[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[PULL 001/113] target/i386: fix operand order for PDEP and PEXT
From: |
Paolo Bonzini |
Subject: |
[PULL 001/113] target/i386: fix operand order for PDEP and PEXT |
Date: |
Wed, 2 Dec 2020 03:06:57 -0500 |
For PDEP and PEXT, the mask is provided in the memory (mod+r/m)
operand, and therefore is loaded in s->T0 by gen_ldst_modrm.
The source is provided in the second source operand (VEX.vvvv)
and therefore is loaded in s->T1. Fix the order in which
they are passed to the helpers.
Reported-by: Lenard Szolnoki <blog@lenardszolnoki.com>
Analyzed-by: Lenard Szolnoki <blog@lenardszolnoki.com>
Fixes: https://bugs.launchpad.net/qemu/+bug/1605123
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
target/i386/translate.c | 8 +++----
tests/tcg/i386/Makefile.target | 3 +++
tests/tcg/i386/test-i386-bmi2.c | 39 +++++++++++++++++++++++++++++++++
3 files changed, 46 insertions(+), 4 deletions(-)
create mode 100644 tests/tcg/i386/test-i386-bmi2.c
diff --git a/target/i386/translate.c b/target/i386/translate.c
index 4c57307e42..e8f5f5803a 100644
--- a/target/i386/translate.c
+++ b/target/i386/translate.c
@@ -3936,14 +3936,14 @@ static void gen_sse(CPUX86State *env, DisasContext *s,
int b,
}
ot = mo_64_32(s->dflag);
gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
- /* Note that by zero-extending the mask operand, we
+ /* Note that by zero-extending the source operand, we
automatically handle zero-extending the result. */
if (ot == MO_64) {
tcg_gen_mov_tl(s->T1, cpu_regs[s->vex_v]);
} else {
tcg_gen_ext32u_tl(s->T1, cpu_regs[s->vex_v]);
}
- gen_helper_pdep(cpu_regs[reg], s->T0, s->T1);
+ gen_helper_pdep(cpu_regs[reg], s->T1, s->T0);
break;
case 0x2f5: /* pext Gy, By, Ey */
@@ -3954,14 +3954,14 @@ static void gen_sse(CPUX86State *env, DisasContext *s,
int b,
}
ot = mo_64_32(s->dflag);
gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
- /* Note that by zero-extending the mask operand, we
+ /* Note that by zero-extending the source operand, we
automatically handle zero-extending the result. */
if (ot == MO_64) {
tcg_gen_mov_tl(s->T1, cpu_regs[s->vex_v]);
} else {
tcg_gen_ext32u_tl(s->T1, cpu_regs[s->vex_v]);
}
- gen_helper_pext(cpu_regs[reg], s->T0, s->T1);
+ gen_helper_pext(cpu_regs[reg], s->T1, s->T0);
break;
case 0x1f6: /* adcx Gy, Ey */
diff --git a/tests/tcg/i386/Makefile.target b/tests/tcg/i386/Makefile.target
index a66232a67d..ad187cb2c9 100644
--- a/tests/tcg/i386/Makefile.target
+++ b/tests/tcg/i386/Makefile.target
@@ -18,6 +18,9 @@ test-i386-pcmpistri: CFLAGS += -msse4.2
run-test-i386-pcmpistri: QEMU_OPTS += -cpu max
run-plugin-test-i386-pcmpistri-%: QEMU_OPTS += -cpu max
+run-test-i386-bmi2: QEMU_OPTS += -cpu max
+run-plugin-test-i386-bmi2-%: QEMU_OPTS += -cpu max
+
#
# hello-i386 is a barebones app
#
diff --git a/tests/tcg/i386/test-i386-bmi2.c b/tests/tcg/i386/test-i386-bmi2.c
new file mode 100644
index 0000000000..d80a859565
--- /dev/null
+++ b/tests/tcg/i386/test-i386-bmi2.c
@@ -0,0 +1,39 @@
+/* See if various BMI2 instructions give expected results */
+#include <assert.h>
+#include <stdint.h>
+
+int main(int argc, char *argv[]) {
+ uint64_t ehlo = 0x202020204f4c4845ull;
+ uint64_t mask = 0xa080800302020001ull;
+ uint64_t result64;
+ uint32_t result32;
+
+ /* 64 bits */
+ asm volatile ("pextq %2, %1, %0" : "=r"(result64) : "r"(ehlo),
"m"(mask));
+ assert(result64 == 133);
+
+ asm volatile ("pdepq %2, %1, %0" : "=r"(result64) : "r"(result64),
"m"(mask));
+ assert(result64 == (ehlo & mask));
+
+ asm volatile ("pextq %2, %1, %0" : "=r"(result64) : "r"(-1ull),
"m"(mask));
+ assert(result64 == 511); /* mask has 9 bits set */
+
+ asm volatile ("pdepq %2, %1, %0" : "=r"(result64) : "r"(-1ull),
"m"(mask));
+ assert(result64 == mask);
+
+ /* 32 bits */
+ asm volatile ("pextl %2, %k1, %k0" : "=r"(result32) : "r"(ehlo),
"m"(mask));
+ assert(result32 == 5);
+
+ asm volatile ("pdepl %2, %k1, %k0" : "=r"(result32) : "r"(result32),
"m"(mask));
+ assert(result32 == (uint32_t)(ehlo & mask));
+
+ asm volatile ("pextl %2, %k1, %k0" : "=r"(result32) : "r"(-1ull),
"m"(mask));
+ assert(result32 == 7); /* mask has 3 bits set */
+
+ asm volatile ("pdepl %2, %k1, %k0" : "=r"(result32) : "r"(-1ull),
"m"(mask));
+ assert(result32 == (uint32_t)mask);
+
+ return 0;
+}
+
--
2.26.2
- [PULL 000/113] First batch of misc (i386, kernel-doc, memory, vl.c) changes for QEMU 6.0, Paolo Bonzini, 2020/12/02
- [PULL 003/113] target/i386: seg_helper: Correct segement selector nullification in the RET/IRET helper, Paolo Bonzini, 2020/12/02
- [PULL 002/113] target/i386: Support up to 32768 CPUs without IRQ remapping, Paolo Bonzini, 2020/12/02
- [PULL 005/113] docs/devel/loads-stores: Add regexp for DMA functions, Paolo Bonzini, 2020/12/02
- [PULL 008/113] dma: Let dma_memory_set() propagate MemTxResult, Paolo Bonzini, 2020/12/02
- [PULL 001/113] target/i386: fix operand order for PDEP and PEXT,
Paolo Bonzini <=
- [PULL 009/113] dma: Let dma_memory_rw() propagate MemTxResult, Paolo Bonzini, 2020/12/02
- [PULL 004/113] WHPX: support for the kernel-irqchip on/off, Paolo Bonzini, 2020/12/02
- [PULL 007/113] dma: Document address_space_map/address_space_unmap() prototypes, Paolo Bonzini, 2020/12/02
- [PULL 013/113] pci: Let pci_dma_read() propagate MemTxResult, Paolo Bonzini, 2020/12/02
- [PULL 014/113] pci: Let pci_dma_write() propagate MemTxResult, Paolo Bonzini, 2020/12/02
- [PULL 006/113] qom: eliminate identical functions, Paolo Bonzini, 2020/12/02
- [PULL 012/113] pci: Let pci_dma_rw() propagate MemTxResult, Paolo Bonzini, 2020/12/02
- [PULL 010/113] dma: Let dma_memory_read() propagate MemTxResult, Paolo Bonzini, 2020/12/02
- [PULL 011/113] dma: Let dma_memory_write() propagate MemTxResult, Paolo Bonzini, 2020/12/02
- [PULL 015/113] hw/ssi/aspeed_smc: Rename 'max_slaves' variable as 'max_peripherals', Paolo Bonzini, 2020/12/02