Skip to contentSkip to conversation

x86 change mov 0, %reg to xor %reg, %reg

PATCH In review Kanna Scarlet 5 participants 13 messages Aug 4, 2022 – Aug 9, 2022

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover x86 change mov 0, %reg to xor %reg, %reg

Hello Linux x86 maintainers,

I'm an informatic student 19 y.o. I am still studying Linux kernel open source in GNU/Weeb community. I want to be a linux kerne dev in the future. This is my first time sending a patch to Linux Kernel, I am still learning the community. I may make a mistake in this email, please correct me if i am wrong

I want to improve x86-64 assembly code with this patch. This patch changes mov $0, %reg with xor %reg, %reg because xor %reg, %reg is smaller so it is good to save space

asm:

ba 00 00 00 00  mov   $0x0,%edx
31 d2           xor   %edx,%edx

Regards,

---

Kanna Scarlet (1):
x86: Change mov $0, %reg with xor %reg, %reg

 arch/x86/boot/compressed/head_64.S     | 2 +-
 arch/x86/boot/compressed/mem_encrypt.S | 2 +-
 arch/x86/kernel/ftrace_32.S            | 4 ++--
 arch/x86/kernel/head_64.S              | 2 +-
 arch/x86/math-emu/div_Xsig.S           | 2 +-
 arch/x86/math-emu/reg_u_sub.S          | 2 +-
 6 files changed, 7 insertions(+), 7 deletions(-)

base-commit: ff89dd08c0f0a3fd330c9ef9d775e880f82c291e

Signature · 1 line
Kanna Scarlet
#1 lore raw

PATCH 1/1 x86: Change mov $0, %reg with xor %reg, %reg

Change mov $0, %reg with xor %reg, %reg because xor %reg, %reg is smaller so it is good to save space

asm:

ba 00 00 00 00  movl   $0x0,%edx
31 d2           xorl   %edx,%edx
---
 arch/x86/boot/compressed/head_64.S     | 2 +-
 arch/x86/boot/compressed/mem_encrypt.S | 2 +-
 arch/x86/kernel/ftrace_32.S            | 4 ++--
 arch/x86/kernel/head_64.S              | 2 +-
 arch/x86/math-emu/div_Xsig.S           | 2 +-
 arch/x86/math-emu/reg_u_sub.S          | 2 +-
 6 files changed, 7 insertions(+), 7 deletions(-)
arch/x86/boot/compressed/head_64.S +1 -1
index d33f060900d2..39442e7f5993 100644
@@ -666,7 +666,7 @@ SYM_CODE_START(trampoline_32bit_src)
 	movl	%cr4, %eax
 	andl	$X86_CR4_MCE, %eax
 #else
-	movl	$0, %eax
+	xorl	%eax, %eax
 #endif
 
 	/* Enable PAE and LA57 (if required) paging modes */
arch/x86/boot/compressed/mem_encrypt.S +1 -1
index a73e4d783cae..d1e4d3aa8395 100644
@@ -111,7 +111,7 @@ SYM_CODE_START(startup32_vc_handler)
 	cmpl	$0x72, 16(%esp)
 	jne	.Lfail
 
-	movl	$0, %eax		# Request CPUID[fn].EAX
+	xorl	%eax, %eax		# Request CPUID[fn].EAX
 	movl	%ebx, %edx		# CPUID fn
 	call	sev_es_req_cpuid	# Call helper
 	testl	%eax, %eax		# Check return code
arch/x86/kernel/ftrace_32.S +2 -2
index a0ed0e4a2c0c..cff7decb58be 100644
@@ -171,7 +171,7 @@ SYM_CODE_START(ftrace_graph_caller)
 	movl	3*4(%esp), %eax
 	/* Even with frame pointers, fentry doesn't have one here */
 	lea	4*4(%esp), %edx
-	movl	$0, %ecx
+	xorl	%ecx, %ecx
 	subl	$MCOUNT_INSN_SIZE, %eax
 	call	prepare_ftrace_return
 	popl	%edx
@@ -184,7 +184,7 @@ SYM_CODE_END(ftrace_graph_caller)
 return_to_handler:
 	pushl	%eax
 	pushl	%edx
-	movl	$0, %eax
+	xorl	%eax, %eax
 	call	ftrace_return_to_handler
 	movl	%eax, %ecx
 	popl	%edx
arch/x86/kernel/head_64.S +1 -1
index d860d437631b..eeb06047e30a 100644
@@ -184,7 +184,7 @@ SYM_INNER_LABEL(secondary_startup_64_no_verify, SYM_L_GLOBAL)
 	movq	%cr4, %rcx
 	andl	$X86_CR4_MCE, %ecx
 #else
-	movl	$0, %ecx
+	xorl	%ecx, %ecx
 #endif
 
 	/* Enable PAE mode, PGE and LA57 */
arch/x86/math-emu/div_Xsig.S +1 -1
index 8c270ab415be..5767b4d23954 100644
@@ -122,7 +122,7 @@ SYM_FUNC_START(div_Xsig)
 	movl	XsigLL(%esi),%eax
 	rcrl	%eax
 	movl	%eax,FPU_accum_1
-	movl	$0,%eax
+	xorl	%eax,%eax
 	rcrl	%eax
 	movl	%eax,FPU_accum_0
 
arch/x86/math-emu/reg_u_sub.S +1 -1
index 4c900c29e4ff..130b49fa1ca2 100644
@@ -212,7 +212,7 @@ L_must_be_zero:
 L_shift_32:
 	movl	%ebx,%eax
 	movl	%edx,%ebx
-	movl	$0,%edx
+	xorl	%edx,%edx
 	subw	$32,EXP(%edi)	/* Can get underflow here */
 
 /* We need to shift left by 1 - 31 bits */
Signature · 1 line
Kanna Scarlet
maintainer
#2 lore raw

Re: [PATCH 1/1] x86: Change mov $0, %reg with xor %reg, %reg

Quoted 2 lines · Kanna Scarlet

Change mov $0, %reg with xor %reg, %reg because xor %reg, %reg is smaller so it is good to save space

Bonus points if you find out what other advantage

XOR reg,reg

has when it comes to clearing integer registers.

Signature · 4 lines
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette
#3 lore raw

Re: [PATCH 1/1] x86: Change mov $0, %reg with xor %reg, %reg

Quoted 5 lines · Borislav Petkov

Bonus points if you find out what other advantage

XOR reg,reg

has when it comes to clearing integer registers.

Hello sir Borislav,

Thank you for your response. I tried to find out other advantages of xor reg,reg on Google and found this:
https://stackoverflow.com/a/33668295/7275114

"xor (being a recognized zeroing idiom, unlike mov reg, 0) has some obvious and some subtle advantages:

1. smaller code-size than mov reg,0. (All CPUs)
2. avoids partial-register penalties for later code. (Intel P6-family and SnB-family).
3. doesn't use an execution unit, saving power and freeing up execution resources. (Intel SnB-family)
4. smaller uop (no immediate data) leaves room in the uop cache-line for nearby instructions to borrow if needed. (Intel SnB-family).
5. doesn't use up entries in the physical register file. (Intel SnB-family (and P4) at least, possibly AMD as well since they use a similar PRF design instead of keeping register state in the ROB like Intel P6-family microarchitectures.)"

Should I add all in the explanation sir? I will send v2 revision tomorrow.

We also find more files to patch with this command:

grep -rE "mov.?\s+\\$\\0\s*," arch/x86

it shows many immediate zero moves to 64-bit register in file arch/x86/crypto/curve25519-x86_64.c, but the next instruction may depend on the previous %rflags value, we are afraid to change this because xor touches %rflags. We will try to change it to movl $0, %r32 to reduce the code size.

Example cmovc needs %rflags

    "  adcx %1, %%r11;"
    "  movq %%r11, 24(%2);"

    /* Step 3: Fold the carry bit back in; guaranteed not to carry at this point */
    "  mov $0, %%rax;"
    "  cmovc %%rdx, %%rax;"

Thanks.

Regards,

Signature · 1 line
Kanna Scarlet
#4 lore raw

RE: [PATCH 1/1] x86: Change mov $0, %reg with xor %reg, %reg

From: Kanna Scarlet

Quoted 18 lines

Sent: 04 August 2022 19:08

Quoted 5 lines · Borislav Petkov

Bonus points if you find out what other advantage

XOR reg,reg

has when it comes to clearing integer registers.

Hello sir Borislav,

Thank you for your response. I tried to find out other advantages of xor reg,reg on Google and found this:
https://stackoverflow.com/a/33668295/7275114

"xor (being a recognized zeroing idiom, unlike mov reg, 0) has some obvious and some subtle advantages:

Quoted 11 lines

1. smaller code-size than mov reg,0. (All CPUs)
2. avoids partial-register penalties for later code. (Intel P6-family and SnB-family).
3. doesn't use an execution unit, saving power and freeing up execution resources. (Intel SnB-family)
4. smaller uop (no immediate data) leaves room in the uop cache-line for nearby instructions to borrow if needed. (Intel SnB-family).
5. doesn't use up entries in the physical register file. (Intel SnB-family (and P4) at least, possibly AMD as well since they use a similar PRF design instead of keeping register state in the ROB like Intel P6-family microarchitectures.)"

You missed one, and an additional change:

Use "xor %rax,%rax" instead of "xor %eax,%eax" to save the 'reg' prefix.

David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)

#5 lore raw

Re: [PATCH 1/1] x86: Change mov $0, %reg with xor %reg, %reg

Quoted 2 lines · David Laight

Use "xor %rax,%rax" instead of "xor %eax,%eax" to save the 'reg' prefix.

Also, some places explicitly use the mov variant to zero a register without touching rflags. Please be careful to not change those.

Regards,

Signature · 10 lines
Jörg Rödel
[email protected]

SUSE Software Solutions Germany GmbH
Frankenstraße 146
90461 Nürnberg
Germany

(HRB 36809, AG Nürnberg)
Geschäftsführer: Ivo Totev, Andrew Myers, Andrew McDonald, Boudien Moerman
#6 lore raw

Re: [PATCH 1/1] x86: Change mov $0, %reg with xor %reg, %reg

Quoted 6 lines · Joerg Roedel
Quoted 2 lines · David Laight

Use "xor %rax,%rax" instead of "xor %eax,%eax" to save the 'reg' prefix.

Also, some places explicitly use the mov variant to zero a register without touching rflags. Please be careful to not change those.

thank you for reminder, i will check again to make myself more sure the patch doesn't break this %rflags dependency situation

Regards,

Signature · 1 line
Kanna Scarlet
#7 lore raw

Re: [PATCH 1/1] x86: Change mov $0, %reg with xor %reg, %reg

Quoted 5 lines · Kanna Scarlet
Quoted 4 lines · Joerg Roedel
Quoted 2 lines · David Laight

Use "xor %rax,%rax" instead of "xor %eax,%eax" to save the 'reg' prefix.

Quoted 7 lines
Quoted 2 lines

Also, some places explicitly use the mov variant to zero a register without touching rflags. Please be careful to not change those.

thank you for reminder, i will check again to make myself more sure the patch doesn't break this %rflags dependency situation

Regards,

In some cases you can hoist the zeroing to avoid that (and sometimes improve performance in the process), but be very careful in general when messing with hand-optimized assembly code like crypto; for those pieces of code benchmarking the change is mandatory.

#8 lore raw

Re: [PATCH 1/1] x86: Change mov $0, %reg with xor %reg, %reg

In reply to David Laight
Quoted 2 lines · David Laight

Use "xor %rax,%rax" instead of "xor %eax,%eax" to save the 'reg' prefix.

hello David Laight

"xor %rax,%rax" is bigger because of rex prefix, "xor %eax,%eax" is smaller because it doesn't need rex prefix.

asm:

   0:   48 31 c0  xor    %rax,%rax
   3:   31 c0     xor    %eax,%eax

so i think to save from rex prefix, use xor %eax,%eax instead of xor %rax,%rax.

Best regards,

Signature · 1 line
Kanna Scarlet
#9 lore raw

Re: [PATCH 1/1] x86: Change mov $0, %reg with xor %reg, %reg

In reply to David Laight
Quoted 33 lines · David Laight

From: Kanna Scarlet

Quoted 18 lines

Sent: 04 August 2022 19:08

Quoted 5 lines · Borislav Petkov

Bonus points if you find out what other advantage

XOR reg,reg

has when it comes to clearing integer registers.

Hello sir Borislav,

Thank you for your response. I tried to find out other advantages of xor reg,reg on Google and found this:
https://stackoverflow.com/a/33668295/7275114

"xor (being a recognized zeroing idiom, unlike mov reg, 0) has some obvious and some subtle advantages:

Quoted 11 lines

1. smaller code-size than mov reg,0. (All CPUs)
2. avoids partial-register penalties for later code. (Intel P6-family and SnB-family).
3. doesn't use an execution unit, saving power and freeing up execution resources. (Intel SnB-family)
4. smaller uop (no immediate data) leaves room in the uop cache-line for nearby instructions to borrow if needed. (Intel SnB-family).
5. doesn't use up entries in the physical register file. (Intel SnB-family (and P4) at least, possibly AMD as well since they use a similar PRF design instead of keeping register state in the ROB like Intel P6-family microarchitectures.)"

You missed one, and an additional change:

Quoted 10 lines

Use "xor %rax,%rax" instead of "xor %eax,%eax" to save the 'reg' prefix.

David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)

You mean the other way around...

#10 lore raw

RE: [PATCH 1/1] x86: Change mov $0, %reg with xor %reg, %reg

From: H. Peter Anvin

Quoted 42 lines

Sent: 08 August 2022 20:00

Quoted 38 lines · David Laight

From: Kanna Scarlet

Quoted 18 lines

Sent: 04 August 2022 19:08

Quoted 5 lines · Borislav Petkov

Bonus points if you find out what other advantage

XOR reg,reg

has when it comes to clearing integer registers.

Hello sir Borislav,

Thank you for your response. I tried to find out other advantages of xor reg,reg on Google and found this:
https://stackoverflow.com/a/33668295/7275114

"xor (being a recognized zeroing idiom, unlike mov reg, 0) has some obvious and some subtle advantages:

Quoted 11 lines

1. smaller code-size than mov reg,0. (All CPUs)
2. avoids partial-register penalties for later code. (Intel P6-family and SnB-family).
3. doesn't use an execution unit, saving power and freeing up execution resources. (Intel SnB-family)
4. smaller uop (no immediate data) leaves room in the uop cache-line for nearby instructions to borrow if needed. (Intel SnB-family).
5. doesn't use up entries in the physical register file. (Intel SnB-family (and P4) at least, possibly AMD as well since they use a similar PRF design instead of keeping register state in the ROB like Intel P6-family microarchitectures.)"

You missed one, and an additional change:

Use "xor %rax,%rax" instead of "xor %eax,%eax" to save the 'reg' prefix.

David

Quoted 7 lines
Quoted 5 lines

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)

You mean the other way around...

Maybe :-(
The 32bit versions are best.
Somehow the register naming convention ended up getting sort of 'backwards'. 'register' is bigger than 'extended'.
I've 'only' been writing x86 asm since 1982!

David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)

maintainer
#11 lore raw

Re: [PATCH 1/1] x86: Change mov $0, %reg with xor %reg, %reg

In reply to Kanna Scarlet
Quoted 1 line · Kanna Scarlet

Hello sir Borislav,

Please, no "sir" - just Boris or Borislav,

Quoted 19 lines

Thank you for your response. I tried to find out other advantages of xor reg,reg on Google and found this:
https://stackoverflow.com/a/33668295/7275114

"xor (being a recognized zeroing idiom, unlike mov reg, 0) has some obvious and some subtle advantages:

1. smaller code-size than mov reg,0. (All CPUs)
2. avoids partial-register penalties for later code. (Intel P6-family and SnB-family).
3. doesn't use an execution unit, saving power and freeing up execution resources. (Intel SnB-family)
4. smaller uop (no immediate data) leaves room in the uop cache-line for nearby instructions to borrow if needed. (Intel SnB-family).
5. doesn't use up entries in the physical register file. (Intel SnB-family (and P4) at least, possibly AMD as well since they use a similar PRF design instead of keeping register state in the ROB like Intel P6-family microarchitectures.)"

Quoted 1 line

Should I add all in the explanation sir?

You should try to understand what this means and write the gist of it in your own words. This is how you can learn something.

Quoted 9 lines

We also find more files to patch with this command:

   grep -rE "mov.?\s+\\$\\0\s*," arch/x86

it shows many immediate zero moves to 64-bit register in file arch/x86/crypto/curve25519-x86_64.c, but the next instruction may depend on the previous %rflags value, we are afraid to change this because xor touches %rflags. We will try to change it to movl $0, %r32 to reduce the code size.

I don't think you need to do that - you can do this one patch in order to go through the whole process of creating and submitting a patch but you should not go on a "let's convert everything" spree just for the sake of it.

Because maintainers barely have time to look at patches, you don't have to send them more when they're not really needed.

Rather, I'd suggest you go and try to fix real bugs. This has some ideas what to do:

https://www.linux.com/news/three-ways-beginners-contribute-linux-kernel/

Looking at the kernel bugzilla and trying to understand and reproduce a bug from there would get you a long way. And you'll learn a lot.

Also, you should peruse

https://www.kernel.org/doc/html/latest/process/index.html

which has a lot of information about how this whole community thing works.

I sincerely hope that helps.

Thx.

Signature · 4 lines
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette
#12 lore raw

Re: [PATCH 1/1] x86: Change mov $0, %reg with xor %reg, %reg

Quoted 4 lines · Borislav Petkov
Quoted 1 line · Kanna Scarlet

Hello sir Borislav,

Please, no "sir" - just Boris or Borislav,

ok, sorry

Quoted 4 lines

I don't think you need to do that - you can do this one patch in order to go through the whole process of creating and submitting a patch but you should not go on a "let's convert everything" spree just for the sake of it.

ok, i will try to finish the process for this one patch for learning the submitting process. After that I will avoid touching similar small improvement and focus on real kernel bugs/issues, i'll send v2 revision with only commit message improvement

Quoted 13 lines

Because maintainers barely have time to look at patches, you don't have to send them more when they're not really needed.

Rather, I'd suggest you go and try to fix real bugs. This has some ideas what to do:

https://www.linux.com/news/three-ways-beginners-contribute-linux-kernel/

Looking at the kernel bugzilla and trying to understand and reproduce a bug from there would get you a long way. And you'll learn a lot.

Also, you should peruse

Quoted 8 lines

https://www.kernel.org/doc/html/latest/process/index.html

which has a lot of information about how this whole community thing works.

I sincerely hope that helps.

Thx.

thank you for the guide, I'm following it

Regards,

Signature · 1 line
Kanna Scarlet