Re: [go-nuts] [Question]: Are there any plans to teach the compiler to emit VEX encoded instructions ?

2018-03-01 Thread Agniva De Sarker
Oops, spoke too soon. Got you now Ian. Thanks. On Mar 2, 2018 1:37 AM, "Agniva De Sarker" wrote: > We can still support all processors. I believe it should be possible to > check for cpuid flags and conditionally emit these instructions. > > On Mar 2, 2018 12:23 AM, "Ian Lance Taylor" wrote: >

Re: [go-nuts] [Question]: Are there any plans to teach the compiler to emit VEX encoded instructions ?

2018-03-01 Thread Agniva De Sarker
We can still support all processors. I believe it should be possible to check for cpuid flags and conditionally emit these instructions. On Mar 2, 2018 12:23 AM, "Ian Lance Taylor" wrote: > On Thu, Mar 1, 2018 at 5:05 AM, Agniva De Sarker > wrote: > > > > I believe using the non-destructive 3 o

Re: [go-nuts] [Question]: Are there any plans to teach the compiler to emit VEX encoded instructions ?

2018-03-01 Thread Ian Lance Taylor
On Thu, Mar 1, 2018 at 5:05 AM, Agniva De Sarker wrote: > > I believe using the non-destructive 3 operand form will help a lot in > reducing the size of binaries. And also, it might give us a good base to add > FMA support and SIMD optimizations later. > > I have been adding fast paths to some mat

[go-nuts] [Question]: Are there any plans to teach the compiler to emit VEX encoded instructions ?

2018-03-01 Thread Agniva De Sarker
Hi, I believe using the non-destructive 3 operand form will help a lot in reducing the size of binaries. And also, it might give us a good base to add FMA support and SIMD optimizations later. I have been adding fast paths to some math functions to use AVX instructions, but I was just wonderin