Open telegraphic opened 1 year ago
Patch coverage: 9.96
% and project coverage change: -2.76
:warning:
Comparison is base (
e9c34da
) 67.04% compared to head (78de6d4
) 64.28%.
:exclamation: Your organization is not using the GitHub App Integration. As a result you may experience degraded service beginning May 15th. Please install the Github App Integration for your organization. Read more.
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Do you have feedback about the report comment? Let us know in this issue.
Tensor core arithmetic also accelerates 16 bit type data.
On 2/14/23 9:49 AM, jaycedowell wrote:
In python/bifrost/blocks/psrdada.py https://github.com/ledatelescope/bifrost/pull/194#discussion_r1105904988:
- Returns:
- s (str): DADA header string with padding to hdrlen
- """
- s = "HDR_VERSION 1.0\n"
- s+= "HDR_SIZE %i\n" % hdrlen
- keys_to_skip = ('HDR_VERSION', 'HDR_SIZE')
update parameters from bifrost tensor
- if '_tensor' in hdr_dict.keys():
- print(hdr_dict['_tensor'])
- dtype = hdr_dict['_tensor']['dtype']
- dtype_vals = {
- 'cf32': { 'NBIT': '32', 'NDIM': '2' },
- 'f32': { 'NBIT': '32', 'NDIM': '1' },
- 'ci8': { 'NBIT': '8', 'NDIM': '2' },
- 'i8': { 'NBIT': '8', 'NDIM': '1' } }
Are additional data types expected/allowed?
I raised the notion of 16-bit to accommodate directly time-series output from many-bit ADCs. i16 would do that. Quick examination suggests that uint16 is recognized by CUDA7(?) but using it may not result in speed advantage over 4-byte integer math. Proper research required.
On 2/15/23 1:58 AM, Danny Price wrote:
@.**** commented on this pull request.
In python/bifrost/blocks/psrdada.py https://github.com/ledatelescope/bifrost/pull/194#discussion_r1106722410:
- Returns:
- s (str): DADA header string with padding to hdrlen
- """
- s = "HDR_VERSION 1.0\n"
- s+= "HDR_SIZE %i\n" % hdrlen
- keys_to_skip = ('HDR_VERSION', 'HDR_SIZE')
update parameters from bifrost tensor
- if '_tensor' in hdr_dict.keys():
- print(hdr_dict['_tensor'])
- dtype = hdr_dict['_tensor']['dtype']
- dtype_vals = {
- 'cf32': { 'NBIT': '32', 'NDIM': '2' },
- 'f32': { 'NBIT': '32', 'NDIM': '1' },
- 'ci8': { 'NBIT': '8', 'NDIM': '2' },
- 'i8': { 'NBIT': '8', 'NDIM': '1' } }
There isn't really a strict definition AFAIK. There is no header keyword for float/integer, but I think it's reasonable to add cf64/f64. Would need to decide between i16/ci16 and f16/cf16 though
— Reply to this email directly, view it on GitHub https://github.com/ledatelescope/bifrost/pull/194#discussion_r1106722410, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACL54MQIQ7KCUU3K6EQ33PDWXR5CFANCNFSM6AAAAAAS7I5RJ4. You are receiving this because you commented.Message ID: @.***>
These updates are from https://github.com/Molonglo/bifrost, which is used for UTMOST-2D.
An alternative to this merge is to move PSRDADA functionality into a plugin.