EPYC 7302P + 3090 2 x C2 in parallel won't work #24
Description
[](2021-02-08T00:08:47.967 INFO bellperson::groth16::prover > ZQ: build provers end: 143.060379629s
2021-02-08T00:08:47.968 INFO bellperson::groth16::prover > ZQ: starting proof timer
2021-02-08T00:08:47.968 INFO bellperson::groth16::prover > ZQ: get params start
2021-02-08T00:10:20.109 INFO bellperson::groth16::prover > ZQ: waiting params...
2021-02-08T00:10:20.119 INFO bellperson::groth16::prover > ZQ: get params end: 92.15143134s
2021-02-08T00:10:20.119 INFO bellperson::groth16::prover > ZQ: a_s start
2021-02-08T00:10:20.372 INFO bellperson::gpu::locks > GPU is available for FFT!
2021-02-08T00:10:21.291 INFO bellperson::gpu::fft > FFT: 1 working device(s) selected.
2021-02-08T00:10:21.291 INFO bellperson::gpu::fft > FFT: Device 0: GeForce RTX 3090
2021-02-08T00:10:21.291 INFO bellperson::domain > GPU FFT kernel instantiated!
2021-02-08T00:20:51.937 WARN bellperson::gpu::locks > GPU FFT failed! Falling back to CPU... Error: OpenCL Error: Ocl Error:
################################ OPENCL ERROR ###############################
Error executing function: clEnqueueWriteBuffer
Status error code: CL_MEM_OBJECT_ALLOCATION_FAILURE (-4)
Please visit the following url for more information:
https://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clEnqueueWriteBuffer.html#errors
#############################################################################
2021-02-08T00:24:33.141 INFO bellperson::groth16::prover > ZQ: a_s end: 853.022306209s
2021-02-08T00:24:33.141 INFO bellperson::groth16::prover > ZQ: h_s start
2021-02-08T00:24:33.142 INFO bellperson::gpu::locks > GPU is available for Multiexp!
2021-02-08T00:24:33.236 INFO bellperson::gpu::multiexp > Multiexp: 1 working device(s) selected. (CPU utilization: 0)
2021-02-08T00:24:33.236 INFO bellperson::gpu::multiexp > Multiexp: Device 0: GeForce RTX 3090 (Chunk-size: 67108864)
2021-02-08T00:24:33.236 INFO bellperson::multiexp > GPU Multiexp kernel instantiated!
2021-02-08T00:24:33.237 INFO bellperson::gpu::multiexp > GABEDEBUG: <G> size:104, <PrimeField> size:32, <Projective> size:144
2021-02-08T00:24:33.237 INFO bellperson::gpu::multiexp > GABEDEBUG: GPU mem need:21511413760byte, 20514Mbyte
2021-02-08T00:24:38.932 INFO bellperson::gpu::multiexp > GABEDEBUG: <G> size:104, <PrimeField> size:32, <Projective> size:144
2021-02-08T00:24:38.932 INFO bellperson::gpu::multiexp > GABEDEBUG: GPU mem need:21511413624byte, 20514Mbyte
2021-02-08T00:24:40.227 WARN bellperson::gpu::locks > GPU Multiexp failed! Falling back to CPU... Error: OpenCL Error: Ocl Error:
################################ OPENCL ERROR ###############################
Error executing function: clEnqueueWriteBuffer
Status error code: CL_MEM_OBJECT_ALLOCATION_FAILURE (-4)
Please visit the following url for more information:
https://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clEnqueueWriteBuffer.html#errors
#############################################################################
****)
On official code, the same server can finish 2 C2 tasks in 40 minutes.
On the modified code here, it throws CL_MEM_OBJECT_ALLOCATION_FAILURE.