diff options
| author | Lynn Boger <laboger@linux.vnet.ibm.com> | 2023-04-17 10:02:48 -0500 |
|---|---|---|
| committer | Lynn Boger <laboger@linux.vnet.ibm.com> | 2023-04-21 16:47:45 +0000 |
| commit | e23322e2ccd19b5802a823d20a089540afef79ce (patch) | |
| tree | df27fc09f5ee19f00d2b1319ddfcc22b1d22d846 /src/math | |
| parent | de788efeac46f18bd3372666cfc2c698de69d8b6 (diff) | |
| download | go-e23322e2ccd19b5802a823d20a089540afef79ce.tar.xz | |
cmd/internal/obj/ppc64: modify PCALIGN to ensure alignment
The initial purpose of PCALIGN was to identify code
where it would be beneficial to align code for performance,
but avoid cases where too many NOPs were added. On p10, it
is now necessary to enforce a certain alignment in some
cases, so the behavior of PCALIGN needs to be slightly
different. Code will now be aligned to the value specified
on the PCALIGN instruction regardless of number of NOPs added,
which is more intuitive and consistent with power assembler
alignment directives.
This also adds 64 as a possible alignment value.
The existing values used in PCALIGN were modified according to
the new behavior.
A testcase was updated and performance testing was done to
verify that this does not adversely affect performance.
Change-Id: Iad1cf5ff112e5bfc0514f0805be90e24095e932b
Reviewed-on: https://go-review.googlesource.com/c/go/+/485056
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
Reviewed-by: Archana Ravindar <aravind5@in.ibm.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
Reviewed-by: Paul Murphy <murp@ibm.com>
Reviewed-by: Bryan Mills <bcmills@google.com>
Diffstat (limited to 'src/math')
| -rw-r--r-- | src/math/big/arith_ppc64x.s | 14 |
1 files changed, 7 insertions, 7 deletions
diff --git a/src/math/big/arith_ppc64x.s b/src/math/big/arith_ppc64x.s index 5fdbf40a24..0613f5c3ad 100644 --- a/src/math/big/arith_ppc64x.s +++ b/src/math/big/arith_ppc64x.s @@ -45,7 +45,7 @@ TEXT ·addVV(SB), NOSPLIT, $0 // gain significant performance as z_len increases (up to // 1.45x). - PCALIGN $32 + PCALIGN $16 loop: MOVD 8(R8), R11 // R11 = x[i] MOVD 16(R8), R12 // R12 = x[i+1] @@ -134,7 +134,7 @@ TEXT ·subVV(SB), NOSPLIT, $0 // gain significant performance as z_len increases (up to // 1.45x). - PCALIGN $32 + PCALIGN $16 loop: MOVD 8(R8), R11 // R11 = x[i] MOVD 16(R8), R12 // R12 = x[i+1] @@ -216,7 +216,7 @@ TEXT ·addVW(SB), NOSPLIT, $0 CMP R0, R9 MOVD R9, CTR // Set up the loop counter BEQ tail // If R9 = 0, we can't use the loop - PCALIGN $32 + PCALIGN $16 loop: MOVD 8(R8), R20 // R20 = x[i] @@ -294,7 +294,7 @@ TEXT ·subVW(SB), NOSPLIT, $0 // we don't need to capture CA every iteration because we've already // done that above. - PCALIGN $32 + PCALIGN $16 loop: MOVD 8(R8), R20 MOVD 16(R8), R21 @@ -365,7 +365,7 @@ TEXT ·shlVU(SB), NOSPLIT, $0 CMP R5, R0 // iterate from i=len(z)-1 to 0 BEQ loopexit // Already at end? MOVD 0(R15),R10 // x[i] - PCALIGN $32 + PCALIGN $16 shloop: SLD R9, R10, R10 // x[i]<<s MOVDU -8(R15), R14 @@ -528,7 +528,7 @@ TEXT ·mulAddVWW(SB), NOSPLIT, $0 CMP R0, R14 MOVD R14, CTR // Set up the loop counter BEQ tail // If R9 = 0, we can't use the loop - PCALIGN $32 + PCALIGN $16 loop: MOVD 8(R8), R20 // R20 = x[i] @@ -611,7 +611,7 @@ TEXT ·addMulVVW(SB), NOSPLIT, $0 MOVD R0, R4 // R4 = c = 0 MOVD R22, CTR // Initialize loop counter BEQ done - PCALIGN $32 + PCALIGN $16 loop: MOVD (R8)(R3), R20 // Load x[i] |