1) dynamic stack alloc
2) loads
3) shifts
4) subtract
5) immediate form of add, and, or, xor
6) change flag from -pattern-isel to -enable-ppc-pattern-isel
Remove dead arguments from getGlobalBaseReg in the simple ISel
llvm-svn: 20810
- unsigned TrueValue = getReg(TrueVal, BB, BB->begin());
+ unsigned TrueValue = getReg(TrueVal);
Fixes the PPC regressions from last night.
The other hunk is just a clarity improvement.
llvm-svn: 19263
While we're at it, improve codegen of select instructions. For this
testcase:
int %test(bool %C, int %A, int %B) {
%D = select bool %C, int %A, int %B
ret int %D
}
We used to generate this code:
_test:
cmpwi cr0, r3, 0
bne .LBB_test_2 ;
.LBB_test_1: ;
b .LBB_test_3 ;
.LBB_test_2: ;
or r5, r4, r4
.LBB_test_3: ;
or r3, r5, r5
blr
Now we emit:
_test:
cmpwi cr0, r3, 0
bne .LBB_test_2 ;
.LBB_test_1: ;
or r4, r5, r5
.LBB_test_2: ;
or r3, r4, r4
blr
-Chris
llvm-svn: 19214
addi r3, r3, -1
instead of
addi r3, r3, 1
for 'sub int X, 1'.
Secondarily, this fixes several cases where we could crash given an unsigned
constant. And fixes a couple of minor missed optimization cases, such as
xor X, ~0U -> not X
llvm-svn: 18379
place to help bring up the PowerPC back end on Darwin. This code is no
longer serves any purpose now that the AsmPrinter does the right thing
all the time printing GlobalValues. --Cruft.
llvm-svn: 18267
static global variables whose addresses are taken. This allows us to
convert the following code for taking the address of a static function foo
addis r2, r30, ha16(Ll1__2E_foo_2$non_lazy_ptr-"L00001$pb")
lwz r3, lo16(Ll1__2E_foo_2$non_lazy_ptr-"L00001$pb")(r2)
which also includes linker stub code emitted at the end of the .s file not
shown here, and replace it with this:
addis r2, r30, ha16(l1__2E_foo_2-"L00001$pb")
la r3, lo16(l1__2E_foo_2-"L00001$pb")(r2)
which in addition to not needing linker help, also has no load instruction.
For those not up on PowerPC mnemonics, la is shorthand for add immediate.
llvm-svn: 18239
reg-reg copies. The necessary conditions for this bug are a GEP that is
used outside the basic block in which it is defined, whose components
other than the pointer are all constant zero, and where the use is
selected before the definition (backwards branch to successsor block).
llvm-svn: 18084
directly rather than making a copy for the register allocator to coalesce.
This kills thousands of live intervals across the testsuite.
llvm-svn: 17946
int test(int x) { return 32768 - x; }
Fixed by teaching the function that checks a constant's validity to be used
as an immediate argument about subtract-from instructions.
llvm-svn: 17476
by the recently committed rlwimi.ll test file. Also commit initial code
for bitfield extract, although it is turned off until fully debugged.
llvm-svn: 17207
as the shift amount operand to a shift instruction. This was causing us to
emit unnecessary clear operations for code such as:
int foo(int x) { return 1 << x; }
llvm-svn: 17175
1. optional shift left
2. and x, immX
3. and y, immY
4. or z, x, y
==> rlwimi z, x, y, shift, mask begin, mask end
where immX == ~immY and immX is a run of set bits. This transformation
fires 32 times on voronoi, once on espresso, and probably several
dozen times on external benchmarks such as gcc.
To put this in terms of actual code generated for
struct B { unsigned a : 3; unsigned b : 2; };
void storeA (struct B *b, int v) { b->a = v;}
void storeB (struct B *b, int v) { b->b = v;}
Old:
_storeA:
rlwinm r2, r4, 0, 29, 31
lwz r4, 0(r3)
rlwinm r4, r4, 0, 0, 28
or r2, r4, r2
stw r2, 0(r3)
blr
_storeB:
rlwinm r2, r4, 3, 0, 28
rlwinm r2, r2, 0, 27, 28
lwz r4, 0(r3)
rlwinm r4, r4, 0, 29, 26
or r2, r2, r4
stw r2, 0(r3)
blr
New:
_storeA:
lwz r2, 0(r3)
rlwimi r2, r4, 0, 29, 31
stw r2, 0(r3)
blr
_storeB:
lwz r2, 0(r3)
rlwimi r2, r4, 3, 27, 28
stw r2, 0(r3)
blr
llvm-svn: 17078
flag rotate left word immediate then mask insert (rlwimi) as a two-address
instruction, and update the ISel usage of the instruction accordingly.
This will allow us to properly schedule rlwimi, and use it to efficiently
codegen bitfield operations.
llvm-svn: 17068
This transformation fires a few dozen times across the testsuite.
For example, int test2(int X) { return X ^ 0x0FF00FF0; }
Old:
_test2:
lis r2, 4080
ori r2, r2, 4080
xor r3, r3, r2
blr
New:
_test2:
xoris r3, r3, 4080
xori r3, r3, 4080
blr
llvm-svn: 17004
of one or more 1 bits (may wrap from least significant bit to most
significant bit) as the rlwinm rather than andi., andis., or some longer
instructons sequence.
int andn4(int z) { return z & -4; }
int clearhi(int z) { return z & 0x0000FFFF; }
int clearlo(int z) { return z & 0xFFFF0000; }
int clearmid(int z) { return z & 0x00FFFF00; }
int clearwrap(int z) { return z & 0xFF0000FF; }
_andn4:
rlwinm r3, r3, 0, 0, 29
blr
_clearhi:
rlwinm r3, r3, 0, 16, 31
blr
_clearlo:
rlwinm r3, r3, 0, 0, 15
blr
_clearmid:
rlwinm r3, r3, 0, 8, 23
blr
_clearwrap:
rlwinm r3, r3, 0, 24, 7
blr
llvm-svn: 16832
1. Fix an illegal argument to getClassB when deciding whether or not to
sign extend a byte load.
2. Initial addition of isLoad and isStore flags to the instruction .td file
for eventual use in a scheduler.
3. Rewrite of how constants are handled in emitSimpleBinaryOperation so
that we can emit the PowerPC shifted immediate instructions far more
often. This allows us to emit the following code:
int foo(int x) { return x | 0x00F0000; }
_foo:
.LBB_foo_0: ; entry
; IMPLICIT_DEF
oris r3, r3, 15
blr
llvm-svn: 16826
integers that we can use as immediate values in instructions.
Example from yacr2:
- lis r10, -1
- ori r10, r10, 65535
- add r28, r28, r10
+ addi r28, r28, -1
addi r7, r7, 1
addi r9, r9, 1
b .LBB_main_9 ; loopentry.1.i214
llvm-svn: 16566
the ISel to use indexed and non-zero immediate offsets for GEPs that have
more than one use. This is common for instruction sequences such as a load
followed by a modify and store to the same address.
llvm-svn: 16493
