Add some ldm/vldm optimized CopyMem routines. Add performance macros to BDS

git-svn-id: https://edk2.svn.sourceforge.net/svnroot/edk2/trunk/edk2@10388 6f19259b-4bc3-4df7-8a09-765794883524

4 files changed, 389 insertions, 0 deletions

diff --git a/ArmPkg/Library/BaseMemoryLibVstm/Arm/CopyMem.S b/ArmPkg/Library/BaseMemoryLibVstm/Arm/CopyMem.S
new file mode 100755
index 0000000000..2ce686c152
--- /dev/null
+++ b/ArmPkg/Library/BaseMemoryLibVstm/Arm/CopyMem.S
@@ -0,0 +1,114 @@
+#------------------------------------------------------------------------------ 

+#

+# CopyMem() worker for ARM

+#

+# This file started out as C code that did 64 bit moves if the buffer was

+# 32-bit aligned, else it does a byte copy. It also does a byte copy for

+# any trailing bytes. Update using VSTM/SLDM to do 128 byte copies.

+#

+# Copyright (c) 2008-2010 Apple Inc. All rights reserved.<BR>

+# All rights reserved. This program and the accompanying materials

+# are licensed and made available under the terms and conditions of the BSD License

+# which accompanies this distribution.  The full text of the license may be found at

+# http://opensource.org/licenses/bsd-license.php

+#

+# THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS,

+# WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED.

+#

+#------------------------------------------------------------------------------

+

+/**

+  Copy Length bytes from Source to Destination. Overlap is OK.

+

+  This implementation 

+

+  @param  Destination Target of copy

+  @param  Source      Place to copy from

+  @param  Length      Number of bytes to copy

+

+  @return Destination

+

+

+VOID *

+EFIAPI

+InternalMemCopyMem (

+  OUT     VOID                      *DestinationBuffer,

+  IN      CONST VOID                *SourceBuffer,

+  IN      UINTN                     Length

+  )

+**/

+.text

+.align 2

+.globl ASM_PFX(InternalMemCopyMem)

+

+ASM_PFX(InternalMemCopyMem):

+	stmfd	sp!, {r4, r9, lr}

+	tst	r0, #3

+	mov	r4, r0

+	mov	r9, r0

+	mov	ip, r2

+	mov	lr, r1

+	movne	r0, #0

+	bne	L4

+	tst	r1, #3

+	movne	r3, #0

+	moveq	r3, #1

+	cmp	r2, #127

+	movls	r0, #0

+	andhi	r0, r3, #1

+L4:

+	cmp	r4, r1

+	bcc	L26

+	bls	L7

+	rsb	r3, r1, r4

+	cmp	ip, r3

+	bcc	L26

+	cmp	ip, #0

+	beq	L7

+	add	r9, r4, ip

+	add	lr, ip, r1

+	b	L16

+L29:

+	sub	ip, ip, #8

+	cmp	ip, #7

+	ldrd	r2, [lr, #-8]!

+	movls	r0, #0

+	cmp	ip, #0

+	strd	r2, [r9, #-8]!

+	beq	L7

+L16:

+	cmp	r0, #0

+	bne	L29

+	sub	r3, lr, #1

+	sub	ip, ip, #1

+	ldrb	r3, [r3, #0]	

+	sub	r2, r9, #1

+	cmp	ip, #0

+	sub	r9, r9, #1

+	sub	lr, lr, #1

+	strb	r3, [r2, #0]

+	bne	L16

+	b   L7

+L11:

+	ldrb	r3, [lr], #1	

+	sub	ip, ip, #1

+	strb	r3, [r9], #1

+L26:

+	cmp	ip, #0

+	beq	L7

+L30:

+	cmp	r0, #0

+	beq	L11

+	sub	ip, ip, #128          // 32

+	cmp	ip, #127              // 31

+	vldm     lr!, {d0-d15}

+	movls	r0, #0

+	cmp	ip, #0

+	vstm  r9!, {d0-d15}

+	bne	L30

+L7:

+  dsb

+  mov	r0, r4

+	ldmfd	sp!, {r4, r9, pc}

+

+  

diff --git a/ArmPkg/Library/BaseMemoryLibVstm/Arm/CopyMem.asm b/ArmPkg/Library/BaseMemoryLibVstm/Arm/CopyMem.asm
new file mode 100755
index 0000000000..ce049d2eed
--- /dev/null
+++ b/ArmPkg/Library/BaseMemoryLibVstm/Arm/CopyMem.asm
@@ -0,0 +1,115 @@
+;------------------------------------------------------------------------------ 

+;

+; CopyMem() worker for ARM

+;

+; This file started out as C code that did 64 bit moves if the buffer was

+; 32-bit aligned, else it does a byte copy. It also does a byte copy for

+; any trailing bytes. Update using VSTM/SLDM to do 128 byte copies.

+;

+; Copyright (c) 2008-2010 Apple Inc. All rights reserved.<BR>

+; All rights reserved. This program and the accompanying materials

+; are licensed and made available under the terms and conditions of the BSD License

+; which accompanies this distribution.  The full text of the license may be found at

+; http://opensource.org/licenses/bsd-license.php

+;

+; THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS,

+; WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED.

+;

+;------------------------------------------------------------------------------

+

+/**

+  Copy Length bytes from Source to Destination. Overlap is OK.

+

+  This implementation 

+

+  @param  Destination Target of copy

+  @param  Source      Place to copy from

+  @param  Length      Number of bytes to copy

+

+  @return Destination

+

+

+VOID *

+EFIAPI

+InternalMemCopyMem (

+  OUT     VOID                      *DestinationBuffer,

+  IN      CONST VOID                *SourceBuffer,

+  IN      UINTN                     Length

+  )

+**/

+	EXPORT InternalMemCopyMem

+
+	AREA AsmMemStuff, CODE, READONLY
+

+InternalMemCopyMem

+	stmfd	sp!, {r4, r9, lr}
+	tst	r0, #3
+	mov	r4, r0
+	mov	r9, r0
+	mov	ip, r2
+	mov	lr, r1
+	movne	r0, #0
+	bne	L4
+	tst	r1, #3
+	movne	r3, #0
+	moveq	r3, #1
+	cmp	r2, #127
+	movls	r0, #0
+	andhi	r0, r3, #1
+L4
+	cmp	r4, r1
+	bcc	L26
+	bls	L7
+	rsb	r3, r1, r4
+	cmp	ip, r3
+	bcc	L26
+	cmp	ip, #0
+	beq	L7
+	add	r9, r4, ip
+	add	lr, ip, r1
+	b	L16
+L29
+	sub	ip, ip, #8
+	cmp	ip, #7
+	ldrd	r2, [lr, #-8]!
+	movls	r0, #0
+	cmp	ip, #0
+	strd	r2, [r9, #-8]!
+	beq	L7
+L16
+	cmp	r0, #0
+	bne	L29
+	sub	r3, lr, #1
+	sub	ip, ip, #1
+	ldrb	r3, [r3, #0]	
+	sub	r2, r9, #1
+	cmp	ip, #0
+	sub	r9, r9, #1
+	sub	lr, lr, #1
+	strb	r3, [r2, #0]
+	bne	L16
+	b   L7
+L11
+	ldrb	r3, [lr], #1	
+	sub	ip, ip, #1
+	strb	r3, [r9], #1
+L26
+	cmp	ip, #0
+	beq	L7
+L30
+	cmp	r0, #0
+	beq	L11
+	sub	ip, ip, #128          // 32
+	cmp	ip, #127              // 31
+	vldm     lr!, {d0-d15}
+	movls	r0, #0
+	cmp	ip, #0
+	vstm  r9!, {d0-d15}
+	bne	L30
+L7
+  dsb
+  mov	r0, r4
+	ldmfd	sp!, {r4, r9, pc}
+

+  END

+  

diff --git a/ArmPkg/Library/BaseMemoryLibVstm/Arm/SetMem.S b/ArmPkg/Library/BaseMemoryLibVstm/Arm/SetMem.S
new file mode 100755
index 0000000000..44c6d4bece
--- /dev/null
+++ b/ArmPkg/Library/BaseMemoryLibVstm/Arm/SetMem.S
@@ -0,0 +1,80 @@
+#------------------------------------------------------------------------------ 

+#

+# SemMem() worker for ARM

+#

+# This file started out as C code that did 64 bit moves if the buffer was

+# 32-bit aligned, else it does a byte copy. It also does a byte copy for

+# any trailing bytes. Update to use VSTM/VLDM to do 128 byte writes.

+#

+# Copyright (c) 2008-2010 Apple Inc. All rights reserved.<BR>

+# All rights reserved. This program and the accompanying materials

+# are licensed and made available under the terms and conditions of the BSD License

+# which accompanies this distribution.  The full text of the license may be found at

+# http://opensource.org/licenses/bsd-license.php

+#

+# THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS,

+# WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED.

+#

+#------------------------------------------------------------------------------

+

+/**

+  Set Buffer to Value for Size bytes.

+

+  @param  Buffer   Memory to set.

+  @param  Length   Number of bytes to set

+  @param  Value    Value of the set operation.

+

+  @return Buffer

+

+VOID *

+EFIAPI

+InternalMemSetMem (

+  OUT     VOID                      *Buffer,

+  IN      UINTN                     Length,

+  IN      UINT8                     Value

+  )

+**/

+	

+.text

+.align 2

+.globl ASM_PFX(InternalMemSetMem)

+

+ASM_PFX(InternalMemSetMem):

+	stmfd	sp!, {r4-r7, lr}

+	tst	  r0, #3

+	movne	r3, #0

+	moveq	r3, #1

+	cmp	  r1, #127

+	movls lr, #0

+	andhi	lr, r3, #1

+	cmp	  lr, #0

+	mov	  r12, r0

+	bne	  L31

+L32:

+	mov	  r3, #0

+	b	    L43

+L31:

+  vdup.8  q0,r2

+  vmov    q1,q0

+  vmov    q2,q0

+  vmov    q3,q0

+  vmov    q4,q0

+  vmov    q5,q0

+  vmov    q6,q0

+  vmov    q7,q0

+	b	      L32

+L34:

+	cmp	    lr, #0

+	streqb	r2, [r12], #1

+	subeq	  r1, r1, #1

+	beq	    L43

+	sub	    r1, r1, #128

+	cmp	    r1, #127

+	cmp	    r1, #31

+	movls	  lr, r3

+	vstm    r12!, {d0-d15}

+L43:

+	cmp	    r1, #0

+	bne	    L34

+	ldmfd	  sp!, {pc}

+	
\ No newline at end of file
diff --git a/ArmPkg/Library/BaseMemoryLibVstm/Arm/SetMem.asm b/ArmPkg/Library/BaseMemoryLibVstm/Arm/SetMem.asm
new file mode 100755
index 0000000000..4e16ac61bf
--- /dev/null
+++ b/ArmPkg/Library/BaseMemoryLibVstm/Arm/SetMem.asm
@@ -0,0 +1,80 @@
+;------------------------------------------------------------------------------ 

+;

+; SetMem() worker for ARM

+;

+; This file started out as C code that did 64 bit moves if the buffer was

+; 32-bit aligned, else it does a byte copy. It also does a byte copy for

+; any trailing bytes. Update to use VSTM/VLDM to do 128 byte writes.

+;

+; Copyright (c) 2008-2010 Apple Inc. All rights reserved.<BR>

+; All rights reserved. This program and the accompanying materials

+; are licensed and made available under the terms and conditions of the BSD License

+; which accompanies this distribution.  The full text of the license may be found at

+; http://opensource.org/licenses/bsd-license.php

+;

+; THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS,

+; WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED.

+;

+

+/**

+  Set Buffer to Value for Size bytes.

+

+  @param  Buffer   Memory to set.

+  @param  Length   Number of bytes to set

+  @param  Value    Value of the set operation.

+

+  @return Buffer

+

+VOID *

+EFIAPI

+InternalMemSetMem (

+  OUT     VOID                      *Buffer,

+  IN      UINTN                     Length,

+  IN      UINT8                     Value

+  )

+**/

+	

+	EXPORT InternalMemSetMem

+	

+	AREA AsmMemStuff, CODE, READONLY
+

+InternalMemSetMem

+	stmfd	sp!, {lr}

+	tst	  r0, #3

+	movne	r3, #0

+	moveq	r3, #1

+	cmp	  r1, #127

+	movls lr, #0

+	andhi	lr, r3, #1

+	cmp	  lr, #0

+	mov	  r12, r0

+	bne	  L31

+L32

+	mov	  r3, #0

+	b	    L43

+L31

+  vdup.8  q0,r2

+  vmov    q1,q0

+  vmov    q2,q0

+  vmov    q3,q0

+  vmov    q4,q0

+  vmov    q5,q0

+  vmov    q6,q0

+  vmov    q7,q0

+	b	      L32

+L34

+	cmp	    lr, #0

+	streqb	r2, [r12], #1

+	subeq	  r1, r1, #1

+	beq	    L43

+	sub	    r1, r1, #128

+	cmp	    r1, #127

+	movls	  lr, r3

+	vstm    r12!, {d0-d15}

+L43

+	cmp	    r1, #0

+	bne	    L34

+	ldmfd	  sp!, {pc}

+	

+  END

+  
\ No newline at end of file