maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/XCore/epilogue_prologue.ll

242 lines
6.9 KiB
LLVM
Raw Normal View History

[XCore] Fix handling of functions where only the LR is spilled. Previously we only checked if the LR required saving if the frame size was non zero. However because the caller reserves 1 word for the callee to use that doesn't count towards our frame size it is possible for the LR to need saving and for the frame size to be 0. We didn't hit when the LR needed saving because of a function calls because the 1 word of stack we must allocate for our callee means the frame size is always non zero in this case. However we can hit this case if the LR is clobbered in inline asm. llvm-svn: 181520
2013-05-10 00:43:42 +08:00
; RUN: llc < %s -march=xcore | FileCheck %s
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
; RUN: llc < %s -march=xcore -disable-fp-elim | FileCheck %s -check-prefix=CHECKFP
[XCore] Fix handling of functions where only the LR is spilled. Previously we only checked if the LR required saving if the frame size was non zero. However because the caller reserves 1 word for the callee to use that doesn't count towards our frame size it is possible for the LR to need saving and for the frame size to be 0. We didn't hit when the LR needed saving because of a function calls because the 1 word of stack we must allocate for our callee means the frame size is always non zero in this case. However we can hit this case if the LR is clobbered in inline asm. llvm-svn: 181520
2013-05-10 00:43:42 +08:00
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; When using SP for small frames, we don't need any scratch registers (SR).
; When using SP for large frames, we may need two scratch registers.
; When using FP, for large or small frames, we may need one scratch register.
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; FP + small frame: spill FP+SR = entsp 2
AsmPrinter: mark the beginning and the end of a function in verbose mode llvm-svn: 303690
2017-05-24 05:22:16 +08:00
; CHECKFP-LABEL: f1:
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
; CHECKFP: entsp 2
; CHECKFP-NEXT: stw r10, sp[1]
; CHECKFP-NEXT: ldaw r10, sp[0]
; CHECKFP: set sp, r10
; CHECKFP-NEXT: ldw r10, sp[1]
; CHECKFP-NEXT: retsp 2
;
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; !FP + small frame: no spills = no stack adjustment needed
AsmPrinter: mark the beginning and the end of a function in verbose mode llvm-svn: 303690
2017-05-24 05:22:16 +08:00
; CHECK-LABEL: f1:
[XCore] Fix handling of functions where only the LR is spilled. Previously we only checked if the LR required saving if the frame size was non zero. However because the caller reserves 1 word for the callee to use that doesn't count towards our frame size it is possible for the LR to need saving and for the frame size to be 0. We didn't hit when the LR needed saving because of a function calls because the 1 word of stack we must allocate for our callee means the frame size is always non zero in this case. However we can hit this case if the LR is clobbered in inline asm. llvm-svn: 181520
2013-05-10 00:43:42 +08:00
; CHECK: stw lr, sp[0]
; CHECK: ldw lr, sp[0]
; CHECK-NEXT: retsp 0
define void @f1() nounwind {
entry:
tail call void asm sideeffect "", "~{lr}"() nounwind
ret void
}
[XCore] Ensure implicit operands aren't lost on the return instruction. Patch by Robert Lytton. llvm-svn: 186500
2013-07-17 18:58:37 +08:00
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; FP + small frame: spill FP+SR+R0+LR = entsp 3 + extsp 1
AsmPrinter: mark the beginning and the end of a function in verbose mode llvm-svn: 303690
2017-05-24 05:22:16 +08:00
; CHECKFP-LABEL: f3:
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
; CHECKFP: entsp 3
; CHECKFP-NEXT: stw r10, sp[1]
; CHECKFP-NEXT: ldaw r10, sp[0]
; CHECKFP-NEXT: stw [[REG:r[4-9]+]], r10[2]
; CHECKFP-NEXT: mov [[REG]], r0
; CHECKFP-NEXT: extsp 1
; CHECKFP-NEXT: bl f2
; CHECKFP-NEXT: ldaw sp, sp[1]
; CHECKFP-NEXT: mov r0, [[REG]]
; CHECKFP-NEXT: ldw [[REG]], r10[2]
; CHECKFP-NEXT: set sp, r10
; CHECKFP-NEXT: ldw r10, sp[1]
; CHECKFP-NEXT: retsp 3
;
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; !FP + small frame: spill R0+LR = entsp 2
AsmPrinter: mark the beginning and the end of a function in verbose mode llvm-svn: 303690
2017-05-24 05:22:16 +08:00
; CHECK-LABEL: f3:
[XCore] Ensure implicit operands aren't lost on the return instruction. Patch by Robert Lytton. llvm-svn: 186500
2013-07-17 18:58:37 +08:00
; CHECK: entsp 2
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; CHECK-NEXT: stw [[REG:r[4-9]+]], sp[1]
; CHECK-NEXT: mov [[REG]], r0
; CHECK-NEXT: bl f2
; CHECK-NEXT: mov r0, [[REG]]
; CHECK-NEXT: ldw [[REG]], sp[1]
; CHECK-NEXT: retsp 2
[XCore] Ensure implicit operands aren't lost on the return instruction. Patch by Robert Lytton. llvm-svn: 186500
2013-07-17 18:58:37 +08:00
declare void @f2()
define i32 @f3(i32 %i) nounwind {
entry:
call void @f2()
ret i32 %i
}
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; FP + large frame: spill FP+SR = entsp 2 + 100000
AsmPrinter: mark the beginning and the end of a function in verbose mode llvm-svn: 303690
2017-05-24 05:22:16 +08:00
; CHECKFP-LABEL: f4:
XCore target: Optimise entsp / retsp selection llvm-svn: 198612
2014-01-06 22:20:47 +08:00
; CHECKFP: entsp 65535
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
; CHECKFP-NEXT: .cfi_def_cfa_offset 262140
XCore target: Optimise entsp / retsp selection llvm-svn: 198612
2014-01-06 22:20:47 +08:00
; CHECKFP-NEXT: .cfi_offset 15, 0
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
; CHECKFP-NEXT: extsp 34467
; CHECKFP-NEXT: .cfi_def_cfa_offset 400008
; CHECKFP-NEXT: stw r10, sp[1]
; CHECKFP-NEXT: .cfi_offset 10, -400004
; CHECKFP-NEXT: ldaw r10, sp[0]
; CHECKFP-NEXT: .cfi_def_cfa_register 10
; CHECKFP-NEXT: set sp, r10
; CHECKFP-NEXT: ldw r10, sp[1]
; CHECKFP-NEXT: ldaw sp, sp[65535]
XCore target: Optimise entsp / retsp selection llvm-svn: 198612
2014-01-06 22:20:47 +08:00
; CHECKFP-NEXT: retsp 34467
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
;
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; !FP + large frame: spill SR+SR = entsp 2 + 100000
AsmPrinter: mark the beginning and the end of a function in verbose mode llvm-svn: 303690
2017-05-24 05:22:16 +08:00
; CHECK-LABEL: f4:
XCore target: Optimise entsp / retsp selection llvm-svn: 198612
2014-01-06 22:20:47 +08:00
; CHECK: entsp 65535
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
; CHECK-NEXT: .cfi_def_cfa_offset 262140
XCore target: Optimise entsp / retsp selection llvm-svn: 198612
2014-01-06 22:20:47 +08:00
; CHECK-NEXT: .cfi_offset 15, 0
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; CHECK-NEXT: extsp 34467
; CHECK-NEXT: .cfi_def_cfa_offset 400008
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
; CHECK-NEXT: ldaw sp, sp[65535]
XCore target: Optimise entsp / retsp selection llvm-svn: 198612
2014-01-06 22:20:47 +08:00
; CHECK-NEXT: retsp 34467
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
define void @f4() {
entry:
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
%0 = alloca [100000 x i32]
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
ret void
}
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; FP + large frame: spill FP+SR+R4+LR = entsp 3 + 200000 + extsp 1
; CHECKFP: .section .cp.rodata.cst4,"aMc",@progbits,4
[MC] Use .p2align instead of .align For historic reasons, the behavior of .align differs between targets. Fortunately, there are alternatives, .p2align and .balign, which make the interpretation of the parameter explicit, and which behave consistently across targets. This patch teaches MC to use .p2align instead of .align, so that people reading code for multiple architectures don't have to remember which way each platform does its .align directive. Differential Revision: http://reviews.llvm.org/D16549 llvm-svn: 258750
2016-01-26 08:03:25 +08:00
; CHECKFP-NEXT: .p2align 2
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; CHECKFP-NEXT: .LCPI[[CNST0:[0-9_]+]]:
; CHECKFP-NEXT: .long 200002
; CHECKFP-NEXT: .LCPI[[CNST1:[0-9_]+]]:
; CHECKFP-NEXT: .long 200001
; CHECKFP-NEXT: .text
AsmPrinter: mark the beginning and the end of a function in verbose mode llvm-svn: 303690
2017-05-24 05:22:16 +08:00
; CHECKFP-LABEL: f6:
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
; CHECKFP: entsp 65535
; CHECKFP-NEXT: .cfi_def_cfa_offset 262140
; CHECKFP-NEXT: .cfi_offset 15, 0
; CHECKFP-NEXT: extsp 65535
; CHECKFP-NEXT: .cfi_def_cfa_offset 524280
; CHECKFP-NEXT: extsp 65535
; CHECKFP-NEXT: .cfi_def_cfa_offset 786420
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; CHECKFP-NEXT: extsp 3398
; CHECKFP-NEXT: .cfi_def_cfa_offset 800012
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
; CHECKFP-NEXT: stw r10, sp[1]
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; CHECKFP-NEXT: .cfi_offset 10, -800008
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
; CHECKFP-NEXT: ldaw r10, sp[0]
; CHECKFP-NEXT: .cfi_def_cfa_register 10
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; CHECKFP-NEXT: ldw r1, cp[.LCPI[[CNST0]]]
; CHECKFP-NEXT: stw [[REG:r[4-9]+]], r10[r1]
; CHECKFP-NEXT: .cfi_offset 4, -4
; CHECKFP-NEXT: mov [[REG]], r0
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
; CHECKFP-NEXT: extsp 1
; CHECKFP-NEXT: ldaw r0, r10[2]
; CHECKFP-NEXT: bl f5
; CHECKFP-NEXT: ldaw sp, sp[1]
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; CHECKFP-NEXT: ldw r1, cp[.LCPI3_1]
; CHECKFP-NEXT: ldaw r0, r10[r1]
; CHECKFP-NEXT: extsp 1
; CHECKFP-NEXT: bl f5
; CHECKFP-NEXT: ldaw sp, sp[1]
; CHECKFP-NEXT: mov r0, [[REG]]
; CHECKFP-NEXT: ldw r1, cp[.LCPI[[CNST0]]]
; CHECKFP-NEXT: ldw [[REG]], r10[r1]
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
; CHECKFP-NEXT: set sp, r10
; CHECKFP-NEXT: ldw r10, sp[1]
; CHECKFP-NEXT: ldaw sp, sp[65535]
; CHECKFP-NEXT: ldaw sp, sp[65535]
; CHECKFP-NEXT: ldaw sp, sp[65535]
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; CHECKFP-NEXT: retsp 3398
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
;
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; !FP + large frame: spill SR+SR+R4+LR = entsp 4 + 200000
; CHECK: .section .cp.rodata.cst4,"aMc",@progbits,4
[MC] Use .p2align instead of .align For historic reasons, the behavior of .align differs between targets. Fortunately, there are alternatives, .p2align and .balign, which make the interpretation of the parameter explicit, and which behave consistently across targets. This patch teaches MC to use .p2align instead of .align, so that people reading code for multiple architectures don't have to remember which way each platform does its .align directive. Differential Revision: http://reviews.llvm.org/D16549 llvm-svn: 258750
2016-01-26 08:03:25 +08:00
; CHECK-NEXT: .p2align 2
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; CHECK-NEXT: .LCPI[[CNST0:[0-9_]+]]:
; CHECK-NEXT: .long 200003
; CHECK-NEXT: .LCPI[[CNST1:[0-9_]+]]:
; CHECK-NEXT: .long 200002
; CHECK-NEXT: .text
AsmPrinter: mark the beginning and the end of a function in verbose mode llvm-svn: 303690
2017-05-24 05:22:16 +08:00
; CHECK-LABEL: f6:
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
; CHECK: entsp 65535
; CHECK-NEXT: .cfi_def_cfa_offset 262140
; CHECK-NEXT: .cfi_offset 15, 0
; CHECK-NEXT: extsp 65535
; CHECK-NEXT: .cfi_def_cfa_offset 524280
; CHECK-NEXT: extsp 65535
; CHECK-NEXT: .cfi_def_cfa_offset 786420
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; CHECK-NEXT: extsp 3399
; CHECK-NEXT: .cfi_def_cfa_offset 800016
; CHECK-NEXT: ldaw r1, sp[0]
; CHECK-NEXT: ldw r2, cp[.LCPI[[CNST0]]]
; CHECK-NEXT: stw [[REG:r[4-9]+]], r1[r2]
; CHECK-NEXT: .cfi_offset 4, -4
; CHECK-NEXT: mov [[REG]], r0
; CHECK-NEXT: ldaw r0, sp[3]
; CHECK-NEXT: bl f5
; CHECK-NEXT: ldaw r0, sp[0]
; CHECK-NEXT: ldw r1, cp[.LCPI[[CNST1]]]
; CHECK-NEXT: ldaw r0, r0[r1]
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
; CHECK-NEXT: bl f5
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; CHECK-NEXT: mov r0, [[REG]]
; CHECK-NEXT: ldaw [[REG]], sp[0]
; CHECK-NEXT: ldw r1, cp[.LCPI[[CNST0]]]
; CHECK-NEXT: ldw [[REG]], [[REG]][r1]
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
; CHECK-NEXT: ldaw sp, sp[65535]
; CHECK-NEXT: ldaw sp, sp[65535]
; CHECK-NEXT: ldaw sp, sp[65535]
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; CHECK-NEXT: retsp 3399
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
declare void @f5(i32*)
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
define i32 @f6(i32 %i) {
entry:
%0 = alloca [200000 x i32]
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-28 03:29:02 +08:00
%1 = getelementptr inbounds [200000 x i32], [200000 x i32]* %0, i32 0, i32 0
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
call void @f5(i32* %1)
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-28 03:29:02 +08:00
%2 = getelementptr inbounds [200000 x i32], [200000 x i32]* %0, i32 0, i32 199999
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
call void @f5(i32* %2)
ret i32 %i
}
[XCore] Don't create invalid MKMSK instructions inside loadImmediate(). Summary: Previously loadImmediate() would produce MKMSK instructions with invalid immediate values such as mkmsk r0, 9. Fix this by checking the mask size is valid. Reviewers: robertlytton Reviewed By: robertlytton CC: llvm-commits Differential Revision: http://reviews.llvm.org/D3289 llvm-svn: 206163
2014-04-14 20:30:35 +08:00
; FP + large frame: spill FP+SR+LR = entsp 2 + 256 + extsp 1
AsmPrinter: mark the beginning and the end of a function in verbose mode llvm-svn: 303690
2017-05-24 05:22:16 +08:00
; CHECKFP-LABEL: f8:
[XCore] Don't create invalid MKMSK instructions inside loadImmediate(). Summary: Previously loadImmediate() would produce MKMSK instructions with invalid immediate values such as mkmsk r0, 9. Fix this by checking the mask size is valid. Reviewers: robertlytton Reviewed By: robertlytton CC: llvm-commits Differential Revision: http://reviews.llvm.org/D3289 llvm-svn: 206163
2014-04-14 20:30:35 +08:00
; CHECKFP: entsp 258
; CHECKFP-NEXT: stw r10, sp[1]
; CHECKFP-NEXT: ldaw r10, sp[0]
; CHECKFP-NEXT: mkmsk [[REG:r[0-9]+]], 8
; CHECKFP-NEXT: ldaw r0, r10{{\[}}[[REG]]{{\]}}
; CHECKFP-NEXT: extsp 1
; CHECKFP-NEXT: bl f5
; CHECKFP-NEXT: ldaw sp, sp[1]
; CHECKFP-NEXT: set sp, r10
; CHECKFP-NEXT: ldw r10, sp[1]
; CHECKFP-NEXT: retsp 258
;
; !FP + large frame: spill SR+SR+LR = entsp 3 + 256
AsmPrinter: mark the beginning and the end of a function in verbose mode llvm-svn: 303690
2017-05-24 05:22:16 +08:00
; CHECK-LABEL: f8:
[XCore] Don't create invalid MKMSK instructions inside loadImmediate(). Summary: Previously loadImmediate() would produce MKMSK instructions with invalid immediate values such as mkmsk r0, 9. Fix this by checking the mask size is valid. Reviewers: robertlytton Reviewed By: robertlytton CC: llvm-commits Differential Revision: http://reviews.llvm.org/D3289 llvm-svn: 206163
2014-04-14 20:30:35 +08:00
; CHECK: entsp 257
; CHECK-NEXT: ldaw r0, sp[254]
; CHECK-NEXT: bl f5
; CHECK-NEXT: retsp 257
define void @f8() nounwind {
entry:
%0 = alloca [256 x i32]
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-28 03:29:02 +08:00
%1 = getelementptr inbounds [256 x i32], [256 x i32]* %0, i32 0, i32 253
[XCore] Don't create invalid MKMSK instructions inside loadImmediate(). Summary: Previously loadImmediate() would produce MKMSK instructions with invalid immediate values such as mkmsk r0, 9. Fix this by checking the mask size is valid. Reviewers: robertlytton Reviewed By: robertlytton CC: llvm-commits Differential Revision: http://reviews.llvm.org/D3289 llvm-svn: 206163
2014-04-14 20:30:35 +08:00
call void @f5(i32* %1)
ret void
}
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; FP + large frame: spill FP+SR+LR = entsp 2 + 32768 + extsp 1
AsmPrinter: mark the beginning and the end of a function in verbose mode llvm-svn: 303690
2017-05-24 05:22:16 +08:00
; CHECKFP-LABEL: f9:
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; CHECKFP: entsp 32770
; CHECKFP-NEXT: stw r10, sp[1]
; CHECKFP-NEXT: ldaw r10, sp[0]
[XCore] Don't create invalid MKMSK instructions inside loadImmediate(). Summary: Previously loadImmediate() would produce MKMSK instructions with invalid immediate values such as mkmsk r0, 9. Fix this by checking the mask size is valid. Reviewers: robertlytton Reviewed By: robertlytton CC: llvm-commits Differential Revision: http://reviews.llvm.org/D3289 llvm-svn: 206163
2014-04-14 20:30:35 +08:00
; CHECKFP-NEXT: ldc [[REG:r[0-9]+]], 32767
; CHECKFP-NEXT: ldaw r0, r10{{\[}}[[REG]]{{\]}}
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; CHECKFP-NEXT: extsp 1
; CHECKFP-NEXT: bl f5
; CHECKFP-NEXT: ldaw sp, sp[1]
; CHECKFP-NEXT: set sp, r10
; CHECKFP-NEXT: ldw r10, sp[1]
; CHECKFP-NEXT: retsp 32770
;
; !FP + large frame: spill SR+SR+LR = entsp 3 + 32768
AsmPrinter: mark the beginning and the end of a function in verbose mode llvm-svn: 303690
2017-05-24 05:22:16 +08:00
; CHECK-LABEL: f9:
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
; CHECK: entsp 32771
; CHECK-NEXT: ldaw r0, sp[32768]
; CHECK-NEXT: bl f5
; CHECK-NEXT: retsp 32771
[XCore] Don't create invalid MKMSK instructions inside loadImmediate(). Summary: Previously loadImmediate() would produce MKMSK instructions with invalid immediate values such as mkmsk r0, 9. Fix this by checking the mask size is valid. Reviewers: robertlytton Reviewed By: robertlytton CC: llvm-commits Differential Revision: http://reviews.llvm.org/D3289 llvm-svn: 206163
2014-04-14 20:30:35 +08:00
define void @f9() nounwind {
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
entry:
XCore target: Make handling of large frames not dependent upon an FP. eliminateFrameIndex() has been reworked to handle both small & large frames with either a FP or SP. An additional Slot is required for Scavenging spills when not using FP for large frames. Reworked the handling of Register Scavenging. Whether we are using an FP or not, whether it is a large frame or not, and whether we are using a large code model or not are now independent. llvm-svn: 196091
2013-12-02 19:05:28 +08:00
%0 = alloca [32768 x i32]
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-28 03:29:02 +08:00
%1 = getelementptr inbounds [32768 x i32], [32768 x i32]* %0, i32 0, i32 32765
XCore target: Enable frames larger than 65535 to be lowered llvm-svn: 196084
2013-12-02 18:18:14 +08:00
call void @f5(i32* %1)
ret void
}