diff options
Diffstat (limited to 'sysdeps/powerpc/powerpc64/le/power9/strcmp.S')
| -rw-r--r-- | sysdeps/powerpc/powerpc64/le/power9/strcmp.S | 44 |
1 files changed, 13 insertions, 31 deletions
diff --git a/sysdeps/powerpc/powerpc64/le/power9/strcmp.S b/sysdeps/powerpc/powerpc64/le/power9/strcmp.S index 83b21c6..f0cde81 100644 --- a/sysdeps/powerpc/powerpc64/le/power9/strcmp.S +++ b/sysdeps/powerpc/powerpc64/le/power9/strcmp.S @@ -28,21 +28,6 @@ The implementation uses unaligned doubleword access for first 32 bytes as in POWER8 patch and uses vectorised loops after that. */ -/* TODO: Change this to actual instructions when minimum binutils is upgraded - to 2.27. Macros are defined below for these newer instructions in order - to maintain compatibility. */ -#define VCTZLSBB(r,v) .long (0x10010602 | ((r)<<(32-11)) | ((v)<<(32-21))) - -#define VEXTUBRX(t,a,b) .long (0x1000070d \ - | ((t)<<(32-11)) \ - | ((a)<<(32-16)) \ - | ((b)<<(32-21)) ) - -#define VCMPNEZB(t,a,b) .long (0x10000507 \ - | ((t)<<(32-11)) \ - | ((a)<<(32-16)) \ - | ((b)<<(32-21)) ) - /* Get 16 bytes for unaligned case. reg1: Vector to hold next 16 bytes. reg2: Address to read from. @@ -61,10 +46,7 @@ 2: \ vperm reg1, v9, reg1, reg3; -/* TODO: change this to .machine power9 when the minimum required binutils - allows it. */ - - .machine power7 + .machine power9 ENTRY_TOCLESS (STRCMP, 4) li r0, 0 @@ -116,7 +98,7 @@ L(align): /* Both s1 and s2 are unaligned. */ GET16BYTES(v4, r7, v10) GET16BYTES(v5, r4, v6) - VCMPNEZB(v7, v5, v4) + vcmpnezb. v7, v5, v4 beq cr6, L(match) b L(different) @@ -136,28 +118,28 @@ L(match): L(s1_align): lvx v4, r7, r0 GET16BYTES(v5, r4, v6) - VCMPNEZB(v7, v5, v4) + vcmpnezb. v7, v5, v4 addi r7, r7, 16 addi r4, r4, 16 bne cr6, L(different) lvx v4, r7, r0 GET16BYTES(v5, r4, v6) - VCMPNEZB(v7, v5, v4) + vcmpnezb. v7, v5, v4 addi r7, r7, 16 addi r4, r4, 16 bne cr6, L(different) lvx v4, r7, r0 GET16BYTES(v5, r4, v6) - VCMPNEZB(v7, v5, v4) + vcmpnezb. v7, v5, v4 addi r7, r7, 16 addi r4, r4, 16 bne cr6, L(different) lvx v4, r7, r0 GET16BYTES(v5, r4, v6) - VCMPNEZB(v7, v5, v4) + vcmpnezb. v7, v5, v4 addi r7, r7, 16 addi r4, r4, 16 beq cr6, L(s1_align) @@ -167,37 +149,37 @@ L(s1_align): L(aligned): lvx v4, 0, r7 lvx v5, 0, r4 - VCMPNEZB(v7, v5, v4) + vcmpnezb. v7, v5, v4 addi r7, r7, 16 addi r4, r4, 16 bne cr6, L(different) lvx v4, 0, r7 lvx v5, 0, r4 - VCMPNEZB(v7, v5, v4) + vcmpnezb. v7, v5, v4 addi r7, r7, 16 addi r4, r4, 16 bne cr6, L(different) lvx v4, 0, r7 lvx v5, 0, r4 - VCMPNEZB(v7, v5, v4) + vcmpnezb. v7, v5, v4 addi r7, r7, 16 addi r4, r4, 16 bne cr6, L(different) lvx v4, 0, r7 lvx v5, 0, r4 - VCMPNEZB(v7, v5, v4) + vcmpnezb. v7, v5, v4 addi r7, r7, 16 addi r4, r4, 16 beq cr6, L(aligned) /* Calculate and return the difference. */ L(different): - VCTZLSBB(r6, v7) - VEXTUBRX(r5, r6, v4) - VEXTUBRX(r4, r6, v5) + vctzlsbb r6, v7 + vextubrx r5, r6, v4 + vextubrx r4, r6, v5 subf r3, r4, r5 extsw r3, r3 blr |