FedoraVmProfileResults-01

FedoraVmProfileResults.txt

Same use case of decode a 1080p video 1000 times.

This was done using the "Perf" tool

Compiler settings
- opt level: -Og
- -g


------------------------------------------------------------------

Run #1:


Expensive Functions:

	11.13% h264dec libc-2.30.so	[.] __memset_sse2_unaligned_erms
	9.87%  h264dec h264dec		[.] WelsDec::DecodeCurrentAccessUnit
	6.12%  h264dec h264dec		[.] WelsDec::ParseSliceHeaderSyntaxs
	5.13%  h264dec h264dec		[.] WelsDec::CavlcGetLevelVal


__memset_sse2_unaligned_erms - Expensive Lines

	20.74 97: movdqa %xmm0, (%rcx)
	66.74 	movdqa	%xmm0, 0x10(%rcx)

	8.09	movdqa	%xmm0, 0x30(%rcx)


WelsDec::DecodeCurrentAccessUnit - Expensive Lines

	30.02	rep	movsq %ds:(%rsi),%es:(%rdi)

	0.92	mov	%r8,0x1500(%r12)	

	1.12	mov	0x894(%r14),%ecx


WelsDec::ParseSliceHeaderSyntaxs - Expensive Lines

	40.30	rep stos %rax,%es:(%rdi)

	1.06	mov %r10b, 0x1321(%rbp)


WelsDec::CavlcGetLevelVal - Expensive Lines
	
	lea 	WelsDec::g_kuiPrefix8BitsTable,%r12

	2.98	mov	%ecx,%eax

	4.89	->jg	444e14 <WelsDec::CavlcGetLevelValPint *, WelsDec::TagReadBitsCache*, unsigned char, unsigned char)+0x294>


------------------------------------------------------------------

Run #2:

Expensive Functions:

	11.45%	h264dec	libc-2.30.so	[.] __memset_sse2_unaligned_erms
	 9.83%	h264dec	h264dec		[.] WelsDec::DecodeCurrentAccessUnit
	 6.13%	h264dec	h264dec		[.] WelsDec::ParseSliceHeaderSyntaxs
	 5.07%	h264dec	h264dec		[.] WelsDec::CavlcGetLevelVal


__memset_sse2_unaligned_erms - Expensive Lines

	20.07	movdqa	%xmm0,(%rcx)
	67.03	movdqa	%xmm0,0x10(%rcx)

	8.64	movdqa	%xmm0,0x30(%rcx)


WelsDec::DecodeCurrentAccessUnit - Expensive Lines

	28.56	rep	movsq %ds:(%rsi),%es:(%rdi)

	29.72	rep	movsq %ds:(%rsi),%es:(%rdi)

	26.80	rep	stos %rax,%es:(%rdi)
		

WelsDec::ParseSliceHeaderSyntaxs - Expensive Lines

	39.73	rep	stos %rax,%es:(%rdi)

	9.78	rep	movsq %ds:(%rsi),%es:(%rdi)


WelsDec::CavlcGetLevelVal - Expensive Lines

	3.44	add	$0x2,%rcx

	2.85	mov	%ecx,%eax

	4.80	->jg	444e14 <WelsDec::CavlcGetLevelVal(int*, WelsDec::TagReadBitsCache*, unsigned char, unsigned char)+0x284>

	4.54	xor	%fs:0x28,%rax