Date: | Tue, 30 May 2006 11:18:05 +0200 (CEST) |
---|---|
From: | Enric Herrero <enricherrero@xxxxxxxx> |
Subject: | Re: [Gems-users] Failed assertion in DirectoryMemory.C |
Ok, we have seen that: g_MEMORY_MODULE_BITS = g_MEMORY_SIZE_BITS - g_DATA_BLOCK_BITS -and that g_MEMORY_SIZE_BYTES is the installed memory. Probably in the mapping function memoryBits() is g_NUM_MEMORIES_BITS. In this case our system should work but we still have a failed assertion. Maybe the problem is that we use a large number of memories since in configurations with 4,8,16 and 32 memories we don't have problems. We have 4096 Mb of main memory divided in 64 memories, so 64Mb/memory ruby0.setparam g_NUM_PROCESSORS 64 ruby0.setparam g_NUM_L2_BANKS 64 ruby0.setparam g_NUM_MEMORIES 64 ruby0.setparam g_PROCS_PER_CHIP 64 ruby0.setparam L1_CACHE_NUM_SETS_BITS 8 ruby0.setparam DIR_CACHE_NUM_SETS_BITS 10 ruby0.setparam L2_CACHE_NUM_SETS_BITS 18 ruby0.setparam g_DATA_BLOCK_BYTES 32 ruby0.setparam g_MEMORY_SIZE_BYTES 4294967296 If we understand well all the memories in ruby can access all positions in dram memory. It is the mapping function that determines which memory has access to each position. In this case a failed assertion when looking an address may not be a problem of the mapping function. Is it right? Any help is highly appreciated, Thanks /Marco and Enric Enric Herrero <enricherrero@xxxxxxxx> escribió: Hi,
LLama Gratis a cualquier PC del Mundo. Llamadas a fijos y móviles desde 1 céntimo por minuto. http://es.voice.yahoo.com |
[← Prev in Thread] | Current Thread | [Next in Thread→] |
---|---|---|
|
Previous by Date: | Re: [Gems-users] question about makePrefetch and makeRequest, Weihang Jiang |
---|---|
Next by Date: | Re: [Gems-users] Shared Counter in LogTM, James Wang |
Previous by Thread: | Re: [Gems-users] Failed assertion in DirectoryMemory.C, Enric Herrero |
Next by Thread: | Re: [Gems-users] Failed assertion in DirectoryMemory.C, Enric Herrero |
Indexes: | [Date] [Thread] |