summaryrefslogtreecommitdiffstats
Commit message (Collapse)AuthorAgeFilesLines
* Separate LowerOperation() into multiple functions, one per opcode.Evan Cheng2006-04-252-1135/+1213
| | | | llvm-svn: 27972
* slightly more useful error messageAndrew Lenharth2006-04-251-2/+3
| | | | llvm-svn: 27971
* better c99 struct handlingAndrew Lenharth2006-04-251-2/+1
| | | | llvm-svn: 27970
* another c99 style problemAndrew Lenharth2006-04-251-0/+22
| | | | llvm-svn: 27969
* Fix a typo.Evan Cheng2006-04-251-1/+1
| | | | llvm-svn: 27968
* Fix a warningNate Begeman2006-04-251-1/+1
| | | | llvm-svn: 27967
* No functionality changes, but cleaner code with correct comments.Nate Begeman2006-04-251-34/+40
| | | | llvm-svn: 27966
* Explicitly specify result type for def : Pat<> patterns (if it produces a vectorEvan Cheng2006-04-251-45/+47
| | | | | | result). Otherwise tblgen will pick the default (v16i8 for 128-bit vector). llvm-svn: 27965
* Added X86 SSE2 intrinsics which can be represented as vector_shuffles. This isEvan Cheng2006-04-242-13/+73
| | | | | | | a temporary workaround for the 2-wide vector_shuffle problem (i.e. its mask would have type v2i32 which is not legal). llvm-svn: 27964
* Add a new entry.Evan Cheng2006-04-241-0/+32
| | | | llvm-svn: 27963
* Added a movq test case.Evan Cheng2006-04-241-0/+10
| | | | llvm-svn: 27962
* Special case handling two wide build_vector(0, x).Evan Cheng2006-04-241-4/+4
| | | | llvm-svn: 27961
* Some missing movlps, movhps, movlpd, and movhpd patterns.Evan Cheng2006-04-241-6/+14
| | | | llvm-svn: 27960
* A little bit more build_vector enhancement for v8i16 cases.Evan Cheng2006-04-241-42/+105
| | | | llvm-svn: 27959
* Remove a completed entry.Evan Cheng2006-04-241-55/+0
| | | | llvm-svn: 27958
* Suggest looking at the SPARC backend. How weird is that?? :)Chris Lattner2006-04-241-2/+3
| | | | llvm-svn: 27957
* Added addJumpTableIndexEvan Cheng2006-04-241-0/+5
| | | | llvm-svn: 27956
* MakeMIInst() should handle jump table index operands.Evan Cheng2006-04-241-0/+3
| | | | llvm-svn: 27955
* Add a noteChris Lattner2006-04-231-0/+7
| | | | llvm-svn: 27954
* MOVL shuffle (i.e. movd or movss / movsd from memory) of undef, V2 == V2Evan Cheng2006-04-231-2/+7
| | | | llvm-svn: 27953
* Fix the updating of the machine CFG when a PHI node was in a successor ofNate Begeman2006-04-232-8/+16
| | | | | | | the jump table's range check block. This re-enables 100% dense jump tables by default on PPC & x86 llvm-svn: 27952
* Keep Visual Studio informed.Jeff Cohen2006-04-231-0/+3
| | | | llvm-svn: 27951
* Code cleanup associated with jump tables, thanks to Chris for noticingNate Begeman2006-04-223-9/+19
| | | | | | these. llvm-svn: 27950
* Turn of jump tables for a bit, there are still some issues to work out withNate Begeman2006-04-221-1/+1
| | | | | | updating the machine CFG. llvm-svn: 27949
* Optimized stores to the constant pool, while cool, are unnecessary.Nate Begeman2006-04-221-3/+0
| | | | llvm-svn: 27948
* JumpTable support! What this represents is working asm and jit support forNate Begeman2006-04-2240-77/+717
| | | | | | | | x86 and ppc for 100% dense switch statements when relocations are non-PIC. This support will be extended and enhanced in the coming days to support PIC, and less dense forms of jump tables. llvm-svn: 27947
* Don't do all the lowering stuff for 2-wide build_vector's. Also, minor ↵Evan Cheng2006-04-221-2/+5
| | | | | | optimization for shuffle of undef. llvm-svn: 27946
* Fix a performance regression. Use {p}shuf* when there are only two distinct ↵Evan Cheng2006-04-221-13/+17
| | | | | | elements in a build_vector. llvm-svn: 27945
* Two more build_vector tests.Evan Cheng2006-04-222-0/+37
| | | | llvm-svn: 27944
* Teach the JIT how to relocate LI, this fixes the JIT on Prolangs-C/TimberWolfMCChris Lattner2006-04-221-0/+1
| | | | llvm-svn: 27943
* This no longer needs the JIT.Chris Lattner2006-04-221-4/+1
| | | | llvm-svn: 27942
* Fix JIT support for static ctors, which was apparently completely broken!Chris Lattner2006-04-221-1/+5
| | | | | | | This allows Prolangs-C++/city and probably a bunch of other stuff to work well with the new front-end llvm-svn: 27941
* Revamp build_vector lowering to take advantage of movss and movd instructions.Evan Cheng2006-04-211-64/+141
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | movd always clear the top 96 bits and movss does so when it's loading the value from memory. The net result is codegen for 4-wide shuffles is much improved. It is near optimal if one or more elements is a zero. e.g. __m128i test(int a, int b) { return _mm_set_epi32(0, 0, b, a); } compiles to _test: movd 8(%esp), %xmm1 movd 4(%esp), %xmm0 punpckldq %xmm1, %xmm0 ret compare to gcc: _test: subl $12, %esp movd 20(%esp), %xmm0 movd 16(%esp), %xmm1 punpckldq %xmm0, %xmm1 movq %xmm1, %xmm0 movhps LC0, %xmm0 addl $12, %esp ret or icc: _test: movd 4(%esp), %xmm0 #5.10 movd 8(%esp), %xmm3 #5.10 xorl %eax, %eax #5.10 movd %eax, %xmm1 #5.10 punpckldq %xmm1, %xmm0 #5.10 movd %eax, %xmm2 #5.10 punpckldq %xmm2, %xmm3 #5.10 punpckldq %xmm3, %xmm0 #5.10 ret #5.10 There are still room for improvement, for example the FP variant of the above example: __m128 test(float a, float b) { return _mm_set_ps(0.0, 0.0, b, a); } _test: movss 8(%esp), %xmm1 movss 4(%esp), %xmm0 unpcklps %xmm1, %xmm0 xorps %xmm1, %xmm1 movlhps %xmm1, %xmm0 ret The xorps and movlhps are unnecessary. This will require post legalizer optimization to handle. llvm-svn: 27939
* Fix the commentNate Begeman2006-04-211-1/+1
| | | | llvm-svn: 27938
* Change the PPC JIT to use a Static relocation modelNate Begeman2006-04-213-29/+4
| | | | llvm-svn: 27937
* typoChris Lattner2006-04-211-1/+1
| | | | llvm-svn: 27936
* fix thinkoChris Lattner2006-04-211-2/+2
| | | | llvm-svn: 27935
* add some low-prio notesChris Lattner2006-04-211-0/+69
| | | | llvm-svn: 27934
* The BFS scheduler is apparently nondeterminstic (causes many llvmgcc bootstrapChris Lattner2006-04-211-2/+5
| | | | | | miscompares). Switch RISC targets to use the list-td scheduler, which isn't. llvm-svn: 27933
* movddup is a SSE3 instruction.Evan Cheng2006-04-211-1/+1
| | | | llvm-svn: 27932
* Remove a hack required by V9.Chris Lattner2006-04-211-4/+2
| | | | llvm-svn: 27931
* Fix a couple more memory issuesChris Lattner2006-04-211-4/+4
| | | | llvm-svn: 27930
* Remove the extraneous --defined-only option to nm. This is the default andReid Spencer2006-04-211-3/+3
| | | | | | some versions of nm don't recognize it (its a gnu option). llvm-svn: 27928
* Add && to each RUN: line (except the last one).Evan Cheng2006-04-2111-13/+14
| | | | llvm-svn: 27927
* More build_vector tests.Evan Cheng2006-04-213-0/+62
| | | | llvm-svn: 27926
* Check for llc crash.Evan Cheng2006-04-212-0/+2
| | | | llvm-svn: 27925
* To be replaced with another test.Evan Cheng2006-04-211-21/+0
| | | | llvm-svn: 27924
* Now generating perfect (I think) code for "vector set" with a single non-zeroEvan Cheng2006-04-213-104/+174
| | | | | | | | | | | | | | | | | | | scalar value. e.g. _mm_set_epi32(0, a, 0, 0); ==> movd 4(%esp), %xmm0 pshufd $69, %xmm0, %xmm0 _mm_set_epi8(0, 0, 0, 0, 0, a, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0); ==> movzbw 4(%esp), %ax movzwl %ax, %eax pxor %xmm0, %xmm0 pinsrw $5, %eax, %xmm0 llvm-svn: 27923
* Fix a really subtle and obnoxious memory bug that caused issues with anChris Lattner2006-04-201-11/+11
| | | | | | | | | llvm-gcc4 boostrap. Whenever a node is deleted by the dag combiner, it *must* be returned by the visit function, or the dag combiner will not know that the node has been processed (and will, e.g., send it to the target dag combine xforms). llvm-svn: 27922
* Don't require the "dot" program if -flat option is given, and don't produceReid Spencer2006-04-201-19/+25
| | | | | | | any of the fancy graphs or other output. Just produce the flat, makefile style output on stdout. llvm-svn: 27921
OpenPOWER on IntegriCloud