1 00:00:00,280 --> 00:00:02,400 Hi, and welcome to the Neil Ashton Podcast. 2 00:00:03,080 --> 00:00:05,920 In each episode, we explain some of the fascinating ways 3 00:00:05,920 --> 00:00:09,080 that science and engineering are changing the world around us. 4 00:00:09,800 --> 00:00:12,760 We talk to leading engineers from elite level sports like 5 00:00:12,840 --> 00:00:16,880 cycling and Formula One to some of the world's top academics to 6 00:00:16,880 --> 00:00:20,450 understand how fluid dynamics, machine learning and 7 00:00:20,450 --> 00:00:23,000 supercomputing are bringing in a new era of discovery. 8 00:00:23,960 --> 00:00:27,080 We also hear some of their life stories, their career advice, 9 00:00:27,640 --> 00:00:30,080 the lessons they've learned on the way that I hope will be 10 00:00:30,080 --> 00:00:33,800 helpful to you too. So sit back and enjoy this 11 00:00:33,800 --> 00:00:41,120 episode. Hi, and welcome back to the Neil 12 00:00:41,120 --> 00:00:44,360 Ashton Podcast. So today I'm speaking with 13 00:00:44,480 --> 00:00:49,520 Professor Mike Giles, somebody who has had an enormous 14 00:00:49,520 --> 00:00:55,428 influence on the CFD community, but more broadly in, in maths, 15 00:00:55,428 --> 00:01:01,680 in high performance computing. And someone that, as I say to to 16 00:01:01,680 --> 00:01:05,245 Mike at the beginning of the of the chat is his name often 17 00:01:05,245 --> 00:01:07,800 comes up when I speak to, to other people. 18 00:01:08,320 --> 00:01:11,760 And I was always really intrigued to, to speak to him, 19 00:01:12,440 --> 00:01:18,200 particularly because when I was at Oxford myself, he was always 20 00:01:18,200 --> 00:01:21,120 a figure that was, that was mentioned and was influential in 21 00:01:21,120 --> 00:01:26,400 so many bits. Probably for most people, he is 22 00:01:26,720 --> 00:01:32,040 most known as the, I guess you'd call it the lead developer 23 00:01:32,040 --> 00:01:36,540 instigator of the Hydra Rolls-Royce CFD code that is 24 00:01:36,540 --> 00:01:41,913 used still used today by I would assume, thousands of engineers 25 00:01:41,913 --> 00:01:46,037 around the world to design the the jet engines that they 26 00:01:46,037 --> 00:01:50,259 produce. And, and I think what makes him 27 00:01:50,259 --> 00:01:57,467 really interesting is his pivot also to then work in finance, 28 00:01:57,475 --> 00:02:04,275 computational maths as applied to, to, to, to finance, where 29 00:02:04,275 --> 00:02:07,674 he has equally made an impression. 30 00:02:07,682 --> 00:02:12,464 And that's what's incredible for someone to make an impression in 31 00:02:12,464 --> 00:02:17,395 two quite different fields, CFD and and finance being quite 32 00:02:17,395 --> 00:02:22,088 different shows the level of the person, the intellect, the 33 00:02:22,088 --> 00:02:26,656 the sort of innovation potential to switch to a 34 00:02:26,656 --> 00:02:29,040 different field and then still make an impact. 35 00:02:29,760 --> 00:02:34,160 Now I only come from the first field, so my appreciation of the 36 00:02:34,160 --> 00:02:36,920 second in the sort of quant side of the world. 37 00:02:36,920 --> 00:02:41,200 But I've read enough to know and you only have to look at do a 38 00:02:41,200 --> 00:02:47,080 quick Google search to see that he's his work on what's called 39 00:02:47,080 --> 00:02:52,280 that multilevel Monte Carlo methods has been very impactful 40 00:02:52,280 --> 00:02:57,040 in that community. And the third thing that I guess 41 00:02:57,040 --> 00:03:01,052 he has really pioneered in some ways, and one one was one of 42 00:03:01,052 --> 00:03:04,276 the early adopters for is on the high, high performance 43 00:03:04,276 --> 00:03:07,120 computing side and particularly around GPUs. 44 00:03:08,280 --> 00:03:12,800 I should state obviously for transparency, I do now work at 45 00:03:12,800 --> 00:03:17,720 that company NVIDIA, but this conversation was in no way 46 00:03:17,720 --> 00:03:22,120 arranged to to promote NVIDIA. This was organized completely 47 00:03:22,120 --> 00:03:25,178 separately. And it just happens to be that I 48 00:03:25,178 --> 00:03:28,520 work there. So you'll hear some mentions of 49 00:03:28,520 --> 00:03:31,560 it, but please trust me, this is not some sort of, you know, 50 00:03:32,000 --> 00:03:35,440 product placement. And and he was actually working 51 00:03:35,440 --> 00:03:38,640 on this and I think he said that he was maybe the number 2 or 52 00:03:38,640 --> 00:03:42,000 like the second ever CUDA fellow and was looking at this in 53 00:03:42,000 --> 00:03:46,200 2006, 2007. You know, well before sort of 54 00:03:46,200 --> 00:03:49,280 everybody knows the name of NVIDIA and GPUs. 55 00:03:49,280 --> 00:03:53,480 So he, he is always, and he says in the discussion, had an 56 00:03:53,480 --> 00:03:56,880 interest in high performance computers throughout the time 57 00:03:57,120 --> 00:04:01,680 and, and still today teaches classes and, and programs and 58 00:04:01,680 --> 00:04:04,160 does everything, which is, I really love that when you see 59 00:04:04,160 --> 00:04:06,752 someone who's gone through their whole career and it's 60 00:04:06,752 --> 00:04:08,020 completely fine if you do change. 61 00:04:08,028 --> 00:04:10,360 And some people as they progress, become more senior, 62 00:04:10,368 --> 00:04:13,855 they get less hands on, you know, and then they're more 63 00:04:13,863 --> 00:04:16,259 about enacting a vision for what they want to do. 64 00:04:16,267 --> 00:04:17,440 And there's nothing wrong with that. 65 00:04:17,447 --> 00:04:21,146 But I always have a special appreciation for, for people who 66 00:04:21,146 --> 00:04:25,384 are, you know, one of the world's leading professors and 67 00:04:25,392 --> 00:04:29,254 they're still hands-on at the keyboard. 68 00:04:29,262 --> 00:04:35,106 So yeah. Mike is a professor of numerical 69 00:04:35,106 --> 00:04:38,680 analysis at the Maths Institute at the University of 70 00:04:38,680 --> 00:04:40,280 Oxford. As we mentioned, beautiful 71 00:04:40,280 --> 00:04:42,320 building, lovely location. I'm very jealous. 72 00:04:44,000 --> 00:04:48,880 And he was at the University of Cambridge, where he read as an 73 00:04:48,880 --> 00:04:51,000 undergraduate. And we mentioned it, he was a 74 00:04:51,000 --> 00:04:53,960 senior Wrangler, which if you look it up on Wikipedia, 75 00:04:53,960 --> 00:04:57,920 basically means the person who graduated top of class for the 76 00:04:57,920 --> 00:05:00,600 whole university in terms of maths as an undergraduate. 77 00:05:01,320 --> 00:05:04,320 And then he went to MIT, he was Kennedy Scholar, taught there, 78 00:05:04,680 --> 00:05:07,120 came back to Oxford. Now we go through all of that, 79 00:05:07,680 --> 00:05:11,200 but one of the things that's probably worth mentioning and 80 00:05:11,200 --> 00:05:15,200 congratulating him on is that actually very recently he was 81 00:05:15,200 --> 00:05:18,280 elected a fellow of the Royal Society, which is one of the 82 00:05:18,280 --> 00:05:21,120 highest honours that can be bestowed on somebody. 83 00:05:21,720 --> 00:05:25,880 So he's had an amazing career, still has an amazing career. 84 00:05:26,200 --> 00:05:30,160 And like any of these episodes, when I talk to someone like 85 00:05:30,160 --> 00:05:33,440 that, there are so many things that I realized I didn't ask him 86 00:05:33,440 --> 00:05:36,920 after I finished the episode. And I don't think we would have 87 00:05:36,920 --> 00:05:39,360 time anyway to, to to go through stuff. 88 00:05:40,280 --> 00:05:43,920 I will just note that if you Google his name and go to his 89 00:05:43,920 --> 00:05:47,133 personal website, I'll put it in the chat for the YouTube side 90 00:05:47,133 --> 00:05:49,720 of things. He has a great link to lots of 91 00:05:49,720 --> 00:05:52,120 presentations courses that he's done. 92 00:05:52,840 --> 00:05:57,854 So I would definitely look there to, yeah, to find out 93 00:05:57,854 --> 00:06:00,000 more. But yeah, I really hope you 94 00:06:00,000 --> 00:06:02,440 enjoy this conversation. I genuinely did. 95 00:06:02,520 --> 00:06:04,720 I hope you can see it from my face if you're watching it. 96 00:06:04,720 --> 00:06:08,240 I was learning and, you know, interested throughout the whole 97 00:06:08,480 --> 00:06:11,160 2 hours. So yeah, please sit back and 98 00:06:11,160 --> 00:06:13,720 enjoy this episode with Professor Mike Giles. 99 00:06:14,480 --> 00:06:16,440 Thank you very much. I really appreciate it. 100 00:06:16,720 --> 00:06:20,192 As I said just before your name comes up a lot in people I 101 00:06:20,192 --> 00:06:25,480 speak to in the in the CFD and HPC world. 102 00:06:25,480 --> 00:06:29,520 And so I was, yeah, really wanting to speak to you to find 103 00:06:29,520 --> 00:06:32,520 out a little bit more how this, you know, connections started 104 00:06:32,520 --> 00:06:38,880 out. But maybe we could start, you 105 00:06:38,880 --> 00:06:41,720 know, I guess towards towards the beginning. 106 00:06:41,720 --> 00:06:44,480 Now you're a professor at, you know, one of the top 107 00:06:44,480 --> 00:06:46,960 universities in the in the in the world and maths department, 108 00:06:46,960 --> 00:06:48,640 but obviously with an engineering background. 109 00:06:50,520 --> 00:06:54,240 Did you always want? Was your interest in maths? 110 00:06:54,360 --> 00:06:56,440 Engineering? Were you a person with planes 111 00:06:56,440 --> 00:06:58,560 and cars? More reading? 112 00:06:58,560 --> 00:07:01,040 Physics What? What was your early days like? 113 00:07:02,680 --> 00:07:07,320 So I would say in school my interests were maths and 114 00:07:07,320 --> 00:07:14,200 physics. And so I actually, you know, 115 00:07:14,400 --> 00:07:19,160 went to Cambridge to study maths, intending after the first 116 00:07:19,160 --> 00:07:22,080 year to transfer into theoretical physics. 117 00:07:24,200 --> 00:07:28,240 But then, you know, enjoyed the maths, stayed, stayed in the 118 00:07:28,240 --> 00:07:32,080 maths. So, so as a child, that's kind 119 00:07:32,080 --> 00:07:37,280 of the direction I suppose I saw myself going in, but I also 120 00:07:37,280 --> 00:07:41,720 wasn't sure, you know, what I would want to do after my 121 00:07:41,720 --> 00:07:45,174 studies. And so in terms of CFD. 122 00:07:45,174 --> 00:07:51,914 I mean really the, the pivotal thing for me was the fact that 123 00:07:51,914 --> 00:07:56,266 in in going to Cambridge in those days, you did the 124 00:07:56,266 --> 00:08:00,753 entrance exam in November, December and then you had from 125 00:08:00,753 --> 00:08:04,028 January to October to do something else. 126 00:08:04,036 --> 00:08:09,769 And some people have travelled then in my case, I went and 127 00:08:09,777 --> 00:08:12,806 worked at Rolls-Royce for that period. 128 00:08:12,814 --> 00:08:18,732 And that, that was really to learn well, well, to see what 129 00:08:18,740 --> 00:08:21,984 engineering was like, see if that's something which 130 00:08:21,992 --> 00:08:27,680 interested me. So at age 18, I was a 131 00:08:27,680 --> 00:08:29,589 Rolls-Royce undergraduate engineering apprentice. 132 00:08:29,597 --> 00:08:34,940 That was my, my job title. And normally they wouldn't, 133 00:08:34,948 --> 00:08:37,756 wouldn't have taken a mathematician on, you know, 134 00:08:37,764 --> 00:08:42,240 usually it, it was engineers, maybe people in materials. 135 00:08:42,248 --> 00:08:46,010 But I was actually third generation Rolls-Royce. 136 00:08:46,018 --> 00:08:49,912 My, my mother was a programmer before I was born. 137 00:08:49,920 --> 00:08:52,876 She was a programmer with Rolls Royce in Derby. 138 00:08:52,884 --> 00:08:57,024 And my grandfather worked, worked for the company for more 139 00:08:57,024 --> 00:09:03,325 than 25 years in Glasgow. So back in those days, you know, 140 00:09:03,325 --> 00:09:06,864 the application forms asked if you had any family members 141 00:09:06,872 --> 00:09:09,946 working in the company. I mean, these days that, that 142 00:09:09,954 --> 00:09:12,392 that would be nepotism that's strictly forbidden. 143 00:09:12,400 --> 00:09:17,226 But back in those days that that was viewed as a positive thing. 144 00:09:17,226 --> 00:09:19,960 And so I think because of that, 145 00:09:19,960 --> 00:09:26,520 they, they, they took me on as an apprentice because I was a 146 00:09:26,520 --> 00:09:28,720 mathematician. They weren't exactly sure what 147 00:09:28,720 --> 00:09:31,760 to do with me. So, so I went through a lot of 148 00:09:31,760 --> 00:09:35,936 the standard training with, with the engineers then, whereas 149 00:09:35,936 --> 00:09:39,528 the engineers had to be moved around different parts of the 150 00:09:39,528 --> 00:09:43,640 company to satisfy, you know, requirements for chartered 151 00:09:43,640 --> 00:09:47,280 engineer status. You know, later on, you know, I 152 00:09:47,280 --> 00:09:50,760 had, I had more flexibility as, as to what I did. 153 00:09:50,760 --> 00:09:55,040 And so I got into various assignments which involved 154 00:09:55,040 --> 00:09:58,220 programming in in various forms. 155 00:09:58,228 --> 00:10:03,420 Oh wow, that's interesting. So this is before you did your 156 00:10:03,428 --> 00:10:06,412 undergraduate? Yes, this is the period after 157 00:10:06,420 --> 00:10:09,320 school. Before undergraduate and then 158 00:10:09,328 --> 00:10:13,100 each summer while I was at Cambridge, each summer I went 159 00:10:13,108 --> 00:10:16,676 back to Rolls-Royce for, for another two months. 160 00:10:16,684 --> 00:10:24,509 So it was after my second year. So that would be 1980 that I 161 00:10:24,517 --> 00:10:29,700 joined what was essentially the CFD group, you know, only a few 162 00:10:29,700 --> 00:10:33,600 months after it was first created, you know, so it was 163 00:10:33,608 --> 00:10:35,680 called the Theoretical Sciences Group. 164 00:10:35,688 --> 00:10:44,134 And that that summer I was doing 2D grid generation using 165 00:10:44,142 --> 00:10:50,200 conformal mapping. So directly using, using my, my, 166 00:10:50,200 --> 00:10:53,360 my coursework in complex variable theory. 167 00:10:53,368 --> 00:10:58,991 That was actually a paper, I think it was written by Bob Ni 168 00:10:58,999 --> 00:11:02,100 at Pratt and Whitney. I don't know whether you 169 00:11:02,108 --> 00:11:06,542 recognise that, that name, but Bob's a really, you know, senior 170 00:11:06,542 --> 00:11:09,180 figure of that of that era. Oh. 171 00:11:09,188 --> 00:11:11,328 Wow. OK, so you did. 172 00:11:11,336 --> 00:11:15,400 So you're going between, but in the during your undergraduate, 173 00:11:15,408 --> 00:11:21,732 did you already therefore get a sense that you wanted to go down 174 00:11:21,732 --> 00:11:25,648 the more CFD route because you were doing in Rolls-Royce? 175 00:11:25,656 --> 00:11:29,060 I imagine because you were doing maths you didn't touch. 176 00:11:29,068 --> 00:11:32,550 Well, I guess would you do fluid dynamics in? 177 00:11:32,558 --> 00:11:35,864 I guess you would do numerical methods, but would would fluids 178 00:11:35,864 --> 00:11:37,543 come up? So. 179 00:11:37,551 --> 00:11:41,410 So we did lots of theoretical fluid dynamics. 180 00:11:41,418 --> 00:11:46,746 So I can remember thinking, you know, on on the one hand here, 181 00:11:46,754 --> 00:11:50,685 here I am taking a course learning about inviscid 182 00:11:50,693 --> 00:11:54,280 incompressible 2D flow over a cylinder. 183 00:11:54,288 --> 00:12:00,664 And on the other hand, here I am at Rolls-Royce looking at these 184 00:12:00,664 --> 00:12:05,504 phenomenally complex, you know, engineering devices. 185 00:12:05,512 --> 00:12:11,730 You're clearly paper and pencil cannot take you very far. 186 00:12:11,738 --> 00:12:16,960 So I was really sold on the idea of numerical simulation at a 187 00:12:16,968 --> 00:12:20,578 very early age. You know, a very, very early 188 00:12:20,586 --> 00:12:24,564 stage. I mean, already by the time I 189 00:12:24,572 --> 00:12:27,908 went to university I'd done some amount of programming. 190 00:12:27,916 --> 00:12:33,214 So I got into programming pretty early in, in, in part through, 191 00:12:33,222 --> 00:12:37,317 through my mother. I mean she, she was IT support 192 00:12:37,325 --> 00:12:41,862 at University of Stirling. So in the holidays I actually 193 00:12:41,870 --> 00:12:46,536 did a little bit of programming on one of the big academic 194 00:12:46,544 --> 00:12:53,418 systems down in Manchester. Where was that in the big 195 00:12:53,426 --> 00:12:58,085 building? When was that in Manchester? 196 00:12:58,093 --> 00:13:02,550 I mean, I, I didn't go there. I mean, this was remote access 197 00:13:02,558 --> 00:13:05,040 from Stirling. Yeah. 198 00:13:05,048 --> 00:13:08,400 So this was in, in, in the sense, you know, the equivalent 199 00:13:08,400 --> 00:13:10,760 of what's now the Edinburgh Parallel Computing Centre. 200 00:13:10,768 --> 00:13:13,948 Back in those days, it was the Manchester centre. 201 00:13:13,956 --> 00:13:18,730 I don't remember what I did on it. 202 00:13:18,738 --> 00:13:23,055 Nothing very significant. But also while at Rolls-Royce 203 00:13:23,055 --> 00:13:26,989 for those eight months before going to Cambridge one one day 204 00:13:26,989 --> 00:13:31,314 a week, they sent us along to what was then called Derby 205 00:13:31,314 --> 00:13:35,970 Tech to sort of keep our, our brains ticking over doing 206 00:13:35,978 --> 00:13:37,740 various classes, including programming. 207 00:13:37,748 --> 00:13:45,480 And so I, I, I worked in an IBM system there, I, I wrote a code 208 00:13:45,480 --> 00:13:50,720 to do project critical path analysis, which was great fun. 209 00:13:50,728 --> 00:13:53,920 So so I've always enjoyed programming. 210 00:13:53,928 --> 00:13:55,399 What? What languages would it be? 211 00:13:55,407 --> 00:13:57,550 I'm sorry if I'm asking a stupid question, I'm trying. 212 00:13:57,558 --> 00:14:00,077 To. Get back to like. 213 00:14:00,085 --> 00:14:04,168 So I think that must have been Fortran. 214 00:14:04,176 --> 00:14:09,482 It was punch cards. So that was my experience with, 215 00:14:09,482 --> 00:14:13,640 with, with punch cards that that time at Derby Tech, it was, 216 00:14:13,640 --> 00:14:16,940 it was a cast off the IBM machine from Rolls-Royce. 217 00:14:16,948 --> 00:14:19,840 They, they, they donated it to Derby Tech. 218 00:14:19,848 --> 00:14:22,755 That's, that's my recollection anyway. 219 00:14:22,763 --> 00:14:29,789 So yes, I think that must have been Fortran in, in Cambridge, 220 00:14:29,797 --> 00:14:36,238 we had the whole teaching lab of desktop machines and that was 221 00:14:36,246 --> 00:14:41,899 basic, I think that we used. So there was a, there was a 222 00:14:41,907 --> 00:14:45,502 numerical projects course in the third year, which was very good, 223 00:14:45,502 --> 00:14:49,520 you know, and so that, that really sort of solidified my, 224 00:14:49,528 --> 00:14:53,720 my, my interest in, in, in computational methods. 225 00:14:53,728 --> 00:15:02,669 And then what about so I, I, I have to ask you one, one 226 00:15:02,677 --> 00:15:04,549 question. When I was doing a bit of 227 00:15:04,557 --> 00:15:08,512 research, I hadn't come across this term before, but it's quite 228 00:15:08,512 --> 00:15:11,728 an esteemed Senior Wrangler. Am I pronouncing it? 229 00:15:11,736 --> 00:15:14,189 Correctly. I guess you've looked at the 230 00:15:14,197 --> 00:15:17,563 Wikipedia. And then I was like, then I 231 00:15:17,571 --> 00:15:21,144 started to look down and I read fantastic stories about people 232 00:15:21,152 --> 00:15:24,005 being paraded around. I don't know if that was the 233 00:15:24,013 --> 00:15:26,780 time. So this was the top 234 00:15:26,788 --> 00:15:29,760 undergraduate of maths, is that correct? 235 00:15:29,768 --> 00:15:34,681 Which which you you've got. Yeah, there there was no fanfare 236 00:15:34,681 --> 00:15:41,882 in my case. So yeah, I was just told 237 00:15:41,890 --> 00:15:46,960 afterwards by by my tutor, yeah. OK, well, it's it's still it 238 00:15:46,968 --> 00:15:49,950 shows, I guess your your ability. 239 00:15:49,958 --> 00:15:57,280 Were you, did you enjoy the, the sort of Cambridge life, the 240 00:15:57,288 --> 00:15:59,824 collegiate life? Was, was that something that 241 00:15:59,832 --> 00:16:03,742 you, I mean, now obviously you know, you're at the other place, 242 00:16:03,742 --> 00:16:07,280 but was that something that you made a, a strong impression 243 00:16:07,280 --> 00:16:10,995 on you and, and sort of motivated you later on to, to 244 00:16:10,995 --> 00:16:14,374 ultimately stay in or, or go back to academia? 245 00:16:14,382 --> 00:16:19,490 I mean, I, I enjoyed my time at Cambridge. 246 00:16:19,498 --> 00:16:25,840 I spent a lot of time doing orienteering while I was there. 247 00:16:25,840 --> 00:16:27,760 So orienteering was something I 248 00:16:27,760 --> 00:16:31,800 did as a child, you know, from the age of about 12, you know, 249 00:16:31,800 --> 00:16:34,280 for. Yeah, for, for those listening 250 00:16:34,280 --> 00:16:36,880 to this who don't know about orienteering, it's effectively 251 00:16:37,160 --> 00:16:41,560 cross country running, using a map to guide yourself through 252 00:16:41,560 --> 00:16:43,720 forests and over hillsides and stuff. 253 00:16:45,400 --> 00:16:50,720 So, yeah, a lot of my time at Cambridge was spent going off 254 00:16:51,280 --> 00:16:59,720 orienteering at weekends. I would say that I was as a 255 00:16:59,720 --> 00:17:03,040 diligent student, but I wasn't particularly hard working. 256 00:17:03,040 --> 00:17:04,640 Let's let's let's put it that way. 257 00:17:04,880 --> 00:17:10,839 You know, I seem to remember my, my maths as being sort of a, a 258 00:17:10,839 --> 00:17:14,079 nine to five activity five days a week. 259 00:17:14,079 --> 00:17:15,880 And then, you know, weekends I was away. 260 00:17:17,319 --> 00:17:20,118 I didn't really start working until I went to MIT. 261 00:17:20,720 --> 00:17:25,118 OK, that, that that sort of sums up the difference between MIT 262 00:17:25,118 --> 00:17:28,680 and Cambridge also, you know, grad student life and undergrad 263 00:17:28,680 --> 00:17:30,920 life, I guess. Yeah, Yeah. 264 00:17:30,920 --> 00:17:34,960 So you said the beginning that you'd, you weren't sure. 265 00:17:34,960 --> 00:17:38,040 You, you were debating around theoretical physics, maths I 266 00:17:38,040 --> 00:17:40,560 presume. Then as you went towards the end 267 00:17:40,560 --> 00:17:45,400 of your undergrad and particularly with the summers in 268 00:17:45,400 --> 00:17:49,360 Rolls-Royce, you'd, you'd put aside the theoretical physics 269 00:17:49,480 --> 00:17:53,360 and you were more moving towards, I guess the engineering 270 00:17:53,360 --> 00:17:55,760 or applied mathematics. Would that be fair? 271 00:17:56,240 --> 00:18:00,120 I mean, I think by the time I finished that initial 8 months 272 00:18:00,120 --> 00:18:07,527 at Rolls-Royce, I think probably the physics ideas had to a 273 00:18:07,527 --> 00:18:13,160 large extent dropped out. Although I think equally I knew 274 00:18:13,160 --> 00:18:15,200 I didn't want to work in industry. 275 00:18:15,760 --> 00:18:19,320 So yeah. So maybe things were still 276 00:18:19,320 --> 00:18:28,400 somewhat open. Yes, I I do remember, I'm trying 277 00:18:28,400 --> 00:18:31,462 to think whether it would be end of my second year or early in 278 00:18:31,462 --> 00:18:37,000 my third year. There was a talk in college by 279 00:18:38,600 --> 00:18:43,120 an academic researcher just talking about the joy of being 280 00:18:43,120 --> 00:18:47,640 an academic researcher and that that did strike me. 281 00:18:48,200 --> 00:18:54,440 So, so that's one of those moments that I think, you know, 282 00:18:54,480 --> 00:18:59,960 confirmed me in going as an academic path or, or at least 283 00:18:59,960 --> 00:19:02,800 taking it further and doing, doing a PhD. 284 00:19:06,720 --> 00:19:13,560 But you know, the move to MIT, my, my tutor told me about the, 285 00:19:13,560 --> 00:19:18,240 the Kennedy scholarship scheme and encouraged me to apply, you 286 00:19:18,240 --> 00:19:22,000 know, and so, you know, initially I went to MIT on, on 287 00:19:22,000 --> 00:19:26,800 this one year scholarship thinking it would be a chance to 288 00:19:26,800 --> 00:19:30,680 see the world. And then I would think about 289 00:19:31,040 --> 00:19:38,200 what to do next. Having gone out to MIT, my, my 290 00:19:38,200 --> 00:19:43,080 supervisor there found funding for my second year to, to finish 291 00:19:43,080 --> 00:19:48,600 up the masters. And in doing that, I also came 292 00:19:48,600 --> 00:19:53,200 up with a, what turned out to be a good idea for a PhD project. 293 00:19:53,640 --> 00:19:57,640 And then, then he got more, more research funding for, for me 294 00:19:59,080 --> 00:20:03,640 actually from US Air Force to, to carry on and do do the PhD. 295 00:20:04,200 --> 00:20:07,480 So it wasn't. It wasn't the master plan by any 296 00:20:07,480 --> 00:20:13,040 means, but yeah, yeah, I went out initially for one year and 297 00:20:13,040 --> 00:20:17,360 ended up staying for 11. So what was the Kennedy 298 00:20:17,360 --> 00:20:18,920 scholarship like? Because I spoke to somebody 299 00:20:18,920 --> 00:20:21,880 else, Anthony. Oh, sorry, I'm confusing. 300 00:20:24,120 --> 00:20:29,240 Oh, got a brain fog. Now the have to rise. 301 00:20:31,000 --> 00:20:36,720 Now I'm just interested because I wonder if you'd actually come 302 00:20:36,960 --> 00:20:39,480 across. Probably you haven't. 303 00:20:40,640 --> 00:20:44,440 My my memory for names is terrible, but there wasn't 304 00:20:44,440 --> 00:20:46,880 Anthony about my time. Yeah, that's why I'm. 305 00:20:47,600 --> 00:20:51,280 Just. Wondering if you've in In 306 00:20:51,280 --> 00:20:54,640 general, most of the Kennedy scholars were at Harvard. 307 00:20:54,640 --> 00:21:00,040 There were very few at MIT and, and, and these days it's really 308 00:21:00,040 --> 00:21:03,920 quite the exception to have anybody at MIT, which is a bit 309 00:21:03,920 --> 00:21:06,640 of a shame I think. Yeah, Tony Purnell. 310 00:21:07,320 --> 00:21:13,200 Did you ever come across Tony? Yeah, that that that name does 311 00:21:13,200 --> 00:21:14,560 sound familiar. So what? 312 00:21:14,600 --> 00:21:15,880 What? What's he doing now? 313 00:21:16,280 --> 00:21:19,800 Yeah, So Tony, who I actually also interviewed, he's a really 314 00:21:19,920 --> 00:21:23,480 great guy. So he he won the Kennedy 315 00:21:23,480 --> 00:21:30,440 scholarship, went to MIT and I think it was in 80, mid 80s, so 316 00:21:30,440 --> 00:21:36,280 similar time. And then he went to he then went 317 00:21:36,280 --> 00:21:37,960 to work. He created his own company, but 318 00:21:37,960 --> 00:21:41,240 he ultimately ended up in Formula One running the what is 319 00:21:41,240 --> 00:21:43,280 now the Red Bull team. But he's now a professor at 320 00:21:43,280 --> 00:21:46,600 Cambridge. Does all the aerodynamics and 321 00:21:46,600 --> 00:21:49,276 CFD. I wonder whether? 322 00:21:49,284 --> 00:21:54,303 Yeah, so, yeah. So I was a Kennedy Scholar 81 to 323 00:21:54,303 --> 00:22:00,606 82. Yeah, and the name does sound 324 00:22:00,614 --> 00:22:03,140 familiar. Basically, we we didn't really 325 00:22:03,148 --> 00:22:07,636 hang out with each other. I didn't hang out with the other 326 00:22:07,636 --> 00:22:12,227 Kennedy scholars. So yes, yes. 327 00:22:12,235 --> 00:22:17,934 I didn't really have have have those connections, but it was a 328 00:22:17,934 --> 00:22:20,460 part time experience. Oh, oh, yes, yeah, yeah. 329 00:22:20,468 --> 00:22:24,549 So, so, you know, it was Kennedy scholarship that took me over to 330 00:22:24,549 --> 00:22:27,075 MIT. And as it happened, and this 331 00:22:27,083 --> 00:22:31,500 this was really sort of coincidence, my supervisor at 332 00:22:31,508 --> 00:22:38,910 MIT had Rolls-Royce funding. This was in the days when 333 00:22:38,910 --> 00:22:43,942 Rolls-Royce was starting to sell engines to the US Marine Corps 334 00:22:43,950 --> 00:22:49,094 for the Harrier, and they wanted to be viewed as more of an 335 00:22:49,102 --> 00:22:54,120 international company. And so as part of that, I think 336 00:22:54,120 --> 00:22:57,622 almost out of their marketing budget, maybe there was a whole 337 00:22:57,622 --> 00:23:00,772 pile of research funding to be spent to MIT. 338 00:23:00,780 --> 00:23:07,000 And and so some of this was was going to my supervisor, a guy 339 00:23:07,008 --> 00:23:14,676 by the name of Tilt Thompkins. So although my going to MIT and, 340 00:23:14,676 --> 00:23:18,660 you know, landing up with Tilt as a supervisor was completely 341 00:23:18,668 --> 00:23:22,680 independent of Rolls-Royce, there was still that sort of 342 00:23:22,688 --> 00:23:24,645 accidental background connection. 343 00:23:24,653 --> 00:23:35,600 So then my, my, my, my graduate history at, at MIT is a bit 344 00:23:35,608 --> 00:23:39,516 curious. So I did did the masters in 18 345 00:23:39,524 --> 00:23:44,328 months, which is a bit faster than normal, but not 346 00:23:44,336 --> 00:23:48,900 exceptional. I then did did my PhD in 2 1/2 347 00:23:48,908 --> 00:23:51,850 years, which is highly unusual for MIT. 348 00:23:51,858 --> 00:23:58,236 So, So what happened there was I was basically two years into my 349 00:23:58,236 --> 00:24:01,988 PhD and you know, it had gone very well. 350 00:24:01,996 --> 00:24:05,739 I mean, this was, you know, the project that me and Mark Drela 351 00:24:05,747 --> 00:24:10,150 did, you know, you know, the ISES code, you know, 2D airfoil 352 00:24:10,150 --> 00:24:15,102 design code. So the project was going ahead 353 00:24:15,110 --> 00:24:20,765 very well, but I got called into the head of department's office 354 00:24:20,765 --> 00:24:26,412 about two years into my PhD and told that my supervisor hadn't 355 00:24:26,420 --> 00:24:30,584 got tenure and would be leaving in six months time and would I 356 00:24:30,592 --> 00:24:35,200 like his job. Wow. 357 00:24:35,208 --> 00:24:41,940 So so I then had to finish up really quickly that that I 358 00:24:41,948 --> 00:24:43,107 wasn't. Expecting that. 359 00:24:43,115 --> 00:24:44,258 Wow. OK, so. 360 00:24:44,266 --> 00:24:48,503 Yeah, yes, that, that, that, that last six months was, yes, 361 00:24:48,511 --> 00:24:52,475 exhausting. So then, so your PhD. 362 00:24:52,483 --> 00:24:55,984 What was the? What was the end title of your 363 00:24:55,992 --> 00:25:00,020 thesis then for the? Something like two-dimensional 364 00:25:00,028 --> 00:25:03,000 transonic aerodynamic design method. 365 00:25:03,008 --> 00:25:09,640 So did you work with or be inspired, I guess by, by 366 00:25:09,648 --> 00:25:13,783 Professor Jameson? Was there, was there any sort of 367 00:25:13,783 --> 00:25:16,310 link into? I just, I always find it 368 00:25:16,318 --> 00:25:18,990 interesting when there's sort of Brits going over to the US. 369 00:25:18,998 --> 00:25:21,104 Yeah, yeah, they go and work on things. 370 00:25:21,112 --> 00:25:23,875 But it was independent. There was no he didn't have a. 371 00:25:23,883 --> 00:25:27,102 Connection. No, there was no connection at 372 00:25:27,110 --> 00:25:32,352 all there, although he, he, he was aware of me. 373 00:25:32,360 --> 00:25:39,628 He was aware of my master's thesis and actually told Tilt, 374 00:25:39,628 --> 00:25:44,078 like, you know, something along the lines of, you know, they 375 00:25:44,078 --> 00:25:48,528 should have given me a PhD for it, which it was flattering. 376 00:25:48,536 --> 00:25:55,960 But but no, my, my, my master's thesis was using some 377 00:25:55,968 --> 00:26:03,127 mathematics WKB analysis to understand some numerical wave 378 00:26:03,135 --> 00:26:07,888 propagation on grids. And what it shows you is that if 379 00:26:07,888 --> 00:26:11,224 you have something like the convection equation, if you have 380 00:26:11,224 --> 00:26:15,718 a poorly resolved wave, it can actually travel in the wrong 381 00:26:15,726 --> 00:26:20,147 direction and you can get these weird wave track wave trapping 382 00:26:20,155 --> 00:26:22,892 phenomenon. Anyway, it, it, it, it, it was 383 00:26:22,900 --> 00:26:24,289 something that intrigued Antony. 384 00:26:24,297 --> 00:26:31,240 And so, so he was aware of me already at at that point. 385 00:26:31,248 --> 00:26:38,918 I mean, I was obviously aware of him, but the, you know, the, the 386 00:26:38,918 --> 00:26:44,696 stream tube idea that I had, which was the basis of the ISES 387 00:26:44,696 --> 00:26:48,870 code, that was completely different to anything, you know, 388 00:26:48,870 --> 00:26:55,296 in, in, in CFD at that time, you know, worked beautifully in 389 00:26:55,296 --> 00:26:58,327 2D. It had had no natural 3D 390 00:26:58,327 --> 00:27:03,426 extension, but, but for, for 2D wing design it, it, it was 391 00:27:03,426 --> 00:27:06,477 really ideal. And so Mark, Mark Drela and I 392 00:27:06,485 --> 00:27:10,832 teamed up on that. And in fact, for a while we were 393 00:27:10,832 --> 00:27:14,882 going to do a joint thesis. We're actually going to write it 394 00:27:14,882 --> 00:27:18,116 up as a single document between the two of us. 395 00:27:18,124 --> 00:27:21,708 And our thesis committee was perfectly happy with this. 396 00:27:21,716 --> 00:27:25,205 And since our thesis committee included the head of department, 397 00:27:25,205 --> 00:27:27,623 I assumed that this was all fine. 398 00:27:27,631 --> 00:27:32,072 And then later on in the process, the central university 399 00:27:32,072 --> 00:27:37,130 somehow, you know, learnt about this and said there is no 400 00:27:37,138 --> 00:27:39,560 precedent for this, there shall be no precedent. 401 00:27:39,568 --> 00:27:44,438 You know, you're not allowed. And so very late in the process, 402 00:27:44,438 --> 00:27:48,282 basically, you know, Mark and I had to sort of carve our work 403 00:27:48,282 --> 00:27:51,880 up into two separate pieces so that we could write up two 404 00:27:51,880 --> 00:27:55,208 separate documents. You know, so in, in, in the end, 405 00:27:55,208 --> 00:27:58,600 I, I finished up early because I had to take over Tilt's 406 00:27:58,608 --> 00:28:01,099 position. Mark, Mark finished up a year 407 00:28:01,107 --> 00:28:04,579 later then, you know, he did, he got an academic position as 408 00:28:04,587 --> 00:28:06,532 well, you know, so we were both hired. 409 00:28:06,540 --> 00:28:09,620 I mean, there was never any question of one of us 410 00:28:09,628 --> 00:28:13,116 freeloading off the other one. I mean, our, our thesis 411 00:28:13,124 --> 00:28:15,948 committee were perfectly content on that point. 412 00:28:15,956 --> 00:28:22,162 So yes, so it's a curious situation, but it meant that I, 413 00:28:22,162 --> 00:28:28,690 I took over Tilt's office, a software engineer, almost all of 414 00:28:28,690 --> 00:28:32,800 his students and all of his research contracts, including 415 00:28:32,808 --> 00:28:34,685 the Rolls-Royce research contract. 416 00:28:34,693 --> 00:28:38,624 So, so at that point I was then back into the Rolls-Royce 417 00:28:38,632 --> 00:28:41,300 family. And how old were you then? 418 00:28:41,308 --> 00:28:43,340 You couldn't have been that old. 25. 419 00:28:43,348 --> 00:28:47,790 OK, that's quite, that's quite young then. 420 00:28:47,798 --> 00:28:53,975 In in the US system and particularly at MIT, they they 421 00:28:53,983 --> 00:29:00,704 do hire a lot of people straight from PhD into assistant 422 00:29:00,712 --> 00:29:05,102 professor positions because given the tenure system if they 423 00:29:05,102 --> 00:29:08,315 decide that they made a mistake, they just flush you out 424 00:29:08,315 --> 00:29:15,020 after seven years. Whereas whereas in, in the 425 00:29:15,028 --> 00:29:19,506 British system, you know, we, we, we, we like to see people 426 00:29:19,514 --> 00:29:23,160 get a good bit of experience under their belt, you know, 427 00:29:23,168 --> 00:29:26,248 before we'll, we'll, we'll hire them in, in Oxford. 428 00:29:26,256 --> 00:29:28,440 Yeah. Very occasionally we'll we'll 429 00:29:28,448 --> 00:29:33,740 take people straight from PhD, but it's very, very rare. 430 00:29:33,748 --> 00:29:36,805 Wow. So you're 25, you're an 431 00:29:36,813 --> 00:29:41,219 assistant professor at MIT, you've got a PhD, all that, and 432 00:29:41,219 --> 00:29:43,890 and now you're taking over the Rolls-Royce. 433 00:29:43,898 --> 00:29:47,263 So what what? How do things progress from from 434 00:29:47,263 --> 00:29:53,489 there? So looking back, actually the, 435 00:29:53,497 --> 00:29:59,089 the, the first thing is it took me about six months, I think, to 436 00:29:59,089 --> 00:30:03,855 recover from burnout from having finished up the PhD so 437 00:30:03,855 --> 00:30:08,240 quickly. But, you know, there were, there 438 00:30:08,480 --> 00:30:11,720 were a certain number of plans that were already in place that 439 00:30:11,760 --> 00:30:14,120 I sort of carried on supervising students. 440 00:30:14,120 --> 00:30:17,440 But I guess during that first six months, I was thinking 441 00:30:17,440 --> 00:30:20,727 about, you know, what, what was the first new thing I wanted 442 00:30:20,727 --> 00:30:24,160 to, to, to do with Rolls-Royce funding. 443 00:30:24,640 --> 00:30:30,080 And you know, I want, I wanted to do something different. 444 00:30:30,720 --> 00:30:34,400 I, I can't remember how much I talked to them to understand 445 00:30:34,400 --> 00:30:37,760 what their concerns were at the time. 446 00:30:37,760 --> 00:30:39,400 So that what, what, what they needed. 447 00:30:42,120 --> 00:30:48,520 But what I decided to do was to develop a 2D unsteady CFD code. 448 00:30:49,960 --> 00:30:57,480 I think that they had had some engineering challenges in I 449 00:30:57,480 --> 00:31:01,520 think it was a military engine where in military engines 450 00:31:01,520 --> 00:31:06,040 there's a smaller gap between the stators and rotors. 451 00:31:06,320 --> 00:31:09,080 So they're they're they're more closely coupled stages. 452 00:31:09,880 --> 00:31:13,480 And as that as such, that means you get a larger level of 453 00:31:13,480 --> 00:31:18,040 unsteady forcing on on on the blades. 454 00:31:18,320 --> 00:31:21,880 And I think they, that there had possibly been some engineering 455 00:31:23,040 --> 00:31:27,040 project where they had major difficulties with that and they 456 00:31:27,480 --> 00:31:31,760 really needed tools to analyse that. 457 00:31:34,720 --> 00:31:37,800 I mean, this, this was still, you know, fairly early days for, 458 00:31:37,800 --> 00:31:42,480 for, for CFD, you know, so, so I think at that point, you know, 459 00:31:42,480 --> 00:31:48,680 people like Bill Dawes and John Denton had developed steady 2D 460 00:31:48,680 --> 00:31:55,720 CFD codes, possibly even 3D, but but not anything unsteady. 461 00:31:56,160 --> 00:32:02,000 So, you know, so I, I, you know, my first code for Rolls-Royce 462 00:32:02,000 --> 00:32:07,080 was one called UNSFLO, which was 2D unsteady. 463 00:32:07,400 --> 00:32:11,520 Initially it was wake rotor interaction. 464 00:32:11,800 --> 00:32:14,760 So you were passing in the wakes through upstream boundary 465 00:32:14,760 --> 00:32:22,000 conditions and then going into doing stator rotor interaction. 466 00:32:22,240 --> 00:32:25,080 So you've actually got the moving blade rows, you know, 467 00:32:25,160 --> 00:32:30,240 moving relative to each other that that was initially 468 00:32:30,240 --> 00:32:33,800 inviscid. I later made it viscous. 469 00:32:35,520 --> 00:32:38,160 I can't remember now what I did for a turbulence model. 470 00:32:38,160 --> 00:32:41,960 It's probably an algebraic turbulence model in those in 471 00:32:41,960 --> 00:32:44,920 those days. Mid 80s, Yeah, yeah, I guess, 472 00:32:45,160 --> 00:32:50,480 yeah. So the the the sort of unique 473 00:32:51,280 --> 00:32:56,480 thing about UNSFLO was this thing called the time inclined 474 00:32:57,680 --> 00:33:03,400 plane. So one of the difficulties in in 475 00:33:03,400 --> 00:33:10,019 doing unsteady analysis in turbo machinery is the number of 476 00:33:10,019 --> 00:33:14,160 rotor blades is different to the number of stator blades. 477 00:33:15,520 --> 00:33:19,320 So you want to do a simulation that just has one blade passage, 478 00:33:20,480 --> 00:33:23,760 but if you do it the natural way, that doesn't work. 479 00:33:23,760 --> 00:33:27,080 You know, you don't have the right periodicity to, to to do 480 00:33:27,080 --> 00:33:30,720 that. So the time inclined plane 481 00:33:33,000 --> 00:33:38,880 involved, you know, usually you know, when you're at time level 482 00:33:38,880 --> 00:33:42,600 N, you know, all the grid points are at the same physical time. 483 00:33:43,160 --> 00:33:47,000 The time inclined plane, you know that, that, that that time 484 00:33:47,000 --> 00:33:52,480 was inclined. So that, and you could incline 485 00:33:52,480 --> 00:33:57,080 it in such a way that you then set up the right periodicity 486 00:33:57,080 --> 00:34:00,960 condition to cope with this arbitrary blade count. 487 00:34:02,320 --> 00:34:07,800 So yeah, that that, that that was kind of the the unique 488 00:34:07,800 --> 00:34:09,679 point. And then, you know, there's also 489 00:34:09,679 --> 00:34:13,080 some maths I did on non reflecting boundary conditions 490 00:34:13,440 --> 00:34:15,920 that when you're doing these unsteady interactions with the 491 00:34:15,920 --> 00:34:21,400 boundaries very close, you want the outgoing waves to go out to 492 00:34:21,400 --> 00:34:24,000 not be artificially reflected from the boundary. 493 00:34:24,600 --> 00:34:27,520 And so there's a whole piece of research on on, on that. 494 00:34:27,920 --> 00:34:30,639 Stuff you take for granted now in a commercial solver. 495 00:34:31,679 --> 00:34:35,840 Yeah, Yeah. So I mean really I was one of 496 00:34:35,840 --> 00:34:40,480 the first people doing that in, in the context of compressible 497 00:34:40,480 --> 00:34:45,760 flow CFD yes. I mean these, these these days 498 00:34:46,320 --> 00:34:49,480 you, you've got things like absorbing boundary methods, 499 00:34:49,480 --> 00:34:53,560 which is probably what you would use if you're doing far field 500 00:34:53,760 --> 00:34:55,215 acoustics and electromagnetics. 501 00:34:55,223 --> 00:35:01,045 But that actually wouldn't work well in in this context of 502 00:35:01,053 --> 00:35:05,002 closely coupled stages with very close in boundaries. 503 00:35:05,010 --> 00:35:08,860 And were you always hands on? Were you, you know, at that 504 00:35:08,868 --> 00:35:12,675 time, were you always that sort of person who was programming it 505 00:35:12,675 --> 00:35:15,372 yourself? You had students, but you were 506 00:35:15,380 --> 00:35:19,880 still very much yes and and sort of a hands on programmer. 507 00:35:19,888 --> 00:35:24,760 Yes, that was hands on. I mean generally the students 508 00:35:24,768 --> 00:35:28,829 were writing their own other codes. 509 00:35:28,837 --> 00:35:36,504 So, so with UNSFLO there was this software engineer Bob 510 00:35:36,512 --> 00:35:38,598 Haimes. So I don't know whether you 511 00:35:38,606 --> 00:35:43,068 recognise the name Bob Haimes. He, he he was responsible for 512 00:35:43,076 --> 00:35:45,170 developing our visualisation software. 513 00:35:45,178 --> 00:35:48,434 So VISUAL2, VISUAL3. Yeah. 514 00:35:48,442 --> 00:35:52,929 Name names from the past. Bob's Bob's still at, at MIT, I 515 00:35:52,929 --> 00:35:57,479 think he still hasn't retired. You know, he must be about 10 516 00:35:57,487 --> 00:36:01,580 years older than me, something like that. 517 00:36:01,588 --> 00:36:07,095 So, yeah. So I did most of the UNSFLO 518 00:36:07,103 --> 00:36:11,782 development, but Bob will have helped me with, with, with bits 519 00:36:11,782 --> 00:36:16,344 of that as well as doing, you know, he did, he certainly did 520 00:36:16,352 --> 00:36:19,292 all, all of the visualisation because if you're doing a, you 521 00:36:19,300 --> 00:36:23,260 know, 2D calculation, you know, you want to have some nice, nice 522 00:36:23,260 --> 00:36:25,583 visualisation. It's true then, true now, isn't 523 00:36:25,583 --> 00:36:29,360 it? So this was, this was in, in the 524 00:36:29,360 --> 00:36:34,010 early days of Silicon Graphics and there was a company called 525 00:36:34,018 --> 00:36:39,414 Stellar which was based just outside, well, in the Boston 526 00:36:39,422 --> 00:36:43,444 suburb. And I, I sort of got involved 527 00:36:43,452 --> 00:36:47,016 with, with them. And so we, we, we had a couple 528 00:36:47,024 --> 00:36:51,685 of their, their machines, beautiful machines, you know, 529 00:36:51,693 --> 00:36:56,720 multiple cores I recall as well, which was unusual. 530 00:36:56,728 --> 00:37:00,400 I mean, it was a very early days of parallel computing. 531 00:37:00,408 --> 00:37:06,250 And back then were you, you know, as much as you were 532 00:37:06,258 --> 00:37:09,982 developing the code for accuracy and the physics side, did you, 533 00:37:09,990 --> 00:37:13,801 did you still have a strong interest in the sort of high 534 00:37:13,809 --> 00:37:16,420 performance computing side? Were you were you always excited 535 00:37:16,420 --> 00:37:19,680 to try out different machines or have access to the machines 536 00:37:19,680 --> 00:37:23,720 or did that come later? No, I was always interested. 537 00:37:23,728 --> 00:37:29,867 So I mean, even even while I was doing, you know, masters and 538 00:37:29,875 --> 00:37:33,955 PhDs. So I think, oh, I'm trying to 539 00:37:33,963 --> 00:37:38,168 remember. So it was so probably soon after 540 00:37:38,168 --> 00:37:43,934 my master's, I spent the summer at NASA Langley at at at 541 00:37:43,934 --> 00:37:50,488 ICASE and down there I think I was doing programming on oh 542 00:37:50,488 --> 00:37:55,960 gosh, what, what what would it be? 543 00:38:00,320 --> 00:38:02,880 My mind's gone blank. There was there was Cray and 544 00:38:02,880 --> 00:38:04,200 then there was the other company. 545 00:38:05,880 --> 00:38:12,840 What was the other one called? So it was, it was before, you 546 00:38:12,840 --> 00:38:21,280 know, the ETA 10. Oh, Cyber 205. 547 00:38:21,640 --> 00:38:24,040 Yeah, that sounds, sounds maybe right. 548 00:38:25,400 --> 00:38:30,840 So I guess coming out to CDC maybe so. 549 00:38:31,080 --> 00:38:36,160 So I I had my first experience of supercomputing at at ICASE. 550 00:38:36,160 --> 00:38:38,676 I don't think I did a lot there. 551 00:38:38,684 --> 00:38:41,800 Who was there at that time? Because I've heard other people 552 00:38:41,800 --> 00:38:45,002 mention about this ICASE they don't do it anymore, I don't 553 00:38:45,010 --> 00:38:47,540 think, but it was this wasn't it. 554 00:38:47,548 --> 00:38:53,483 So the person who led it, I think when I was there was Milt 555 00:38:53,483 --> 00:38:57,766 Rose, but I think he may have retired not long after I was 556 00:38:57,774 --> 00:39:02,897 there. And then for many years it was 557 00:39:02,905 --> 00:39:06,898 led by somebody whose surname is Hussaini. 558 00:39:06,906 --> 00:39:10,000 I'm trying to remember what his first name is. 559 00:39:10,008 --> 00:39:15,572 Possibly Yousuff. No, I'm not sure. 560 00:39:15,580 --> 00:39:21,548 Yes, yes, ICASE. I mean, you know, there are lots 561 00:39:21,548 --> 00:39:27,504 of academics there. Who else do I remember? 562 00:39:27,512 --> 00:39:30,880 Eli Turkel. I remember he was, he was there, 563 00:39:30,880 --> 00:39:35,056 I think, you know, the summer that I was there. 564 00:39:35,064 --> 00:39:41,482 And then, you know, we sometimes saw the people in, in the CFD 565 00:39:41,490 --> 00:39:48,362 group there and oh, there was a, there was a great, great person 566 00:39:48,362 --> 00:39:54,398 who headed up the CFD group. Oh, gosh, my, my, my memory 567 00:39:54,406 --> 00:39:58,400 today is, is is poor. Yeah. 568 00:39:58,408 --> 00:40:02,010 So, yeah. So, yeah. 569 00:40:02,018 --> 00:40:10,960 So I had, I had experience with this, I think it was a Cyber 205 570 00:40:10,960 --> 00:40:16,744 at Langley. At MIT, I did a little bit of 571 00:40:16,744 --> 00:40:20,800 work on on the Thinking Machines CM-5, you know. 572 00:40:20,808 --> 00:40:26,525 That I remember at the time, you know, there was a lot of 573 00:40:26,533 --> 00:40:29,320 discussion about, you know, this is the future of massively 574 00:40:29,328 --> 00:40:32,820 parallel computing, you know so I think it was 64,000 575 00:40:32,828 --> 00:40:36,570 processors, but each of the processors was incredibly 576 00:40:36,578 --> 00:40:41,945 elementary. And then it got blown away by 577 00:40:41,953 --> 00:40:45,918 RISC computing, you know, and, and people putting together PC 578 00:40:45,926 --> 00:40:50,876 clusters, you know, so I think DARPA basically bankrolled it 579 00:40:50,884 --> 00:40:54,774 for maybe five years, eight years. 580 00:40:54,782 --> 00:40:59,095 But then, yeah, it wasn't capable of of sustaining itself. 581 00:40:59,095 --> 00:41:01,560 And then and then, you know, 582 00:41:01,560 --> 00:41:04,680 Cray, Cray really got established, you know, And so 583 00:41:05,000 --> 00:41:11,280 Cray, Cray was the winner. Yeah. 584 00:41:11,280 --> 00:41:15,240 I so Rolls-Royce at one stage had the Cray. 585 00:41:15,520 --> 00:41:20,160 So I don't remember the time scales. 586 00:41:20,160 --> 00:41:25,320 So I don't remember whether UNSFLO was ever run on the Cray 587 00:41:25,320 --> 00:41:31,800 or not. I think it's possible it was. 588 00:41:35,760 --> 00:41:38,560 And were you travelling back when you were MIT working for 589 00:41:38,560 --> 00:41:42,200 Rolls-Royce, Did you come back to the UK, to Derby to sort of 590 00:41:42,840 --> 00:41:47,000 have meetings with it or was it slightly sort of separated? 591 00:41:48,520 --> 00:41:53,440 I mean, I probably came back twice a year at most. 592 00:41:54,480 --> 00:41:59,080 So. And, you know, to some extent I 593 00:41:59,080 --> 00:42:02,040 would be coming home to come home and see people. 594 00:42:02,760 --> 00:42:05,000 And then while I was here, I would visit Rolls-Royce. 595 00:42:05,000 --> 00:42:08,720 And to some extent, you know, a trip might, might be motivated, 596 00:42:09,120 --> 00:42:11,600 you know, more more primarily because of Rolls-Royce, you 597 00:42:11,600 --> 00:42:16,000 know. So, yeah, I guess probably twice 598 00:42:16,000 --> 00:42:19,640 a year was the norm in those days. 599 00:42:21,280 --> 00:42:24,040 So how did things progress then at MIT? 600 00:42:24,040 --> 00:42:28,120 So you were working on the Rolls-Royce projects that, Yeah. 601 00:42:28,120 --> 00:42:30,560 How did, how did, how did things evolve during that time? 602 00:42:32,200 --> 00:42:37,320 I mean, as well as working on the Rolls-Royce projects, I had 603 00:42:37,320 --> 00:42:41,640 some amount of funding from US sources. 604 00:42:41,960 --> 00:42:46,920 Not a lot, but things I think. Office of Naval Research Air 605 00:42:46,920 --> 00:42:53,400 Force Oh, we, we, we bought the stellar machines with a grant 606 00:42:53,400 --> 00:42:56,720 from DARPA. So again, DARPA was really 607 00:42:56,720 --> 00:43:01,880 active in funding new technologies to see whether, you 608 00:43:01,880 --> 00:43:03,520 know, these really were useful or not. 609 00:43:04,880 --> 00:43:10,440 That that is probably the worst proposal I have ever written. 610 00:43:12,520 --> 00:43:16,400 But but, but my excuse is by the time I submitted the proposal, 611 00:43:16,720 --> 00:43:21,760 my temperature was 102 or 103. I was going down with glandular 612 00:43:21,760 --> 00:43:24,120 fever. Oh, I've had that. 613 00:43:24,320 --> 00:43:25,960 That's bad glandular fever, isn't it? 614 00:43:26,000 --> 00:43:30,960 Yeah. So in in, in the way I was lucky 615 00:43:30,960 --> 00:43:34,440 mine was sufficiently bad that I had to be admitted into 616 00:43:34,440 --> 00:43:37,400 hospital. And so they then pumped, pumped 617 00:43:37,400 --> 00:43:41,680 me full of, you know, various antibiotics and stuff. 618 00:43:41,680 --> 00:43:47,520 And so I recovered well. I was ill for like 6 weeks or 619 00:43:47,560 --> 00:43:50,080 something, it was horrible. Yeah, I was. 620 00:43:50,880 --> 00:43:54,726 I was probably in hospital for a week or two and then I was 621 00:43:54,726 --> 00:43:59,570 sent home and told to stay at home and and recuperate for 622 00:43:59,570 --> 00:44:06,280 like 2 months or something. And and So what I did was to 623 00:44:06,280 --> 00:44:13,520 write up a 60 page document documenting all of UNSFLO so 624 00:44:13,640 --> 00:44:19,240 that that that was the best documented code I ever wrote it. 625 00:44:19,720 --> 00:44:25,360 It's always this, this rule that people hate writing detailed 626 00:44:25,360 --> 00:44:28,880 documentation and, you know, just just laying out all the 627 00:44:28,880 --> 00:44:33,480 details of the numerics, all of the things that your future 628 00:44:33,760 --> 00:44:35,840 people modifying the code need to do. 629 00:44:35,960 --> 00:44:39,480 You know, it's very hard getting getting students to do that. 630 00:44:40,600 --> 00:44:47,640 So anyway, I yeah, yes, I'm trying to trying to remember 631 00:44:47,640 --> 00:44:50,360 exactly when, when all these different things happened. 632 00:44:50,480 --> 00:44:52,440 Yeah. So. 633 00:44:54,640 --> 00:44:58,400 So you're progressing at MIT, you're an assistant professor. 634 00:44:58,960 --> 00:45:02,800 Did you feel that you would always stay there or what 635 00:45:02,800 --> 00:45:06,440 started to get into your head about coming back? 636 00:45:07,240 --> 00:45:14,280 To so I, I, I had a good, good job. 637 00:45:14,280 --> 00:45:16,160 Obviously they're, they're at MIT. 638 00:45:16,160 --> 00:45:19,440 I was part of the gas turbine lab and there were wonderful 639 00:45:19,440 --> 00:45:23,320 experimentalists there. And so, you know, we, we, we did 640 00:45:23,320 --> 00:45:27,320 lots of good works of comparing numerics with experiment. 641 00:45:27,320 --> 00:45:32,400 And there's one paper we, we, we have an unsteady heat transfer 642 00:45:32,680 --> 00:45:37,480 where in the sense neither the experimentalists nor me had 643 00:45:37,520 --> 00:45:40,960 great faith in our own research and yet the results matched 644 00:45:41,280 --> 00:45:43,760 wonderfully. We were delighted. 645 00:45:46,160 --> 00:45:51,840 I, I did feel to some extent, not quite a fish out the water, 646 00:45:51,840 --> 00:45:54,560 but really I am at heart a mathematician. 647 00:45:55,040 --> 00:45:58,400 You know, you, you asked me at, at the beginning, you know, was 648 00:45:58,400 --> 00:46:01,720 I the kind of kid that tinkered with devices? 649 00:46:01,960 --> 00:46:06,880 No, I wasn't, you know, so I, I, I really wasn't by nature an 650 00:46:06,880 --> 00:46:11,880 engineer in that sense, by by nature, I'm an applied 651 00:46:11,880 --> 00:46:16,720 mathematician who enjoys mathematics and enjoys seeing it 652 00:46:16,720 --> 00:46:19,144 being useful in the real world. 653 00:46:19,152 --> 00:46:22,176 I'm not fundamentally at heart an engineer. 654 00:46:22,184 --> 00:46:28,085 And so in that sense, I did feel a little bit constrained being 655 00:46:28,093 --> 00:46:30,748 in the aeronautics and astronautics department, which 656 00:46:30,756 --> 00:46:34,072 is, you know, where I was at MIT. 657 00:46:34,080 --> 00:46:40,322 And that that I was sort of most aware of when doing things like 658 00:46:40,322 --> 00:46:42,856 the non-reflecting boundary condition theory, because that 659 00:46:42,864 --> 00:46:47,932 was very much maths theory, but publishing it in in engineering 660 00:46:47,932 --> 00:46:52,145 CFD, So journals, you know, AIAA Journal. 661 00:46:52,153 --> 00:47:01,680 So that yeah, I think also, you know, getting married and 662 00:47:01,688 --> 00:47:04,560 thinking about where, where you want to have a family and raise 663 00:47:04,560 --> 00:47:07,854 kids and things like that. You know, that that that was 664 00:47:07,862 --> 00:47:12,617 part of it. And then the other part of it. 665 00:47:12,625 --> 00:47:18,947 And I forget the exact sequence of all of this was on one of my 666 00:47:18,947 --> 00:47:24,322 trips to Rolls-Royce, I was called in to see the chief 667 00:47:24,330 --> 00:47:28,756 engineer who said, oh, by the way, please let us know whenever 668 00:47:28,756 --> 00:47:32,795 you want to come back to this country and we'll sort it. 669 00:47:32,803 --> 00:47:37,939 So I had that standing offer from Rolls-Royce that they would 670 00:47:37,939 --> 00:47:42,720 organise it for me or that I could basically choose where I 671 00:47:42,728 --> 00:47:48,181 came back to. So yeah, so I decided, yes, I 672 00:47:48,189 --> 00:47:54,722 wanted to, to, to come home and, and, and I chose Oxford rather 673 00:47:54,730 --> 00:47:59,615 than Cambridge as the place to come to partly. 674 00:47:59,623 --> 00:48:04,468 Well, you see in, in Cambridge I'd have probably been in 675 00:48:04,476 --> 00:48:09,600 engineering, not in maths. You know, Cambridge maths was 676 00:48:09,608 --> 00:48:13,372 always kind of anti numerical analysis. 677 00:48:13,380 --> 00:48:20,742 There was certainly a chunk of the faculty who felt that you 678 00:48:20,750 --> 00:48:24,180 know the computer is what you used if you weren't clever 679 00:48:24,188 --> 00:48:29,526 enough to do it properly. Old school. 680 00:48:29,534 --> 00:48:38,640 Old school, yes, whereas Oxford really had had embraced 681 00:48:38,648 --> 00:48:43,576 numerical methods and had very, very strong group here and in 682 00:48:43,584 --> 00:48:46,860 those days the numerical analysis group, although they 683 00:48:46,868 --> 00:48:50,486 were a group of mathematicians, they were in the computer 684 00:48:50,494 --> 00:48:53,105 science department. And so there were also people in 685 00:48:53,105 --> 00:48:54,920 computer science on the parallel computing side. 686 00:48:54,928 --> 00:48:58,310 And so that that was an attraction to me as well, you 687 00:48:58,318 --> 00:49:00,675 know, to have them as sort of neighbors. 688 00:49:00,683 --> 00:49:05,835 As it turned out that that didn't work out because soon 689 00:49:05,843 --> 00:49:11,285 after arriving in 92, it was the first dot-com boom and all, all 690 00:49:11,285 --> 00:49:15,945 of the parallel computing people left to set up companies. 691 00:49:15,945 --> 00:49:19,280 So anyway, but but that was part 692 00:49:19,280 --> 00:49:21,240 of the motivation of choosing Oxford. 693 00:49:21,240 --> 00:49:25,040 Ah, OK. So yeah. 694 00:49:25,040 --> 00:49:27,800 So, you know, Rolls-Royce kind of organized it. 695 00:49:28,560 --> 00:49:33,520 I mean it, it had to be a properly advertised and competed 696 00:49:33,520 --> 00:49:36,040 for a position. So I was in this strange 697 00:49:36,040 --> 00:49:40,680 position of helping to write the job description, possibly even 698 00:49:40,680 --> 00:49:44,560 the advert for a position that I then applied for. 699 00:49:44,680 --> 00:49:50,040 You know, there are various parts of my career which I, I, I 700 00:49:50,040 --> 00:49:52,480 look back on. And now, now I realise just how 701 00:49:52,480 --> 00:49:55,510 peculiar they were at the time. 702 00:49:55,518 --> 00:49:59,761 That's right. So you technically came into the 703 00:49:59,761 --> 00:50:01,610 computer science department? Yeah. 704 00:50:01,618 --> 00:50:04,850 So in those days it was called the Computing Laboratory. 705 00:50:04,858 --> 00:50:08,255 Yeah. And that was a Keble Road always 706 00:50:08,255 --> 00:50:12,623 OK. And so that was like a 707 00:50:12,623 --> 00:50:16,055 Rolls-Royce. So it was a Rolls-Royce 708 00:50:16,063 --> 00:50:20,944 readership in CFD and then and then they funded me to set up a 709 00:50:20,944 --> 00:50:23,934 whole research group. So I, I have a lot of funding 710 00:50:23,942 --> 00:50:27,140 from them for a, you know, prolonged period. 711 00:50:27,148 --> 00:50:34,681 So really from 92 through to 2008 is when I moved to maths, 712 00:50:34,689 --> 00:50:40,350 you know, so, so, so for those sort of 15 years, you know, I, I 713 00:50:40,350 --> 00:50:42,800 had a lot of Rolls-Royce funding. 714 00:50:42,808 --> 00:50:48,024 So this is the bit that I was wondering about. 715 00:50:48,032 --> 00:50:53,042 So you came into computer science for CFD and is this the 716 00:50:53,042 --> 00:50:58,308 time then when I guess what most people recognize as the 717 00:50:58,308 --> 00:51:01,928 Hydra code is, is this where it begins or did it actually 718 00:51:01,928 --> 00:51:04,940 begin even when you were MIT? When, when would when would you 719 00:51:04,940 --> 00:51:07,870 say was the start of that? Journey. 720 00:51:07,878 --> 00:51:14,562 So I would say that Hydra proper started in about 96, but there 721 00:51:14,570 --> 00:51:20,496 was there were various bits of research that in the sense laid 722 00:51:20,496 --> 00:51:27,048 the foundations for Hydra. So while I was still at MIT, you 723 00:51:27,048 --> 00:51:32,900 know, so I was thinking about what to do after UNSFLO. 724 00:51:32,908 --> 00:51:39,319 And so my plan after UNSFLO was that we really wanted something 725 00:51:39,319 --> 00:51:45,464 that would be a design tool for doing complete engines, you 726 00:51:45,472 --> 00:51:50,120 know, so no no longer single stage, really looking 727 00:51:50,120 --> 00:51:56,238 multistage. I was also interested in the 728 00:51:56,246 --> 00:52:00,068 idea that the unsteadiness could be done from a linear 729 00:52:00,076 --> 00:52:03,216 perturbation point of view rather than doing nonlinear 730 00:52:03,224 --> 00:52:07,480 unsteady. So so also linearise the 731 00:52:07,488 --> 00:52:11,140 unsteady equations, look at, you know, harmonics. 732 00:52:11,148 --> 00:52:16,104 And so doing this for both flutter and forced response. 733 00:52:16,112 --> 00:52:23,160 The key technical issue there is whether it was legitimate to do 734 00:52:23,160 --> 00:52:26,790 linearised harmonic analysis of shock capturing. 735 00:52:26,798 --> 00:52:33,580 And so I had a student who did a project at MIT to prove that, 736 00:52:33,588 --> 00:52:37,762 show that that was was a legitimate thing to do that as 737 00:52:37,770 --> 00:52:42,634 long as you so slightly smeared the shock over a few grid points 738 00:52:42,634 --> 00:52:46,980 that that the linearised analysis did do the right thing 739 00:52:46,980 --> 00:52:52,362 and did, yeah, you know, you could do it on that basis. 740 00:52:52,370 --> 00:52:57,218 So that was the precursor work at MIT. 741 00:52:57,226 --> 00:53:04,535 And then early on in Oxford, I don't think I'd started this 742 00:53:04,543 --> 00:53:10,350 coding before I moved. I did a code called 743 00:53:10,350 --> 00:53:13,360 SLIQ—S-L-I-Q. So steady, linear and quadratic. 744 00:53:13,880 --> 00:53:17,186 So the idea was you did you know the non linear steady state, 745 00:53:17,186 --> 00:53:21,720 you did a linear perturbation analysis for the unsteady 746 00:53:21,720 --> 00:53:26,320 effects and then the quadratic was to it's kind of a formal 747 00:53:26,320 --> 00:53:31,760 asymptotic expansion to get the mean flow changes due to the 748 00:53:31,760 --> 00:53:34,944 second order quadratic effects. 749 00:53:34,952 --> 00:53:42,821 So, yeah, so I developed SLIQ early in, in in my Oxford days 750 00:53:42,829 --> 00:53:47,552 and and you know, I had a student who who, who worked on 751 00:53:47,560 --> 00:53:53,804 that with with me. So that was one piece of work. 752 00:53:53,812 --> 00:54:02,035 Another piece of work was back in the 1992–93 era, there was 753 00:54:02,043 --> 00:54:04,782 funding. I think this came from, you 754 00:54:04,790 --> 00:54:07,988 know, the UK Department of Trade and Industry DTI. 755 00:54:07,996 --> 00:54:12,580 In those days there was an initiative of setting up 756 00:54:12,588 --> 00:54:16,808 parallel application centres in various parts of the country. 757 00:54:16,816 --> 00:54:22,424 And so I, I had a colleague on the computer science side in 758 00:54:22,432 --> 00:54:25,815 Oxford, Bill McColl, who'd applied for that funding even 759 00:54:25,823 --> 00:54:31,795 before I arrived. And so had funding for an IBM 760 00:54:31,803 --> 00:54:40,282 machine, something called an SP2 and and also there was funding 761 00:54:40,290 --> 00:54:44,961 there for, for research and also also match funding. 762 00:54:44,969 --> 00:54:48,925 I think so. So I had 50:50 matching from 763 00:54:48,933 --> 00:54:53,824 Rolls-Royce and so did the project there on developing. 764 00:54:53,832 --> 00:55:00,268 So in the sense of support layer for doing distributed memory 765 00:55:00,276 --> 00:55:03,160 parallel computing. So this was something called, 766 00:55:03,168 --> 00:55:06,880 you know, we called OPlus Oxford parallel library for 767 00:55:06,888 --> 00:55:12,580 unstructured solvers. Although although orally OPlus 768 00:55:12,588 --> 00:55:18,085 doesn't doesn't sound right, but written down it looks good. 769 00:55:18,093 --> 00:55:24,220 So, yeah, so, so I think Bill, yeah, must have been in Oxford 770 00:55:24,228 --> 00:55:29,698 after I arrived from maybe two or three years before he left in 771 00:55:29,698 --> 00:55:33,144 that.com boom. He was one of the people I hoped 772 00:55:33,144 --> 00:55:36,116 to work, worked with a bit more on the parallel computing 773 00:55:36,116 --> 00:55:39,520 side. So, so we got a whole pile of 774 00:55:39,520 --> 00:55:44,880 funding half and Rolls-Royce half from DTI to develop this 775 00:55:44,880 --> 00:55:47,694 parallel application framework. 776 00:55:47,702 --> 00:55:55,740 There was another code that at the postdoc Paul Crumpton, who 777 00:55:55,748 --> 00:56:00,598 actually did most of the development of that software and 778 00:56:00,598 --> 00:56:05,545 wrote the CFD code as basically a test bed to check 779 00:56:05,545 --> 00:56:09,885 everything worked correctly. But that was never intended as 780 00:56:09,893 --> 00:56:17,492 a CFD code for Rolls-Royce. So then that's what led into 781 00:56:17,500 --> 00:56:20,984 Hydra. So basically there were the the 782 00:56:20,984 --> 00:56:25,168 ideas of doing steady post linear perturbation tested out 783 00:56:25,176 --> 00:56:30,500 in SLIQ and there was the parallel framework in OPlus 784 00:56:30,508 --> 00:56:34,360 that we'd developed thoroughly tested out. 785 00:56:34,368 --> 00:56:39,896 And so Hydra then what was built on those foundations now. 786 00:56:39,904 --> 00:56:44,592 So Hydra it dropped the idea of doing the quadratic piece. 787 00:56:44,600 --> 00:56:50,904 So, so it's non linear steady, or at least in its original 788 00:56:50,912 --> 00:56:57,800 incarnation, non linear, steady linear perturbation for flutter 789 00:56:57,800 --> 00:57:05,224 forced response and then adjoints of all of those for for 790 00:57:05,224 --> 00:57:08,840 the design optimization, both steady and and the unsteady 791 00:57:08,848 --> 00:57:12,355 aspects, you know, so the adjoints was completely 792 00:57:12,363 --> 00:57:17,318 motivated by the work that Antony Jameson was doing, you 793 00:57:17,326 --> 00:57:23,720 know, on on, you know, the aircraft side, except that I 794 00:57:23,728 --> 00:57:29,940 chose a different technical approach in as much as Anthony 795 00:57:29,948 --> 00:57:35,128 always viewed the adjoint being developed at the PDE level, you 796 00:57:35,128 --> 00:57:39,775 know, formulating the adjoint PDE and then thinking about how 797 00:57:39,775 --> 00:57:44,020 to discretize it. Whereas I followed the the 798 00:57:44,028 --> 00:57:48,824 so-called discrete adjoint approach, where where you take 799 00:57:48,832 --> 00:57:53,480 the non linear discrete equations, you linearize those 800 00:57:53,488 --> 00:57:58,322 sort of element by element and then you take the transpose of 801 00:57:58,330 --> 00:58:01,320 the matrix to define the the discrete adjoint. 802 00:58:01,328 --> 00:58:04,749 You know and. Yes. 803 00:58:04,757 --> 00:58:09,005 How did you, just out of interest at this time, maybe 804 00:58:09,013 --> 00:58:15,125 also MIT, but Oxford, how active were you in, you know, the AIAA, 805 00:58:15,125 --> 00:58:21,408 the the sort of turbo machinery conferences were you, were you 806 00:58:21,408 --> 00:58:25,754 sort of always, were you at these events and saw these 807 00:58:25,754 --> 00:58:30,352 people or was this more of a like direct industrial 808 00:58:30,360 --> 00:58:32,280 engagement? I'm always interested, like now 809 00:58:32,280 --> 00:58:34,680 I go to the AIAA conference. I'm always, just always 810 00:58:34,688 --> 00:58:38,280 wondering what it was like, you know, before and was that still 811 00:58:38,280 --> 00:58:43,488 the main venue I guess to go. Yes, yes, that, that, that, that 812 00:58:43,488 --> 00:58:49,342 was the main venue. I, I sometimes went to the 813 00:58:49,350 --> 00:58:56,627 ASME IGTI conference, but I went more, more to the AIAA 814 00:58:56,635 --> 00:59:00,474 conferences. You know, I think there was more 815 00:59:00,474 --> 00:59:05,040 discussion of CFD at at AIAA and in, in, in particular, you 816 00:59:05,048 --> 00:59:09,680 know, the AIAA CFD conference. I mean that that was really my 817 00:59:09,680 --> 00:59:13,160 home, you know, sort of during this period. 818 00:59:13,160 --> 00:59:18,736 I would say that, you know, the ASME was more on the 819 00:59:18,736 --> 00:59:23,360 application side. Yeah. 820 00:59:23,360 --> 00:59:26,585 So I, I, I, I can remember going to two or three of those, but 821 00:59:26,585 --> 00:59:29,240 it was the AIAA ones which really were. 822 00:59:29,240 --> 00:59:30,880 Where was was this Reno? No. 823 00:59:30,880 --> 00:59:36,080 Where was the CFD? So Reno was the January 824 00:59:36,080 --> 00:59:39,480 conference, the CFD conference, what was in the summer. 825 00:59:40,000 --> 00:59:47,016 So I I remember well, I remember one in Snowmass in, in 826 00:59:47,016 --> 00:59:48,454 Colorado. Yeah. 827 00:59:48,462 --> 00:59:49,654 Beautiful venue. Yeah. 828 00:59:49,662 --> 00:59:51,640 Good, good. Orienteering, right? 829 00:59:51,648 --> 00:59:56,080 Well, yeah, yeah, except it's high, high enough up that you 830 00:59:56,088 --> 00:59:58,516 really wouldn't be wanting to run at that altitude. 831 00:59:58,524 --> 01:00:00,928 No, Yeah, yeah. I mean, this is this is a ski 832 01:00:00,936 --> 01:00:02,498 resort that in the in the summer. 833 01:00:02,506 --> 01:00:06,476 You know, they, you know, are used for conferences because 834 01:00:06,484 --> 01:00:10,855 there aren't that many people who want to go hiking. 835 01:00:10,863 --> 01:00:15,600 But I remember Snowmass. I remember Hawaii. 836 01:00:15,608 --> 01:00:21,778 I remember there was 1 in LA. Yeah. 837 01:00:21,786 --> 01:00:25,784 I mean, it just hopped, hopped all over the place. 838 01:00:25,792 --> 01:00:26,980 Yeah. So. 839 01:00:26,988 --> 01:00:31,048 Yeah. So I was really mainly in those 840 01:00:31,048 --> 01:00:36,775 days going to engineering conferences, not so many, I 841 01:00:36,783 --> 01:00:42,255 guess some maths conferences, but, but in those days, yeah, 842 01:00:42,263 --> 01:00:46,548 most of my publishing was in in engineering journal still at 843 01:00:46,556 --> 01:00:48,540 that point. But did you struggle? 844 01:00:48,548 --> 01:00:52,332 I'm interested because I, because you have such a strong 845 01:00:52,340 --> 01:00:55,392 maths background, did you ever struggle getting accepted or 846 01:00:55,400 --> 01:01:00,655 being in that blur of what journal, what conference is the 847 01:01:00,655 --> 01:01:03,895 right place for maths? And then I always find this 848 01:01:03,903 --> 01:01:06,807 interesting that, you know, it's something too applied or too 849 01:01:06,815 --> 01:01:08,400 fundamental. Did you find a sweet spot or was 850 01:01:08,400 --> 01:01:11,720 there still a little bit of a straight frustration that your 851 01:01:11,728 --> 01:01:14,730 deep maths wasn't understood by everybody? 852 01:01:14,738 --> 01:01:19,872 Do do you know what I'm getting at or was it not an issue then? 853 01:01:19,872 --> 01:01:23,080 I mean, I, I think I was 854 01:01:23,080 --> 01:01:27,840 probably in those days still viewed more as an engineer than 855 01:01:27,880 --> 01:01:29,912 than the mathematics, you know. 856 01:01:29,920 --> 01:01:35,810 I mean, you know, if you have a PhD in in aeronautics from MIT, 857 01:01:35,810 --> 01:01:38,020 you know, you're, you're an engineer. 858 01:01:38,028 --> 01:01:41,684 Yeah. So, I mean, there was never a 859 01:01:41,692 --> 01:01:44,828 question of the engineering community not accepting me. 860 01:01:44,836 --> 01:01:50,120 You know, I think, you know, my, my evolution has, has been one 861 01:01:50,128 --> 01:01:52,360 of the of the maths community accepting me. 862 01:01:52,368 --> 01:01:56,227 OK, OK. Yeah. 863 01:01:56,235 --> 01:02:00,215 So, yeah. And that, that, that I guess, 864 01:02:00,223 --> 01:02:04,800 you know, happened more once once I moved over into the Maths 865 01:02:04,800 --> 01:02:10,133 Institute. But yes, I guess I felt, I mean, 866 01:02:10,133 --> 01:02:15,080 having made the, the move to the numerical analysis group in 867 01:02:15,080 --> 01:02:19,681 Oxford, then in a sense I felt I was back amongst 868 01:02:19,681 --> 01:02:23,160 mathematicians. But but I was still very much at 869 01:02:23,160 --> 01:02:26,200 the engineering end of of of the group. 870 01:02:26,600 --> 01:02:32,729 You know, I think I was probably 40 before I wrote my first 871 01:02:32,729 --> 01:02:38,366 paper that had a theorem and a proof in it, you know, So I 872 01:02:38,366 --> 01:02:41,440 mean that that's kind of needed to be a mathematician. 873 01:02:42,600 --> 01:02:45,040 So what was the link to the engineering department at that 874 01:02:45,040 --> 01:02:50,120 time like because there's was, was it the case that that was 875 01:02:50,120 --> 01:02:55,680 more experimental work and the sort of CFD was mainly computer 876 01:02:55,680 --> 01:02:58,480 science? Is, is that sort of, I always 877 01:02:58,480 --> 01:03:01,160 found that unique in Oxford that, you know, like I said, MIT 878 01:03:01,160 --> 01:03:05,280 engineering where like, yeah, I guess Oxford's a bit different. 879 01:03:05,960 --> 01:03:11,200 So in in Oxford at that time we had three UTCs. 880 01:03:11,400 --> 01:03:14,920 So the UTCs are the university technology centres that, that 881 01:03:14,920 --> 01:03:22,840 Rolls-Royce set up. And so I had mine in CFD, there 882 01:03:22,840 --> 01:03:30,160 was an experimental one in heat transfer in, in, in engineering 883 01:03:30,160 --> 01:03:33,040 science. And then there was one in 884 01:03:33,040 --> 01:03:39,808 materials which I didn't didn't have any interaction with 885 01:03:39,808 --> 01:03:43,120 except one of the profs. There used to be an orienteer 886 01:03:43,120 --> 01:03:44,840 back in my case. So there was. 887 01:03:46,200 --> 01:03:51,280 A UTC just for CFD. Yes, yes, yes. 888 01:03:51,280 --> 01:03:55,560 So this was my, my own little UTC. 889 01:03:55,560 --> 01:04:03,000 I was the only academic in it. So and, and possibly within the 890 01:04:03,000 --> 01:04:07,040 Rolls-Royce family, that's slightly unusual to have a UTC 891 01:04:07,040 --> 01:04:12,720 that only has one academic. Generally they're they're bigger 892 01:04:12,720 --> 01:04:15,920 than that. But it was a funding mechanism, 893 01:04:15,920 --> 01:04:17,680 I guess, and you were doing the work. 894 01:04:18,600 --> 01:04:22,840 Yes, the funding mechanism it it, it involved me in all the 895 01:04:22,840 --> 01:04:26,040 UTC directors meetings. I mean, you were part of the 896 01:04:26,040 --> 01:04:31,080 family. And I mean that that's kind of 897 01:04:31,080 --> 01:04:35,640 important in the sense that Rolls-Royce really knew how to 898 01:04:35,640 --> 01:04:43,680 work well with academics, that you were part of the family in 899 01:04:43,680 --> 01:04:46,840 the sense that you knew all the problems as well as, you know, 900 01:04:46,840 --> 01:04:49,440 the, the achievements of, of, of Rolls-Royce. 901 01:04:49,440 --> 01:04:54,040 You know, they, they, they, you know, they didn't hide anything 902 01:04:54,040 --> 01:04:57,010 from you so that you could think about what you might 903 01:04:57,010 --> 01:05:00,680 potentially do to, to, to help them address some of their 904 01:05:00,680 --> 01:05:06,560 challenges and things. So, you know, it really was a 905 01:05:06,560 --> 01:05:13,040 very good collaborative experience when, when you're 906 01:05:13,040 --> 01:05:16,320 working with industry, you know, it's important that both sides 907 01:05:16,320 --> 01:05:20,200 realise that what the other wants out of the relationship 908 01:05:20,200 --> 01:05:23,440 is, is different, you know, and, and, and so you're always 909 01:05:23,440 --> 01:05:26,459 looking for this sort of win-win arrangement, you know, the, 910 01:05:26,459 --> 01:05:29,720 you know, so they understood that, you know, for us, it was 911 01:05:29,720 --> 01:05:33,680 important for the students to publish papers, to write 912 01:05:33,760 --> 01:05:37,080 dissertations, you know, that there would be times when they 913 01:05:37,080 --> 01:05:39,880 would be utterly focused on writing their dissertation and 914 01:05:40,160 --> 01:05:44,800 not doing any more research, you know, but equally, I, I 915 01:05:44,800 --> 01:05:48,280 understood what Rolls-Royce needed out of the relationship, 916 01:05:49,200 --> 01:05:53,120 which was primarily software, but occasionally if, if there 917 01:05:53,120 --> 01:05:56,840 was a particular engineering thing to be investigated, you 918 01:05:56,840 --> 01:06:01,840 know, we may occasionally do some, no special calculations 919 01:06:01,840 --> 01:06:05,520 just for them, as it were, rather than as, as part of the 920 01:06:05,520 --> 01:06:07,680 research. Yeah, I always found that 921 01:06:07,680 --> 01:06:13,280 interesting that some it takes a special company to understand 922 01:06:13,280 --> 01:06:17,000 the value of academic engagement and have the patience and the 923 01:06:17,000 --> 01:06:20,760 long term vision that it's not just cheap labour. 924 01:06:21,440 --> 01:06:23,564 Yes, you know that requires you. 925 01:06:23,572 --> 01:06:27,618 You have to be very careful to make sure it's never just cheap 926 01:06:27,618 --> 01:06:32,920 labour. I think Rolls-Royce was possibly 927 01:06:32,920 --> 01:06:36,266 slightly disappointed that they never ended up being able 928 01:06:36,266 --> 01:06:39,870 to hire any of my students. You know, that's, that's, that's 929 01:06:39,870 --> 01:06:43,252 the other thing that, you know, Rolls-Royce would ideally 930 01:06:43,252 --> 01:06:49,125 like from a UTC is as a source of, you know, people to be 931 01:06:49,125 --> 01:06:53,960 employed, you know, and that that never quite happened. 932 01:06:53,968 --> 01:06:57,840 But, yeah. But certainly they, they, they, 933 01:06:57,840 --> 01:07:01,200 they got their money's worth in terms of CFD codes. 934 01:07:01,208 --> 01:07:06,044 So, yeah, and I guess like in today's world probably, you 935 01:07:06,052 --> 01:07:10,375 know, you'd create a start up or something, write a code, you 936 01:07:10,383 --> 01:07:15,375 know, I guess at that time there was a more traditional link to 937 01:07:15,383 --> 01:07:18,454 the to the company, right? You know, in terms of. 938 01:07:18,462 --> 01:07:20,720 Yeah. I mean in, in, in this area, it 939 01:07:20,720 --> 01:07:22,600 would be tough to do a start up. 940 01:07:22,640 --> 01:07:27,520 I mean, I guess Juan Alonso's done, done the startup. 941 01:07:27,520 --> 01:07:30,880 I haven't talked to him recently as to how, how that's going. 942 01:07:32,240 --> 01:07:38,240 I mean it, yeah. At, at, at one point in parallel 943 01:07:38,240 --> 01:07:43,120 computing, I, you know, once I moved out of CFD into into 944 01:07:43,120 --> 01:07:47,560 mathematical finance, I, I tried setting up a spin off. 945 01:07:48,760 --> 01:07:52,280 No, it's just more hard work. Yeah, I'm, I'm, I'm, I'm 946 01:07:52,320 --> 01:07:56,520 fundamentally an academic, not, not a start up guy. 947 01:07:57,440 --> 01:08:01,960 The reason I say it is because, you know, speaking now to 948 01:08:01,960 --> 01:08:08,400 yourself or, or to Antony, and I guess that like now CFD has 949 01:08:08,400 --> 01:08:13,160 become very dominated by these huge multibillion dollar 950 01:08:13,560 --> 01:08:18,720 commercial companies. I guess in the 80s, that was 951 01:08:18,920 --> 01:08:23,200 before the time, wasn't it? It was before the, the Fluent 952 01:08:23,200 --> 01:08:28,359 and, and, and the, the OpenFOAM and it was still, you wrote your 953 01:08:28,359 --> 01:08:31,600 own code. I assume that's old. 954 01:08:32,359 --> 01:08:35,261 Was that one of the reasons for the Hydra development that they 955 01:08:35,261 --> 01:08:37,479 couldn't just buy something off the shelf that wasn't a 956 01:08:37,479 --> 01:08:40,120 company that could just sell them a capability they felt they 957 01:08:40,120 --> 01:08:41,160 needed? Yes. 958 01:08:41,160 --> 01:08:44,520 And even today I don't think there's a company that could 959 01:08:44,520 --> 01:08:47,652 sell them what they need because the turbo machinery 960 01:08:47,652 --> 01:08:50,160 requirements are really very specific. 961 01:08:50,167 --> 01:08:54,627 So I mean I've I've not kept up with the discipline. 962 01:08:54,635 --> 01:09:00,161 So I don't know what ANSYS Fluent has as a capability these 963 01:09:00,161 --> 01:09:06,183 days, but you know, the ability, for example, to have 964 01:09:06,183 --> 01:09:10,652 flutter calculations being performed on a single blade 965 01:09:10,652 --> 01:09:13,910 passage with an inter blade phase angle between the 966 01:09:13,910 --> 01:09:16,359 passages. You know, that's such a unique 967 01:09:16,560 --> 01:09:21,612 requirement of turbo machinery that I'm not sure ANSYS views 968 01:09:21,612 --> 01:09:24,560 the market as being big enough to develop that capability. 969 01:09:24,568 --> 01:09:28,564 And then, and then you get into things like real gas effects, 970 01:09:28,573 --> 01:09:32,290 you know, so, so, so Hydra doesn't assume a fixed gamma. 971 01:09:32,298 --> 01:09:36,486 You know, it, it, it has a general, you know, energy 972 01:09:36,493 --> 01:09:38,904 temperature relationship in there. 973 01:09:38,912 --> 01:09:45,160 So again, I mean, I guess things like that ANSYS could add in, 974 01:09:45,167 --> 01:09:49,109 but they would be additions for particular customers and, and 975 01:09:49,109 --> 01:09:54,720 they would charge accordingly. I mean, one of the reasons I got 976 01:09:54,720 --> 01:10:01,480 out of the CFD business is it's not clear to me long term, you 977 01:10:01,480 --> 01:10:06,480 know, what Rolls-Royce will do for their next CFD code. 978 01:10:06,920 --> 01:10:12,040 So I mean, Hydra's just celebrated its 25th anniversary 979 01:10:12,040 --> 01:10:17,600 at, at, at Rolls-Royce. I would say it'll remain the 980 01:10:17,600 --> 01:10:21,920 corporate, you know, primary code for at least another 10 981 01:10:21,920 --> 01:10:25,360 years because it takes a long time to, to introduce a brand 982 01:10:25,360 --> 01:10:29,160 new code. And you know, they, you know, 983 01:10:29,160 --> 01:10:36,215 they're doing work with Spencer Sherwin on his Nektar++ code 984 01:10:36,215 --> 01:10:40,993 with the thought that they will use that in, in applications 985 01:10:40,993 --> 01:10:47,124 where they want to really do DNS or or at least high resolution 986 01:10:47,124 --> 01:10:53,520 LES, but, but that won't be part of their sort of standard, you 987 01:10:53,520 --> 01:11:00,600 know, design process. So I think it's, it's harder to 988 01:11:00,600 --> 01:11:06,040 see now. Yeah, companies funding a brand 989 01:11:06,040 --> 01:11:10,240 new code development. That's kind of what I was 990 01:11:10,280 --> 01:11:15,480 hinting at, that it does seem as if that that generation in the 991 01:11:15,480 --> 01:11:21,760 80s and the 90s throughout aerospace was a time of real 992 01:11:21,760 --> 01:11:23,840 innovation and development. It was. 993 01:11:23,840 --> 01:11:27,560 Themselves. Yeah, I mean when, when, when I, 994 01:11:27,880 --> 01:11:32,280 you know, joined Rolls-Royce's first CFD group in 1980. 995 01:11:32,680 --> 01:11:36,520 You know, most of the design methods were one-dimensional 996 01:11:36,520 --> 01:11:42,720 design methods, you know, with sort of mean, mean line, sort 997 01:11:42,720 --> 01:11:46,560 of, you know, I mean really mathematical models rather than 998 01:11:46,560 --> 01:11:51,760 numerics. So, yeah, I mean it, it, it's 999 01:11:51,760 --> 01:11:57,240 been wonderful just being part of that whole process of, you 1000 01:11:57,240 --> 01:12:01,880 know, you know, the development of computational engineering 1001 01:12:01,880 --> 01:12:07,294 and, and seeing it go all the way from 1D, 2D and 3D, from 1002 01:12:07,294 --> 01:12:12,040 inviscid to viscous steady to unsteady and then the adjoints 1003 01:12:12,040 --> 01:12:16,160 and everything. You know, it is interesting. 1004 01:12:16,160 --> 01:12:21,120 So looking back over your lifetime and looking at the 1005 01:12:21,120 --> 01:12:24,360 progress in computational engineering and then of course 1006 01:12:24,360 --> 01:12:27,040 the progress in parallel computing and the high 1007 01:12:27,040 --> 01:12:30,080 performance computing, I mean. Both of them, yeah. 1008 01:12:30,360 --> 01:12:33,040 It's really stunning looking back over the years. 1009 01:12:34,000 --> 01:12:38,720 Yeah, that's what I was wondering about is so you, so 1010 01:12:38,720 --> 01:12:40,920 you came into Oxford doing the computer science. 1011 01:12:42,280 --> 01:12:45,600 I was just looking at the and I guess this is where we have a 1012 01:12:45,600 --> 01:12:48,600 slight shared connection into some of the people. 1013 01:12:51,240 --> 01:12:54,720 How did and you mentioned also about the government setting up 1014 01:12:54,720 --> 01:12:56,440 some, you know, parallel computing. 1015 01:12:57,400 --> 01:13:02,496 I guess how did that evolve the HPC angle, Oxford, maybe some 1016 01:13:02,496 --> 01:13:06,080 of the e-Research ideas, you moved into maths. 1017 01:13:06,080 --> 01:13:09,360 When did how did that time? And I guess that probably aligns 1018 01:13:09,360 --> 01:13:14,280 to also when you had a declining interest in CFD and more into 1019 01:13:14,280 --> 01:13:17,120 the computing and other areas. Is that fair to say? 1020 01:13:17,600 --> 01:13:20,280 Were they aligned a little bit or were they sort of separate? 1021 01:13:21,000 --> 01:13:25,160 OK, so yeah, yeah, let's let let's. 1022 01:13:25,480 --> 01:13:26,800 Try to. I've asked you like 3 questions 1023 01:13:26,800 --> 01:13:29,320 in one. Yeah, yeah, yeah. 1024 01:13:29,360 --> 01:13:32,280 So get get the timeline straight in my mind. 1025 01:13:32,880 --> 01:13:36,240 OK. So the main Hydra research 1026 01:13:36,240 --> 01:13:47,440 period was sort of 96 to maybe 2004, 2006, something like that. 1027 01:13:48,840 --> 01:13:55,920 And the Hydra code was built on top of this OPlus parallel 1028 01:13:55,920 --> 01:13:59,360 layer that we had developed earlier as part of the DTI 1029 01:13:59,360 --> 01:14:05,480 funded activity. And so incidentally, both OPlus 1030 01:14:05,480 --> 01:14:09,480 and Hydra, you're the IPR belongs to Rolls-Royce. 1031 01:14:12,480 --> 01:14:20,960 So OPlus was very much developed in the time of single 1032 01:14:20,960 --> 01:14:27,280 core CPUs, RISC-based CPUs, you know, so, so there, you 1033 01:14:27,280 --> 01:14:30,160 know, there was good performance there, but single core. 1034 01:14:30,840 --> 01:14:35,000 And then what was happening over time was that CPUs were, were 1035 01:14:35,000 --> 01:14:41,920 going multicore and, and then we had GPUs that came along. 1036 01:14:42,840 --> 01:14:47,960 So, so again, I've, I've always been interested in the latest 1037 01:14:47,960 --> 01:14:53,920 computing technology. So I think it must have been 1038 01:14:56,320 --> 01:15:06,160 sort of late 2006, early 2007. An ex colleague, Mike Rodger, I 1039 01:15:06,160 --> 01:15:08,000 don't know if the name means anything to you. 1040 01:15:08,720 --> 01:15:16,120 He, he set up a company spin off from Warwick actually, which 1041 01:15:16,200 --> 01:15:22,040 sold parallel, you know, systems, you know, parallel 1042 01:15:22,040 --> 01:15:26,280 clusters. And he was the one who said to 1043 01:15:26,280 --> 01:15:32,720 me, there's this new thing called the GPU and this, this, 1044 01:15:32,720 --> 01:15:36,520 this new language CUDA. Actually, I'm not even sure if 1045 01:15:36,520 --> 01:15:38,840 he told me about CUDA. He may just have told me, you 1046 01:15:38,840 --> 01:15:41,000 know, there's this new hardware called GPUs. 1047 01:15:41,200 --> 01:15:43,720 It's really impressive. You know, you ought to have a 1048 01:15:43,720 --> 01:15:51,640 look at it. And at about the same time there 1049 01:15:51,640 --> 01:15:57,440 was a company based in Bristol called ClearSpeed, which had 1050 01:15:57,440 --> 01:16:01,300 people in it who I think had the background coming from the 1051 01:16:01,300 --> 01:16:07,720 INMOS transputer days, you know, so, so I first tried the the 1052 01:16:07,720 --> 01:16:14,400 ClearSpeed card and I was impressed by what it was capable 1053 01:16:14,400 --> 01:16:17,800 of, but I wasn't blown away by it. 1054 01:16:20,240 --> 01:16:27,320 And then actually I had the visiting student, a Chinese 1055 01:16:27,320 --> 01:16:31,760 student, I forget how, how he came to me, but he, he did the 1056 01:16:31,760 --> 01:16:35,040 ClearSpeed work. And then I said, well, now, now 1057 01:16:35,040 --> 01:16:40,120 let's try this GPU. And he came back to me with it 1058 01:16:40,120 --> 01:16:45,240 within like a couple of weeks with the code performing just 1059 01:16:45,240 --> 01:16:48,120 incredibly fast. And I was saying, OK, you've 1060 01:16:48,120 --> 01:16:55,840 clearly messed up the timing. And so it took took me a week to 1061 01:16:55,840 --> 01:16:59,320 convince myself, no, I mean it really was performing that that 1062 01:16:59,320 --> 01:17:03,760 well. So, so this was right at the 1063 01:17:03,760 --> 01:17:08,840 beginning of CUDA. So I think we, we started with 1064 01:17:09,360 --> 01:17:12,440 the 0.9 beta release or something like that. 1065 01:17:14,080 --> 01:17:18,120 And, and so I was just stunned at, you know, the power of, of 1066 01:17:18,120 --> 01:17:20,680 the GPU. Now this was when I was already 1067 01:17:20,680 --> 01:17:23,840 starting to pivot into mathematical finance. 1068 01:17:23,840 --> 01:17:27,320 So I was actually looking at it for doing Monte Carlo 1069 01:17:28,360 --> 01:17:34,240 simulations in, in finance, which is why NVIDIA then got 1070 01:17:34,240 --> 01:17:37,800 interested in me because they saw that as a potential market. 1071 01:17:40,320 --> 01:17:46,800 So, yeah, so, so I was pivoting to finance, but I was also 1072 01:17:46,800 --> 01:17:49,160 interested in the power of the GPUs. 1073 01:17:49,520 --> 01:17:53,680 So after two or three years of that, I'm trying to think 1074 01:17:53,680 --> 01:18:00,200 exactly on the timing. Again, I was interested in doing 1075 01:18:00,200 --> 01:18:07,240 an upgrade to OPlus which could then incorporate both GPUs and 1076 01:18:07,240 --> 01:18:12,240 multicore CPUs. But by the way, just before you 1077 01:18:12,240 --> 01:18:15,656 go on, did so did you if it was 2006, 2007, did you interact 1078 01:18:15,656 --> 01:18:19,972 with Ian Buck at all then on the CUDA stuff or was it on 1079 01:18:19,972 --> 01:18:23,440 people more in the UK who? You no, no. 1080 01:18:23,440 --> 01:18:30,200 So Massimiliano I, I interacted with Ian. 1081 01:18:30,360 --> 01:18:37,200 Ian, I probably met in in the same way that I met Jensen at, 1082 01:18:37,320 --> 01:18:39,720 you know, drinks things at GTC. 1083 01:18:39,728 --> 01:18:44,240 So, so I was, I think the number 2 CUDA fellow. 1084 01:18:44,248 --> 01:18:47,400 I think. I think the first CUDA fellow 1085 01:18:47,408 --> 01:18:52,285 was somebody in India. And then I was #2 like a month 1086 01:18:52,293 --> 01:18:56,396 later. So yeah, yeah, I mean, yeah. 1087 01:18:56,404 --> 01:19:00,146 So I was going along to the GTC, you know. 1088 01:19:00,154 --> 01:19:05,336 I have to say to the, I guess to the people is that a bit of like 1089 01:19:05,336 --> 01:19:08,448 self-proclaimed interest just because now the team I'm in has 1090 01:19:08,448 --> 01:19:10,840 those people in it. So I'm always intrigued when 1091 01:19:10,848 --> 01:19:13,530 they're like, oh, you, you know that person and this person 1092 01:19:13,538 --> 01:19:15,900 knows this person. So it's kind of interesting, 1093 01:19:15,908 --> 01:19:18,910 especially because you were, I mean, now everybody knows NVIDIA 1094 01:19:18,910 --> 01:19:22,521 and GPUs. But 2006 seven was really early 1095 01:19:22,521 --> 01:19:25,240 on, wasn't it? Yeah, yeah. 1096 01:19:25,248 --> 01:19:29,730 And, and in fact, some of the people from ClearSpeed moved to 1097 01:19:29,730 --> 01:19:36,024 NVIDIA in part in influenced by my feedback to them. 1098 01:19:36,032 --> 01:19:39,385 Yeah. They then saw the writing on the 1099 01:19:39,385 --> 01:19:41,855 wall and moved, moved ship accordingly. 1100 01:19:41,863 --> 01:19:48,160 So, but yes, I don't think I, yeah, I don't remember talking 1101 01:19:48,168 --> 01:19:52,415 to Ian Buck at all. Massimiliano was was was one of 1102 01:19:52,415 --> 01:19:58,707 my early contacts in in in the Bay Area. 1103 01:19:58,715 --> 01:20:03,859 Yeah. So, yeah, yes, I sort of came 1104 01:20:03,867 --> 01:20:08,770 back then to do this upgrade from OPlus to OP2. 1105 01:20:08,778 --> 01:20:16,040 So that was with Gihan Mudalige and István Reguly—you know 1106 01:20:16,440 --> 01:20:23,400 both of them. So that had funding from both 1107 01:20:23,400 --> 01:20:30,831 Rolls-Royce and, you know, the UK EPSRC, you know, government 1108 01:20:30,831 --> 01:20:36,880 funding. So that has in a sense, you 1109 01:20:36,880 --> 01:20:41,120 know, protected Hydra's future by, by up upgrading the, the, 1110 01:20:41,320 --> 01:20:45,240 the underlying computing harness for, for, you know, 1111 01:20:45,240 --> 01:20:51,560 modern systems. But but I haven't really done 1112 01:20:51,920 --> 01:20:58,372 CFD as such since about 2007. I think that was the last 1113 01:20:58,380 --> 01:21:01,078 time. So what was, yeah, maybe that's 1114 01:21:01,078 --> 01:21:05,996 the elephant in the room. Why what, why was the, I mean, 1115 01:21:06,004 --> 01:21:10,542 you sort of hinted towards it earlier, but what was the reason 1116 01:21:10,542 --> 01:21:15,580 to maybe move away from CFD and move into the, you know, the 1117 01:21:15,580 --> 01:21:18,342 financial side? Was that the maths angle moving 1118 01:21:18,342 --> 01:21:21,780 to the maths department sort of just wanting new challenges? 1119 01:21:21,788 --> 01:21:25,720 The computing side What? What was the ingredients to 1120 01:21:25,728 --> 01:21:35,820 that? Yeah, I'm just, I'm just 1121 01:21:35,828 --> 01:21:38,180 thinking what, what, what what to say. 1122 01:21:38,188 --> 01:21:42,200 Yeah, let's, let's let's be be open about it because I have 1123 01:21:42,208 --> 01:21:47,250 been been open about this, you know, in, in various settings. 1124 01:21:47,258 --> 01:21:53,950 I got, you know, more than burnt out developing the Hydra code. 1125 01:21:53,958 --> 01:21:58,682 You know that there were lots of aspects of the Hydra code that 1126 01:21:58,690 --> 01:22:02,840 worked very well, but they had major problems early on with 1127 01:22:02,848 --> 01:22:08,690 numerical stability which got me horrendously stressed out to to 1128 01:22:08,690 --> 01:22:12,790 the point of serious consequences. 1129 01:22:12,798 --> 01:22:18,072 So, you know, for about 6 months it took to, to recover from all 1130 01:22:18,072 --> 01:22:24,062 of that. And I decided at that point that 1131 01:22:24,062 --> 01:22:30,040 it's a got to a situation where in a sense, too much 1132 01:22:30,048 --> 01:22:32,080 perspiration and too little inspiration. 1133 01:22:32,088 --> 01:22:38,525 It, it, it's not being fun, you know, managing a large software 1134 01:22:38,525 --> 01:22:45,340 project is, is tiring, you know, and, and, you know, I'd, 1135 01:22:45,340 --> 01:22:52,201 yeah, I'd, I'd had enough. So, you know, I sort of finished 1136 01:22:52,201 --> 01:22:57,250 up the things that needed to be finished up, transferred 1137 01:22:57,250 --> 01:23:01,276 things to Rolls-Royce and they've, they've continued, you 1138 01:23:01,276 --> 01:23:02,920 know, the development subsequently. 1139 01:23:05,000 --> 01:23:10,440 So I decided I had to had to do something fresh different. 1140 01:23:10,920 --> 01:23:15,920 I had to get out of big codes. Yeah. 1141 01:23:16,760 --> 01:23:24,827 So in, in engineering style, I, I thought, OK, where's the 1142 01:23:24,827 --> 01:23:28,400 money in terms of where, where's the research funding? 1143 01:23:28,840 --> 01:23:32,880 What do, what do people care about health and wealth? 1144 01:23:33,360 --> 01:23:38,880 So, so computational biology and computational finance, those, 1145 01:23:39,120 --> 01:23:42,760 those were the two areas I contemplated moving into. 1146 01:23:43,280 --> 01:23:48,840 And computational biology, I, I know no biology. 1147 01:23:48,840 --> 01:23:51,840 So I'd have been starting from ground zero on that. 1148 01:23:52,240 --> 01:23:57,200 And I also wasn't convinced there was the research funding 1149 01:23:57,200 --> 01:24:00,960 in in that area, although I was proved wrong on that, you know, 1150 01:24:01,040 --> 01:24:03,080 you know, there is actually plenty of funding there. 1151 01:24:04,240 --> 01:24:07,583 Computational finance, there was a very good mathematical 1152 01:24:07,583 --> 01:24:10,720 finance group in the maths department but they didn't 1153 01:24:10,720 --> 01:24:12,856 have a lot of numerical expertise. 1154 01:24:12,864 --> 01:24:16,728 So, so for me that that was a natural fit. 1155 01:24:16,736 --> 01:24:21,654 And So what I did was I actually transferred from computer 1156 01:24:21,662 --> 01:24:26,752 science into maths to join the mathematical finance group. 1157 01:24:26,760 --> 01:24:31,700 And then a couple of years later the rest of the numerical 1158 01:24:31,708 --> 01:24:35,975 analysis group kind of moved, moved across behind me into 1159 01:24:35,983 --> 01:24:43,496 maths as well. And so now, now I'm, I'm head of 1160 01:24:43,496 --> 01:24:46,920 the, the numerical analysis group now, you know, so, so in a 1161 01:24:46,920 --> 01:24:50,400 sense in, in, in the last few years I've sort of transferred 1162 01:24:50,408 --> 01:24:54,304 back from the finance group into the numerical analysis group. 1163 01:24:54,312 --> 01:24:59,769 As which ultimately sort of came with you in the fullness of time 1164 01:24:59,769 --> 01:25:02,496 from the computer science into the into the. 1165 01:25:02,504 --> 01:25:05,200 Maths, yeah. When when when maths got its new 1166 01:25:05,200 --> 01:25:07,824 building, there was the opportunity for the numerical 1167 01:25:07,832 --> 01:25:11,388 analysis group to move and it was a take-it-or-leave-it kind 1168 01:25:11,396 --> 01:25:13,990 of opportunity. And you know, they, they, they, 1169 01:25:13,990 --> 01:25:16,580 they moved. And for people listening, it's a 1170 01:25:16,580 --> 01:25:18,420 lovely building. It's a very nice. 1171 01:25:18,428 --> 01:25:23,118 It's a very nice. Building also, you know, when I 1172 01:25:23,118 --> 01:25:27,340 joined computer science in 92, the numerical analysis group was 1173 01:25:27,340 --> 01:25:31,136 half of the whole department, you know, Wow yeah. 1174 01:25:31,144 --> 01:25:34,740 So what happened over time was the computer science side grew 1175 01:25:34,748 --> 01:25:36,665 and the numerical analysis side didn't. 1176 01:25:36,673 --> 01:25:42,816 And so by by the time the numerical analysis group moved 1177 01:25:42,824 --> 01:25:49,620 over in 2010, yeah, they were basically pushed out 1178 01:25:49,628 --> 01:25:54,302 effectively, you know, you know, you know, that it didn't make 1179 01:25:54,310 --> 01:25:57,460 sense for them being in computer science any longer. 1180 01:25:57,468 --> 01:26:01,460 I mean, historically they were there because that's where the 1181 01:26:01,468 --> 01:26:05,896 computers were, you know, and, and that, you know, that is a 1182 01:26:05,904 --> 01:26:07,467 history that has happened elsewhere. 1183 01:26:07,475 --> 01:26:10,746 I mean, you know, numerical analysis at Stanford for a long 1184 01:26:10,746 --> 01:26:12,640 time was based in computer science. 1185 01:26:12,648 --> 01:26:15,465 And I think, again, historically, it's because 1186 01:26:15,473 --> 01:26:19,980 that's where the computers were. I do always find this 1187 01:26:19,988 --> 01:26:22,564 interesting and that's why I just wanted to, you know, get 1188 01:26:22,572 --> 01:26:26,740 the I get the quick history of like places like the e-Research 1189 01:26:26,740 --> 01:26:30,440 Centre and others because in some ways I understand the logic 1190 01:26:30,440 --> 01:26:33,428 that where does some of these places fit? 1191 01:26:33,436 --> 01:26:36,100 You know, there's numerical analysis, there's computers, 1192 01:26:36,108 --> 01:26:38,481 there's high performance computing, there's an 1193 01:26:38,489 --> 01:26:43,438 engineering application, there's a pure application was the was 1194 01:26:43,446 --> 01:26:47,910 was that sort of e-Research centre and I believe there were 1195 01:26:47,910 --> 01:26:52,022 others around the country was the initiative to try and bring 1196 01:26:52,022 --> 01:26:56,460 them together to in a more collaborative way Was that was 1197 01:26:56,468 --> 01:27:00,570 that sort of the initiative? I know that they've subsequently 1198 01:27:00,570 --> 01:27:03,336 largely folded into other departments now, but. 1199 01:27:03,344 --> 01:27:09,170 Yes, I mean, it it there was a massive funding initiative to 1200 01:27:09,178 --> 01:27:16,610 fund this e-Science and yeah, so, so Tony Hey was was the 1201 01:27:16,618 --> 01:27:25,841 person in charge of that. And I'm not sure what exactly 1202 01:27:25,849 --> 01:27:30,740 the intent was. How, how explicitly they wanted 1203 01:27:30,740 --> 01:27:33,740 it to be an interdisciplinary effort. 1204 01:27:33,748 --> 01:27:39,560 Maybe. Maybe they did so in Oxford. 1205 01:27:39,568 --> 01:27:42,786 Yeah. When, when, when they round up 1206 01:27:42,794 --> 01:27:44,576 the usual suspects for high performance computing. 1207 01:27:44,584 --> 01:27:47,009 I was, I was one of the usual suspects. 1208 01:27:47,017 --> 01:27:52,942 And so I was one of the four people who put in the Oxford 1209 01:27:52,950 --> 01:28:00,664 bid, you know, to get OERC initially, I mean, at that point 1210 01:28:00,664 --> 01:28:05,865 Anne Trefethen was Tony's deputy and then she later, you 1211 01:28:05,865 --> 01:28:10,985 know, joined Oxford and and and became head of OERC. 1212 01:28:10,993 --> 01:28:18,696 So I think certainly computer science in those days in Oxford 1213 01:28:18,696 --> 01:28:22,036 was very theoretical, especially after the parallel 1214 01:28:22,036 --> 01:28:25,840 computing people like Bill McColl had left during the 1215 01:28:25,840 --> 01:28:31,840 dot-com era to some extent in Oxford there were tensions 1216 01:28:31,840 --> 01:28:36,422 between engineering and computer science as to where some of 1217 01:28:36,422 --> 01:28:40,590 the more applied computer science activities should go, 1218 01:28:40,590 --> 01:28:46,320 should should go. And then OERC was, was just, 1219 01:28:46,400 --> 01:28:50,440 yeah, another, another location to have such things, you know, 1220 01:28:50,440 --> 01:28:54,560 so, so eventually it made sense for OERC to be merged into 1221 01:28:54,560 --> 01:28:59,640 engineering. I think at the time computer 1222 01:28:59,640 --> 01:29:03,240 science expressed a strong view that they they did not want to 1223 01:29:03,240 --> 01:29:06,480 be the destination. So. 1224 01:29:06,760 --> 01:29:11,200 Anyway, OK, but maybe on to then more the maths side. 1225 01:29:11,240 --> 01:29:18,120 I mean, I'm, I'm intrigued because maybe to explain at more 1226 01:29:18,120 --> 01:29:22,560 of a higher level, if you can, what what are the similarities 1227 01:29:22,560 --> 01:29:26,280 between some of the mathematical finances and maybe CFD? 1228 01:29:26,280 --> 01:29:29,760 What are the, and obviously you're known for which, you 1229 01:29:30,640 --> 01:29:33,320 know, I only understand the basic basics of it, but the sort 1230 01:29:33,320 --> 01:29:36,120 of multilevel Monte Carlo, which I as soon as I saw the 1231 01:29:36,120 --> 01:29:37,560 description, always like a multigrid. 1232 01:29:37,800 --> 01:29:40,400 It always makes me interested that there's some of these, you 1233 01:29:40,400 --> 01:29:44,000 see it with AI today that some of these CFD solutions to 1234 01:29:44,000 --> 01:29:46,440 problems are now being applied to new areas. 1235 01:29:46,680 --> 01:29:48,960 So what? Yeah, what are the sort of high 1236 01:29:48,960 --> 01:29:52,444 level things that makes the link between CFD and maths, I 1237 01:29:52,444 --> 01:29:56,880 guess. So in in mathematical finance, 1238 01:29:56,888 --> 01:30:00,690 there's basically two kinds of methodology. 1239 01:30:00,698 --> 01:30:05,961 You can approach it from a PDE point of view, where in in one 1240 01:30:05,969 --> 01:30:11,085 sense you you have a PDE that describes the evolution of the 1241 01:30:11,093 --> 01:30:14,942 probability density function for, you know, a stock having a 1242 01:30:14,942 --> 01:30:17,320 certain value at the time in the future. 1243 01:30:17,328 --> 01:30:23,350 And there's a corresponding sort of adjoint of that to give the 1244 01:30:23,358 --> 01:30:28,045 value of of various financial options or there's the Monte 1245 01:30:28,053 --> 01:30:32,485 Carlo approach where you simulate lots of these different 1246 01:30:32,485 --> 01:30:36,968 possible future trajectories of of the stock and then and 1247 01:30:36,968 --> 01:30:40,854 then say, OK, given that family of solutions, what's the 1248 01:30:40,854 --> 01:30:43,920 average pay off of your financial option? 1249 01:30:43,928 --> 01:30:50,730 So when I initially moved into finance, I was focused on the 1250 01:30:50,738 --> 01:30:53,974 PDE side because it's basically convection diffusion PDEs. 1251 01:30:53,982 --> 01:30:59,681 And I thought, hey, you know, this is, this is a no brainer. 1252 01:30:59,689 --> 01:31:04,090 I can just bring all the CFD techniques over and, and, and do 1253 01:31:04,090 --> 01:31:09,444 things here. What I quickly found was other 1254 01:31:09,452 --> 01:31:15,272 people have beaten me to it in terms of moving over from CFD. 1255 01:31:15,280 --> 01:31:19,336 So particularly Peter Forsyth, University of Waterloo in 1256 01:31:19,344 --> 01:31:24,304 Canada, who I think Peter had come out of the oil reservoir 1257 01:31:24,312 --> 01:31:26,844 CFD area if I remember correctly. 1258 01:31:26,852 --> 01:31:33,456 And so he he did a lot of the pioneering work in terms of PDE 1259 01:31:33,456 --> 01:31:37,760 methods, you know, numerical methods for for finance. 1260 01:31:37,768 --> 01:31:42,590 And so there wasn't, as it happened so much, you know, left 1261 01:31:42,590 --> 01:31:45,720 for, for me, as I had maybe thought. 1262 01:31:45,728 --> 01:31:52,624 And then I mean this, this was over a relatively short period 1263 01:31:52,632 --> 01:31:55,768 of time. I thought that I would probably 1264 01:31:55,768 --> 01:31:59,672 have to teach a course on Monte Carlo methods, you know, 1265 01:31:59,680 --> 01:32:04,868 because, you know, we, we, we had still have an MSc in 1266 01:32:04,876 --> 01:32:06,785 mathematical and computational finance. 1267 01:32:06,793 --> 01:32:11,605 And I, I teach numerics on that. And you know, we would need to 1268 01:32:11,613 --> 01:32:17,387 teach them about both sides. And so I took a 2 day course put 1269 01:32:17,387 --> 01:32:22,504 on by a couple of professors from Columbia University in, in, 1270 01:32:22,504 --> 01:32:28,344 in the US, put on in, in London for, for London finance. 1271 01:32:28,344 --> 01:32:31,120 People managed to, to, to 1272 01:32:31,120 --> 01:32:35,960 convince the, the, the department to, to pay my fees 1273 01:32:35,960 --> 01:32:39,040 for that. They, they, they, they gave me a 1274 01:32:39,040 --> 01:32:44,040 50% discount as an academic, but Even so, it was a costlier 1275 01:32:44,200 --> 01:32:45,960 course to attend. Anyway. 1276 01:32:46,560 --> 01:32:50,600 So I went to this course and they were teaching me about, 1277 01:32:50,720 --> 01:32:56,200 about, about Monte Carlo methods and talking about doing 1278 01:32:56,320 --> 01:33:01,520 sensitivity calculations. And I went up and talked to Paul 1279 01:33:01,520 --> 01:33:04,880 Glasserman, the lead guy in in a coffee break and said, well, 1280 01:33:05,240 --> 01:33:09,360 this is all fascinating, but I presume that of course you, you 1281 01:33:09,400 --> 01:33:13,160 actually use adjoint methods to to compute these sensitivities 1282 01:33:13,320 --> 01:33:17,520 more more efficiently. To which he said, what? 1283 01:33:19,920 --> 01:33:24,840 So yeah. So just pure, pure luck. 1284 01:33:24,840 --> 01:33:29,920 I was able to introduce adjoint methods to the financial Monte 1285 01:33:29,920 --> 01:33:34,400 Carlo community for doing sensitivity calculations. 1286 01:33:34,640 --> 01:33:40,806 So I did a paper with with, with Paul Glasserman in a finance, 1287 01:33:40,806 --> 01:33:45,983 so industry magazine really, rather than as a proper 1288 01:33:45,983 --> 01:33:48,400 academic journal. Because from my point of view 1289 01:33:48,400 --> 01:33:50,680 there was absolutely nothing new mathematically. 1290 01:33:50,680 --> 01:33:53,746 This was just a new application. 1291 01:33:53,754 --> 01:34:00,776 This is a paper that went, went, went by the title of Smoking 1292 01:34:00,784 --> 01:34:03,996 Adjoints, which helped helped. Its notoriety. 1293 01:34:04,004 --> 01:34:09,909 This is a journal or, or trade journal that liked puns in their 1294 01:34:09,909 --> 01:34:13,215 titles. And so I would never dare do 1295 01:34:13,223 --> 01:34:16,640 that in an academic journal. But anyway, so yeah. 1296 01:34:16,648 --> 01:34:21,619 So I so I got known for, you know, the adjoint work, which is 1297 01:34:21,619 --> 01:34:26,205 really sort of taken over in, in in the finance sector. 1298 01:34:26,213 --> 01:34:29,286 And then the multilevel Monte Carlo. 1299 01:34:29,294 --> 01:34:33,366 Yes, I mean, you're right to take, you know, the analogy to 1300 01:34:33,374 --> 01:34:36,420 multigrid. I mean, multigrid is such a 1301 01:34:36,428 --> 01:34:41,260 fundamental part of CFD that it was natural in getting into 1302 01:34:41,268 --> 01:34:44,824 Monte Carlo methods to think, well, is there anything 1303 01:34:44,832 --> 01:34:48,108 analogous that that that we can do here? 1304 01:34:48,116 --> 01:34:54,087 And so, yeah, I came up with with, you know, the multilevel 1305 01:34:54,095 --> 01:34:56,600 idea. And it's one of those things 1306 01:34:56,608 --> 01:35:00,562 that like multigrid itself, I mean, it's such a simple idea. 1307 01:35:00,570 --> 01:35:04,391 It really ought to have been thought of ages before. 1308 01:35:04,399 --> 01:35:08,154 But because I came in from a different background, you know, 1309 01:35:08,154 --> 01:35:12,982 this was, this was part of my toolkit, you know, it, it, it 1310 01:35:12,990 --> 01:35:17,230 was a fairly natural thing for, for, for me to do. 1311 01:35:17,238 --> 01:35:21,226 And so I've kind of been living off that and extensions of that, 1312 01:35:21,226 --> 01:35:23,680 you know, for the last 15 years. 1313 01:35:26,880 --> 01:35:30,200 And what about the computing side that the GPU side, you 1314 01:35:30,200 --> 01:35:33,819 know, is that is that being just because of, of, of an 1315 01:35:33,819 --> 01:35:36,448 interest? How much of that and forgive me 1316 01:35:36,448 --> 01:35:40,760 for for not knowing, but how, how much is that shaped Also on 1317 01:35:40,760 --> 01:35:43,046 the maths side, how, how much is that acceleration? 1318 01:35:43,054 --> 01:35:47,757 You know GPUs to CFD is well known but is it a similar idea 1319 01:35:47,757 --> 01:35:53,535 on the maths side? So, I mean, most maths research 1320 01:35:53,535 --> 01:35:56,560 just doesn't need lots of compute power. 1321 01:35:56,568 --> 01:36:01,340 I mean, these these days more and more, you know, our, our 1322 01:36:01,348 --> 01:36:06,975 students do things in Python And, you know, maybe some will, 1323 01:36:06,975 --> 01:36:11,800 will use the JAX package within Python to get performance. 1324 01:36:12,120 --> 01:36:17,280 But a lot of work. No, nobody worries about, you 1325 01:36:17,280 --> 01:36:22,960 know, performance, you know, so I'm, I'm kind of an unusual, 1326 01:36:23,560 --> 01:36:27,200 yeah. So I, I, I still teach my, my 1327 01:36:27,200 --> 01:36:30,240 CUDA course every year with, with Wes. 1328 01:36:31,840 --> 01:36:35,160 You know, a few weeks ago I taught in, you know, a one day 1329 01:36:35,160 --> 01:36:40,760 OpenMP, you know, some mini course for, for PhD students 1330 01:36:40,760 --> 01:36:44,360 just to introduce them to this forgetting performance. 1331 01:36:44,560 --> 01:36:48,640 But there's very, very few students who are particularly 1332 01:36:48,640 --> 01:36:52,640 interested in, in that. It is an interesting question. 1333 01:36:52,640 --> 01:36:55,280 You know, where, where does that kind of work belong? 1334 01:36:55,280 --> 01:36:59,920 You know, to what extent does it belong in computer science or 1335 01:36:59,920 --> 01:37:09,080 maths or engineering? You know, I guess I'm somehow a 1336 01:37:09,080 --> 01:37:14,720 product of, of my time and it's not clear that there will be a 1337 01:37:14,720 --> 01:37:19,440 new generation of people like me, at least not in maths. 1338 01:37:19,680 --> 01:37:21,600 Yes. I'm not sure where, where the 1339 01:37:21,600 --> 01:37:26,680 next generation of me sort of lives, you know, so, so there's 1340 01:37:26,680 --> 01:37:29,200 Wes. Wes is in, in engineering now, 1341 01:37:29,640 --> 01:37:33,160 you know, having moved with, with, with OERC. 1342 01:37:33,760 --> 01:37:39,600 So, you know, Wes is now kind of my successor within the 1343 01:37:39,600 --> 01:37:47,600 university. He's, he's now Mr. HPC, you 1344 01:37:47,640 --> 01:37:53,320 know, there will be individuals in departments like physics and 1345 01:37:53,320 --> 01:37:57,784 chemistry and biochemistry who, who have their expertise in 1346 01:37:57,784 --> 01:38:01,897 HPC, but we don't really, you know, we're not a community as 1347 01:38:01,897 --> 01:38:09,160 such now, I would say, and we're still struggling a bit to 1348 01:38:10,000 --> 01:38:13,480 organise graduate teaching across the university. 1349 01:38:13,880 --> 01:38:19,680 I mean, this is something where Oxford and I think Cambridge 1350 01:38:19,680 --> 01:38:24,360 are, are poor compared to our American counterparts, where 1351 01:38:24,720 --> 01:38:27,160 you'll, you'll have graduate courses offered by one 1352 01:38:27,160 --> 01:38:30,880 department taken by people from across the university. 1353 01:38:31,520 --> 01:38:33,640 Yeah. We, we don't do enough of that 1354 01:38:34,160 --> 01:38:37,280 Our, our, our, our CUDA course is, is unusual. 1355 01:38:37,280 --> 01:38:43,040 I mean, this this year, I think we're currently up to about 1356 01:38:43,040 --> 01:38:48,800 well over 100 people signed up, of whom 65 are Oxford people 1357 01:38:48,800 --> 01:38:54,914 and another 35 externals, you know, and so the Oxford people 1358 01:38:54,914 --> 01:38:59,708 do come from across the university, but that, that's 1359 01:38:59,708 --> 01:39:03,188 very unusual in, in the Oxford setup. 1360 01:39:03,196 --> 01:39:07,544 I'm, I'm, I'm trying to get more of that happening. 1361 01:39:07,552 --> 01:39:13,810 You know that we we have a much more systematic training in, in 1362 01:39:13,810 --> 01:39:18,886 advanced computing because you know there are needs across the 1363 01:39:18,886 --> 01:39:22,519 university. I mean, I have to ask the 1364 01:39:22,527 --> 01:39:25,450 obvious question, which is you know, you've you've pivoted once 1365 01:39:25,450 --> 01:39:28,600 I guess to the maths were you, were you? 1366 01:39:28,608 --> 01:39:33,112 And are you still tempted on the AI given that it seems your 1367 01:39:33,120 --> 01:39:36,420 skills are probably absolutely perfectly aligned, which is 1368 01:39:36,428 --> 01:39:39,588 strong maths and tech, engineering and HPC? 1369 01:39:39,596 --> 01:39:46,558 Yeah, So the quick answer is no. So it, it, it is curious. 1370 01:39:46,566 --> 01:39:51,376 I've I've got so three points of contact with with AI. 1371 01:39:51,384 --> 01:39:58,590 I've, I've got the, the GPUs I've got, you know, stochastic- 1372 01:39:58,590 --> 01:40:04,035 gradient methods are very close to the stuff I'm doing in Monte 1373 01:40:04,035 --> 01:40:07,498 Carlo. And there's the adjoint, you 1374 01:40:07,506 --> 01:40:11,436 know, so. I mean, certainly at one point 1375 01:40:11,444 --> 01:40:15,492 in the past, if you looked at the source code for PyTorch, 1376 01:40:15,500 --> 01:40:20,416 there were more references to my adjoint publications than than 1377 01:40:20,424 --> 01:40:26,822 any other academic, you know, So it is, it is interesting how, 1378 01:40:26,830 --> 01:40:30,896 how, how the adjoint stuffs got got, got picked up. 1379 01:40:30,904 --> 01:40:37,802 But, but no, no, it, it AI is for, for for a new generation. 1380 01:40:37,810 --> 01:40:40,970 No, I'm, I'm, I'm not doing another pivot. 1381 01:40:40,978 --> 01:40:47,760 I'm, I'm I'm happy to keep things going on on the HPC side. 1382 01:40:47,760 --> 01:40:50,160 So I will continue doing things 1383 01:40:50,160 --> 01:40:55,120 with, with, with GPUs I've, I'm, I'm interested in the 1384 01:40:55,120 --> 01:40:59,560 potential of FPGAs for doing finance calculations. 1385 01:41:01,520 --> 01:41:06,040 It it it's amusing when when I arrived in in 92 in computer 1386 01:41:06,040 --> 01:41:11,280 science, one of my colleagues then in in computer science told 1387 01:41:11,280 --> 01:41:14,917 me, you know, this is wonderful new technology Mike called 1388 01:41:14,917 --> 01:41:18,080 FPGAs is going to absolutely revolutionize everything that 1389 01:41:18,080 --> 01:41:19,720 you're doing. You know, you really need to 1390 01:41:19,720 --> 01:41:25,600 learn about it, and that's kind of still the message. 1391 01:41:25,840 --> 01:41:27,680 Yeah. I was just about to say they 1392 01:41:27,680 --> 01:41:29,960 haven't been right in the last 30 years. 1393 01:41:30,160 --> 01:41:32,600 Doesn't mean that they may not be right at some point in the 1394 01:41:32,600 --> 01:41:33,920 next 30 years. So. 1395 01:41:35,680 --> 01:41:37,400 And there's a story for GPUs from that. 1396 01:41:37,400 --> 01:41:39,760 It is that, you know, you were there early in the day, But I, I 1397 01:41:39,760 --> 01:41:43,200 guess it's Jensen always says that, you know, it's, it's taken 1398 01:41:43,200 --> 01:41:46,800 like 30 years to sort of get to this point where there's been 1399 01:41:46,800 --> 01:41:50,400 this massive, you know, use because of because of AI, but I 1400 01:41:50,400 --> 01:41:56,240 guess it was because of people who are early adopters it like 1401 01:41:56,240 --> 01:41:57,560 yourself. It's it's been a slow 1402 01:41:57,560 --> 01:42:00,640 progression, but suddenly it's all come together. 1403 01:42:00,640 --> 01:42:02,280 Hasn't. It, I mean, it has been 1404 01:42:02,280 --> 01:42:08,280 fascinating being sort of in, in, in the company of, of NVIDIA 1405 01:42:08,280 --> 01:42:13,640 during this whole evolution. You know, that, you know, I went 1406 01:42:14,160 --> 01:42:21,920 to all the GTCs in, in the early days and I, I remember being 1407 01:42:21,920 --> 01:42:29,040 there, you know, when ImageNet— you know, sorry, not 1408 01:42:29,040 --> 01:42:37,944 ImageNet—when AlexNet won the competition, yeah, for, you 1409 01:42:37,944 --> 01:42:40,940 know, image recognition, image classification. 1410 01:42:40,948 --> 01:42:46,815 And then the next year, the top 10 competitors won with, with 1411 01:42:46,823 --> 01:42:51,900 GPUs, you know, and, and you know, the way that NVIDIA 1412 01:42:51,908 --> 01:42:55,960 pivoted was, was impressive, but they were always on the lookout 1413 01:42:55,960 --> 01:43:00,226 for the killer application. You know, so right at the 1414 01:43:00,234 --> 01:43:04,200 beginning they thought, at least this is my impression that they 1415 01:43:04,200 --> 01:43:06,994 thought that computational finance might be the killer 1416 01:43:07,002 --> 01:43:09,484 application. And so that's why they were 1417 01:43:09,492 --> 01:43:12,980 interested in in what I did in implementing, you know, the 1418 01:43:12,988 --> 01:43:16,308 first random number generator on, on, on CUDA. 1419 01:43:16,316 --> 01:43:21,800 You know, I think in, I think at that point, the value 1420 01:43:21,808 --> 01:43:25,580 proposition wasn't sufficient to persuade the banks to take 1421 01:43:25,588 --> 01:43:30,118 highly paid quants to, to rewrite all of the software. 1422 01:43:30,126 --> 01:43:34,735 And so the adoption rate was, was, was slow. 1423 01:43:34,743 --> 01:43:39,980 But yeah, yeah, the company was always on the lookout for that 1424 01:43:39,988 --> 01:43:42,740 killer application. And, you know, when when they 1425 01:43:42,748 --> 01:43:47,082 saw, you know, those early days in AI, they thought, yeah, let's 1426 01:43:47,082 --> 01:43:49,680 let's let's double down on this. 1427 01:43:49,680 --> 01:43:53,040 And, you know, first, first on the software side and then 1428 01:43:53,040 --> 01:43:55,040 increasingly on the hardware side as well. 1429 01:43:55,360 --> 01:43:57,320 Yeah, I mean, it is. Fascinating. 1430 01:43:58,200 --> 01:44:05,800 I just find that the CFD in some ways is still a niche industry 1431 01:44:05,800 --> 01:44:08,200 at a global level. You know the amount of money and 1432 01:44:08,200 --> 01:44:11,000 the amount of people. But I do always find it 1433 01:44:11,000 --> 01:44:16,360 fascinating that so many things have originated in CFD. 1434 01:44:16,440 --> 01:44:19,800 I mean, Ian Buck's PhD was CFD, right? 1435 01:44:19,800 --> 01:44:22,240 CUDA, as far as I'm aware, there's like a fluid dynamics 1436 01:44:23,440 --> 01:44:27,259 application, you know, like with yourself with the early days 1437 01:44:27,259 --> 01:44:29,600 of CFD. OK, now you've it's. 1438 01:44:29,600 --> 01:44:32,760 It's funny how I guess because it was one of the original hard 1439 01:44:32,760 --> 01:44:34,400 problems to solve. Yes. 1440 01:44:34,400 --> 01:44:42,720 I mean CFD, I guess was a prime driver for HPC for for a long 1441 01:44:42,720 --> 01:44:54,480 period, you know, So yeah, I mean, both in terms of as it 1442 01:44:54,480 --> 01:45:00,720 were open, open CFD and all, all, all of you know, the 1443 01:45:00,720 --> 01:45:04,400 nuclear weapons stuff as well. Yeah, which is sort of related 1444 01:45:04,400 --> 01:45:10,200 very closely, you know. So whereas now you know, within 1445 01:45:10,200 --> 01:45:15,520 the UK the biggest computers or the most IT spend is in the 1446 01:45:15,520 --> 01:45:20,280 banks, you know, you know, So you know, the amount of money 1447 01:45:20,280 --> 01:45:25,000 that Rolls-Royce spends on compute per year, I would 1448 01:45:25,000 --> 01:45:29,680 imagine is, is less than any any one of the big banks in 1449 01:45:29,680 --> 01:45:32,952 London. I don't know that for a fact, 1450 01:45:32,960 --> 01:45:34,486 but no. Probably. 1451 01:45:34,494 --> 01:45:39,140 That's the case, you know, of sense. 1452 01:45:39,148 --> 01:45:43,142 Yeah. You know, so these days it it's 1453 01:45:43,142 --> 01:45:45,788 the money in AI which is driving the hardware 1454 01:45:45,788 --> 01:45:47,520 development, you know, clearly. 1455 01:45:47,840 --> 01:45:53,960 And so, you know, you know, CFD is no longer that driver. 1456 01:45:57,120 --> 01:46:00,720 I guess CFD was also driving a lot of development of numerical 1457 01:46:00,720 --> 01:46:05,080 methods within academia, you know, and academia tends to be 1458 01:46:05,080 --> 01:46:08,240 at the sort of bleeding edge of the technology. 1459 01:46:08,320 --> 01:46:13,480 You know, again, all of those cheap bodies, Yeah. 1460 01:46:13,920 --> 01:46:18,520 Bright, bright young minds and well, seeing just how much they 1461 01:46:18,520 --> 01:46:21,320 they can squeeze out of this, this new hardware. 1462 01:46:25,560 --> 01:46:28,360 But yes, what are the big drivers these days other than 1463 01:46:28,360 --> 01:46:29,560 AI? Yeah, Yeah. 1464 01:46:29,560 --> 01:46:32,920 I mean that so much is is is based on that now. 1465 01:46:34,720 --> 01:46:38,840 Yeah, that's why I find it interesting that the and it's a 1466 01:46:38,840 --> 01:46:41,720 bit more of a controversial topic, which is the convergence 1467 01:46:42,240 --> 01:46:45,600 or the use of AI techniques for some of these disciplines. 1468 01:46:45,600 --> 01:46:53,720 You know, whether just as there's influences of using CFD 1469 01:46:53,720 --> 01:47:01,520 or numerical methods for AI, is AI got any use within CFD or or 1470 01:47:01,640 --> 01:47:05,960 you know, do you see it in the maths world that the actual use 1471 01:47:05,960 --> 01:47:11,000 of AI techniques or is it controversial or not sort of 1472 01:47:11,000 --> 01:47:12,960 proven? You know, just as you were 1473 01:47:12,960 --> 01:47:17,000 bringing in like techniques from CFD to AI applications in the 1474 01:47:17,000 --> 01:47:23,231 finance. So I mean, in terms of AI within 1475 01:47:23,231 --> 01:47:33,450 CFD in general, I would say I'm a sceptic, but but to some 1476 01:47:33,450 --> 01:47:36,612 extent that may be that I just haven't spent enough time to 1477 01:47:36,612 --> 01:47:39,080 actually see what is is happening. 1478 01:47:40,120 --> 01:47:45,840 I think potentially the idea of an AI-based turbulence model 1479 01:47:46,160 --> 01:47:49,960 might, might make sense. I mean, turbulence modelling is 1480 01:47:49,960 --> 01:47:54,360 such a challenging topic and there's so little progress I 1481 01:47:54,360 --> 01:48:01,120 think has been made in the last, you know, 25 years that AI may 1482 01:48:01,120 --> 01:48:08,080 may be the answer there. I think in I think, you know, 1483 01:48:08,080 --> 01:48:10,820 for detailed CFD. I think it it, it's going to 1484 01:48:10,820 --> 01:48:14,695 need to continue to be traditional CFD methods. 1485 01:48:14,703 --> 01:48:21,060 You know that whether AI can then get trained on the results 1486 01:48:21,060 --> 01:48:26,205 of a large number of simulations such that it does a 1487 01:48:26,205 --> 01:48:30,680 pretty good proxy for the purposes of of design 1488 01:48:30,680 --> 01:48:37,160 optimization. Maybe I guess that doesn't 1489 01:48:37,160 --> 01:48:40,880 particularly interest or excite me. 1490 01:48:41,400 --> 01:48:46,240 But again, maybe I'm, I'm just showing my, my, my age. 1491 01:48:47,680 --> 01:48:57,520 I, I am interested in the impact of AI on research more 1492 01:48:57,520 --> 01:49:03,000 generally. So, you know, I'm really very 1493 01:49:03,000 --> 01:49:10,320 impressed by the capabilities of the latest ChatGPT, you know, 1494 01:49:10,400 --> 01:49:15,520 and you know, I think it really has improved hugely, you know, 1495 01:49:15,680 --> 01:49:18,480 over the last two years, say, you know, so the pace of 1496 01:49:18,480 --> 01:49:25,208 development is, is fascinating, you know, so, so I'm doing 1497 01:49:25,208 --> 01:49:30,136 quite a few experiments in different settings just just 1498 01:49:30,136 --> 01:49:36,880 to understand what it can do. So, so you know, I mean, just 1499 01:49:36,880 --> 01:49:44,000 yesterday I was getting it to, you know, tell me all about, you 1500 01:49:44,080 --> 01:49:50,400 know, Brownian motion, spatial white noise, space-time white 1501 01:49:50,400 --> 01:49:53,520 noise for, you know, for stochastic modelling. 1502 01:49:54,400 --> 01:49:58,523 And so put together a whole lot of information for me to, to, 1503 01:49:58,523 --> 01:50:03,600 to give to an MSc student who's doing a project with me, you 1504 01:50:03,600 --> 01:50:08,360 know, all all the way to it, you know, telling me about the 1505 01:50:08,360 --> 01:50:13,640 stochastic heat equation, providing me with some sample 1506 01:50:13,640 --> 01:50:19,400 code, implementing it in, in a simple, you know, finite 1507 01:50:19,400 --> 01:50:25,560 difference approximation. It, it tried to set it up to, to 1508 01:50:25,560 --> 01:50:28,240 use multilevel. So understood about multilevel 1509 01:50:28,520 --> 01:50:35,040 it it didn't quite get things right, but but it, you know, did 1510 01:50:35,040 --> 01:50:38,960 surprisingly well, you know, so I really wonder where we're 1511 01:50:38,960 --> 01:50:42,560 going to be in five or ten years time, you know, so. 1512 01:50:42,880 --> 01:50:48,360 Yeah. I think it's entirely possible. 1513 01:50:48,360 --> 01:50:52,760 I might even go as far as to say likely, that 10 years from now 1514 01:50:53,120 --> 01:50:59,720 you will have a proof assistant that is capable of looking at a 1515 01:50:59,720 --> 01:51:04,320 theorem and proof given the necessary background 1516 01:51:04,320 --> 01:51:09,200 information, and it won't be able to say with certainty that 1517 01:51:09,200 --> 01:51:16,920 the proof is correct. But it will on many occasions be 1518 01:51:16,920 --> 01:51:21,280 able to highlight bits in the proof that look dodgy or don't 1519 01:51:21,280 --> 01:51:29,280 like wrong. So yeah, So I, I, I kind of 1520 01:51:29,280 --> 01:51:33,400 think that AI, you know, you should absolutely never trust it 1521 01:51:33,400 --> 01:51:35,800 100%. You should always look at what 1522 01:51:35,800 --> 01:51:38,800 it produces. But the fact that it gets some 1523 01:51:38,800 --> 01:51:42,480 things wrong doesn't matter if it also gets some things right. 1524 01:51:42,880 --> 01:51:47,760 And it is, you know, it's very helpful. 1525 01:51:47,760 --> 01:51:53,666 I mean, another another thing I did with it a few days ago, 1526 01:51:53,666 --> 01:51:59,960 it's always irritated me with programming Intel CPUs that at 1527 01:51:59,960 --> 01:52:03,480 times their compiler is very poor at vectorization. 1528 01:52:03,760 --> 01:52:07,960 And so if you really want to get performance, you have to work 1529 01:52:07,960 --> 01:52:11,720 with the vector intrinsics, you know, the AVX-512 intrinsics. 1530 01:52:13,040 --> 01:52:18,200 And it's irritated me the fact that Intel doesn't provide you 1531 01:52:18,800 --> 01:52:23,680 with a C++ class that has operator overloading, you know, 1532 01:52:24,080 --> 01:52:26,720 to go along with all of those intrinsics. 1533 01:52:28,480 --> 01:52:32,200 And so I asked ChatGPT about this and it, it produced me 1534 01:52:32,200 --> 01:52:37,568 with, you know, it gave me a C++ class definition with, with all 1535 01:52:37,568 --> 01:52:40,932 of the operator overloading, you know, and I'm, I'm not a 1536 01:52:40,932 --> 01:52:43,530 good enough C++ programmer that I could have done that 1537 01:52:43,530 --> 01:52:45,800 myself. So just saved me, you know, 1538 01:52:46,680 --> 01:52:56,040 incredible amount of time. So things that in a sense are 1539 01:52:56,720 --> 01:53:01,160 are routine enough for a large group of people now become 1540 01:53:01,160 --> 01:53:05,800 automatically available to everyone else who isn't part of 1541 01:53:05,800 --> 01:53:11,080 that, you know, So, so as a productivity tool, you know, it 1542 01:53:11,080 --> 01:53:17,440 really is capable. So I'm, I'm doing this, I mean, 1543 01:53:17,440 --> 01:53:21,520 partly from, for my own benefit, partly also to just spread the 1544 01:53:21,520 --> 01:53:25,840 word amongst my colleagues, you know, to give them a range of 1545 01:53:25,840 --> 01:53:30,920 examples of, look, this is what AI can do for you. 1546 01:53:31,680 --> 01:53:35,640 Yeah. I mean, I agree. 1547 01:53:35,640 --> 01:53:40,360 It's, it's amazing what it, it's and and it's, it almost does 1548 01:53:40,360 --> 01:53:43,480 need people to evangelise or show it because if you haven't 1549 01:53:43,480 --> 01:53:45,680 seen it, you don't know. But once you see that, Oh my 1550 01:53:45,680 --> 01:53:47,760 God, this is. Yeah. 1551 01:53:47,760 --> 01:53:51,320 So I'm, I'm, I'm now kind of getting into that evangelism 1552 01:53:51,680 --> 01:53:56,960 group. Yes, Yeah. 1553 01:53:57,240 --> 01:54:02,560 It it it, it is fascinating. So here's maybe a question for 1554 01:54:02,560 --> 01:54:08,440 you as, as we get towards the end of this discussion, which 1555 01:54:08,440 --> 01:54:13,400 I'm sure we could carry on for hours, because I, I, yeah, 1556 01:54:13,400 --> 01:54:16,200 you've got so many interesting things that we didn't talk 1557 01:54:16,200 --> 01:54:18,000 about. But I'm I'm conscious of like 1558 01:54:18,000 --> 01:54:21,636 your, your time, but maybe more philosophical question, which 1559 01:54:21,636 --> 01:54:27,360 is if you were if you were now an undergraduate. 1560 01:54:27,960 --> 01:54:31,680 Yeah. In today's world, what would be 1561 01:54:31,680 --> 01:54:38,640 your advice on a career trajectory, what to focus on? 1562 01:54:38,640 --> 01:54:41,240 Like, I know it's a very difficult question, but seeing 1563 01:54:41,240 --> 01:54:44,120 everything you've done, is there anything you would advise now, 1564 01:54:44,720 --> 01:54:48,920 somebody who is, yeah, 18, 19 or doing an undergraduate and is 1565 01:54:48,920 --> 01:54:54,040 thinking about, you know, academia, industry, Is it good 1566 01:54:54,040 --> 01:54:57,400 to work with industry to get that understanding? 1567 01:54:59,480 --> 01:55:11,920 Yeah. Yeah, it's tough because I mean, 1568 01:55:11,920 --> 01:55:14,920 I think the main advice is to try to do something that you 1569 01:55:14,920 --> 01:55:17,120 enjoy doing. You know, you know, you know, 1570 01:55:17,320 --> 01:55:20,280 you know, we don't want everyone to be doing the same things. 1571 01:55:20,280 --> 01:55:25,760 Everyone should kind of pursue, you know what, what, what they 1572 01:55:25,760 --> 01:55:31,640 like. If, if I mean if it was me again 1573 01:55:31,640 --> 01:55:37,474 now, you know, thinking of, you know, where I was as a 1574 01:55:37,474 --> 01:55:42,662 teenager, you know, I would, I would still probably be, be 1575 01:55:42,662 --> 01:55:50,120 heading in into mathematics. You know, I would. 1576 01:55:50,440 --> 01:55:53,440 I think by my nature I was always interested in 1577 01:55:53,440 --> 01:55:57,320 applications. So I think as a teenager I was 1578 01:55:57,320 --> 01:56:00,360 interested in maths because it helped me with applications in 1579 01:56:00,360 --> 01:56:03,600 physics that that was kind of the motivation. 1580 01:56:05,520 --> 01:56:09,640 I mean maths now one of the areas of growth I think is in 1581 01:56:09,640 --> 01:56:14,160 terms of mathematical modelling in medicine. 1582 01:56:14,800 --> 01:56:20,680 So, so I do think if, if, if I was maybe an undergraduate now 1583 01:56:20,800 --> 01:56:26,760 and, and thinking about where I wanted to head for PhD, you 1584 01:56:26,760 --> 01:56:32,080 know, computational methods applied in, in the sort of 1585 01:56:32,080 --> 01:56:37,360 medical area, I think may, may be an area right right now. 1586 01:56:37,360 --> 01:56:40,040 It's still challenging with the whole funding situation. 1587 01:56:40,040 --> 01:56:42,960 So, you know, we have difficulties linking up medical 1588 01:56:42,960 --> 01:56:47,120 departments with science departments and, you know, the 1589 01:56:47,120 --> 01:56:51,160 whole way the UK funding mechanisms work, but at least we 1590 01:56:51,160 --> 01:56:54,080 don't have the problems that, that, that the US has for us. 1591 01:56:57,240 --> 01:57:01,140 So yes, I mean that that's maybe what, what I would end up 1592 01:57:01,140 --> 01:57:05,320 doing if, you know, I was, I was 18 again. 1593 01:57:05,320 --> 01:57:09,240 Now I don't know. But I think the main thing is, 1594 01:57:09,360 --> 01:57:14,960 is to try to enjoy what it is you do. 1595 01:57:14,960 --> 01:57:23,160 And, and you know, don't don't worry too much about, about the 1596 01:57:23,160 --> 01:57:26,520 future, you know, you know, you know, when I started as, as an 1597 01:57:26,520 --> 01:57:28,400 undergraduate at Cambridge, I really didn't know what I was 1598 01:57:28,400 --> 01:57:31,520 going to be doing at the end. I think, I think now students 1599 01:57:31,520 --> 01:57:34,560 start as undergrads with much more of an idea of what they 1600 01:57:34,560 --> 01:57:38,120 want to do at the end of it. And maybe there's. 1601 01:57:38,200 --> 01:57:40,440 Pressure. Yeah, there is more pressure 1602 01:57:40,440 --> 01:57:43,360 now. I mean, back when I was an 1603 01:57:43,360 --> 01:57:48,240 undergraduate, so few people went to university that in a 1604 01:57:48,240 --> 01:57:53,520 sense you were, you felt, you know, guaranteed that you you 1605 01:57:53,520 --> 01:57:55,120 would get a good job at the end of it. 1606 01:57:55,120 --> 01:58:00,160 And people do do not feel that guarantee now, you know, so, so, 1607 01:58:00,160 --> 01:58:05,000 so I do, you know, I do understand and appreciate that, 1608 01:58:05,000 --> 01:58:12,920 you know, I certainly wouldn't discourage people from doing, 1609 01:58:13,280 --> 01:58:15,880 you know, you know, pursuing interests in programming. 1610 01:58:16,160 --> 01:58:18,960 I know, I know there's talk about AI is going to do away 1611 01:58:18,960 --> 01:58:21,640 with all these programmers. No, I don't think so. 1612 01:58:21,640 --> 01:58:26,560 I think, you know, you, you, you, you still need people with 1613 01:58:27,280 --> 01:58:30,200 with programming skills. I think I would definitely 1614 01:58:30,200 --> 01:58:34,400 encourage people to develop their AI using skills, you know, 1615 01:58:34,520 --> 01:58:41,600 so and engage with these AI tools, you know, because they 1616 01:58:41,880 --> 01:58:48,120 they are hugely useful, you know, and yet at the same time, 1617 01:58:48,360 --> 01:58:51,880 you have to have the critical skills to look at what they 1618 01:58:51,880 --> 01:58:55,600 produce and ask whether it's right or not, you know, so, so 1619 01:58:55,600 --> 01:59:00,040 you absolutely still do need to have very strong understanding 1620 01:59:00,040 --> 01:59:04,240 of your technical area, but given that, you know, the AI 1621 01:59:04,240 --> 01:59:09,040 tools can, you know, make you more productive. 1622 01:59:10,760 --> 01:59:14,320 Would would you agree as well that I mean it's easy for both 1623 01:59:14,320 --> 01:59:21,520 of us to say this, but that the if you want to guarantee a job 1624 01:59:21,520 --> 01:59:26,960 for the next 60 years, having a maths and/or engineering 1625 01:59:26,960 --> 01:59:30,360 background allows you to turn your hand to almost any 1626 01:59:30,360 --> 01:59:33,600 problems. And it's, it's quite a general 1627 01:59:34,400 --> 01:59:38,200 skill set that is desirable because you could turn to maths, 1628 01:59:38,200 --> 01:59:40,564 you could turn to CFD, you could turn to biology, you could 1629 01:59:40,564 --> 01:59:43,870 turn, you know, there's always a need for the sort of 1630 01:59:43,870 --> 01:59:46,668 mathematical simulation side of things. 1631 01:59:46,676 --> 01:59:53,412 I mean, I, I, I think maths is rightly viewed as, as a subject 1632 01:59:53,412 --> 01:59:56,910 that has, you know, lots of real world applications. 1633 01:59:56,918 --> 02:00:02,220 I think, I think it's viewed more so now than probably 30 1634 02:00:02,228 --> 02:00:06,792 years ago, you know, so you know, you know, because of data 1635 02:00:06,792 --> 02:00:10,325 science, because of AI. You know, the importance of 1636 02:00:10,333 --> 02:00:12,885 maths I think is much better understood. 1637 02:00:12,893 --> 02:00:18,608 We still at some time, you know, have have trouble convincing 1638 02:00:18,616 --> 02:00:22,760 politicians of, of, of that from the point of view of funding. 1639 02:00:22,768 --> 02:00:27,769 You know, we, we, we do feel a bit hard done by, shall we say, 1640 02:00:27,769 --> 02:00:31,186 in terms of supporting, you know, you know, the underpinning 1641 02:00:31,186 --> 02:00:34,428 maths that's so important for so many applications. 1642 02:00:34,436 --> 02:00:41,540 So yes, I think, you know, a maths education gives you a very 1643 02:00:41,540 --> 02:00:48,715 firm foundation for, for life. But again, I would, I would 1644 02:00:48,723 --> 02:00:53,332 emphasise doing things that you enjoy, you know, you know, I 1645 02:00:53,340 --> 02:00:58,400 think, you know, there are people on, on the, you know, the 1646 02:00:58,400 --> 02:01:04,318 creative side where I don't see AI sort of taking over. 1647 02:01:04,326 --> 02:01:10,760 I mean, whether it's creative in the arts or being a chef or 1648 02:01:10,768 --> 02:01:16,242 being a singer, you know, I mean, yes, I mean, we'll see 1649 02:01:16,250 --> 02:01:21,180 where, where, where AI gets to in terms of, you know, 1650 02:01:21,188 --> 02:01:23,488 generating your pop songs and things. 1651 02:01:23,496 --> 02:01:26,942 I don't know, yes. I mean, maybe, maybe it'll turn 1652 02:01:26,942 --> 02:01:30,749 out yeah, but people will, will, will still want live 1653 02:01:30,757 --> 02:01:31,836 performances and stuff. No. 1654 02:01:31,844 --> 02:01:36,202 So, so I think there's a lot of stuff on the creative side which 1655 02:01:36,202 --> 02:01:39,629 will will be important for the future. 1656 02:01:39,637 --> 02:01:45,216 But, you know, there's only so much that AI can do. 1657 02:01:45,224 --> 02:01:50,496 You know, I, I think it, I think it's best to think of it as a 1658 02:01:50,504 --> 02:01:53,245 productivity tool that it's important you engage with so 1659 02:01:53,253 --> 02:01:57,121 that you you have those skills to be productive. 1660 02:01:57,129 --> 02:02:03,295 But don't, don't think that it's going to suddenly eliminate huge 1661 02:02:03,295 --> 02:02:06,380 numbers of jobs. Yeah. 1662 02:02:06,388 --> 02:02:11,082 Well, and you still you need at least at the moment there's a 1663 02:02:11,090 --> 02:02:15,090 huge market for people with HPC, CFD, programming, maths 1664 02:02:15,098 --> 02:02:18,055 backgrounds to develop these AI models. 1665 02:02:18,063 --> 02:02:23,305 So it's actually a a great time in some ways to have those skill 1666 02:02:23,305 --> 02:02:25,608 sets. Yes, I'm not sure that we're, 1667 02:02:25,616 --> 02:02:30,589 we're developing enough new people in HPC, you know, and I'm 1668 02:02:30,589 --> 02:02:34,832 not sure which degree programs they're coming out of it. 1669 02:02:34,840 --> 02:02:37,768 It's definitely not maths. It's definitely not computer 1670 02:02:37,776 --> 02:02:40,940 science. Well, not, not Oxford computer 1671 02:02:40,948 --> 02:02:43,529 science. You know, places like Warwick 1672 02:02:43,537 --> 02:02:48,464 maybe have a bit more or Bristol maybe have a bit more of a focus 1673 02:02:48,464 --> 02:02:52,782 on on HPC. Now it's, it's a very actually 1674 02:02:52,790 --> 02:02:57,575 good question because I guess by definition, what is HPC is part 1675 02:02:57,575 --> 02:03:01,355 of the problem. And I, I would agree with you 1676 02:03:01,363 --> 02:03:04,830 that. And again, to link to the AI, 1677 02:03:04,838 --> 02:03:09,628 it's even more important because you know, as you know, big AI 1678 02:03:09,636 --> 02:03:13,152 training clusters are essentially what people would 1679 02:03:13,160 --> 02:03:17,687 have called HPC clusters before. There is essentially no 1680 02:03:17,687 --> 02:03:20,618 difference. Networking, you use Slurm or, 1681 02:03:20,618 --> 02:03:25,280 now, Kubernetes and things. But but you're right, what 1682 02:03:25,288 --> 02:03:29,800 course teaches? Yeah. 1683 02:03:29,808 --> 02:03:35,175 Yes, I mean, I know some of the inside story on, on in 1684 02:03:35,183 --> 02:03:40,746 Microsoft's development of large GPU clusters and how they in, 1685 02:03:40,754 --> 02:03:45,075 you know, involved a consultant who is one of the world's 1686 02:03:45,083 --> 02:03:48,544 leading HPC experts, you know, so, so yes, absolutely. 1687 02:03:48,552 --> 02:03:53,254 You know, doing, doing these really large systems is is a 1688 02:03:53,262 --> 02:03:56,600 massive HPC challenge. It, it, it would be interesting 1689 02:03:56,600 --> 02:04:01,175 to know, you know, the people that NVIDIA hires as DevTechs, 1690 02:04:01,183 --> 02:04:07,500 what is their background? You know, to what extent have 1691 02:04:07,508 --> 02:04:13,106 they been formally trained in HPC or to what extent is, has it 1692 02:04:13,106 --> 02:04:17,964 been a passion? And they've learnt on the job in 1693 02:04:17,964 --> 02:04:21,568 various application areas and they've, they've proved their 1694 02:04:21,576 --> 02:04:25,120 skills and, you know, been hired on that basis. 1695 02:04:25,128 --> 02:04:29,448 That's that's a good question. And I think maybe this goes not 1696 02:04:29,448 --> 02:04:35,155 full circle, but half to our discussion about some of these 1697 02:04:35,155 --> 02:04:37,506 OERC research software engineers. 1698 02:04:37,506 --> 02:04:42,600 At least what I see is that people get their HPC skills 1699 02:04:42,600 --> 02:04:46,600 often during their PhD when they're using an HPC facility 1700 02:04:46,600 --> 02:04:50,534 and they sort of become best buddies with the admin because 1701 02:04:50,534 --> 02:04:52,868 they want to get higher up in the queue where they have to 1702 02:04:52,868 --> 02:04:54,200 sort of figure out some courses. 1703 02:04:54,680 --> 02:04:58,080 And so they haven't done necessary HPC course, but 1704 02:04:58,080 --> 02:05:00,760 they've had to do it to get access to the compute. 1705 02:05:02,040 --> 02:05:04,800 But somebody was there to manage the system. 1706 02:05:04,800 --> 02:05:07,000 So it's probably those, like RSEs, I think they call it. 1707 02:05:07,120 --> 02:05:08,840 Is it RSEs now research software? 1708 02:05:08,840 --> 02:05:12,440 Engineers. Are sort of like this lifeblood 1709 02:05:12,640 --> 02:05:15,360 who support the people and are probably helping them. 1710 02:05:15,360 --> 02:05:22,880 That maybe is an underrated skill or need in the community. 1711 02:05:23,520 --> 02:05:29,040 But you're right, I'm I'm not sure of any HPC official 1712 02:05:29,040 --> 02:05:34,320 training, but maybe I need to do a bit of digging around to we 1713 02:05:34,320 --> 02:05:36,440 sort of take for granted stuff that we have. 1714 02:05:36,440 --> 02:05:39,000 We never realized that. What if all those people retire? 1715 02:05:39,000 --> 02:05:41,002 I suppose is your point, isn't? 1716 02:05:41,010 --> 02:05:44,820 It like, yeah, yes, yes, I wonder yes, where, where, where, 1717 02:05:44,820 --> 02:05:48,915 where the next generation of academics is, is, is coming 1718 02:05:48,923 --> 02:05:51,584 from, but. Yeah, yeah. 1719 02:05:51,592 --> 02:05:55,584 But I, yeah, really appreciate you talking. 1720 02:05:55,592 --> 02:06:00,782 And I, I will put some links to the the sort of episode notes 1721 02:06:00,790 --> 02:06:03,284 because you, you know, I saw you've got a great website as 1722 02:06:03,292 --> 02:06:06,540 well where you link some of the, you know, full list of your 1723 02:06:06,548 --> 02:06:09,426 papers, some of the courses that you're doing so that people 1724 02:06:09,434 --> 02:06:12,360 could maybe read up because we didn't get into all of your 1725 02:06:12,368 --> 02:06:14,261 academic papers. But I know you've done a good 1726 02:06:14,269 --> 02:06:16,920 job of linking and some of the presentations and things like 1727 02:06:16,928 --> 02:06:18,509 that. So I'll put it through. 1728 02:06:18,517 --> 02:06:21,496 But yeah, I just want to say thank you also collectively 1729 02:06:21,504 --> 02:06:25,840 thank you because all the work that you did in throughout your 1730 02:06:25,840 --> 02:06:29,074 career has actually helped people like me and others who 1731 02:06:29,082 --> 02:06:32,916 work in industry to be able to do CFD in an easy way that we 1732 02:06:32,924 --> 02:06:36,680 take for granted now and don't have to program 2D grids by hand 1733 02:06:36,680 --> 02:06:39,942 like like you did. So thank you from all the CFD 1734 02:06:39,950 --> 02:06:41,983 people who take it for granted now. 1735 02:06:41,991 --> 02:06:45,854 And yeah, thanks for taking the time to to speak to me. 1736 02:06:45,862 --> 02:06:48,508 You're very welcome. It's fun. 1737 02:06:48,516 --> 02:06:53,047 Fun reminiscing about these these things from days past. 1738 02:06:53,055 --> 02:06:54,630 Yeah. Great. 1739 02:06:54,638 --> 02:06:55,440 Thanks very much.