From: Richard Henderson <rth@twiddle.net>
To: qemu-devel@nongnu.org
Cc: peter.maydell@linaro.org, claudio.fontana@gmail.com
Subject: [Qemu-devel] [PATCH 04/26] tcg-aarch64: Use MOVN in tcg_out_movi
Date: Fri, 14 Mar 2014 19:48:30 -0700 [thread overview]
Message-ID: <1394851732-25692-5-git-send-email-rth@twiddle.net> (raw)
In-Reply-To: <1394851732-25692-1-git-send-email-rth@twiddle.net>
When profitable, initialize the register with MOVN instead of MOVZ,
before setting the remaining lanes with MOVK.
Signed-off-by: Richard Henderson <rth@twiddle.net>
---
tcg/aarch64/tcg-target.c | 62 ++++++++++++++++++++++++++++++++++++++----------
1 file changed, 49 insertions(+), 13 deletions(-)
diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
index 47f4708..a7b6796 100644
--- a/tcg/aarch64/tcg-target.c
+++ b/tcg/aarch64/tcg-target.c
@@ -531,24 +531,60 @@ static void tcg_out_movi(TCGContext *s, TCGType type, TCGReg rd,
tcg_target_long value)
{
AArch64Insn insn;
-
- if (type == TCG_TYPE_I32) {
+ int i, wantinv, shift;
+ tcg_target_long svalue = value;
+ tcg_target_long ivalue, imask;
+
+ /* For 32-bit values, discard potential garbage in value. For 64-bit
+ values within [2**31, 2**32-1], we can create smaller sequences by
+ interpreting this as a negative 32-bit number, while ensuring that
+ the high 32 bits are cleared by setting SF=0. */
+ if (type == TCG_TYPE_I32 || (value & ~0xffffffffull) == 0) {
+ svalue = (int32_t)value;
value = (uint32_t)value;
+ type = TCG_TYPE_I32;
+ }
+ ivalue = ~svalue;
+
+ /* Would it take fewer insns to begin with MOVN? For the value and its
+ inverse, count the number of 16-bit lanes that are 0. */
+ for (i = wantinv = imask = 0; i < (32 << type); i += 16) {
+ tcg_target_long mask = 0xffffull << i;
+ if ((value & mask) == 0) {
+ wantinv -= 1;
+ }
+ if ((ivalue & mask) == 0) {
+ wantinv += 1;
+ imask |= mask;
+ }
}
- /* count trailing zeros in 16 bit steps, mapping 64 to 0. Emit the
- first MOVZ with the half-word immediate skipping the zeros, with a shift
- (LSL) equal to this number. Then all next instructions use MOVKs.
- Zero the processed half-word in the value, continue until empty.
- We build the final result 16bits at a time with up to 4 instructions,
- but do not emit instructions for 16bit zero holes. */
+ /* If we had more 0xffff than 0x0000, invert VALUE and use MOVN. */
insn = I3405_MOVZ;
- do {
- unsigned shift = ctz64(value) & (63 & -16);
- tcg_out_insn_3405(s, insn, shift >= 32, rd, value >> shift, shift);
+ if (wantinv > 0) {
+ value = ivalue;
+ insn = I3405_MOVN;
+ }
+
+ /* Find the lowest lane that is not 0x0000. */
+ shift = ctz64(value) & (63 & -16);
+ tcg_out_insn_3405(s, insn, type, rd, value >> shift, shift);
+
+ if (wantinv > 0) {
+ /* Re-invert the value, so MOVK sees non-inverted bits. */
+ value = ~value;
+ /* Clear out all the 0xffff lanes. */
+ value ^= imask;
+ }
+ /* Clear out the lane that we just set. */
+ value &= ~(0xffffUL << shift);
+
+ /* Iterate until all lanes have been set, and thus cleared from VALUE. */
+ while (value) {
+ shift = ctz64(value) & (63 & -16);
+ tcg_out_insn(s, 3405, MOVK, type, rd, value >> shift, shift);
value &= ~(0xffffUL << shift);
- insn = I3405_MOVK;
- } while (value);
+ }
}
static inline void tcg_out_ldst_r(TCGContext *s,
--
1.8.5.3
next prev parent reply other threads:[~2014-03-15 2:49 UTC|newest]
Thread overview: 52+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-03-15 2:48 [Qemu-devel] [PATCH 00/26] tcg/aarch64 improvements, part 3 Richard Henderson
2014-03-15 2:48 ` [Qemu-devel] [PATCH 01/26] tcg-aarch64: Properly detect SIGSEGV writes Richard Henderson
2014-03-24 11:05 ` Claudio Fontana
2014-03-24 15:33 ` Richard Henderson
2014-03-24 11:45 ` Claudio Fontana
2014-03-24 12:17 ` Peter Maydell
2014-03-24 12:41 ` Peter Maydell
2014-03-24 15:27 ` Richard Henderson
2014-03-15 2:48 ` [Qemu-devel] [PATCH 02/26] tcg-aarch64: Use intptr_t apropriately Richard Henderson
2014-03-24 12:12 ` Claudio Fontana
2014-03-15 2:48 ` [Qemu-devel] [PATCH 03/26] tcg-aarch64: Use TCGType and TCGMemOp constants Richard Henderson
2014-03-24 12:52 ` Claudio Fontana
2014-03-15 2:48 ` Richard Henderson [this message]
2014-03-24 14:06 ` [Qemu-devel] [PATCH 04/26] tcg-aarch64: Use MOVN in tcg_out_movi Claudio Fontana
2014-03-15 2:48 ` [Qemu-devel] [PATCH 05/26] tcg-aarch64: Use ORRI " Richard Henderson
2014-03-24 14:06 ` Claudio Fontana
2014-03-15 2:48 ` [Qemu-devel] [PATCH 06/26] tcg-aarch64: Special case small constants " Richard Henderson
2014-03-24 14:08 ` Claudio Fontana
2014-03-15 2:48 ` [Qemu-devel] [PATCH 07/26] tcg-aarch64: Use adrp " Richard Henderson
2014-03-24 14:05 ` Claudio Fontana
2014-03-24 15:36 ` Richard Henderson
2014-03-26 9:34 ` Claudio Fontana
2014-03-15 2:48 ` [Qemu-devel] [PATCH 08/26] tcg-aarch64: Use symbolic names for branches Richard Henderson
2014-03-24 15:31 ` Claudio Fontana
2014-03-15 2:48 ` [Qemu-devel] [PATCH 09/26] tcg-aarch64: Create tcg_out_brcond Richard Henderson
2014-03-24 15:31 ` Claudio Fontana
2014-03-15 2:48 ` [Qemu-devel] [PATCH 10/26] tcg-aarch64: Use CBZ and CBNZ Richard Henderson
2014-03-24 15:32 ` Claudio Fontana
2014-03-15 2:48 ` [Qemu-devel] [PATCH 11/26] tcg-aarch64: Reuse FP and LR in translated code Richard Henderson
2014-03-28 9:48 ` Claudio Fontana
2014-03-28 13:23 ` Richard Henderson
2014-03-15 2:48 ` [Qemu-devel] [PATCH 12/26] tcg-aarch64: Introduce tcg_out_insn_3314 Richard Henderson
2014-03-15 2:48 ` [Qemu-devel] [PATCH 13/26] tcg-aarch64: Rearrange prologue insn order Richard Henderson
2014-03-15 2:48 ` [Qemu-devel] [PATCH 14/26] tcg-aarch64: Implement tcg_register_jit Richard Henderson
2014-03-15 2:48 ` [Qemu-devel] [PATCH 15/26] tcg-aarch64: Avoid add with zero in tlb load Richard Henderson
2014-03-26 9:36 ` Claudio Fontana
2014-03-15 2:48 ` [Qemu-devel] [PATCH 16/26] tcg-aarch64: Use tcg_out_call for qemu_ld/st Richard Henderson
2014-03-26 9:37 ` Claudio Fontana
2014-03-15 2:48 ` [Qemu-devel] [PATCH 17/26] tcg-aarch64: Use ADR to pass the return address to the ld/st helpers Richard Henderson
2014-03-26 9:38 ` Claudio Fontana
2014-03-15 2:48 ` [Qemu-devel] [PATCH 18/26] tcg-aarch64: Use TCGMemOp in qemu_ld/st Richard Henderson
2014-03-26 9:39 ` Claudio Fontana
2014-03-15 2:48 ` [Qemu-devel] [PATCH 19/26] tcg-aarch64: Implement TCG_TARGET_HAS_new_ldst Richard Henderson
2014-03-26 9:40 ` Claudio Fontana
2014-03-15 2:48 ` [Qemu-devel] [PATCH 20/26] tcg-aarch64: Introduce tcg_out_insn_3507 Richard Henderson
2014-03-26 9:40 ` Claudio Fontana
2014-03-15 2:48 ` [Qemu-devel] [PATCH 21/26] tcg-aarch64: Merge aarch64_ldst_get_data/type into tcg_out_op Richard Henderson
2014-03-15 2:48 ` [Qemu-devel] [PATCH 22/26] tcg-aarch64: Replace aarch64_ldst_op_data with TCGMemOp Richard Henderson
2014-03-15 2:48 ` [Qemu-devel] [PATCH 23/26] tcg-aarch64: Replace aarch64_ldst_op_data with AArch64LdstType Richard Henderson
2014-03-15 2:48 ` [Qemu-devel] [PATCH 24/26] tcg-aarch64: Prefer unsigned offsets before signed offsets for ldst Richard Henderson
2014-03-15 2:48 ` [Qemu-devel] [PATCH 25/26] tcg-aarch64: Merge tcg_out_movr with tcg_out_mov Richard Henderson
2014-03-15 2:48 ` [Qemu-devel] [PATCH 26/26] tcg-aarch64: Support stores of zero Richard Henderson
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1394851732-25692-5-git-send-email-rth@twiddle.net \
--to=rth@twiddle.net \
--cc=claudio.fontana@gmail.com \
--cc=peter.maydell@linaro.org \
--cc=qemu-devel@nongnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).