International Journal of Innovative Research in Science, Engineering and Technology

International Journal of Innovative Research in Science, Engineering and Technology

Menu

ISSN ONLINE(2319-8753)PRINT(2347-6710)

All submissions of the EM system will be redirected to Online Manuscript Submission System. Authors are requested to submit articles directly to Online Manuscript Submission System of respective journal.

Design of Low Power L2 Cache Architecture Using Way Tag Information

A. Divya jebaseeli1, M.Kiruba2
  1. PG Scholar, Kalaingnar karunanidhi institute of technology, Coimbatore, India
  2. Assistant Professor of ECE, Kalaingnar karunanidhi institute of technology, Coimbatore, India
Related article at Pubmed, Scholar Google

Visit for more related articles at International Journal of Innovative Research in Science, Engineering and Technology

International Journal of Innovative Research in Science, Engineering and Technology

Abstract

Today high-performance microprocessors make use of cache write-through policy for performance improvement and achieving good tolerance to soft errors in on-chip cache. However write through policy incurs large power utilization, while accessing the cache at low level (L2 cache) during write operation. In new method, way_tagged cache was used under write-through policy, it’s consumed more energy. By maintaining the wag tag of L2 cache in the L1 cache during read operation. The proposed technique enables L2 cache to work in direct mapping manner during write hit and reducing tag comparison of cache miss prediction, if cache miss is predicted there is no need to access the L2 cache. So that significant portion of energy will be reduced, without performance degradation. Simulation results are obtained both L1 and L2 cache configuration. The proposed technique achieves 70.7%energy saving in L2 cache on average with only 0.02% area overhead and no performance degradation, when compare with existing methods.

Index Terms
Cache, low power, write-through policy, way tag

INTRODUCTION
Multi-level on-chip cache systems have been commonly adopted in high-performance microprocessor [2]-[4]. Cache is the highest or first level of the memory hierarchy, it store the frequently access data.L2 caches are unified caches, which has L1 and further data .to keep the data reliable throughout the memory hierarchy, writethrough and write-back policies are commonly employed. Under the write-through Policy, all copies of a cache block are updated immediately both L1 and L2 after The cache block is modified at the current cache, even though The block might not be evicted. As a result, the write-through Policy maintains identical data copies at all levels of the cache (L1 and L2) hierarchy throughout most of their life time of execution.
While decreasing device sizes radiation produced soft error, become a serious problem for advanced CMOS circuits. As CMOS technology is scaled into the nanometer range, where Soft errors in microelectronic components are induced either by alpha material or by external irradiation [1]. These errors, which can occur more often than permanent errors [5], arise from Single Event multi-bit Upsets, are getting poorer in on-chip memories [7],[8]. At the architecture level, an effective solution is to keep data consistent among different levels of the memory hierarchy to prevent the system from soft errors [10]–[12]. Benefited from immediate update, cache write-through policy is effectively tolerant to soft errors because the data at all related levels of the cache hierarchy are always kept consistent. Due to this aspect many high performance micro-processor designs have adapted write-through policy.
Consider a two-level (i.e., L1 and L2) cache system for example. If the L1 data cache implements the write-back policy, a write hit in the L1 cache does not need to access the L2 cache. In contrast, if the L1 cache is writethrough, then both L1 and L2 caches need to be accessed for every write operation. Obviously, the write-through policy incurs more write accesses in the L2 cache then L1 cache, which in turn increases the energy consumption of the cache system. So that power consumption is reduced because of power consumption is a one of the serious issues in designing cache system [13]-[14].
In this paper, we propose new cache architecture, referred to as partial way-tagged cache, to improve the energy efficiency of write-through cache systems with minimal area overhead and without performance degradation. Consider a two-level cache hierarchy, where the L1 data cache is write-through and the L2 cache is inclusive for high performance. It is observed that all the data present in the L1 cache will have copies in the L2 cache. In addition, the locations of these copies in the L2 cache will not change until they are ejected from the L2 cache. Thus, we can attach a way tag in the L2 cache and send this tag information to the L1 cache when the data is loaded to the L1 cache. During the subsequent accesses when there is a write hit in the L1 cache, we can access the L2 cache in an equivalent direct-mapping manner because the way tag of the data copy in the L2 cache is available. As this operation accounts for the majority of L2 cache accesses in most applications, the energy consumption of L2 cache can be reduced significantly.
In this paper, the work is extending by making the following contributions. First, a detailed VLSI architecture of the proposed way tagged cache is developed, where various design issues regarding timing, control logic, operating mechanisms, and area overhead. Second, the idea of way tagging can be extended to many existing low-power cache design techniques, so that better tradeoffs of performance and energy efficiency can be achieved. Finally, a complete simulations is performed with new results covering the effectiveness of the proposed technique under different cache configurations, further improve energy efficiency.

BACKGROUND AND RELATED WORKS
Many techniques have been developed to reduce cache power dissipation. In this section, we briefly analysis some existing work related to the proposed technique. In [6] partitioned cache data arrays into several sub banks. During each access only the desired subbank is activated and corresponding data is activated. Ghose et al. further divided cache bitlines into small segmentations [9]. When a memory cell is accessed, only the connected bitline segmentations are evaluated. These techniques effectively reduce the energy per access by modifying the structure of cache systems. At the architecture level, set-associative caches mostly preferred due to their low miss rates. In conventional set-associative caches, all tag and data arrays are accessed simultaneously for performance improvement. This, however, comes at the cost of energy overhead. Many techniques have been proposed to reduce the energy consumption of set-associative cache. The fundamental idea is to trigger fewer tags and data arrays during an access, so that cache power dissipation can be reduced. In the phased cache [15] proposed by Hasegawa et al., two phase of cache access. Cache tag arrays are accessed in the first phase and the second phase only the data array is access corresponding to the matched tag. Energy consumption can be reduced due to only one data array is accessed compare to the N-data array. Due to the increase in access cycles to reduce the performance on phased cache this usually employed in the lower level memory. Another technique referred to as way concatenation was proposed by Zhang et al. [16] to reduce the cache energy in embedded systems. According to the system requirement the cache can be configured as direct-mapping, four-way, two-way setassociative with the necessary software support. By accessing fewer tag and data arrays, better energy efficiency is attained. This technique is effectively suitable for embedded system, but not suitable for high-performance microprocessors. Other techniques include way-predicting set-associative caches, proposed by Inoue et al. [17]–[19], that make a prediction on the ways of both tag and data arrays in which the desired date might be located in cache.
If the prediction is correct, the corresponding way of data array is accessed to complete the operation; otherwise, the rest ways of the cache are accessed to collect the desired data. Another similar approach proposed by Min et al. [20] employs redundant cache to predict the incoming cache references. The location cache needs to be trigged for every operation in the L1 cache which wastes energy if the hit rate of L1 cache is high. Among the above related work, phased caches and way-predicting caches are commonly employed in high-performance Microprocessors. Compared with these techniques, the proposed way-tagged cache achieves better power efficiency without performance degradation. Specifically, the basic idea of way-predicting caches is to store a small number of the most recently used (MRU) addresses. Since L2 caches are usually unified caches, so the MRU-based prediction has a poor prediction rate [21], [22], it reduce the performance. For phased caches, the energy consumption of accessing tag arrays is nearly portion of total L2 cache energy. As Shown in Section III, applying the proposed technique of way tagging can reduce this energy consumption. Section IV provides more details comparing the proposed technique with twolevel cache.

WAY-TAGGED CACHE
In this section, we propose a way-tagged cache that utilizes the way information in L2 cache to improve energy efficiency. We consider a conventional set-associative cache system when the L1 data cache loads/writes data from/into the L2 cache; all ways in the L2 cache are activated simultaneously for performance consideration at the cost of energy overhead. Fig. 1 illustrates the architecture of the conventional two-level cache. Only the L1 data cache and L2 unified cache are shown.
image
The L2 cache always maintains the most recent copy of the data under the write through policy. Thus, the L2 cache is a unified cache, whenever a data is updated in the L1 cache, the L2 cache is updated with the same data as well. This will increase the write access in L2 cache and becomes more energy consumption.
image
The read accesses in the L2 cache are initiated by the read misses in the L1 cache, which usually occurs much less frequently. Normally write access consumes more power in the total power. The location of L1 data copies in the L2 cache will not change until the data are evicted from the L2 cache. The proposed way-tagged cache reduces the number of ways accessed during L2 cache accesses. The way-tag arrays stores the way tags of the data in the L2 cache; the L1 data cache loads a data from the L2 cache. These way tags provide the key information for the consequent write accesses to the L2 cache.
In general, both write and read accesses in the L1 cache may need to access the L2 cache. These accesses lead to different operations as summarized in Table I in proposed method all write operations of the L1 cache need to access the L2 cache under the write through policy. In the case of a write hit in the L1 cache, the L2 cache will be activated only one way because the way tag information of the L2 cache is available in way-tag array.
While for a write miss in the L1 cache, the corresponding data is not stored in the L1 cache. As a result, its corresponding way tag information is not available in the way-tag arrays. So, all ways in the L2 cache need to be activated simultaneously. By this approach all L1 write operation need to access the correct location in L2 cache in order to avoid the performance degradation. Note that waytag arrays are very small in size. For L1 read operation, if the cache hit means directly access the L1 data no need to access way-tag array so that significant portion of the energy is reduced. If cache miss means check the way-tag array the tag is matched and select correct is activated in L2 cache or else all way are active in L2 cache. Write accesses account for the majority of L2 cache accesses in most applications; so that energy overhead is occur in high performance microprocessor. In addition, write hits are dominant among all write operations. Therefore, by activating fewer ways in most of the L2 write accesses, the proposed way tagged cache is very effective in reducing memory energy consumption.
Fig. 2 shows the system diagram of proposed waytagged cache. New components are introduced: way-tag arrays, way-tag buffer, way decoder, and way register, all shown in the dotted line. The way tags of each cache line in the L2 cache are maintained in the way-tag arrays, located with the L1 data cache. Note that write buffers are commonly employed in write through caches to improve the performance. With a write buffer, the data to be written into the L1 cache is also sent to the write buffer. This avoids write stalls for write operations to be completed in the L2 cache. A small way-tag buffer is introduced to buffer the way tags read from the way-tag arrays. A way decoder is working to decode way tags and generate the enable signals for the L2 cache, which activate only the desired ways in the L2 cache. Each way in the L2 cache is encoded into a way tag. A way register stores way tags and provides this information to the way-tag arrays.

PROPOSED WAY TAG CACHE
Fig.2 shows the diagram of proposed way-tagged cache architecture. In this paper, introduce several new components: way-tag arrays, way-tag buffer, way decoder, and way register, all shown in the dotted line.
image
A. Way-Tag Arrays
Way tag arrays have way tags of a data is loaded from the L2 cache to the L1 cache, shown in Fig 3. Note that the data arrays in the L1 data cache and the way-tag arrays share the same address from CPU. The WRITEH_W signal of way-tag arrays is generated from the write/read signal of the data arrays in the L1 data cache as shown in Fig. 8. A UPDATE is control signal, obtained from the cache controller. When a L1 write miss, UPDATE will be declared and allow WRITEH_W to enable the write operation to the way-tag arrays (UPDATE=1 and WRITEH_W, See Table II). UPDATE keeps invalid and WRITEH_W =1, a read operation to the way-tag arrays. During the read operations of the L1 cache, the way-tag arrays do not need to be accessed and thus, reduce energy overhead. To minimize the overhead of way-tag arrays, the size of a way-tag array can be expressed as
image
Where SL1, Sline,L1 and Nway,L1 are the size of the L1 data cache, cache line size and number of the ways in the L1 data cache respectively.imae is a binary code.
image
The way-tag arrays are operated in parallel with the L1 data cache for avoiding the performance degradation. Due to their small size, the access delay is much smaller than that of the L1 cache.
image
Way-tag buffer is temporarily stores the Way tags from the way-tag arrays in the L1 cache. It has the same number of entries as the write buffer of the L2 cache and shares the control signals with it. Note that write buffers are commonly employed, the data to be written into the L1 cache is also sent to the write buffer to improve the performance. This avoids write stalls when the processor waits for write operations to be completed in the L2 cache. When a write miss occurs in L1 cache, all the ways in the L2 cache need to be activated as the way information is not available. Otherwise, only the desired way is activated. Way tag buffer is small in to avoid area overhead.
C. Way decoder
The function of the way decoder is employed to decode way tags and generate the enable signal, which activate only desired ways in L2 cache. This avoids the additional wires and the chip area is negligible. A write hit in the L1 cache, the way decoder works as an n -to- N decoder that selects one way-enable signal. For a write miss or a read miss in the L1 cache, the way decoder assert all way-enable signals, so that all ways in the L2 cache are activated. A Fig.4 shows that write and read miss signals decides the way decoder operation.
image
D. way register
The way tags for the way-tag arrays is Provided by Way register. A 4-way L2 cache is consider, that labels “00”, “01”, “10”, and“11”.This are stored in the way register. When the L1 cache loads a data from the L2 cache, the corresponding way tag in the way register is sent to the way-tag arrays by this manner the corresponding way tags are stored in way-tag array. The proposed way-tagged caches way operates under different modes during read and write operations. Only the way containing the desired data is activated in the L2 cache for a write hit in the L1 cache, working the L2 cache equivalently a direct-mapping cache to reduce energy consumption without performance overhead under the write-through policy.

RESULT AND DISCUSSION
The way-tag cache has been implemented using VHDL Language. Fig.3 shows the simulation results for two level caches. Clock, array of tag and data are given as input, corresponding data are return into the memory or else read the data from the memory depending upon read and write operation.
image
Fig. 6 shows the simulation results for way tag array. Clock, array of tag and data are given as input, corresponding data are return into L1 or else read the data from theL2 cache depending upon operation cache hit and cache miss occur in L1 cache.
image
Fig. 7 shows the simulation results for way tag cache. Clock, array of tag and data of L1, L2 and way tag array are given as input, corresponding data are return into L1 or else read the data from theL2 cache depending upon operation cache hit and cache miss occur in L1 cache and way tag array.
image
Fig. 8 shows that power analysis of two-level cache. This will consumes power 222mW.
image
Fig. 9 shows that power analysis of way tag caches. This will consumes power 114mW.the operations way-tagged cache consumes less power compare the twolevel cache.
image
Fig. 9 Power analysis of way tag cache
image
Table 5.1 shows that power and delay comparison for both two-level cache and way-tag cache. Here, way-tag cache consumes less power when compare the two-level cache as well as delay also reduced. Nearly 50% of the power will be reduced and 10% of delay also reduced.

CONCLUSION
This paper presents a new energy-efficient cache technique for high-performance microprocessors employing the write-through policy. The proposed technique has a way-tag array, it stores to each way in the L2 cache. This way tag is sent to the way-tag arrays in the L1 cache when the data is loaded from the L2 cache to the L1 cache. Utilizing the way tags stored in the way-tag arrays, the L2 cache can be accessed as a direct-mapping cache during the subsequent write hits, thereby reducing cache energy consumption. Simulation results demonstrate significantly reduction in cache energy consumption with minimal area overhead and no performance degradation. Furthermore, the idea of way tagging can be applied to many existing lowpower cache techniques such as the phased access cache to further reduce cache energy consumption. Future work is being directed towards extending this technique with partial tag comparison to reducing the energy consumption of cache operations.

References