4 ms·
The M1 MacBook Pro gets just over 1TFlops. best: 1022.61 gflops [numthreads(2, 16, 1)] compute void main(constant float4[] A : register(u0),
by zimpenfish 6y ago
The M1 MacBook Pro gets just over 1TFlops.
best: 1022.61 gflops
[numthreads(2, 16, 1)]
compute void main(constant float4[] A : register(u0),
constant float4[] B : register(u1),
device float4[] C : register(u2),
float3 threadID : SV_DispatchThreadID) {
[...etc...]
- jeffbee 6y agoThere must be some quantization happening here for us to both get the same gflops to 6 significant digits.
- brrrrrm 6y agoI looked into this in the past: it's the timing API (and the fact that I'm not running that many iterations of the kernels). Browsers restrict the resolution to prevent side-channel attacks
- Jonovono 6y agoThat's exactly what I got on my air ``` best: 1022.61 gflops [numthreads(2, 16, 1)] compute void main(constant float4[] A : register(u0), constant float4[] B : register(u1), device float4[] C : register(u2), float3 threadID : SV_DispatchThreadID) { uint m = uint(threadID.x); uint n = uint(threadID.y); ```
- fazulzip 6y agoM1 Mac Mini best: 1022.61 gflops [numthreads(2, 4, 1)] compute void main(constant float4[] A : register(u0), constant float4[] B : register(u1), device float4[] C : register(u2), float3 threadID : SV_DispatchThreadID)
- thebigspacefuck 6y agoMy iPhone XS is faster?? best: 2684.35 gflops [numthreads(16, 16, 1)] compute void main(constant float4[] A : register(u0), constant float4[] B : register(u1), device float4[] C : register(u2), float3 threadID : SV_DispatchThreadID) { uint m = uint(threadID.x); uint n = uint(threadID.y);