IRIX Base Documentation 2001 May

home *** CD-ROM | disk | FTP | other *** search

/ IRIX Base Documentation 2001 May / SGI IRIX Base Documentation 2001 May.iso / usr / share / catman / p_man / cat5 / lno.z / lno

Wrap

Text File | 1998-10-30 | 27.9 KB | 562 lines

LNO(5) Last changed: 3-10-98 NNAAMMEE LNO - Compiler loop nest optimization option group SSYYNNOOPPSSIISS --LLNNOO:: ... IIMMPPLLEEMMEENNTTAATTIIOONN IRIX systems DDEESSCCRRIIPPTTIIOONN This man page describes the loop nest optimization options accepted by the ff9900(1), ff7777(1), CCCC(1), cccc(1), and cc8899(1) commands. The --LLNNOO:: option group specifies options and transformations performed on loop nests. The --LLNNOO:: option group is enabled only if the --OO33 option is also specified on the compiler command line. For information on the LNO options that are in effect during a compilation, use the --LLIISSTT::ooppttiioonnss==OONN option. You can specify more than one suboption to the --LLNNOO:: option either by using colons to separate each suboption or by specifying multiple options on the command line. For example, the following command lines are equivalent: f90 -LNO:auto_dist=ON:outer=OFF b.f f90 -LNO:auto_dist=ON -LNO:outer=OFF b.f Some --LLNNOO:: suboptions are specified with a setting that either enables or disables the feature. To enable a feature, specify the argument either alone or with ==11, ==OONN, or ==TTRRUUEE. To disable a feature, specify the suboption with either ==00, ==OOFFFF, or ==FFAALLSSEE. For example, the following command lines are equivalent: f90 -LNO:auto_dist:blocking=OFF:oinvar=FALSE a.f f90 -LNO:auto_dist=1:blocking=0:oinvar=OFF a.f For brevity, this man page shows only the OONN or OOFFFF settings to suboptions, but 00, 11, TTRRUUEE, and FFAALLSSEE are also allowed as settings. In addition, this man page shows the abbreviated form for some of the suboption names. You can use either the abbreviation or the complete suboption name when using the suboptions. The following is a list of the abbreviations and the complete suboption names: Complete name Abbreviation oouutteerr__uunnrroollll oouu aassssoocciiaattiivviittyy aassssoocc cclleeaann__mmiissss__ppeennaallttyy ccmmpp ddiirrttyy__mmiissss__ppeennaallttyy ddmmpp ccaacchhee__ssiizzee ccss iiss__mmeemmoorryy__lleevveell iiss__mmeemm lliinnee__ssiizzee llss ttllbb__eennttrriieess ttllbb ttllbb__cclleeaann__mmiissss__ppeennaallttyy ttllbbccmmpp pprreeffeettcchh__lleevveell ppff See "F77 LNO Directives" at the end of this man page for a summary of the F77 directives for LNO. See the _M_I_P_S_P_r_o _7 _F_o_r_t_r_a_n _9_0 _C_o_m_m_a_n_d_s _a_n_d _D_i_r_e_c_t_i_v_e_s _R_e_f_e_r_e_n_c_e _M_a_n_u_a_l, publication SR-3907, for a discussion of the Fortran 90 LNO directives. See _M_I_P_S_p_r_o _C _a_n_d _C++ _P_r_a_g_m_a_s, for descriptions of the C and C++ LNO #pragma directives. The descriptions to the suboptions to --LLNNOO:: are divided into the following categories: * General options * Transformation options * Cache memory management options * TLB options * Prefetch options The --LLNNOO option accepts the following general suboptions: SSuubbooppttiioonn AAccttiioonn aauuttoo__ddiisstt[[ == (( OONN||OOFFFF ))]] Distributes local arrays in common blocks that are accessed in parallel. The default is OOFFFF. This optimization works with either automatic parallelism or parallelism using directives; it is always safe, and does not affect the layout of arrays in virtual space, and does not incur addressing overhead. ffiissssiioonn==_n Controls loop fission. _n can be one of the following: 0 Disables loop fission. 1 Performs normal fission as necessary. This is the default. 2 Specifies that fission be tried before fusion. If --LLNNOO::ffiissssiioonn==_n and --LLNNOO::ffuussiioonn==_n are both set to 1 or to 2, fusion is performed. ffuussiioonn==_n Controls loop fusion. _n can be one of the following: 0 Disables loop fusion. 1 Performs standard outer loop fusion. This is the default. 2 Specifies that outer loops should be fused, even if it means partial fusion. The compiler attempts fusion before fission. The compiler performs partial fusion if not all levels can be fused in the multiple-level fusion. If --LLNNOO==ffiissssiioonn==_n and --LLNNOO::ffuussiioonn==_n are both set to 1 or to 2, fusion is performed. ffuussiioonn__ppeeeelliinngg__lliimmiitt==_n Sets the limit for the number of iterations allowed to be peeled in fusion, where _n >= 0. By default, _n=5. ggaatthheerr__ssccaatttteerr==_n Performs gather-scatter optimizations. _n can be one of the following: 0 Disables all gather-scatter optimization. 1 Performs gather-scatter optimizations on non-nested IIFF statements. This is the default. 2 Performs multi-level gather-scatter optimizations. iiggnnoorree__pprraaggmmaass[[ == (( OONN||OOFFFF ))]] Specifies that the command line options override directives in the source file. The default is OOFFFF. nnoonn__bblloocckkiinngg__llooaaddss[[ == (( OONN||OOFFFF ))]] (C/C++ and F77 only) Specifies whether the processor blocks on loads. If not set, the default of the current processor is used. ooiinnvvaarr[[ == (( OONN||OOFFFF ))]] Controls outer loop hoisting. The default is OONN. oopptt==_n Controls the LNO optimization level. _n can be one of the following: 0 Disables nearly all loop nest optimization. 1 Peforms full loop nest transformations. This is the default. oouutteerr[[ == (( OONN||OOFFFF ))]] Enables or disables outer loop fusion. The default is OONN. vviinnttrr[[ == (( OONN||OOFFFF ))]] Specifies that vectorizable versions of the math intrinsic functions should be used. The default is OONN. For information on the math intrinsic functions, see mmaatthh(3M). The loop transformation arguments allow you to control cache blocking, loop unrolling, and loop interchange. They are as follows: bblloocckkiinngg[[ == (( OONN||OOFFFF ))]] Specify bblloocckkiinngg==OOFFFF to disable the cache blocking transformation. The default is OONN. bblloocckkiinngg__ssiizzee==[_n_1][,,_n_2] Specifies a block size that the compiler must use when performing any blocking. When using the MIPSpro 7 Fortran 90 compiler, specify a value for _n_2 when using a 2-level cache. For _n_1 or _n_2, specify a positive integer number that represents the number of iterations. iinntteerrcchhaannggee[[ == (( OONN||OOFFFF ))]] Specifies whether or not loop interchange optimizations are performed. The default is OONN. oouu==_n Indicates that all outer loops for which unrolling is legal should be unrolled by _n, where _n is a positive integer. The compiler unrolls loops by this amount or not at all. oouu__ddeeeepp[[ == (( OONN||OOFFFF ))]] Specifies that for loops with 3-deep, or deeper, loop nests, the compiler should outer unroll the wind-down loops that result from outer unrolling loops further out. This results in large code size, but it generates much faster code whenever wind-down loop execution costs are important. The default is OONN. oouu__ffuurrtthheerr==_n Specifies whether or not the compiler performs outer loop unrolling on wind-down loops. Specify an integer for _n. oouu__mmaaxx==_n Indicates that the compiler can unroll as many as _n copies per loop, but no more. oouu__pprroodd__mmaaxx==_n Indicates that the product of unrolling of the various outer loops in a given loop nest is not to exceed _n, where _n is a positive integer. The default is 16. ppwwrr22[[ == (( OONN||OOFFFF ))]] (C/C++ and F77 only) Specifies whether to ignore the leading dimension (set to OOFFFF to ignore). You can disable additional unrolling by specifying --LLNNOO::oouu__ffuurrtthheerr==999999999999. Unrolling is enabled as much as is sensible by specifying --LLNNOO::oouu__ffuurrtthheerr==33. Certain arguments allow you to describe the target cache memory system. The numbering in the following arguments starts with the cache level closest to the processor and works outward: aassssoocc11==_n, aassssoocc22==_n, aassssoocc33==_n, aassssoocc44==_n Specifies the cache set associativity. For a fully associative cache, such as main memory, set _n to any sufficiently large number, such as 128. Specify a positive integer for _n. Specifying _n=00 indicates that there is no cache at that level. ccmmpp11==_n, ccmmpp22==_n, ccmmpp33==_n, ccmmpp44==_n ddmmpp11==_n, ddmmpp22==_n, ddmmpp33==_n, ddmmpp44==_n Specifies, in processor cycles, the time for a clean miss (ccmmpp_x==) or dirty miss (ddmmpp_x==) to the next outer level of the memory hierarchy. This number is approximate because it depends upon a clean or dirty line, read or write miss, etc. Specify a positive integer for _n. Specifying _n=00 indicates that there is no cache at that level. ccss11==_n, ccss22==_n, ccss33==_n, ccss44==_n Specifies the cache size. The value _n can be 0, or it can be a positive integer followed by one of the following letters: kk, KK, mm, or MM. This specifies the cache size in Kbytes or Mbytes. Specifying 0 indicates that there is no cache at that level. The default cache size depends on your system. You can use the --LLIISSTT::ooppttiioonnss==OONN option to see the default cache sizes used during your compilation. iiss__mmeemm11[[ == (( OONN||OOFFFF ))]] iiss__mmeemm22[[ == (( OONN||OOFFFF ))]] iiss__mmeemm33[[ == (( OONN||OOFFFF ))]] iiss__mmeemm44[[ == (( OONN||OOFFFF ))]] Specifies that certain memory hierarchies should be modeled as memory, not cache. The default is OOFFFF for each option. Blocking can be attempted for this memory hierarchy level, and blocking appropriate for memory, rather than cache, is applied. No prefetching is performed, and any prefetching options are ignored. If an --OOPPTT::iiss__mmeemm_x[[ == (( OONN||OOFFFF ))]] option is specified, the corresponding aassssoocc_x==_n specification is ignored, any ccmmpp_x==_n and ddmmpp_x==_n options on the command line are ignored. llss11==_n, llss22==_n, llss33==_n, llss44==_n Specifies the line size, in bytes. This is the number of bytes, specified in the form of a positive integer number, _n, that are moved from the memory hierarchy level further out to this level on a miss. Specifying _n=0 indicates that there is no cache at that level. Certain arguments control the TLB. The TLB is a cache for the page table, and it is assumed to be fully associative. The TLB control arguments are as follows: ppss11==_n, ppss22==_n, ppss33==_n, ppss44==_n Specifies the number of bytes in a page. Specify a positive integer for _n. The default _n depends on your system hardware. ttllbb11==_n, ttllbb22==_n, ttllbb33==_n, ttllbb44==_n Specifies the number of entries in the TLB for this cache level. Specify a positive integer for _n. The default _n depends on your system hardware. ttllbbccmmpp11==_n, ttllbbccmmpp22==_n, ttllbbccmmpp33==_n, ttllbbccmmpp44==_n ttllbbddmmpp11==_n, ttllbbddmmpp22==_n, ttllbbddmmpp33==_n, ttllbbddmmpp44==_n Specifies the number of processor cycles it takes to service a clean TLB miss (the ttllbbccmmpp_x== options) or dirty TLB miss (the ttllbbddmmpp_n== options). Specify a positive integer for _n. The default _n depends on your system hardware. The following arguments control the prefetch operation: ppff11[[ == (( OONN||OOFFFF ))]] ppff22[[ == (( OONN||OOFFFF ))]] ppff33[[ == (( OONN||OOFFFF ))]] ppff44[[ == (( OONN||OOFFFF ))]] Selectively disables and enables prefetching for cache level _x, for ppff_x[[ == (( OONN||OOFFFF ))]] When --rr1100000000 is in effect, ppff11==OONN and ppff22==OONN by default. At any other --rr_n setting, OOFFFF is in effect for all cache levels. pprreeffeettcchh==_n Specifies levels of prefetching. _n can be one of the following: 0 Disables all prefetching. This is the default when --rr44000000, --rr55000000, or --rr88000000 is in effect. 1 Enables conservative prefetching. This is the default when --rr1100000000 is in effect. 2 Enables aggressive prefetching. pprreeffeettcchh__aahheeaadd==_n Prefetches the specified number of cache lines ahead of the reference. Specify a positive integer for _n. The default is 2. pprreeffeettcchh__mmaannuuaall[[ == (( OONN||OOFFFF ))]] Specifies whether manual prefetches (through directives) should be respected or ignored. pprreeffeettcchh__mmaannuuaall==OOFFFF ignores manual prefetches. This is the default when --rr88000000, --rr55000000, or --rr44000000 is in effect. pprreeffeettcchh__mmaannuuaall==OONN respects manual prefetches. This is the default when --rr1100000000 is in effect. FF7777 LLNNOO DDiirreeccttiivveess Directives within a program unit apply only to that program unit, reverting to the default values at the end of the program unit. Directives that occur outside of a program unit alter the default value, and therefore apply to the rest of the file from that point on, until overridden by a subsequent directive. Directives within a file override the command line options by default. To have the command line options override directives, use the command line option: -LNO:ignore_pragmas FFiissssiioonn aanndd FFuussiioonn DDiirreeccttiivveess * CC**$$** AAGGGGRREESSSSIIVVEE IINNNNEERR LLOOOOPP FFIISSSSIIOONN: Fission this loop in inner_fission phase to as many loops as possible. This must be followed by a inner loop and has no effect if that loop is not inner any more after the SNL phase. * CC**$$** FFIISSSSIIOONN [[((_n))]] or CC**$$** FFIISSSSIIOONNAABBLLEE: Fission the enclosing _n level of loops after this directive. Perform legality test unless a ffiissssiioonnaabbllee directive is also specified. Does not re-order statements. * CC**$$** FFUUSSEE [[((_n [[,,_l_e_v_e_l]] ))]] or CC**$$** FFUUSSAABBLLEE: Fuse the following _n immediately adjacent loops. Fusion is attempted on each pair of adjacent loops and the _l_e_v_e_l, by default, is the determined by the maximal SNL levels of the fused loops, although partial fusion is allowed. Iterations may be peeled as needed during fusion; the peeling limit is 5 or the number specified by the --LLNNOO::ffuussiioonn__ppeeeelliinngg__lliimmiitt flag. When the FFUUSSAABBLLEE directive is present, no legality test is done and the fusion is done up to the maximal SNL levels where the iteration numbers matched for each pair of loops to be matched. The default value for _n is 2. * CC**$$** NNOO FFIISSSSIIOONN: The loop following this directive should not be fissioned in either fiz_fuse phase or inner_fission phase. Its inner loops, however, are allowed to be fissioned. * CC**$$** NNOO FFUUSSIIOONN: The loop following this directive should not be fused with other loops. SSNNLL TTrraannssffoorrmmaattiioonn DDiirreeccttiivveess The parallelizing preprocessor may do some transformation for parallelism that violate some of these directives. * CC**$$** IINNTTEERRCCHHAANNGGEE ((_I,, _J [[,,_K ......]] )): Loops _I, _J and _K (in any order) must directly follow this directive and be perfectly nested inside each other. If they are not perfectly nested, the compiler may perform loop distribution to make them so, or may ignore the annotation, or may apply imperfect interchange (this is not likely). The compiler attempts to reorder loops so that _I is outermost, then _J, then _K. The compiler may ignore this directive. There must be a minimum of 2 indexes in the directive. * CC**$$** NNOO IINNTTEERRCCHHAANNGGEE: Prevents the compiler from involving the loop directly following this directive in a permutation, or any loop nested within this loop. * CC**$$** BBLLOOCCKKIINNGG SSIIZZEE ((_n_1,,_n_2)) or CC**$$** BBLLOOCCKKIINNGG SSIIZZEE ((_n_1)) or CC**$$** BBLLOOCCKKIINNGG SSIIZZEE ((,,_n_2)): If the specified loop is involved in a blocking for the primary or secondary cache, it will have a blocksize of _n_1 or _n_2. The compiler will try to include this loop within such a block. If a blocking size is specified as 0, the loop is not actually stripped, but the entire loop is inside the block. * CC**$$** NNOO BBLLOOCCKKIINNGG: Prevent the compiler from involving this loop in a cache blocking. * CC**$$** UUNNRROOLLLL ((_n [[,,_n_2]] )): This directive suggests that _n-_1 copies of the loop body be added to the inner loop. If the loop that this directive directly preceeds is an inner loop, then it indicates standard unrolling. If the loop that this directive directly preceeds is not innermost, then outer loop unrolling is performed. _n must be at least 1. If _n=1 then no unrolling will be performed. If _n=0, then the default unrolling should be applied. _n_2 is ignored. * CC**$$** BBLLOOCCKKAABBLLEE ((_I,,_J [[,,_K ......]] )): The _I, _J and _K loops must be adjacent and nested within each other, although not necessarily perfectly nested. This directive informs the compiler that these loops may legally be involved in a blocking with each other, even if the compiler would consider such a transformation illegal. The loops are also interchangeable and unrollable. This directive does not instruct the compiler which of these transformations to apply. You must specify at least 2 loop indexes in the directive. PPrreeffeettcchh DDiirreeccttiivveess * CC**$$** PPRREEFFEETTCCHH ((_n[[,,_n]])): Specify prefetching for each level of the cache. The scope is the entire function containing the directive. _n can be one of the following values: 00 prefetching off (default for all processors except R10000) 11 prefetching on, but conservative 22 prefetching on, and aggressive (default when prefetch is on) * CC**$$** PPRREEFFEETTCCHH__MMAANNUUAALL ((_n)): Specify if manual prefetches (through directives) should be respected or ignored. Scope: Entire function containing the directive. _n can be one of the following values: 00 ignore manual prefetches (default for mmiippss33 and earlier) 11 respect manual prefetches (default for mmiippss44) * CC**$$** PPRREEFFEETTCCHH__RREEFF__DDIISSAABBLLEE==_A [[,, ssiizzee==_n_u_m]]: This directive explicitly disables prefetching all references to array _A in the current function. The auto-prefetcher runs (if enabled) ignoring array _A. The ssiizzee is used for volume analysis. Scope: Entire function containing the directive. ssiizzee==_n_u_m is the size of the array references in this loop, in Kbytes. This is an optional argument and must be a constant. * CC**$$** PPRREEFFEETTCCHH__RREEFF==_a_r_r_a_y-_r_e_f,,[[ssttrriiddee==[[_s_t_r]] [[,,_s_t_r]]]],, [[lleevveell==[[_l_e_v]] [[,,_l_e_v]]]],, [[kkiinndd==[[_r_d/_w_r]]]],, [[ssiizzee==[[_s_z]]]]: This directive generates a single prefetch instruction to the specified memory location. It searches for array references that match the supplied reference in the current loop-nest. If such a reference is found, that reference is connected to this prefetch node with the specified latency. If no such reference is found, this prefetch node stays free-floating and is scheduled "loosely". All references to this array in this loop-nest are ignored by the automatic prefetcher (if enabled). If the size is supplied, then the auto-prefetcher (if enabled) reduces the effective cache size by that amount in its calculations. The compiler tries to issue one prefetch per stride iteration, but cannot guarantee it. Redundant prefetches are preferred to transformations (such as inserting conditionals) which incur other overhead. Scope: No scope. Just generates a prefetch instruction. The following arguments are used with this option: _a_r_r_a_y-_r_e_f Required. The reference itself, for example, AA((ii,, jj)). _s_t_r Optional. Prefetch every _s_t_r iterations of this loop. The default is 1. _l_e_v Optional. The level in memory hierarchy to prefetch. The default is 2. If _l_e_v=1, prefetch from L2 to L1 cache. If _l_e_v=2, prefetch from memory to L1 cache. _r_d/_w_r Optional. The default is read/write. _s_z Optional. The size (in Kbytes) of the array referenced in this loop. This must be a constant. DDeeppeennddeennccee AAnnaallyyssiiss DDiirreeccttiivveess * CCDDIIRR$$ IIVVDDEEPP: This applies only to inner loops. Liberalize dependence analysis. Given two memory references, where at least one is loop variant, ignore any loop-carried dependences between the two references. The following are examples of this directive. do i = 1,n b(k) = b(k) + a(i) enddo IIVVDDEEPP does not break the dependence because bb((kk)) is not loop-variant. do i=1,n a(i) = a(i-1) + 3 enddo IIVVDDEEPP does break the dependence but the compiler warns the user that it is breaking an obvious dependence. do i=1,n a(b(i)) = a(b(i)) + 3. enddo IIVVDDEEPP does break the dependence. do i = 1,n a(i) = b(i) c(i) = a(i) + 3. enddo IIVVDDEEPP does not break the dependence on aa[[ii]] because it is within an iteration. If --OOPPTT::ccrraayy__iivvddeepp==OONN,, Cray semantics are used and all lexically backwards dependences are broken. The following are examples: do i=1,n a(i) = a(i-1) + 3. enddo IIVVDDEEPP does break the dependence but the compiler warns the user that it's breaking an obvious dependence. do i=1,n a(i) = a(i+1) + 3. enddo IIVVDDEEPP does not break the dependence because the dependence is from the load to the store, and the load comes lexically before the store. If --OOPPTT::lliibbeerraall__iivvddeepp==OONN,, all dependences are broken. SSEEEE AALLSSOO cccc(1), CCCC(1), ccoorrdd(1), ddssoo(1), ff7777(1), ff9900(1), ffppmmooddee(1), hhiinnvv(1), lldd(1), mmaakkee(1), ppiixxiiee(1), ppmmaakkee(1), pprrooff(1), rrlldd(1), ssmmaakkee(1). mmaatthh(3M). aauuttoo__pp(5), ggpp__oovveerrffllooww(5), iippaa(5), oopptt(5), ppee__eennvviirroonn(5). _M_I_P_S_p_r_o _C _a_n_d _C++ _P_r_a_g_m_a_s, publication 007-3587-001 _C _L_a_n_g_u_a_g_e _R_e_f_e_r_e_n_c_e _M_a_n_u_a_l, publication 007-0701-120 _C_o_m_p_i_l_e_r _I_n_f_o_r_m_a_t_i_o_n _F_i_l_e (_C_I_F) _R_e_f_e_r_e_n_c_e _M_a_n_u_a_l, publication SR-2401 _M_I_P_S_P_r_o _7 _F_o_r_t_r_a_n _9_0 _C_o_m_m_a_n_d_s _a_n_d _D_i_r_e_c_t_i_v_e_s _R_e_f_e_r_e_n_c_e _M_a_n_u_a_l, publication SR-3907 This man page is available only online.