Basic Memory Allocator Info needed- Hoard and SLAB - memory-management

Recently I have been reading about memory allocator, like Hoard and SLAB. However I didn't get few things:
a. Are these allocator managing physical memory or virtual memory. {If (your answer is Physical Memory) please read point b, else, read point c}
b. If they manage physical memory, since both these alloctor make use of per-CPU data structures, don't they end-up giving space from same physical page to different processes. For e.g. consider T1 starts on CPU C and request for an int. After this let T1 get preempted and T2 starts executing and it also ask for an int. Since our structures are per-CPU, is it not the case that we will end up satisfying both request from same physical page.
c. And if they manage virtual memory, then why we say that all data structures are per-CPU, rather we should say that they are per-process, because on every context switch we have to repopulate these data structures.

Related

Variable allocation and tracking

I started searching and reading about ALDS and memory management recently after I got a doubt about memory allocation, and after a couple of days of study I learnt a lot of things about memory management but the actual doubt remains unsolved.
So the doubt is, while allocating memory to a variable, how exactly does the system know which block of memory is available and which is free, and similarly when we destruct an object or set a variable as null or when GC frees up some memory, what exactly does it do with that block of memory, as I know the actual data is never erased on deletion, that block just gets marked as free somewhere in some table, but does that table keep track of each and every bit on the memory, if yes then wouldn't that become a lot of data in itself to store?
For an example, if I declare a linked list, then a block will be allocated in heap with it's next block having null value as there is no other node to reference, now as I keep adding more nodes into it, system will keep allocating more blocks each containing reference to next one. Now these blocks can be present on random locations depending on the availability of memory at allocation time, and can only be accessed through their proceeding nodes.
So now, for any given block of memory, how the system will know if its free and has just garbage value in it, or its actually a node of some linked list.
On a modern operating system the process has a logical, linear address space. Part of that address space is reserved for the system and is common to all processes. Some of the address space may be reserved but most of the remainder is available to the process.
The address space is defined by PAGE TABLES. The structure of the page table is defined by the processor but the operating system maintains a table for each process. Memory is allocated to a process in PAGES. The smallest I am aware of is 512 bytes but the size can go up to a megabyte or even larger in some processors and some processor configurations.The size is always a power of 2.
The page table defines:
Whether an page has actually been mapped to the process
Whether the pages has a corresponding physical memory location
If so, the mapping to that physical location.
There operating system only knows about pages.
At the next level down there are memory managers. These are not part of the operating system. Memory managers manage heaps that consist of pages allocated by the operating system. The memory manage has to keep track of the heap size and what memory has been allocated within it.
Memory managers operate is a huge number of different ways. There are malloc/free implementations galore that you can link into your code to get different behaviors.

virtual memory effects and relations between paging and segmentation

This is my first post. I want to ask about how are virtual memory related to paging and segmentation. I am searching internet for few days, but still can't manage to put that information into right order. Here is what I know so far:
We can talk about addresses (we could say they are levels of memory abstraction) in memory:
physical level (CPU talking to memory controller, "hey give me contents of address 0xFFEABCD", these adresses are adresses of cells in RAM, so cell 0xABCD has physical address 0xABCD. memory controller can only use physical adresses, so if adress is not physical it must be changed to physical.
logical level.This is abstraction over physical addresses. Here processes if ask for memory, (assume successfull allocation) are given address which has no direct relation to cells in RAM. We can say these addresses are from different pool (world?) than physical addresses. As I said before memory controller only understand physical adresses, so to use logical addresses, we need to convert them to physical addresses. There are two ways for OS to be able to create logical adresses:
paging - in which physical memory (RAM) is divided into continous blocks of memory (called frames), and logical memory (this other world) is also divided in same in length blocks (called pages). Now OS keep in RAM data structure called page table. It's an associative array (map) and it's primary goal of existence is to translate logical level addresses to physical level adresses. Paging has following effect: memory allocated by process in RAM (so in frames in physical memory belonging to program) may not be in contingous manner (so there may be holes inside).
segmentation - program is divided into parts called segments. Segments sizes are not fixed, so different segments may have different sizes. Program is divided in few segments and each segment will have its own place in RAM (physical) memory. So one segment (call it sementA), and another (call it segmentB) may not be near each other. In other words segmentA don't have to has segmentB as a neighbour.
internal fragmentation - when memory which belongs to process isn't used in 100%. So if process want to have 2 bytes for its use, OS need to allocate page/pages which total size need to be greater or equal than amount of memory requested by program. Typical size of page is 4KB. Unit in which OS gives memory to process are pages. So it can't give less than 4KB. So if we use 2 bytes, 4KB - 2B = 4094 bytes are wasted (memory is associated with our process so other processes can't use it. Only we can use it, but we only need 2B).
external fragmentation - when allocated blocks of memory are one near another, but there is a little hole between them. Its free, so other programs, can use it, but it is unlikly because it is very small. That holes with high probability will be wasted. More holes - more wasted memory.
Paging may cause effect of internal fragmentation. Segmentation may cause effect of external fragmentation.
virtual level - addresses used in virtual memory. This is extension of logical memory level. Now program don't even need to have all of it's allocated pages in RAM to start execution. It can be implemented with following techniques:
paged segmentation - method in which segments are divided into pages.
segmented paging - less used method but also possible.
Combining them takes a positive aspects from both solutions.
What i have read about pros and cons of virtual memory:
PROS:
processes have their own address space which mean if we have two processes A and B, and both of them have a pointer to address eg. 17 processA pointer will be showing to different frame than pointer in processB. this results in greater process isolation. Processes are protected from each other (so one process can't do things with another process memory if it isn't shared memory because in its mapping don't exist such mapping entry), and OS is more protected from processes.
have more memory than you physical first order memory(RAM, due to swapping to secondary order memory).
better use of memory due to:
swapping unused parts of programs to secondary memory.
making sharings pages possible, also make possible "copy on write".
improved multiprogram capability (when not needed parts of programs are swapped out to secondary memory, they made free space in ram which could be used for new procesess.)
improved CPU utilisation (if you can have more processes loaded into memory you have bigger probability than there exist some program that now need do CPU stuff, not IO stuff. In such cases you can better utilise CPU).
CONS:
virtual memory has it's overhead because we need to get access to memory twice (but here a lot of improvment can be achieved using TLB buffers)
it makes OS part managing memory more complicated.
So here we came to parts which I don't really understand:
Why in some sources logical address and virtual addresses are described as synonymes? Do I get something wrong?
Is really virtual memory making protection to processes? I mean, in segmentation for example there was also check if process do not acces other memory (resulting in segfault if it does), paging also has a protection bit in a page table, so doesn't the protection come from simply extending abstraction of logic level addresses? If VM (Virtual Memory) brings extended protection features, what are they and how they work? In other words: does creating separate address space for each process, bring extended memory protection. If so, what can't be achieved is paging without VM?
How really differ paged segmentation from segmented paging. I know that the difference between these two will be how a address is constructed (a page number, segment number, that stuff..), but I suppose it isn't enough to develop 2 strategies. This reason is like nothing. I read that segmented paging is less elastic, and that's the reason why it is rarely used. But why it it less elastic? Is the reason for that, that in program you can have only few segments instead a lot of pages. If thats the case paging indeed allow better "granularity".
If VM make separate address space for each process, does it mean, paging without VM use logic addresses from "one pool" (is then every logic address globally unique in that case?).
Any help on that topic would be appreciated.
Edit: #1
Ok. I finally understood that paging not on demand is also a virtual memory. I just found some clarification was helpful to understand the topic. Below is link to image which I made to visualize differences. Thanks for help.
differences between paging, demand paging and swapping
Why in some sources logical address and virtual addresses are described as synonymes? Do I get something wrong?
Many sources conflate logical and virtual memory translation. In ye olde days, logical address translation never took place without virtual address translation so processor documentation referred to them as the same.
Now we have large memory systems that use logical memory translation without virtual memory.
Is really virtual memory making protection to processes?
It is the logical memory translation that implements page protections.
How really differ paged segmentation from segmented paging.
You can really ignore segments. No rationally designed processor architecture designed after 1970 used segments and they are finally dying out.
If VM make separate address space for each process, does it mean, paging without VM use logic addresses from "one pool"
It is logical memory that creates the separate address space for each process. Paging is virtual memory. You cannot have one without the other.

Could someone help me understand VkPhysicalDeviceMemoryProperties?

I'm trying to figure it out, but I'm getting a little stuck.
The way the types and heaps are related is simple, if a bit strange. (why not just give VkMemoryHeap a VkMemoryType member?)
I think I understand what all the VkMemoryPropertyFlags mean, they seem fairly straightforward.
But what's with the VkMemoryHeap.flags member? It apparently only has one non-zero valid value, VkMemoryHeapFlagBits.VK_MEMORY_HEAP_DEVICE_LOCAL_BIT, and though that wouldn't be too odd on it's own, but there's also a VkMemoryPropertyFlagBits.VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT that could be present on the memory type of the heap.
What does the VkMemoryHeap.flags member mean and how does it relate to the VkMemoryType.flags member?
Vulkan recognizes two distinct concepts when it comes to memory. There are the actual physical pieces of RAM that the device can talk to. Then there are ways to allocate memory from one of those pools of RAM.
A heap represents a specific piece of RAM. VkMemoryHeap is the object that describes one of the available heaps of RAM that the device can talk to. There really aren't that many things that define a particular heap. Just two: the number of bytes of that RAMs storage and the storage's location relative to the Vulkan device (local vs. non-local).
A memory type is a particular means of allocating memory from a specific heap. VkMemoryType is the object that describes a particular way of allocating memory. And there are a lot more descriptive flags for how you can allocate memory from a heap.
For a more concrete example, consider a standard PC setup with a discrete GPU. The device has its own local RAM, but the discrete GPU can also access CPU memory. So a Vulkan device will have two heaps: one of them will be local, the other non-local.
However, there will usually be more than two memory types. You usually have one memory type that represents local memory, which does not have the VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT set. That means you can't map the memory; you can only access it via transfer operations from some other memory type (or from rendering operations or whatever).
But you will often have two memory types that both use the same non-local heap. They will both be VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT, thus allowing mapping. However, one of them will likely have the VK_MEMORY_PROPERTY_HOST_CACHED_BIT flag set, while the other will be VK_MEMORY_PROPERTY_HOST_COHERENT_BIT. This allows you to choose whether you want cached CPU access (thus requiring an explicit flush of ranges of modified memory) or uncached CPU access.
But while they are two separate memory types, they both allocate from the same heap. Which is why VkMemoryType has an index that refers to the heap who's memory it is allocating from.
Only thing I'm not getting is how the two DEVICE_LOCAL flags interact.
Did you look at the specification? It's not exactly hiding how this works:
if propertyFlags has the VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT bit set, memory allocated with this type is the most efficient for device access. This property will only be set for memory types belonging to heaps with the VK_MEMORY_HEAP_DEVICE_LOCAL_BIT set.
Is it saying that if the memory is local then all types corresponding to that memory are local, or that they can be local?
You seem to be trying to impose the wrong meaning to these things. Just look at what the specification says and take it at face value.
PROPERTY_DEVICE_LOCAL denotes a memory type which will achieve the best device access performance. The only connection between this and MEMORY_DEVICE_LOCAL is that memory types with PROPERTY_DEVICE_LOCAL will only be associated with memory heaps that use MEMORY_DEVICE_LOCAL.
That's the only relevant meaning here.
If you want an example of when a memory heap would be device local, yet have memory types that aren't, consider a GPU that has no memory of its own. There's only one heap, which is therefore MEMORY_DEVICE_LOCAL.
However, allocating memory from that pool in a way that makes it host-visible may decrease the performance of device access to that memory. Therefore, for such hardware, the host-visible memory types for the same heap will not use PROPERTY_DEVICE_LOCAL.
Then again, other hardware doesn't lose performance from making memory host-visible. So they only have one memory type, which has all of the available properties. For Intel, their on-chip GPUs apparently have access to some level of the CPU's caches.

CUDA: When to use shared memory and when to rely on L1 caching?

After Compute Capability 2.0 (Fermi) was released, I've wondered if there are any use cases left for shared memory. That is, when is it better to use shared memory than just let L1 perform its magic in the background?
Is shared memory simply there to let algorithms designed for CC < 2.0 run efficiently without modifications?
To collaborate via shared memory, threads in a block write to shared memory and synchronize with __syncthreads(). Why not simply write to global memory (through L1), and synchronize with __threadfence_block()? The latter option should be easier to implement since it doesn't have to relate to two different locations of values, and it should be faster because there is no explicit copying from global to shared memory. Since the data gets cached in L1, threads don't have to wait for data to actually make it all the way out to global memory.
With shared memory, one is guaranteed that a value that was put there remains there throughout the duration of the block. This is as opposed to values in L1, which get evicted if they are not used often enough. Are there any cases where it's better too cache such rarely used data in shared memory than to let the L1 manage them based on the usage pattern that the algorithm actually has?
2 big reasons why automatic caching is less efficient than manual scratch pad memory (applies to CPUs as well)
parallel accesses to random addresses are more efficient. Example: histogramming. Let's say you want to increment N bins, and each are > 256 bytes apart. Then due to coalescing rules, that will result in N serial reads/writes since global and cache memory is organized in large ~256byte blocks. Shared memory doesn't have that problem.
Also to access global memory, you have to do virtual to physical address translation. Having a TLB that can do lots of translations in || will be quite expensive. I haven't seen any SIMD architecture that actually does vector loads/stores in || and I believe this is the reason why.
avoids writing back dead values to memory, which wastes bandwidth & power. Example: in an image processing pipeline, you don't want your intermediate images to get flushed to memory.
Also, according to an NVIDIA employee, current L1 caches are write-through (immediately writes to L2 cache), which will slow down your program.
So basically, the caches get in the way if you really want performance.
As far as i know, L1 cache in a GPU behaves much like the cache in a CPU. So your comment that "This is as opposed to values in L1, which get evicted if they are not used often enough" doesn't make much sense to me
Data on L1 cache isn't evicted when it isn't used often enough. Usually it is evicted when a request is made for a memory region that wasn't previously in cache, and whose address resolves to one that is already in use. I don't know the exact caching algorithm employed by NVidia, but assuming a regular n-way associative, then each memory entry can only be cached in a small subset of the entire cache, based on it's address
I suppose this may also answer your question. With shared memory, you get full control as to what gets stored where, while with cache, everything is done automatically. Even though the compiler and the GPU can still be very clever in optimizing memory accesses, you can sometimes still find a better way, since you're the one who knows what input will be given, and what threads will do what (to a certain extent of course)
Caching data through several memory layers always needs to follow a cache-coherency protocol. There are several such protocols and the decision on which one is the most suitable is always a trade off.
You can have a look at some examples:
Related to GPUs
Generally for computing units
I don't want to get in many details, because it is a huge domain and I am not an expert. What I want to point out is that in a shared-memory system (here the term shared does not refer to the so called shared memory of GPUs) where many compute-units (CUs) need data concurrently there is a memory protocol that attempts to keep the data close to the units so that can fetch them as fast as possible. In the example of a GPU when many threads in the same SM (symmetric multiprocessor) access the same data there should be a coherency in the sense that if thread 1 reads a chunk of bytes from the global memory and in the next cycle thread 2 is going to access these data, then an efficient implementation would be such that thread 2 is aware that data are found already in L1 cache and can access it fast. This is what the cache coherency protocol attempts to achieve, to let all compute units be up to date with what data exist in caches L1, L2 and so on.
However, keeping threads up to date, or else, keeping threads in coherent states, comes at some cost which is essentially missing cycles.
In CUDA by defining the memory as shared rather than L1-cache you free it from that coherency protocol. So access to that memory (which is physically the same piece of whatever material it is) is direct and does not implicitly call the functionality of coherency protocol.
I don't know how fast should this be, I didn't perform any such benchmark but the idea is that since you don't pay anymore for this protocol the access should be faster!
Of course, the shared memory on NVIDIA GPUs is split in banks and if someone wants to use it for performance improvement should have a look at this before. The reason is bank conflicts that occur when two threads access the same bank and this causes serialization of the access..., but that's another thing link

How to pin a shared memory segment into physical memory

I use boost::interprocess::managed_shared_memory to load a data structure in shared memory. I need to pin the shared memory segment into physical memory (for example similar to system call mlock for mapped files).
In linux, sooner or later my data structure gets swapped out of physical memory. In my case this incurs a forbidding cost for the next process accessing the structure, after it has been swapped out.
Is there any way to pin shared memory into physical memory? I am interested in any solution, even if it means that I cannot use boost::interprocess.
Using basic_managed_xsi_shared_memory (apparently available since boost 1.46), you can access the underlying shmid (from the get_shmid member) which should allow you to control the shmid using shmctl. With shmctl you can prevent the swapping of shared memory pages by applying the SHM_LOCK command to the shmid.
Other types of locking (which you refer to as 'pinning'), such as locking memory mapped files into memory, can be realized by supplying return values obtained from mapped_region's get_address and get_size member functions to the mlock command.

Resources