From: Richard Henderson <richard.henderson@linaro.org>
To: qemu-devel@nongnu.org
Cc: qemu-arm@nongnu.org, Peter Maydell <peter.maydell@linaro.org>
Subject: [PATCH v6 20/45] target/arm: Implement SME LD1, ST1
Date: Fri, 8 Jul 2022 20:45:15 +0530 [thread overview]
Message-ID: <20220708151540.18136-21-richard.henderson@linaro.org> (raw)
In-Reply-To: <20220708151540.18136-1-richard.henderson@linaro.org>
We cannot reuse the SVE functions for LD[1-4] and ST[1-4],
because those functions accept only a Zreg register number.
For SME, we want to pass a pointer into ZA storage.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/helper-sme.h | 82 +++++
target/arm/sme.decode | 9 +
target/arm/sme_helper.c | 595 +++++++++++++++++++++++++++++++++++++
target/arm/translate-sme.c | 70 +++++
4 files changed, 756 insertions(+)
diff --git a/target/arm/helper-sme.h b/target/arm/helper-sme.h
index 154bc73d2e..95f6e88bdd 100644
--- a/target/arm/helper-sme.h
+++ b/target/arm/helper-sme.h
@@ -33,3 +33,85 @@ DEF_HELPER_FLAGS_4(sme_mova_cz_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
DEF_HELPER_FLAGS_4(sme_mova_zc_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
DEF_HELPER_FLAGS_4(sme_mova_cz_q, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
DEF_HELPER_FLAGS_4(sme_mova_zc_q, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+
+DEF_HELPER_FLAGS_5(sme_ld1b_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1b_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1b_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1b_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+
+DEF_HELPER_FLAGS_5(sme_ld1h_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1h_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1h_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1h_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1h_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1h_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1h_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1h_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+
+DEF_HELPER_FLAGS_5(sme_ld1s_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1s_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1s_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1s_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1s_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1s_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1s_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1s_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+
+DEF_HELPER_FLAGS_5(sme_ld1d_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1d_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1d_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1d_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1d_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1d_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1d_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1d_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+
+DEF_HELPER_FLAGS_5(sme_ld1q_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1q_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1q_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1q_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1q_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1q_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1q_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_ld1q_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+
+DEF_HELPER_FLAGS_5(sme_st1b_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1b_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1b_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1b_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+
+DEF_HELPER_FLAGS_5(sme_st1h_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1h_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1h_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1h_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1h_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1h_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1h_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1h_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+
+DEF_HELPER_FLAGS_5(sme_st1s_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1s_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1s_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1s_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1s_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1s_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1s_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1s_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+
+DEF_HELPER_FLAGS_5(sme_st1d_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1d_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1d_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1d_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1d_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1d_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1d_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1d_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+
+DEF_HELPER_FLAGS_5(sme_st1q_be_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1q_le_h, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1q_be_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1q_le_v, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1q_be_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1q_le_h_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1q_be_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
+DEF_HELPER_FLAGS_5(sme_st1q_le_v_mte, TCG_CALL_NO_WG, void, env, ptr, ptr, tl, i32)
diff --git a/target/arm/sme.decode b/target/arm/sme.decode
index 241b4895b7..900e3f2a07 100644
--- a/target/arm/sme.decode
+++ b/target/arm/sme.decode
@@ -37,3 +37,12 @@ MOVA 11000000 esz:2 00001 0 v:1 .. pg:3 0 za_imm:4 zr:5 \
&mova to_vec=1 rs=%mova_rs
MOVA 11000000 11 00001 1 v:1 .. pg:3 0 za_imm:4 zr:5 \
&mova to_vec=1 rs=%mova_rs esz=4
+
+### SME Memory
+
+&ldst esz rs pg rn rm za_imm v:bool st:bool
+
+LDST1 1110000 0 esz:2 st:1 rm:5 v:1 .. pg:3 rn:5 0 za_imm:4 \
+ &ldst rs=%mova_rs
+LDST1 1110000 111 st:1 rm:5 v:1 .. pg:3 rn:5 0 za_imm:4 \
+ &ldst esz=4 rs=%mova_rs
diff --git a/target/arm/sme_helper.c b/target/arm/sme_helper.c
index e88244423d..10fd1ed910 100644
--- a/target/arm/sme_helper.c
+++ b/target/arm/sme_helper.c
@@ -19,10 +19,14 @@
#include "qemu/osdep.h"
#include "cpu.h"
+#include "internals.h"
#include "tcg/tcg-gvec-desc.h"
#include "exec/helper-proto.h"
+#include "exec/cpu_ldst.h"
+#include "exec/exec-all.h"
#include "qemu/int128.h"
#include "vec_internal.h"
+#include "sve_ldst_internal.h"
/* ResetSVEState */
void arm_reset_sve_state(CPUARMState *env)
@@ -233,3 +237,594 @@ void HELPER(sme_mova_zc_q)(void *vd, void *za, void *vg, uint32_t desc)
}
#undef DO_MOVA_Z
+
+/*
+ * Clear elements in a tile slice comprising len bytes.
+ */
+
+typedef void ClearFn(void *ptr, size_t off, size_t len);
+
+static void clear_horizontal(void *ptr, size_t off, size_t len)
+{
+ memset(ptr + off, 0, len);
+}
+
+static void clear_vertical_b(void *vptr, size_t off, size_t len)
+{
+ for (size_t i = 0; i < len; ++i) {
+ *(uint8_t *)(vptr + tile_vslice_offset(i + off)) = 0;
+ }
+}
+
+static void clear_vertical_h(void *vptr, size_t off, size_t len)
+{
+ for (size_t i = 0; i < len; i += 2) {
+ *(uint16_t *)(vptr + tile_vslice_offset(i + off)) = 0;
+ }
+}
+
+static void clear_vertical_s(void *vptr, size_t off, size_t len)
+{
+ for (size_t i = 0; i < len; i += 4) {
+ *(uint32_t *)(vptr + tile_vslice_offset(i + off)) = 0;
+ }
+}
+
+static void clear_vertical_d(void *vptr, size_t off, size_t len)
+{
+ for (size_t i = 0; i < len; i += 8) {
+ *(uint64_t *)(vptr + tile_vslice_offset(i + off)) = 0;
+ }
+}
+
+static void clear_vertical_q(void *vptr, size_t off, size_t len)
+{
+ for (size_t i = 0; i < len; i += 16) {
+ memset(vptr + tile_vslice_offset(i + off), 0, 16);
+ }
+}
+
+/*
+ * Copy elements from an array into a tile slice comprising len bytes.
+ */
+
+typedef void CopyFn(void *dst, const void *src, size_t len);
+
+static void copy_horizontal(void *dst, const void *src, size_t len)
+{
+ memcpy(dst, src, len);
+}
+
+static void copy_vertical_b(void *vdst, const void *vsrc, size_t len)
+{
+ const uint8_t *src = vsrc;
+ uint8_t *dst = vdst;
+ size_t i;
+
+ for (i = 0; i < len; ++i) {
+ dst[tile_vslice_index(i)] = src[i];
+ }
+}
+
+static void copy_vertical_h(void *vdst, const void *vsrc, size_t len)
+{
+ const uint16_t *src = vsrc;
+ uint16_t *dst = vdst;
+ size_t i;
+
+ for (i = 0; i < len / 2; ++i) {
+ dst[tile_vslice_index(i)] = src[i];
+ }
+}
+
+static void copy_vertical_s(void *vdst, const void *vsrc, size_t len)
+{
+ const uint32_t *src = vsrc;
+ uint32_t *dst = vdst;
+ size_t i;
+
+ for (i = 0; i < len / 4; ++i) {
+ dst[tile_vslice_index(i)] = src[i];
+ }
+}
+
+static void copy_vertical_d(void *vdst, const void *vsrc, size_t len)
+{
+ const uint64_t *src = vsrc;
+ uint64_t *dst = vdst;
+ size_t i;
+
+ for (i = 0; i < len / 8; ++i) {
+ dst[tile_vslice_index(i)] = src[i];
+ }
+}
+
+static void copy_vertical_q(void *vdst, const void *vsrc, size_t len)
+{
+ for (size_t i = 0; i < len; i += 16) {
+ memcpy(vdst + tile_vslice_offset(i), vsrc + i, 16);
+ }
+}
+
+/*
+ * Host and TLB primitives for vertical tile slice addressing.
+ */
+
+#define DO_LD(NAME, TYPE, HOST, TLB) \
+static inline void sme_##NAME##_v_host(void *za, intptr_t off, void *host) \
+{ \
+ TYPE val = HOST(host); \
+ *(TYPE *)(za + tile_vslice_offset(off)) = val; \
+} \
+static inline void sme_##NAME##_v_tlb(CPUARMState *env, void *za, \
+ intptr_t off, target_ulong addr, uintptr_t ra) \
+{ \
+ TYPE val = TLB(env, useronly_clean_ptr(addr), ra); \
+ *(TYPE *)(za + tile_vslice_offset(off)) = val; \
+}
+
+#define DO_ST(NAME, TYPE, HOST, TLB) \
+static inline void sme_##NAME##_v_host(void *za, intptr_t off, void *host) \
+{ \
+ TYPE val = *(TYPE *)(za + tile_vslice_offset(off)); \
+ HOST(host, val); \
+} \
+static inline void sme_##NAME##_v_tlb(CPUARMState *env, void *za, \
+ intptr_t off, target_ulong addr, uintptr_t ra) \
+{ \
+ TYPE val = *(TYPE *)(za + tile_vslice_offset(off)); \
+ TLB(env, useronly_clean_ptr(addr), val, ra); \
+}
+
+/*
+ * The ARMVectorReg elements are stored in host-endian 64-bit units.
+ * For 128-bit quantities, the sequence defined by the Elem[] pseudocode
+ * corresponds to storing the two 64-bit pieces in little-endian order.
+ */
+#define DO_LDQ(HNAME, VNAME, BE, HOST, TLB) \
+static inline void HNAME##_host(void *za, intptr_t off, void *host) \
+{ \
+ uint64_t val0 = HOST(host), val1 = HOST(host + 8); \
+ uint64_t *ptr = za + off; \
+ ptr[0] = BE ? val1 : val0, ptr[1] = BE ? val0 : val1; \
+} \
+static inline void VNAME##_v_host(void *za, intptr_t off, void *host) \
+{ \
+ HNAME##_host(za, tile_vslice_offset(off), host); \
+} \
+static inline void HNAME##_tlb(CPUARMState *env, void *za, intptr_t off, \
+ target_ulong addr, uintptr_t ra) \
+{ \
+ uint64_t val0 = TLB(env, useronly_clean_ptr(addr), ra); \
+ uint64_t val1 = TLB(env, useronly_clean_ptr(addr + 8), ra); \
+ uint64_t *ptr = za + off; \
+ ptr[0] = BE ? val1 : val0, ptr[1] = BE ? val0 : val1; \
+} \
+static inline void VNAME##_v_tlb(CPUARMState *env, void *za, intptr_t off, \
+ target_ulong addr, uintptr_t ra) \
+{ \
+ HNAME##_tlb(env, za, tile_vslice_offset(off), addr, ra); \
+}
+
+#define DO_STQ(HNAME, VNAME, BE, HOST, TLB) \
+static inline void HNAME##_host(void *za, intptr_t off, void *host) \
+{ \
+ uint64_t *ptr = za + off; \
+ HOST(host, ptr[BE]); \
+ HOST(host + 1, ptr[!BE]); \
+} \
+static inline void VNAME##_v_host(void *za, intptr_t off, void *host) \
+{ \
+ HNAME##_host(za, tile_vslice_offset(off), host); \
+} \
+static inline void HNAME##_tlb(CPUARMState *env, void *za, intptr_t off, \
+ target_ulong addr, uintptr_t ra) \
+{ \
+ uint64_t *ptr = za + off; \
+ TLB(env, useronly_clean_ptr(addr), ptr[BE], ra); \
+ TLB(env, useronly_clean_ptr(addr + 8), ptr[!BE], ra); \
+} \
+static inline void VNAME##_v_tlb(CPUARMState *env, void *za, intptr_t off, \
+ target_ulong addr, uintptr_t ra) \
+{ \
+ HNAME##_tlb(env, za, tile_vslice_offset(off), addr, ra); \
+}
+
+DO_LD(ld1b, uint8_t, ldub_p, cpu_ldub_data_ra)
+DO_LD(ld1h_be, uint16_t, lduw_be_p, cpu_lduw_be_data_ra)
+DO_LD(ld1h_le, uint16_t, lduw_le_p, cpu_lduw_le_data_ra)
+DO_LD(ld1s_be, uint32_t, ldl_be_p, cpu_ldl_be_data_ra)
+DO_LD(ld1s_le, uint32_t, ldl_le_p, cpu_ldl_le_data_ra)
+DO_LD(ld1d_be, uint64_t, ldq_be_p, cpu_ldq_be_data_ra)
+DO_LD(ld1d_le, uint64_t, ldq_le_p, cpu_ldq_le_data_ra)
+
+DO_LDQ(sve_ld1qq_be, sme_ld1q_be, 1, ldq_be_p, cpu_ldq_be_data_ra)
+DO_LDQ(sve_ld1qq_le, sme_ld1q_le, 0, ldq_le_p, cpu_ldq_le_data_ra)
+
+DO_ST(st1b, uint8_t, stb_p, cpu_stb_data_ra)
+DO_ST(st1h_be, uint16_t, stw_be_p, cpu_stw_be_data_ra)
+DO_ST(st1h_le, uint16_t, stw_le_p, cpu_stw_le_data_ra)
+DO_ST(st1s_be, uint32_t, stl_be_p, cpu_stl_be_data_ra)
+DO_ST(st1s_le, uint32_t, stl_le_p, cpu_stl_le_data_ra)
+DO_ST(st1d_be, uint64_t, stq_be_p, cpu_stq_be_data_ra)
+DO_ST(st1d_le, uint64_t, stq_le_p, cpu_stq_le_data_ra)
+
+DO_STQ(sve_st1qq_be, sme_st1q_be, 1, stq_be_p, cpu_stq_be_data_ra)
+DO_STQ(sve_st1qq_le, sme_st1q_le, 0, stq_le_p, cpu_stq_le_data_ra)
+
+#undef DO_LD
+#undef DO_ST
+#undef DO_LDQ
+#undef DO_STQ
+
+/*
+ * Common helper for all contiguous predicated loads.
+ */
+
+static inline QEMU_ALWAYS_INLINE
+void sme_ld1(CPUARMState *env, void *za, uint64_t *vg,
+ const target_ulong addr, uint32_t desc, const uintptr_t ra,
+ const int esz, uint32_t mtedesc, bool vertical,
+ sve_ldst1_host_fn *host_fn,
+ sve_ldst1_tlb_fn *tlb_fn,
+ ClearFn *clr_fn,
+ CopyFn *cpy_fn)
+{
+ const intptr_t reg_max = simd_oprsz(desc);
+ const intptr_t esize = 1 << esz;
+ intptr_t reg_off, reg_last;
+ SVEContLdSt info;
+ void *host;
+ int flags;
+
+ /* Find the active elements. */
+ if (!sve_cont_ldst_elements(&info, addr, vg, reg_max, esz, esize)) {
+ /* The entire predicate was false; no load occurs. */
+ clr_fn(za, 0, reg_max);
+ return;
+ }
+
+ /* Probe the page(s). Exit with exception for any invalid page. */
+ sve_cont_ldst_pages(&info, FAULT_ALL, env, addr, MMU_DATA_LOAD, ra);
+
+ /* Handle watchpoints for all active elements. */
+ sve_cont_ldst_watchpoints(&info, env, vg, addr, esize, esize,
+ BP_MEM_READ, ra);
+
+ /*
+ * Handle mte checks for all active elements.
+ * Since TBI must be set for MTE, !mtedesc => !mte_active.
+ */
+ if (mtedesc) {
+ sve_cont_ldst_mte_check(&info, env, vg, addr, esize, esize,
+ mtedesc, ra);
+ }
+
+ flags = info.page[0].flags | info.page[1].flags;
+ if (unlikely(flags != 0)) {
+#ifdef CONFIG_USER_ONLY
+ g_assert_not_reached();
+#else
+ /*
+ * At least one page includes MMIO.
+ * Any bus operation can fail with cpu_transaction_failed,
+ * which for ARM will raise SyncExternal. Perform the load
+ * into scratch memory to preserve register state until the end.
+ */
+ ARMVectorReg scratch = { };
+
+ reg_off = info.reg_off_first[0];
+ reg_last = info.reg_off_last[1];
+ if (reg_last < 0) {
+ reg_last = info.reg_off_split;
+ if (reg_last < 0) {
+ reg_last = info.reg_off_last[0];
+ }
+ }
+
+ do {
+ uint64_t pg = vg[reg_off >> 6];
+ do {
+ if ((pg >> (reg_off & 63)) & 1) {
+ tlb_fn(env, &scratch, reg_off, addr + reg_off, ra);
+ }
+ reg_off += esize;
+ } while (reg_off & 63);
+ } while (reg_off <= reg_last);
+
+ cpy_fn(za, &scratch, reg_max);
+ return;
+#endif
+ }
+
+ /* The entire operation is in RAM, on valid pages. */
+
+ reg_off = info.reg_off_first[0];
+ reg_last = info.reg_off_last[0];
+ host = info.page[0].host;
+
+ if (!vertical) {
+ memset(za, 0, reg_max);
+ } else if (reg_off) {
+ clr_fn(za, 0, reg_off);
+ }
+
+ while (reg_off <= reg_last) {
+ uint64_t pg = vg[reg_off >> 6];
+ do {
+ if ((pg >> (reg_off & 63)) & 1) {
+ host_fn(za, reg_off, host + reg_off);
+ } else if (vertical) {
+ clr_fn(za, reg_off, esize);
+ }
+ reg_off += esize;
+ } while (reg_off <= reg_last && (reg_off & 63));
+ }
+
+ /*
+ * Use the slow path to manage the cross-page misalignment.
+ * But we know this is RAM and cannot trap.
+ */
+ reg_off = info.reg_off_split;
+ if (unlikely(reg_off >= 0)) {
+ tlb_fn(env, za, reg_off, addr + reg_off, ra);
+ }
+
+ reg_off = info.reg_off_first[1];
+ if (unlikely(reg_off >= 0)) {
+ reg_last = info.reg_off_last[1];
+ host = info.page[1].host;
+
+ do {
+ uint64_t pg = vg[reg_off >> 6];
+ do {
+ if ((pg >> (reg_off & 63)) & 1) {
+ host_fn(za, reg_off, host + reg_off);
+ } else if (vertical) {
+ clr_fn(za, reg_off, esize);
+ }
+ reg_off += esize;
+ } while (reg_off & 63);
+ } while (reg_off <= reg_last);
+ }
+}
+
+static inline QEMU_ALWAYS_INLINE
+void sme_ld1_mte(CPUARMState *env, void *za, uint64_t *vg,
+ target_ulong addr, uint32_t desc, uintptr_t ra,
+ const int esz, bool vertical,
+ sve_ldst1_host_fn *host_fn,
+ sve_ldst1_tlb_fn *tlb_fn,
+ ClearFn *clr_fn,
+ CopyFn *cpy_fn)
+{
+ uint32_t mtedesc = desc >> (SIMD_DATA_SHIFT + SVE_MTEDESC_SHIFT);
+ int bit55 = extract64(addr, 55, 1);
+
+ /* Remove mtedesc from the normal sve descriptor. */
+ desc = extract32(desc, 0, SIMD_DATA_SHIFT + SVE_MTEDESC_SHIFT);
+
+ /* Perform gross MTE suppression early. */
+ if (!tbi_check(desc, bit55) ||
+ tcma_check(desc, bit55, allocation_tag_from_addr(addr))) {
+ mtedesc = 0;
+ }
+
+ sme_ld1(env, za, vg, addr, desc, ra, esz, mtedesc, vertical,
+ host_fn, tlb_fn, clr_fn, cpy_fn);
+}
+
+#define DO_LD(L, END, ESZ) \
+void HELPER(sme_ld1##L##END##_h)(CPUARMState *env, void *za, void *vg, \
+ target_ulong addr, uint32_t desc) \
+{ \
+ sme_ld1(env, za, vg, addr, desc, GETPC(), ESZ, 0, false, \
+ sve_ld1##L##L##END##_host, sve_ld1##L##L##END##_tlb, \
+ clear_horizontal, copy_horizontal); \
+} \
+void HELPER(sme_ld1##L##END##_v)(CPUARMState *env, void *za, void *vg, \
+ target_ulong addr, uint32_t desc) \
+{ \
+ sme_ld1(env, za, vg, addr, desc, GETPC(), ESZ, 0, true, \
+ sme_ld1##L##END##_v_host, sme_ld1##L##END##_v_tlb, \
+ clear_vertical_##L, copy_vertical_##L); \
+} \
+void HELPER(sme_ld1##L##END##_h_mte)(CPUARMState *env, void *za, void *vg, \
+ target_ulong addr, uint32_t desc) \
+{ \
+ sme_ld1_mte(env, za, vg, addr, desc, GETPC(), ESZ, false, \
+ sve_ld1##L##L##END##_host, sve_ld1##L##L##END##_tlb, \
+ clear_horizontal, copy_horizontal); \
+} \
+void HELPER(sme_ld1##L##END##_v_mte)(CPUARMState *env, void *za, void *vg, \
+ target_ulong addr, uint32_t desc) \
+{ \
+ sme_ld1_mte(env, za, vg, addr, desc, GETPC(), ESZ, true, \
+ sme_ld1##L##END##_v_host, sme_ld1##L##END##_v_tlb, \
+ clear_vertical_##L, copy_vertical_##L); \
+}
+
+DO_LD(b, , MO_8)
+DO_LD(h, _be, MO_16)
+DO_LD(h, _le, MO_16)
+DO_LD(s, _be, MO_32)
+DO_LD(s, _le, MO_32)
+DO_LD(d, _be, MO_64)
+DO_LD(d, _le, MO_64)
+DO_LD(q, _be, MO_128)
+DO_LD(q, _le, MO_128)
+
+#undef DO_LD
+
+/*
+ * Common helper for all contiguous predicated stores.
+ */
+
+static inline QEMU_ALWAYS_INLINE
+void sme_st1(CPUARMState *env, void *za, uint64_t *vg,
+ const target_ulong addr, uint32_t desc, const uintptr_t ra,
+ const int esz, uint32_t mtedesc, bool vertical,
+ sve_ldst1_host_fn *host_fn,
+ sve_ldst1_tlb_fn *tlb_fn)
+{
+ const intptr_t reg_max = simd_oprsz(desc);
+ const intptr_t esize = 1 << esz;
+ intptr_t reg_off, reg_last;
+ SVEContLdSt info;
+ void *host;
+ int flags;
+
+ /* Find the active elements. */
+ if (!sve_cont_ldst_elements(&info, addr, vg, reg_max, esz, esize)) {
+ /* The entire predicate was false; no store occurs. */
+ return;
+ }
+
+ /* Probe the page(s). Exit with exception for any invalid page. */
+ sve_cont_ldst_pages(&info, FAULT_ALL, env, addr, MMU_DATA_STORE, ra);
+
+ /* Handle watchpoints for all active elements. */
+ sve_cont_ldst_watchpoints(&info, env, vg, addr, esize, esize,
+ BP_MEM_WRITE, ra);
+
+ /*
+ * Handle mte checks for all active elements.
+ * Since TBI must be set for MTE, !mtedesc => !mte_active.
+ */
+ if (mtedesc) {
+ sve_cont_ldst_mte_check(&info, env, vg, addr, esize, esize,
+ mtedesc, ra);
+ }
+
+ flags = info.page[0].flags | info.page[1].flags;
+ if (unlikely(flags != 0)) {
+#ifdef CONFIG_USER_ONLY
+ g_assert_not_reached();
+#else
+ /*
+ * At least one page includes MMIO.
+ * Any bus operation can fail with cpu_transaction_failed,
+ * which for ARM will raise SyncExternal. We cannot avoid
+ * this fault and will leave with the store incomplete.
+ */
+ reg_off = info.reg_off_first[0];
+ reg_last = info.reg_off_last[1];
+ if (reg_last < 0) {
+ reg_last = info.reg_off_split;
+ if (reg_last < 0) {
+ reg_last = info.reg_off_last[0];
+ }
+ }
+
+ do {
+ uint64_t pg = vg[reg_off >> 6];
+ do {
+ if ((pg >> (reg_off & 63)) & 1) {
+ tlb_fn(env, za, reg_off, addr + reg_off, ra);
+ }
+ reg_off += esize;
+ } while (reg_off & 63);
+ } while (reg_off <= reg_last);
+ return;
+#endif
+ }
+
+ reg_off = info.reg_off_first[0];
+ reg_last = info.reg_off_last[0];
+ host = info.page[0].host;
+
+ while (reg_off <= reg_last) {
+ uint64_t pg = vg[reg_off >> 6];
+ do {
+ if ((pg >> (reg_off & 63)) & 1) {
+ host_fn(za, reg_off, host + reg_off);
+ }
+ reg_off += 1 << esz;
+ } while (reg_off <= reg_last && (reg_off & 63));
+ }
+
+ /*
+ * Use the slow path to manage the cross-page misalignment.
+ * But we know this is RAM and cannot trap.
+ */
+ reg_off = info.reg_off_split;
+ if (unlikely(reg_off >= 0)) {
+ tlb_fn(env, za, reg_off, addr + reg_off, ra);
+ }
+
+ reg_off = info.reg_off_first[1];
+ if (unlikely(reg_off >= 0)) {
+ reg_last = info.reg_off_last[1];
+ host = info.page[1].host;
+
+ do {
+ uint64_t pg = vg[reg_off >> 6];
+ do {
+ if ((pg >> (reg_off & 63)) & 1) {
+ host_fn(za, reg_off, host + reg_off);
+ }
+ reg_off += 1 << esz;
+ } while (reg_off & 63);
+ } while (reg_off <= reg_last);
+ }
+}
+
+static inline QEMU_ALWAYS_INLINE
+void sme_st1_mte(CPUARMState *env, void *za, uint64_t *vg, target_ulong addr,
+ uint32_t desc, uintptr_t ra, int esz, bool vertical,
+ sve_ldst1_host_fn *host_fn,
+ sve_ldst1_tlb_fn *tlb_fn)
+{
+ uint32_t mtedesc = desc >> (SIMD_DATA_SHIFT + SVE_MTEDESC_SHIFT);
+ int bit55 = extract64(addr, 55, 1);
+
+ /* Remove mtedesc from the normal sve descriptor. */
+ desc = extract32(desc, 0, SIMD_DATA_SHIFT + SVE_MTEDESC_SHIFT);
+
+ /* Perform gross MTE suppression early. */
+ if (!tbi_check(desc, bit55) ||
+ tcma_check(desc, bit55, allocation_tag_from_addr(addr))) {
+ mtedesc = 0;
+ }
+
+ sme_st1(env, za, vg, addr, desc, ra, esz, mtedesc,
+ vertical, host_fn, tlb_fn);
+}
+
+#define DO_ST(L, END, ESZ) \
+void HELPER(sme_st1##L##END##_h)(CPUARMState *env, void *za, void *vg, \
+ target_ulong addr, uint32_t desc) \
+{ \
+ sme_st1(env, za, vg, addr, desc, GETPC(), ESZ, 0, false, \
+ sve_st1##L##L##END##_host, sve_st1##L##L##END##_tlb); \
+} \
+void HELPER(sme_st1##L##END##_v)(CPUARMState *env, void *za, void *vg, \
+ target_ulong addr, uint32_t desc) \
+{ \
+ sme_st1(env, za, vg, addr, desc, GETPC(), ESZ, 0, true, \
+ sme_st1##L##END##_v_host, sme_st1##L##END##_v_tlb); \
+} \
+void HELPER(sme_st1##L##END##_h_mte)(CPUARMState *env, void *za, void *vg, \
+ target_ulong addr, uint32_t desc) \
+{ \
+ sme_st1_mte(env, za, vg, addr, desc, GETPC(), ESZ, false, \
+ sve_st1##L##L##END##_host, sve_st1##L##L##END##_tlb); \
+} \
+void HELPER(sme_st1##L##END##_v_mte)(CPUARMState *env, void *za, void *vg, \
+ target_ulong addr, uint32_t desc) \
+{ \
+ sme_st1_mte(env, za, vg, addr, desc, GETPC(), ESZ, true, \
+ sme_st1##L##END##_v_host, sme_st1##L##END##_v_tlb); \
+}
+
+DO_ST(b, , MO_8)
+DO_ST(h, _be, MO_16)
+DO_ST(h, _le, MO_16)
+DO_ST(s, _be, MO_32)
+DO_ST(s, _le, MO_32)
+DO_ST(d, _be, MO_64)
+DO_ST(d, _le, MO_64)
+DO_ST(q, _be, MO_128)
+DO_ST(q, _le, MO_128)
+
+#undef DO_ST
diff --git a/target/arm/translate-sme.c b/target/arm/translate-sme.c
index 79c33d35c2..42d14b883a 100644
--- a/target/arm/translate-sme.c
+++ b/target/arm/translate-sme.c
@@ -173,3 +173,73 @@ static bool trans_MOVA(DisasContext *s, arg_MOVA *a)
return true;
}
+
+static bool trans_LDST1(DisasContext *s, arg_LDST1 *a)
+{
+ typedef void GenLdSt1(TCGv_env, TCGv_ptr, TCGv_ptr, TCGv, TCGv_i32);
+
+ /*
+ * Indexed by [esz][be][v][mte][st], which is (except for load/store)
+ * also the order in which the elements appear in the function names,
+ * and so how we must concatenate the pieces.
+ */
+
+#define FN_LS(F) { gen_helper_sme_ld1##F, gen_helper_sme_st1##F }
+#define FN_MTE(F) { FN_LS(F), FN_LS(F##_mte) }
+#define FN_HV(F) { FN_MTE(F##_h), FN_MTE(F##_v) }
+#define FN_END(L, B) { FN_HV(L), FN_HV(B) }
+
+ static GenLdSt1 * const fns[5][2][2][2][2] = {
+ FN_END(b, b),
+ FN_END(h_le, h_be),
+ FN_END(s_le, s_be),
+ FN_END(d_le, d_be),
+ FN_END(q_le, q_be),
+ };
+
+#undef FN_LS
+#undef FN_MTE
+#undef FN_HV
+#undef FN_END
+
+ TCGv_ptr t_za, t_pg;
+ TCGv_i64 addr;
+ int svl, desc = 0;
+ bool be = s->be_data == MO_BE;
+ bool mte = s->mte_active[0];
+
+ if (!dc_isar_feature(aa64_sme, s)) {
+ return false;
+ }
+ if (!sme_smza_enabled_check(s)) {
+ return true;
+ }
+
+ t_za = get_tile_rowcol(s, a->esz, a->rs, a->za_imm, a->v);
+ t_pg = pred_full_reg_ptr(s, a->pg);
+ addr = tcg_temp_new_i64();
+
+ tcg_gen_shli_i64(addr, cpu_reg(s, a->rm), a->esz);
+ tcg_gen_add_i64(addr, addr, cpu_reg_sp(s, a->rn));
+
+ if (mte) {
+ desc = FIELD_DP32(desc, MTEDESC, MIDX, get_mem_index(s));
+ desc = FIELD_DP32(desc, MTEDESC, TBI, s->tbid);
+ desc = FIELD_DP32(desc, MTEDESC, TCMA, s->tcma);
+ desc = FIELD_DP32(desc, MTEDESC, WRITE, a->st);
+ desc = FIELD_DP32(desc, MTEDESC, SIZEM1, (1 << a->esz) - 1);
+ desc <<= SVE_MTEDESC_SHIFT;
+ } else {
+ addr = clean_data_tbi(s, addr);
+ }
+ svl = streaming_vec_reg_size(s);
+ desc = simd_desc(svl, svl, desc);
+
+ fns[a->esz][be][a->v][mte][a->st](cpu_env, t_za, t_pg, addr,
+ tcg_constant_i32(desc));
+
+ tcg_temp_free_ptr(t_za);
+ tcg_temp_free_ptr(t_pg);
+ tcg_temp_free_i64(addr);
+ return true;
+}
--
2.34.1
next prev parent reply other threads:[~2022-07-08 15:42 UTC|newest]
Thread overview: 47+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-07-08 15:14 [PATCH v6 00/45] target/arm: Scalable Matrix Extension Richard Henderson
2022-07-08 15:14 ` [PATCH v6 01/45] target/arm: Handle SME in aarch64_cpu_dump_state Richard Henderson
2022-07-08 15:14 ` [PATCH v6 02/45] target/arm: Add infrastructure for disas_sme Richard Henderson
2022-07-08 15:14 ` [PATCH v6 03/45] target/arm: Trap non-streaming usage when Streaming SVE is active Richard Henderson
2022-07-08 15:14 ` [PATCH v6 04/45] target/arm: Mark ADR as non-streaming Richard Henderson
2022-07-08 15:15 ` [PATCH v6 05/45] target/arm: Mark RDFFR, WRFFR, SETFFR " Richard Henderson
2022-07-08 15:15 ` [PATCH v6 06/45] target/arm: Mark BDEP, BEXT, BGRP, COMPACT, FEXPA, FTSSEL " Richard Henderson
2022-07-08 15:15 ` [PATCH v6 07/45] target/arm: Mark PMULL, FMMLA " Richard Henderson
2022-07-08 15:15 ` [PATCH v6 08/45] target/arm: Mark FTSMUL, FTMAD, FADDA " Richard Henderson
2022-07-08 15:15 ` [PATCH v6 09/45] target/arm: Mark SMMLA, UMMLA, USMMLA " Richard Henderson
2022-07-08 15:15 ` [PATCH v6 10/45] target/arm: Mark string/histo/crypto " Richard Henderson
2022-07-08 15:15 ` [PATCH v6 11/45] target/arm: Mark gather/scatter load/store " Richard Henderson
2022-07-08 15:15 ` [PATCH v6 12/45] target/arm: Mark gather prefetch " Richard Henderson
2022-07-08 15:15 ` [PATCH v6 13/45] target/arm: Mark LDFF1 and LDNF1 " Richard Henderson
2022-07-08 15:15 ` [PATCH v6 14/45] target/arm: Mark LD1RO " Richard Henderson
2022-07-08 15:15 ` [PATCH v6 15/45] target/arm: Add SME enablement checks Richard Henderson
2022-07-08 15:15 ` [PATCH v6 16/45] target/arm: Handle SME in sve_access_check Richard Henderson
2022-07-08 15:15 ` [PATCH v6 17/45] target/arm: Implement SME RDSVL, ADDSVL, ADDSPL Richard Henderson
2022-07-08 15:15 ` [PATCH v6 18/45] target/arm: Implement SME ZERO Richard Henderson
2022-07-08 15:15 ` [PATCH v6 19/45] target/arm: Implement SME MOVA Richard Henderson
2022-07-08 15:15 ` Richard Henderson [this message]
2022-07-08 15:15 ` [PATCH v6 21/45] target/arm: Export unpredicated ld/st from translate-sve.c Richard Henderson
2022-07-08 15:15 ` [PATCH v6 22/45] target/arm: Implement SME LDR, STR Richard Henderson
2022-07-08 15:15 ` [PATCH v6 23/45] target/arm: Implement SME ADDHA, ADDVA Richard Henderson
2022-07-08 15:15 ` [PATCH v6 24/45] target/arm: Implement FMOPA, FMOPS (non-widening) Richard Henderson
2022-07-08 15:15 ` [PATCH v6 25/45] target/arm: Implement BFMOPA, BFMOPS Richard Henderson
2022-07-08 15:15 ` [PATCH v6 26/45] target/arm: Implement FMOPA, FMOPS (widening) Richard Henderson
2022-07-08 15:15 ` [PATCH v6 27/45] target/arm: Implement SME integer outer product Richard Henderson
2022-07-08 15:15 ` [PATCH v6 28/45] target/arm: Implement PSEL Richard Henderson
2022-07-08 15:15 ` [PATCH v6 29/45] target/arm: Implement REVD Richard Henderson
2022-07-08 15:15 ` [PATCH v6 30/45] target/arm: Implement SCLAMP, UCLAMP Richard Henderson
2022-07-08 15:15 ` [PATCH v6 31/45] target/arm: Reset streaming sve state on exception boundaries Richard Henderson
2022-07-08 15:15 ` [PATCH v6 32/45] target/arm: Enable SME for -cpu max Richard Henderson
2022-07-08 15:15 ` [PATCH v6 33/45] linux-user/aarch64: Clear tpidr2_el0 if CLONE_SETTLS Richard Henderson
2022-07-08 15:15 ` [PATCH v6 34/45] linux-user/aarch64: Reset PSTATE.SM on syscalls Richard Henderson
2022-07-08 15:15 ` [PATCH v6 35/45] linux-user/aarch64: Add SM bit to SVE signal context Richard Henderson
2022-07-08 15:15 ` [PATCH v6 36/45] linux-user/aarch64: Tidy target_restore_sigframe error return Richard Henderson
2022-07-08 15:15 ` [PATCH v6 37/45] linux-user/aarch64: Do not allow duplicate or short sve records Richard Henderson
2022-07-08 15:15 ` [PATCH v6 38/45] linux-user/aarch64: Verify extra record lock succeeded Richard Henderson
2022-07-08 15:15 ` [PATCH v6 39/45] linux-user/aarch64: Move sve record checks into restore Richard Henderson
2022-07-08 15:15 ` [PATCH v6 40/45] linux-user/aarch64: Implement SME signal handling Richard Henderson
2022-07-08 15:15 ` [PATCH v6 41/45] linux-user: Rename sve prctls Richard Henderson
2022-07-08 15:15 ` [PATCH v6 42/45] linux-user/aarch64: Implement PR_SME_GET_VL, PR_SME_SET_VL Richard Henderson
2022-07-08 15:15 ` [PATCH v6 43/45] target/arm: Only set ZEN in reset if SVE present Richard Henderson
2022-07-08 15:15 ` [PATCH v6 44/45] target/arm: Enable SME for user-only Richard Henderson
2022-07-08 15:15 ` [PATCH v6 45/45] linux-user/aarch64: Add SME related hwcap entries Richard Henderson
2022-07-11 12:50 ` [PATCH v6 00/45] target/arm: Scalable Matrix Extension Peter Maydell
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=20220708151540.18136-21-richard.henderson@linaro.org \
--to=richard.henderson@linaro.org \
--cc=peter.maydell@linaro.org \
--cc=qemu-arm@nongnu.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).