Commits · 4eac5f59e64b96b2ec0e0377cbf746fda5b587f4 · Roger Ferrer / llvm-epi-0.8

Nov 07, 2006

Fixed a bug which causes x86 be to incorrectly match · 922e1911

Evan Cheng authored Nov 07, 2006

shuffle v, undef, <2, ?, 3, ?>
to movhlps
It should match to unpckhps instead.

Added proper matching code for
shuffle v, undef, <2, 3, 2, 3>

llvm-svn: 31519

922e1911

Nov 04, 2006
- remove dead/redundant vars · 9ac6442d
  Chris Lattner authored Nov 03, 2006
```
llvm-svn: 31435
```
  9ac6442d
Nov 01, 2006
- Fix ldmxcsr JIT encoding. · 94e5bc9e
  Evan Cheng authored Nov 01, 2006
```
llvm-svn: 31343
```
  94e5bc9e
Oct 27, 2006
- Fixed a significant bug where unpcklpd is incorrectly used to extract element 1 from a v2f64 value. · e056dd59
  Evan Cheng authored Oct 27, 2006
```
llvm-svn: 31228
```
  e056dd59
Oct 25, 2006
- X86ISD::PEXTRW 3rd operand type is always target pointer type. · 7e065ff7
  Evan Cheng authored Oct 25, 2006
```
llvm-svn: 31185
```
  7e065ff7
Oct 11, 2006
- ComplexPatterns sse_load_f32 and sse_load_f64 returns in / out chain operands. · 4090dc47
  Evan Cheng authored Oct 11, 2006
```
llvm-svn: 30892
```
  4090dc47
Oct 09, 2006
- Don't go too crazy with these AddComplexity. Try matching shufps with load · 57ccb6d3
  Evan Cheng authored Oct 09, 2006
```
folding first.

llvm-svn: 30848
```
  57ccb6d3
- Reflects ISD::LOAD / ISD::LOADX / LoadSDNode changes. · e71fe34d
  Evan Cheng authored Oct 09, 2006
```
llvm-svn: 30844
```
  e71fe34d
Oct 07, 2006
- completely disable folding of loads into scalar sse instructions and provide · 398195eb
  Chris Lattner authored Oct 07, 2006
```
a framework for doing it right.  This fixes
CodeGen/X86/2006-10-07-ScalarSSEMiscompile.ll.

Once X86DAGToDAGISel::SelectScalarSSELoad is implemented right, this task
will be done.

llvm-svn: 30817
```
  398195eb
- convert packed FP add/sub/mul/div to use a multiclass. · 942009fe
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30815
```
  942009fe
- one multiclass now defines all 8 variants of binary-scalar-sse-fp operations. · 4005f4e4
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30814
```
  4005f4e4
- Switch ADD/MUL/DIV/SUB scalarsse fp ops to a multiclass · 6eaee2c8
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30813
```
  6eaee2c8
- Random acts of shrinkage · c8c64418
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30812
```
  c8c64418
- Convert pand/por/pxor to use multiclass · b5df7e55
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30811
```
  b5df7e55
- Convert some more instructions over to use a new multiclass. · 6138cba5
  Chris Lattner authored Oct 07, 2006
```
Fix a bug where the asmstring for PSUBQrm was wrong.

llvm-svn: 30810
```
  6138cba5
- Fix a bug where PADDQrm printed paddd instead of paddq. · 662ba43f
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30809
```
  662ba43f
- Add multiclass for SSE2 instructions that correspond to simple binops. · 29c62a3c
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30808
```
  29c62a3c
- rename: · e0928d9d
  Chris Lattner authored Oct 07, 2006
```
  PDI_binop_rm -> PDI_binop_rm_int
  PDI_binop_rmi -> PDI_binop_rmi_int

to make it clear that these are for use with intrinsics.

llvm-svn: 30807
```
  e0928d9d
- Convert saturating PADD/PSUB's to use a multiclass · 489b6308
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30806
```
  489b6308
- Convert PAVG*, PMADDWD, and PMUL* to use multiclasses. · fa2ce882
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30805
```
  fa2ce882
- Fix typo in packsswb instr definition, where the load had the wrong type. · cab92e4c
  Chris Lattner authored Oct 07, 2006
```
This allows us to use the multiclass for other packs.

llvm-svn: 30804
```
  cab92e4c
- handle pmin/pmax with multiclasses · e746a9cd
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30800
```
  e746a9cd
- simplify pack and shift intrinsics with multiclasses · b14e6a0f
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30797
```
  b14e6a0f
- Use a multiclass to simplify 'SSE2 Integer comparison' · 521fc4e3
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30796
```
  521fc4e3
- move class defns close to uses to make it easier to read · c6138cec
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30795
```
  c6138cec
- simplify horizontal op definitions · 87e69232
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30794
```
  87e69232
- remove more unneeded type info · 3e9fc374
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30793
```
  3e9fc374
- remove unneeded definitions and type info · 807be0a7
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30792
```
  807be0a7
- remove some unneeded type info · 5b1358a8
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30791
```
  5b1358a8
- simplify patterns by merging in operand info · 3414c022
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30790
```
  3414c022
- Factor operands into packed unary classes · ca21ce5f
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30789
```
  ca21ce5f
- remove dead/duplicate instructions · 0052c3ff
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30788
```
  0052c3ff
- Pull operand info up into parent class for scalar sse intrinsics. · 904c6e9c
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30787
```
  904c6e9c
- convert the sole sd unary intrinsic to a multiclass for consistency · e698c90e
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30786
```
  e698c90e
- pull operand string into the multiclass · 2bb2f050
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30785
```
  2bb2f050
- Remove RSQRTSS[rm] RCPSS[rm], which are dead. · 069679c7
  Chris Lattner authored Oct 07, 2006
```
Introduce SS_IntUnary, a multiclass to replace SS_Int[rm].

llvm-svn: 30784
```
  069679c7
- eliminate redundancy · f13a7b37
  Chris Lattner authored Oct 07, 2006
```
llvm-svn: 30783
```
  f13a7b37
Oct 03, 2006
- These don't have immediate operands. · a36e6cf4
  Evan Cheng authored Oct 03, 2006
```
llvm-svn: 30694
```
  a36e6cf4
Sep 11, 2006
- X86ISD::CMP now produces a chain as well as a flag. Make that the chain · 4259a0f6
  Evan Cheng authored Sep 11, 2006
```
operand of a conditional branch to allow load folding into CMP / TEST
instructions.

llvm-svn: 30241
```
  4259a0f6
Sep 05, 2006
- JIT encoding bug. · 17c28b2e
  Evan Cheng authored Sep 05, 2006
```
llvm-svn: 30112
```
  17c28b2e