmpl $3, (%esp)
|
|
; CHECK-NEXT: jg .LBB0_3
|
|
; CHECK-NEXT: .p2align 4, 0x90
|
|
; CHECK-NEXT: .LBB0_2: # %forbody
|
|
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
|
|
; CHECK-NEXT: movl (%esp), %eax
|
|
; CHECK-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
; CHECK-NEXT: movl {{[0-9]+}}(%esp), %edx
|
|
; CHECK-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
|
|
; CHECK-NEXT: psubw %xmm0, %xmm1
|
|
; CHECK-NEXT: movq %xmm1, (%ecx,%eax,8)
|
|
; CHECK-NEXT: incl (%esp)
|
|
; CHECK-NEXT: cmpl $3, (%esp)
|
|
; CHECK-NEXT: jle .LBB0_2
|
|
; CHECK-NEXT: .LBB0_3: # %afterfor
|
|
; CHECK-NEXT: popl %eax
|
|
; CHECK-NEXT: retl
|
|
;
|
|
; ATOM-LABEL: convert:
|
|
; ATOM: # %bb.0: # %entry
|
|
; ATOM-NEXT: pushl %eax
|
|
; ATOM-NEXT: pcmpeqd %xmm0, %xmm0
|
|
; ATOM-NEXT: movl $0, (%esp)
|
|
; ATOM-NEXT: cmpl $3, (%esp)
|
|
; ATOM-NEXT: jg .LBB0_3
|
|
; ATOM-NEXT: .p2align 4, 0x90
|
|
; ATOM-NEXT: .LBB0_2: # %forbody
|
|
; ATOM-NEXT: # =>This Inner Loop Header: Depth=1
|
|
; ATOM-NEXT: movl (%esp), %eax
|
|
; ATOM-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
; ATOM-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
|
|
; ATOM-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
; ATOM-NEXT: psubw %xmm0, %xmm1
|
|
; ATOM-NEXT: movq %xmm1, (%ecx,%eax,8)
|
|
; ATOM-NEXT: incl (%esp)
|
|
; ATOM-NEXT: cmpl $3, (%esp)
|
|
; ATOM-NEXT: jle .LBB0_2
|
|
; ATOM-NEXT: .LBB0_3: # %afterfor
|
|
; ATOM-NEXT: popl %eax
|
|
; ATOM-NEXT: retl
|
|
entry:
|
|
%dst.addr = alloca <2 x i32>*
|
|
%src.addr = alloca <4 x i16>*
|
|
%i = alloca i32, align 4
|
|
store <2 x i32>* %dst, <2 x i32>** %dst.addr
|
|
store <4 x i16>* %src, <4 x i16>** %src.addr
|
|
store i32 0, i32* %i
|
|
br label %forcond
|
|
|
|
forcond:
|
|
%tmp = load i32, i32* %i
|
|
%cmp = icmp slt i32 %tmp, 4
|
|
br i1 %cmp, label %forbody, label %afterfor
|
|
|
|
forbody:
|
|
%tmp1 = load i32, i32* %i
|
|
%tmp2 = load <2 x i32>*, <2 x i32>** %dst.addr
|
|
%arrayidx = getelementptr <2 x i32>, <2 x i32>* %tmp2, i32 %tmp1
|
|
%tmp3 = load i32, i32* %i
|
|
%tmp4 = load <4 x i16>*, <4 x i16>** %src.addr
|
|
%arrayidx5 = getelementptr <4 x i16>, <4 x i16>* %tmp4, i32 %tmp3
|
|
%tmp6 = load <4 x i16>, <4 x i16>* %arrayidx5
|
|
%add = add <4 x i16> %tmp6, < i16 1, i16 1, i16 1, i16 1 >
|
|
%conv = bitcast <4 x i16> %add to <2 x i32>
|
|
store <2 x i32> %conv, <2 x i32>* %arrayidx
|
|
br label %forinc
|
|
|
|
forinc:
|
|
%tmp7 = load i32, i32* %i
|
|
%inc = add i32 %tmp7, 1
|
|
store i32 %inc, i32* %i
|
|
br label %forcond
|
|
|
|
afterfor:
|
|
ret void
|
|
}
|