mirror of
https://github.com/RPCSX/llvm.git
synced 2024-12-02 16:56:50 +00:00
a04e9e4a0a
There is not an official documented ABI for frame pointers in Thumb2, but we should try to emit something which is useful. We use r7 as the frame pointer for Thumb code, which currently means that if a function needs to save a high register (r8-r11), it will get pushed to the stack between the frame pointer (r7) and link register (r14). This means that while a stack unwinder can follow the chain of frame pointers up the stack, it cannot know the offset to lr, so does not know which functions correspond to the stack frames. To fix this, we need to push the callee-saved registers in two batches, with the first push saving the low registers, fp and lr, and the second push saving the high registers. This is already implemented, but previously only used for iOS. This patch turns it on for all Thumb2 targets when frame pointers are required by the ABI, and the frame pointer is r7 (Windows uses r11, so this isn't a problem there). If frame pointer elimination is enabled we still emit a single push/pop even if we need a frame pointer for other reasons, to avoid increasing code size. We must also ensure that lr is pushed to the stack when using a frame pointer, so that we end up with a complete frame record. Situations that could cause this were rare, because we already push lr in most situations so that we can return using the pop instruction. Differential Revision: https://reviews.llvm.org/D23516 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@279506 91177308-0d34-0410-b5e6-96231b3b80d8
113 lines
5.2 KiB
LLVM
113 lines
5.2 KiB
LLVM
; RUN: llc < %s -mtriple=arm-apple-ios -mattr=+v6,+vfp2 | FileCheck %s
|
|
; RUN: llc < %s -mtriple=arm-apple-ios -mattr=+v6,+vfp2 | FileCheck --check-prefix=DOMAIN %s
|
|
|
|
; The execution domain checking code would translate vmovs to vorr whether or not
|
|
; we had NEON instructions. Verify we don't if we're not compiled with NEON.
|
|
; DOMAIN-NOT: vorr
|
|
@quant_coef = external global [6 x [4 x [4 x i32]]] ; <[6 x [4 x [4 x i32]]]*> [#uses=1]
|
|
@dequant_coef = external global [6 x [4 x [4 x i32]]] ; <[6 x [4 x [4 x i32]]]*> [#uses=1]
|
|
@A = external global [4 x [4 x i32]] ; <[4 x [4 x i32]]*> [#uses=1]
|
|
|
|
; CHECK-LABEL: dct_luma_sp:
|
|
define fastcc i32 @dct_luma_sp(i32 %block_x, i32 %block_y, i32* %coeff_cost) "no-frame-pointer-elim"="true" {
|
|
entry:
|
|
; Make sure to use base-updating stores for saving callee-saved registers.
|
|
; CHECK: push
|
|
; CHECK-NOT: sub sp
|
|
; CHECK: push
|
|
%predicted_block = alloca [4 x [4 x i32]], align 4 ; <[4 x [4 x i32]]*> [#uses=1]
|
|
br label %cond_next489
|
|
|
|
cond_next489: ; preds = %cond_false, %bb471
|
|
%j.7.in = load i8, i8* null ; <i8> [#uses=1]
|
|
%i.8.in = load i8, i8* null ; <i8> [#uses=1]
|
|
%i.8 = zext i8 %i.8.in to i32 ; <i32> [#uses=4]
|
|
%j.7 = zext i8 %j.7.in to i32 ; <i32> [#uses=4]
|
|
%tmp495 = getelementptr [4 x [4 x i32]], [4 x [4 x i32]]* %predicted_block, i32 0, i32 %i.8, i32 %j.7 ; <i32*> [#uses=2]
|
|
%tmp496 = load i32, i32* %tmp495 ; <i32> [#uses=2]
|
|
%tmp502 = load i32, i32* null ; <i32> [#uses=1]
|
|
%tmp542 = getelementptr [6 x [4 x [4 x i32]]], [6 x [4 x [4 x i32]]]* @quant_coef, i32 0, i32 0, i32 %i.8, i32 %j.7 ; <i32*> [#uses=1]
|
|
%tmp543 = load i32, i32* %tmp542 ; <i32> [#uses=1]
|
|
%tmp548 = ashr i32 0, 0 ; <i32> [#uses=3]
|
|
%tmp561 = sub i32 0, %tmp496 ; <i32> [#uses=3]
|
|
%abscond563 = icmp sgt i32 %tmp561, -1 ; <i1> [#uses=1]
|
|
%abs564 = select i1 %abscond563, i32 %tmp561, i32 0 ; <i32> [#uses=1]
|
|
%tmp572 = mul i32 %abs564, %tmp543 ; <i32> [#uses=1]
|
|
%tmp574 = add i32 %tmp572, 0 ; <i32> [#uses=1]
|
|
%tmp576 = ashr i32 %tmp574, 0 ; <i32> [#uses=7]
|
|
%tmp579 = icmp eq i32 %tmp548, %tmp576 ; <i1> [#uses=1]
|
|
br i1 %tmp579, label %bb712, label %cond_next589
|
|
|
|
cond_next589: ; preds = %cond_next489
|
|
%tmp605 = getelementptr [6 x [4 x [4 x i32]]], [6 x [4 x [4 x i32]]]* @dequant_coef, i32 0, i32 0, i32 %i.8, i32 %j.7 ; <i32*> [#uses=1]
|
|
%tmp606 = load i32, i32* %tmp605 ; <i32> [#uses=1]
|
|
%tmp612 = load i32, i32* null ; <i32> [#uses=1]
|
|
%tmp629 = load i32, i32* null ; <i32> [#uses=1]
|
|
%tmp629a = sitofp i32 %tmp629 to double ; <double> [#uses=1]
|
|
%tmp631 = fmul double %tmp629a, 0.000000e+00 ; <double> [#uses=1]
|
|
%tmp632 = fadd double 0.000000e+00, %tmp631 ; <double> [#uses=1]
|
|
%tmp642 = call fastcc i32 @sign( i32 %tmp576, i32 %tmp561 ) ; <i32> [#uses=1]
|
|
%tmp650 = mul i32 %tmp606, %tmp642 ; <i32> [#uses=1]
|
|
%tmp656 = mul i32 %tmp650, %tmp612 ; <i32> [#uses=1]
|
|
%tmp658 = shl i32 %tmp656, 0 ; <i32> [#uses=1]
|
|
%tmp659 = ashr i32 %tmp658, 6 ; <i32> [#uses=1]
|
|
%tmp660 = sub i32 0, %tmp659 ; <i32> [#uses=1]
|
|
%tmp666 = sub i32 %tmp660, %tmp496 ; <i32> [#uses=1]
|
|
%tmp667 = sitofp i32 %tmp666 to double ; <double> [#uses=2]
|
|
call void @levrun_linfo_inter( i32 %tmp576, i32 0, i32* null, i32* null )
|
|
%tmp671 = fmul double %tmp667, %tmp667 ; <double> [#uses=1]
|
|
%tmp675 = fadd double %tmp671, 0.000000e+00 ; <double> [#uses=1]
|
|
%tmp678 = fcmp oeq double %tmp632, %tmp675 ; <i1> [#uses=1]
|
|
br i1 %tmp678, label %cond_true679, label %cond_false693
|
|
|
|
cond_true679: ; preds = %cond_next589
|
|
%abscond681 = icmp sgt i32 %tmp548, -1 ; <i1> [#uses=1]
|
|
%abs682 = select i1 %abscond681, i32 %tmp548, i32 0 ; <i32> [#uses=1]
|
|
%abscond684 = icmp sgt i32 %tmp576, -1 ; <i1> [#uses=1]
|
|
%abs685 = select i1 %abscond684, i32 %tmp576, i32 0 ; <i32> [#uses=1]
|
|
%tmp686 = icmp slt i32 %abs682, %abs685 ; <i1> [#uses=1]
|
|
br i1 %tmp686, label %cond_next702, label %cond_false689
|
|
|
|
cond_false689: ; preds = %cond_true679
|
|
%tmp739 = icmp eq i32 %tmp576, 0 ; <i1> [#uses=1]
|
|
br i1 %tmp579, label %bb737, label %cond_false708
|
|
|
|
cond_false693: ; preds = %cond_next589
|
|
ret i32 0
|
|
|
|
cond_next702: ; preds = %cond_true679
|
|
ret i32 0
|
|
|
|
cond_false708: ; preds = %cond_false689
|
|
ret i32 0
|
|
|
|
bb712: ; preds = %cond_next489
|
|
ret i32 0
|
|
|
|
bb737: ; preds = %cond_false689
|
|
br i1 %tmp739, label %cond_next791, label %cond_true740
|
|
|
|
cond_true740: ; preds = %bb737
|
|
%tmp761 = call fastcc i32 @sign( i32 %tmp576, i32 0 ) ; <i32> [#uses=1]
|
|
%tmp780 = load i32, i32* null ; <i32> [#uses=1]
|
|
%tmp785 = getelementptr [4 x [4 x i32]], [4 x [4 x i32]]* @A, i32 0, i32 %i.8, i32 %j.7 ; <i32*> [#uses=1]
|
|
%tmp786 = load i32, i32* %tmp785 ; <i32> [#uses=1]
|
|
%tmp781 = mul i32 %tmp780, %tmp761 ; <i32> [#uses=1]
|
|
%tmp787 = mul i32 %tmp781, %tmp786 ; <i32> [#uses=1]
|
|
%tmp789 = shl i32 %tmp787, 0 ; <i32> [#uses=1]
|
|
%tmp790 = ashr i32 %tmp789, 6 ; <i32> [#uses=1]
|
|
br label %cond_next791
|
|
|
|
cond_next791: ; preds = %cond_true740, %bb737
|
|
%ilev.1 = phi i32 [ %tmp790, %cond_true740 ], [ 0, %bb737 ] ; <i32> [#uses=1]
|
|
%tmp796 = load i32, i32* %tmp495 ; <i32> [#uses=1]
|
|
%tmp798 = add i32 %tmp796, %ilev.1 ; <i32> [#uses=1]
|
|
%tmp812 = mul i32 0, %tmp502 ; <i32> [#uses=0]
|
|
%tmp818 = call fastcc i32 @sign( i32 0, i32 %tmp798 ) ; <i32> [#uses=0]
|
|
unreachable
|
|
}
|
|
|
|
declare i32 @sign(i32, i32)
|
|
|
|
declare void @levrun_linfo_inter(i32, i32, i32*, i32*)
|