RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-22 18:54:02 +01:00
Code Issues Projects Releases Wiki Activity
bedf13b1a7
llvm-mirror/test/CodeGen/AArch64/machine-outliner-regsave.mir

130 lines
4.0 KiB
Plaintext
Raw Normal View History

[MachineOutliner][AArch64] Add support for saving LR to a register This teaches the outliner to save LR to a register rather than the stack when possible. This allows us to avoid bumping the stack in outlined functions in some cases. By doing this, in a later patch, we can teach the outliner to do something like this: f1: ... bl OUTLINED_FUNCTION ... f2: ... move LR's contents to a register bl OUTLINED_FUNCTION move the register's contents back instead of falling back to saving LR in both cases. llvm-svn: 338278
2018-07-30 19:45:28 +02:00
# RUN: llc -mtriple=aarch64-apple-darwin -run-pass=prologepilog \
# RUN: -run-pass=machine-outliner -verify-machineinstrs %s -o - | FileCheck %s
# Check that we save LR to a callee-saved register when possible.
# foo() should use a callee-saved register. However, bar() should not.
[MachineOutliner] Outline both register save calls + no LR save calls together Instead of treating the outlined functions for these as distinct frames, they should be combined into one case. Neither allows for stack fixups, and both generate the same frame. Thus, they ought to be considered one case. This makes the code far easier to understand, for one thing. It also offers some small code size improvements. It's fairly rare to see a class of outlined functions that doesn't fall entirely into one variant (on CTMark anyway). It does happen from time to time though. This mostly offers some serious simplification. Also update the test to show the added functionality. llvm-svn: 348036
2018-11-30 22:14:58 +01:00
--- |
[MachineOutliner][AArch64] Add support for saving LR to a register This teaches the outliner to save LR to a register rather than the stack when possible. This allows us to avoid bumping the stack in outlined functions in some cases. By doing this, in a later patch, we can teach the outliner to do something like this: f1: ... bl OUTLINED_FUNCTION ... f2: ... move LR's contents to a register bl OUTLINED_FUNCTION move the register's contents back instead of falling back to saving LR in both cases. llvm-svn: 338278
2018-07-30 19:45:28 +02:00
define void @foo() #0 {
ret void
}
define void @bar() #0 {
ret void
}
Migrate function attribute "no-frame-pointer-elim" to "frame-pointer"="all" as cleanups after D56351
2019-12-25 00:52:21 +01:00
attributes #0 = { minsize noinline noredzone "frame-pointer"="all" }
[MachineOutliner][AArch64] Add support for saving LR to a register This teaches the outliner to save LR to a register rather than the stack when possible. This allows us to avoid bumping the stack in outlined functions in some cases. By doing this, in a later patch, we can teach the outliner to do something like this: f1: ... bl OUTLINED_FUNCTION ... f2: ... move LR's contents to a register bl OUTLINED_FUNCTION move the register's contents back instead of falling back to saving LR in both cases. llvm-svn: 338278
2018-07-30 19:45:28 +02:00
...
---
# Make sure that when we outline and a register is available, we
# use it to save + restore LR instead of SP.
[MachineOutliner] Outline both register save calls + no LR save calls together Instead of treating the outlined functions for these as distinct frames, they should be combined into one case. Neither allows for stack fixups, and both generate the same frame. Thus, they ought to be considered one case. This makes the code far easier to understand, for one thing. It also offers some small code size improvements. It's fairly rare to see a class of outlined functions that doesn't fall entirely into one variant (on CTMark anyway). It does happen from time to time though. This mostly offers some serious simplification. Also update the test to show the added functionality. llvm-svn: 348036
2018-11-30 22:14:58 +01:00
# Also make sure that we can call functions that require no save as the same
# outlined function.
[MachineOutliner][AArch64] Add support for saving LR to a register This teaches the outliner to save LR to a register rather than the stack when possible. This allows us to avoid bumping the stack in outlined functions in some cases. By doing this, in a later patch, we can teach the outliner to do something like this: f1: ... bl OUTLINED_FUNCTION ... f2: ... move LR's contents to a register bl OUTLINED_FUNCTION move the register's contents back instead of falling back to saving LR in both cases. llvm-svn: 338278
2018-07-30 19:45:28 +02:00
# CHECK: name: foo
[MachineOutliner] Outline both register save calls + no LR save calls together Instead of treating the outlined functions for these as distinct frames, they should be combined into one case. Neither allows for stack fixups, and both generate the same frame. Thus, they ought to be considered one case. This makes the code far easier to understand, for one thing. It also offers some small code size improvements. It's fairly rare to see a class of outlined functions that doesn't fall entirely into one variant (on CTMark anyway). It does happen from time to time though. This mostly offers some serious simplification. Also update the test to show the added functionality. llvm-svn: 348036
2018-11-30 22:14:58 +01:00
# CHECK-DAG: bb.1
[MachineOutliner][AArch64] Add support for saving LR to a register This teaches the outliner to save LR to a register rather than the stack when possible. This allows us to avoid bumping the stack in outlined functions in some cases. By doing this, in a later patch, we can teach the outliner to do something like this: f1: ... bl OUTLINED_FUNCTION ... f2: ... move LR's contents to a register bl OUTLINED_FUNCTION move the register's contents back instead of falling back to saving LR in both cases. llvm-svn: 338278
2018-07-30 19:45:28 +02:00
# CHECK-DAG: $x[[REG:[0-9]+]] = ORRXrs $xzr, $lr, 0
# CHECK-NEXT: BL
# CHECK-NEXT: $lr = ORRXrs $xzr, $x[[REG]], 0
[MachineOutliner] Outline both register save calls + no LR save calls together Instead of treating the outlined functions for these as distinct frames, they should be combined into one case. Neither allows for stack fixups, and both generate the same frame. Thus, they ought to be considered one case. This makes the code far easier to understand, for one thing. It also offers some small code size improvements. It's fairly rare to see a class of outlined functions that doesn't fall entirely into one variant (on CTMark anyway). It does happen from time to time though. This mostly offers some serious simplification. Also update the test to show the added functionality. llvm-svn: 348036
2018-11-30 22:14:58 +01:00
# CHECK-DAG: bb.2
[MachineOutliner][AArch64] Add support for saving LR to a register This teaches the outliner to save LR to a register rather than the stack when possible. This allows us to avoid bumping the stack in outlined functions in some cases. By doing this, in a later patch, we can teach the outliner to do something like this: f1: ... bl OUTLINED_FUNCTION ... f2: ... move LR's contents to a register bl OUTLINED_FUNCTION move the register's contents back instead of falling back to saving LR in both cases. llvm-svn: 338278
2018-07-30 19:45:28 +02:00
# CHECK-DAG: $x[[REG]] = ORRXrs $xzr, $lr, 0
# CHECK-NEXT: BL
# CHECK-NEXT: $lr = ORRXrs $xzr, $x[[REG]], 0
[MachineOutliner] Outline both register save calls + no LR save calls together Instead of treating the outlined functions for these as distinct frames, they should be combined into one case. Neither allows for stack fixups, and both generate the same frame. Thus, they ought to be considered one case. This makes the code far easier to understand, for one thing. It also offers some small code size improvements. It's fairly rare to see a class of outlined functions that doesn't fall entirely into one variant (on CTMark anyway). It does happen from time to time though. This mostly offers some serious simplification. Also update the test to show the added functionality. llvm-svn: 348036
2018-11-30 22:14:58 +01:00
# CHECK-DAG: bb.3
[MachineOutliner][AArch64] Add support for saving LR to a register This teaches the outliner to save LR to a register rather than the stack when possible. This allows us to avoid bumping the stack in outlined functions in some cases. By doing this, in a later patch, we can teach the outliner to do something like this: f1: ... bl OUTLINED_FUNCTION ... f2: ... move LR's contents to a register bl OUTLINED_FUNCTION move the register's contents back instead of falling back to saving LR in both cases. llvm-svn: 338278
2018-07-30 19:45:28 +02:00
# CHECK-DAG: $x[[REG]] = ORRXrs $xzr, $lr, 0
# CHECK-NEXT: BL
# CHECK-NEXT: $lr = ORRXrs $xzr, $x[[REG]], 0
[MachineOutliner] Outline both register save calls + no LR save calls together Instead of treating the outlined functions for these as distinct frames, they should be combined into one case. Neither allows for stack fixups, and both generate the same frame. Thus, they ought to be considered one case. This makes the code far easier to understand, for one thing. It also offers some small code size improvements. It's fairly rare to see a class of outlined functions that doesn't fall entirely into one variant (on CTMark anyway). It does happen from time to time though. This mostly offers some serious simplification. Also update the test to show the added functionality. llvm-svn: 348036
2018-11-30 22:14:58 +01:00
# CHECK-DAG: bb.4
# CHECK-NOT: $x[[REG]] = ORRXrs $xzr, $lr, 0
# CHECK-DAG: BL
[MachineOutliner][AArch64] Add support for saving LR to a register This teaches the outliner to save LR to a register rather than the stack when possible. This allows us to avoid bumping the stack in outlined functions in some cases. By doing this, in a later patch, we can teach the outliner to do something like this: f1: ... bl OUTLINED_FUNCTION ... f2: ... move LR's contents to a register bl OUTLINED_FUNCTION move the register's contents back instead of falling back to saving LR in both cases. llvm-svn: 338278
2018-07-30 19:45:28 +02:00
name: foo
tracksRegLiveness: true
[MachineOutliner] Outline both register save calls + no LR save calls together Instead of treating the outlined functions for these as distinct frames, they should be combined into one case. Neither allows for stack fixups, and both generate the same frame. Thus, they ought to be considered one case. This makes the code far easier to understand, for one thing. It also offers some small code size improvements. It's fairly rare to see a class of outlined functions that doesn't fall entirely into one variant (on CTMark anyway). It does happen from time to time though. This mostly offers some serious simplification. Also update the test to show the added functionality. llvm-svn: 348036
2018-11-30 22:14:58 +01:00
fixedStack:
[MachineOutliner][AArch64] Add support for saving LR to a register This teaches the outliner to save LR to a register rather than the stack when possible. This allows us to avoid bumping the stack in outlined functions in some cases. By doing this, in a later patch, we can teach the outliner to do something like this: f1: ... bl OUTLINED_FUNCTION ... f2: ... move LR's contents to a register bl OUTLINED_FUNCTION move the register's contents back instead of falling back to saving LR in both cases. llvm-svn: 338278
2018-07-30 19:45:28 +02:00
body: |
bb.0:
[AArch64] Fix MIR tests with invalid live-ins. A register can't be live if it isn't defined; fix issues in various testcases. Differential Revision: https://reviews.llvm.org/D78531
2020-04-21 03:27:07 +02:00
$x9 = ORRXri $xzr, 1
[MachineOutliner] Outline both register save calls + no LR save calls together Instead of treating the outlined functions for these as distinct frames, they should be combined into one case. Neither allows for stack fixups, and both generate the same frame. Thus, they ought to be considered one case. This makes the code far easier to understand, for one thing. It also offers some small code size improvements. It's fairly rare to see a class of outlined functions that doesn't fall entirely into one variant (on CTMark anyway). It does happen from time to time though. This mostly offers some serious simplification. Also update the test to show the added functionality. llvm-svn: 348036
2018-11-30 22:14:58 +01:00
$lr = ORRXri $xzr, 1
bb.1:
liveins: $lr, $w9
[MachineOutliner][AArch64] Add support for saving LR to a register This teaches the outliner to save LR to a register rather than the stack when possible. This allows us to avoid bumping the stack in outlined functions in some cases. By doing this, in a later patch, we can teach the outliner to do something like this: f1: ... bl OUTLINED_FUNCTION ... f2: ... move LR's contents to a register bl OUTLINED_FUNCTION move the register's contents back instead of falling back to saving LR in both cases. llvm-svn: 338278
2018-07-30 19:45:28 +02:00
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 2
[MachineOutliner] Outline both register save calls + no LR save calls together Instead of treating the outlined functions for these as distinct frames, they should be combined into one case. Neither allows for stack fixups, and both generate the same frame. Thus, they ought to be considered one case. This makes the code far easier to understand, for one thing. It also offers some small code size improvements. It's fairly rare to see a class of outlined functions that doesn't fall entirely into one variant (on CTMark anyway). It does happen from time to time though. This mostly offers some serious simplification. Also update the test to show the added functionality. llvm-svn: 348036
2018-11-30 22:14:58 +01:00
bb.2:
[MachineOutliner][AArch64] Add support for saving LR to a register This teaches the outliner to save LR to a register rather than the stack when possible. This allows us to avoid bumping the stack in outlined functions in some cases. By doing this, in a later patch, we can teach the outliner to do something like this: f1: ... bl OUTLINED_FUNCTION ... f2: ... move LR's contents to a register bl OUTLINED_FUNCTION move the register's contents back instead of falling back to saving LR in both cases. llvm-svn: 338278
2018-07-30 19:45:28 +02:00
liveins: $lr, $w9
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 2
[MachineOutliner] Outline both register save calls + no LR save calls together Instead of treating the outlined functions for these as distinct frames, they should be combined into one case. Neither allows for stack fixups, and both generate the same frame. Thus, they ought to be considered one case. This makes the code far easier to understand, for one thing. It also offers some small code size improvements. It's fairly rare to see a class of outlined functions that doesn't fall entirely into one variant (on CTMark anyway). It does happen from time to time though. This mostly offers some serious simplification. Also update the test to show the added functionality. llvm-svn: 348036
2018-11-30 22:14:58 +01:00
bb.3:
liveins: $lr, $w9
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 2
bb.4:
[MachineOutliner][AArch64] Add support for saving LR to a register This teaches the outliner to save LR to a register rather than the stack when possible. This allows us to avoid bumping the stack in outlined functions in some cases. By doing this, in a later patch, we can teach the outliner to do something like this: f1: ... bl OUTLINED_FUNCTION ... f2: ... move LR's contents to a register bl OUTLINED_FUNCTION move the register's contents back instead of falling back to saving LR in both cases. llvm-svn: 338278
2018-07-30 19:45:28 +02:00
liveins: $lr, $w9
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 1
$w9 = ORRWri $wzr, 2
[MachineOutliner] Outline both register save calls + no LR save calls together Instead of treating the outlined functions for these as distinct frames, they should be combined into one case. Neither allows for stack fixups, and both generate the same frame. Thus, they ought to be considered one case. This makes the code far easier to understand, for one thing. It also offers some small code size improvements. It's fairly rare to see a class of outlined functions that doesn't fall entirely into one variant (on CTMark anyway). It does happen from time to time though. This mostly offers some serious simplification. Also update the test to show the added functionality. llvm-svn: 348036
2018-11-30 22:14:58 +01:00
bb.5:
liveins: $w9
[MachineOutliner][AArch64] Add support for saving LR to a register This teaches the outliner to save LR to a register rather than the stack when possible. This allows us to avoid bumping the stack in outlined functions in some cases. By doing this, in a later patch, we can teach the outliner to do something like this: f1: ... bl OUTLINED_FUNCTION ... f2: ... move LR's contents to a register bl OUTLINED_FUNCTION move the register's contents back instead of falling back to saving LR in both cases. llvm-svn: 338278
2018-07-30 19:45:28 +02:00
RET undef $lr
...
---
# Convoluted case that shows that we'll still save to the stack when there are
# no approprate registers available.
# The live-in lists do not contain x16 or x17 since including them would cause
# nothing to be outlined.
# They also deliberately don't contain x18 to show that on Darwin we won't store
# to that.
# CHECK-LABEL: name: bar
# CHECK: early-clobber $sp = STRXpre $lr, $sp, -16
# CHECK-NEXT: BL
# CHECK-DAG: early-clobber $sp, $lr = LDRXpost $sp, 16
# CHECK: early-clobber $sp = STRXpre $lr, $sp, -16
# CHECK-NEXT: BL
# CHECK-DAG: early-clobber $sp, $lr = LDRXpost $sp, 16
# CHECK: early-clobber $sp = STRXpre $lr, $sp, -16
# CHECK-NEXT: BL
# CHECK-NEXT: early-clobber $sp, $lr = LDRXpost $sp, 16
name: bar
tracksRegLiveness: true
body: |
bb.0:
liveins: $lr, $x0, $x1, $x2, $x3, $x4, $x5, $x6, $x7, $x8, $x9, $x10, $x11, $x12, $x13, $x14, $x15, $x19, $x20, $x21, $x22, $x23, $x20, $x21, $x22, $x23, $x24, $x25, $x26, $x27, $x28
[AArch64] Save FP for leaf functions when disabling frame pointer elimination The change allows clang -mno-omit-leaf-frame-pointer to disable frame pointer elimination. This behavior matches X86 and Mips, and also GCC AArch64. Reviewed By: efriedma Differential Revision: https://reviews.llvm.org/D71168
2019-12-07 19:23:30 +01:00
$lr = ORRXri $xzr, 1
[MachineOutliner][AArch64] Add support for saving LR to a register This teaches the outliner to save LR to a register rather than the stack when possible. This allows us to avoid bumping the stack in outlined functions in some cases. By doing this, in a later patch, we can teach the outliner to do something like this: f1: ... bl OUTLINED_FUNCTION ... f2: ... move LR's contents to a register bl OUTLINED_FUNCTION move the register's contents back instead of falling back to saving LR in both cases. llvm-svn: 338278
2018-07-30 19:45:28 +02:00
$w10 = ORRWri $wzr, 1
$w10 = ORRWri $wzr, 1
$w10 = ORRWri $wzr, 1
$w10 = ORRWri $wzr, 1
$w10 = ORRWri $wzr, 1
$w12 = ORRWri $wzr, 2
bb.1:
liveins: $lr, $x0, $x1, $x2, $x3, $x4, $x5, $x6, $x7, $x8, $x9, $x10, $x11, $x12, $x13, $x14, $x15, $x19, $x20, $x21, $x22, $x23, $x20, $x21, $x22, $x23, $x24, $x25, $x26, $x27, $x28
$w10 = ORRWri $wzr, 1
$w10 = ORRWri $wzr, 1
$w10 = ORRWri $wzr, 1
$w10 = ORRWri $wzr, 1
$w10 = ORRWri $wzr, 1
$w12 = ORRWri $wzr, 2
bb.2:
liveins: $lr, $x0, $x1, $x2, $x3, $x4, $x5, $x6, $x7, $x8, $x9, $x10, $x11, $x12, $x13, $x14, $x15, $x19, $x20, $x21, $x22, $x23, $x20, $x21, $x22, $x23, $x24, $x25, $x26, $x27, $x28
$w10 = ORRWri $wzr, 1
$w10 = ORRWri $wzr, 1
$w10 = ORRWri $wzr, 1
$w10 = ORRWri $wzr, 1
$w10 = ORRWri $wzr, 1
$w12 = ORRWri $wzr, 2
bb.3:
liveins: $lr, $x0, $x1, $x2, $x3, $x4, $x5, $x6, $x7, $x8, $x9, $x10, $x11, $x12, $x13, $x14, $x15, $x19, $x20, $x21, $x22, $x23, $x20, $x21, $x22, $x23, $x24, $x25, $x26, $x27, $x28
RET undef $lr
Reference in New Issue Copy Permalink