Skip to contentSkip to conversation

target/mips: Translate MSA vector load/store opcodes

RFC PATCH Superseded Philippe Mathieu-Daudé 2 participants 4 messages Apr 14 – Apr 15

v4 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

v4 cover target/mips: Translate MSA vector load/store opcodes

Hi Richard,

I tried to implement your suggestion from v3 [*]. Unfortunately the box where I had the MSA test suite got erased, so it will take me some time to restore the setup. I was surprised these patches pass our CI tests so I added a pair of abort() which didn't fire, so this code path isn't tested at all there. Anyhow, do you mind having a quick look to see this is going in the right direction?

Thanks!

Phil.

[*] https://lore.kernel.org/qemu-devel/[email protected]/

Philippe Mathieu-Daudé (2):
target/mips: Translate MSA vector load opcode (LD.df) target/mips: Translate MSA vector store opcode (ST.df)

 target/mips/tcg/msa_helper.h.inc |   9 --
 target/mips/tcg/msa_helper.c     | 173 -------------------------------
 target/mips/tcg/msa_translate.c  |  96 ++++++++++++++---
 3 files changed, 84 insertions(+), 194 deletions(-)
Signature · 1 line
2.53.0
#1 lore raw

RFC PATCH v4 1/2 target/mips: Translate MSA vector load opcode (LD.df)

Replace helpers by translation.

Remove legacy cpu_ld*_data_ra() calls, replacing by tcg_gen_qemu_ld() which allow to respect atomicity.

---
 target/mips/tcg/msa_helper.h.inc |  1 -
 target/mips/tcg/msa_helper.c     | 69 --------------------------------
 target/mips/tcg/msa_translate.c  | 44 +++++++++++++++++++-
 3 files changed, 43 insertions(+), 71 deletions(-)
target/mips/tcg/msa_helper.h.inc +0 -1
index 4963d1553a0..7ede16f327a 100644
@@ -434,7 +434,6 @@ DEF_HELPER_4(msa_ffint_s_df, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_ffint_u_df, void, env, i32, i32, i32)
 
 #define MSALDST_PROTO(type)                         \
-DEF_HELPER_3(msa_ld_ ## type, void, env, i32, tl)   \
 DEF_HELPER_3(msa_st_ ## type, void, env, i32, tl)
 MSALDST_PROTO(b)
 MSALDST_PROTO(h)
target/mips/tcg/msa_helper.c +0 -69

75 lines not on this page · raw message

target/mips/tcg/msa_translate.c +43 -1

56 lines not on this page · raw message

Signature · 1 line
2.53.0
#2 lore raw

Re: [RFC PATCH v4 1/2] target/mips: Translate MSA vector load opcode (LD.df)

Quoted 19 lines · Philippe Mathieu-Daudé
target/mips/tcg/msa_translate.c +19 -0
+static bool trans_LD(DisasContext *ctx, arg_msa_i *a)
+{
+    TCGv_va addr;
+    TCGv_i128 d16;
+    MemOp mop;
+    int d0 = a->wd << 1;
+    int d1 = d0 + 1;
+
+    if (!check_msa_enabled(ctx)) {
+        return true;
+    }
+
+    addr = tcgv_va_temp_new();
+    gen_base_offset_addr(ctx, addr, a->ws, a->sa << a->df);
+
+    mop = MO_128 | MO_LE;
+    if (a->df == 0) {
+        mop |= MO_ATOM_NONE;
+    } else if (a->df == 3) {

target/mips/tcg/msa_translate.c TRANS(FTINT_U, trans_msa_2rf, gen_helper_msa_ftint_u_df); · line 789

Quoted 10 lines
target/mips/tcg/msa_translate.c +10 -0
+        mop |= MO_ATOM_IFALIGN_PAIR;
+    } else {
+        mop |= MO_ATOM_SUBALIGN; /* slightly stronger than required */
+    }
+    mop |= a->df << MO_ASHIFT; /* MO_ALIGN */
+
+    d16 = tcg_temp_new_i128();
+    tcg_gen_qemu_ld_i128(d16, addr, ctx->mem_idx, mop);
+
+    tcg_gen_st_i128(d16, tcg_env, offsetof(CPUMIPSState, active_fpu.fpr[d0]));

You cannot write to slots that have ...

target/mips/tcg/msa_translate.c TRANS(FTINT_U, trans_msa_2rf, gen_helper_msa_ftint_u_df); · line 794

Quoted 5 lines
target/mips/tcg/msa_translate.c +5 -0
+
+    if (mo_endian(ctx) != MO_LE) {
+        if (a->df == 1) {
+            tcg_gen_hswap_i64(msa_wr_d[d0], msa_wr_d[d0]);
+            tcg_gen_hswap_i64(msa_wr_d[d1], msa_wr_d[d1]);

... tcg globals associated with them.

You need

	tcg_gen_extr_i128_i64(msa_wr_d[lo], msa_wr_d[hi], d16);

to extract the halves into existing TCGv_i64.

I can't follow the endianness assumptions in msa_helper.c. I suspect it can't handle both big and little-endian host.

But hswap and wswap don't match bswap16x4 and bswap32x2,

Before:
LE load, then bswap elements without swapping element ordering.

After:
LE load, then swap element ordering without bswap elements.

I think you were looking for:
BE load, then swap element ordering without bswap elements.

But in any case, it's an algorithm change.

r~

#3 lore raw

RFC PATCH v4 2/2 target/mips: Translate MSA vector store opcode (ST.df)

In reply to Philippe Mathieu-Daudé

Replace helpers by translation.

Remove legacy cpu_st*_data_ra() calls, replacing by tcg_gen_qemu_st() which allow to respect atomicity.

Remove the ensure_writable_pages() hack and the bswap NxM helpers.

---
 target/mips/tcg/msa_helper.h.inc |   8 ---
 target/mips/tcg/msa_helper.c     | 104 -------------------------------
 target/mips/tcg/msa_translate.c  |  54 ++++++++++++----
 3 files changed, 42 insertions(+), 124 deletions(-)
target/mips/tcg/msa_helper.h.inc +0 -8
index 7ede16f327a..e994353056f 100644
@@ -432,11 +432,3 @@ DEF_HELPER_4(msa_ftint_s_df, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_ftint_u_df, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_ffint_s_df, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_ffint_u_df, void, env, i32, i32, i32)
-
-#define MSALDST_PROTO(type)                         \
-DEF_HELPER_3(msa_st_ ## type, void, env, i32, tl)
-MSALDST_PROTO(b)
-MSALDST_PROTO(h)
-MSALDST_PROTO(w)
-MSALDST_PROTO(d)
-#undef MSALDST_PROTO
target/mips/tcg/msa_helper.c +0 -104

107 lines not on this page · raw message

target/mips/tcg/msa_translate.c +42 -12

83 lines not on this page · raw message

Signature · 1 line
2.53.0