#8256: adding locality levels to prefetch# and friends -------------------------------------+------------------------------------ Reporter: carter | Owner: carter Type: task | Status: patch Priority: normal | Milestone: 7.8.1 Component: Compiler | Version: 7.7 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Unknown/Multiple Type of failure: None/Unknown | Difficulty: Unknown Test Case: | Blocked By: Blocking: | Related Tickets: #8252 ,#8107 -------------------------------------+------------------------------------ Comment (by carter): I wrote some C code to see what the code gen in clang does for that asm {{{ // void __builtin_prefetch (const void *addr, ...) void prefetchnta(const void *addr){ __builtin_prefetch(addr, 0,0); } void prefetch2(const void *addr){ __builtin_prefetch(addr, 0,1); } void prefetch1(const void *addr){ __builtin_prefetch(addr, 0,2); } void prefetch0(const void *addr){ __builtin_prefetch(addr, 0,3); } }} the asm generated is {{{ .section __TEXT,__text,regular,pure_instructions .globl _prefetchnta .align 4, 0x90 _prefetchnta: ## @prefetchnta .cfi_startproc ## BB#0: pushq %rbp Ltmp2: .cfi_def_cfa_offset 16 Ltmp3: .cfi_offset %rbp, -16 movq %rsp, %rbp Ltmp4: .cfi_def_cfa_register %rbp movq %rdi, -8(%rbp) movq -8(%rbp), %rdi prefetchnta (%rdi) popq %rbp ret .cfi_endproc .globl _prefetch2 .align 4, 0x90 _prefetch2: ## @prefetch2 .cfi_startproc ## BB#0: pushq %rbp Ltmp7: .cfi_def_cfa_offset 16 Ltmp8: .cfi_offset %rbp, -16 movq %rsp, %rbp Ltmp9: .cfi_def_cfa_register %rbp movq %rdi, -8(%rbp) movq -8(%rbp), %rdi prefetcht2 (%rdi) popq %rbp ret .cfi_endproc .globl _prefetch1 .align 4, 0x90 _prefetch1: ## @prefetch1 .cfi_startproc ## BB#0: pushq %rbp Ltmp12: .cfi_def_cfa_offset 16 Ltmp13: .cfi_offset %rbp, -16 movq %rsp, %rbp Ltmp14: .cfi_def_cfa_register %rbp movq %rdi, -8(%rbp) movq -8(%rbp), %rdi prefetcht1 (%rdi) popq %rbp ret .cfi_endproc .globl _prefetch0 .align 4, 0x90 _prefetch0: ## @prefetch0 .cfi_startproc ## BB#0: pushq %rbp Ltmp17: .cfi_def_cfa_offset 16 Ltmp18: .cfi_offset %rbp, -16 movq %rsp, %rbp Ltmp19: .cfi_def_cfa_register %rbp movq %rdi, -8(%rbp) movq -8(%rbp), %rdi prefetcht0 (%rdi) popq %rbp ret .cfi_endproc .subsections_via_symbols }}} -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/8256#comment:17 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler